For courses in Finite Element Analysis, offered in departments of Mechanical or Civil and Environmental Engineering. Wh
1,146 133 46MB
Pages 928 [929] Year 2015
Table of contents :
Cover
Title
Copyright
Contents
Preface
Acknowledgments
1 Introduction
1.1 Engineering Problems
1.2 Numerical Methods
1.3 A Brief History of the Finite Element Method and Ansys
1.4 Basic Steps in the Finite Element Method
1.5 Direct Formulation
1.6 Minimum Total Potential Energy Formulation
1.7 Weighted Residual Formulations
1.8 Verification of Results
1.9 Understanding the Problem
Summary
References
Problems
2 Matrix Algebra
2.1 Basic Definitions
2.2 Matrix Addition or Subtraction
2.3 Matrix Multiplication
2.4 Partitioning of a Matrix
2.5 Transpose of a Matrix
2.6 Determinant of a Matrix
2.7 Solutions of Simultaneous Linear Equations
2.8 Inverse of a Matrix
2.9 Eigenvalues and Eigenvectors
2.10 Using Matlab to Manipulate Matrices
2.11 Using Excel to Manipulate Matrices
Summary
References
Problems
3 Trusses
3.1 Definition of a Truss
3.2 Finite Element Formulation
3.3 Space Trusses
3.4 Overview of the Ansys Program
3.5 Examples Using Ansys
3.6 Verification of Results
Summary
References
Problems
4 Axial Members, Beams, and Frames
4.1 Members Under Axial Loading
4.2 Beams
4.3 Finite Element Formulation of Beams
4.4 Finite Element Formulation of Frames
4.5 ThreeDimensional Beam Element
4.6 An Example Using Ansys
4.7 Verification of Results
Summary
References
Problems
5 OneDimensional Elements
5.1 Linear Elements
5.2 Quadratic Elements
5.3 Cubic Elements
5.4 Global, Local, and Natural Coordinates
5.5 Isoparametric Elements
5.6 Numerical Integration: Gauss–Legendre Quadrature
5.7 Examples of OneDimensional Elements in Ansys
Summary
References
Problems
6 Analysis of OneDimensional Problems
6.1 Heat Transfer Problems
6.2 A Fluid Mechanics Problem
6.3 An Example Using Ansys
6.4 Verification of Results
Summary
References
Problems
7 TwoDimensional Elements
7.1 Rectangular Elements
7.2 Quadratic Quadrilateral Elements
7.3 Linear Triangular Elements
7.4 Quadratic Triangular Elements
7.5 Axisymmetric Elements
7.6 Isoparametric Elements
7.7 TwoDimensional Integrals: Gauss–Legendre Quadrature
7.8 Examples of TwoDimensional Elements in Ansys
Summary
References
Problems
8 More Ansys
8.1 Ansys Program
8.2 Ansys Database and Files
8.3 Creating a Finite Element Model with Ansys: Preprocessing
8.4 hMethod Versus pMethod
8.5 Applying Boundary Conditions, Loads, and the Solution
8.6 Results of Your Finite Element Model: Postprocessing
8.7 Selection Options
8.8 Graphics Capabilities
8.9 ErrorEstimation Procedures
8.10 An Example Problem
Summary
References
9 Analysis of TwoDimensional Heat Transfer Problems
9.1 General Conduction Problems
9.2 Formulation with Rectangular Elements
9.3 Formulation with Triangular Elements
9.4 Axisymmetric Formulation of ThreeDimensional Problems
9.5 Unsteady Heat Transfer
9.6 Conduction Elements Used by Ansys
9.7 Examples Using Ansys
9.8 Verification of Results
Summary
References
Problems
10 Analysis of TwoDimensional Solid Mechanics Problems
10.1 Torsion of Members with Arbitrary CrossSection Shape
10.2 PlaneStress Formulation
10.3 Isoparametric Formulation: Using a Quadrilateral Element
10.4 Axisymmetric Formulation
10.5 Basic Failure Theories
10.6 Examples Using Ansys
10.7 Verification of Results
Summary
References
Problems
11 Dynamic Problems
11.1 Review of Dynamics
11.2 Review of Vibration of Mechanical and Structural Systems
11.3 Lagrange’s Equations
11.4 Finite Element Formulation of Axial Members
11.5 Finite Element Formulation of Beams and Frames
11.6 Examples Using Ansys
Summary
References
Problems
12 Analysis of Fluid Mechanics Problems
12.1 Direct Formulation of Flow Through Pipes
12.2 Ideal Fluid Flow
12.3 Groundwater Flow
12.4 Examples Using Ansys
12.5 Verification of Results
Summary
References
Problems
13 ThreeDimensional Elements
13.1 The FourNodeTetrahedral Element
13.2 Analysis of ThreeDimensional Solid Problems Using FourNodeTetrahedral Elements
13.3 The EightNode Brick Element
13.4 The TenNode Tetrahedral Element
13.5 The TwentyNode Brick Element
13.6 Examples of ThreeDimensional Elements in Ansys
13.7 Basic SolidModeling Ideas
13.8 A Thermal Example Using Ansys
13.9 A Structural Example Using Ansys
Summary
References
Problems
14 Design and Material Selection
14.1 Engineering Design Process
14.2 Material Selection
14.3 Electrical, Mechanical, and Thermophysical Properties of Materials
14.4 Common Solid Engineering Materials
14.5 Some Common Fluid Materials
Summary
References
Problems
15 Design Optimization
15.1 Introduction to Design Optimization
15.2 The Parametric Design Language of Ansys
15.3 Examples of Batch Files
Summary
References
Problems
Appendix A Mechanical Properties of Some Materials
Appendix B Thermophysical Properties of Some Materials
Appendix C Properties of Common Line and Area Shapes
Appendix D Geometrical Properties of Structural Steel Shapes
Appendix E Conversion Factors
Appendix F An Introduction to MATLAB
Index
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
W
Y
Finite Element Analysis
Theory and Application with ANSYS
For these Global Editions, the editorial team at Pearson has collaborated with educators across the world to address a wide range of subjects and requirements, equipping students with the best possible learning tools. This Global Edition preserves the cuttingedge approach and pedagogy of the original, but also features alterations, customization, and adaptation from the North American version.
Global edition
Global edition
Global edition
Finite Element Analysis Theory and Application with ANSYS fourth edition
Saeed Moaveni
fourth edition Moaveni
This is a special edition of an established title widely used by colleges and universities throughout the world. Pearson published this exclusive edition for the benefit of students outside the United States and Canada. If you purchased this book within the United States or Canada you should be aware that it has been imported without the approval of the Publisher or Author. Pearson Global Edition
Moaveni_0273774301_mech.indd 1
26/11/14 2:48 pm
Finite Element Analysis
A01_MOAV4303_04_GE_FM.INDD 1
27/11/14 8:07 AM
A01_MOAV4303_04_GE_FM.INDD 2
27/11/14 8:07 AM
Finite Element Analysis Theory and Application with ANSYS Fourth Edition Global Edition
Saeed Moaveni Minnesota State University, Mankato
Boston Columbus Indianapolis New York San Francisco Hoboken Amsterdam Cape Town Dubai London Madrid Milan Munich Paris Montreal Toronto Delhi Mexico City São Paulo Sydney Hong Kong Seoul Singapore Taipei Tokyo
A01_MOAV4303_04_GE_FM.INDD 3
27/11/14 8:07 AM
Vice President and Editorial Director, ECS: Marcia J. Horton Acquisitions Editor: Norrin Dias Marketing Assistant: Jon Bryant Senior Managing Editor: Scott Disanno Production Program Manager: Clare Romeo Production Project Manager: Jennifer Sargunar Director of Operations: Nick Sklitsis Operations Specialist: Linda Sager Cover Designer: Lumina Datamatics Ltd. Cover Photo: George Spade/123RF
Manager, Rights and Permissions: Rachel Youdelman Photo Permission Coordinator: Rachel Youdelman Image Permission Coordinator: Paul Sarkis FullService Project Management: Jouve India Head of Learning Asset Acquisition, Global Edition: Laura Dent Senior Manufacturing Controller, Global Edition: Trudy Kimber Senior Acquisitions Editor, Global Edition: Priyanka Ahuja Project Editor, Global Edition: Aaditya Bugga
Many of the designations by manufacturers and seller to distinguish their products are claimed as trademarks. Where those designations appear in this book, and the publisher was aware of a trademark claim, the designations have been printed in initial caps or all caps. MATLAB® is a registered trademark of The MathWorks, Inc., 3 Apple Hill Drive, Natick, MA 017602098. Pearson Education Limited Edinburgh Gate Harlow Essex CM20 2JE England and Associated Companies throughout the world Visit us on the World Wide Web at: www.pearsonglobaleditions.com © Pearson Education Limited, 2015 The right of Saeed Moaveni to be identified as the author of this work has been asserted by him in accordance with the Copyright, Designs and Patents Act 1988. Authorized adaptation from the United States edition, entitled Finite Element Analysis: Theory and Application with ANSYS, 4th edition, ISBN 9780133840803, by Saeed Moaveni, published by Pearson Education © 2015. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without either the prior written permission of the publisher or a license permitting restricted copying in the United Kingdom issued by the Copyright Licensing Agency Ltd, Saffron House, 6–10 Kirby Street, London EC1N 8TS. All trademarks used herein are the property of their respective owners. The use of any trademark in this text does not vest in the author or publisher any trademark ownership rights in such trademarks, nor does the use of such trademarks imply any affiliation with or endorsement of this book by such owners. ISBN 10: 0273774301 ISBN 13: 9780273774303 British Library CataloguinginPublication Data A catalogue record for this book is available from the British Library 10 9 8 7 6 5 4 3 2 1 19 18 17 16 15 Typeset in 10 Times Ten LT Std by Jouve India. Printed in Great Britain by CPI Group (UK) Ltd, Croydon, CRO 4YY
A01_MOAV4303_04_GE_FM.INDD 4
27/11/14 8:07 AM
To memories of my mother and father
A01_MOAV4303_04_GE_FM.INDD 5
27/11/14 8:07 AM
A01_MOAV4303_04_GE_FM.INDD 6
27/11/14 8:07 AM
Contents Preface 13 Acknowledgments 17 1 Introduction 21 1.1 Engineering Problems 22 1.2 Numerical Methods 25 1.3 A Brief History of the Finite Element Method and Ansys 26 1.4 Basic Steps in the Finite Element Method 26 1.5 Direct Formulation 28 1.6 Minimum Total Potential Energy Formulation 57 1.7 Weighted Residual Formulations 63 1.8 Verification of Results 68 1.9 Understanding the Problem 69 Summary 74 References 74 Problems 74
2 Matrix Algebra 86 2.1 Basic Definitions 86 2.2 Matrix Addition or Subtraction 89 2.3 Matrix Multiplication 89 2.4 Partitioning of a Matrix 93 2.5 Transpose of a Matrix 97 2.6 Determinant of a Matrix 101 2.7 Solutions of Simultaneous Linear Equations 106 2.8 Inverse of a Matrix 114 2.9 Eigenvalues and Eigenvectors 118 2.10 Using Matlab to Manipulate Matrices 122 2.11 Using Excel to Manipulate Matrices 126 Summary 140 References 141 Problems 141
3 Trusses 145 3.1 Definition of a Truss 145 3.2 Finite Element Formulation 146 3.3 Space Trusses 171 7
A01_MOAV4303_04_GE_FM.INDD 7
27/11/14 8:07 AM
8 Contents
3.4 Overview of the Ansys Program 173 3.5 Examples Using Ansys 181 3.6 Verification of Results 213 Summary 215 References 215 Problems 215
4
Axial Members, Beams, and Frames 225
4.1 Members Under Axial Loading 225 4.2 Beams 233 4.3 Finite Element Formulation of Beams 238 4.4 Finite Element Formulation of Frames 254 4.5 ThreeDimensional Beam Element 260 4.6 An Example Using Ansys 262 4.7 Verification of Results 287 Summary 289 References 290 Problems 291
5 OneDimensional Elements 303 5.1 Linear Elements 303 5.2 Quadratic Elements 307 5.3 Cubic Elements 309 5.4 Global, Local, and Natural Coordinates 312 5.5 Isoparametric Elements 314 5.6 Numerical Integration: Gauss–Legendre Quadrature 316 5.7 Examples of O neDimensional Elements in Ansys 321 Summary 321 References 321 Problems 321
6
Analysis of OneDimensional Problems 328
6.1 Heat Transfer Problems 328 6.2 A Fluid Mechanics Problem 347 6.3 An Example Using Ansys 351 6.4 Verification of Results 366 Summary 367 References 367 Problems 368
7 TwoDimensional Elements 371 7.1 Rectangular Elements 371 7.2 Quadratic Quadrilateral Elements 375
A01_MOAV4303_04_GE_FM.INDD 8
27/11/14 8:07 AM
Contents 9
7.3 Linear Triangular Elements 380 7.4 Quadratic Triangular Elements 385 7.5 Axisymmetric Elements 389 7.6 Isoparametric Elements 394 7.7 TwoDimensional Integrals: Gauss–Legendre Quadrature 397 7.8 Examples of TwoDimensional Elements in Ansys 398 Summary 399 References 399 Problems 400
8 More Ansys 407 8.1 Ansys Program 407 8.2 Ansys Database and Files 408 8.3 Creating a Finite Element Model with Ansys: Preprocessing 410 8.4 hMethod Versus pMethod 424 8.5 Applying Boundary Conditions, Loads, and the Solution 424 8.6 Results of Your Finite Element Model: Postprocessing 427 8.7 Selection Options 432 8.8 Graphics Capabilities 433 8.9 ErrorEstimation Procedures 435 8.10 An Example Problem 437 Summary 451 References 452
9
Analysis of TwoDimensional Heat Transfer Problems 453
9.1 General Conduction Problems 453 9.2 Formulation with Rectangular Elements 460 9.3 Formulation with Triangular Elements 471 9.4 Axisymmetric Formulation of ThreeDimensional Problems 490 9.5 Unsteady Heat Transfer 497 9.6 Conduction Elements Used by Ansys 507 9.7 Examples Using Ansys 508 9.8 Verification of Results 548 Summary 548 References 550 Problems 550
10
Analysis of TwoDimensional Solid Mechanics Problems 562
10.1 Torsion of Members with Arbitrary CrossSection Shape 562 10.2 PlaneStress Formulation 578 10.3 Isoparametric Formulation: Using a Quadrilateral Element 586 10.4 Axisymmetric Formulation 593 10.5 Basic Failure Theories 595
A01_MOAV4303_04_GE_FM.INDD 9
27/11/14 8:07 AM
10 Contents
10.6 Examples Using Ansys 596 10.7 Verification of Results 618 Summary 618 References 620 Problems 620
11
Dynamic Problems 629
11.1 Review of Dynamics 629 11.2 Review of Vibration of Mechanical and Structural Systems 643 11.3 Lagrange’s Equations 660 11.4 Finite Element Formulation of Axial Members 662 11.5 Finite Element Formulation of Beams and Frames 671 11.6 Examples Using Ansys 685 Summary 704 References 704 Problems 704
12 Analysis of Fluid Mechanics Problems 711 12.1 Direct Formulation of Flow Through Pipes 711 12.2 Ideal Fluid Flow 723 12.3 Groundwater Flow 729 12.4 Examples Using Ansys 732 12.5 Verification of Results 753 Summary 754 References 755 Problems 756
13 ThreeDimensional Elements 761 13.1 The FourNode Tetrahedral Element 761 13.2 Analysis of ThreeDimensional Solid Problems Using FourNode Tetrahedral Elements 764 13.3 The EightNode Brick Element 769 13.4 The TenNode Tetrahedral Element 771 13.5 The TwentyNode Brick Element 772 13.6 Examples of ThreeDimensional Elements in Ansys 774 13.7 Basic SolidModeling Ideas 778 13.8 A Thermal Example Using Ansys 789 13.9 A Structural Example Using Ansys 806 Summary 819 References 819 Problems 819
A01_MOAV4303_04_GE_FM.INDD 10
27/11/14 8:07 AM
Contents 11
14 Design and Material Selection 828 14.1 Engineering Design Process 829 14.2 Material Selection 832 14.3 Electrical, Mechanical, and Thermophysical Properties of Materials 833 14.4 Common Solid Engineering Materials 835 14.5 Some Common Fluid Materials 842 Summary 844 References 844 Problems 844
15
Design Optimization 846
15.1 Introduction to Design Optimization 846 15.2 The Parametric Design Language of Ansys 850 15.3 Examples of Batch Files 852 Summary 863 References 864 Problems 864
Appendix A Mechanical Properties of Some Materials 865 Appendix B Thermophysical Properties of Some Materials 869 Appendix C Properties of Common Line and Area Shapes 871 Appendix D Geometrical Properties of Structural Steel Shapes 875 Appendix E Conversion Factors 879 Appendix F An Introduction to MATLAB 881 Index 915
A01_MOAV4303_04_GE_FM.INDD 11
27/11/14 8:07 AM
A01_MOAV4303_04_GE_FM.INDD 12
27/11/14 8:07 AM
Preface
Changes in the Fourth Edition The fourth edition, consisting of 15 chapters, includes a number of new additions and changes that were incorporated in response to ANSYS revisions and suggestions and requests made by professors, students, and professionals using the third edition of the book. The major changes include:
• Explanation of the changes that were made in the ANSYS’s newest release (Chapters 3 and 8) • Explanation of new element type capabilities (Chapters 3, 4, 6, 8 through 13, and 15) • A new comprehensive example problem that demonstrates the use of BEAM188 element in modeling beam and frame problems (Chapter 4) • Modification of twenty example problems to incorporate new ANSYS element types (Chapters 3, 4, 6, 8 through 13, and 15) • Eight new comprehensive example problems that show in great detail how to use Excel to solve different types of finite element problems (Chapters 2 through 6 and 9 through 12) • More detail on theory and expanded derivations • Explanation of new MATLAB revisions in Appendix F Organization There are many good textbooks already in existence that cover the theory of finite element methods for advanced students. However, none of these books incorporate ANSYS as an integral part of their materials to introduce finite element modeling to undergraduate students and newcomers. In recent years, the use of finite element analysis (FEA) as a design tool has grown rapidly. E asytouse, comprehensive packages such as ANSYS, a generalpurpose finite element computer program, have become common tools in the hands of design engineers. Unfortunately, many engineers who lack the proper training or understanding of the underlying concepts have been using these tools. This introductory book is written to assist engineering students and practicing engineers new to the field of finite element modeling to gain a clear understanding of the basic c oncepts. The text offers insight into the theoretical aspects of FEA and also covers some practical aspects of modeling. Great care has been exercised to avoid overwhelming students with theory, yet enough theoretical background is offered to allow individuals to use ANSYS intelligently and effectively. ANSYS is an 13
A01_MOAV4303_04_GE_FM.INDD 13
27/11/14 8:07 AM
14 Preface
integral part of this text. In each chapter, the relevant basic theory is discussed first and demonstrated using simple problems with hand calculations. These problems are followed by examples that are solved using ANSYS. Exercises in the text are also presented in this manner. Some exercises require manual calculations, while others, more complex in nature, require the use of ANSYS. The simpler h andcalculation problems will enhance students’ understanding of the concepts by encouraging them to go through the necessary steps in a FEA. Design problems are also included at the end of Chapters 3, 4, 6, and 9 through 14. Various sources of error that can contribute to incorrect results are discussed. A good engineer must always find ways to check the results. While experimental testing of models may be the best way, such testing may be expensive or time consuming. Therefore, whenever possible, throughout this text emphasis is placed on doing a “sanity check” to verify one’s FEA. A section at the end of each appropriate chapter is devoted to possible approaches for verifying ANSYS results. Another unique feature of this book is that the last two chapters are devoted to the introduction of design, material selection, optimization, and parametric programming with ANSYS. The book is organized into 15 chapters. Chapter 1 reviews basic ideas in finite element analysis. Common formulations, such as direct, potential energy, and weighted residual methods, are discussed. Chapter 2 provides a comprehensive review of matrix algebra. Chapter 3 deals with the analysis of trusses, because trusses offer economical solutions to many engineering structural problems. An overview of the ANSYS program is given in Chapter 3 so that students can begin to use ANSYS right away. Finite element formulation of members under axial loading, beams, and frames are introduced in Chapter 4. Chapter 5 lays the foundation for analysis of onedimensional problems by introducing o nedimensional linear, quadratic, and cubic elements. Global, local, and natural coordinate systems are also discussed in detail in Chapter 5. An introduction to isoparametric formulation and numerical integration by Gauss– Legendre formulae is also presented in Chapter 5. Chapter 6 considers Galerkin formulation of onedimensional heat transfer and fluid problems. Twodimensional linear and higher order elements are introduced in Chapter 7. Gauss–Legendre formulae for twodimensional integrals are also presented in Chapter 7. In Chapter 8 the essential capabilities and the organization of the ANSYS program are covered. The basic steps in creating and a nalyzing a model with ANSYS is discussed in detail. Chapter 9 includes the analysis of twodimensional heat transfer problems with a section devoted to unsteady situations. Chapter 10 provides an analysis of torsion of noncircular shafts and plane stress problems. Dynamic problems are explored in Chapter 11. Review of dynamics and vibrations of mechanical and structural systems are also given in this chapter. In Chapter 12, t wodimensional, ideal fluidmechanics problems are analyzed. Direct formulation of the piping network problems and underground seepage flow are also discussed. Chapter 13 provides a discussion on threedimensional elements and formulations. This chapter also presents basic ideas regarding topdown and bottomup solid modeling methods. The last two chapters of the book are devoted to design and
A01_MOAV4303_04_GE_FM.INDD 14
27/11/14 8:07 AM
Preface 15
optimization ideas. Design process and m aterial selection are explained in Chapter 14. Design optimization ideas and parametric p rogramming are discussed in Chapter 15. Examples of ANSYS batch files are also given in Chapter 15. Each chapter begins by stating the objectives and c oncludes by summarizing what the reader should have gained from studying that chapter. The examples that are solved using ANSYS show in great detail how to use ANSYS to model and analyze a variety of engineering problems. Chapter 8 is also written such that it can be taught right away if the instructor sees the need to start with ANSYS. A brief review of appropriate fundamental principles in solid mechanics, heat transfer, dynamics, and fluid mechanics is also provided throughout the book. Additionally, when appropriate, students are warned about becoming too quick to generate finite element models for problems for which there exist simple analytical solutions. Mechanical and thermophysical properties of some common materials used in e ngineering are given in Appendices A and B. Appendices C and D give properties of c ommon area shapes and properties of structural steel shapes, respectively. A comprehensive introduction to MATLAB is given in Appendix F. Finally, a Web site at http://www.pearsonglobaleditions.com/moaveni will be maintained for the following purposes: (1) to share any changes in the upcoming versions of ANSYS; (2) to share additional information on upcoming text revisions; (3) to provide additional homework problems and design problems; and (4) although I have done my best to eliminate errors and mistakes, as is with most books, some errors may still exist. I will post the corrections that are brought to my attention at the site. The Web site will be accessible to all instructors and students. Thank you for considering this book and I hope you enjoy the fourth edition. Saeed Moaveni
A01_MOAV4303_04_GE_FM.INDD 15
27/11/14 8:07 AM
A01_MOAV4303_04_GE_FM.INDD 16
27/11/14 8:07 AM
Acknowledgments
I would like to express my sincere gratitude to ANSYS, Inc. for providing the photographs that appear on page 27 of this book. Descriptions for these photographs are given in Chapter 1 with the images. I would also like to thank ANSYS, Inc. for giving me permission to adapt material from various ANSYS documents, related to capabilities and the organization of ANSYS. The essential capabilities and organizations of ANSYS are covered in Chapters 3, 8, 13, and 15. As I have mentioned in the Preface, there are many good published books in finite element analysis. When writing this book, several of these books were consulted. They are cited at the end of each appropriate chapter. The reader can benefit from referring to these books and articles. I am also thankful to all reviewers who offered general and specific comments. Global Edition Pearson would like to thank and acknowledge the following people for their work on the Global Edition: Contributor Jagadeesha T, National Institute of Technology—Calicut Reviewers Swarup Bag, Indian Institute of Technology—Guwahati Avinash Parashar, Indian Institute of Technology—Roorkee Mokhtar B Awang, Universiti Teknologi Petronas—Malaysia Yash Parikh, Symbiosis Institute of Technology—Pune
17
A01_MOAV4303_04_GE_FM.INDD 17
27/11/14 8:07 AM
A01_MOAV4303_04_GE_FM.INDD 18
27/11/14 8:07 AM
Finite Element Analysis
A01_MOAV4303_04_GE_FM.INDD 19
27/11/14 8:07 AM
A01_MOAV4303_04_GE_FM.INDD 20
27/11/14 8:07 AM
C h a p t e r
1
Introduction
The finite element method is a numerical procedure that can be used to obtain solutions to a large class of engineering problems involving stress analysis, heat transfer, electromagnetism, and fluid flow. This book was written to help you gain a clear understanding of the fundamental concepts of finite element modeling. Having a clear understanding of the basic concepts will enable you to use a generalpurpose finite element software, such as ANSYS, effectively. ANSYS is an integral part of this text. In each chapter, the relevant basic theory behind each respective concept is discussed first. This discussion is followed by examples that are solved using ANSYS. Throughout this text, emphasis is placed on methods by which you may verify your findings from finite element analysis (FEA). In addition, at the end of particular chapters, a section is devoted to the approaches you should consider to verify results generated by using ANSYS. Some of the exercises provided in this text require manual calculations. The purpose of these exercises is to enhance your understanding of the concepts by encouraging you to go through the necessary steps of FEA. This book can also serve as a reference text for readers who may already be design engineers who are beginning to get involved in finite element modeling and need to know the underlying concepts of FEA. The objective of this chapter is to introduce you to basic concepts in finite element formulation, including direct formulation, the minimum potential energy theorem, and the weighted residual methods. The main topics of Chapter 1 include the following: 1.1 Engineering Problems 1.2 Numerical Methods 1.3 A Brief History of the Finite Element Method and ANSYS 1.4 Basic Steps in the Finite Element Method 1.5 Direct Formulation 1.6 Minimum Total Potential Energy Formulation 1.7 Weighted Residual Formulations 1.8 Verification of Results 1.9 Understanding the Problem
M01_MOAV4303_04_GE_C01.INDD 21
27/11/14 9:32 AM
22 Chapter 1 Introduction
1.1 Engineering Problems In general, engineering problems are mathematical models of physical situations. Mathematical models of many engineering problems are differential equations with a set of corresponding boundary and/or initial conditions. The differential equations are derived by applying the fundamental laws and principles of nature to a system or a control volume. These governing equations represent balance of mass, force, or energy. When possible, the exact solution of these equations renders detailed behavior of a system under a given set of conditions, as shown by some examples in Table 1.1. The analytical solutions are composed of two parts: (1) a homogenous part and (2) a particular part. In any given engineering problem, there are two sets of design parameters that influence the way in which a system behaves. First, there are those parameters that
Table 1.1 Examples of governing differential equations, boundary conditions, initial conditions, and exact solutions for some engineering problems
Problem Type A beam:
Governing Equation, Boundary Conditions, or Initial Conditions wX(L  X) d 2Y = 2 2 dX Boundary conditions: at X = 0, Y = 0 and at X = L, Y = 0
Deflection of the beam Y as the function of distance X:
d 2y
The position of the mass y as the function of time:
EI
Y
w
X
E, I L
An elastic system:
dt y0
y
m k
A fin:
Tq , h
Tbase
AC
Solution
2
+ v2n y = 0
Y =
w (  X 4 + 2LX 3  L3X) 24EI
k m Initial conditions: at time t = 0, y = y0 and dy at time t = 0, = 0 dt
y(t) = y0 cos vnt
hp d 2T (T  T∞) = 0 kAc dX 2
Temperature distribution along the fin as the function of X:
Boundary conditions:
T = T ∞ + (Tbase  T ∞ )e2kA X
where v2n =
hp
c
at X = 0, T = Tbase as L S ∞, T = T∞
X
P = Perimeter L
M01_MOAV4303_04_GE_C01.INDD 22
27/11/14 9:32 AM
Section 1.1 Engineering Problems 23
provide information regarding the natural behavior of a given system. These parameters include material and geometric properties such as modulus of elasticity, thermal conductivity, viscosity, and area, and second moment of area. Table 1.2 summarizes the physical properties that define the natural characteristics of various problems.
Table 1.2 Physical properties characterizing various engineering systems Problem Type
Examples of Parameters That Characterize a System
Solid Mechanics Examples Load
E, A
Modulus of elasticity, E; member length, L; crosssectional area, A
A truss Load
E
Modulus of elasticity, E; length, L; crosssectional area, A
An elastic plate Load
Modulus of elasticity, E; member length, L; second moment of area, I
E,I A beam Torque
G,J A shaft
Modulus of rigidity, G; member length, L; polar moment of inertia of the area, J
Heat Transfer Examples High temperature
Heat flow
K Low temperature A wall
Thermal conductivity, K; thickness, L; area, A continued
M01_MOAV4303_04_GE_C01.INDD 23
27/11/14 9:32 AM
24 Chapter 1 Introduction
Table 1.2 Continued Problem Type
Examples of Parameters That Characterize a System
K
Thermal conductivity, K; perimeter, P; crosssectional area, A
Fins Fluid Flow Examples L D High pressure
Low pressure
Pipe networks
Water
Fluid viscosity, m; pipe roughness, e; pipe diameter, D; pipe length, L
Water
Concrete dam Porous medium
k
A concrete dam
Soil permeability, k
Electrical and Magnetism Problems + Voltage
R2
R1

Electrical network
Resistance, R
Stator Rotor
Magnetic field of an electric motor
M01_MOAV4303_04_GE_C01.INDD 24
Permeability, m
27/11/14 9:32 AM
Section 1.2 Numerical Methods 25
Table 1.3 Parameters causing disturbances in various engineering systems Problem Type
Examples of Parameters that Produce Disturbances in a System
Solid Mechanics
External forces and moments; support excitation
Heat Transfer
Temperature difference; heat input
Fluid Flow and Pipe Networks
Pressure difference; rate of flow
Electrical Network
Voltage difference
On the other hand, there are parameters that produce disturbances in a system. These types of parameters are summarized in Table 1.3. Examples of these parameters include external forces, moments, temperature difference across a medium, and pressure difference in fluid flow. The system characteristics as shown in Table 1.2 dictate the natural behavior of a system, and they always appear in the homogenous part of the solution of a governing differential equation. In contrast, the parameters that cause the disturbances appear in the particular solution. It is important to understand the role of these parameters in finite element modeling in terms of their respective appearances in stiffness or conductance matrices and load or forcing matrices. The system characteristics will always show up in the stiffness matrix, conductance matrix, or resistance matrix, whereas the disturbance parameters will always appear in the load matrix. We will explain the concepts of stiffness, conductance, and load matrices in Section 1.5. 1.2 Numerical Methods There are many practical engineering problems for which we cannot obtain exact solutions. This inability to obtain an exact solution may be attributed to either the complex nature of governing differential equations or the difficulties that arise from dealing with the boundary and initial conditions. To deal with such problems, we resort to numerical approximations. In contrast to analytical solutions, which show the exact behavior of a system at any point within the system, numerical solutions approximate exact solutions only at discrete points, called nodes. The first step of any numerical procedure is discretization. This process divides the medium of interest into a number of small subregions (elements) and nodes. There are two common classes of numerical methods: (1) finite difference methods and (2) finite element methods. With finite difference methods, the differential equation is written for each node, and the derivatives are replaced by difference equations. This approach results in a set of simultaneous linear equations. Although finite difference methods are easy to understand and employ in simple problems, they become difficult to apply to problems with complex geometries or complex boundary conditions. This situation is also true for problems with nonisotropic material properties. In contrast, the finite element method uses integral formulations rather than difference equations to create a system of algebraic equations. Moreover, a continuous function is assumed to represent the approximate solution for each element. The complete solution is then generated by connecting or assembling the individual solutions, allowing for continuity at the interelemental boundaries.
M01_MOAV4303_04_GE_C01.INDD 25
27/11/14 9:32 AM
26 Chapter 1 Introduction
1.3 A Brief History* of the Finite Element Method and ANSYS The finite element method is a numerical procedure that can be applied to obtain solutions to a variety of problems in engineering. Steady, transient, linear, or nonlinear problems in stress analysis, heat transfer, fluid flow, and electromagnetism problems may be analyzed with finite element methods. The origin of the modern finite element method may be traced back to the early 1900s when some investigators approximated and modeled elastic continua using discrete equivalent elastic bars. However, Courant (1943) has been credited with being the first person to develop the finite element method. In a paper published in the early 1940s, Courant used piecewise polynomial interpolation over triangular subregions to investigate torsion problems. The next significant step in the utilization of finite element methods was taken by Boeing in the 1950s when Boeing, followed by others, used triangular stress elements to model airplane wings. Yet, it was not until 1960 that Clough made the term finite element popular. During the 1960s, investigators began to apply the finite element method to other areas of engineering, such as heat transfer and seepage flow problems. Zienkiewicz and Cheung (1967) wrote the first book entirely devoted to the finite element method in 1967. In 1971, ANSYS was released for the first time. ANSYS is a comprehensive generalpurpose finite element computer program that contains more than 100,000 lines of code. ANSYS is capable of performing static, dynamic, heat transfer, fluid flow, and electromagnetism analyses. ANSYS has been a leading FEA program for over 40 years. The current version of ANSYS has a completely new look, with multiple windows incorporating a graphical user interface (GUI), pulldown menus, dialog boxes, and a tool bar. Today, you will find ANSYS in use in many engineering fields, including aerospace, automotive, electronics, and nuclear. In order to use ANSYS or any other “canned” FEA computer program intelligently, it is imperative that one first fully understands the underlying basic concepts and limitations of the finite element methods. ANSYS is a very powerful and impressive engineering tool that may be used to solve a variety of problems (see Table 1.4). However, a user without a basic understanding of the finite element methods will find himself or herself in the same predicament as a computer technician with access to many impressive instruments and tools, but who cannot fix a computer because he or she does not understand the inner workings of a computer! 1.4 Basic Steps in the Finite Element Method The basic steps involved in any finite element analysis consist of the following: Preprocessing Phase 1. Create and discretize the solution domain into finite elements; that is, subdivide the problem into nodes and elements.
*See Cook et al. (1989) for more detail.
M01_MOAV4303_04_GE_C01.INDD 26
27/11/14 9:32 AM
Section 1.4 Basic Steps in the Finite Element Method 27
Table 1.4 Examples of the capabilities of ANSYS*
A V6 engine used in frontwheeldrive automobiles analyses were conducted by Analysis & Design Appl. Co. Ltd. (ADAPCO) on behalf of a major U.S. automobile manufacturer to improve product performance. Contours of thermal stress in the engine block are shown in the figure above.
Large deflection capabilities of ANSYS were utilized by engineers at Today’s Kids, a toy manufacturer, to confirm failure locations on the company’s play slide, shown in the figure above, when the slide is subjected to overload. This nonlinear analysis capability is required to detect these stresses because of the product’s structural behavior.
Electromagnetic capabilities of ANSYS, which include the use of both vector and scalar potentials interfaced through a specialized element, as well as a threedimensional graphics representation of farfield decay through infinite boundary elements, are depicted in this analysis of a bath plate, shown in the figure above. Isocontours are used to depict the intensity of the Hfield.
Structural Analysis Engineering Corporation used ANSYS to determine the natural frequency of a rotor in a diskbrake assembly. In this analysis, 50 modes of vibration, which are considered to contribute to brake squeal, were found to exist in the lighttruck brake rotor.
*Photographs courtesy of ANSYS, Inc., Canonsburg, PA.
M01_MOAV4303_04_GE_C01.INDD 27
27/11/14 9:32 AM
28 Chapter 1 Introduction
2. Assume a shape function to represent the physical behavior of an element; that is, a continuous function is assumed to represent the approximate behavior (solution) of an element. 3. Develop equations for an element. 4. Assemble the elements to present the entire problem. Construct the global stiffness matrix. 5. Apply boundary conditions, initial conditions, and loading. Solution Phase 6. Solve a set of linear or nonlinear algebraic equations simultaneously to obtain nodal results, such as displacement values at different nodes or temperature values at different nodes in a heat transfer problem. Postprocessing Phase 7. Obtain other important information. At this point, you may be interested in values of principal stresses, heat fluxes, and so on. In general, there are several approaches to formulating finite element problems: (1) direct formulation, (2) the minimum total potential energy formulation, and (3) weighted residual formulations. Again, it is important to note that the basic steps involved in any finite element analysis, regardless of how we generate the finite element model, will be the same as those listed above. 1.5 Direct Formulation The following problem illustrates the steps and the procedure involved in direct formulation. Example 1.1 Consider a bar with a variable cross section supporting a load P, as shown in Figure 1.1. The bar is fixed at one end and carries the load P at the other end. Let us designate the width of the bar at the top by w1, at the bottom by w2, its thickness by t, and its length by L. The bar’s modulus of elasticity will be denoted by E. We are interested in approximating how much the bar will deflect at various points along its length when it is subjected to the load P. We will neglect the weight of the bar in the following analysis, assuming that the applied load is considerably larger than the weight of the bar: Preprocessing Phase 1. Discretize the solution domain into finite elements. We begin by subdividing the problem into nodes and elements. In order to highlight the basic steps in a finite element analysis, we will keep this problem simple and thus represent it by a model that has five nodes and four elements, as shown in Figure 1.2. However, note that we can increase the accuracy of our results by generating a model with additional nodes and elements. This task is left as an exercise for you to complete. (See Problem 1 at the end of this chapter.) The given bar
M01_MOAV4303_04_GE_C01.INDD 28
27/11/14 9:32 AM
Section 1.5 Direct Formulation 29 w1
L y w2
P
Figure 1.1 A bar under axial loading.
is modeled using four individual segments (elements), with each segment having a uniform cross section. The crosssectional area of each element is represented by an average area of the cross sections at the nodes that make up (define) the element. This model is shown in Figure 1.2. 2. Assume a solution that approximates the behavior of an element. In order to study the behavior of a typical element, let’s consider the deflection of a solid member with a uniform cross section A that has a length / when subjected to a force F, as shown in Figure 1.3. The average stress s in the member is given by
s =
F (1.1) A
The average normal strain e of the member is defined as the change in length ∆/ per unit original length / of the member:
e =
1
L
4
/1
A1
2 3
∆/ (1.2) /
/2
A2 A3 A4
5
/3 /4
u1
1 Element 1
u2
2 Element 2
u3
3 Element 3
u4
4 Element 4 5
u5 P
P
P
Figure 1.2 Subdividing the bar into elements and nodes.
M01_MOAV4303_04_GE_C01.INDD 29
27/11/14 9:32 AM
30 Chapter 1 Introduction
/ ∆/
kequivalent x Figure 1.3 A solid member of uniform cross section subjected to a force F.
F
F
Over the elastic region, the stress and strain are related by Hooke’s law, according to the equation
s = Ee (1.3)
where E is the modulus of elasticity of the material. Combining Eqs. (1.1), (1.2), and (1.3) and simplifying, we have F = ¢
AE ≤ ∆/ (1.4) /
Note that Eq. (1.4) is similar to the equation for a linear spring, F = kx. Therefore, a centrally loaded member of uniform cross section may be modeled as a spring with an equivalent stiffness of
keq =
AE (1.5) /
Turning our attention to Example 1.1, we note once again that the bar’s cross section varies in the ydirection. As a first approximation, we model the bar as a series of centrally loaded members with different cross sections, as shown in Figure 1.2. Thus, the bar is represented by a model consisting of four elastic springs (elements) in series, and the elastic behavior of an element with nodes i and i + 1 is modeled by an equivalent linear spring according to the equation
f = keq(ui + 1  ui) =
Aavg E /
(ui + 1  ui) =
(Ai + 1 + Ai)E (ui + 1  ui) (1.6) 2/
where ui + 1 and ui are the deflections at nodes i + 1 and i, and the equivalent element stiffness is given by
keq =
(Ai + 1 + Ai)E (1.7) 2/
Ai and Ai + 1 are the crosssectional areas of the member at nodes i and i + 1 respectively, and / is the length of the element. Employing the above model, let us consider the forces acting on each node. The freebody diagram of nodes, which shows the forces acting on nodes 1 through 5 of this model, is depicted in Figure 1.4.
M01_MOAV4303_04_GE_C01.INDD 30
27/11/14 9:32 AM
Section 1.5 Direct Formulation 31 R1 Node 1: k1(u2 – u1) k1(u2 – u1) Node 2: k2(u3 – u2) k2(u3 – u2) Node 3: k3(u4 – u3) k3(u4 – u3) Node 4: k4(u5 – u4) k4(u5 – u4) Node 5: P
Figure 1.4 Free body diagram of the nodes in Example 1.1.
Static equilibrium requires that the sum of the forces acting on each node be zero. This requirement creates the following five equations:
node 1: R1  k1(u2  u1) = 0
node 2: k1(u2  u1)  k2(u3  u2) = 0
node 3: k2(u3  u2)  k3(u4  u3) = 0
node 4: k3(u4  u3)  k4(u5  u4) = 0
node 5: k4(u5  u4)  P = 0
(1.8)
Rearranging the equilibrium equations given by Eq. (1.8) by separating the reaction force R1 and the applied external force P from the internal forces, we have k1 u1 k1u1
M01_MOAV4303_04_GE_C01.INDD 31
k1u2 +k1u2
+k2u2 k2u2
k2u3 +k2u3 +k3u3 k3u3
k3u4 +k3u4
+k4u4 k4u4
k4u5 +k4u5
= = = = =
R1 0 0 (1.9) 0 P
27/11/14 9:32 AM
32 Chapter 1 Introduction
Presenting the equilibrium equations of Eq. (1.9) in a matrix form, we have k1 k1 E 0 0 0
k1 k 1 + k2 k2 0 0
0 k2 k 2 + k3 k3 0
0 0 k3 k 3 + k4 k4
0 u1 R1 0 u2 0 0 U e u3 u = e 0 u (1.10) k4 u4 0 k4 u5 P
It is also important to distinguish between the reaction forces and the applied loads in the load matrix. To do so, the matrix relation of Eq. (1.10) is written as R1 k1 0 k1 e 0 u = E 0 0 0 0 0
k1 k 1 + k2 k2 0 0
0 k2 k 2 + k3 k3 0
0 0 k3 k 3 + k4 k4
0 u1 0 0 u2 0 0 U e u3 u  e 0 u (1.11) k4 u4 0 k4 u5 P
We can readily show that under additional nodal loads and other fixed boundary conditions, the relationship given by Eq. (1.11) can be put into the general form which stands for
5 R 6 = [K] 5 u 6  5 F6 (1.12)
5 reaction matrix6 = [stiffness matrix]{displacement matrix}  5 load matrix6
Note the difference between applied load matrix 5 F6 and the reaction force matrix 5 R 6 . Turning our attention to Example 1.1 again, we find that because the bar is fixed at the top, the displacement of node 1 is zero. Hence, there are only four unknown nodal displacement values, u2, u3, u4, and u5. The reaction force at node 1, R1, is also unknown—all together, there are five unknowns. Because there are five equilibrium equations, as given by Eq. (1.11), we should be able to solve for all of the unknowns. However, it is important to note that even though the number of unknowns match the number of equations, the system of equations contains two different types of unknowns—displacement and reaction force. In order to eliminate the need to consider the unknown reaction force simultaneously and focus first on unknown displacements, we make use of the known boundary condition and replace the first row of Eq. (1.10) with a row that reads u1 = 0. The application of the boundary condition u1 = 0 eliminates the need to consider the unknown reaction force in our system of equations and creates a set of equations with the displacements being the only unknowns.
M01_MOAV4303_04_GE_C01.INDD 32
27/11/14 9:32 AM
Section 1.5 Direct Formulation 33
Thus, application of the boundary condition leads to the following matrix equation:
1 k1 E 0 0 0
0 k 1 + k2 k2 0 0
0 k2 k 2 + k3 k3 0
0 0 k3 k 3 + k4 k4
0 u1 0 0 u2 0 0 U e u3 u = e 0 u (1.13) k4 u4 0 k4 u5 P
The solution of the above matrix yields the nodal displacement values. It should be clear from the above explanation and examining Eq. (1.13) that for solid mechanics problems, the application of boundary conditions to the finite element formulations transforms the system of equations as given by Eq. (1.11) to a new general form that is made up of only the stiffness matrix, the displacement matrix, and the load matrix: [stiffness matrix]5 displacement matrix6 = 5 load matrix6
After we solve for the nodal displacement values, from the above relationship, we use Eq. (1.12) to solve for the reaction force(s). In the next section, we will develop the general elemental stiffness matrix and discuss the construction of the global stiffness matrix by inspection. 3. Develop equations for an element. Because each of the elements in Example 1.1 has two nodes, and with each node we have associated a displacement, we need to create two equations for each element. These equations must involve nodal displacements and the element’s stiffness. Consider the internally transmitted forces fi and fi + 1 and the end displacements ui and ui + 1 of an element, which are shown in Figure 1.5. Static equilibrium conditions require that the sum of fi and fi + 1 be zero. Note that the sum of fi and fi + 1 is zero regardless of which representation of Figure 1.5 is selected. However, for the sake of consistency in the forthcoming derivation, we will use the representation given by Figure 1.5(b), so that fi and fi + 1 are given in fi = keq(ui+1  ui)
fi = keq(ui  ui+1)
Node i
ui
Node i
ui
y
OR
ui+1
Node i + 1 fi+1 = keq(ui+1  ui) (a)
ui+1
Node i + 1 fi+1 = keq(ui+1  ui) (b)
Figure 1.5 Internally transmitted forces through an arbitrary element.
M01_MOAV4303_04_GE_C01.INDD 33
27/11/14 9:32 AM
34 Chapter 1 Introduction
the positive ydirection. Thus, we write the transmitted forces at nodes i and i + 1 according to the following equations: fi = keq(ui  ui + 1) fi + 1 = keq(ui + 1  ui) (1.14)
Equation (1.14) can be expressed in a matrix form by
b
keq u R b i r (1.15) keq ui + 1
fi k r = J eq fi + 1 keq
4. Assemble the elements to present the entire problem. Applying the elemental description given by Eq. (1.15) to all elements and assembling them (putting them together) will lead to the formation of the global stiffness matrix. The stiffness matrix for element (1) is given by [K](1) = J
k1 R k1
k1 k1
and its position in the global stiffness matrix is given by
[K](1G)
k1 k1 0 0 0
k1 k1 = E 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0U 0 0
u1 u2 u3 u4 u5
The nodal displacement matrix is shown alongside the position of element 1 in the global stiffness matrix to aid us to observe the contribution of a node to its neighboring elements. Similarly, for elements (2), (3), and (4), we have [K](2) = J
k2 R k2
k2 k2
and its position in the global matrix
[K](2G)
0 0 = E0 0 0
[K](3) = J
0 k2 k2 0 0
0 k2 k2 0 0
0 0 0 0 0
0 0 0U 0 0
u1 u2 u3 u4 u5
0 0 k3 k3 0
0 0 0U 0 0
u1 u2 u3 u4 u5
k3 R k3
k3 k3
and its position in the global matrix
[K](3G)
M01_MOAV4303_04_GE_C01.INDD 34
0 0 = E0 0 0
0 0 0 0 0
0 0 k3 k3 0
27/11/14 9:32 AM
Section 1.5 Direct Formulation 35
and [K](4) = J
k4 R k4
k4 k4
and its position in the global matrix
[K](4G)
0 0 = E0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 k4 k4
0 0 0 U k4 k4
u1 u2 u3 u4 u5
The final global stiffness matrix is obtained by assembling, or adding together, each element’s position in the global stiffness matrix:
[K](G) = [K](1G) + [K](2G) + [K](3G) + [K](4G)
[K](G)
k1 k1 = E 0 0 0
k1 k 1 + k2 k2 0 0
0 k2 k 2 + k3 k3 0
0 0 k3 k 3 + k4 k4
0 0 0 U (1.16) k4 k4
Note that the global stiffness matrix obtained using elemental description, as given by Eq. (1.16), is identical to the global stiffness matrix we obtained earlier from the analysis of the freebody diagrams of the nodes, as given by the lefthand side of Eq. (1.10). 5. Apply boundary conditions and loads. The bar is fixed at the top, which leads to the boundary condition u1 = 0. The external load P is applied at node 5. Applying these conditions results in the following set of linear equations.
1 k1 E 0 0 0
0 k 1 + k2 k2 0 0
0 k2 k 2 + k3 k3 0
0 0 k3 k 3 + k4 k4
0 u1 0 0 u2 0 0 U e u3 u = e 0 u (1.17) k4 u4 0 k4 u5 P
Again, note that the first row of the matrix in Eq. (1.17) must contain a 1 followed by four 0s to read u1 = 0, the given boundary condition. As explained earlier, also note that in solid mechanics problems, the finite element formulation will always lead to the following general form: [stiffness matrix]5 displacement matrix6 = 5 load matrix6 M01_MOAV4303_04_GE_C01.INDD 35
27/11/14 9:32 AM
36 Chapter 1 Introduction
Solution Phase 6. Solve a system of algebraic equations simultaneously. In order to obtain numerical values of the nodal displacements, let us assume that E = 10.4 * 106 lb/in2 (aluminum), w1 = 2 in, w2 = 1 in, t = 0.125 in, L = 10 in, and P = 1000 lb. You may consult Table 1.5 while working toward the solution. Table 1.5 Properties of the elements in Example 1.1
Element
Nodes
Average CrossSectional Area (in2)
1 2 3 4
1 2 2 3 3 4 4 5
0.234375 0.203125 0.171875 0.140625
Length (in) 2.5 2.5 2.5 2.5
Modulus of Elasticity (lb/in2) 10.4 10.4 10.4 10.4
* * * *
Element’s Stiffness Coefficient (lb/in)
106 106 106 106
975 845 715 585
* * * *
103 103 103 103
The variation of the crosssectional area of the bar in the ydirection can be expressed by: A(y) = ¢ w1 + ¢
(1  2) w2  w1 ≤y≤t = ¢ 2 + y≤(0.125) = 0.25  0.0125y (1.18) L 10
Using Eq. (1.18), we can compute the crosssectional areas at each node: A1 = 0.25 in2
A2 = 0.25  0.0125(2.5) = 0.21875 in2
A3 = 0.25  0.0125(5.0) = 0.1875 in2
A4 = 0.25  0.0125(7.5) = 0.15625 in2
A5 = 0.125 in2 Next, the equivalent stiffness coefficient for each element is computed from the equations keq =
M01_MOAV4303_04_GE_C01.INDD 36
(Ai + 1 + Ai)E 2/
k1 =
(0.21875 + 0.25)(10.4 * 106) lb = 975 * 103 2(2.5) in
k2 =
(0.1875 + 0.21875)(10.4 * 106) lb = 845 * 103 2(2.5) in
k3 =
(0.15625 + 0.1875)(10.4 * 106) lb = 715 * 103 2(2.5) in
k4 =
(0.125 + 0.15625)(10.4 * 106) lb = 585 * 103 2(2.5) in
27/11/14 9:32 AM
Section 1.5 Direct Formulation 37
and the elemental matrices are [K](1) = J
k1 k1
k1 975 R = 103 J k1 975
975 R 975
[K](2) = J
k2 k2
k2 845 R = 103 J k2 845
845 R 845
[K](3) = J
k3 k3
k3 715 R = 103 J k3 715
715 R 715
[K](4) = J
k4 k4
k4 585 R = 103 J k4 585
585 R 585
Assembling the elemental matrices leads to the generation of the global stiffness matrix:
[K](G)
975 975 + 845 845 0 0
975 975 = 103 E 0 0 0
0 845 845 + 715 715 0
0 0 715 715 + 585 585
0 0 0 U 585 585
Applying the boundary condition u1 = 0 and the load P = 1000 lb, we get 1 975 103 E 0 0 0
0 1820 845 0 0
0 845 1560 715 0
0 0 715 1300 585
0 u1 0 0 u2 0 0 U e u3 u = e 0 u 585 u4 0 585 u5 103
Because in the second row, the 975 coefficient gets multiplied by u1 = 0, we need only to solve the following 4 * 4 matrix: 1820 845 103 D 0 0
845 1560 715 0
0 715 1300 585
0 u2 0 0 u 0 T d 3t = d t 585 u4 0 585 u5 103
The displacement solution is u1 = 0, u2 = 0.001026 in, u3 = 0.002210 in, u4 = 0.003608 in, and u5 = 0.005317 in. Postprocessing Phase 7. Obtain other information. For Example 1.1, we may be interested in obtaining other information, such as the average normal stresses in each element. These values can be determined from the equation AavgE (ui + 1  ui) keq(ui + 1  ui) f u i + 1  ui / s = = = = E¢ ≤ (1.19) Aavg Aavg Aavg /
M01_MOAV4303_04_GE_C01.INDD 37
27/11/14 9:33 AM
38 Chapter 1 Introduction
Since the displacements of different nodes are known, Eq. (1.19) could have been obtained directly from the relationship between the stresses and strains, s = Ee = E ¢
u i + 1  ui ≤ (1.20) /
Employing Eq. (1.20) in Example 1.1, we compute the average normal stress for each element as s(1) = E ¢
(10.4 * 106)(0.001026  0) u 2  u1 lb ≤ = = 4268 2 / 2.5 in
s(2) = E ¢
(10.4 * 106)(0.002210  0.001026) u 3  u2 lb ≤ = = 4925 2 / 2.5 in
s(3) = E ¢
(10.4 * 106)(0.003608  0.002210) u 4  u3 lb ≤ = = 5816 2 / 2.5 in
s(4) = E ¢
(10.4 * 106)(0.005317  0.003608) u 5  u4 lb ≤ = = 7109 2 / 2.5 in
In Figure 1.6, we note that for the given problem, regardless of where we cut a section through the bar, the internal force at the section is equal to 1000 lb. So, s(1) =
f 1000 lb = = 4267 2 Aavg 0.234375 in
s(2) =
f 1000 lb = = 4923 2 Aavg 0.203125 in
s(3) =
f 1000 lb = = 5818 2 Aavg 0.171875 in
s(4) =
f 1000 lb = = 7111 2 Aavg 0.140625 in f=P f=P
L y
P = 1000 lb
P
P
Figure 1.6 The internal forces in Example 1.1.
M01_MOAV4303_04_GE_C01.INDD 38
27/11/14 9:33 AM
Section 1.5 Direct Formulation 39
Ignoring the errors we get from rounding off our answers, we find that these results are identical to the element stresses computed from the displacement information. This comparison tells us that our displacement calculations are good for this problem. Reaction Forces For Example 1.1, the reaction force may be computed in a number of ways. First, referring to Figure 1.4, we note that the statics equilibrium at node 1 requires R1 = k1(u2  u1) = 975 * 103(0.001026  0) = 1000 lb The statics equilibrium for the entire bar also requires that R1 = P = 1000 lb As you may recall, we can also compute the reaction forces from the general reaction equation or
5 R 6 = [K]5 u 6  5 F6
5 reaction matrix6 = [stiffness matrix]5 displacement matrix6  5 load matrix6
Because Example 1.1 is a simple problem, we do not actually need to go through the matrix operations in the aforementioned general equation to compute the reaction forces. However, as a demonstration, the procedure is shown here. From the general equation, we get R1 975 R2 975 e R3 u = 103 E 0 R4 0 R5 0
975 1820 845 0 0
0 845 1560 715 0
0 0 715 1300 585
0 0 0 0 0.001026 0 0 U e 0.002210 u  e 0 u 585 0.003608 0 585 0.005317 103
where R1, R2, R3, R4, and R5 represent the reactions forces at nodes 1 through 5 respectively. Performing the matrix operation, we have R1 1000 R2 0 e R3 u = e 0 u R4 0 R5 0 The negative value of R1 simply means that the direction of the reaction force is up (because we assumed that the positive ydirection points down). Of course, as expected, the outcome is the same as in our earlier calculations because the rows of the above matrix represent the static equilibrium conditions at each node. When solving for reaction forces, it is important to note that you must use the complete stiffness matrix, without the influence of boundary conditions, as
M01_MOAV4303_04_GE_C01.INDD 39
27/11/14 9:33 AM
40 Chapter 1 Introduction
shown in E quations (1.11) and (1.12). Next, we will consider finite element formulation of a heat transfer problem. Example 1.1 also is solved using Excel. See Section 2.11. Example 1.2 A typical exterior frame wall (made up of 2 * 4 studs) of a house contains the materials shown in the table below. Let us assume an inside room temperature of 70°F and an outside air temperature of 20°F, with an exposed area of 150 ft2. We are interested in determining the temperature distribution through the wall. Resistance hr # ft2 # °F/Btu
Ufactor Btu/hr # ft2 # °F
1. O utside film resistance (winter, 15mph wind)
0.17
5.88
2. Siding, wood (1/2 * 8 lapped)
0.81
1.23
3. Sheathing (1/2 in regular)
1.32
0.76
4. Insulation batt (3  3½ in)
11.0
0.091
5. Gypsum wall board (1/2 in)
0.45
2.22
6. Inside film resistance (winter)
0.68
1.47
Items
1
2
3
5
4
6
Preprocessing Phase 1. Discretize the solution domain into finite elements. We will represent this problem by a model that has seven nodes and six elements, as shown in Figure 1.7. 2. Assume a solution that approximates the behavior of an element. For Example 1.2, there are two modes of heat transfer (conduction and convection) that we must first understand before we can proceed with formulating the conductance matrix and the thermal load matrix. The steadystate thermal behavior of the elements (2), (3), (4), and (5) may be modeled using Fourier’s law. When there exists a temperature gradient in a medium, conduction heat transfer occurs, as shown in Figure 1.8. The energy is transported from the hightemperature region
T1 = 205F 1
(1)
2
Siding
Sheathing
Insulation Batt
Gypsum Board
Element 2
Element 3
Element4
Element 5
(2)
3
(3)
4
(4)
5
(5)
T7 = 705F 6
(6)
7
Figure 1.7 Finite element model of Example 1.2.
M01_MOAV4303_04_GE_C01.INDD 40
27/11/14 9:33 AM
Section 1.5 Direct Formulation 41 /
k
Ti+1
Ti
qX
Figure 1.8 Heat transfer in a medium by conduction.
X
to the lowtemperature region by molecular activities. The heat transfer rate is given by Fourier’s law:
qX = kA
0T (1.21) 0X
qX is the Xcomponent of the heat transfer rate, k is the thermal conductivity of 0T the medium, A is the area normal to heat flow, and is the temperature gradient. 0X The minus sign in Eq. (1.21) is due to the fact that heat flows in the direction of decreasing temperature. Equation (1.21) can be written in a difference form in terms of the spacing between the nodes (length of the element) / and the respective temperatures of the nodes i and i + 1, Ti and Ti + 1, according to the equation
q =
kA(Ti + 1  Ti) (1.22) /
In the field of heat transfer, it is also common to write Eq. (1.22) in terms of the thermal transmittance coefficient U, or, as it is often called, the Ufactor (U = k/ ). The Ufactor represents thermal transmission through a unit area and has the units of Btu/hr # ft2 # °F. It is the reciprocal of thermal resistance. So, Equation (1.22) becomes
q = UA(Ti + 1  Ti) (1.23)
The steadystate thermal behavior of elements (1) and (6) may be modeled using Newton’s law of cooling. Convection heat transfer occurs when a fluid in motion comes into contact with a surface whose temperature differs from the moving fluid. The overall heat transfer rate between the fluid and the surface is governed by Newton’s law of cooling, according to the equation
M01_MOAV4303_04_GE_C01.INDD 41
q = hA(Ts  Tf) (1.24)
27/11/14 9:33 AM
42 Chapter 1 Introduction
where h is the heat transfer coefficient, Ts is the surface temperature, and Tf r epresents the temperature of the moving fluid. Newton’s law of cooling can also be written in terms of the Ufactor, such that q = UA(Ts  Tf) (1.25)
where U = h, and it represents the reciprocal of thermal resistance due to convection boundary conditions. Under steadystate conduction, the application of energy balance to a surface, with a convective heat transfer, requires that the energy transferred to this surface via conduction must be equal to the energy transfer by convection. This principle, kA
0T = hA[Ts  Tf] (1.26) 0X
is depicted in Figure 1.9. Now that we understand the two modes of heat transfer involved in this problem, we can apply the energy balance to the various surfaces of the wall, starting with the wall’s exterior surface located at node 2. The heat loss through the wall due to conduction must equal the heat loss to the surrounding cold air by convection. That is, U2A(T3  T2) = U1A(T2  T1) The application of energy balance to surfaces located at nodes 3, 4, and 5 yields the equations U3A(T4  T3) = U2A(T3  T2) U4A(T5  T4) = U3A(T4  T3) U5A(T6  T5) = U4A(T5  T4) For the interior surface of the wall, located at node 6, the heat loss by convection of warm air is equal to the heat transfer by conduction through the gypsum board, according to the equation U6A(T7  T6) = U5A(T6  T5) / qconvection = hA[Ts  Tf ]
qconduction = kA
Ti = Ts
0T 0X
Ti+1
k
Tf , h X
M01_MOAV4303_04_GE_C01.INDD 42
Figure 1.9 Energy balance at a surface with a convective heat transfer.
27/11/14 9:33 AM
Section 1.5 Direct Formulation 43
Separating the known temperatures from the unknown temperatures, we have +(U1 + U2)AT2 U2AT3 U2AT2 +(U2 + U3)AT3 U3AT4 U3AT3 +(U3 + U4)AT4 U4AT5 U4AT4 +(U4 + U5)AT5 U5AT6 U5AT5 +(U5 + U6)AT6
= U1AT1 = 0 = 0 = 0 = U6AT7
The above relationships can be represented in matrix form as U1 + U2 U2 0 AE 0 0
U2 U2 + U3 U3 0 0
0 U3 U3 + U4 U4 0
0 0 U4 U4 + U5 U5
0 T2 U1AT1 0 T3 0 0 U e T4 u = e 0 u (1.27) U5 T5 0 U5 + U6 T6 U6AT7
Note that the relationship given by Eq. (1.27) was developed by applying the conservation of energy to the surfaces located at nodes 2, 3, 4, 5, and 6. Next, we will consider the elemental formulation of this problem, which will lead to the same results. 3. Develop equations for an element. In general, for conduction problems, the heat transfer rates qi and qi + 1 and the nodal temperatures Ti and Ti + 1 for an element are related according to the equations
kA (Ti  Ti + 1) / kA = (Ti + 1  Ti) (1.28) /
qi =
qi + 1
The heat flow through nodes i and i + 1 is depicted in Figure 1.10. Because each of the elements in Example 1.2 has two nodes, and we have associated a temperature with each node, we want to create two equations for each element. These equations must involve nodal temperatures and the element’s thermal conductivity or Ufactor, based on Fourier’s law. Under steadystate conditions, the application of the conservation of energy requires that the sum of qi and qi + 1 into an element be zero; that is, the energy flowing into node i + 1 must be equal to the energy flowing out of node i. Note that the sum of qi and qi + 1 is zero regardless of which representation of Figure 1.10 is selected. However, for the sake of consistency in the forthcoming derivation, we will use the representation given by Figure 1.10(b). Elemental description given by Eq. (1.28) can be expressed in matrix form by
M01_MOAV4303_04_GE_C01.INDD 43
b
qi kA 1 r = J qi + 1 / 1
1 T R b i r (1.29) 1 Ti + 1
27/11/14 9:33 AM
44 Chapter 1 Introduction /
/
k
k Ti+1
qi = kA (Ti+1 – Ti) /
qi+1 = kA (Ti+1 – Ti) /
Ti
Ti+1
qi = kA (Ti – Ti+1) /
qi+1 = kA (Ti+1 – Ti) /
Ti
X
X
(a)
(b)
Figure 1.10 Heat flow through nodes i and i + 1.
The thermal conductance matrix for an element is [K](e) =
kA 1 J / 1
1 R (1.30) 1
The conductance matrix can also be written in terms of the Ufactor ¢ U = [K](e) = UA J
1 1
k ≤: /
1 R (1.31) 1
Similarly, under steadystate conditions, the application of the conservation of energy to the nodes of a convective element gives = hA(Ti  Ti + 1)
qi
qi + 1 = hA(Ti + 1  Ti) (1.32)
Equation (1.32) expressed in a matrix form is b
qi 1 r = hA J qi + 1  1
1 T R b i r 1 Ti + 1
The thermal conductance matrix for a convective element then becomes
[K](e) = hA J
1 1
1 R (1.33) 1
Equation (1.33) can also be written in terms of the Ufactor (U = h):
M01_MOAV4303_04_GE_C01.INDD 44
[K](e) = UA J
1 1
1 R (1.34) 1
27/11/14 9:33 AM
Section 1.5 Direct Formulation 45
4. Assemble the elements to present the entire problem. Applying the elemental description given by Eqs. (1.31) and (1.34) to all of the elements in Example 1.2 and assembling leads to the formation of the global stiffness matrix. So, [K](1) = A J
U1 U1
U1 R U1
and its position in the global matrix is
[K](1G)
U1 U1 0 = AG 0 0 0 0
U1 U1 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0W 0 0 0
T1 T2 T3 T4 T5 T6 T7
The nodal temperature matrix is shown along with the global thermal conductance matrix to help you observe the contribution of a node to its neighboring elements:
[K](2) = A J
[K](3) = A J
[K](4) = A J
M01_MOAV4303_04_GE_C01.INDD 45
U2 U2
U3 U3
U4 U4
U2 R and [K](2G) U2
0 0 0 = AG 0 0 0 0
0 U2 U2 0 0 0 0
U3 R and [K](3G) U3
0 0 0 = AG 0 0 0 0
0 0 0 0 0 0 0
U4 R and [K](4G) U4
0 0 0 = AG 0 0 0 0
0 0 0 0 0 0 0
0 U2 U2 0 0 0 0
0 0 U3 U3 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0W 0 0 0
T1 T2 T3 T4 T5 T6 T7
0 0 U3 U3 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0W 0 0 0
T1 T2 T3 T4 T5 T6 T7
0 0 0 U4 U4 0 0
0 0 0 0 0 0 0
0 0 0 0W 0 0 0
T1 T2 T3 T4 T5 T6 T7
0 0 0 U4 U4 0 0
27/11/14 9:33 AM
46 Chapter 1 Introduction
[K](5) = A J
[K](6) = A J
U5 U5
U6 U6
U5 R and [K](5G) U5
0 0 0 = AG 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
U6 R and [K](6G) U6
0 0 0 = AG 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 U5 U5 0 0 0 0 0 0 0 0
0 0 0 0 U5 U5 0
0 0 0 0 0 U6 U6
0 0 0 0W 0 0 0
T1 T2 T3 T4 T5 T6 T7
0 0 0 0 W 0 U6 U6
T1 T2 T3 T4 T5 T6 T7
The global conductance matrix is [K](G) = [K](1G) + [K](2G) + [K](3G) + [K](4G) + [K](5G) + [K](6G)
[K](G)
U1 U1 0 = AG 0 0 0 0
U1 U1 + U2 U2 0 0 0 0
0 U2 U2 + U3 U3 0 0 0
0 0 U3 U3 + U4 U4 0 0
0 0 0 U4 U4 + U5 U5 0
0 0 0 0 U5 U5 + U6 U6
0 0 0 0 W (1.35) 0 U6 U6
5. Apply boundary conditions and thermal loads. For the given problem, the exterior of the wall is exposed to a known air temperature T1, and the room temperature, T7, is also known. Thus, we want the first row to read T1 = 20°F and the last row to read T7 = 70°F. So, we have 1
A U1 0 AG 0 0 0 0
0 U1 + U2 U2 0 0 0 0
M01_MOAV4303_04_GE_C01.INDD 46
0  U2 U2 + U3  U3 0 0 0
0 0  U3 U3 + U4 U4 0 0
0 0 0  U4 U4 + U5  U5 0
0 0 0 0 U5 U5 + U6 0
0 T1 20°F 0 T2 0 0 T3 0 0 W g T4 w = g 0 w (1.36) 0 T5 0 U6 T6 0 1/A T7 70°F
27/11/14 9:33 AM
Section 1.5 Direct Formulation 47
Note that the finite element formulation of heat transfer problems will always lead to an equation of the form [K] 5 T6 = 5 q 6
[conductance matrix]5 temperature matrix6 = 5 heat flow matrix6
Also note that for Example 1.2, the heat transfer rate through each element was caused by temperature differences across the nodes of a given element. Thus, the external nodal heat flow values are zero in the heat flow matrix. An example of a situation in which external nodal heat values are not zero is a heating strip attached to a solid surface (e.g., the base of a pressing iron); for such a situation, the external nodal heat value is equal to the amount of heat being generated by the heating strip over the surface. Turning our attention to the matrices given by Eq. (1.36) and incorporating the known boundary conditions into rows 2 and 6 of the conductance matrix, we can reduce Eq. (1.36) to U1 + U2 U2 0 AE 0 0
U2 U2 + U3 U3 0 0
0 U3 U3 + U4 U4 0
0 0 U4 U4 + U5 U5
0 T2 U1AT1 0 T3 0 0 U e T4 u = e 0 u U5 T5 0 U5 + U6 T6 U6AT7
Keep in mind that the above matrix was obtained by assembling the elemental description and applying the boundary conditions. Moreover, the results of this approach are identical to the relations we obtained earlier by balancing the heat flows at the nodes, as given by Eq. (1.27). This equality in the outcome is expected because the elemental formulations are based on the application of energy balance as well. Referring to the original global matrix, substituting for the Uvalues and employing the given boundary conditions, we have 1 0 0 0 0 0 0 150 T1 20°F 5.885.88 + 1.23 1.23 0 0 0 0 T2 0 0 1.23 1.23 + 0.76 0.76 0 0 0 T3 0 0 0.76 0.76 + 0.091 0.091 0 0 Y g T4 w = g 0 w 150 I 0 0 0 0 0.091 0.091 + 2.22 2.22 0 T5 0 0 0 0 0 2.22 2.22 + 1.47 1.47 0 T6 1 T7 70°F 0 0 0 0 0 0 150
M01_MOAV4303_04_GE_C01.INDD 47
27/11/14 9:33 AM
48 Chapter 1 Introduction
Simplifying, we obtain 7.11 1.23 E 0 0 0
1.23 1.99 0.76 0 0
0 0.76 0.851 0.091 0
0 0 0.091 2.311 2.22
0 T2 (5.88)(20) 0 T3 0 0 U e T4 u = e 0 u 2.22 T5 0 3.69 T6 (1.47)(70)
Solution Phase 6. Solve a system of algebraic equations simultaneously. Solving the previous matrix yields the temperature distribution along the wall: T1 20.00 T2 20.59 T3 23.41 g T4 w = g 27.97 w °C T5 66.08 T6 67.64 T7 70.00 For problems similar to the type discussed here, the knowledge of temperature distribution within the wall is important in determining where condensation may occur in the wall and thus where one should place a vapor barrier to avoid moisture condensation. To demonstrate this concept, let us assume that moisture can diffuse through the gypsum board and that the inside air has a relative humidity of 40%. With the help of a psychometric chart, using a dry bulb temperature of 70°F and the value f = 40,, we identify the condensation temperature to be 44°F. Therefore, the water vapor in the air at any surface whose temperature is 44°F or below will condense. In the absence of a vapor barrier, the water vapor in the air will condense somewhere between surface 5 and 4 for the assumed conditions in this problem. Postprocessing Phase 7. Obtain other information. For this example, we may be interested in obtaining other information, such as heat loss through the wall. Such information is important in computing the heat load for a building. Because we have assumed steadystate conditions, the heat loss through the wall should be equal to the heat transfer through each element. This value can be determined from the equation
M01_MOAV4303_04_GE_C01.INDD 48
q = UA(Ti + 1  Ti) (1.37)
27/11/14 9:33 AM
Section 1.5 Direct Formulation 49
The heat transfer through each element is q = UA(Ti + 1  Ti) = (1.47)(150)(70  67.64) = (2.22)(150)(67.64  66.08) = g Btu = (5.88)(150)(20.59  20) = 520 hr We also could have calculated the heat loss through the wall using the overall U‑factor in the following manner: 1 A(Tinside  Toutside) Σ Resistance Btu = (0.0693)(150)(70  20) = 520 hr
q = Uoverall A(Tinside  Toutside) =
This problem is just another example of how we can generate finite element models using the direct method. A Torsional Problem: Direct Formulation Example 1.3 Consider the torsion of a circular shaft, shown in Figure 1.11. Recall from your previous study of the mechanics of materials that the angle of twist u for a shaft with a constant crosssectional area with a polar moment of inertia J and length /, made of homogenous material with a shear modulus of elasticity G, subject to a torque T is given by u =
T/ JG
Using direct formulation, equilibrium conditions, and u =
T/ JG
we can show that for an element comprising two nodes, the stiffness matrix, the angle of twists, and the torques are related according to the equation JG 1 J / 1
1 u T R b 1 r = b 1 r (1.38) 1 u2 T2 T2
T1 / Figure 1.11 A torsion of circular shaft.
M01_MOAV4303_04_GE_C01.INDD 49
27/11/14 9:33 AM
50 Chapter 1 Introduction
We will discuss torsional problems in much more detail in Chapter 10. For now, let us consider a shaft that is made of two parts, as shown in Figure 1.12. Part AB is made of material with a shear modulus of elasticity of GAB = 3.9 * 106 lb/in2 and has a dia meter of 1.5 in. Segment BC is made of slightly different material with a shear modulus of elasticity of GBC = 4.0 * 106 lb/in2 and with a diameter of 1 in. The shaft is fixed at both ends. A torque of 200 lb # ft is applied at D. Using three elements, let us determine the angle of twist at D and B, and the torsional reactions at the boundaries. We will represent this problem by a model that has four nodes at A, B, C, and D, respectively, and three elements (AD, DB, BC). The polar moment of inertia for each element is given by 4
1 1 1.5 in ≤ = 0.497 in4 J1 = J2 = pr4 = p¢ 2 2 2 4
1 1 1.0 J3 = pr4 = p¢ in ≤ = 0.0982 in4 2 2 2 The stiffness matrix for each element is computed from Eq. (1.38) as [K](e) =
JG 1 J / 1
1 R 1
So, for element (1), the stiffness matrix is [K](1) =
(0.497 in4)(3.9 * 106 lb/in2) 1 J (12 * 2.5) in 1
1 64610 R = J 1 64610
64610 R lb # in 64610
and its position in the global stiffness matrix is
[K](1G)
1.5 in
64610 64610 = D 0 0
64610 64610 0 0
0 0 0 0
0 0 T 0 0
u1 u2 u3 u4
T = 200 lb ? ft 1 in B
A
C
D
2.5 ft
1 ft
2 ft
Figure 1.12 A schematic of the shaft in Example 1.3.
M01_MOAV4303_04_GE_C01.INDD 50
27/11/14 9:33 AM
Section 1.5 Direct Formulation 51
Similarly, for elements (2) and (3), their respective stiffness matrices and positions in the global stiffness matrix are as follows: [K](2) =
[K](2G)
0 0 = D 0 0
[K](3) =
[K](3G)
(3.9 * 106 lb/in2)(0.497 in4) 1 J (12 * 1.0) in 1 0 161525 161525 0
0 161525 161525 0
0 0 T 0 0
0 0 0 0
0 0 16367 16367
0 0 T 16367 16367
161525 R lb # in 161525
1 16367 R = J 1 16367
16367 R lb # in 16367
u1 u2 u3 u4
(4.0 * 106 lb/in2)(0.0982 in4) 1 J (12 * 2.0) in 1
0 0 = D 0 0
1 161525 R = J 1 161525
u1 u2 u3 u4
The final global matrix is obtained simply by assembling, or adding, elemental descriptions: [K](G) = [K](1G) + [K](2G) + [K](3G) [K](G)
64610 64610 = D 0 0
64610 64610 + 161525 161525 0
0 161525 161525 + 16367 16367
0 0 T 16367 16367
Applying the fixed boundary conditions at points A and C and applying the external torque, we have 1 64610 D 0 0
0 226135 161525 0
0 161525 177892 0
0 u1 0 0 u2 (200 * 12) lb # in T d t = d t 16367 u3 0 1 u4 0
Solving the above set of equations, we obtain u1 0 u2 0.03020 rad d t = d t u3 0.02742 rad u4 0
M01_MOAV4303_04_GE_C01.INDD 51
27/11/14 9:33 AM
52 Chapter 1 Introduction
The reaction moments at boundaries A and C can be determined as follows: 5 R 6 = [K]5 U 6  5 T6
RA 64610 64610 0 0 0 0 R 64610 226135 161525 0 0.03020 rad (200 * 12) lb # in d Dt = D Td t d t RB 0 161525 177892 16367 0.02742 rad 0 RC 0 0 16367 16367 0 0 RA 1951 lb # in R 0 d Dt = d t RB 0 RC 449 lb # in Note that the sum of RA and RC is equal to the applied torque of 2400 lb # in. Also note that the change in the diameter of the shafts will give rise to stress concentrations that are not accounted for by the model we used here. Example 1.4 A steel plate is subjected to an axial load, as shown in Figure 1.13. Approximate the deflections and average stresses along the plate. The plate is 1/16 in thick and has a modulus of elasticity E = 29 * 106 lb/in2. We may model this problem using four nodes and four elements, as shown in Figure 1.13. Next, we compute the equivalent stiffness coefficient for each element: k1 =
(5)(0.0625)(29 * 106) A 1E = = 9,062,500 lb/in /1 1
k 2 = k3 = k4 =
(2)(0.0625)(29 * 106) A 2E = = 906,250 lb/in /2 4
(5)(0.0625)(29 * 106) A 4E = = 4,531,250 lb/in /4 2
The stiffness matrix for element (1) is [K](1) = J
k1 k1
k1 R k1
and its position in the global stiffness matrix is
[K](1G)
M01_MOAV4303_04_GE_C01.INDD 52
k1 k = D 1 0 0
k1 k1 0 0
0 0 0 0
0 0 T 0 0
u1 u2 u3 u4
27/11/14 9:33 AM
Section 1.5 Direct Formulation 53
2" 800 lb
1" 2" 1"
4"
2"
(2) 800 lb
(4)
(1) (3)
k2 k4
k1 1
2
k3
3
800 lb
4
Figure 1.13 A schematic of the steel plate in Example 1.4.
Similarly, the respective stiffness matrices and positions in the global stiffness matrix for elements (2), (3), and (4) are k2 k2
k2 R k2
k = J 3 k3
k3 R k3
[K](2) = J
[K](3)
[K](4) = J
M01_MOAV4303_04_GE_C01.INDD 53
k4 k4
k4 R k4
[K](2G)
0 0 = D 0 0
0 k2 k2 0
0 k2 k2 0
0 0 T 0 0
u1 u2 u3 u4
[K](3G)
0 0 = D 0 0
0 k3 k3 0
0 k3 k3 0
0 0 T 0 0
u1 u2 u3 u4
[K](4G)
0 0 = D 0 0
0 0 T k4 k4
u1 u2 u3 u4
0 0 0 0
0 0 k4 k4
27/11/14 9:33 AM
54 Chapter 1 Introduction
The final global matrix is obtained simply by assembling, or adding, the individual elemental matrices: [K](G) = [K](1G) + [K](2G) + [K](3G) + [K](4G)
[K](G)
k1 k1 = D 0 0
k1 k 1 + k2 + k3 k2 k3 0
0 k2 k3 k 2 + k3 + k4 k4
0 0 T k4 k4
Substituting for the elements’ respective stiffness coefficients, the global stiffness matrix becomes
[K](G)
9,062,500 9,062,500 = D 0 0
9,062,500 10,875,000 1,812,500 0
0 1,812,500 6,343,750 4,531,250
0 0 T 4,531,250 4,531,250
Applying the boundary condition u1 = 0 and the load to node 4, we obtain 1 9,062,500 D 0 0
0 10,875,000 1,812,500 0
0 1,812,500 6,343,750 4,531,250
0 u1 0 0 u2 0 T d t = d t 4,531,250 u3 0 4,531,250 u4 800
Solving the system of equations yields the displacement solution as u1 0 u 8.827 * 105 d 2t = d t in u3 5.296 * 104 u4 7.062 * 104 and the stresses in each element are s(1) = E ¢
(29 * 106)(8.827 * 105  0) u 2  u1 lb ≤ = = 2560 2 / 1 in
s(2) = s(3) = E ¢ s(4) = E ¢
(29 * 106)(5.296 * 104  8,827 * 105) u 3  u2 lb ≤ = = 3200 2 / 4 in
(29 * 106)(7.062 * 104  5.296 * 104) u4  u3 lb ≤ = = 2560 2 / 2 in
Note that the model used to analyze this problem consisted of springs in parallel as well as in series. The two springs in parallel could have been combined and represented by a single spring having a stiffness equal to k2 + k3 (see Problem 25). Also note that because of the hole, the abrupt changes in the cross section of the strip will give rise to stress concentrations with values exceeding those average values we computed here. After you study planestress finite element formulation (discussed in Chapter 10), you
M01_MOAV4303_04_GE_C01.INDD 54
27/11/14 9:33 AM
Section 1.5 Direct Formulation 55
Figure 1.14 The xcomponent of stress distribution for the plate in Example 1.4, as computed by ANSYS.
will revisit this problem (see Problem 10.13) and be asked to solve it using ANSYS. Furthermore, you will be asked to plot the components of the stress distributions in the plate and thus identify the location and magnitude of maximum stresses. To give you just a taste of what is to come in Chapter 10 and also to shed more light on our discussion about the stress concentration regions, we have solved Example 1.4 using ANSYS and have determined the xcomponent of the stress distribution in the plate, as shown in Figure 1.14. In the results shown in Figure 1.14, the load was applied as a pressure over the entire right surface of the bar. Note the variation of the stresses at section A–A from approximately 3000 psi to 3500 psi. At section B–B, the xcomponent of the stresses varies from approximately 2300 psi to 2600 psi. These values are not that far off from the average stress values obtained using the direct model. Also note that the maximum and minimum stress values given by ANSYS could change, depending upon how we apply the load to the bar, especially in the regions near the point of load application and the regions near the hole. Keeping in mind Example 1.4 and Figure 1.13, remember that in a real situation, the load would be applied over an area, not at a single point. Thus, remember that how you apply the external load to your finite element model will influence the stress distribution results, particularly in the region near where the load is applied. This principle is especially true in Example 1.4 because it deals with a short plate with a hole. For the sake of convenience, the results of Section 1.5 is summarized in Table 1.6. In the table, carefully examine what constitutes an element, its degrees of freedom, and the physical balance requirements.
M01_MOAV4303_04_GE_C01.INDD 55
27/11/14 9:33 AM
56 Chapter 1 Introduction
Table 1.6 Examples of elements and nodes Degrees of Freedom
Physical Balance Requirement
Nodal displacements: ui, ui + 1
Force balance: fi + fi+1 = 0
Nodal angle of twist: ui, ui + 1
Torque balance: Ti + Ti+1 = 0
Nodal temperatures: Ti, Ti + 1
Energy balance: qi + qi+1 = 0
Qi = C(Pi  Pi + 1) Qi+1 = C(Pi+1  Pi)
Nodal pressures: Pi, Pi + 1
Flow balance: Qi + Qi+1 = 0
1 (V  Vi+1) R i
Nodal voltages: Vi, Vi + 1
Electric current balance: Ii + Ii+1 = 0
Element Linear Elastic Element (linear spring) fi i
ui fi = k(ui  ui+1) fi+1 = k(ui+1  ui)
k i+1
ui+1 fi+1
Torsional Elastic Element (torsional spring) Ti+1
ui+1
Ti ui
Ti =
J, G
i +1
JG (u  ui + 1) / i
Ti + 1 =
/
i
JG (u  ui) / i+1
Conduction Element / qi
Ti i
k
Ti+1 i+1
qi+1
qi = qi + 1
kA (Ti  Ti+1) / kA = (Ti+1  Ti) /
Laminar Pipe Flow Element (See Section 12.1) Pi+1
Qi+1
Pi Qi
i+1
i Electrical Resistance Element Vi Ii
R
i
M01_MOAV4303_04_GE_C01.INDD 56
Vi+1 i+1
Ii+1
Ii =
Ii + 1 =
1 (V  Vi) R i+1
27/11/14 9:33 AM
Section 1.6 Minimum Total Potential Energy Formulation 57
1.6 Minimum Total Potential Energy Formulation The minimum total potential energy formulation is a common approach in generating finite element models in solid mechanics. External loads applied to a body will cause the body to deform. During the deformation, the work done by the external forces is stored in the material in the form of elastic energy, called strain energy. Let us consider the strain energy in a solid member when it is subjected to a central force F, as shown in Figure 1.15. Also shown in Figure 1.15 is a piece of material from the member in the form of differential volume and the normal stresses acting on the surfaces of this volume. Earlier, it was shown that the elastic behavior of the member may be modeled as a linear spring. In Figure 1.15 note that y′ is a variable measuring deformation of the member and its value varies from 0 to ∆/. When the member is stretched by a differential amount dy′, the stored energy in the material is Λ =
L0
y′
Fdy′ =
F
L0
y′
ky′dy′ =
F y'
y'
/
/
1 1 ky′2 = ¢ ky′≤y′ (1.39) 2 2
F = _ AE _ / = ky' /
Unstretched length (no load) y
y
sy
dy'
dy
x dz dx z
sy Figure 1.15 The elastic behavior of a member subjected to a central load.
M01_MOAV4303_04_GE_C01.INDD 57
27/11/14 9:33 AM
58 Chapter 1 Introduction
We can write Eq. (1.39)—for a piece of material from the member in the form of differential volume—in terms of the normal stress (s) and strain (e): force elastic force dy′ 1 elastic $%& 1 $+%1& 1 $%& dy′ = (sydxdz) edy = se dV (ky′) 2 2 2 Therefore, for a member or an element under axial loading, the strain energy Λ(e) is obtained by adding up the stored energy in all pieces (differential volumes) making up the member: se Ee2 dΛ = dV = dV (1.40) Λ(e) = L LV 2 LV 2
dΛ =
where V is the volume of the member and s = Ee. The total potential energy Π for a body consisting of n elements and m nodes is the difference between the total strain energy and the work done by the external forces: Π = a Λ(e)  a Fiui (1.41)
n
m
e=1
i=1
The minimum total potential energy principle simply states that for a stable system, the displacement at the equilibrium position occurs such that the value of the system’s total potential energy is a minimum. 0Π 0 n (e) 0 m = Λ Fiui = 0 for i = 1, 2, 3, c, n (1.42) a 0ui 0ui e = 1 0ui ia =1
The following examples offer insight into the physical meaning of Eq. (1.42). Example 1.5 Consider the following situations: (a) We have applied a force F to a linear spring as shown in Figure 1.16. Depending on the stiffness value of the spring, the spring stretches by a certain amount x. The static equilibrium requires that the applied force F be equal to the internal force in the spring kx. F = kx or x =
F k
Now, let us consider the total potential energy of the system as defined by Eq. (1.41). The stored elastic energy in the spring is Λ = 12 kx2 and the work done by the external force F is Fx (force times displacement). Thus, the total potential energy of the system is Π =
1 2 kx  Fx 2
k
F
M01_MOAV4303_04_GE_C01.INDD 58
x
Figure 1.16 A linear spring subjected to a force F.
27/11/14 9:33 AM
Section 1.6 Minimum Total Potential Energy Formulation 59 8 cm A
2 cm G
B
P = 12 N C
k = 20 N/cm Figure 1.17 The rod of Example 1.5.
Minimizing Π with respect to x, we have dΠ d 1 2 = ¢ kx  Fx≤ = kx  F = 0 dx dx 2 F . k (b) The slender rod shown in Figure 1.17 weighs 8 N and is supported by a spring with a stiffness k = 20 N/cm. A force P = 12 N is applied to the end of the rod at point C. We are interested in determining the deflection of the spring. First, we solve this problem by applying the static equilibrium conditions and then apply the minimum total potential energy concept. Static equilibrium requires that sum of the moments of the forces acting on the rod about point A be zero. Considering the freebody diagram of the rod shown in Figure 1.18, we find which results in x =
+ a MA = 0 ⤿
(8N)(5 cm) + Fs (8 cm)  (12 N)(10 cm) = 0 Fs = 20 N and kx = (20 N/cm)(x) = 20 N x = 1 cm
Now, we solve the problem using the minimum total potential energy approach. We note that elastic energy stored in the system is predominantly due to elastic energy of the spring and is given by Λ =
1 2 1 kx = (20 N/cm)(x2) = 10x2 2 2
The work done by the external forces is calculated by multiplying the weight of the rod by the displacement of point G, and force P by the displacement of endpoint C. Through 5 cm Rx
A
3 cm G
2 cm B
xG Ry
W=8N
P = 12 N C
x
xc
FS
Figure 1.18 The freebody diagram of the rod in Example 1.5.
M01_MOAV4303_04_GE_C01.INDD 59
27/11/14 9:33 AM
60 Chapter 1 Introduction
similar triangles, we can relate the displacements of points G and C to the displacement of the spring (point B) according to xG x = 8 5 xC x = 8 10
5 x 8 5 or xC = x 4
or xG =
Thus, the work done by the external forces is given by 5 5 a Fiui = (8 N) a 8 xb + (12 N) a 4 xb = 5x + 15x = 20x
The total potential energy of the system is
Π = a Λ  a Fiui = 10x2  20x and dΠ d = ¢ 10x2  20x≤ = 20x  20 = 0 dx dx Solving the above equation for x, we find x = 1 cm. Because there is only one unknown displacement, note that when we employed Eqs. (1.41) and (1.42), we replaced the displacement ui with x and the partial derivative symbol with the ordinary symbol. We have plotted the total potential energy Π = 10x2  20x as a function of displacement x in Figure 1.19. It is clear from examining Figure 1.19 that the minimum total potential energy occurs at x = 1 cm. Now, let us turn our attention back to Example 1.1. The strain energy for an arbitrary element (e) can be determined from Eq. (1.40) as Λ(e) =
AavgE Ee2 dV = (u2i + 1 + u2i  2ui + 1 ui) (1.43) 2/ L 2
Total Potential Energy (N.cm)
V
8 3 0  0.5
2
0.5
1
1.5
2
2.5
7 12 Displacement of Spring x (cm) Figure 1.19 Total potential energy versus displacement x.
M01_MOAV4303_04_GE_C01.INDD 60
27/11/14 9:33 AM
Section 1.6 Minimum Total Potential Energy Formulation 61
where e = (ui + 1  ui)//, and V = Aavg/ were substituted for the axial strain and volume respectively. Minimizing the strain energy with respect to ui and ui + 1 leads to AavgE 0Λ(e) = (ui  ui + 1) (1.44) 0ui / AavgE 0Λ(e) = (ui + 1  ui) 0ui + 1 /
and, in matrix form,
0Λ(e) 0ui k d (e) t = J eq 0Λ keq 0ui + 1
keq u R b i r (1.45) keq ui + 1
where keq = (AavgE)//. Minimizing the work done by the external forces at nodes i and i + 1 of an arbitrary element (e), we get
0 (F u ) = Fi 0ui i i
0 (F u ) = Fi + 1 0ui + 1 i + 1 i + 1
(1.46)
For Example 1.1, the minimum total potential energy formulation leads to a global stiffness matrix that is identical to the one obtained from direct formulation:
[K](G)
k1 k1 = E 0 0 0
k1 k 1 + k2 k2 0 0
0 k2 k 2 + k3 k3 0
0 0 k3 k 3 + k4 k4
0 0 0 U k4 k4
Furthermore, application of the boundary condition and the load results in
1 k1 E 0 0 0
0 k 1 + k2 k2 0 0
0 k2 k 2 + k3 k3 0
0 0 k3 k 3 + k4 k4
0 u1 0 0 u2 0 0 U e u3 u = e 0 u (1.47) k4 u4 0 k4 u5 P
The displacement results will be identical to the ones obtained earlier from the direct method, as given by Eq. (1.17). The concepts of strain energy and minimum total potential energy will be used to formulate solid mechanics problems in Chapters 4, 10, and 13. Therefore, spending a little extra time now to understand the basic ideas will benefit you enormously later.
M01_MOAV4303_04_GE_C01.INDD 61
27/11/14 9:33 AM
62 Chapter 1 Introduction
Example 1.1: Exact Solution* In this section, we will derive the exact solution to Example 1.1 and compare the finite element formulation displacement results for this problem to the exact displacement solutions. As shown in Figure 1.20, the statics equilibrium requires the sum of the forces in the ydirection to be zero. This requirement leads to the relation P  (savg)A(y) = 0 (1.48)
Once again, using Hooke’s law (s = Ee) and substituting for the average stress in terms of the strain, we have P  Ee A(y) = 0 (1.49)
Recall that the average normal strain is the change in length du per unit original length of the differential segment dy. So, e =
du dy
If we substitute this relationship into Eq. (1.49), we now have P  EA(y)
du = 0 (1.50) dy
Rearranging Eq. (1.50), we get
du =
Pdy (1.51) EA(y)
w1
L
(savg) A(y)
y w2
dy du
P P Figure 1.20 The relationship between the external force P and the average stresses for the bar in Example 1.1.
*The contribution of shear stresses is neglected.
M01_MOAV4303_04_GE_C01.INDD 62
27/11/14 9:33 AM
Section 1.7 Weighted Residual Formulations 63
Table 1.7 Comparison of displacement results Results from the Exact Displacement Method (in) Eq. (1.53)
Location of a Point Along the Bar (in)
Results from the Direct Method (in)
Results from the Energy Method (in)
y = 0
0
0
0
y = 2.5
0.001027
0.001026
0.001026
y = 5.0
0.002213
0.002210
0.002210
y = 7.5
0.003615
0.003608
0.003608
y = 10
0.005333
0.005317
0.005317
The exact solution is then obtained by integrating Eq. (1.51) over the length of the bar u
L
Pdy L0 L0 EA(y) y y Pdy u(y) = = L0 EA(y) L0
du =
Pdy w2  w1 E ¢ w1 + ¢ ≤y≤t L
(1.52)
where the area is A(y) = ¢ w1 + ¢
w2  w1 ≤ y≤t L
The deflection profile along the bar is obtained by integrating Eq. (1.52), resulting in
u(y) =
w2  w1 PL Jln ¢ w1 + ¢ ≤y≤  ln w1 R (1.53) Et(w2  w1) L
Equation (1.53) can be used to generate displacement values at various points along the bar. It is now appropriate to examine the accuracy of the direct and potential energy methods by comparing their displacement results with the values. Table 1.7 shows nodal displacements computed using direct and energy methods. It is clear from examination of Table 1.7 that all of the results are in agreement with each other. 1.7 Weighted Residual Formulations The weighted residual methods are based on assuming an approximate solution for the governing differential equation. The assumed solution must satisfy the initial and boundary conditions of the given problem. Because the assumed solution is not exact, substitution of the solution into the differential equation will lead to some residuals or errors. Simply stated, each residual method requires the error to vanish over some selected intervals or at some points. To demonstrate this concept, let’s turn our attention
M01_MOAV4303_04_GE_C01.INDD 63
27/11/14 9:33 AM
64 Chapter 1 Introduction
to Example 1.1. The governing differential equation and the corresponding boundary condition for this problem are as follows:
A(y)E
du  P = 0 subject to the boundary condition u(0) = 0 (1.54) dy
Next, we need to assume an approximate solution. Again, keep in mind that the assumed solution must satisfy the boundary condition. We choose u(y) = c1y + c2y2 + c3y3 (1.55)
where c1, c2, and c3 are unknown coefficients. Equation (1.55) certainly satisfies the fixed boundary condition represented by u(0) = 0. Substitution of the assumed solution, Eq. (1.55), into the governing differential equation, Eq. (1.54), yields the error function ℛ: A(y)
du
$++++%+++& dy (1.56) $+++%+++& w2  w1 2 (c + 2c y + 3c y ) ¢ w1 + ¢ ≤ y≤t E 1  P = ℛ 2 3 L
Substituting for values of w1, w2, L, t, and E in Example 1.1 and simplifying, we get ℛ/E = (0.25  0.0125y)(c1 + 2c2y + 3c3 y2)  96.154 * 106 Collocation Method In the collocation method the error, or residual function ℛ is forced to be zero at as many points as there are unknown coefficients. Because the assumed solution in this example has three unknown coefficients, we will force the error function to equal zero at three points. We choose the error function to vanish at y = L/3, y = 2L/3, and y = L: ℛ(c, y) 2 y = L = 0 3
ℛ = ¢ 0.25  0.0125¢
10 10 10 2 ≤ ≤ ¢ c1 + 2c2 ¢ ≤ + 3c3 ¢ ≤ ≤  96.154 * 106 = 0 3 3 3
ℛ(c, y) 2 y = 2L = 0 3
ℛ = ¢ 0.25  0.0125¢ ℛ(c, y) 2
y=L
20 20 20 2 ≤ ≤ ¢ c1 + 2c2 ¢ ≤ + 3c3 ¢ ≤ ≤  96.154 * 106 = 0 3 3 3
= 0
ℛ = (0.25  0.0125(10))(c1 + 2c2(10) + 3c3(10)2)  96.154 * 106 = 0
M01_MOAV4303_04_GE_C01.INDD 64
27/11/14 9:33 AM
Section 1.7 Weighted Residual Formulations 65
This procedure creates three linear equations that we can solve to obtain the unknown coefficients c1, c2, and c3: 20 100 c + c = 461.539 * 106 3 2 3 3 40 400 c1 + c + c = 576.924 * 106 3 2 3 3 c1 + 20c2 + 300c3 = 769.232 * 106 c1 +
Solving the above equations yields c1 = 423.0776 * 106, c2 = 21.65 * 1015, and c3 = 1.153848 * 106. Substitution of the ccoefficients into Eq. (1.55) yields the approximate displacement profile: u(y) = 423.0776 * 106y + 21.65 * 1015y2 + 1.153848 * 106y3 (1.57)
In order to get an idea of how accurate the collocation approximate results are, we will compare them to the exact results later in this chapter. Subdomain Method In the subdomain method, the integral of the error function over some selected subintervals is forced to be zero. The number of subintervals chosen must equal the number of unknown coefficients. Thus, for our assumed solution, we will have three integrals:
L0
L 3
L0
L 3
ℛ dy = 0
(1.58)
[(0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106]dy = 0
2L 3
LL
ℛ dy = 0
3
2L 3
LL
[(0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106]dy = 0
3 L
L2L
ℛ dy = 0
L2L
[(0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106]dy = 0
3 L
3
Integration of equations given by Eq. (1.58) results in three linear equations that we can solve to obtain the unknown coefficients c1, c2, and c3: 763.88889 * 10  3c1 + 2.4691358c2 + 8.1018519c3 = 320.513333 * 106 0.625c1 + 6.1728395c2 + 47.4537041c3 = 3.2051333 * 104 0.4861111c1 + 8.0246917c2 + 100.694444c3 = 3.2051333 * 104
M01_MOAV4303_04_GE_C01.INDD 65
27/11/14 9:33 AM
66 Chapter 1 Introduction
Solving the above equations yields c1 = 391.35088 * 106, c2 = 6.075 * 106, and c3 = 809.61092 * 109. Substitution of the ccoefficients into Eq. (1.55) yields the approximate displacement profile:
u(y) = 391.35088 * 106y + 6.075 * 106y2 + 809.61092 * 109y3 (1.59) We will compare the displacement results obtained from the subdomain method to the exact results later in this chapter. Galerkin Method The Galerkin method requires the error to be orthogonal to some weighting functions Φi, according to the integral La
b
Φi ℛ dy = 0
i = 1, 2, c, N (1.60)
The weighting functions are chosen to be members of the approximate solution. Because there are three unknowns in the assumed approximate solution for Example 1.1, we need to generate three equations. Recall that the assumed solution is u(y) = c1y + c2y2 + c3y3; thus, the weighting functions are selected to be Φ1 = y, Φ2 = y2, and Φ3 = y3. This selection leads to the following equations:
L0 L0
y[(0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106]dy = 0 (1.61)
L
L0
L
y2[(0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106]dy = 0 L
y3[0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106]dy = 0
Integration of Eq. (1.61) results in three linear equations that we can solve to obtain the unknown coefficients c1, c2, and c3: 8.333333c1 + 104.1666667c2 + 1125c3 = 0.0048077 52.083333c1 + 750c2 + 8750c3 = 0.0320513333 375c1 + 5833.3333c2 + 71428.57143c3 = 0.240385 Solving the above equations yields c1 = 400.642 * 106, c2 = 4.006 * 106, and c3 = 0.935 * 106. Substitution of the ccoefficients into Eq. (1.55) yields the approximate displacement profile: u(y) = 400.642 * 106y + 4.006 * 106y2 + 0.935 * 106y3 (1.62)
We will compare the displacement results obtained from the Galerkin method to the exact results later in this chapter.
M01_MOAV4303_04_GE_C01.INDD 66
27/11/14 9:33 AM
Section 1.7 Weighted Residual Formulations 67
LeastSquares Method The leastsquares method requires the error to be minimized with respect to the unknown coefficients in the assumed solution, according to the relationship Minimize¢
La
b
ℛ2dy≤
which leads to La
b
ℛ
0ℛ dy = 0 0ci
i = 1, 2, c, N (1.63)
Because there are three unknowns in the approximate solution of Example 1.1, Eq. (1.63) generates three equations. Recall that the error function is ℛ/E = (0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106 Differentiating the error function with respect to c1, c2, and c3 and substituting into Eq. (1.63), we have:
L 0 L 0 L 0
0ℛ 0c1 ℛ 10 $+++++++++++++++++%++++++++++++++++++& $++1++%1++++&
[(0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106] (0.25  0.0125y) dy = 0
0ℛ 0c2 ℛ 10 $+++++++++++++++++%++++++++++++++++++& $++1+++%1+++++&
[(0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106] (0.25  0.0125y)2y dy = 0
0ℛ 0c3 ℛ 10 $+++++++++++++++++%++++++++++++++++++& $++1+++%1+++++&
[(0.25  0.0125y)(c1 + 2c2y + 3c3y2)  96.154 * 106] (0.25  0.0125y)3y2 dy = 0
Integration of the above equations results in three linear equations that we can solve to obtain the unknown coefficients c1, c2, and c3: 0.364583333c1 + 2.864583333c2 + 25c3 = 0.000180289 2.864583333c1 + 33.333333c2 + 343.75c3 = 0.001602567 25c1 + 343.75c2 + 3883.928571c3 = 0.015024063 Solving the set of equations simultaneously yields c1 = 389.773 * 106, c2 = 6.442 * 106, and c3 = 0.789 * 106. Substitution of the ccoefficients into Eq. (1.55) yields the approximate displacement profile: u(y) = 389.733 * 106y + 6.442 * 106y2 + 0.789 * 106y3 (1.64)
Next, we will compare the displacement results obtained from the leastsquares method and the other weighted residual methods to the exact results.
M01_MOAV4303_04_GE_C01.INDD 67
27/11/14 9:33 AM
68 Chapter 1 Introduction
Table 1.8 Comparison of weighted residual results Location of a Point Along the Bar (in)
Displacement Results from the Exact Solution Eq. (1.53) (in)
Displacement Results from the Collocation Method Eq. (1.57) (in)
Displacement Results from the Subdomain Method Eq. (1.59) (in)
Displacement Results from the Galerkin Method Eq. (1.62) (in)
Displacement Results from the LeastSquares Method Eq. (1.64) (in)
y = 0
0
0
0
0
0
y = 2.5
0.001027
0.001076
0.001029
0.001041
0.001027
y = 5.0
0.002213
0.002259
0.002209
0.002220
0.002208
y = 7.5
0.003615
0.003660
0.003618
0.003624
0.003618
y = 10
0.005333
0.005384
0.005330
0.005342
0.005331
Comparison of Weighted Residual Solutions Now we will examine the accuracy of weighted residual methods by comparing their displacement results with the exact values. Table 1.8 shows nodal displacements computed using the exact, collocation, subdomain, Galerkin, and leastsquares methods. It is clear from an examination of Table 1.8 that the results are in agreement with each other. It is also important to note here that the primary purpose of Section 1.7 was to introduce you to the general concepts of weighted residual methods and the basic procedures in the simplest possible way. Because the Galerkin method is one of the most commonly used procedures in finite element formulations, more detail and an indepth view of the Galerkin method will be offered later in Chapters 6 and 9. We will employ the Galerkin method to formulate one and twodimensional problems once you have become familiar with the ideas of one and twodimensional elements. Also note that in the above examples of the use of weighted residual methods, we assumed a solution that was to provide an approximate solution over the entire domain of the given problem. As you will see later, we will use piecewise solutions with the Galerkin method. That is to say, we will assume linear or nonlinear solutions that are valid only over each element and then combine, or assemble, the elemental solutions. 1.8 Verification of Results In recent years, the use of finite element analysis as a design tool has grown rapidly. Easytouse, comprehensive packages such as ANSYS have become a common tool in the hands of design engineers. Unfortunately, many engineers without the proper training or a solid understanding of the underlying concepts have been using finite element analysis. Engineers who use finite element analysis must understand the limitations of the finite element procedures. There are various sources of error that can contribute to incorrect results. They include 1. Wrong input data, such as physical properties and dimensions This mistake can be corrected by simply listing and verifying physical properties and coordinates of nodes or keypoints (points defining the vertices of an object;
M01_MOAV4303_04_GE_C01.INDD 68
27/11/14 9:33 AM
Section 1.9 Understanding the Problem 69
they are covered in more detail in Chapters 8 and 13) before proceeding any further with the analysis. 2. Selecting inappropriate types of elements Understanding the underlying theory will benefit you the most in this respect. You need to fully grasp the limitations of a given type of element and understand to which type of problems it applies. 3. Poor element shape and size after meshing This area is a very important part of any finite element analysis. Inappropriate element shape and size will influence the accuracy of your results. It is important that the user understands the difference between free meshing (using mixedarea element shapes) and mapped meshing (using all quadrilateral area elements or all hexahedral volume elements) and the limitations associated with them. These concepts will be explained in more detail in Chapters 8 and 13. 4. Applying wrong boundary conditions and loads This step is usually the most difficult aspect of modeling. It involves taking an actual problem and estimating the loading and the appropriate boundary conditions for a finite element model. This step requires good judgment and some experience. You must always find ways to check your results. While experimental testing of your model may be the best way to do so, it may be expensive or time consuming. You should always start by applying equilibrium conditions and energy balance to different portions of a model to ensure that the physical laws are not violated. For example, for static models, the sum of the forces acting on a freebody diagram of your model must be zero. This concept will allow you to check for the accuracy of computed reaction forces. You may want to consider defining and mapping stresses along an arbitrary cross section and integrating this information. The resultant internal forces computed in this manner must balance against external forces. In a heat transfer problem under steadystate conditions, apply conservation of energy to a control volume surrounding an arbitrary node. Are the energies flowing into and out of a node balanced? At the end of particular chapters in this text, a section is devoted to verifying the results of your models. In these sections, problems will be solved using ANSYS, and the steps for verifying results will be shown. 1.9 Understanding the Problem You can save lots of time and money if you first spend a little time with a piece of paper and a pencil to try to understand the problem you are planning to analyze. Before initiating numerical modeling on the computer and generating a finite element model, it is imperative that you develop a sense of or a feel for the problem. There are many questions that a good engineer will ask before proceeding with the modeling process: Is the material under axial loading? Is the body under bending moments or twisting moments or a combination of the two? Do we need to worry about buckling? Can we approximate the behavior of the material with a twodimensional model? Does heat transfer play a significant role in the problem? Which modes of heat transfer are influential? If you choose to employ FEA, “backoftheenvelope” calculations will greatly enhance your understanding of the problem, in turn helping you to develop a good, reasonable finite element model, particularly in terms of your selection of element types. Some practicing
M01_MOAV4303_04_GE_C01.INDD 69
27/11/14 9:33 AM
70 Chapter 1 Introduction
engineers still use finite element analysis to solve a problem that could have been solved more easily by hand by someone with a good grasp of the fundamental concepts of the mechanics of materials and heat transfer. To shed more light on this very important point, consider the following examples. Example 1.6 Imagine that by mistake, an empty coffee pot has been left on a heating element. Assuming that the heater puts approximately 20 Watts (typically, a heater creates a lower wattage) into the bottom of the pot, determine the temperature distribution within the glass if the surrounding air is at 25°C, with a corresponding heat transfer coefficient h = 15 W/m2 # K. The pot is cylindrical in shape, with a diameter of 14 cm and height of 14 cm, and the glass is 3 mm thick.
Heating plate
This problem is first analyzed using a finite element model. After you study threedimensional thermalsolid elements (discussed in Chapter 13), you will revisit this problem (see Problem 13.11) and be asked to solve it using ANSYS. As you will learn later, a solid model of the pot is created and meshed and the appropriate boundary conditions are applied and the temperature solutions is then obtained. The results of this analysis is shown in Figure 1.21. From the results of finite element analysis we find that the maximum temperature of 113.18°C occurs at the bottom of the pot in the center location, as shown in Figure 1.21. This is a good example of a problem that could have been solved more easily by hand by someone with a good grasp of the fundamental concepts of heat transfer. We can approximate the temperature of the glass by applying the energy balance to the bottom of the pot and assuming a onedimensional model. Because the pot is made of thin glass, we can neglect the spatial temperature variation within the glass. Under steadystate
M01_MOAV4303_04_GE_C01.INDD 70
27/11/14 9:33 AM
Section 1.9 Understanding the Problem 71
Figure 1.21 The temperature distribution in the pot of Example 1.6.
conditions, the heat flux added into the bottom of the glass is approximately equal to the rate of energy convected away by air. Thus, we employ Newton’s law of cooling q″ = h(Ts  Tf) (1.65)
where q″ h Ts Tf
= = = =
heat flux, W/m2 heat transfer coefficient, W/m2 # °C (W/m2 # K) surface temperature of the coffee pot,°C surrounding air temperature,°C
We can estimate the heat flux into the bottom of the pot: q″ =
20 W p (0.14 m)2 4
= 1299 W/m2
and substituting for heat flux, h, and Tf into Eq. (1.65), and solving for Ts, 1299 W/m2 = (15 W/m # °C)(Ts  25)
M01_MOAV4303_04_GE_C01.INDD 71
S
Ts = 111.6°C
27/11/14 9:33 AM
72 Chapter 1 Introduction
As you can see, the temperature result obtained by hand calculation (Ts = 111.6°C) is very close to the result of our finite element model (Tmax = 113.18°C). Thus, there was no need to resort to finite element formulation to solve the above problem. Example 1.7 Consider the torsion of a steel bar (G = 11 * 103 ksi) having a rectangular cross section, as shown in the accompanying figure. Under the loading shown, the angle of twist is measured to be u = 0.0005 rad/in. We are interested in determining the location(s) and magnitude of the maximum shear stress. Steel bar G = 11 * 103 ksi P
0.5
in 1 in
1 in
P
Again, we have analyzed this problem using a finite element model. All of the steps leading to the ANSYS solution are given in Example 10.1 (revisited). The results of this analysis are shown in Figure 1.22. The results of finite element analysis show that the maximum shear stress of 2558 lb/in2 occurs at the midsection of the rectangle. This is another example of a problem that could have been solved more easily by hand by someone with a good grasp of the fundamental concepts of mechanics of materials. As you will learn in Chapter 10, Section 10.1, there are analytical solutions that we could employ to solve problems dealing with torsion of members with rectangular crosssectional area. When a torque is applied to a straight bar with a rectangular crosssectional area, within the elastic region of the material, the maximum shearing stress and an angle of twist caused by the torque are given by tmax =
T c1wh2
where tmax = maximum shear stress, lb/in2 T = applied torque, lb # in
M01_MOAV4303_04_GE_C01.INDD 72
27/11/14 9:33 AM
Section 1.9 Understanding the Problem 73
Figure 1.22 The shear stress distribution for the steel bar of Example 1.7.
w = width of the rectangular [email protected], in h = height of the rectangular [email protected], in c1 = a constant coefficient that depends on aspect ratio of the cross section, 0.246; see Table 10.1 and TL c2 Gwh3 L = length of the bar, in G = shear modulus or modulus of rigidity of material, lb/in2 c2 = a constant coefficient that depends on aspect ratio of the cross section, 0.229; see Table 10.1 u =
Substituting into the above equations appropriate values, we get T(1 in) TL = 0.0005 rad/in = 1 T = 157.5 lb # in 3 c2 Gwh 0.229(11 * 106 lb/in2)(1 in)(0.5 in)3 T 157.5 lb # in = = = 2560 lb/in2 2 c1 wh 0.246(1 in)(0.5 in)2
u = tmax
When comparing 2560 lb/in2 to the FEA results of 2558 lb/in2, you see that we could have saved lots of time by calculating the maximum shear stress using the analytical solution and avoided generating a finite element model.
M01_MOAV4303_04_GE_C01.INDD 73
27/11/14 9:33 AM
74 Chapter 1 Introduction
Summary At this point you should
1. have a good understanding of the physical properties and the parameters that characterize the behavior of an engineering system. Examples of these properties and design parameters are given in Tables 1.2 and 1.3. 2. realize that a good understanding of the fundamental concepts of the finite element method will benefit you by enabling you to use ANSYS more effectively. 3. know the seven basic steps involved in any finite element analysis, as discussed in Section 1.4. 4. understand the differences among direct formulation, minimum total potential energy formulation, and the weighted residual methods (particularly the Galerkin formulation). 5. know that it is wise to spend some time to gain a full understanding of a problem before initiating a finite element model of the problem. There may even exist a reasonable closedform solution to the problem, and thus you can save lots of time and money. 6. realize that you must always find a way to verify your FEA results.
References ASHRAE Handbook, Fundamental Volume, American Society of Heating, Refrigerating, and AirConditioning Engineers, Atlanta, 1993. Bickford, B. W., A First Course in the Finite Element Method, Richard D. Irwin, Burr Ridge, 1989. Clough, R. W., “The Finite Element Method in Plane Stress Analysis, Proceedings of American Society of Civil Engineers, 2nd Conference on Electronic Computations,” Vol. 23, 1960, pp. 345–378. Cook, R. D., Malkus, D. S., and Plesha, M. E., Concepts and Applications of Finite Element Analysis, 3rd. ed., New York, John Wiley and Sons, 1989. Courant, R., “Variational Methods for the Solution of Problems of Equilibrium and Vibrations,” Bulletin of the American Mathematical Society, Vol. 49, 1943, pp. 1–23. Hrennikoff, A., “Solution of Problems in Elasticity by the Framework Method,” J. Appl. Mech., Vol. 8, No. 4, 1941, pp. A169–A175. Levy, S., “Structural Analysis and Influence Coefficients for Delta Wings,” Journal of the Aeronautical Sciences, Vol. 20, No. 7, 1953, pp. 449–454. Patankar, S. V., Numerical Heat Transfer and Fluid Flow, New York, McGrawHill, 1991. Zienkiewicz, O. C., and Cheung, Y. K. K., The Finite Element Method in Structural and Continuum Mechanics, London, McGrawHill, 1967. Zienkiewicz, O. C., The Finite Element Method, 3d. ed., London, McGrawHill, 1979.
Problems 1. Solve Example 1.1 using: (a) two elements, and (b) eight elements. Compare your results to the exact values.
2. A concrete table columnsupport with the profile shown in the accompanying figure is to carry a load of approximately 500 lb. Using the direct method discussed in Section 1.5,
M01_MOAV4303_04_GE_C01.INDD 74
27/11/14 9:33 AM
Chapter 1 Problems 75 determine the deflection and average normal stresses along the column. Divide the column into five elements. (E = 3.27 * 103 ksi) 500 lb
9 in.
28 in.
4 in.
3 in.
1 ft
3. An aluminum strap with a thickness of 5 mm and the profile shown in the accompanying figure is to carry a load of 2000 N. Using the direct method discussed in Section 1.5, determine the deflection and the average normal stress along the strap. Divide the strap into three elements. This problem may be revisited again in Chapter 10, where a more indepth analysis may be sought. (E = 70 GPa) 10 cm
3 cm
1.5 cm
12 cm
3 cm
2000 N
M01_MOAV4303_04_GE_C01.INDD 75
27/11/14 9:33 AM
76 Chapter 1 Introduction
4. A thin steel plate with the profile shown in the accompanying figure is subjected to an axial load. Approximate the deflection and the average normal stresses along the plate using the model shown in the figure. The plate has a thickness of 0.125 in and a modulus of elasticity E = 28 * 103 ksi. You will be asked to use ANSYS to analyze this problem again in Chapter 10. 4 in
2 in
k1
–12 in 12 in
k2
–12 in
–12 in
2 in
500 lb
k3
k4
k5
k6
500 lb
5. Apply the statics equilibrium conditions directly to each node of the thin steel plate (using a finite element model) in Problem 4.
6. For the spring system shown in the accompanying figure, determine the displacement of each node. Start by identifying the size of the global matrix. Write down elemental stiffness matrices, and show the position of each elemental matrix in the global matrix. Apply the boundary conditions and loads. Solve the set of linear equations. Also compute the reaction forces. lb k2 = 9 — in lb k1 = 5 — in
20 lb lb k3 = 5 — in
lb k5 = 10 — in 20 lb
lb k6 = 20 — in
lb k4 = 20 — in
M01_MOAV4303_04_GE_C01.INDD 76
27/11/14 9:33 AM
Chapter 1 Problems 77
7. A typical exterior masonry wall of a house, shown in the accompanying figure, consists of the items in the accompanying table. Assume an inside room temperature of 68°F and an outside air temperature of 10°F, with an exposed area of 150 ft2. Determine the temperature distribution through the wall. Also calculate the heat loss through the wall.
Resistance hr # ft2 # °F/Btu
Ufactor Btu/hr # ft2 # °F
1. Outside film resistance (winter, 15mph wind)
0.17
5.88
2. Face brick (4 in)
0.44
2.27
3. Cement mortar (1/2 in)
0.1
4. Cinder block (8 in)
1.72
0.581
5. Air space (3/4 in)
1.28
0.781
6. Gypsum board (1/2 in)
0.45
2.22
7. Inside film resistance (winter)
0.68
1.47
Items
10.0
1
2 3
4
5
6
7
8. In order to increase the thermal resistance of a typical exterior frame wall, such as the one in Example 1.2, it is customary to use 2 * 6 studs instead of 2 * 4 studs to allow for placement of more insulation within the wall cavity. A typical exterior (2 * 6) frame wall of a house consists of the materials shown in the accompanying figure. Assume an inside room temperature of 68°F and an outside air temperature of 20°F, with an exposed area of 150 ft2. Determine the temperature distribution through the wall.
Items
Resistance hr # ft2 # °F/Btu
Ufactor Btu/hr # ft2 # °F
1. Outside film resistance (winter, 15mph wind)
0.17
2. Siding, wood (1/2 * 8 lapped)
0.81
1.23
3. Sheathing (1/2 in regular)
1.32
0.76
4. Insulation batt (512 in)
19.0
5.88
0.053
5. Gypsum wall board (1/2 in)
0.45
2.22
6. Inside film resistance (winter)
0.68
1.47
1
2
3
4
5
6
9. Assuming the moisture can diffuse through the gypsum board in Problem 8, where should you place a vapor barrier to avoid moisture condensation? Assume an indoor air temperature of 68°F with relative humidity of 40%. 10. A typical ceiling of a house consists of the items in the accompanying table. Assume an inside room temperature of 755F and an attic air temperature of 255F, with an exposed area of 1000 ft2. Determine the temperature distribution through the ceiling. Also calculate heat loss through the ceiling.
M01_MOAV4303_04_GE_C01.INDD 77
27/11/14 9:33 AM
78 Chapter 1 Introduction
Resistance hr # ft2 # °F/Btu
Ufactor Btu/hr # ft2 # °F
0.7
1.5
19.0
0.05
3. Gypsum board (1/2 in)
0.5
2.3
4. Inside film resistance (winter)
0.7
1.5
Items 1. Inside attic film resistance 2. Insulation batt (6 in)
1
2
3
4
11. A typical [email protected] solid wood core door exposed to winter conditions has the characteristics shown in the accompanying table. Assume an inside room temperature of 505F and an outside air temperature of 255F, with an exposed area of 25 ft2. (a) Determine the inside and outside temperatures of the door’s surface. (b) Determine heat loss through the door. Resistance hr # ft2 # °F/Btu
Items
Ufactor Btu/hr # ft2 # °F
1. Outside film resistance (winter, 15mph wind)
0.15
6
2. 138 @in
0.4
2.5
0.68
1.5
solid wood core
3. Inside film resistance (winter)
12. The concrete table columnsupport in Problem 2 is reinforced with three [email protected] steel rods,
as shown in the accompanying figure. Determine the deflection and average normal stresses along the column under a load of 1000 lb. Divide the column into five elements. (EC = 3.27 * 103 ksi; Es = 29 * 103 ksi) 1000 lb
6 in
9 in
28 in
4 in
1 in
1 in 3 in
3 in 12 in
1 ft
13. Compute the total strain energy for the concrete table columnsupport in Problem 12. 14. A 10in slender rod weighing 10 lb is supported by a spring with a stiffness k = 50 lb/in.
A force P = 50 lb is applied to the rod at the location shown in the accompanying figure. Determine the deflection of the spring (a) by drawing a freebody diagram of the rod and
M01_MOAV4303_04_GE_C01.INDD 78
27/11/14 9:33 AM
Chapter 1 Problems 79 applying the statics equilibrium conditions, and (b) by applying the minimum total potential energy concept. P = 50 lb 6 in 2 in
— k = 50 lb in
W = 10 lb
15. In a DC electrical circuit, Ohm’s law relates the voltage drop V2  V1 across a resistor to
a current I flowing through the element and the resistance R according to the equation V2  V1 = RI. R
V2
V1
I
Using direct formulation, show that for a resistance element comprising two nodes, the conductance matrix, the voltage drop, and the currents are related according to the equation 1 1 J R 1
1 V I R b 1r = b 1r 1 V2 I2
16. Use the results of Problem 15 to set up and solve for the voltage drop in each branch of the circuit shown in the accompanying figure. 6Æ
20 mA
M01_MOAV4303_04_GE_C01.INDD 79
2
12 Æ
18 Æ
27/11/14 9:33 AM
80 Chapter 1 Introduction
17. The deformation of a simply supported beam under a distributed load, shown in the accompanying figure, is governed by the relationship M(X ) d 2Y = 2 EI dX where M(X) is the internal bending moment and is given by M(X ) =
wX(L  X ) 2
Y w
X
L
Derive the equation for the exact deflection. Assume an approximate deflection solution of the form X 2 X Y(X) = c1 J ¢ ≤  ¢ ≤ R L L Use the following methods to evaluate c1: (a) the collocation method and (b) the subdomain method. Also, using the approximate solutions, determine the maximum deflection of the beam if a W27 * 84 (wide flange shape) with a span of L = 22 ft supports a distributed load of w = 6 kips/ft. 18. For the example problem used throughout Section 1.7, assume an approximate solution of the form u(y) = c1y + c2y2 + c3y3 + c4y4. Using the collocation, subdomain, Galerkin, and leastsquares methods, determine the unknown coefficients c1, c2, c3, and c4. Compare your results to those obtained in Section 1.7.
19. The leakage flow of hydraulic fluid through the gap between a piston–cylinder arrangement may be modeled as laminar flow of fluid between infinite parallel plates, as shown in the accompanying figure. This model offers reasonable results for relatively small gaps. The differential equation governing the flow is m
dp d2u = 2 dx dy
u H
y x
M01_MOAV4303_04_GE_C01.INDD 80
27/11/14 9:33 AM
Chapter 1 Problems 81 dp is the dx pressure drop and is constant. Derive the equation for the exact fluid velocities. Assume an py approximate fluid velocity solution of the form u(y) = c1 Jsin¢ ≤ R . Use the following H methods to evaluate c1: (a) the collocation method and (b) the subdomain method. Compare the approximate results to the exact solution. where m is the dynamic viscosity of the hydraulic fluid, u is the fluid velocity, and
20. Use the Galerkin and leastsquares methods to solve Problem 19. Compare the approximate results to the exact solution.
21. For the cantilever beam shown in the accompanying figure, the deformation of the beam under a load P is governed by the relationship M(X ) d 2Y = EI dX 2 where M(X) is the internal bending moment and is M(X ) = PX P Y
X
x L
Derive the equation for the exact deflection. Assume an appropriate form of a polynomial function. Keep in mind that the assumed solution must satisfy the given boundary conditions. Use the subdomain method and the Galerkin method to solve for the unknown coefficients of the assumed solution. 22. A shaft is made of three parts, as shown in the accompanying figure. Parts AB and CD are made of the same material with a modulus of rigidity of G = 9.8 * 103 ksi, and each has a diameter of 1.5 in. Segment BC is made of a material with a modulus of rigidity of G = 11.2 * 103 ksi and has a diameter of 1 in. The shaft is fixed at both ends. A torque of 2400 lb # in is applied at C. Using three elements, determine the angle of twist at B and C and the torsional reactions at the boundaries.
A
B
2 ft
M01_MOAV4303_04_GE_C01.INDD 81
C
1.5 ft
D
2 ft
27/11/14 9:34 AM
82 Chapter 1 Introduction
23. For the shaft in Problem 22, replace the torque at C by two equal torques of 1500 lb # in
at B and C. Compute the angle of twist at B and C and the torsional reactions at the boundaries. 24. Consider a plate with a variable cross section supporting a load of 1500 lb, as shown in the accompanying figure. Using direct formulation, determine the deflection of the bar at locations y = 2.5 in, y = 7.5 in, and y = 10 in. The plate is made of a material with a modulus of elasticity E = 10.6 * 103 ksi. 4 in thickness = 0.125 in
2.5 in
10 in
–12 in
5 in
2 in
1500 lb
25. Consider the springs in parallel and in series, as shown in the accompanying figure. Realizing that deformation of each spring in parallel is the same, and the applied force must equal the sum of forces in individual springs, show that for the springs in parallel the equivalent spring constant ke is ke = k1 + k2 + k3 For the springs in series, realizing that the total deformation of the springs is the sum of the deformations of the individual springs, and the force in each spring equals the applied force, show that for the springs in series, the equivalent spring constant is ke =
M01_MOAV4303_04_GE_C01.INDD 82
1 1 1 1 + + k1 k2 k3
27/11/14 9:34 AM
Chapter 1 Problems 83
k1
k2
k1
k3
k2
F
k3
F
26. Use the results of Problem 25 and determine the equivalent spring constant for the system of the springs shown in the accompanying figure.
2k
k
2k
3k
3k
3k
F
27. Determine the equivalent spring constant for the cantilever beam shown in the accompanying figure. F E, I
x L
M01_MOAV4303_04_GE_C01.INDD 83
27/11/14 9:34 AM
84 Chapter 1 Introduction
28. Use the results of Problem 27 and Eq. (1.5) to determine the equivalent spring constant for the system shown in the accompanying figure.
E2, A2 L2 E1, I1 F
L1 L3
E3, A3
29. Determine the equivalent spring constant for the system shown. Determine the deflection of point A, using the minimum total potential energy concept.
E2, A2
L2 E1, I1
F A
L1
L1
2
2
L3
E3, A3
M01_MOAV4303_04_GE_C01.INDD 84
27/11/14 9:34 AM
Chapter 1 Problems 85
30. Neglect the mass of the connecting rod and determine the deflection of each spring for the system shown in the accompanying figure (a) by applying the statics equilibrium conditions, and (b) by applying the minimum total potential energy concept.
k
L
F
L
L 2k
M01_MOAV4303_04_GE_C01.INDD 85
27/11/14 9:34 AM
C h a p t e r
2
Matrix Algebra
In Chapter 1 we discussed the basic steps involved in any finite element analysis. These steps include discretizing the problem into elements and nodes, assuming a function that represents behavior of an element, developing a set of equations for an element, assembling the elemental formulations to present the entire problem, and applying the boundary conditions and loading. These steps lead to a set of linear (nonlinear for some problems) algebraic equations that must be solved simultaneously. A good understanding of matrix algebra is essential in formulation and solution of finite element models. As is the case with any topic, matrix algebra has its own terminology and follows a set of rules. We provide an overview of matrix terminology and matrix algebra in this chapter. The main topics discussed in Chapter 2 include 2.1 Basic Definitions 2.2 Matrix Addition or Subtraction 2.3 Matrix Multiplication 2.4 Partitioning of a Matrix 2.5 Transpose of a Matrix 2.6 Determinant of a Matrix 2.7 Solutions of Simultaneous Linear Equations 2.8 Inverse of a Matrix 2.9 Eigenvalues and Eigenvectors 2.10 Using MATLAB to Manipulate Matrices 2.11 Using Excel to Manipulate Matrices 2.1 Basic Definitions A matrix is an array of numbers or mathematical terms. The numbers or the mathematical terms that make up the matrix are called the elements of matrix. The size of a matrix is defined by its number of rows and columns. A matrix may consist of m rows and n
M02_MOAV4303_04_GE_C02.INDD 86
27/11/14 9:18 AM
Section 2.1 Basic Definitions 87
columns. For example, 6 [N ] = C 1 5
5 26 8
9 14 S 0
sin u cos u 0 0
cos u sin u [T] = D 0 0
0f (x, y, z) 0x 0f (x, y, z) 5 L6 = f v 0y 0f (x, y, z) 0z
L
L
[I] = F
0 0 cos u sin u W
x dx
L
0
0
L
W
x2 dx LL
0
0 0 T sin u cos u
y dy V
y2 dy LL
0
Matrix [N] is a 3 by 3 (or 3 * 3) matrix whose elements are numbers, [T] is a 4 * 4 that has sine and cosine terms as its elements, 5 L6 is a 3 * 1 matrix with its elements representing partial derivatives, and [I] is a 2 * 2 matrix with integrals for its elements. The [N], [T], and [I] are square matrices. A square matrix has the same number of rows and columns. The element of a matrix is denoted by its location. For example, the element in the first row and the third column of a matrix [B] is denoted by b13, and an element occurring in matrix [A] in row 2 and column 3 is denoted by the term a23. In this book, we denote the matrix by a boldface letter in brackets [] and 5 6 , for example: [K], [T], 5 F6 , and the elements of matrices are represented by regular lowercase letters. The 5 6 is used to distinguish a column matrix. Column Matrix and Row Matrix
A column matrix is defined as a matrix that has one column but could have many rows. On the other hand, a row matrix is a matrix that has one row but could have many columns. Examples of column and row matrices follow. 0f (x, y, z) 1 0x x1 0f (x, y, z) 5 5A6 = d t , 5 X 6 = c x2 s , and 5 L6 = f v are examples of column matrices, 2 0y x3 0f (x, y, z) 3 0z whereas [C] = [5 0 2
3] and [Y] = [y1 y2 y3] are examples of row matrices.
Diagonal, Unit, and Band (Banded) Matrix A diagonal matrix is one that has elements only along its principal diagonal; the elements are zero everywhere else (aij = 0 when i ≠ j). An example of a 4 * 4 diagonal
M02_MOAV4303_04_GE_C02.INDD 87
27/11/14 9:19 AM
88 Chapter 2 Matrix Algebra
matrix follows: a1 0 [A] = D 0 0
0 a2 0 0
0 0 a3 0
0 0 T 0 a4
The diagonal along which a1, a2, a3, and a4 lies is called the principal diagonal. An identity or unit matrix is a diagonal matrix whose elements consist of a value 1. An example of an identity matrix follows: 1 0 0 [I] = G . . 0 0
0 1 0 . . 0 0
0 0 1 . . 0 0
. . . . . . .
. . . . . . .
0 0 0 . . 1 0
0 0 0 .W . 0 1
A banded matrix is a matrix that has a band of nonzero elements parallel to its principal diagonal. As shown in the example that follows, all other elements outside the band are zero. b11 b21 0 [B] = G 0 0 0 0
b12 b22 b32 0 0 0 0
0 b23 b33 b43 0 0 0
0 0 b34 b44 b54 0 0
0 0 0 b45 b55 b65 0
0 0 0 0 b56 b66 b76
0 0 0 0 W 0 b67 b77
Upper and Lower Triangular Matrix An upper triangular matrix is one that has zero elements below the principal diagonal (uij = 0 when i 7 j), and the lower triangular matrix is one that has zero elements above the principal diagonal (lij = 0 when i 6 j). Examples of upper triangular and lower triangular matrices are shown below. u11 0 [U] = D 0 0
M02_MOAV4303_04_GE_C02.INDD 88
u12 u22 0 0
u13 u23 u33 0
u14 u24 T u34 u44
l11 l [L] = D 21 l31 l41
0 l22 l32 l42
0 0 l33 l43
0 0 T 0 l44
27/11/14 9:19 AM
Section 2.3 Matrix Multiplication 89
2.2 Matrix Addition or Subtraction Two matrices can be added together or subtracted from each other provided that they are of the same size—each matrix must have the same number of rows and columns. We can add matrix [A]m * n of dimension m by n to matrix [B]m * n of the same dimension by adding the like elements. Matrix subtraction follows a similar rule, as shown. a11 a21 [A] { [B] = E . . am1
a12 a22 . . am2
. . . . .
(a11 { b11) (a21 { b21) = E . . (am1 { bm1)
. . . . .
a1n b11 a2n b21 . U { E . . . amn bm1
(a12 { b12) (a22 { b22) . . (am2 { bm2)
. . . . .
b12 b22 . . bm2 . . . . .
. . . . .
. . . . .
b1n b2n . U . bmn
(a1n { b1n) (a2n { b2n) . U . (amn { bmn)
The rule for matrix addition or subtraction can be generalized in the following manner. Let us denote the elements of matrix [A] by aij and the elements of matrix [B] by bij, where the number of rows i varies from 1 to m and the number of columns j varies from 1 to n. If we were to add matrix [A] to matrix [B] and denote the resulting matrix by [C], it follows that [A] + [B] = [C] and cij = aij + bij for i = 1, 2, c, m and j = 1, 2, c, n (2.1)
2.3 Matrix Multiplication In this section we discuss the rules for multiplying a matrix by a scalar quantity and by another matrix. Multiplying a Matrix by a Scalar Quantity When a matrix [A] of size m * n is multiplied by a scalar quantity such as b, the operation results in a matrix of the same size m * n, whose elements are the product of elements in the original matrix and the scalar quantity. For example, when we multiply matrix [A] of size m * n by a scalar quantity b, this operation results in another matrix
M02_MOAV4303_04_GE_C02.INDD 89
27/11/14 9:19 AM
90 Chapter 2 Matrix Algebra
of size m * n, whose elements are computed by multiplying each element of matrix [A] by b, as shown below.
a11 a21 b[A] = b E . . am1
a12 a22 . . am2
. . . . .
. . . . .
a1n ba11 a2n ba21 . U = E . . . amn bam1
ba12 ba22 . . bam2
. . . . .
. . . . .
ba1n ba2n . U (2.2) . bamn
Multiplying a Matrix by Another Matrix Whereas any size matrix can be multiplied by a scalar quantity, matrix multiplication can be performed only when the number of columns in the premultiplier matrix is equal to the number of rows in the postmultiplier matrix. For example, matrix [A] of size m * n can be premultiplied by matrix [B] of size n * p because the number of columns n in matrix [A] is equal to number of rows n in matrix [B]. Moreover, the multiplication results in another matrix, say [C], of size m * p. Matrix multiplication is carried out according to the following rule: [A]m * n[B]n * p = [C]m * p must match
a11 a12 a21 a22 [A][B] = E . . . . am1 am2
. . . . .
. a1n b11 . a2n b21 . . UE . . . . . amn bn1
b12 b22 . . bn2
. . . . .
. . . . .
b1p c11 b2p c21 . U = E . . . bnp cm1
c12 c22 . . cm2
. . . . .
. . . . .
c1p c2p . U . cmp
where the elements in the first column of the [C] matrix are computed from c11 = a11b11 + a12b21 + c + a1nbn1 c21 = a21b11 + a22b21 + c + a2nbn1 .................................. cm1 = am1b11 + am2b21 + c + amnbn1 and the elements in the second column of the [C] matrix are c12 = a11b12 + a12b22 + c + a1nbn2 c22 = a21b12 + a22b22 + c + a2nbn2 .................................. cm2 = am1b12 + am2b22 + c + amnbn2 and similarly, the elements in the other columns are computed, leading to the last column of the [C] matrix c1p = a11b1p + a12b2p + c + a1nbnp c2p = a21b1p + a22b2p + c + a2nbnp ................................... cmp = am1b1p + am2b2p + c + amnbnp
M02_MOAV4303_04_GE_C02.INDD 90
27/11/14 9:19 AM
Section 2.3 Matrix Multiplication 91
The multiplication procedure that leads to the values of the elements in the [C] matrix may be represented in a compact summation form by cmp = a amkbkp (2.3) n
k=1
When multiplying matrices, keep in mind the following rules. Matrix multiplication is not commutative except for very special cases. [A][B] ≠ [B][A] (2.4)
Matrix multiplication is associative; that is
[A]([B][C]) = ([A][B])[C] (2.5) The distributive law holds true for matrix multiplication; that is ([A] + [B])[C] = [A][C] + [B][C] (2.6)
or
[A]([B] + [C]) = [A][B] + [A][C] (2.7)
For a square matrix, the matrix may be raised to an integer power n in the following manner: n
n times 6
[A] = [A][A]c[A] (2.8)
This may be a good place to point out that if [I] is an identity matrix and [A] is a square matrix of matching size, then it can be readily shown that the product of [I ][ A] = [ A][ I ] = [ A]. See Example 2.1 for the proof. Example 2.1 Given matrices 0 [A] = C 8 9
5 3 2
0 4 7 S , [B] = C 7 9 1
6 2 3
2 1 3 S , and 5C6 = c 2 s 4 5
perform the following operations: a. [A] + [B] = ? b. [A]  [B] = ? c. 3[A] = ? d. [A][B] = ? e. [A]5C6 = ? f. [A]2 = ? g. Show that [I ][A] = [A][I ] = [A]
M02_MOAV4303_04_GE_C02.INDD 91
27/11/14 9:19 AM
92 Chapter 2 Matrix Algebra
We will use the operation rules discussed in the preceding sections to answer these questions. a. [A] + [B] = ? 0 [A] + [B] = C 8 9
2 3 S 4
5 3 2
0 4 7S + C7 9 1
6 2 3
(0 + 4) = C (8 + 7) (9 + 1)
(5 + 6) (3 + 2) ( 2 + 3)
(0 + ( 2)) 4 (7 + 3) S = C 15 (9 + ( 4)) 10
11 5 1
2 10 S 5
b. [A]  [B] = ? 0 [A]  [B] = C 8 9
5 3 2
(04) = C (87) (9 1)
0 4 7S  C7 9 1 (56) (32) ( 23)
6 2 3
2 3 S 4
(0( 2)) 4 (73) S = C 1 (9 ( 4)) 8
1 1 5
2 4S 13
c. 3[A] = ? 0 3[A] = 3C 8 9
5 3 2
0 0 7 S = C (3) (8) 9 (3) (9)
(3) (5) (3) (3) (3) ( 2)
0 0 (3) (7) S = C 24 (3) (9) 27
15 9 6
0 21 S 27
d. [A][B] = ? 0 [A][B] = C 8 9
5 3 2
0 4 7S C7 9 1
6 2 3
2 3 S = 4
(0)(4) + (5)(7) + (0)(1) (0)(6) + (5)(2) + (0)(3) (0)( 2) + (5)(3) + (0)( 4) C (8)(4) + (3)(7) + (7)(1) (8)(6) + (3)(2) + (7)(3) (8)( 2) + (3)(3) + (7)( 4) S (9)(4) + ( 2)(7) + (9)(1) (9)(6) + ( 2)(2) + (9)(3) (9)( 2) + ( 2)(3) + (9)( 4)
35 = C 60 31
10 75 77
15 35 S 60
e. [A]5 C6 = ? 0 [A] 5C6 = C 8 9
M02_MOAV4303_04_GE_C02.INDD 92
5 3 2
0 1 (0) ( 1) + (5) (2) + (0) (5) 10 7 S c 2 s = c (8) ( 1) + (3) (2) + (7) (5) s = c 33 s 9 5 (9) ( 1) + ( 2) (2) + (9) (5) 32
27/11/14 9:19 AM
Section 2.4 Partitioning of a Matrix 93
f. [A]2 = ? 0 [A] = [A][A] = C 8 9
5 3 2
2
0 0 7S C8 9 9
5 3 2
0 40 7 S = C 87 9 65
15 35 21
35 84 S 67
5 3 2
0 0 7S = C8 9 9
5 3 2
0 7 S and 9
0 1 0
0 0 0S = C8 1 9
5 3 2
0 7S 9
g. Show that [I ][A] = [A][I ] = [A]
1 [I ][A] = C 0 0
0 1 0
0 [A][I ] = C 8 9
5 3 2
0 0 0S C8 1 9 0 1 7S C0 9 0
2.4 Partitioning of a Matrix Finite element formulation of complex problems typically involves relatively large sized matrices. For these situations, when performing numerical analysis dealing with matrix operations, it may be advantageous to partition the matrix and deal with a subset of elements. The partitioned matrices require less computer memory to perform the operations. Traditionally, dashed horizontal and vertical lines are used to show how a matrix is partitioned. For example, we may partition matrix [A] into four smaller matrices in the following manner: a11 a21 [A] = E a31 a41 a51 and in terms of submatrices [A] = J
a12 a22 a32 a42 a52
a13 a23 a33 a43 a53
A11 A21
A12 R, A22
a14 a24 a34 a44 a54
a15 a25 a35 a45 a55
a16 a26 a36 U a46 a56
where a11 a21
a12 a22
a13 R a23
[A12] = J
a14 a24
a15 a25
a16 R a26
a31 [A21] = C a41 a51
a32 a42 a52
a33 a43 S a53
a34 [A22] = C a44 a54
a35 a45 a55
a36 a46 S a56
[A11] = J
It is important to note that matrix [A] could have been partitioned in a number of other ways, and the way a matrix is partitioned would define the size of submatrices.
M02_MOAV4303_04_GE_C02.INDD 93
27/11/14 9:19 AM
94 Chapter 2 Matrix Algebra
Addition and Subtraction Operations Using Partitioned Matrices Now let us turn our attention to matrix operations dealing with addition, subtraction, or multiplication of two matrices that are partitioned. Consider matrix [B] having the same size (5 * 6) as matrix [A]. If we partition matrix [B] in exactly the same way we partitioned [A] previously, then we can add the submatrices in the following manner: b11 b21 [B] = E b31 b41 b51
b12 b22 b32 b42 b52
b13 b23 b33 b43 b53
b14 b24 b34 b44 b54
b15 b25 b35 b45 b55
b16 b26 b36 U b46 b56
where b11 b21
b12 b22
b13 R b23
[B12] = J
b14 b24
b15 b25
b16 R b26
b31 [B21] = C b41 b51
b32 b42 b52
b33 b43 S b53
b34 [B22] = C b44 b54
b35 b45 b55
b36 b46 S b56
[B11] = J
Then, using submatrices we can write [A] + [B] = J
A11 + B11 A21 + B21
A12 + B12 R A22 + B22
Matrix Multiplication Using Partitioned Matrices As mentioned earlier, matrix multiplication can be performed only when the number of columns in the premultiplier matrix is equal to the number of rows in the postmultiplier matrix. Referring to [A] and [B] matrices of the preceding section, because the number of columns in matrix [A] does not match the number of rows of matrix [B], then matrix [B] cannot be premultiplied by matrix [A]. To demonstrate matrix multiplication using submatrices, consider matrix [C] of size 6 * 3, which is partitioned in the manner shown below. c11 c21 c [C] = F 31 c41 c51 c61 [C] = J
M02_MOAV4303_04_GE_C02.INDD 94
C11 C21
c12 c22 c32 c42 c52 c62
c13 c23 c33 V c43 c53 c63
C12 R C22
27/11/14 9:19 AM
Section 2.4 Partitioning of a Matrix 95
where c11 [C11] = C c21 c31
c12 c22 S c32
c13 and 5 C12 6 = c c23 s c33
c41 [C21] = C c51 c61
c42 c52 S c62
c43 and 5 C22 6 = c c53 s c63
Next, consider premultiplying matrix [C] by matrix [A]. Let us refer to the results of this multiplication by matrix [D] of size 5 * 3. In addition to paying attention to the size requirement for matrix multiplication, to carry out the multiplication using partitioned matrices, the premultiplying and postmultiplying matrices must be partitioned in such a way that the resulting submatrices conform to the multiplication rule. That is, if we partition matrix [A] between the third and the fourth columns, then matrix [C] must be partitioned between the third and the fourth rows. However, the column partitioning of matrix [C] may be done arbitrarily, because regardless of how the columns are partitioned, the resulting submatrices will still conform to the multiplication rule. In other words, instead of partitioning matrix [C] between columns two and three, we could have partitioned the matrix between columns one and two and still carried out the multiplication using the resulting submatrices. [A][C] = [D] = J
A11 A21
A12 C11 RJ A22 C21
= J
D11 D21
D12 R D22
where
C12 A C + A12C21 R = J 11 11 C22 A21C11 + A22C21
A11C12 + A12C22 R A21C12 + A22C22
[D11] = [A11][C11] + [A12][C21] and [D12] = [A11]5 C12 6 + [A12]5 C22 6 [D21] = [A21][C11] + [A22][C21] [D22] = [A21]5 C12 6 + [A22]5 C22 6
Example 2.2
Given the following partitioned matrices, calculate the product of [A][B] = [C] using the submatrices shown. 5 3 [A] = E 1 0 2 [A] = J
M02_MOAV4303_04_GE_C02.INDD 95
A11 A21
7 8 4 10 5
2 3 0 5 9
0 5 7 12 2
3 0 15 3 18
5 8 9 U and 1 10
2 8 5 [B] = F 4 3 1
10 7 2 8 12 5
0 5 4 V 13 0 7
A12 R A22
27/11/14 9:19 AM
96 Chapter 2 Matrix Algebra
where [A11] = J
5 3
1 [A21] = C 0 2
7 8
2 R 3
4 10 5
[A12] = J
0 5S 9
0 5
7 [A22] = C 12 2
3 0 15 3 18
5 R 8 9 1 S 10
and [B] = J
B11 B21
B12 R B22
where 2 [B11] = C 8 5 4 [B21] = C 3 1
10 7S 2 8 12 S 5
0 and 5 B12 6 = c 5 s 4 13 and 5 B22 6 = c 0 s 7
A A B B A B + A12B21 A11B12 + A12B22 C C [A][B] = [C] = J 11 12 R J 11 12 R = J 11 11 R = J 11 12 R A21 A22 B21 B22 A21B11 + A22B21 A21B12 + A22B22 C21 C22 where 10 7S 2 4 8 0 3 5 70 164 + J R C 3 12 S = J R 5 0 8 73 80 1 5 0 5 7 2 [C12] = [A11]5 B12 6 + [A12]5 B22 6 = J Rc 5 s 3 8 3 4 13 0 3 5 62 + J Rc 0 s = b r 5 0 8 43 7 1 4 0 2 10 [C21] = [A21][B11] + [A22][B21] = C 0 10 5 S C 8 7S 2 5 9 5 2 7 15 9 4 8 116 319 + C 12 3 1 S C 3 12 S = C 111 207 S 2 18 10 1 5 29 185 5 [C11] = [A11][B11] + [A12][B21] = J 3
M02_MOAV4303_04_GE_C02.INDD 96
7 8
2 2 RC 8 3 5
27/11/14 9:19 AM
Section 2.5 Transpose of a Matrix 97
1 [C22] = [A21]5 B12 6 + [A22]5 B22 6 = C 0 2
4 0 0 10 5 S c 5 s 5 9 4 7 15 9 13 174 + C 12 3 1 S c 0 s = c 179 s 2 18 10 7 105
The final result is given by
[A][B] = [C] = J
A11 A21
A12 B11 RJ A22 B21
B12 C R = J 11 B22 C21
70 73 C12 R = E 116 C22 111 29
164 80 319 207 185
62 43 174 U 179 105
As explained earlier, the column partitioning of matrix [B] may be done arbitrarily because the resulting submatrices still conform to the multiplication rule. It is left as an exercise for you (see Problem 3 at the end of this chapter) to show that we could have partitioned matrix [B] between columns one and two and used the resulting submatrices to compute [A][B] = [C]. 2.5 Transpose of a Matrix As you will see in the following chapters, the finite element formulation lends itself to situations wherein it is desirable to rearrange the rows of a matrix into the columns of another matrix. To demonstrate this idea, let us go back and consider step 4 in Example 1.1. In step 4 we assembled the elemental stiffness matrices to obtain the global stiffness matrix. You will recall that we constructed the stiffness matrix for each element with its position in the global stiffness matrix by inspection. Let’s recall the stiffness matrix for element (1), which is shown here again for the sake of continuity and convenience. [K](1) = J
[K](1G)
k1 k1
k1 k1 = E 0 0 0
k1 R and its position in the global matrix k1 k1 k1 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0U 0 0
Instead of putting together [K](1G) by inspection as we did, we could obtain [K](1G) using the following procedure:
M02_MOAV4303_04_GE_C02.INDD 97
[K](1G) = [A1]T[K](1)[A1] (2.9)
27/11/14 9:19 AM
98 Chapter 2 Matrix Algebra
where [A1] = J
1 0
0 1
0 0
0 0
0 R 0
and 1 0 [A1]T = E 0 0 0
0 1 0U 0 0
[A1]T, called the transpose of [A1], is obtained by taking the first and the second rows of [A1] and making them into the first and the second columns of the transpose matrix. It is easily verified that by carrying out the multiplication given by Eq. (2.9), we will arrive at the same result that was obtained by inspection.
[K](1G)
1 0 = E0 0 0
0 1 k 0U J 1 k1 0 0
k1 1 RJ k1 0
0 1
0 0
k1 k1 0 R = E 0 0 0 0
0 0
k1 k1 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0U 0 0
0 0 0 0 0
0 0 0U 0 0
Similarly, we could have performed the following operation to obtain [K](2G): [K](2G) = [A2]T[K](2)[A2] where [A2] = J
0 0
1 0
0 1
0 0
0 R 0
and 0 1 T [A2] = E 0 0 0
0 0 1U 0 0
and
[K](2G)
0 1 = E0 0 0
M02_MOAV4303_04_GE_C02.INDD 98
0 0 k 1U J 2 k2 0 0
k2 0 RJ k2 0
1 0
0 1
0 0
0 0 0 R = E0 0 0 0
0 k2 k2 0 0
0 k2 k2 0 0
27/11/14 9:19 AM
Section 2.5 Transpose of a Matrix 99
As you have seen from the previous examples, we can use a positioning matrix, such as [A] and its transpose, and methodically create the global stiffness matrix for finite element models. In general, to obtain the transpose of a matrix [B] of size m * n, the first row of the given matrix becomes the first column of the [B]T, the second row of [B] becomes the second column of [B]T, and so on, leading to the mth row of [B] becoming the mth column of the [B]T, resulting in a matrix with the size of n * m. Clearly, if you take the transpose of [B]T, you will end up with [B]. That is, ([B]T)T = [B] (2.10)
As you will see in succeeding chapters, in order to save space, we write the solution matrices, which are column matrices, as row matrices using the transpose of the solution—another use for transpose of a matrix. For example, we represent the displacement solution U1 U2 5 U 6 = e U3 u by [U ]T = [U1
#
U2
U3
#
Un]
Un
When performing matrix operations dealing with transpose of matrices, the following identities are true: ([A] + [B] + c + [N ])T = [A]T + [B]T + c + [N ]T (2.11) ([A][B]c[N ])T = [N ]T c[B]T[A]T (2.12)
In Eq. (2.12), note the change in the order of multiplication. This is a good place to define a symmetric matrix. A symmetric matrix is a square matrix whose elements are symmetrical with respect to its principal diagonal. An example of a symmetric matrix follows: 1 4 [A] = D 2 5
4 5 15 20
2 15 3 8
5 20 T 8 0
Note that for a symmetric matrix, element amn is equal to anm. That is, amn = anm for all values of n and m. Therefore, for a symmetric matrix, [A] = [A]T. Example 2.3 Given the following matrices: 0 [A] = C 8 9
M02_MOAV4303_04_GE_C02.INDD 99
5 3 2
0 4 7 S and [B] = C 7 9 1
6 2 3
2 3 S 4
27/11/14 9:19 AM
100 Chapter 2 Matrix Algebra
perform the following operations: a. [A]T = ? and [B]T = ? b. verify that ([A] + [B])T = [A]T + [B]T c. verify that ([A][B])T = [B]T[A]T a. [A]T = ? and [B]T = ? As explained earlier, the first, second, third, . . . , and the mth rows of a matrix become the first, second, third, . . . , and the mth columns of the transpose matrix respectively. 0 8 9 [A]T = C 5 3 2 S 0 7 9 Similarly, 4 7 1 T [B] = C 6 2 3 S 2 3 4 b. Verify that ([A] + [B])T = [A]T + [B]T. 4 [A] + [B] = C 15 10 0 [A]T + [B]T = C 5 0
2 4 T 10 S and ([A] + [B]) = C 11 5 2
11 5 1 8 3 7
9 4 2 S + C 6 9 2
7 2 3
1 4 3 S = C 11 4 2
15 5 10
10 1S 5
15 5 10
10 1S 5
Comparing results, it is clear that the given identity is true. c. Verify that ([A][B])T = [B]T[A]T. In Example 2.2 we computed the product of [A][B]: 0 [A][B] = C 8 9
5 3 2
0 4 7S C7 9 1
6 2 3
2 35 3 S = C 60 4 31
10 75 77
15 35 S 60
and using the results we get 35 ([A][B]) = C 10 15 Alternatively, T
60 75 35
31 77 S 60
4 7 1 0 8 9 35 60 31 [B] [A] = C 6 2 3 S C 5 3 2 S = C 10 75 77 S 2 3 4 0 7 9 15 35 60 Again, by comparing results we can see that the given identity is true. T
M02_MOAV4303_04_GE_C02.INDD 100
T
27/11/14 9:19 AM
Section 2.6 Determinant of a Matrix 101
2.6 Determinant of a Matrix Up to this point we have defined essential matrix terminology and discussed basic matrix operations. In this section we define what is meant by a determinant of a matrix. As you will see in the succeeding sections, determinant of a matrix is used in solving a set of simultaneous equations, obtaining the inverse of a matrix, and forming the characteristic equations for a dynamic problem (eigenvalue problem). Let us consider the solution to the following set of simultaneous equations: a11x1 + a12x2 = b1 (2.13a) a21x1 + a22x2 = b2 (2.13b)
We can represent Eqs. (2.13a) and (2.13b) in a matrix form by J
a11 a21
a12 x b R b 1 r = b 1 r (2.14) a22 x2 b2
or in a compact form by [A]5 X6 = 5 B6
To solve for the unknowns x1 and x2, we may first solve for x2 in terms of x1, using Eq. (2.13b), and then substitute that relationship into Eq. (2.13a). These steps are shown next. x2 =
b2  a21x1 a22
a11x1 + a12 ¢
1
b2  a21x1 ≤ = b1 a22
Solving for x1
x1 =
b1a22  a12b2 (2.15a) a11a22  a12a21
After we substitute for x1 in either Eq. (2.13a) or (2.13b), we get
x2 =
a11b2  b1a21 (2.15b) a11a22  a12a21
Referring to the solutions given by Eqs. (2.15a) and (2.15b), we see that the denominators in these equations represent the product of coefficients in the main diagonal minus the product of the coefficient in the other diagonal of the [A] matrix. The a11a22  a12a21 is the determinant of the 2 * 2 [A] matrix and is represented in one of following ways: Det[A] or det[A] or 2
a11 a21
a12 2 = a11a22  a12a21 (2.16) a22
Only the determinant of a square matrix is defined. Moreover, keep in mind that the determinant of the [A] matrix is a single number. That is, after we substitute for the
M02_MOAV4303_04_GE_C02.INDD 101
27/11/14 9:19 AM
102 Chapter 2 Matrix Algebra
values of a11, a22, a12, and a21 into a11a22  a12a21, we get a single number. In general, the determinant of a matrix is a single value. However, as you will see later, for dynamic problems the determinant of a matrix resulting from equations of motions is a polynomial expression. Cramer’s rule is a numerical technique that can be used to obtain solutions to a relatively small set of equations similar to the previous example. Using Cramer’s rule we can represent the solutions to the set of simultaneous equations given by Eqs. (2.13a) and (2.13b) with the following determinants:
x1 =
2 b1 b2
a12 2 a22
2 a11 a21
a12 2 a22
and x2 =
2 a11 a21
b1 2 b2
2 a11 a21
a12 2 a22
(2.17)
Let us now consider the determinant of a 3 * 3 matrix such as c11 [C] = C c21 c31
c12 c22 c32
c13 c23 S c33
which is computed in the following manner: c11 3 c21
c31
c12 c22 c32
c13 c c c + c12c23c31 + c13c21c32  c13c22c31 c23 3 = 11 22 33 (2.18)  c11c23c32  c12c21c33 c33
There is a simple procedure called direct expansion, which you can use to obtain the results given by Eq. (2.18). Direct expansion proceeds in the following manner. First, we repeat and place the first and the second columns of the matrix [C] next to the third column, as shown in Figure 2.1. Then we add the products of the diagonal elements lying on the solid arrows and subtract them from the products of the diagonal elements lying on the dashed arrows. This procedure, shown in Figure 2.1, results in the determinant value given by Eq. (2.18). The direct expansion procedure cannot be used to obtain higher order determinants. Instead, we resort to a method that first reduces the order of the determinant—to what is called a minor—and then evaluates the lower order determinants. To demonstrate this method, let’s consider the righthand side of Eq. (2.18) and factor out c11,c12, and c13 from it. This operation is shown below. c11c22c33 + c12c23c31 + c13c21c32  c13c22c31  c11c23c32  c12c21c33 = c11(c22c33  c23c32)  c12(c21c33  c23c31) + c13(c21c32  c22c31) As you can see, the expressions in the parentheses represent the determinants of reduced 2 * 2 matrices. Thus, we can express the determinant of the given 3 * 3 matrix
M02_MOAV4303_04_GE_C02.INDD 102
27/11/14 9:19 AM
Section 2.6 Determinant of a Matrix 103
a11
a12
c11
c12
c13
c11
c12
a21
a22
c21
c22
c23
c21
c22
c31
c32
c33
c31
c32
(a)
Figure 2.1 Direct expansion procedure for computing the determinant of (a) 2 * 2 matrix, and (b) 3 * 3 matrix.
(b)
in terms of the determinants of the reduced 2 * 2 matrices (minors) in the following manner: c11 3 c21 c31
c12 c22 c32
c13 c c23 3 = c11 2 22 c32 c33
c23 2 c  c12 2 21 c33 c31
c23 2 c + c13 2 21 c33 c31
c22 2 c32
A simple way to visualize this reduction of a thirdorder determinant into three secondorder minors is to draw a line through the first row, first column, second column, and third column. Note that the elements shown in the box—the common element contained in elimination rows and columns—are factors that get multiplied by the lower order minors. The plus and the minus signs before these factors are assigned based on the following procedure. We add the row and the column number of the factor, and if the sum is an even number, we assign a positive sign, and if the sum is an odd number we assign a negative sign to the factor. For example, when deleting the first row and the second column (1 + 2), a negative sign should then appear before the c12 factor. c11 c21 c31
c12 c22 c32
c13 c23 c33
c11 c21 c31
c12 c22 c32
c13 c23 c33
c11 c21 c31
c12 c22 c32
c13 c23 c33
It is important to note that alternatively to compute the determinant of [A] we could have eliminated the second or the third row—instead of the first row—to reduce the given determinant into three other secondorder minors. This point is demonstrated in Example 2.4. Our previous discussion demonstrates that the order of a determinant may be reduced into other lower order minors, and the lower order determinants may be used to evaluate the value of the higher order determinant. Here are two useful properties of determinants: (1) The determinant of a matrix [A] is equal to the determinant of its transpose [A]T. This property may be readily verified (see Example 2.4). (2) If you multiply the elements of a row or a column of a matrix by a scalar quantity, then its determinant gets multiplied by that quantity as well.
M02_MOAV4303_04_GE_C02.INDD 103
27/11/14 9:19 AM
104 Chapter 2 Matrix Algebra
Example 2.4 Given the following matrix: 1 [A] = C 8 6
5 3 2
0 7S 9
calculate a. determinant of [A] b. determinant of [A]T a. For this example, we use both the direct expansion and the minor methods to compute the determinant of [A]. As explained earlier, using the direct expansion method, we repeat and place the first and the second columns of the matrix next to the third column as shown, and compute the products of the elements along the solid arrows, then subtract them from the products of elements along the dashed arrow. 1 8 6 1 38 6
5 3 2 5 3 2
0 7 9
1 8 6
5 3 2
0 (1)(3)(9) + (5)(7)(6) + (0)(8)( 2)  (5)(8)(9) 73 =  (1)(7)( 2)  (0)(3)(6) = 109 9
Next, we use the minor to compute the determinant of [A]. For this example, we eliminate the elements in the first row and in first, second, and third columns, as shown. 1 8 6
5 3 2
0 7 9
1 8 6
5 3 2
0 7 9
1 8 6
5 3 2
0 7 9
1 38
5 0 3 72 8 72 8 3 2 3 73 = 12 5 2 + 02 2 9 6 9 6 2 6 2 9 = (1)[(3)(9) (7)( 2)](5)[(8)(9) (7)(6)] = 109 Alternatively, to compute the determinant of the [A] matrix, we can eliminate the elements in the second row, and first, second, and third column as shown: 1 8 6
M02_MOAV4303_04_GE_C02.INDD 104
5 3 2
0 7 9
1 8 6
5 3 2
0 7 9
1 8 6
5 3 2
0 7 9
27/11/14 9:19 AM
Section 2.6 Determinant of a Matrix 105
1 38 6
5 3 2
0 5 7 3 = 8 2 2 9
02 1 + 32 9 6
02 1  72 9 6
5 2 2
= (8)[(5)(9) (0)( 2)]+(3)[(1)(9) (0)(6)] (7)[(1)( 2) (5)(6)] = 109 b. As already mentioned, the determinant of [A]T is equal to the determinant of [A]. Therefore, there is no need to perform any additional calculations. However, as a means of verifying this identity, we will compute and compare determinant of [A]T to the determinant of [A]. Recall that [A]T is obtained by interchanging the first, second, and third rows of [A] into the first, second, and third columns of the 1 [A]T, leading to [A]T = C 5 0 1 5 0
8 3 7
6 2 9
1 35
8 3 7
6 3 2 3 = 1 2 7 9
0
1 5 0
8 3 7
8 3 7
6 2 9
6 2 S . Using minors, we get 9 1 5 0
2 2 5  82 9 0
8 3 7
6 2 9
2 2 5 + 62 9 0
32 7
= (1)[(3)(9) ( 2)(7)](8)[(5)(9) ( 2)(0)] + (6)[(5)(7) (3)(0)] = 109 When the determinant of a matrix is zero, the matrix is called a singular. A singular matrix results when the elements in two or more rows of a given matrix are identical. For example, consider the following matrix: 2 [A] = C 2 1
1 1 3
4 4S 5
whose rows one and two are identical. As shown below, the determinant of [A] is zero. 2 32 1
1 1 3
4 (2)(1)(5) + (1)(4)(1) + (4)(2)(3)  (1)(2)(5) 43 =  (2)(4)(3)  (4)(1)(1) = 0 5
Matrix singularity can also occur when the elements in two or more rows of a matrix are linearly dependent. For example, if we multiply the elements of the second row of matrix [A] by a scalar factor such as 7, then the resulting matrix, 2 [A] = C 14 1
M02_MOAV4303_04_GE_C02.INDD 105
1 7 3
4 28 S 5
27/11/14 9:19 AM
106 Chapter 2 Matrix Algebra
is singular because rows one and two are now linearly dependent. As shown below, the determinant of the new [A] matrix is zero. 2 3 14 1
1 7 3
4 (2)(7)(5) + (1)(28)(1) + (4)(14)(3)  (1)(14)(5) 28 3 =  (2)(28)(3)  (4)(7)(1) = 0 5
2.7 Solutions of Simultaneous Linear Equations As you saw in Chapter 1, the finite element formulation leads to a system of algebraic equations. Recall that for Example 1.1, the bar with a variable cross section supporting a load, the finite element approximation and the application of the boundary condition and the load resulted in a set of four linear equations: 1820 845 103 D 0 0
845 1560 715 0
0 715 1300 585
0 u2 0 0 u 0 Td 3t = d t 585 u4 0 585 u5 103
In the sections that follow we discuss two methods that you can use to obtain solutions to a set of linear equations. Gauss Elimination Method We begin our discussion by demonstrating the Gauss elimination method using an example. Consider the following three linear equations with three unknowns, x1, x2, and x3. 2x1 + x2 + x3 = 13 (2.19a) 3x1 + 2x2 + 4x3 = 32 (2.19b) 5x1  x2 + 3x3 = 17 (2.19c)
1. We begin by dividing the first equation, Eq. (2.19a), by 2, the coefficient of x1 term. This operation leads to
x1 +
1 1 13 x2 + x3 = (2.20) 2 2 2
2. We multiply Eq. (2.20) by 3, the coefficient of x1 in Eq. (2.19b).
M02_MOAV4303_04_GE_C02.INDD 106
3x1 +
3 3 39 x2 + x3 = (2.21) 2 2 2
27/11/14 9:19 AM
Section 2.7 Solutions of Simultaneous Linear Equations 107
We then subtract Eq. (2.21) from Eq. (2.19b). This step eliminates x1 from Eq. (2.19b). This operation leads to
3x1 + 2x2 + 4x3 = 32  ¢ 3x1 +
3 3 39 x2 + x3 = ≤ 2 2 2
1 5 25 x + x3 = (2.22) 2 2 2 2
3. Similarly, to eliminate x1 from Eq. (2.19c), we multiply Eq. (2.20) by 5, the coefficient of x1 in Eq. (2.19c) 5 5 65 5x1 + x2 + x3 = (2.23) 2 2 2 We then subtract the above equation from Eq. (2.19c), which eliminates x1 from Eq. (2.19c). This operation leads to
5x1  x2 + 3x3 = 17  ¢ 5x1 +
5 5 65 x + x3 = ≤ 2 2 2 2
7 1 31  x2 + x3 =  (2.24) 2 2 2
Let us summarize the results of the operations performed during steps 1 through 3. These operations eliminated the x1 from Eqs. (2.19b) and (2.19c).
1 1 x2 + x3 = 2 2 1 5 x + x3 = 2 2 2 7 1  x2 + x3 = 2 2
x1 +
13 (2.25a) 2 25 (2.25b) 2 31  (2.25c) 2
4. To eliminate x2 from Eq. (2.25c), first we divide Eq. (2.25b) by 1/2, the coefficient of x2.
x2 + 5x3 = 25 (2.26)
Then we multiply Eq. (2.26) by 7/2, the coefficient of x2 in Eq. (2.25c), and subtract that equation from Eq. (2.25c). These operations lead to
M02_MOAV4303_04_GE_C02.INDD 107
7 1 31  x2 + x3 = 2 2 2 7 35 175  ¢  x2 x = ≤ 2 2 3 2 18x3 = 72 (2.27)
27/11/14 9:19 AM
108 Chapter 2 Matrix Algebra
Dividing both sides of Eq. (2.27) by 18, we get x3 = 4 Summarizing the results of the previous steps, we have
1 1 13 x2 + x3 = (2.28) 2 2 2 x2 + 5x3 = 25 (2.29)
x3 = 4 (2.30)
x1 +
Note Eqs. (2.28) and (2.29) are the same as Eqs. (2.25a) and (2.26), which are renumbered for convenience. Now we can use back substitution to compute the values of x2 and x1. We substitute for x3 in Eq. (2.29) and solve for x2. x2 + 5(4) = 25
S
x2 = 5
Next, we substitute for x3 and x2 in Eq. (2.28) and solve for x1. x1 +
1 1 13 (5) + (4) = 2 2 2
S
x1 = 2
The Lower Triangular, Upper Triangular (LU) Decomposition Method When designing structures, it is often necessary to change the load and consequently the load matrix to determine its effect on the resulting displacements and stresses. Some heat transfer analysis also requires experimenting with the heat load in reaching the desirable temperature distribution within the medium. The Gauss elimination method requires full implementation of the coefficient matrix (the stiffness or the conductance matrix) and the righthand side matrix (load matrix) in order to solve for the unknown displacements (or temperatures). When using Gauss elimination, the entire process must be repeated each time a change in the load matrix is made. Whereas the Gauss elimination is not well suited for such situations, the LU method handles any changes in the load matrix much more efficiently. The LU method consists of two major parts: a decomposition part and a solution part. We explain the LU method using the following three equations a11x1 + a12x2 + a13x3 = b1 (2.31) a21x1 + a22x2 + a23x3 = b2 (2.32) a31x1 + a32x2 + a33x3 = b3 (2.33)
or in a matrix form,
M02_MOAV4303_04_GE_C02.INDD 108
a11 C a21 a31
a12 a22 a32
a13 x1 b1 a23 S c x2 s = c b2 s or [A] 5 x6 = 5 b6 (2.34) a33 x3 b3
27/11/14 9:19 AM
Section 2.7 Solutions of Simultaneous Linear Equations 109
Decomposition Part The main idea behind the 1 pose the coefficient matrix [A] into lower [L] = C l21 u11 u12 u13 l31 [U] = C 0 u22 u23 S matrices so that 0 0 u33 a11 C a21 a31
a13 1 a23 S = C l21 a33 l31
a12 a22 a32
0 1 l32
LU 0 1 l32
0 u11 0S C 0 1 0
method is first to decom0 0 S and upper triangular 1
u12 u22 0
u13 u23 S (2.35) u33
u12 u22 0
u13 u23 S u33
Carrying out the multiplication operation, we get
a11 C a21 a31
a12 a22 a32
u11 = C l21u11 l31u11
a13 1 a23 S = C l21 a33 l31
u12 l21u12 + u22 l31u12 + l32u22
0 1 l32
0 u11 0S C 0 1 0
u13 l21u13 + u23 S l31u13 + l32u23 + u33
(2.36)
Now let us compare the elements in the first row of [A] matrix in Eq. (2.36) to the elements in the first row of the [L][U] multiplication results. From this comparison we can see the following relationships: u11 = a11 and u12 = a12 and u13 = a13 Now, by comparing the elements in the first column of [A] matrix in Eq. (2.36) to the elements in the first column of the [L][U] product, we can obtain the values of l21 and l31:
l21u11 = a21
S
l21 =
a21 a21 = (2.37) u11 a11
l31u11 = a31
S
l31 =
a31 a31 = (2.38) u11 a11
Note the value of u11 was determined in the previous step. That is, u11 = a11. We can obtain the values of u22 and u23 by comparing the elements in the second rows of the matrices in Eq. (2.36).
l21u12 + u22 = a22
S
u22 = a22  l21u12 (2.39)
l21u13 + u23 = a23
S
u23 = a23  l21u13 (2.40)
When examining Eqs. (2.39) and (2.40) remember that the values of l21, u12, and u13 are known from previous steps. Now we compare the elements in the second columns of
M02_MOAV4303_04_GE_C02.INDD 109
27/11/14 9:19 AM
110 Chapter 2 Matrix Algebra
Eq. (2.36). This comparison leads to the value of l32. Note, we already know values of u12, l21, u22, and l31 from previous steps. S
l31u12 + l32u22 = a32
l32 =
a32  l31u12 (2.41) u22
Finally, the comparison of elements in the third rows leads to the value of u33. S
l31u13 + l32u23 + u33 = a33
u33 = a33  l31u13  l32u23 (2.42)
We used a simple 3 * 3 matrix to show how the LU decomposition is performed. We can now generalize the scheme for a square matrix of any size n in the following manner: Step 1. The values of the elements in the first row of the [U] matrix are obtained from
u1j = a1j
for j = 1 to n (2.43)
Step 2. The unknown values of the elements in the first column of the [L] matrix are obtained from
li1 =
ai1 u11
for i = 2 to n (2.44)
Step 3. The unknown values of the elements in the second row of the [U] matrix are computed from u2j = a2j  l21u1j
for j = 2 to n (2.45)
Step 4. The values of the elements in the second column of [L] matrix are calculated from
ai2  li1u12 u22
li2 =
for i = 3 to n (2.46)
Next, we determine the unknown values of the elements in the third row of the [U] matrix and the third column of [L]. By now you should see a clear pattern. We evaluate the values of the elements in a row first and then switch to a column. This procedure is repeated until all the unknown elements are computed. We can generalize the above steps in the following way. To obtain the values of the elements in the kth row of [U] matrix, we use ukj = akj  a lkpupj k1
p=1
for j = k to n (2.47)
We will then switch to the kth column of [L] and determine the unknown values in that column. aik  a lipupk k1
lik =
p=1
ukk
for i = k + 1 to n (2.48)
Solution Part So far you have seen how to decompose a square coefficient matrix [A] into lower and upper triangular [L] and [U] matrices. Next, we use the [L] and the
M02_MOAV4303_04_GE_C02.INDD 110
27/11/14 9:19 AM
Section 2.7 Solutions of Simultaneous Linear Equations 111
[U] matrices to solve a set of linear equations. Let’s turn our attention back to the three equations and three unknowns example and replace the coefficient matrix [A] with the [L] and [U] matrices: [A]5 x6 = 5 b6 (2.49)
[L][U]5 x6 = 5 b6 (2.50)
We now replace the product of [U ] 5 x6 by a column matrix 5 z 6 such that {z}
2
[U]5 x6 = 5 z 6 (2.51)
[L][U]5x6 = 5 b 6 S [L]5 z6 = 5 b 6
(2.52)
Because [L] is a lower triangular matrix, we can easily solve for the values of the elements in the 5 z6 matrix, and then use the known values of the 5 z6 matrix to solve for the unknowns in the 5 x6 from the relationship [U] 5 x6 = 5 z6 . These steps are demonstrated next. 1 C l21 l31
0 1 l32
0 z1 b1 0 S c z2 s = c b2 s (2.53) 1 z3 b3
From Eq. (2.53), it is clear that
z1 = b1
(2.54)
z2 = b2  l21z1
(2.55)
z3 = b3  l31z1  l32z2 (2.56) Now that the values of the elements in the 5 z6 matrix are known, we can solve for the unknown matrix 5 x6 using u11 C 0 0
u12 u22 0
u13 x1 z1 u23 S c x2 s = c z2 s (2.57) u33 x3 z3
z3 (2.58) u33 z2  u23x3 x2 = (2.59) u22 z1  u12x2  u13x3 x1 = (2.60) u11
x3 =
Here we used a simple three equations and three unknowns to demonstrate how best to proceed to obtain solutions; we can now generalize the scheme to obtain the solutions for a set of n equations and n unknown.
M02_MOAV4303_04_GE_C02.INDD 111
27/11/14 9:19 AM
112 Chapter 2 Matrix Algebra
z1 = b1 and zi = bi  a lijzj i1
zi  a uijxj
j=1
for i = 2, 3, 4, . . . , n (2.61)
n
xn =
zn and xi = unn
j=i+1
uii
for i = n  1, n  2, n  3, c, 3, 2, 1 (2.62)
Next, we apply the LU method to the set of equations that we used to demonstrate the Gauss elimination method. Example 2.5 Apply the LU decomposition method to the following three equations and three unknowns set of equations: 2x1 + x2 + x3 = 13 3x1 + 2x2 + 4x3 = 32 5x1  x2 + 3x3 = 17 2 [A] = C 3 5
1 2 1
1 13 4 S and 5 b6 = c 32 s 3 17
Note that for the given problem, n = 3. Decomposition Part Step 1. The values of the elements in the first row of the [U] matrix are obtained from u1j = a1j
for j = 1 to n
u11 = a11 = 2
u12 = a12 = 1
u13 = a13 = 1
Step 2. The unknown values of the elements in the first column of the [L] matrix are obtained from ai1 li1 = for i = 2 to n u11 a31 a21 3 5 l21 = = l31 = = u11 u11 2 2 Step 3. The unknown values of the elements in the second row of the [U] matrix are computed from u2j = a2j  l21u1j
for j = 2 to n
3 1 u22 = a22  l21u12 = 2  ¢ ≤(1) = 2 2 3 5 u23 = a23  l21u13 = 4  ¢ ≤(1) = 2 2
M02_MOAV4303_04_GE_C02.INDD 112
27/11/14 9:19 AM
Section 2.7 Solutions of Simultaneous Linear Equations 113
Step 4. The unknown values of the elements in the second column of the [L] matrix are determined from li2 =
ai2  li1u12 u22
for i = 3 to n
a32  l31u12 = u22
5 1 ¢ ≤(1) 2
= 7 1 2 Step 5. Compute the remaining unknown elements in the [U] and [L] matrices. l32 =
ukj = akj  a lkpupj k1
p=1
for j = k to n
5 5 u33 = a33  (l31u13 + l32u23) = 3  ¢¢ ≤(1) + ( 7) ¢ ≤ ≤ = 18 2 2 Because of the size of this problem (n = 3) and the fact that the elements along the main diagonal of the [L] matrix have values of 1, that is, l33 = 1, we do not need to proceed any further. Therefore, the application of the last step aik  a lipupk k1
lik =
p=1
for i = k + 1 to n
ukk
is omitted. We have now decomposed the coefficient matrix [A] into the following lower and upper triangular [L] and [U] matrices: 2 C3 5
1 2 1
1 1 3 4S = E 2 3 5 2
0
0
1
0
7
1
2 U D0 0
1 1 2 0
1 5 T 2 18
When performing this method by hand, here is a good place to check the decomposition results by premultiplying the [L] matrix by the [U] matrix to see if the [A] matrix is recovered. We now proceed with the solution phase of the LU method, Eq. (2.61). Solution Part z1 = b1 and zi = bi  a lijzj for i = 2 to n i1
j=1
z1 = 13
3 25 z2 = b2  l21z1 = 32  ¢ ≤(13) = 2 2
5 25 z3 = b3  (l31z1 + l32z2) = 17  ¢¢ ≤(13) + ( 7) ¢ ≤ ≤ = 72 2 2
M02_MOAV4303_04_GE_C02.INDD 113
27/11/14 9:19 AM
114 Chapter 2 Matrix Algebra
The solution is obtained from Eq. (2.62). zi  a uijxj n
xn =
zn and xi = unn
j=i+1
for i = n  1, n  2, n  3, c, 3, 2, 1
uii
Note for this problem n = 3, therefore i = 2, 1. x3 =
z3 72 = = 4 u33 18 25 5  ¢ ≤(4) 2 2
x2 =
z2  u23x3 = u22
x1 =
13  ((1)(5) + (1)(4)) z1  u12x2  u13x3 = = 2 u11 2
1 2
= 5
2.8 Inverse of a Matrix In the previous sections we discussed matrix addition, subtraction, and multiplication, but you may have noticed that we did not say anything about matrix division. That is because such an operation is not formally defined. Instead, we define an inverse of a matrix in such a way that when it is multiplied by the original matrix, the identity matrix is obtained. [A]1[A] = [A][A]1 = [I] (2.63)
In Eq. (2.63), [A]1 is called the inverse of [A]. Only a square and nonsingular matrix has an inverse. In Section 2.7 we explained the Gauss elimination and the LU methods that you can use to obtain solutions to a set of linear equations. Matrix inversion allows for yet another way of solving for the solutions of a set of linear equations. Once again, recall from our discussion in Chapter 1 that the finite element formulation of an engineering problem leads to a set of linear equations, and the solution of these equations render the nodal values. For instance, formulation of the problem in Example 1.1 led to the set of linear equations given by [K]5 u 6 = 5 F6 (2.64)
To obtain the nodal displacement values 5 u 6, we premultiply Eq. (2.64) by [K]1, which leads to [I]
2 1
[K] [K] 5 u 6 = [K]1 5 F6
(2.65)
[I ]5 u 6 = [K]1 5 F6 (2.66)
and noting that [I ]5 u 6 = 5 u 6 and simplifying,
M02_MOAV4303_04_GE_C02.INDD 114
5 u 6 = [K]1 5 F6 (2.67)
27/11/14 9:19 AM
Section 2.8 Inverse of a Matrix 115
From the matrix relationship given by Eq. (2.67), you can see that the nodal solutions can be easily obtained, provided the value of [K]1 is known. This example shows the important role of the inverse of a matrix in obtaining the solution to a set of linear equations. Now that you see why the inverse of a matrix is important, the next question is, How do we compute the inverse of a square and nonsingular matrix? There are a number of established methods that we can use to determine the inverse of a matrix. Here we discuss a procedure based on the LU decomposition method. Let us refer back to the relationship given by Eq. (2.63), and decompose matrix [A] into lower and upper triangular [L] and [U] matrices. [A]
2
[L][U] [A]1 = [I]
Next, we represent the product of [U][A]
1
(2.68)
by another matrix, say matrix [Y]:
[U][A]1 = [Y] (2.69)
and substitute for [U ][A]1 in terms of [Y] in Eq. (2.68), which leads to
[L][Y] = [I ] (2.70) We then use the relationships given by Eq. (2.70) to solve for the unknown values of elements in matrix [Y], and then use Eq. (2.69) to solve for the values of the elements in matrix [A]1. These steps are demonstrated using Example 2.6. Example 2.6 2 Given [A] = C 3 5
1 2 1
1 4 S , compute [A]1. 3
Step 1. Decompose the given matrix into lower and upper triangular matrices. In Example 2.5 we showed the procedure for decomposing the [A] matrix into lower and upper triangular [L] and [U] matrices. 1 3 [L] = E 2 5 2
0
0
1
0
7
1
2 U
and [U ] = D 0 0
1 1 2 0
1 5 T 2 18
Step 2. Use Eq. (2.70) to determine the unknown values of the elements in the [Y] matrix. [L]
6
1 3 E2 5 2
M02_MOAV4303_04_GE_C02.INDD 115
0 1
7
[Y] 0 6 y11 y12 y13 1 0 U C y21 y22 y23 S = C 0 y31 y32 y33 0 1
0 1 0
0 0S 1
27/11/14 9:19 AM
116 Chapter 2 Matrix Algebra
First, let us consider the multiplication results pertaining to the first column of the [Y] matrix, as shown. 1 3 E2 5 2
0
0
1
0
7
1
y11 1 U c y21 s = c 0 s y31 0
The solution of this system of equations leads to y21 = 
y11 = 1
3 2
y31 = 13
Next, consider the multiplication results pertaining to the second column of [Y]: 1 3 E2 5 2
0
0
1
0
7
1
y12 0 U c y22 s = c 1 s y32 0
The solution of this system of equation yields y12 = 0
y22 = 1
y32 = 7
Similarly, solve for the unknown values of the elements in the remaining column of the [Y] matrix: 1 3 E2 5 2
0
0
1
0
7
1
y13 = 0
y13 0 U c y23 s = c 0 s y33 1
y23 = 0
y33 = 1
Now that the values of elements of the [Y ] are known, we can proceed with calculation of the values of the elements comprising the [A]1, denoted by x11, x12, . . . , and so on, as shown. Using the relationship given by Eq. (2.69), we have [U][A]1 = [Y] 2 D0 0
M02_MOAV4303_04_GE_C02.INDD 116
1 1 2 0
1 x11 5 T C x21 2 x31 18
x12 x22 x32
1 x13 3 x23 S = D 2 x33 13
0
0
1
0T
7
1
27/11/14 9:19 AM
Section 2.8 Inverse of a Matrix 117
Again, we consider multiplication results pertaining to one column at a time. Considering the first column of the [x] matrix, 2
1 1 2 0
D0 0 x31 = 
1 1 x11 5 3 T c x21 s = d  t 2 2 x31 18 13
13 18
x21 =
11 18
x11 =
10 18
Multiplication results for the second column render 2 D0 0 x32 =
1 1 2 0
7 18
1 x12 0 5 T c x22 s = c 1 s 2 x32 7 18 x22 =
1 18
x12 = 
4 18
The multiplication results of the third column yield 2 D0 0 x33 =
1 1 2 0
1 18
1 x13 0 5 T c x23 s = c 0 s 2 x33 1 18 x23 = 
5 18
x13 =
2 18
Therefore, the inverse of the [A] matrix is [A]1 =
10 1 C 11 18 13
4 1 7
2 5 S 1
We can check the result of our calculations by verifying that [A][A]1 = [I ]. 2 1 C3 18 5
1 2 1
1 10 4 S C 11 3 13
4 1 7
2 1 5 S = C 0 1 0
0 1 0
0 0 S Q.E.D. 1
Finally, it is worth noting that the inversion of a diagonal matrix is computed simply by inversing its elements. That is, the inverse of a diagonal matrix is also a diagonal matrix
M02_MOAV4303_04_GE_C02.INDD 117
27/11/14 9:19 AM
118 Chapter 2 Matrix Algebra
with its elements being the inverse of the elements of the original matrix. For example, the inverse of the 4 * 4 diagonal matrix 1 a1 0 a2 0 0
a1 0 [A] = D 0 0
0 0 a3 0
0
0
0
0 0 0 T is [A]1 = H 0 0 a4
1 a2
0
0
0
1 a3
0
0
0
0
1 a4
X
This property of a diagonal matrix should be obvious because [A]1[A] = [I ]. 1 a1
0
0
0
1 a2
0
0
0
1 a3
0
0
0
[A]1[A] = H
0 a1 0 XD 0 0 0 1 a4 0
0 a2 0 0
0 0 a3 0
0 1 0 0 T = D 0 0 a4 0
0 1 0 0
0 0 1 0
0 0 T 0 1
2.9 Eigenvalues and Eigenvectors Up to this point we have discussed some of the methods that you can use to solve a set of linear equations of the form [A]5 x6 = 5 b6 (2.71)
For the set of linear equations that we have considered so far, the values of the elements of the 5 b6 matrix were typically nonzero. This type of system of linear equations is commonly referred to as nonhomogenous. For a nonhomogenous system, unique solutions exist as long as the determinant of the coefficient matrix [A] is nonzero. We now discuss the type of problems that render a set of linear equations of the form [A]5 X6  l 5 X6 = 0 (2.72)
This type of problem, called an eigenvalue problem, occurs in analysis of buckling problems, vibration of elastic structures, and electrical systems. In general, this class of problems has nonunique solutions. That is, we can establish relationships among the unknowns, and many values can satisfy these relationships. It is common practice to write Eq. (2.72) as [[A]  l[I ]]5 X6 = 0 (2.73)
where [I] is the identity matrix having the same dimension as the [A] matrix. In Eq. (2.73), the unknown matrix 5 X6 is called the eigenvector. We demonstrate how to obtain the eigenvectors using the following vibration example.
M02_MOAV4303_04_GE_C02.INDD 118
27/11/14 9:19 AM
Section 2.9 Eigenvalues and Eigenvectors 119
Example 2.7 Consider the two degrees of freedom system shown in Figure 2.2. We are interested in determining the natural frequencies of the system shown. We will discuss in detail the formulation and analysis of multiple degrees of freedom systems in Chapter 11. For the sake of presentation continuity, the derivation of the set of linear equations are shown here as well. Using the freebody diagrams shown, the equations of motion $
m1x1 + 2kx1  kx2 = 0 (2.74)
m2x2  kx1 + 2kx2 = 0 (2.75)
$
or in a matrix form, m J 1 0
$
0 x 2k R b 1$ r + J m2 x2 k
k x 0 R b 1r = b r 2k x2 0
Note that Eqs. (2.74) and (2.75) are secondorder homogenous differential equations. Also note that these equations are coupled, because both x1 and x2 appear in each equation. This type of system is called elastically coupled and may be represented in the general matrix form by $ [M]5 x 6 + [K]5 x6 = 0 (2.76) where [M] and [K] are the mass and the stiffness matrices respectively. We can simplify Eqs. (2.74) and (2.75) by dividing both sides of each equation by the values of the respective masses: 2k k $ x x = 0 (2.77) x1 + m1 1 m1 2 $
x2 
k 2k x + x = 0 (2.78) m2 1 m2 2
Assuming x2 7 x1 k
kx1
m1
m1 x1 k
k(x2  x1) k(x2  x1)
m2
m2 x2 k
..
kx1 + k(x2  x1) = m1 x1
..
k(x2  x1)  kx2 = m2 x2
kx2
Figure 2.2 A schematic diagram of an elastic system with two degrees of freedom.
M02_MOAV4303_04_GE_C02.INDD 119
27/11/14 9:19 AM
120 Chapter 2 Matrix Algebra
Using matrix notation, we premultiply the matrix form of the equations of motion by the inverse of the mass matrix [M]1, which leads to $
5 x 6 + [M]1[K] 5 x6 = 0
As a next step, we assume a harmonic solution of the form x1(t) = X1 sin(vt + f) and x2(t) = X2 sin(vt + f)—or in matrix form,5 x6 = 5 X 6 sin(vt + f)—and substitute the assumed solutions into the differential equations of motion, Eqs. (2.77) and (2.78), to create a set of linear algebraic equations. This step leads to v2X1sin(vt + f) +
2k k X sin(vt + f) X sin(vt + f) = 0 m1 1 m1 2
v2X2sin(vt + f) 
k 2k X sin(vt + f) + X sin(vt + f) = 0 m2 1 m2 2
After simplifying the sin(vt + f) terms, 2k m X 1 v2 b 1 r + D X2 k m2
k m1 X 0 T b 1 r = b r (2.79) 2k X2 0 m2

or in a general matrix form, v2 5 X 6 + [M]1[K] 5 X 6 = 0 (2.80)
Note that 5 x6 = b the 5 X 6 = b
x1(t) r represents the position of each mass as the function of time, x2(t)
X1 r matrix denotes the amplitudes of each vibrating mass, and f is the X2
phase angle. Equation (2.79) may be written as
1 v2 J 0
1 m1 0 X1 Rb r + D 1 X2 0
0 1 m2
TJ
2k k
k X R b 1 r = 0 (2.81) 2k X2
or by
2k m1 DD k m2
k m1 1 T  v2 J 2k 0 m2

0 X 0 R T b 1 r = b r (2.82) 1 X2 0
Comparing Eq. (2.82) to Eq. (2.73), [[A]  l[I]]5 X 6 = 0 we note that v2 = l. Simplifying Eq. (2.82) further, we have
M02_MOAV4303_04_GE_C02.INDD 120
27/11/14 9:19 AM
Section 2.9 Eigenvalues and Eigenvectors 121
v2 +
D
k m2
2k m1

k m1
2k v2 + m2
Tb
X1 r = 0 (2.83) X2
Problems with governing equations of the type (2.73) or (2.83) have nontrivial solutions only when the determinant of the coefficient matrix is zero. Let’s assign some numerical values to the above example problem and proceed with the solution. Let m1 = m2 = 0.1 kg and k = 100 N/m. Forming the determinant of the coefficient matrix and setting it equal to zero, we get 2
2 v + 2000 1000
1000 2 = 0 (2.84) v2 + 2000
( v2 + 2000)(  v2 + 2000)  ( 1000)( 1000) = 0 (2.85)
Simplifying Eq. (2.85), we have v4 4000v2 + 3,000,000 = 0 (2.86)
Equation (2.86) is called the characteristic equation, and its roots are the natural frequencies of the system. v21 = l1 = 1000 (rad/s)2 and v1 = 31.62 rad/s v22 = l2 = 3000 (rad/s)2 and v2 = 54.77 rad/s Once the v2 values are known, they can be substituted back into Eq. (2.83) to solve for the relationship between X1 and X2. The relationship between the amplitudes of mass oscillating at natural frequencies is called natural modes. We can use either relationship (rows) in Eq. (2.83). ( v2 + 2000)X1  1000X2 = 0 ( 1000 + 2000)X1  1000X2 = 0
and substituting for v21 = 1000 S
Or using the second row,
X2 = 1 X1
1000X1 + ( v2 + 2000)X2 = 0 and substituting for v21 = 1000 1000X1 + ( 1000 + 2000)X2 = 0
S
X2 = 1 X1
As expected, the results are identical. The second mode is obtained in a similar manner by substituting for v22 = 3000 in Eq. (2.83). ( v2 + 2000)X1  1000X2 = 0 and substituting for v21 = 3000 ( 3000 + 2000)X1  1000X2 = 0
S
X2 = 1 X1
It is important to note again that the solution of the eigenvalue problems leads to establishing a relationship among the unknowns, not specific values.
M02_MOAV4303_04_GE_C02.INDD 121
27/11/14 9:19 AM
122 Chapter 2 Matrix Algebra
2.10 Using Matlab to Manipulate Matrices MATLAB is mathematical software available in most university computational labs today. MATLAB is a very powerful tool for manipulating matrices; in fact, it was originally designed for that purpose. There are many good textbooks that discuss the capabilities of MATLAB to solve a full range of problems. Here, we introduce only some basic ideas so that you can perform essential matrix operations. For more detail see Appendix F. Once in a MATLAB environment, you can assign values to a variable or define elements of a matrix. For example, to assign a value 5 to the variable x, you simply type x = 5 1 or to define the element of the matrix [A] = C 8 6
5 3 2
0 7 S , you type 9
A = [1 5 0; 8 3 7; 6 2 9] Note that the elements of the matrix are enclosed in brackets and are separated by blank space, and the elements of each row are separated by a semicolon. The basic scalar operations are shown in Table 2.1. MATLAB offers many tools for matrix operations and manipulations. Table 2.2 shows examples of these capabilities. Next, we demonstrate a few MATLAB commands with the aid of some examples. Table 2.1 MATLAB’s basic scalar operators Operation
Symbol
Example: x = 5 and y = 3
Result 8
Addition
+
x + y
Subtraction

x  y
2
Multiplication
*
x*y
15
Division
/
(x + y)>2
4
Raised to a power
¿
x¿2
25
Table 2.2 Examples of MATLAB’s matrix operations Operation
Symbols or Commands
Example: A and B are matrices that you have defined
Addition
+
A + B
Subtraction

A  B
Multiplication
*
A*B
Transpose
matrix name′
A′
Inverse
inv(matrix name)
inv(A)
Determinant
det(matrix name)
det(A)
Eigenvalues
eig(matrix name)
eig(A)
\
see Example 2.8
Matrix left division (uses Gauss elimination to solve a set of linear equations)
M02_MOAV4303_04_GE_C02.INDD 122
27/11/14 9:19 AM
Section 2.10 Using Matlab to Manipulate Matrices 123
Example 2.1 Revisited Given the following matrices: 0 [A] = C 8 9
5 3 2
0 4 7 S , [B] = C 7 9 1
6 2 3
2 1 3 S , and [C] = c 2 s 4 5
using MATLAB, perform the following operations: a. [A] + [B] = ?, b. [A]  [B] = ?, c. 3[A] = ?, d. [A][B] = ?, e. [A]5 C 6 = ? f. [A]2 = ?
Also compute [A]T and the determinant of [A]. A MATLAB session is shown in Figure 2.3. When studying MATLAB examples, note that the response given by MATLAB is shown in boldface. Information that the user must type is shown in regular print. Hit the Enter key after you finish typing data. The MATLAB’s prompt for input is W .
W A = [0 5 0;8 3 7;9 @2 9] A = 0 8 9
5 3 @2
0 7 9
W B = [4 6 @2;7 2 3;1 3 @ 4] B = 4 7 1
6 2 3
@2 3 @4
W C = [@1;2;5] C = @1 2 5 W A+B
ans = 4 11 15 5 10 1
@2 10 5
Figure 2.3 Example of a MATLAB session.
M02_MOAV4303_04_GE_C02.INDD 123
27/11/14 9:19 AM
124 Chapter 2 Matrix Algebra
W [email protected] ans = @4 @1 °2 1 1 °4 8 5 13 W3*A ans = 0 15 24 °9 27 °6
°0 21 27
W A*B ans = 35 10 15 60 75 @35 31 77 @60 W A*C ans = 10 33 32 W A¿2 ans = 40 15 87 35 65 21
35 84 67
W A′ ans = 0 8 9 5 3 @2 0 7 9 W det(A) ans =  45 W
Figure 2.3 (Continued)
M02_MOAV4303_04_GE_C02.INDD 124
27/11/14 9:19 AM
Section 2.10 Using Matlab to Manipulate Matrices 125
EXAMPLE 2.8 Solve the following set of equations using the Gauss elimination and by inverting the [A] matrix and multiplying it by the 5 b 6 matrix. 2x1 + x2 + x3 = 13
3x1 + 2x2 + 4x3 = 32 5x1  x2 + 3x3 = 17 2 [A] = C 3 5
1 2 1
1 13 4 S and 5 b 6 = c 32 s 3 17
We first use the MATLAB matrix left division operator \ to solve this problem. The \ operator solves the problem using Gauss elimination. We then solve the problem using the inv command.
WA =[2 1 1;3 2 4;5 @1 3]
A = 2 1 1 3 2 4 5 @1 3 W b =[13;32;17] b = 13 32 17 W x=A\b x = 2.0000 5.0000 4.0000 W x = inv(A)*b x = 2.0000 5.0000 4.0000
M02_MOAV4303_04_GE_C02.INDD 125
27/11/14 9:19 AM
126 Chapter 2 Matrix Algebra
2.11 Using Excel to Manipulate Matrices Using the following two examples (Example 2.1 (revisited) and Example 2.8 (revisited)), we will show how to use Excel to perform matrix operations. Example 2.1 Revisited 0 Given matrices: [A] = C 8 9
5 3 2
0 4 7 S , [B] = C 7 9 1
6 2 3
1 2 3 S , and {C} = • 2 ¶ , Using 5 4
Excel perform the following operations. a. [A] + [B] = ? b. [A]  [B] = ? c. [A][B] = ? d. [A]{C} = ? 1. In the cells shown in Figure 2.4 type the appropriate characters and values. Use the Format Cells and Font option to create the bold variables as shown. Note that you do not have to create the characters ([A] = , [B] = , and so on) and format them to carry out the matrix operations. This is done merely as a visual aide for presentation purposes. 2. In cell A10 type [A] + [B] = , and then using the left mouse button pick cells B9 through D11, as shown in Figure 2.5. 3. Next, in the formula bar type = B3:D5 + G3:I5 and while holding down the Ctrl and the Shift keys press the Enter key. Note that you could also pick the range B3:D5 or G3:I5, instead of typing them. This sequence of operations will create the result shown in Figure 2.6. You can follow similar procedure as outlined in step 2 to perform [A] − [B], except in the formula bar type = B3:D5 − G3:I5.
Figure 2.4 Example 2.1 Revisited with Excel (step 1).
M02_MOAV4303_04_GE_C02.INDD 126
27/11/14 9:19 AM
Section 2.11 Using Excel to Manipulate Matrices 127
Figure 2.5 Example 2.1 Revisited with Excel (step 2).
Figure 2.6 Example 2.1 Revisited with Excel (step 3).
M02_MOAV4303_04_GE_C02.INDD 127
27/11/14 9:19 AM
128 Chapter 2 Matrix Algebra
Figure 2.7 Example 2.1 Revisited with Excel (step 4).
4. To carry out the matrix multiplication, first type [A][B] = in cell A18, as shown in Figure 2.7. Then pick cells B17 through D19. 5. In the formula bar type = MMULT(B3:D5,G3:I5), and while holding down the Ctrl and the Shift keys press the Enter key. Similarly you can perform the matrix operation [A] {C}. First you pick cells B21 through B23 and in the formula bar type = MMULT(B3:D5, L3:L5), and while holding down the Ctrl and the Shift keys press the Enter key. This sequence of operations will create the result shown in Figure 2.8.
M02_MOAV4303_04_GE_C02.INDD 128
27/11/14 9:19 AM
Section 2.11 Using Excel to Manipulate Matrices 129
Figure 2.8 Example 2.1 Revisited with Excel (step 5).
Example 2.8 Revisited Consider the following 3 linear equations with three unknowns, x1, x2, and x3. Our intent here is to show you how to use Excel to solve a set of linear equations. 2x1 + x2 + x3 = 13 3x1 + 2x2 + 4x3 = 32 5x1  x2 + 3x3 = 17 1. In the cells shown in Figure 2.9 type the appropriate characters and values. Use the Format Cells and Font option of create the bold and subscript variables. 2. In cell A10 type [A]1 = and then using the left mouse button pick cells B9 through D11 as shown in Figure 2.10. 3. Next, in the formula bar type = MINVERSE(B3:D5) and while holding down the Ctrl and the Shift keys press the Enter key. This sequence of operations will create the result shown in Figure 2.11. The inverse of matrix [A] is now computed.
M02_MOAV4303_04_GE_C02.INDD 129
27/11/14 9:19 AM
130 Chapter 2 Matrix Algebra
Figure 2.9 Example 2.8 Revisited with Excel (step 1).
Figure 2.10 Example 2.8 Revisited with Excel (step 2).
M02_MOAV4303_04_GE_C02.INDD 130
27/11/14 9:19 AM
Section 2.11 Using Excel to Manipulate Matrices 131
Figure 2.11 Example 2.8 Revisited with Excel (step 3).
Figure 2.12 Example 2.8 Revisited with Excel (step 4).
M02_MOAV4303_04_GE_C02.INDD 131
27/11/14 9:19 AM
132 Chapter 2 Matrix Algebra
Figure 2.13 Example 2.8 Revisited with Excel (step 5).
4. Type the information shown in cells A14, B13 through B15, C14, D14, and E14 as shown in Figure 2.12. 5. Then pick cells F13 through F15 and in the formula bar type = MMULT (B9:D11,K3:K5), and while holding down the Ctrl and the Shift keys press the Enter key. This sequence of operations will create the result shown in Figure 2.13. The values of x1, x2, and x3 are now calculated. Example 2.9 Using Example 1.1, we will show how to use Excel to set up and solve a finite element problem. Pay close attention to the way cells and a range of cells are named and used in the formulas. When formatting, analyzing, plotting data, or using formulas it is often convenient to select a number of cells simultaneously. The cells that are selected simultaneously are called a range. To define a range, begin with the first cell that you want included in the range and then drag the mouse (while pressing down the left button) to the last cell that should be included in the range. Note that in spreadsheet language, a range is defined by the cell address of the topleft selected cell in the range followed by a colon, :, and ends with the address of the bottomright cell in the range. For example, to select cells A3 through B10, we first select A3 and then drag the mouse diagonally to B10. In spreadsheet language, this range is specified in the following manner—A3:B10. There are situations where you may want to select a number of cells that are not side by side. In such cases, you must first select the contiguous cells, and then while holding (pressing) the Ctrl key select the other noncontiguous cells by dragging the mouse button.
M02_MOAV4303_04_GE_C02.INDD 132
27/11/14 9:20 AM
Section 2.11 Using Excel to Manipulate Matrices 133
Excel allows the user to assign a name to a cell or a range (selected cells). To name a cell or a range, first select the cell or the range as just described, and then click on the Name box in the Formula bar and type in the name you want to assign to the range. You can use upper or lowercase letters along with numbers, but no spaces are allowed between the characters or the numbers. Let us now turn our attention to Example 1.1. 1. In cell A1 type Example 1.1, and in cells A3 and A4 type E = and L = as shown. After inputting value of E in cell B3, select B3 and in the “Name Box” type E and press the Return key. Similarly, after inputting value of L in cell B4, select B4 and in the “Name Box” type L and press the Return key.
2. Create the tables shown, excluding the values of the area. In cell E7, type =0.25−(0.0125 , 2)*(I7+I6) Copy the formula in cells E8 through E10. Next, name the values in cells E7:E10 as Area1, Area2, Area3, and Area4, respectively.
M02_MOAV4303_04_GE_C02.INDD 133
27/11/14 9:20 AM
134 Chapter 2 Matrix Algebra
3. Type in the information in rows 12 through 16 as shown. Select cells G15:H16 and type = (Area1*E>L)*D15:E16 and while holding the Ctrl and Shift keys, press the Return key. Name the values in G15:H16 as Kelement1.
4. In a similar way create Kelement2, Kelement3, and Kelement4 as shown. You may copy [K1], rows 15 and 16 into rows 18 and 19; 21 and 22; and 24 and 25 and then modify them accordingly.
M02_MOAV4303_04_GE_C02.INDD 134
27/11/14 9:20 AM
Section 2.11 Using Excel to Manipulate Matrices 135
5. Next, create [A1], [A2], [A3], and [A4] matrices and name them Aelement1, Aelement2, Aelment3, and Aelement4, as shown. See Section 2.5, Equation 2.9. First create [A1] and then to create [A2], [A3], and [A4], copy [A1], rows 27 through 29 into rows 31 through 33; 35 through 37; and 39 through 41 and modify them accordingly. The nodal displacement U1, U2, U3, U4, U5, and Ui and Uj are shown alongside the [A1], [A2], [A3], and [A4] matrices to aid us observe the contribution of node to its neighboring elements.
M02_MOAV4303_04_GE_C02.INDD 135
27/11/14 9:20 AM
136 Chapter 2 Matrix Algebra
6. Next, create the stiffness matrix for each element (with their proper positions in the global matrix) and name them K1G, K2G, K3G, and K4G. See Equation 2.9. For example, to create [K]1G, select B43:F47 and type = MMULT(TRANSPOSE (Aelement1),MMULT(Kelement1,Aelement1)) and while holding down the Ctrl and Shift keys, press the Return key. In a similar way, create [K]2G, [K]3G, and [K]4G as shown.
7. The final global matrix is created next. Select the range B67:F71 and type =K1G+K2G+K3G+K4G and while holding down the Ctrl and Shift keys, press the Return key. Name the range B67:F71, KG.
M02_MOAV4303_04_GE_C02.INDD 136
27/11/14 9:20 AM
Section 2.11 Using Excel to Manipulate Matrices 137
8. Create the load matrix as shown and name it FG.
9. Apply the boundary conditions. Copy the appropriate portion of the KG matrix and paste it in the range C79:F82 as values only. Name the range KwithappliedBC. Similarly, create the corresponding load matrix in the range C84:C87 and name it FwithappliedBC.
M02_MOAV4303_04_GE_C02.INDD 137
27/11/14 9:20 AM
138 Chapter 2 Matrix Algebra
10. Select the range C89:C92 and type =MMULT(MINVERSE(KwithappliedBC), FwithappliedBC) and while holding down the Ctrl and Shift keys press the Return key.
11. As shown, select cells C94:C98, and copy the values of U partial and add the U1 = 0. Name this matrix, UG.
M02_MOAV4303_04_GE_C02.INDD 138
27/11/14 9:20 AM
Section 2.11 Using Excel to Manipulate Matrices 139
12. Calculate the reaction forces. Select the range C100:C104 and type = (MMULT(KG,UG)FG) and while pressing the Ctrl and Shift keys press the Return key.
The complete Excel sheet is shown next.
M02_MOAV4303_04_GE_C02.INDD 139
27/11/14 9:20 AM
140 Chapter 2 Matrix Algebra
Summary At this point you should 1. know the essential matrix terminology and the basic matrix operations. 2. know how to construct the transpose of a matrix.
M02_MOAV4303_04_GE_C02.INDD 140
27/11/14 9:20 AM
Chapter 2 Problems 141
3. know how to evaluate the determinant of a matrix. 4. know how to use Cramer’s rule, Gauss elimination, and the LU decomposition methods to solve a set of linear equations. 5. know how to compute the inverse of a matrix. 6. be familiar with the solution of an eigenvalue problem. References Gere, James M., and Weaver, William, Matrix Algebra for Engineers, New York, Van Nostrand, 1965. James, M. L., Smith, G. M., and Wolford, J. C., Applied Numerical Methods for Digital Computation, 4th ed., New York, Harper Collins, 1993. MATLAB Manual, Learning MATLAB 6, Release 12. Nakamura, Shoichiro, Applied Numerical Methods with Software, Upper Saddle River, NJ, Prentice Hall, 1991.
Problems 1. Identify the size and the type of the given matrices. Denote whether the matrix is a square, column, diagonal, row, unit (identity), triangular, banded, or symmetric. 1 a. C 3 0
1 e. C 0 0
3 7 8
0 1 0
0 8S 3
0 0S 1
p p2 b. d 3 t p p4 2 4 f. E 0 0 0
c. J
2 0
1 0 3 0 0
0 5 1 8 0
0 0 3 2 8
p1 0 h. D 0 0
0 p2 0 0
0 0 p3 0
0 R 4
0 0 0U 4 7
d. [1
p
p2
1 0 g. D 0 0
4 1 0 0
4 2 1 0
p3]
4 2 T 3 1
0 0 T 0 p4
2. Given matrices 4 [A] = C 7 1
2 0 5
1 1 7 S , [B] = C 5 3 4
2 3 5
1 1 3 S , and {C } = c 2 s 7 4
perform the following operations: a. [A] + [B] = ? b. [A]  [B] = ? c. 3[A] = ? d. [A][B] = ?
M02_MOAV4303_04_GE_C02.INDD 141
27/11/14 9:20 AM
142 Chapter 2 Matrix Algebra e. [A]{C } = ? f. [A]2 = ? g. Show that [I][A] = [A][I] = [A] 3. Given the following partitioned matrices, calculate the product of [A][B] = [C ] using the submatrices given. 5 3 [A] = E 1 0 2
7 8 4 10 5
2 3 0 5 9
0 5 7 12 2
3 0 15 3 18
2 5 8 8 5 9 U and [B] = F 4 1 3  10 1
10 7 2 8 12 5
0 5 4 V 13 0 7
4. Given the matrices 1 [A] = C 3 0
5 9 8
0 0 6 S and [B] = C 4 3 1
6 7 2
1 5 S 2
10 20 4
0 0S 18
perform the following operations: a. [A]T = ? and [B]T = ? b. verify that ([A] + [B])T = [A]T + [B]T c. verify that ([A][B])T = [B]T[A]T
5. Given the following matrices 2 [A] = C 16 12
10 6 4
0 2 14 S , [B] = C 4 18 12
calculate a. determinant of [A] and [B] by direct expansion and by minor lower order determinants methods. b. determinant of [A]T c. determinant of 5[A] Which matrix is singular?
6. Given the following matrix: 0 [A] = C 3 3
5 9 8
0 6S 3
calculate determinant of [A] and determinant of [A]T.
7. Solve the following set of equations using the Gauss elimination method, and compare your answer to the results of Example 1.4. 10875000 C 1812500 0
M02_MOAV4303_04_GE_C02.INDD 142
 1812500 6343750  4531250
0 u2 0 4531250 S c u3 s = c 0 s 4531250 u4 800
27/11/14 9:20 AM
Chapter 2 Problems 143
8. Decompose the coefficient matrix in problem 6 into lower and upper triangular matrices. 9. Solve the following set of equations using the LU method, and compare your answer to the results of Example 1.4. 10875000 C  1812500 0
 1812500 6343750  4531250
0 u2 0  4531250 S c u3 s = c 0 s 4531250 u4 800
10. Solve the following set of equations by finding the inverse of the coefficient matrix first. Compare your answer to the results of Example 1.4. 10875000 C  1812500 0
 1812500 6343750  4531250
0 u2 0  4531250 S c u3 s = c 0 s 4531250 u4 800
11. Solve the following set of equations (a) using the Gaussian method, (b) using the LU decomposition method, and (c) by finding the inverse of the coefficient matrix. 1 C 2 3
1 5 1
1 x1 6 1 S c x2 s = c 15 s 5 x3 14
12. Determine the inverse of the following matrices: 2 4 [A] = D 3 1
1 0 2 3
1 2 2 2
2 1 T 0 1
2 [B] = C 3 8
2 8 4
2 1S 5
[C ] = J
k11 k21
k12 R k22
13. Show that if we multiply the elements of a 4 * 4 matrix by a scalar quantity a, then the value of the determinant of the original matrix gets multiplied by a4. That is: a11 a det § a D 21 a31 a41
a12 a22 a32 a42
a13 a23 a33 a43
a14 a11 a24 a T ¥ = a4 4 21 a34 a31 a44 a41
a12 a22 a32 a42
a13 a23 a33 a43
a14 a24 4 a34 a44
Also, show that if we multiply the elements of a 3 * 3 matrix by the same scalar quantity a, then the value of the determinant of the original matrix gets multiplied by a3? How would you generalize the results of this example? 14. Determine the natural frequencies and the natural modes for the example problem in Section 2.9 for m1 = 0.1 kg, m2 = 0.2 kg, k = 100 N/m.
15. Solve Problem 2 using MATLAB and Excel. 16. Solve Problem 4 using MATLAB and Excel. 17. Solve Problem 5 using MATLAB. 18. Solve Problem 6 using MATLAB. 19. Solve Problem 7 using MATLAB. 20. Solve Problem 8 using MATLAB.
M02_MOAV4303_04_GE_C02.INDD 143
27/11/14 9:20 AM
144 Chapter 2 Matrix Algebra
21. Solve Problem 9 using MATLAB. 22. Solve Problem 10 using MATLAB and Excel. 23. Solve Problem 11 using MATLAB. 24. Solve Problem 13 using MATLAB. 25. Solve the following set of equations, resulting from a model of composite wall, using MATLAB by employing the matrix left division command: 7.11 1.23 E 0 0 0
1.23 1.99  0.76 0 0
0 0.76 0.851  0.091 0
0 0 0.091 2.31  2.22
0 T2 (5.88)(20) T3 0 0 u 0 U e T4 u = e 0 T5 2.22 0 3.69 T6 (1.47)(70)
and compare your solution to the results of Example 1.2.
26. Using MATLAB, solve Problem 25 by first finding the inverse of the coefficient matrix and carrying out the necessary operations afterward.
27. Using MATLAB, solve Problem 25 by first decomposing the coefficient matrix into lower and upper triangular matrices and carrying out the necessary operations afterward.
28. Solve the following set of equations, resulting from a model of a fin, using MATLAB: 1 0.04 E 0 0 0
0 0.09  0.04 0 0
0  0.04 0.09 0.04 0
0 0 0.04 0.09  0.04
0 T1 200 T2 0 0.15 0 U e T3 u = e 0.15 u T4  0.04 0.15 0.045 T5 0.08
29. Solve the following set of equations, resulting from a model of a truss, using MATLAB, and compare your solution to the results of Example 3.1: 7.2 0 0 105 F 0  1.49  1.49
0 7.2 0  4.22  1.49  1.49
0 0 8.44 0  4.22 0
0  4.22 0 4.22 0 0
1.49 1.49 4.22 0 5.71 1.49
1.49 U2X 0 U2Y 1.49 0 0 U4X 0 Vf v = f v 0 U4Y 500 1.49 U5X 0 1.49 U5Y  500
30. Solve Problem 14 using MATLAB. 31. Solve Problem 25 using Excel. 32. Solve Problem 28 using Excel. 33. Solve Problem 29 using Excel. 34. Use Excel and the method shown in Example 2.9 to setup and solve Example 1.2. 35. Use Excel and the method shown in Example 2.9 to setup and solve Example 1.3. 36. Use Excel and the method shown in Example 2.9 to setup and solve Example 1.4.
M02_MOAV4303_04_GE_C02.INDD 144
27/11/14 9:20 AM
C h a p t e r
3
Trusses
The objectives of this chapter are to introduce the basic concepts in finite element formulation of trusses and to provide an overview of the ANSYS program. A major section of this chapter is devoted to the Launcher, the Graphical User Interface, and the organization of the ANSYS program. The main topics discussed in Chapter 3 include the following: 3.1 Definition of a Truss 3.2 Finite Element Formulation 3.3 Space Trusses 3.4 Overview of the ANSYS Program 3.5 Examples Using ANSYS 3.6 Verification of Results
3.1 Definition of a Truss A truss is an engineering structure consisting of straight members connected at their ends by means of bolts, rivets, pins, or welding. The members found in trusses may consist of steel or aluminum tubes, wooden struts, metal bars, angles, and channels. Trusses offer practical solutions to many structural problems in engineering, such as power transmission towers, bridges, and roofs of buildings. A plane truss is defined as a truss whose members lie in a single plane. The forces acting on such a truss must also lie in this plane. Members of a truss are generally considered to be twoforce members. This term means that internal forces act in equal and opposite directions along the members, as shown in Figure 3.1. In the analysis that follows, it is assumed that the members are connected by smooth pins and by a ballandsocket joint in threedimensional trusses. Moreover, it can be shown that as long as the center lines of the joining members intersect at a common point, trusses with bolted or welded joints may be treated as having smooth pins (no bending). Another important assumption deals with the way loads are applied. All loads must be applied at the joints. This assumption is true for most situations because
M03_MOAV4303_04_GE_C03.INDD 145
27/11/14 9:37 AM
146 Chapter 3 Trusses
Load 3
Compression
3
3
Twoforce members 1 1
2
2 1
Tension
2
Figure 3.1 A simple truss subjected to a load.
trusses are designed so that the majority of the load is applied at the joints. Usually, the weights of members are negligible compared to those of the applied loads. However, if the weights of the members are to be considered, then half of the weight of each member is applied to the connecting joints. Statically determinate truss problems are covered in many elementary mechanics text. This class of problems is analyzed by the methods of joints or sections. These methods do not provide information on deflection of the joints because the truss members are treated as rigid bodies. Because the truss members are assumed to be rigid bodies, statically indeterminate problems are impossible to analyze. The finite element method allows us to remove the rigid body restriction and solve this class of problems. Figure 3.2 depicts examples of statically determinate and statically indeterminate problems.
3.2 Finite Element Formulation Let us consider the deflection of a single member when it is subjected to force F, as shown in Figure 3.3. The forthcoming derivation of the stiffness coefficient is identical to the analysis of a centrally loaded member that was presented in Chapter 1, Section 1.5. As a review and for the sake of continuity and convenience, the steps to derive the elements’ equivalent stiffness coefficients are presented here again. Recall that the average stresses in any twoforce member are given by
s =
F (3.1) A
The average strain of the member can be expressed by
e =
∆L (3.2) L
Over the elastic region, the stress and strain are related by Hooke’s law,
M03_MOAV4303_04_GE_C03.INDD 146
s = Ee (3.3)
27/11/14 9:37 AM
Section 3.2 Finite Element Formulation 147 Statically Determinate
Statically Indeterminate Y
Load
Load
3
1
R1X
2
1
2
Load
Load
3
3
1
R1Y
3
X
2
R1X
R2Y
1
R1Y
2
R2X
R2Y
4 unknown reactions 3 equilibrium equations ©FX = 0 ©FY = 0 ©M = 0
3 unknown reactions 3 equilibrium equations ©FX = 0 ©FY = 0 ©M = 0
Figure 3.2 Examples of statically determinate and statically indeterminate problems.
L
∆L F
M03_MOAV4303_04_GE_C03.INDD 147
Figure 3.3 A twoforce member sub jected to a force F.
27/11/14 9:38 AM
148 Chapter 3 Trusses Load
Load
(3)
(6)
3
4 (2)
(4)
(1) 1
5 (5)
u 2
Figure 3.4 A balcony truss.
Combining Eqs. (3.1), (3.2), and (3.3) and simplifying, we have F = ¢
AE ≤ ∆L (3.4) L
Note that Eq. (3.4) is similar to the equation of a linear spring, F = kx. Therefore, a centrally loaded member of uniform cross section may be modeled as a spring with an equivalent stiffness of
keq =
AE (3.5) L
A relatively small balcony truss with five nodes and six elements is shown in Figure 3.4. From this truss, consider isolating a member with an arbitrary orientation. Let us select element (5). In general, two frames of reference are required to describe truss problems: a global coordinate system and a local frame of reference. We choose a fixed global coordinate system, XY (1) to represent the location of each joint (node) and to keep track of the orientation of each member (element), using angles such as u; (2) to apply the constraints and the applied loads in terms of their respective global components; and (3) to represent the solution—that is, the displacement of each joint in global directions. We will also need a local, or an elemental, coordinate system xy, to describe the twoforce member behavior of individual members (elements). The relationship between the local (element) descriptions and the global descriptions is shown in Figure 3.5. The global displacements (UiX, UiY at node i and UjX, UjY at node j) are related to the local displacements (uix, uiy at node i and ujx, ujy at node j) according to the equations
M03_MOAV4303_04_GE_C03.INDD 148
UiX UiY UjX UjY
= = = =
uix cos u  uiy sin u (3.6) uix sin u + uiy cos u ujx cos u  ujy sin u ujx sin u + ujy cos u
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 149 UjY Y
ujx
ujy X
Global coordinate system
x
y
UjX
j
Local coordinate system
UiY uix
uiy
FjY u
fjx
fjy
UiX
i x
y
j
FjX
FiY fix
fiy
u
FiX
i Figure 3.5 Relationship between local and global coordinates. Note that local coordinate x points from node i toward j.
If we write Eqs. (3.6) in matrix form, we have 5 U6 = [T]5 u 6 (3.7)
where
sin u cos u 0 0
UiX cos u UiY sin u 5 U6 = d t , [T] = D UjX 0 UjY 0
0 0 cos u sin u
0 uix 0 u T , and 5 u 6 = d iy t sin u ujx cos u ujy
5 U6 and 5 u 6 represent the displacements of nodes i and j with respect to the global XY and the local xy frame of references, respectively. [T] is the transformation matrix that allows for the transfer of local deformations to their respective global values. In a similar way, the local and global forces may be related according to the equations
M03_MOAV4303_04_GE_C03.INDD 149
FiX FiY FjX FjY
= = = =
fix cos u  fiy sin u fix sin u + fiy cos u (3.8) fjx cos u  fjy sin u fjx sin u + fjy cos u
27/11/14 9:38 AM
150 Chapter 3 Trusses
or, in matrix form, where
5 F6 = [T]5 f6 (3.9) FiX F 5 F6 = d iY t FjX FjY
are components of forces acting at nodes i and j with respect to global coordinates, and fix f 5 f 6 = d iy t fjx fjy represent the local components of the forces at nodes i and j. A general relationship between the local and the global properties was derived in the preceding steps. However, we need to keep in mind that for a given member the forces in the local ydirection are zero. This fact is simply because under the twoforce assumption, a member can only be stretched or shortened along its longitudinal axis (local xaxis). In other words, the internal forces act only in the local xdirection as shown in Figure 3.6. We do not initially set these terms equal to zero in order to maintain a general matrix description that will make the derivation of the element stiffness matrix easier. This process will become clear when we set the ycomponents of the displacements and forces equal to zero. The local internal forces and displacements are related through the stiffness matrix fix k f 0 d iy t = D fjx k fjy 0
where k = keq =
0 0 0 0
k 0 k 0
0 0 Td 0 0
uix uiy t (3.10) ujx ujy
AE , and using matrix form we can write L 5 f6 = [k] 5 u 6 (3.11)
After substituting for 5 f6 and 5 u 6 in terms of 5 F6 and 5 U6, we have
M03_MOAV4303_04_GE_C03.INDD 150
5f6
5u6
$+%+& $+%+& (3.12) 1 [T] 5 F6 = [k][T]1 5 U6
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 151
u jx
f jx
=
k(

) u ix x
f jx
j
=
ui k(
) u jx
j u jx
y

u jx
x or
)
u ix
f ix
=
k(

i
u jx
)
u ix u ix
f ix
=
k(

i u ix
u jx
fix = k(uix  ujx)
fix
fiy = 0
fiy
fjx = k(ujx  uix) = k(uix  ujx) or in a matrix form
fjx
fjy = 0
fjy
=
k
0 k 0
uix
0
0
0
0
uiy
k 0
k
0
ujx
0
0
0
ujy
0
Figure 3.6 Internal forces for an arbitrary truss element. Note that the static equilibrium conditions require that the sum of fix and fjx be zero. Also note that the sum of fix and fjx is zero regardless of which representation is selected.
where [T]1 is the inverse of the transformation matrix [T] and is
[T]1
cos u sin u = D 0 0
sin u cos u 0 0
0 0 cos u sin u
0 0 T (3.13) sin u cos u
Multiplying both sides of Eq. (3.12) by [T] and simplifying, we obtain
5 F6 = [T][k][T]1 5 U6 (3.14)
Substituting for values of the [T], [k], [T]1, and 5 U6 matrices in Eq. (3.14) and multiplying, we are left with FiX cos2 u F sin u cos u d iY t = k D FjX cos2 u FjY sin u cos u
M03_MOAV4303_04_GE_C03.INDD 151
sin u cos u sin2 u sin u cos u sin2 u
cos2 u sin u cos u cos2 u sin u cos u
sin u cos u UiX 2 sin u U T d iY t (3.15) sin u cos u UjX sin2 u UjY
27/11/14 9:38 AM
152 Chapter 3 Trusses
Equations (3.15) express the relationship between the applied forces, the element stiffness matrix [K](e), and the global deflection of the nodes of an arbitrary element. The stiffness matrix [K](e) for any member (element) of the truss is
[K](e)
cos2 u sin u cos u = kD cos2 u sin u cos u
cos2 u sin u cos u cos2 u sin u cos u
sin u cos u sin2 u sin u cos u sin2 u
sin u cos u sin2 u T (3.16) sin u cos u sin2 u
The next few steps involve assembling, or connecting, the elemental stiffness matrices, applying boundary conditions and loads, solving for displacements, and obtaining other information, such as normal stresses. These steps are best illustrated through an example problem. Example 3.1 Consider the balcony truss in Figure 3.4, shown here with dimensions. We are interested in determining the deflection of each joint under the loading shown in the figure. All members are made from Douglasfir wood with a modulus of elasticity of E = 1.90 * 106 lb/in2 and a crosssectional area of 8 in2. We are also interested in calculating average stresses in each member. First, we will solve this problem manually. Later, once we learn how to use ANSYS, we will revisit this problem and solve it using ANSYS. 500 lb 3 ft (3)
500 lb 3 ft (6)
3
4 (2)
3 ft
(4)
5 (5)
(1) 1
2
As discussed in Chapter 1, Section 1.4, there are seven steps involved in any finite element analysis. Here, these steps are discussed again to emphasize the three phases (preprocessing, solution, and postprocessing) associated with the analysis of truss problems. Preprocessing Phase 1. Discretize the problem into nodes and elements. Each truss member is considered an element, and each joint connecting members is a node. Therefore, the given truss can be modeled with five nodes and six elements. Consult Table 3.1 while following the solution.
M03_MOAV4303_04_GE_C03.INDD 152
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 153
Table 3.1 The relationship between the elements and their corresponding nodes Element
Node i
Node j
(1)
1
2
0
(2)
2
3
135
(3)
3
4
0
(4)
2
4
90
(5)
2
5
45
(6)
4
5
0
u See Figures 3.7–3.10
2. Assume a solution that approximates the behavior of an element. As discussed in Section 3.2, we will model the elastic behavior of each element as a spring with an equivalent stiffness of k as given by Eq. (3.5). Since elements (1), (3), (4), and (6) have the same length, crosssectional area, and modulus of elasticity, the equivalent stiffness constant for these elements (members) is lb (8 in2) ¢ 1.90 * 106 2 ≤ in AE k = = = 4.22 * 105 lb/in. L 36 in The stiffness constant for elements (2) and (5) is AE k = = L
(8 in2) ¢ 1.90 * 106
lb ≤ in2
50.9 in
= 2.98 * 105 lb/in.
3. Develop equations for elements. For elements (1), (3), and (6), the local and the global coordinate systems are aligned, which means that u = 0. This relationship is shown in Figure 3.7. Using Eq. (3.16), we find that the stiffness matrices are
[K](e)
cos2 u sin u cos u = kD cos2 u sin u cos u
[K](1)
cos2(0) sin(0) cos(0) = 4.22 * 105 D cos2(0) sin(0) cos(0)
[K](1)
1 0 = 4.22 * 105 D 1 0
M03_MOAV4303_04_GE_C03.INDD 153
sin u cos u sin2 u sin u cos u sin2 u
0 0 0 0
1 0 1 0
cos2 u sin u cos u cos2 u sin u cos u
sin u cos u sin2 u T sin u cos u sin2 u
sin(0) cos(0) sin2(0) sin(0) cos(0) sin2(0)
cos2(0) sin(0) cos(0) cos2(0) sin(0) cos(0)
sin(0) cos(0) sin2(0) T sin(0) cos(0) sin2(0)
0 U1X 0 U1Y T 0 U2X 0 U2Y
27/11/14 9:38 AM
154 Chapter 3 Trusses y
Y
X
x i=1
Element (1) or Element (3) or Element (6)
3 4
j=2 4
Figure 3.7 The orientation of the local coordinates with respect to the global coordinates for elements (1), (3), and (6).
5
and the position of element (1)’s stiffness matrix in the global matrix is
[K](1G)
4.22 0 4.22 0 0 = 105 I 0 0 0 0 0
4.22 0 4.22 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 U1X 0 U1Y 0 U2X 0 U2Y 0 U3X Y 0 U3Y 0 U4X 0 U4Y 0 U5X 0 U5Y
Note that the nodal displacement matrix is shown alongside element (1)’s position in the global matrix to aid us in observing the location of element (1)’s stiffness matrix in the global matrix. Similarly, the stiffness matrix for element (3) is
[K](3)
1 0 = 4.22 * 105 D 1 0
0 0 0 0
1 0 1 0
0 U3X 0 U3Y T 0 U4X 0 U4Y
and its position in the global matrix is
[K](3G) = 105 I
M03_MOAV4303_04_GE_C03.INDD 154
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 4.22 0 4.22 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 4.22 0 4.22 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 Y 0 0 0 0 0
U1X U1Y U2X U2Y U3X U3Y U4X U4Y U5X U5Y
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 155
The stiffness matrix for element (6) is [K](6)
1 0 = 4.22 * 105 D 1 0
1 0 1 0
0 0 0 0
0 U4X 0 U4Y T 0 U5X 0 U5Y
and its position in the global matrix is
[K](6G)
[K](4)
0 0 0 0 0 = 105 I 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 4.22 0 4.22 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 4.22 0 4.22 0
0 U1X 0 U1Y 0 U2X 0 U2Y 0 U3X Y 0 U3Y 0 U4X 0 U4Y 0 U5X 0 U5Y
For element (4), the orientation of the local coordinate system with respect to the global coordinates is shown in Figure 3.8. Thus, for element (4), u = 90, which leads to the stiffness matrix cos2(90) sin(90) cos(90) cos2(90) sin(90) cos(90) 2 sin(90) cos(90) sin (90) sin(90) cos(90) sin2(90) = 4.22 * 105 D T cos2(90) sin(90) cos(90) cos2(90) sin(90) cos(90) sin(90) cos(90) sin2(90) sin(90) cos(90) sin2(90)
0 0 [K](4) = 4.22 * 105 D 0 0
0 1 0 1
0 0 0 0
0 U2X 1 U2Y T 0 U4X 1 U4Y
and its global position
[K](4G) = 10 5 I
M03_MOAV4303_04_GE_C03.INDD 155
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 4.22 0 0 0 4.22 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 4.22 0 0 0 4.22 0 0
0 0 0 0 0 0 0 0 0 0
0 U1X 0 U1Y 0 U2X 0 U2Y 0 U3X Y 0 U3Y 0 U4X 0 U4Y 0 U5X 0 U5Y
27/11/14 9:38 AM
156 Chapter 3 Trusses 4 x
Y
X
Element (4)
y
Figure 3.8 The orientation of the local coordinates with respect to the global coordinates for element (4).
2
For element (2), the orientation of the local coordinate system with respect to the global coordinates is shown in Figure 3.9. Thus, for element (2), yielding the stiffness matrix
[K](2)
cos2(135) sin(135) cos(135) = 2.98 * 105 D cos2(135) sin(135) cos(135)
sin(135) cos(135) sin2(135) sin(135) cos(135) sin2(135)
cos2(135) sin(135) cos(135) cos2(135) sin(135) cos(135)
[K](2)
.5 .5 .5 .5
.5 .5 = 2.98 * 105 D .5 .5
.5 .5 .5 .5
sin(135) cos(135) sin2(135) T sin(135) cos(135) sin2(135)
.5 U2X .5 U2Y T .5 U3X .5 U3Y
3 x x
Y 1355
Element (2) X y
y 2
M03_MOAV4303_04_GE_C03.INDD 156
Figure 3.9 The orientation of the local coordinates with respect to the global coordinates for element (2).
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 157
Simplifying, we get
[K](2)
1 1 = 1.49 * 105 D 1 1
1 1 1 1
1 1 1 1
1 U2X 1 U2Y T 1 U3X 1 U3Y
and its position in the global matrix is 0 0 0 0 0 [K](2G) = 105 I 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 1.49  1.49  1.49 1.49 0 0 0 0
0 0  1.49 1.49 1.49  1.49 0 0 0 0
0 0  1.49 1.49 1.49  1.49 0 0 0 0
0 0 1.49  1.49  1.49 1.49 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 Y 0 0 0 0 0
U1X U1Y U2X U2Y U3X U3Y U4X U4Y U5X U5Y
For element (5), the orientation of the local coordinate system with respect to the global coordinates is shown in Figure 3.10. Thus, for element (5), u = 45, yielding the stiffness matrix cos2(45) sin(45) cos(45) cos2(45) sin(45) cos(45) 2 sin(45) cos(45) sin (45) sin(45) cos(45) sin2(45) [K](5) = 2.98 * 105 D T 2 2 cos (45) sin(45) cos(45) cos (45) sin(45) cos(45) sin(45) cos(45) sin2(45) sin(45) cos(45) sin2(45) .5 .5 [K](5) = 2.98 * 105 D .5 .5
.5 .5 .5 .5
.5 .5 .5 .5
.5 U2X .5 U2Y T .5 U5X .5 U5Y
and its position in the global stiffness matrix is Y y
x 455
y x
5
X Element (5)
2
M03_MOAV4303_04_GE_C03.INDD 157
Figure 3.10 The orientation of the local coordinates with respect to the global coordinates for element (5).
27/11/14 9:38 AM
158 Chapter 3 Trusses
0 0 0 0 0 [K](5G) = 105 I 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 1.49 1.49 0 0 0 0 1.49 1.49
0 0 1.49 1.49 0 0 0 0 1.49 1.49
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 1.49 1.49 0 0 0 0 1.49 1.49
0 0 1.49 1.49 0 Y 0 0 0 1.49 1.49
U1X U1Y U2X U2Y U3X U3Y U4X U4Y U5X U5Y
It is worth noting again that the nodal displacements associated with each element are shown next to each element’s stiffness matrix. This practice makes it easier to connect (assemble) the individual stiffness matrices into the global stiffness matrix for the truss. 4. Assemble elements. The global stiffness matrix is obtained by assembling, or adding together, the individual elements’ matrices: (G) [K] = [K](1G) + [K](2G) + [K](3G) + [K](4G) + [K](5G) + [K](6G)
[K](G) = 105 I
4.22 0 4.22 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
4.22 0 4.22 + 1.49 + 1.49 1.491.49 1.49 1.49 0 0 1.49 1.49 0 0 1.49 1.49 1.49 1.49 0 0 0 0
M03_MOAV4303_04_GE_C03.INDD 158
0 0 0 0 4.22 0 4.22 + 4.22 0 4.22 0
0 0 1.49 + 1.49 4.22 + 1.49 + 1.49 1.49 1.49 0 4.22 1.49 1.49 0 0 0 4.22 0 0 0 4.22 0 0
0 0 1.49 1.49 0 0 4.22 0 4.22 + 1.49 1.49
0 0 1.49 1.49 4.22 + 1.49 1.49 4.22 0 0 0 0 U1X 0 U1Y 1.49 U2X 1.49 U2Y 0 U Y 3X 0 U3Y 0 U4X 0 U4Y 1.49 U5X 1.49 U5Y
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 159
Simplifying, we get 4.22 0 4.22 0 0 [K](G) = 105 I 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
4.22 0 7.2 0 1.49 1.49 0 0 1.49 1.49
0 0 0 7.2 1.49 1.49 0 4.22 1.49 1.49
0 0 1.49 1.49 5.71 1.49 4.22 0 0 0
0 0 0 0 0 0 0 0 0 0 1.49 0 0 1.49 1.49 1.49 0 4.22 1.49 1.49 1.49 4.22 0 0 0 Y 1.49 0 0 0 0 0 8.44 0 4.22 0 0 0 4.22 0 0 0 4.22 0 5.71 1.49 0 0 0 1.49 1.49
5. Apply the boundary conditions and loads. The following boundary conditions apply to this problem: nodes 1 and 3 are fixed, which implies that U1X = 0, U1Y = 0, U3X = 0, and U3Y = 0. Incorporating these conditions into the global stiffness matrix and applying the external loads at nodes 4 and 5 such that F4Y = 500 lb and F5Y = 500 lb results in a set of linear equations that must be solved simultaneously:
105 I
1 0 4.22 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 U1X 1 0 0 0 0 0 0 0 0 U1Y 0 7.2 0 1.49 1.49 0 0 1.49 1.49 U2X 0 0 7.2 1.49 1.49 0 4.22 1.49 1.49 U2Y 0 0 0 1 0 0 0 0 0 U3X Y i y = 0 0 0 0 1 0 0 0 0 U3Y 0 0 0 4.22 0 8.44 0 4.22 0 U4X 0 0 4.22 0 0 0 4.22 0 0 U4Y 0 1.49 1.49 0 0 4.22 0 5.71 1.49 U5X 0 1.49 1.49 0 0 0 0 1.49 1.49 U5Y
0 0 0 0 0 i y 0 0  500 0  500
Because U1X = 0, U1Y = 0, U3X = 0, and U3Y = 0, we can eliminate the first, second, fifth, and sixth rows and columns from our calculation such that we need only solve a 6 * 6 matrix:
M03_MOAV4303_04_GE_C03.INDD 159
27/11/14 9:38 AM
160 Chapter 3 Trusses
7.2 0 0 105 F 0 1.49 1.49
0 7.2 0 4.22 1.49 1.49
0 0 8.44 0 4.22 0
0 4.22 0 4.22 0 0
1.49 1.49 4.22 0 5.71 1.49
1.49 U2X 0 1.49 U2Y 0 0 U 0 V f 4X v = f v 0 U4Y 500 1.49 U5X 0 1.49 U5Y 500
Solution Phase 6. Solve a system of algebraic equations simultaneously. Solving the above matrix for the unknown displacements yields U2X = 0.00355 in, U2Y = 0.01026 in, U4X = 0.00118 in, U4Y = 0.0114 in, U5X = 0.00240 in, and U5Y = 0.0195 in. Thus, the global displacement matrix is
i
U1X U1Y U2X U2Y U3X U3Y U4X U4Y U5X U5Y
y = i
0 0  0.00355  0.01026 0 0 0.00118  0.0114 0.00240  0.0195
y in.
It is important to recognize that the displacements of the nodes are given with respect to the global coordinate system. Postprocessing Phase 7. Obtain other information. Reaction Forces As discussed in Chapter 1, the reaction forces can be computed from
such that
M03_MOAV4303_04_GE_C03.INDD 160
5 R 6 = [K](G) 5 U6  5 F6
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 161
i
4.22 R1X 0 R1Y 4.22 R2X 0 R2Y 0 R3X y = 105 I 0 R3Y 0 R4X 0 R4Y 0 R5X 0 R5Y
0 0 0 0 0 0 0 0 0 0
4.22 0 7.2 0 1.49 1.49 0 0 1.49 1.49
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1.49 1.49 0 0 1.49 1.49 7.2 1.49 1.49 0 4.22 1.49 1.49 1.49 5.71 1.49 4.22 0 0 0 Y 1.49 1.49 1.49 0 0 0 0 0 4.22 0 8.44 0 4.22 0 4.22 0 0 0 4.22 0 0 1.49 0 0 4.22 0 5.71 1.49 1.49 0 0 0 0 1.49 1.49 0 0 0 0 0.00355 0 0.01026 0 0 0 i yi y 0 0 0.00118 0 0.0114  500 0.00240 0 0.0195  500
Note that the entire stiffness, displacement, and load matrices are used. Performing matrix operations yields the reaction results R1X R1Y R2X R2Y R3X i y = R3Y R4X R4Y R5X R5Y
1500 0 0 0  1500 i y lb 1000 0 0 0 0
Internal Forces and Normal Stresses Now let us compute internal forces and the average normal stresses in each member. The member internal forces fix and fjx, which are equal and opposite in direction, are
M03_MOAV4303_04_GE_C03.INDD 161
27/11/14 9:38 AM
162 Chapter 3 Trusses Y ujx
ujx
fjx = k(uix  ujx)
fjx = k(ujx – uix)
X y x
y
j
x or
uix
j
uix
i
i
fix = k(uix  ujx)
fix = k(uix  ujx) Figure 3.11 Internal forces in a truss member.
fix = k(uix  ujx) (3.17) fjx = k(ujx  uix)
Note that the sum of fix and fjx is zero regardless of which representation of Figure 3.11 we select. However, for the sake of consistency in the forthcoming derivation, we will use the second representation so that fix and fjx are given in the positive local xdirection. In order to use Eq. (3.17) to compute the internal force in a given element, we must know the displacements of the element’s end nodes, uix and ujx, with respect to the local coordinate system, x, y. Recall that the global displacements are related to the local displacements through a transformation matrix, according to Eq. (3.7), repeated here for convenience, 5 U6 = [T]5 u 6
and the local displacements in terms of the global displacements: 5 u 6 = [T]1 5 U6
uix cos u uiy sin u d t = D ujx 0 ujy 0
sin u cos u 0 0
0 0 cos u sin u
0 UiX 0 U T D iY T sin u UjX cos u UjY
Once the internal force in each member is computed, the normal stress in each member can be determined from the equation s =
f internal force = area A
or alternatively, we can compute the normal stresses from AE (u  ujx) u ) uix  ujx k(u f L ix ix jx s = = = = E¢ ≤ (3.18) A A A L
M03_MOAV4303_04_GE_C03.INDD 162
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 163
As an example, let us compute the internal force and the normal stress in element (5). For element (5), u = 45, U2X = 0.00355 in, U2Y = 0.01026 in, U5X = 0.0024 in, and U5Y = 0.0195 in. First, we solve for local displacements of nodes 2 and 5 from the relation u2x cos 45 u2y sin 45 d t = D u5x 0 u5y 0
sin 45 cos 45 0 0
0 0 cos 45 sin 45
0 0.00355 0 0.01026 Td t sin 45 0.00240 cos 45 0.01950
which reveals that u2x = 0.00976 in and u5x = 0.01209 in. Upon substitution of these values into Eqs. (3.17) and (3.18), the internal force and the normal stress in element (5) are 696 lb and 87 lb/in2, respectively. Similarly, the internal forces and stresses can be obtained for other elements. This problem will be revisited later in this chapter and solved using ANSYS. The verification of these results will also be discussed in detail later in Section 3.6. Example 3.1 Revisited We will now show how to use Excel to set up and solve Example 3.1. 1. In cell A1 type Example 3.1, and in cells A3 and A4 type E= and A= as shown. After inputting the value of E in cell B3, select B3 and in the “Name Box” type E and hit the Return key. Similarly, after inputting the value of A in cell B4, select B4 and in the “Name Box” type A and hit the Return key.
2. Create the table shown with element and node numbers, length, area, and modulus of elasticity for each member. In cells G7:G12 input Θ for each element and name them Theta1, Theta2, Theta3, Theta4, Theta5, and Theta6, respectively. Also, name the values in D7:D12, Length1, Length2, Length3, Length4, Length5, and Length6.
M03_MOAV4303_04_GE_C03.INDD 163
27/11/14 9:38 AM
164 Chapter 3 Trusses
3. Next compute the Cosine and Sine terms for [K1] as shown and name the selected range CSelement1.
4. In a similar way, create the Cosine and Sine terms for [K2], [K3], [K4], [K5], and [K6] and name the selected ranges CSelement2, CSelement3, CSelement4, CSelement5, and CSelement6.
5. Next, create the [A1] matrix and name it Aelement1, as shown. If you have forgotten what the A matrices represents, see Section 2.5, Equation (2.9). The nodal displacement U1X, U1Y, U2X, U2Y, U3X, U3Y, U4X, U4Y, U5X, U5Y, and UiX, UiY, UjX, and UjY are shown alongside the [A1] matrix to aid us observe the contribution of node to its neighboring elements.
M03_MOAV4303_04_GE_C03.INDD 164
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 165
6. Next, create the [A2], [A3], [A4], [A5], and [A6] matrices and name them Aelement2, Aelement3, Aelement4, Aelement5, and Aelement6, as shown.
7. We now create the stiffness matrix for each element (with their proper positions in the global matrix) and name them K1G, K2G, K3G, K4G, K5G, and K6G. See Equation (2.9). For example, to create [K]1G, select B80:K89 and type =MMULT (TRANSPOSE(Aelement1),MMULT(((A*E/Length1)*CSelement1), Aelement1)) and while holding down the Ctrl and Shift keys hit the Return key. In a similar way, create [K]2G, [K]3G, [K]4G, [K]5G, and [K]6G as shown.
M03_MOAV4303_04_GE_C03.INDD 165
27/11/14 9:38 AM
166 Chapter 3 Trusses
M03_MOAV4303_04_GE_C03.INDD 166
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 167
8. The final global matrix is created next. Select the range B146:K155 and type =K1G+K2G+K3G+K4G+K5G+K6G and while holding down the Ctrl and Shift keys hit the Return key. Name the range B146:K155, KG. In a similar way, create the global load matrix.
9. Apply the boundary conditions. Copy the appropriate portion of the KG matrix and paste it in the range C168:H173 as values only. Name the range KwithappliedBC. Similarly, create the corresponding load matrix in the range C175:C180 and name it FwithappliedBC.
M03_MOAV4303_04_GE_C03.INDD 167
27/11/14 9:38 AM
168 Chapter 3 Trusses
10. Select the range C182:C187 and type =MMULT(MINVERSE(KwithappliedBC),FwithappliedBC) and while holding down the Ctrl and Shift keys hit the Return key. Moreover, as shown, select C189:C198, and copy the values of U partial and add the U1X=0, U1Y=0, U3X=0, and U3Y=0. Name this matrix, UG.
11. Calculate the reaction forces, next. Select the range C200:C209 and type =(MMULT(KG,UG)@FG) and while holding down the Ctrl and Shift keys, hit the Return key.
M03_MOAV4303_04_GE_C03.INDD 168
27/11/14 9:38 AM
Section 3.2 Finite Element Formulation 169
The complete Excel sheet is shown next.
M03_MOAV4303_04_GE_C03.INDD 169
27/11/14 9:38 AM
170 Chapter 3 Trusses
M03_MOAV4303_04_GE_C03.INDD 170
27/11/14 9:38 AM
Section 3.3 Space Trusses 171
3.3 Space Trusses A threedimensional truss is often called a space truss. A simple space truss has six members joined together at their ends to form a tetrahedron, as shown in Figure 3.12. We can create more complex structures by adding three new members to a simple truss. This addition should be done in a manner where one end of each new member is connected to a separate existing joint, attaching the other ends of the new members together to form a new joint. This structure is shown in Figure 3.13. As mentioned earlier, members of a truss are generally considered to be twoforce members. In the analysis of space trusses, it is assumed that the members are connected together by ballandsocket joints. It can be shown that as long as the center lines of the adjacent bolted members intersect at a common point, trusses with bolted or welded joints may also be C
B
C
B D
D
A E A Figure 3.12 A simple truss.
M03_MOAV4303_04_GE_C03.INDD 171
Figure 3.13 Addition of new elements to a simple truss to form complex structures.
27/11/14 9:38 AM
172 Chapter 3 Trusses
treated under the ballandsocket joints assumption (negligible bending moments at the joints). Another restriction deals with the assumption that all loads must be applied at the joints. This assumption is true for most situations. As stated earlier, the weights of members are usually negligible compared to the applied loads. However, if the weights of the members are to be considered, then half of the weight of each member is applied to the connecting joints. Finite element formulation of space trusses is an extension of the analysis of plane trusses. In a space truss, the global displacement of an element is represented by six unknowns, UiX, UiY, UiZ, UjX, UjY, and UjZ, because each node (joint) can move in three directions. Moreover, the angles uX, uY, and uZ define the orientation of a member with respect to the global coordinate system, as shown in Figure 3.14. The directional cosines can be written in terms of the difference between the coordinates of nodes j and i of a member and the member’s length according to the relationships
cos uX =
cos uY =
cos uZ =
Xj  Xi L Yj  Yi L Zj  Zi L
(3.19)
(3.20) (3.21)
j Z
Y uY uZ
uX i
X
M03_MOAV4303_04_GE_C03.INDD 172
Figure 3.14 The angles formed by a member with the X, Y, and Zaxis.
27/11/14 9:38 AM
Section 3.4 Overview of the ANSYS Program 173
where L is the length of the member and is given by L = 2(Xj  Xi)2 + (Yj  Yi)2 + (Zj  Zi)2 (3.22)
The procedure for obtaining the element stiffness matrix for a spacetruss member is identical to the one we followed to derive the twodimensional truss element stiffness. We start the procedure by relating the global displacements and forces to local displacements and forces through a transformation matrix. We then make use of the twoforcemember property of a member. We use a matrix relationship similar to the one given by Eq. (3.14). This relationship leads to the stiffness matrix [K](e) for an element. However, it is important to realize that the elemental stiffness matrix for a spacetruss element is a 6 * 6 matrix rather than the 4 * 4 matrix that we obtained for the twodimensional truss element. For a spacetruss member, the elemental stiffness matrix is
[K](e)
cos2 uX cos uX cos uY cos uX cos uZ = kF cos2 uX cos uX cos uY cos uX cos uZ
cos uX cos uY cos2 uY cos uY cos uZ cos uX cos uY cos2 uY cos uY cos uZ cos2 uX cos uX cos uY cos uX cos uZ cos2 uX cos uX cos uY cos uX cos uZ
cos uX cos uZ cos uY cos uZ cos2 uZ cos uX cos uZ cos uY cos uZ cos2 uZ cos uX cos uY cos2 uY cos uY cos uZ cos uX cos uY cos2 uY cos uY cos uZ
cos uX cos uZ cos uY cos uZ cos2 uZ V (3.23) cos uX cos uZ cos uY cos uZ cos2 uZ
The procedure for the assembly of individual elemental matrices for a spacetruss member—applying boundary conditions, loads, and solving for displacements—is exactly identical to the one we followed for a twodimensional truss. 3.4 Overview of the ANSYS* Program Entering ANSYS This section provides a brief overview of the ANSYS program. More detailed information about how you should go about using ANSYS to model a physical problem is provided in Chapter 8. But for now, enough information will be provided to get you started. One way to enter the ANSYS program is through the ANSYS Launcher, shown in Figure 3.15. The Launcher has a menu that provides the choices you need to run the ANSYS program and other auxiliary programs.
*Materials were adapted with permission from ANSYS documents.
M03_MOAV4303_04_GE_C03.INDD 173
27/11/14 9:38 AM
174 Chapter 3 Trusses
Figure 3.15 The ANSYS Product Launcher for a PC version.
When using the Launcher to enter ANSYS, follow these basic steps: 1. Activate the Launcher by issuing the proper command at the system prompt if you are running ANSYS on a UNIX Platform. On a PC platform, go to: Start S Programs S ANSYS 15.0 S ANSYS Product Launcher®. 2. Select the ANSYS option from the Launcher menu by positioning the cursor of the mouse over it and clicking the left mouse button. This command brings up a dialog box containing interactive entry options. a. Working directory: This directory is the one in which the ANSYS run will be executed. If the directory displayed is not the one you want to work in, pick the Browse button to the right of the directory name and specify the desired directory. b. Jobname: This jobname is the one that will be used as the prefix of the file name for all files generated by the ANSYS run. Type the desired jobname in this field of the dialog box.
M03_MOAV4303_04_GE_C03.INDD 174
27/11/14 9:38 AM
Section 3.4 Overview of the ANSYS Program 175
3. Move the mouse cursor over the Run button at the bottom of the window and press it. The Graphical User Interface (GUI) will then be activated, and you are ready to begin. Program Organization Before introducing the GUI, we will discuss some basic concepts of the ANSYS program. The ANSYS program is organized into two levels: (1) the Begin level and (2) the Processor level. When you first enter the program, you are at the Begin level. From this level, you can enter the ANSYS processors, as shown in Figure 3.16. You may have more or fewer processors available to you than the ones shown in Figure 3.16. The actual processors available depend on the particular ANSYS product you have. The Begin level acts as a gateway into and out of the ANSYS program. It is also used to access certain global program controls. At the Processor level, several routines (processors) are available; each accomplishes a specific task. Most of your analysis will be done at the Processor level. A typical analysis in ANSYS involves three distinct steps: 1. Preprocessing: Using the PREP7 processor, you provide data such as the geometry, materials, and element type to the program. 2. Solution: Using the Solution processor, you define the type of analysis, set boundary conditions, apply loads, and initiate finite element solutions. 3. Postprocessing: Using POST1 (for static or steadystate problems) or POST26 (for transient problems), you review the results of your analysis through graphical displays and tabular listings. You enter a processor by selecting it from the ANSYS main menu in the GUI. You can move from one processor to another by simply choosing the processor you want from the ANSYS main menu. The next section presents a brief overview of the GUI. Enter ANSYS
Exit ANSYS
BEGIN LEVEL
PREP7 General Preprocessor
SOLUTION Processor
POST1 General Postprocessor
POST26 Time–History Postprocessor
Etc.
PROCESSOR LEVEL
Figure 3.16 The organization of ANSYS.
M03_MOAV4303_04_GE_C03.INDD 175
27/11/14 9:38 AM
176 Chapter 3 Trusses
The Graphical User Interface (GUI) The simplest way to communicate with ANSYS is by using the ANSYS menu system, called the Graphical User Interface. The GUI provides an interface between you and the ANSYS program. The program is internally driven by ANSYS commands. However, by using the GUI, you can perform an analysis with little or no knowledge of ANSYS commands. This process works because each GUI function ultimately produces one or more ANSYS commands that are automatically executed by the program. Layout of the GUI The ANSYS GUI consists of six main regions, or windows, as shown in Figure 3.17. A
B
Utility Menu: Contains utility functions that are available throughout the ANSYS session, such as file controls, selecting, and graphics controls. You also exit the ANSYS program through this menu. Main Menu: Contains the primary ANSYS functions, organized by processors. These functions include preprocessor, solution, general postprocessor, Design Xplorer, and so on.
D A C
B
E
Figure 3.17 The ANSYS GUI.
M03_MOAV4303_04_GE_C03.INDD 176
27/11/14 9:38 AM
Section 3.4 Overview of the ANSYS Program 177
C
Toolbar: Contains push buttons that execute commonly used ANSYS commands and functions. You may add your own push buttons by defining abbreviations.
D
Input Window: Allows you to type in commands directly. All previously typedin commands also appear in this window for easy reference and access.
E
Graphics Window: A window where graphics displays are drawn. Output Window: Receives text output from the program. It is usually positioned behind the other windows and can be brought to the front when necessary.
The ANSYS main menu and the ANSYS utility menu, both of which you will use most often, are discussed next. The Main Menu The main menu, shown in Figure 3.18(a), contains main ANSYS functions such as preprocessing, solution, and postprocessing. Each menu topic on the main menu either brings up a submenu or performs an action. The ANSYS main menu has a tree structure. Each menu topic can be expanded to reveal other menu options. The expansion of menu options is indicated by +. You click on the + or the topic name until you reach the desired action. As you reveal other subtopics, the + will turn into , as shown in Figure 3.18(b). For example, to create a rectangle, you click on Preprocessor, then on Modeling, Create, Areas, and Rectangle. As you can see from Figure 3.18(b), you now have three options to create the rectangle: By 2 Corners, or By Centr&Cornr, or By Dimensions. Note that each time you revealed another subtopic, the + turned into . The left mouse button is used to select a topic from the main menu. The submenus in the main menu stay in place until you choose a different menu topic higher up in the hierarchy. The Utility Menu The utility menu, shown in Figure 3.19, contains ANSYS utility functions such as file controls, selecting, and graphic controls. Most of these functions are modeless; that is, they can be executed at any time during the ANSYS session. The modeless nature of the utility menu greatly enhances the productivity and user friendliness of the GUI. Each menu topic on the utility menu activates a pulldown menu of subtopics, which in turn will either cascade to a submenu, indicated by a ▶ after the topic, or perform an action. The symbol to the right of the topic indicates the action: no symbol for immediate execution of the function c for a dialog box + for a picking menu.
M03_MOAV4303_04_GE_C03.INDD 177
27/11/14 9:38 AM
178 Chapter 3 Trusses
(a)
(b) Figure 3.18 The main menu.
Figure 3.19 The utility menu.
Clicking the left mouse button on a menu topic on the utility menu is used to “pull down” the menu topic. Dragging the cursor of the mouse allows you to move the cursor to the desired subtopic. The menus will disappear when you click on an action subtopic or elsewhere in the GUI.
M03_MOAV4303_04_GE_C03.INDD 178
27/11/14 9:38 AM
Section 3.4 Overview of the ANSYS Program 179
Graphical Picking In order to use the GUI effectively, it is important to understand graphical picking. You can use the mouse to identify model entities and coordinate locations. There are two types of graphicalpicking operations: locational picking, where you locate the coordinates of a new point, and retrieval picking, where you identify existing entities. For example, creating key points by picking their locations on the working plane is a locationalpicking operation, whereas picking alreadyexisting key points to apply a load on them is a retrievalpicking operation. Whenever you use graphical picking, the GUI brings up a picking menu. Figure 3.20 shows the picking menus for locational and retrieval picking. The features of the picking menu that are used most frequently in upcoming examples are described in detail below. B
A
Function Title
B
Picking Mode
Picking Mode: Allows you to pick or unpick a location or entity. You can use either these toggle buttons or the right mouse button to switch between pick and unpick modes. The mouse pointer is an up arrow for picking and a down arrow for unpicking. For retrieval picking, you also have the option to choose from single pick, box, circle, and polygon mode.
A
A B
C
Picking Status
D
Picked Data
E
Keyboard Entry Options
F
B
C
C
D
D
Action Buttons E
E
F
F
Figure 3.20 Picking menu for locational and retrieval picking.
M03_MOAV4303_04_GE_C03.INDD 179
27/11/14 9:38 AM
180 Chapter 3 Trusses
D
Picked Data: Shows information about the item being picked. For locational picking, the working plane and global Cartesian coordinates of the point are shown. For retrieval picking, this area shows the entity number. You can see this data by pressing the mouse button and dragging the cursor of the mouse into the graphics area. This procedure allows you to preview the information before releasing the mouse button and picking the item.
F
Action Buttons: This area of the menu contains buttons that take certain actions on the picked entities, as follows: OK: Applies the picked items to execute the function and closes the picking menu. Apply: Applies the picked items to execute the function. Reset: Unpicks all picked entities. Cancel: Cancels the function and closes the picking menu. Pick All: Picks all entities available for retrieval picking only. Help: Brings up help information for the function being performed.
MouseButton Assignments for Picking A summary of the mousebutton assignments used during a picking operation is given below: The left button picks or unpicks the entity or location closest to the cursor of the mouse. Pressing the left mouse button and dragging the cursor of the mouse allows you to preview the items being picked or unpicked. The middle button applies the picked items to execute the function. Its function is the same as that of the Apply button on the picking menu. The right button toggles between pick and unpick mode. Its function is the same as that of the toggle buttons on the picking menu.
The Help System The ANSYS help system gives you information for virtually any component in the GUI and any ANSYS command or concept. It can be accessed within the GUI via the help topic on the utility menu or by pressing the help button from within a dialog box. You can access a help topic by choosing from a manual’s table of contents or index. Other features of the help system include hypertext links, word search, and the ability to print out help topics. An indepth explanation of the capabilities and the organization of the ANSYS program is offered in Chapter 8.
M03_MOAV4303_04_GE_C03.INDD 180
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 181
3.5 Examples Using ANSYS In this section, ANSYS is used to solve truss problems. To analyze truss problems, ANSYS offers a threedimensional spar element. This element, denoted by LINK180, offers three degrees of freedom (UX, UY, UZ) at each node. Input data must include node locations, crosssectional area of the member, and modulus of elasticity. If a member is prestressed, then the initial strain should be included in the input data as well. As we learned previously in our discussion on the theory of truss element, we cannot apply surface loads to this element; thus, all loads must be applied directly at the nodes. To get additional information about these elements, run the ANSYS online help menu.
Example 3.2 Revisited Consider the balcony truss from Example 3.1, as shown in the accompanying figure. We are interested in determining the deflection of each joint under the loading shown in the figure. All members are made from Douglasfir wood with a modulus of elasticity of E = 1.90 * 106 lb/in2 and a crosssectional area of 8 in2. We can now analyze this problem using ANSYS. 500 lb 3 ft (3)
500 lb 3 ft (6)
3
4 (2)
3 ft
(4)
5 (5)
(1) 1
2
The following steps demonstrate how to create the truss geometry, choose the appropriate element type, apply boundary conditions and loads, and obtain results: Enter the ANSYS program by using the Launcher. Type Truss (or a file name of your choice) in the Jobname entry field of the dialog box.
M03_MOAV4303_04_GE_C03.INDD 181
27/11/14 9:38 AM
182 Chapter 3 Trusses
Pick Run to start the GUI. Create a title for the problem. This title will appear on ANSYS display windows to provide a simple way of identifying the displays. To create a title, issue the command utility menu: File S ChangeTitle c
M03_MOAV4303_04_GE_C03.INDD 182
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 183
Define the element type and material properties: main menu: Preprocessor S Element type S Add/Edit/Delete
M03_MOAV4303_04_GE_C03.INDD 183
27/11/14 9:38 AM
184 Chapter 3 Trusses
Assign the crosssectional area of the truss members: main menu: Preprocessor S Real Constants S Add/Edit/Delete
Assign the value of the modulus of elasticity: main menu: Preprocessor S Material Props S Material Models S Structural S Linear S Elastic S Isotropic Note: Doubleclick on Structural and then on Linear, Elastic, and Isotropic.
M03_MOAV4303_04_GE_C03.INDD 184
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 185
Note: EX: Modulus of Elasticity PRXY: Poisson’s Ratio Poisson’s Ratio may be omitted for link elements.
Close the Define Material Model Behavior window.
M03_MOAV4303_04_GE_C03.INDD 185
27/11/14 9:38 AM
186 Chapter 3 Trusses
Save the input data: ANSYS Toolbar: SAVE_DB Set up the graphics area (i.e., workplane, zoom, etc.): utility menu: WorkPlane S WP Settings c
Toggle on the workplane by the following sequence: utility menu: Workplane S Display Working Plane Bring the workplane to view using the following sequence: utility menu: PlotCtrls S Pan, Zoom, Rotate c
M03_MOAV4303_04_GE_C03.INDD 186
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 187
Click on the small circle until you bring the workplane to view. You can also use the arrow buttons to move the workplane in a desired direction. Then, create nodes by picking points on the workplane: main menu: Preprocessor S Modeling S Create S Nodes S On Working Plane On the workplane, pick the location of joints (nodes) and apply them:
M03_MOAV4303_04_GE_C03.INDD 187
27/11/14 9:38 AM
188 Chapter 3 Trusses
[WP = 0,0] [WP = 36,0] [WP = 0,36] [WP = 36,36] [WP = 72,36]
OK
You may want to turn off the workplane now and turn on node numbering instead: utility menu: Workplane S Display Working Plane utility menu: PlotCtrls S Numbering c
M03_MOAV4303_04_GE_C03.INDD 188
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 189
You may want to list nodes at this point in order to check your work: utility menu: List S Nodes c
Close ANSYS Toolbar: SAVE_DB Define elements by picking nodes: main menu: Preprocessor S Modeling S Create S Elements S AutoNumbered S Thru Nodes
M03_MOAV4303_04_GE_C03.INDD 189
27/11/14 9:38 AM
190 Chapter 3 Trusses
[node 1 and then node 2]
[use the middle button anywhere in the ANSYS graphics window to apply]
[node 2 and then node 3]
[anywhere in the ANSYS graphics window]
[node 3 and then node 4]
[anywhere in the ANSYS graphics window]
[node 2 and then node 4]
[anywhere in the ANSYS graphics window]
[node 2 and then node 5]
[anywhere in the ANSYS graphics window]
[node 4 and then node 5]
[anywhere in the ANSYS graphics window]
M03_MOAV4303_04_GE_C03.INDD 190
OK
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 191
ANSYS Toolbar: SAVE_DB Apply boundary conditions and loads: main menu: Solution S Define Loads S Apply S Structural S Displacement S On Nodes
M03_MOAV4303_04_GE_C03.INDD 191
[node 1]
[node 3]
[anywhere in the ANSYS graphics window]
27/11/14 9:38 AM
192 Chapter 3 Trusses
main menu: Solution S Define Loads S Apply S Structural S Force/Moment S On Nodes
[node 4]
[node 5]
[anywhere in the ANSYS graphics window]
ANSYS Toolbar: SAVE_DB
M03_MOAV4303_04_GE_C03.INDD 192
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 193
Solve the problem: main menu: Solution S Solve S Current LS
Close (the solution is done!) window. Close (the /STAT Command) window. For the postprocessing phase, first plot the deformed shape: main menu: General Postproc S Plot Results S Deformed Shape
M03_MOAV4303_04_GE_C03.INDD 193
27/11/14 9:38 AM
194 Chapter 3 Trusses
main menu: General Postproc S List Results S Nodal Solution
Close
M03_MOAV4303_04_GE_C03.INDD 194
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 195
To review other results, such as axial forces and axial stresses, we must copy these results into element tables. These items are obtained using item label and sequence numbers, as given in the ANSYS elements manual. For truss elements, the values of internal forces and stresses, which ANSYS computes from the nodal displacement results, may be looked up and assigned to userdefined labels. For Example 3.1, we have assigned the internal force, as computed by ANSYS, in each member to a user defined label “Axforce.” However, note that ANSYS allows up to eight characters to define such labels. Similarly, the axial stress result for each member is assigned to the label “Axstress.” We now run the following sequence: main menu: General Postproc S Element Table S Define Table
M03_MOAV4303_04_GE_C03.INDD 195
27/11/14 9:38 AM
196 Chapter 3 Trusses
main menu: General Postproc S Element Table S Plot Element Table or main menu: General Postproc S Element Table S List Element Table
M03_MOAV4303_04_GE_C03.INDD 196
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 197
Close List reaction solutions: main menu: General Postproc S List Results S Reaction Solu
M03_MOAV4303_04_GE_C03.INDD 197
27/11/14 9:38 AM
198 Chapter 3 Trusses
Exit ANSYS and save everything, including element tables and reaction forces: ANSYS Toolbar: QUIT
If, for any reason, you need to modify a model, first launch ANSYS and then type the file name of the model in the Jobname entry field of the Launcher dialog box. Then press Run. From the File menu, choose Resume Jobname.DB. Now you have complete access to your model. You can plot nodes, elements, and so on to make certain that you have chosen the right problem. Example 3.3 Consider the threedimensional truss shown in the accompanying figure. We are interested in determining the deflection of joint 2 under the loading shown in the figure. The Cartesian coordinates of the joints with respect to the coordinate system shown in the figure are given in feet. All members are made from aluminum with a modulus of elasticity of E = 10.6 * 106 lb/in2 and a crosssectional area of 1.56 in2. Y UX = 0 UY = 0 4 (0, 6', 0) UX = 0 3 (0, 0, 3')
2 UX = 0 UY = 0 UZ = 0 Z
M03_MOAV4303_04_GE_C03.INDD 198
1 (0, 0, 3')
(6', 0, 0)
X
200 lb
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 199
To solve this problem using ANSYS, we employ the following steps: Enter the ANSYS program by using the Launcher. Type Truss3D (or a file name of your choice) in the Jobname entry field of the dialog box. Pick Run to start the GUI. Create a title for the problem. This title will appear on ANSYS display windows to provide a simple way of identifying the displays: utility menu: File S Change Title c
Define the element type and material properties: main menu: Preprocessor S Element Type S Add/Edit/Delete
M03_MOAV4303_04_GE_C03.INDD 199
27/11/14 9:38 AM
200 Chapter 3 Trusses
Assign the crosssectional area of the truss members: main menu: Preprocessor S Real Constant S Add/Edit/Delete
M03_MOAV4303_04_GE_C03.INDD 200
27/11/14 9:38 AM
Section 3.5 Examples Using ANSYS 201
Assign the value of the modulus of elasticity: main menu: Preprocessor S Material Props S Material Models S Structural S Linear S Elastic S Isotropic
Close the Define Material Model Behavior window. ANSYS Toolbar: SAVE_DB Create nodes in active coordinate system: main menu: Preprocessor S Modeling S Create S Nodes S In Active CS
M03_MOAV4303_04_GE_C03.INDD 201
27/11/14 9:38 AM
202 Chapter 3 Trusses
You may want to turn on node numbering: utility menu: PlotCtrls S Numbering c
M03_MOAV4303_04_GE_C03.INDD 202
27/11/14 9:39 AM
Section 3.5 Examples Using ANSYS 203
You may want to list nodes at this point in order to check your work: utility menu: List S Nodes c
M03_MOAV4303_04_GE_C03.INDD 203
27/11/14 9:39 AM
204 Chapter 3 Trusses
Close ANSYS Toolbar: SAVE_DB Define elements by picking nodes. But first set the view angle: utility menu: PlotCtrls S Pan, Zoom, Rotate c Select the oblique (Obliq) or isometric (Iso) viewing. main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes
[node 1 and then node 2]
[use the middle button anywhere in the ANSYS graphics window to apply]
[node 1 and then node 3]
[anywhere in the ANSYS graphics window]
[node 1 and then node 4]
[anywhere in the ANSYS graphics window]
M03_MOAV4303_04_GE_C03.INDD 204
27/11/14 9:39 AM
Section 3.5 Examples Using ANSYS 205
[node 2 and then node 3]
[anywhere in the ANSYS graphics window]
[node 2 and then node 4]
[anywhere in the ANSYS graphics window]
[node 3 and then node 4]
[anywhere in the ANSYS graphics window]
OK
ANSYS Toolbar: SAVE_DB Apply boundary conditions and loads: main menu: Solution S Define Loads S Apply S Structural S Displacement S On Nodes
M03_MOAV4303_04_GE_C03.INDD 205
[node 1]
[node 3]
[node 4]
[anywhere in the ANSYS graphics window]
27/11/14 9:39 AM
206 Chapter 3 Trusses
main menu: Solution S Define Loads S Apply S Structural S Displacement S On Nodes
[node 1]
[anywhere in the ANSYS graphics window]
M03_MOAV4303_04_GE_C03.INDD 206
27/11/14 9:39 AM
Section 3.5 Examples Using ANSYS 207
main menu: Solution S Define Loads S Apply S Structural S Displacement S On Nodes
[node 1]
[node 4]
[anywhere in the ANSYS graphics window]
main menu: Solution S Define Loads S Apply S Structural S Force/Moment S On Nodes
[node 2]
[anywhere in the ANSYS graphics window]
ANSYS Toolbar: SAVE_DB
M03_MOAV4303_04_GE_C03.INDD 207
27/11/14 9:39 AM
208 Chapter 3 Trusses
Solve the problem: main menu: Solution S Solve S Current LS OK Close (the solution is done!) window. Close (the /STAT Command) window. Now we run the postprocessing phase by listing nodal solutions (displacements): main menu: General Postproc S List Results S Nodal Solution
M03_MOAV4303_04_GE_C03.INDD 208
27/11/14 9:39 AM
Section 3.5 Examples Using ANSYS 209
To review other results, such as axial forces and axial stresses, we must copy these results into element tables. These items are obtained using item label and sequence numbers, as given in the ANSYS elements manual. So, we run the following sequence: main menu: General Postproc S Element Table S Define Table
M03_MOAV4303_04_GE_C03.INDD 209
27/11/14 9:39 AM
210 Chapter 3 Trusses
M03_MOAV4303_04_GE_C03.INDD 210
27/11/14 9:39 AM
Section 3.5 Examples Using ANSYS 211
main menu: General Postproc S Element Table S List Element Table
Close List reaction solutions: main menu: General Postproc S List Results S Reaction Solu
M03_MOAV4303_04_GE_C03.INDD 211
27/11/14 9:39 AM
212 Chapter 3 Trusses
Close Exit ANSYS and save everything, including element tables and reaction forces: ANSYS Toolbar: QUIT
M03_MOAV4303_04_GE_C03.INDD 212
27/11/14 9:39 AM
Section 3.6 Verification of Results 213
3.6 Verification of Results There are various ways to verify your findings. 1. Check the reaction forces. We can use the computed reaction forces and the external forces to check for statics equilibrium: Σ FX = 0 Σ FY = 0
and
Σ Mnode = 0
The reaction forces computed by ANSYS are F1X = 1500 lb; F1Y = 0; F3X = 1500 lb; and F3Y = 1000 lb. Using the freebody diagram shown in the accompanying figure and applying the static equilibrium equations, we have: + S Σ FX = 0 15001500 = 0 + c Σ FY = 0
1000500500 = 0
+ Σ Mnode1 = 0 (1500) (3)  (500) (3)  (500) (6) = 0 ⤿ 1000 lb 500 lb 3 ft (3)
500 lb 3 ft (6)
1500 lb 3
4 (2)
3 ft
(4)
5 (5)
(1)
1500 lb 1
2
Now consider the internal forces of Example 3.1 as computed by ANSYS, shown in Table 3.2. Table 3.2 Internal forces in each element as computed by ANSYS Element Number
M03_MOAV4303_04_GE_C03.INDD 213
Internal Forces (lb)
1
 1500
2
1414
3
500
4
 500
5
 707
6
500
27/11/14 9:39 AM
214 Chapter 3 Trusses
2. The sum of the forces at each node should be zero. Choose an arbitrary node and apply the equilibrium conditions. As an example, let us choose node 5. Using the freebody diagram shown in the accompanying figure, we have + S Σ FX = 0 500 + 707 cos 45 = 0 + c Σ FY = 0
500 + 707 sin 45 = 0 500 lb
500 lb
5 455 707 lb
3. Pass an arbitrary section through the truss. Another way of checking for the validity of your FEA findings is by arbitrarily cutting a section through the truss and applying the statics equilibrium conditions. For example, consider cutting a section through elements (1), (2), and (3), as shown in the accompanying figure. 500 lb
500 lb 3 ft (6)
(3) 500 lb 4
5
1414 lb (4) (2)
(5)
3 ft
1500 lb (1)
2
+ S Σ FX = 0 500 + 1500  1414 cos 45 = 0 + c Σ FY = 0 500  500 + 1414 cos 45 = 0 + Σ Mnode2 = 0 (500) (3) + (500) (3) = 0 ⤿ Again, the validity of the computed internal forces is verified. Moreover, it is important to realize that when you analyze statics problems, statics equilibrium conditions must always be satisfied.
M03_MOAV4303_04_GE_C03.INDD 214
27/11/14 9:39 AM
Chapter 3 Problems 215
Summary At this point you should 1. have a good understanding of the underlying assumptions in truss analysis. 2. understand the significance of using global and local coordinate systems in describing a problem. You should also have a clear understanding of their role in describing nodal displacements and how information presented with respect to each frame of reference is related through the transformation matrix. 3. know the difference between the elemental stiffness matrix and the global stiffness matrix and know how to assemble elemental stiffness matrices to obtain a truss’s global stiffness matrix. 4. know how to apply the boundary conditions and loads to a global matrix to obtain the nodal displacement solution. 5. know how to obtain internal forces and stresses in each member from displacement results. 6. have a good grasp of the basic concepts and commands of ANSYS. You should realize that a typical analysis using ANSYS involves the preprocessing phase, where you provide data such as geometry, materials, and element type to the program; the solution phase, where you apply boundary conditions, apply loads, and initiate a finite element solution; and the postprocessing phase, where you review the results of the analysis through graphics displays, tabular listings, or both. 7. know how to verify the results of your truss analysis. References ANSYS User’s Manual: Procedures, Vol. I, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Commands, Vol. II, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Elements, Vol. III, Swanson Analysis Systems, Inc. Beer, F. P., and Johnston, E. R., Vector Mechanics for Engineers: Statics, 5th ed., New York, McGrawHill, 1988. Segrlind, L., Applied Finite Element Analysis, 2nd ed., New York, John Wiley and Sons, 1984.
Problems 1. Starting with the transformation matrix, show that the inverse of the transformation matrix is its transpose. That is, show that
[T]1
M03_MOAV4303_04_GE_C03.INDD 215
cos u sin u = D 0 0
sin u cos u 0 0
0 0 cos u  sin u
0 0 T sin u cos u
27/11/14 9:39 AM
216 Chapter 3 Trusses
2. Starting with Eq. (3.14), 5F 6 = [T][K][T]1 5U6, and substituting for values of the [T], [K], [T]1, and 5 U6 matrices in Eq. (3.14), verify the elemental relationship FiX cos2 u F sin u cos u d iY t = kD FjX  cos2 u FjY  sin u cos u
 cos2 u  sin u cos u cos2 u sin u cos u
sin u cos u sin2 u  sin u cos u  sin2 u
 sin u cos u UiX 2  sin u U T d iY t sin u cos u UjX sin2 u UjY
3. The members of the truss shown in the accompanying figure have a crosssectional area of
2.5 in2 and are made of aluminum alloy (E = 10.0 * 106 lb/in2). Using hand calculations, determine the deflection of joint A, the stress in each member, and the reaction forces. Verify your results. A
4f
t
200 lb
605
605
4. The members of the truss shown in the accompanying figure have a crosssectional area of 10 cm2 and are made of steel (E = 200 GPa). Using hand calculations, determine the deflection of each joint, the stress in each member, and the reaction forces. Verify your results. 1m
1m
1
2
3
4 kN 1.25 m
155
4
155 4 kN
M03_MOAV4303_04_GE_C03.INDD 216
27/11/14 9:39 AM
Chapter 3 Problems 217
5. The members of the truss shown in the accompanying figure have a crosssectional area of
20 cm2 and are made of aluminum alloy (E = 70 GPa). Using hand calculations, determine the deflection of each joint, the stress in each member, and the reaction forces. Verify your results. 10,000 N
3
4
1m 1 2 1.5 m
6. The members of the truss shown in the accompanying figure have a crosssectional area of
3 in2 and are made of structural steel (E = 30.0 * 106 lb/in2). Using hand calculations, determine the deflection of each joint, the stress in each member, and the reaction forces. Verify your results. 4
5
t
8f
2
3
1 8 ft
8 ft 5000 lb
7. The members of the threedimensional truss shown in the accompanying figure have a crosssectional area of 1.5 in2 and are made of structural steel (E = 30.0 * 106 lb/in2). Using hand calculations, determine the deflection of joint A, the stress in each member, and the reaction forces. Verify your results.
M03_MOAV4303_04_GE_C03.INDD 217
27/11/14 9:39 AM
218 Chapter 3 Trusses Y 5000 lb A
10 ft
6 ft 6 ft
6 ft
X
6 ft Z
8. The members of the threedimensional truss shown in the accompanying figure have a cross
sectional area of 15 cm2 and are made of aluminum alloy (E = 70 GPa). Using hand calculations, determine the deflection of joint A, the stress in each member, and the reaction forces. Verify your results. Y
D
1.5 m
B 1.5 m
A
1.5 m X
C Z
M03_MOAV4303_04_GE_C03.INDD 218
2m
5000 N
27/11/14 9:39 AM
Chapter 3 Problems 219
9. Consider the power transmissionline tower shown in the accompanying figure. The members
have a crosssectional area of 10 in2 and a modulus of elasticity of E = 29 * 106 lb/in2. Using ANSYS, determine the deflection of each joint, the stress in each member, and the reaction forces at the base. Verify your results. 5 at 10 ft 11
13
12
15
14
16 10 ft
10
9 8
7
10 ft 1000 lb
1000 lb 5
6 10 ft
3
4
15 ft 1
2
15 ft
10. Consider the staircase truss shown in the accompanying figure. There are 14 steps, each with a
rise of 8 in and a run of 12 in. The members have a crosssectional area of 4 in2 and are made of steel with a modulus of elasticity of E = 29 * 106 lb/in2. Using ANSYS, determine the deflection of each joint, the stress in each member, and the reaction forces. Verify your results.
200 lb
200 lb 100 lb
100 lb 200 lb 8 in. 200 lb 12 in
M03_MOAV4303_04_GE_C03.INDD 219
200 lb
27/11/14 9:39 AM
220 Chapter 3 Trusses
11. The members of the roof truss shown in the accompanying figure have a crosssectional
area of approximately 25 in2 and are made of Douglasfir wood with a modulus of elasticity of E = 1.9 * 106 lb/in2. Using ANSYS, determine the deflection of each joint, the stresses in each member, and the reaction forces. Verify your results. Also, replace one of the fixed boundary conditions with rollers and obtain the stresses in each member. Discuss the difference in results. 500 lb 500 lb
500 lb
500 lb
500 lb 4.4 ft
8.7 ft 24 ft
12. The members of the floor truss shown in the accompanying figure have a crosssectional
area of approximately 21.5 in2 and are made of Douglasfir wood with a modulus of elasticity of E = 1.9 * 106 lb/in2. Using ANSYS, determine the deflection of each joint, the stresses in each member, and the reaction forces. Verify your results. Also, replace one of the fixed boundary conditions with rollers and solve the problem again to obtain the stresses in each member. Discuss the difference in results. 500 lb
500 lb
500 lb
500 lb
1 ft
30 ft
13. The threedimensional truss shown in the accompanying figure is made of aluminum alloy
(E = 10.9 * 106 psi) and is to support a load of 500 lb. The Cartesian coordinates of the joints with respect to the coordinate system shown in the figure are given in feet. The crosssectional area of each member is 2.246 in2. Using ANSYS, determine the deflection of each joint, the stress in each member, and the reaction forces. Knowing that the second moment of area is 4.090 in4, do you think that buckling is a concern for this truss? Verify your results.
M03_MOAV4303_04_GE_C03.INDD 220
27/11/14 9:39 AM
Chapter 3 Problems 221 Y
3 ft 7 60 605 500 lb
6 4
5 3
9 ft 3 ft
1
2
3 ft
Z
3 ft
4.5 ft X
14. The threedimensional truss shown in the accompanying figure is made of aluminum alloy
(E = 10.4 * 106 lb/in2) and is to support a sign weighing 1000 lb. The Cartesian coordinates of the joints with respect to the coordinate system shown in the figure are given in feet. The crosssectional area of each member is 3.14 in2. Using ANSYS, determine the deflection of joint E, the stresses in each member, and the reaction forces. Verify your results. D (0.5, 2.0, 2.0)
1000 lb
B (0, 0, 2) C (0.5, 2.0, 0) Z
E (1.0, 2.5, 0) Y A (0, 0, 0) X
M03_MOAV4303_04_GE_C03.INDD 221
27/11/14 9:39 AM
222 Chapter 3 Trusses
15. The threedimensional truss shown in the accompanying figure is made of steel
(E = 29 * 106 psi) and is to support the load shown in the figure. The Cartesian coordinates of the joints with respect to the system shown in the figure are given in feet. The crosssectional area of each member is 3.093 in2. Using ANSYS, determine the deflection of each joint, the stresses in each member, and the reaction forces. Verify your results. Y
D (0, 4, 0)
B (0, 0, –2) E (0, 0, 0)
b 0l
A (4, 0, 0)
15 C (0, 0, 2)
0
15
250 lb
Z
X
lb 250 lb
16. During a maintenance process on the threedimensional truss in Problem 15, the AB member
is replaced with a member with the following properties: E = 28 * 106 psi and A = 2.246 in2. Using ANSYS, determine the deflection of each joint and the stresses in each member. Hint: you may need to ask your instructor for some help with this problem or you may want to study Example 6.2 (revisited) on your own to learn about how to assign different attributes to an element in ANSYS. 17. During a maintenance process on the threedimensional truss in Problem 13, members 4–5, 4–6, and 5–6 are replaced with steel members with the following properties: E = 29 * 106 psi and A = 1.25 in2. Member 1–5 is also replaced with a steel member with a crosssectional area of 1.35 in2. Using ANSYS, determine the deflection of each joint and the stresses in each member. See the hint given for Problem 16.
18. Derive the transformation matrix for an arbitrary member of a space truss, shown in the accompanying figure. The directional cosines, in terms of the difference between the coordinates of nodes j and i of a member and its length, are cos uX =
Xj  Xi L
;
cos uY =
Yj  Yi L
;
cos uZ =
Zj  Zi L
where L is the length of the member and is L = 2(Xj  Xi)2 + (Yj  Yi)2 + (Zj  Zi)2
M03_MOAV4303_04_GE_C03.INDD 222
27/11/14 9:39 AM
Chapter 3 Problems 223 j Z
Y
wY wZ
wX i
X
19. The threedimensional truss shown in the accompanying figure is made of steel
(E = 29 * 106 psi) and is to support the load shown in the figure. Dimensions are given in feet. The crosssectional area of each member is 3.25 in2. Using ANSYS, determine the deflection of each joint, the stresses in each member, and the reaction forces. Verify your results. Y
3
2 2 ft 3 ft
2 ft
7 4
2 ft 6 ft
M03_MOAV4303_04_GE_C03.INDD 223
5
2 ft 1
Z
6
b 0l
100 lb
X
10
300 lb
27/11/14 9:39 AM
224 Chapter 3 Trusses
20. Design Problem Size the cross section of each member for the outdoor truss structure shown in the accompanying figure so that the end deflection of the truss is kept under 1 in. Select appropriate material and discuss how you arrived at your final design. Y
2 ft 3 ft Z
M03_MOAV4303_04_GE_C03.INDD 224
250 lb
3 ft 3 ft
250 lb
2 ft
X
27/11/14 9:39 AM
C h a p t e r
4
Axial Members, Beams, and Frames
The objective of this chapter is to introduce you to the analysis of members under axial loading, beams, and frames. Structural members and machine components are generally subject to a push–pull, bending, or twisting type of loading. We will discuss twisting or torsion of structural members and plane stress formulation of machine components in Chapter 10. The main topics discussed in this chapter include the following: 4.1 Members Under Axial Loading 4.2 Beams 4.3 Finite Element Formulation of Beams 4.4 Finite Element Formulation of Frames 4.5 ThreeDimensional Beam Element 4.6 An Example Using ANSYS 4.7 Verification of Results
4.1 Members Under Axial Loading In this section, we use the minimum total potential energy formulation to generate finite element models for members under axial loading. However, before we proceed with finite element formulation of axial members, we should define what we mean by an axial element and corresponding shape functions and their properties. A Linear Element The structural example in this section is employed to introduce the basic ideas of onedimensional element and shape functions. Steel columns are commonly used to support loads from various floors of multistory buildings, as shown in Figure 4.1. The column shown in the figure may be divided into four elements and five nodes to generate a
M04_MOAV4303_04_GE_C04.INDD 225
27/11/14 9:42 AM
226 Chapter 4 Axial Members, Beams, and Frames
Load
Y
Load
Actual deflection profile
5 (4) 4 (3) 3 Approximate deflection profile
(2) 2 (1) 1 Deflection value
u
Figure 4.1 Deflection of a steel column subject to floor loading.
finite element model. The loading from the floors causes vertical displacements of various points along the column. Assuming axial central loading, we may approximate the actual deflection of the column by using a series of linear functions, describing the deflection over each element or each section of the column. Note that the deflection profile u represents the vertical (not the lateral) displacement of the column at various points along the column. The profile is merely plotted as a function of Y. We have modeled the example problem shown in Figure 4.1 by five nodes and four elements. Let us focus our attention on a typical element, as shown in Figure 4.2. Y
Yj
Yi
ui
M04_MOAV4303_04_GE_C04.INDD 226
uj
u
Figure 4.2 Linear approximation of deflection variation for an element.
27/11/14 9:42 AM
Section 4.1 Members Under Axial Loading 227
The linear deflection distribution for a typical element may be expressed as u(e) = c1 + c2Y (4.1)
In order to solve for the unknown coefficients c1 and c2, we make use of the element’s end deflection values which are given by the nodal deflections ui and uj, according to the conditions
at at
u = ui u = uj
Y = Yi (4.2) Y = Yj
Substitution of nodal values into Eq. (4.1) results in two equations and two unknowns: ui = c1 + c2Yi uj = c1 + c2Yj (4.3)
Solving for the unknowns c1 and c2, we get
c1 =
c2 =
uiYj  ujYi Yj  Yi u j  ui
(4.4) (4.5)
Yj  Yi
The element’s deflection distribution in terms of its nodal values is uiYj  ujYi
u(e) =
Yj  Yi
+
u j  ui Yj  Yi
Y (4.6)
Grouping the ui terms together and the uj terms together, Eq. (4.6) becomes u(e) = ¢
Yj  Y Yj  Yi
≤ui + ¢
Y  Yi ≤u (4.7) Yj  Yi j
We now define the shape functions, Si and Sj using the terms in parentheses appearing before ui and uj, according to the equations
Si =
Sj =
Yj  Y Yj  Yi
=
Yj  Y /
(4.8)
Y  Yi Y  Yi = (4.9) Yj  Yi /
where / is the length of the element. Thus, the deflection for an element in terms of the shape functions and the nodal deflection values can be written as u(e) = Siui + Sjuj (4.10)
Equation (4.10) can also be expressed in matrix form as
M04_MOAV4303_04_GE_C04.INDD 227
u(e) = [Si
Sj]b
ui r (4.11) uj
27/11/14 9:42 AM
228 Chapter 4 Axial Members, Beams, and Frames
It will become clear to you that we can use the same approach to approximate the spatial variation of any unknown variable, such as temperature or velocity, in the same manner. We will discuss the concept of onedimensional elements and their properties in more detail in Chapter 5. As we discussed in Chapter 3, in finite element modeling, it is convenient to use two frames of references: (1) a global coordinate system to represent the location of each node, orientation of each element, and to apply boundary conditions and loads. The nodal solutions of finite element models are generally expressed with respect to global coordinates as well. On the other hand, we employ (2) a local coordinate system to take advantage of the local characteristics of the system behavior. For the onedimensional element shown in Figure 4.2, the relationship between a global coordinate Y and a local coordinate y is given by Y = Yi + y. This relationship is shown in Figure 4.3. Substituting for Y in terms of the local coordinate y in Eqs. (4.8) and (4.9), we get
Si =
Sj =
Yj  Y /
=
Yj  (Yi + y) /
= 1 
(Yi + y)  Yi y Y  Yi = = / / /
y (4.12) / (4.13)
where the local coordinate y varies from 0 to /; that is, 0 … y … /. This is a good place to say a few words about the shape functions Si and Sj. They possess unique properties that, once understood, can simplify the derivation of stiffness matrices. We now refer to Eqs. (4.12) and (4.13) and note that Si and Sj each has a value of unity at its corresponding node and zero at the other adjacent node. For example, if we evaluate Si at node i by substituting y = 0 in Eq. (4.12), we find that Si = 1. Similarly, we can show that the value of Sj at node j(y = /) is also 1. The value of the shape function Si, Eq. (4.12), at the adjacent node j(y = /) and Sj, Eq. (4.13), at its adjacent node i(y = 0) are zero. We discuss the properties of shape functions in more detail in Chapter 5.
Y
Yj
j
y Yi
i Figure 4.3 The relationship between a global coordinate Y and a local coordinate y.
M04_MOAV4303_04_GE_C04.INDD 228
27/11/14 9:42 AM
Section 4.1 Members Under Axial Loading 229
Example 4.1 Consider a fourstory building with steel columns. One column is subjected to the loading shown in Figure 4.4. Under axial loading assumption and using linear elements, the vertical displacements of the column at various floorcolumn connection points were determined to be u1 0 u2 0.03283 e u3 u =  e 0.05784 uin u4 0.07504 u5 0.08442 The modulus of elasticity of E = 29 * 106 lb/in2 and area of A = 39.7 in2 were used in the calculations. A detailed analysis of this problem is given in the next section. For now, given the nodal displacement values, we are interested in determining the deflections of points A and B. a. Using the global coordinate Y, the displacement of element (1): Y2  Y Y (1) u(1) = S (1) u1 + 1 u1 + S 2 u2 = / 15  10 10  0 u = (0) + ( 0.03283) 15 15 30,000 lb
point A is represented by  Y1 u2 / = 0.02188 in
30,000 lb 5
25,000 lb
25,000 lb
(4) B 4
(3)
8 ft 25,000 lb
15 ft
25,000 lb
15 ft
25,000 ft
15 ft
3
25,000 lb
(2) 2
Y
10 ft
A (1)
15 ft
1 Figure 4.4 The column in Example 4.1.
M04_MOAV4303_04_GE_C04.INDD 229
27/11/14 9:42 AM
230 Chapter 4 Axial Members, Beams, and Frames
b. The displacement of point B is represented by element (4): (4) u(4) = S(4) 4 u4 + S 5 u5 =
u =
Y5  Y Y  Y4 u4 + u5 / /
60  52 52  45 ( 0.07504) + ( 0.08442) = 0.07941 in 15 15
Stiffness and Load Matrices In this section, we use the minimum total potential energy formulation to g enerate the stiffness and load matrices for members under axial loading. Previously, we showed that under axial loading, we can approximate the exact deflection of the column shown in Figure 4.1 by a series of linear functions. Moreover, as discussed in Section 1.6, applied external loads cause a body to deform. During the deformation, the work done by the external forces is stored in the material in the form of elastic energy, called strain energy. For a member (element) under axial loading, the strain energy Λ(e) is given by Λ(e) =
se Ee2 dV = dV (4.14) LV 2 LV 2
The total potential energy Π for a body consisting of n elements and m nodes is the difference between the total strain energy and the work done by the external forces: Π = a Λ(e)  a Fiui (4.15)
n
m
e=1
i=1
The minimum total potential energy principle states that for a stable system, the displacement at the equilibrium position occurs such that the value of the system’s total potential energy is a minimum. That is, 0Π 0 n (e) 0 m = Λ Fiui = 0 for i = 1, 2, 3,c, m (4.16) 0ui 0ui ea 0ui ia =1 =1
where i takes on different values of node numbers. Recall that the deflection for an arbitrary element with nodes i and j in terms of local shape functions is given by u(e) = Siui + Sjuj (4.17)
y
y
where Si = 1  / and Sj = / and y is the element’s local coordinate, with its origin at node i. The strain in each member can be computed using the relation e = du dy as
e =
M04_MOAV4303_04_GE_C04.INDD 230
ui + uj y y du d d = [Siui + Sjuj] = J ¢ 1  ≤ui + uj R = (4.18) dy dy dy / / /
27/11/14 9:42 AM
Section 4.1 Members Under Axial Loading 231
Incorporating Eq. (4.18) into Eq. (4.14) yields the strain energy for an arbitrary element (e): Λ(e) =
Ee2 AE 2 dV = (u + u2i  2uj ui) (4.19) 2/ j LV 2
Minimizing the strain energy with respect to ui and uj leads to
0Λ(e) AE = (ui  uj) 0ui /
0Λ(e) AE = (uj  ui) (4.20) 0uj / or, in matrix form, 0Λ(e) 0ui AE 1 d (e) t = J 0Λ / 1 0uj
1 u k R b ir = J 1 uj k
k u R b i r (4.21) k uj
(AE) . Minimizing the work done by external forces, the second term on the / righthand side of Eq. (4.16) results in the load matrix
where k =
5 F6 (e) = b
Fi r (4.22) Fj
Computing individual elemental stiffness and load matrices and connecting them leads to global stiffness and load matrices. This step is demonstrated by the next example. Example 4.2 A Column Problem Consider a fourstory building with steel columns. One column is subjected to the loading shown in Figure 4.5. Assuming axial loading, determine (a) vertical displacements of the column at various floorcolumn connection points and (b) the stresses in each portion of the column. E = 29 * 106 lb/in2, A = 39.7 in2. Because all elements have the same length, crosssectional area, and physical properties, the elemental stiffness for elements (1),(2),(3), and (4) is given by [K](e) =
AE 1 J / 1
1 39.7 * 29 * 106 1 R = J 1 15 * 12 1
1 1 R = 6.396 * 106 J 1 1
[K](1) = [K](2) = [K](3) = [K](4) = 6.396 * 106 J
M04_MOAV4303_04_GE_C04.INDD 231
1 1
1 R 1
1 lb R 1 in
27/11/14 9:42 AM
232 Chapter 4 Axial Members, Beams, and Frames 30,000 lb 5
25,000 lb
(4)
15 ft
4
25,000 lb
(3)
15 ft
3
25,000 lb
(2)
15 ft
2
Y
(1)
15 ft
1 Figure 4.5 A schematic of the column in Example 4.2.
The global stiffness matrix is obtained by assembling the elemental matrices:
[K](G)
1 1 = 6.396 * 106 E 0 0 0
1 1 + 1 1 0 0
0 1 1 + 1 1 0
0 0 1 1 + 1 1
0 0 0 U 1 1
The global load matrix is obtained from F1 0 F2 50000 0Fiui 5 F6(G) = b r = e F3 u =  e 50000 u lb 0ui i = 1,5 F4 50000 F5 60000 Note all applied forces act in the negative Y direction. Application of the boundary condition, u1 = 0, and loads results in 1 1 6.396 * 106 E 0 0 0
M04_MOAV4303_04_GE_C04.INDD 232
0 2 1 0 0
0 1 2 1 0
0 0 1 2 1
0 u1 0 0 u2 50000 0 U e u3 u =  e 50000 u 1 u4 50000 1 u5 60000
27/11/14 9:42 AM
Section 4.2 Beams 233
Solving for displacements, we have u1 0 u2 0.03283 e u3 u =  e 0.05784 u in u4 0.07504 u5 0.08442 The axial stresses in each element are determined from s(1) =
E(uj  ui) /
=
29 * 106( 0.03283  0) = 5289 lb/in2 15 * 12
s(2) =
29 * 106(0.05784  (0.03283)) = 4029 lb/in2 15 * 12
s(3) =
29 * 106(0.07504  ( 0.05784)) = 2771 lb/in2 15 * 12
s(4) =
29 * 106(0.08442  ( 0.07504)) = 1511 lb/in2 15 * 12
4.2 Beams Beams play significant roles in many engineering applications, including buildings, bridges, automobiles, and airplane structures. A beam is defined as a structural member whose crosssectional dimensions are relatively smaller than its length. Beams are commonly subjected to transverse loading, which is a type of loading that creates bending in the beam. A beam subjected to a distributed load is shown in Figure 4.6. In the previous chapter we defined trusses as structures consisting of twoforce members. Moreover, recall that when using a truss model to analyze a physical problem, all loads are assumed to apply at the joints or the nodes of the truss. Therefore, no bending of the members are allowed. Note that for a structural member that is considered as a beam, loads may be applied anywhere along the beam and the loading will create bending in the beam. It is important to make these distinctions when modeling a physical problem. The deflection of the neutral axis of a beam at any location x is represented by the variable v. For small deflections, the relationship between the normal stress s at a y Load y x
Neutral axis
Figure 4.6 A beam subjected to a distributed load.
M04_MOAV4303_04_GE_C04.INDD 233
27/11/14 9:42 AM
234 Chapter 4 Axial Members, Beams, and Frames y
M
M
y
x
x
M
M Positive bending and positive curvature
Negative moment and negative curvature
Figure 4.7 The positive and negative bending moments and curvature sign convention.
section, the bending moment at that section M, and the second moment of area I is given by the flexure formula. The flexure formula is the equation s = 
My (4.23) I
where y locates a point in the cross section of the beam and represents the lateral distance from the neutral axis to that point. The deflection of the neutral axis v is also related to the internal bending moment M(x), the transverse shear V(x), and the load w(x) according to the equations
EI
d 2v = M(x) dx2
EI
dM(x) d3 v = = V(x) (4.25) 3 dx dx
EI
dV(x) d4 v = = w(x) (4.26) 4 dx dx
(4.24)
Note that the standard beam sign convention is assumed in the previous equations. The positive and negative bending moments and curvatures are shown in Figure 4.7. For your reference, the deflections and slopes of beams under some typical loads for simply supported and cantilevered supports are summarized in Table 4.1. If you come across problems that can be analyzed using equations (4.24), (4.25), and (4.26) and Table 4.1, solve them as such.
Example 4.3 The cantilevered balcony beam shown in the accompanying figure is a wideflange W18 * 35, with a crosssectional area of 10.3 in2 and a depth of 17. 7 in. The second moment of area is 510 in4. The beam is subjected to a uniformly distributed load of 1000 lb/ft. The modulus of elasticity of the beam is E = 29 * 106 lb/in2. Using the review materials
M04_MOAV4303_04_GE_C04.INDD 234
27/11/14 9:42 AM
235
M04_MOAV4303_04_GE_C04.INDD 235
27/11/14 9:42 AM
y
y
y
w0
L
L
L
w
P
Beam Support and Load
x
x
x
v =
v =
v =
(  x3 + 5Lx2  10L2x + 10L3)
 Px2 (3L  x) 6EI
120LEI
 w0x2
 wx2 2 (x  4Lx + 6L2) 24EI
Equation of Elastic Curve
Table 4.1 Deflections and slopes of beams under some typical loads and supports
vmax =
vmax =
vmax =
 PL3 3EI
30EI
 w0 L4
 wL4 8EI
Maximum Deflection
umax =
umax =
umax =
Slope
 PL2 2EI
24EI
 w0 L3
 wL3 6EI
236
M04_MOAV4303_04_GE_C04.INDD 236
27/11/14 9:42 AM
y
y
L — 2
P
L
w
L — 2
Beam Support and Load
x
x
v =
v =
 Px L (3L2  4x2) for ¢ x … ≤ 48EI 2
 wx 3 (x  2Lx2 + L3) 24EI
Equation of Elastic Curve
Table 4.1 (continued) Deflections and slopes of beams under some typical loads and supports
vmax =
vmax =
 PL3 48EI
 5wL4 384EI
Maximum Deflection
umax =
umax =
Slope
 PL2 16EI
 wL3 24EI
Section 4.2 Beams 237
presented in this section, we will determine the deflection of the beam at the midpoint B and the endpoint C. We will also compute the slope of the beam at point C. 1000 lb/ft
A
B
C
10 ft
The deflection equation for a cantilever beam is given in Table 4.1. wx2 2 (x  4Lx + 6L2) 24EI
v =
L is 2
The deflection of the beam at midpoint corresponding to x = vB = =
wx2 2 (x  4Lx + 6L2) 24EI (1000 lb/ft) (5 ft)2 6
2
4
24(29 * 10 lb/in ) (510 in )
((52  4(10) (5) + 6(10)2)ft2) ¢
12 in 3 ≤ = 0.052 in 1 ft
And the deflection of point C is
vc =
wL4 = 8EI
(1000 lb/ft) (10 ft)4 ¢
12 in 3 ≤ 1 ft
8(29 * 106 lb/in2) (510 in4)
= 0.146 in
The maximum slope occurs at point C. umax =
wL3 = 6EI
(1000 lb/ft) (10 ft)3 1 ft 2 6(29 * 10 lb/in ) (510 in ) ¢ ≤ 12 in 6
2
= 0.00163 rad
4
Let us also calculate the maximum bending stress in the beam. Because the maximum bending moment occurs at point A, the maximum bending stress in the beam will occur at point A. The resulting maximum bending stress at outer fiber of the beam at A is y 2
s =
M04_MOAV4303_04_GE_C04.INDD 237
My = I
M 7 12 in 17.7 (1000 lb/ft)(10 ft)(5 ft) ¢ ≤¢ in ≤ 1 ft 2
510 in4
= 10411 lb/in2
27/11/14 9:42 AM
238 Chapter 4 Axial Members, Beams, and Frames
4.3 Finite Element Formulation of Beams Before we proceed with finite element formulation of beams, we should define what we mean by a beam element. A simple beam element consists of two nodes. At each node, there are two degrees of freedom, a vertical displacement, and a rotation angle (slope), as shown in Figure 4.8. There are four nodal values associated with a beam element. Therefore, we will use a thirdorder polynomial with four unknown coefficients to represent the displacement field. Moreover, we want the first derivatives of the shape functions to be continuous. The resulting shape functions are commonly referred to as Hermite shape functions. As you will see, they differ in some ways from the linear shape functions you have already studied. We start with the thirdorder polynomial v = c1 + c2x + c3x2 + c4x3 (4.27)
The element’s end conditions are given by the following nodal values: For node i: The vertical displacement at x = 0 v = c1 = Ui1 dv 2 For node i: The slope at x = 0 = c2 = Ui2 dx x = 0 For node j: The vertical displacement at x = L v = c1 + c2L + c3L2 + c4L3 = Uj1 dv 2 For node j: The slope at x = L = c2 + 2c3L + 3c4L2 = Uj2 dx x = L We now have four equations with four unknowns. Solving for c1, c2, c3, and c4; substituting into Eq. (4.27); and regrouping the Ui1, Ui2, Uj1, Uj2 terms results in the equation v = Si1Ui1 + Si2Ui2 + Sj1Uj1 + Sj2Uj2 (4.28)
where the shape functions are given by
Si1 = 1 
3x2 2x3 + 3 (4.29) 2 L L
Si2 = x 
2x2 x3 + 2 (4.30) L L
Ui1
Uj1
y Y
Uj2
Ui2 x
X
i
j L
Figure 4.8 A beam element.
M04_MOAV4303_04_GE_C04.INDD 238
27/11/14 9:42 AM
Section 4.3 Finite Element Formulation of Beams 239
3x2 2x3  3 (4.31) 2 L L
Sj1 =
Sj2 = 
x2 x3 + 2 (4.32) L L
It is clear that if we evaluate the shape functions, as given in Eqs. (4.29 through 4.32), at node i at x = 0, we find that Si1 = 1 and Si2 = Sj1 = Sj2 = 0. Also, if we evaluate the slopes of the shape functions at x = 0, we find that
dSj1 dSj2 dSi2 dSi1 = 1 and = = = 0. dx dx dx dx
If we evaluate the shape functions at node j at x = L, we find that Sj1 = 1 and Si1 = Si2 = Sj2 = 0, and if we evaluate the slopes of the shape functions at x = L, dSj1 dSi1 dSi2 = = = 0. These values are the properdx dx dx dx ties of the Hermite thirdorder polynomials. Now that you know what we mean by a beam element, we proceed with derivation of stiffness matrix. In the following derivation, we neglect the contribution of shear stresses to the strain energy. The strain energy for an arbitrary beam element (e) then becomes we determine that
dSj2
= 1 and
Λ(e) =
se Ee2 E d2v 2 (4.33) dV = dV = ¢ y ≤ dV LV 2 LV 2 2 LV dx2
Λ(e) =
E d2 v 2 E d2 v 2 ¢ y 2 ≤ dV = ¢ 2 ≤ dx y2 dA (4.34) 2L 2 L0 dx dx LA V
L
Recognizing the integral
L A
y2dA as the second moment of the area I, we have L
Λ(e) =
EI d2 v 2 ¢ 2 ≤ dx (4.35) 2 L dx 0
Next, we substitute for the displacement field v in terms of the shape functions and the nodal values. Let us begin by evaluating the equation
M04_MOAV4303_04_GE_C04.INDD 239
Ui1 d2v d2 U = [Si1 Si2 Sj1 Sj2] d i2 t (4.36) 2 2 Uj1 dx dx Uj2
27/11/14 9:42 AM
240 Chapter 4 Axial Members, Beams, and Frames
To simplify the next few steps of derivation and to avoid unnecessary mathematical operations, let us make use of matrix notations. First, let the second derivatives of the shape functions be defined in terms of the following relationships:
Di1 =
Di2 =
d2Si1 6 12x =  2 + 3 dx2 L L
d2Si2 4 6x =  + 2 2 L dx L d2Sj1 6 12x = 2  3 Dj1 = dx2 L L 2 d Sj2 2 6x Dj2 = =  + 2 2 L dx L
(4.36a)
Then, Eq. (4.36) takes on the compactmatrix form of d2 v = [D]5 U6 (4.37) dx2
where [D] = [Di1 Di 2 Dj1
The ¢
d2 v 2 ≤ term can be represented in terms of the 5 U6 and [D] matrices as dx2 ¢
Ui1 U Dj 2] and 5 U6 = µ i 2 ∂ . Uj1 Uj 2
d2 v 2 ≤ = ([D] 5 U6)([D] 5 U6) = 5 U6 T[D]T[D]5 U6 (4.38) dx2
In Eq. (4.38), note that [D]5 U6 = 5 U6 T[D]T. The proof of this identity is left as an exercise for you to complete. See Problem 26 at the end of this chapter. Thus, using Eq. (4.38) the strain energy for an arbitrary beam element is L
Λ(e) =
EI 5 U6 T[D]T[D]5 U6 dx (4.39) 2 L0
Recall that the total potential energy Π for a body is the difference between the total strain energy and the work done by the external forces:
Π = ΣΛ(e)  ΣFU (4.40)
Also recall that the minimum total potential energy principle states that for a stable system, the displacement at the equilibrium position occurs such that the value of the system’s total potential energy is a minimum. Thus, for a beam element, we have 0Π 0 0 = ΣΛ(e) ΣFU = 0 for k = 1, 2, 3, 4 (4.41) 0Uk 0Uk 0Uk
where Uk takes on the values of the nodal degrees of freedom Ui1, Ui2, Uj1, and Uj2. Equation (4.40) has two main parts: the strain energy, and the work done by external forces. Differentiation of the strain energy with respect to the nodal degrees of freedom
M04_MOAV4303_04_GE_C04.INDD 240
27/11/14 9:42 AM
Section 4.3 Finite Element Formulation of Beams 241
leads to the formulation of the beam’s stiffness matrix and differentiation of the work done by external forces results in the load matrix. We begin minimizing the strain energy with respect to Ui1, Ui2, Uj1, and Uj2 to obtain the stiffness matrix. Starting with the strain energy part of the total potential energy, we get L
0Λ(e) = EI [D]T[D]dx5 U6 (4.42) 0Uk L0
Evaluating Eq. (4.42) leads to the expression (e)
0Λ 0Uk
12 EI 6L = EI [D]T[D]dx5 U6 = 3 D L 12 L0 6L L
6L 4L2 6L 2L2
12 6L 12 6L
6L Ui1 2 2L U T d i2 t 6L Uj1 2 4L Uj2
The stiffness matrix for a beam element with two degrees of freedom at each node—the vertical displacement and rotation—is
[K](e)
12 EI 6L = 3D L 12 6L
6L 4L2 6L 2L2
12 6L 12 6L
6L 2L2 T (4.43) 6L 4L2
Starting with Eqs. (4.39) and (4.41), proof of steps leading to Eqs. (4.42) and (4.43) is left as an exercise for you to perform. See Problem 27. Load Matrix There are two ways in which we can formulate the nodal load matrices: (1) by minimizing the work done by the load as stated above, and (2) alternatively by computing the beam’s reaction forces. Consider a uniformly distributed load acting on a beam of length L, as shown in Figure 4.9. The reaction forces and moments at the endpoints are also shown in the figure. Using the first approach, we can compute the work done by this type of loading from 1L wv dx. The next step involves substituting for the displacement function in terms of the shape functions and nodal values, and then integrating and differentiating the work term with respect to the nodal displacements. This approach will be w
L
M1
M2
R1
M04_MOAV4303_04_GE_C04.INDD 241
R2
Figure 4.9 A beam element subjected to a uniform distributed load.
27/11/14 9:42 AM
242 Chapter 4 Axial Members, Beams, and Frames
demonstrated in detail when we formulate the load matrix for a plane stress situation. In order to expose you to as many finite element formulations as possible, let us develop the load matrix using the alternate approach, starting with Eq. (4.26): EI
dV(x) d4 v = = w(x) 4 dx dx
For a uniformly distributed load, w(x) is constant. Integrating this equation, we get
EI
d3 v = wx + c1 (4.44) dx3
Applying the boundary condition (at x = 0, V(x) = R1, and using Eq. (4.25)) dv EI 3 ` = R1, we find that c1 = R1. Substituting for the value of c1 and integrating dx x = 0 Eq. (4.44) we obtain 3
EI
d2 v wx2 = + R1x + c2 (4.45) 2 2 dx
Applying the boundary condition (at x = 0, M(x) = M1, and using Eq. (4.24)) d2 v ` = M1, we find that c2 = M1. Substituting for the value of c2 and integratdx2 x = 0 ing, we obtain EI
EI
R1x2 dv wx3 = +  M1x + c3 (4.46) dx 6 2
Applying the boundary condition (zero slope at x = 0) dv dx 0 x = 0 = 0, we find that c3 = 0. Integrating one last time, we have EIv = 
R1x3 M1x2 wx4 + + c4 (4.47) 24 6 2
Applying the boundary condition (zero deflection at x = 0) v(0) = 0, we determine that c4 = 0. To obtain the values of R1 and M1, we can apply two additional boundary conditions to this problem: dv dx 0 x = L = 0 and v(L) = 0. Applying these conditions, we get R1L2 dv 2 wL3 = +  M1L = 0 (4.48) dx x = L 6 2 v(L) = 
R1L3 M 1L2 wL4 + = 0 (4.49) 24 6 2
wL wL2 and M1 = . From the 2 12 symmetry of the problem—that is, applying the statics equilibrium conditions—we find wL wL2 that the reactions at the other end of the beam are R2 = and M2 = . All of the 2 12 reactions are shown in Figure 4.10. Solving these equations simultaneously, we get R1 =
M04_MOAV4303_04_GE_C04.INDD 242
27/11/14 9:43 AM
Section 4.3 Finite Element Formulation of Beams 243
M1 =
wL2 12
R1 =
M2 =
wL 2
wL2 12
R2 =
wL 2
Figure 4.10 Reaction results for a beam subjected to a uniformly distributed load.
If we reverse the signs of the reactions at the endpoints, we can now represent the effect of a uniformly distributed load in terms of its equivalent nodal loads. Similarly, we can obtain the nodal load matrices for other loading situations. The relationships between the actual load and its equivalent nodal loads for some typical loading situations are summarized in Table 4.2. Table 4.2 Equivalent nodal loading of beams Loading
Equivalent Nodal Loading
w
L
w
L P
L — 2
L — 2
wL 2
wL 2
wL2 12
wL2 12
3wL 20
7wL 20
wL2 30
wL2 20
P — 2
P — 2
M= PL 8
M= PL 8
Example 4.3 Revisited Let us consider the cantilevered balcony beam of Example 4.3 again and solve it using a single beam element. Recall that the beam is a wideflange W18 * 35, with a cross sectional area of 10.3 in2 and a depth of 17.7 in. The second moment of area is 510 in4. The beam is subjected to a uniformly distributed load of 1000 lb/ft. The modulus of
M04_MOAV4303_04_GE_C04.INDD 243
27/11/14 9:43 AM
244 Chapter 4 Axial Members, Beams, and Frames
elasticity of the beam E = 29 * 106 lb/in2. We are interested in determining the deflection of the beam at the midpoint B and the endpoint C. Also, we will compute the maximum slope that will occur at point C. 1000 lb/ft
A
B
C
10 ft
Because we are using a single element to model this problem, the elemental stiffness and load matrices are the same as the global matrices. wL 2 wL2 12 = h x wL 2 wL2 12 
[K](e) = [K](G)
12 EI 6L = 3D L 12 6L
6L 4L2 6L 2L2
12 6L 12 6L
6L 2L2 T 6L 4L2
5 F6 (e) = 5 F6 (G)
wL 2 6L U11 wL2 2L2 U12 12 Td t = h x 6L U21 wL 4L2 U22 2 wL2 12 
12 EI 6L D L3 12 6L
6L 4L2 6L 2L2
12 6L 12 6L
Applying the boundary conditions U11 = 0 and U12 = 0 at node 1, we have 1 EI 0 D L3 12 6L
M04_MOAV4303_04_GE_C04.INDD 244
0 1 6L 2L2
0 0 12 6L
0 0 U11 0 0 U wL T d 12 t = f v 6L U21 2 4L2 U22 wL2 12
27/11/14 9:43 AM
Section 4.3 Finite Element Formulation of Beams 245
And simplifying, we get 12 J 6L
J
12 6(10 ft)
wL 6L U21 L3 2 r = d t 2 Rb 4L U22 EI wL2 12
1000(10) (10 ft) 6(10 ft) U 2 R b 21 r = d t 2 (1000) (10)2 4(10 ft)2 U22 1 ft (29 * 106 lb/in2) (510 in4) ¢ ≤ 12 12 in. 3

The deflection and the slope at endpoint C is U21 = 0.01217 ft = 0.146 in and U22 = 0.00163 rad To determine the deflection at point B, we use the deflection equation for the beam L element and evaluate the shape functions at x = . 2 v = S11U11 + S12U12 + S21U21 + S22U22 = S11(0) + S12(0) + S21( 0.146) + S22( 0.00163) Computing the values of the shape functions at point B, we have S21 =
3x2 2x3 3 L 2 2 L 3 1  3 = 2¢ ≤  3¢ ≤ = 2 2 L L L 2 L 2
S22 = 
2
3
x x + 2 = L L
¢
L 2 ≤ 2 L
¢ +
L 3 ≤ 2 L
2
= 
L 8
1 120 in. vB = ¢ ≤ ( 0.146 in) + ¢ ≤ ( 0.00163 rad) = 0.048 in 2 8 Comparing results of our finite element model to the exact solutions given in Example 4.3, we note that they are in good agreement. We could improve our results for the midpoint deflection by using a model that uses two elements. We have left this as an exercise for you. EXAMPLE 4.4 The beam shown in Figure 4.11 is a wideflange W310 * 52 with a crosssectional area of 6650 mm2 and depth of 317 mm. The second moment of the area is 118.6 * 106 mm4. The beam is subjected to a uniformly distributed load of 25,000 N/m. The modulus of elasticity of the beam is E = 200 GPa. Determine the vertical displacement at node 3
M04_MOAV4303_04_GE_C04.INDD 245
27/11/14 9:43 AM
246 Chapter 4 Axial Members, Beams, and Frames 25,000 N/m
1
2
5m
3
Figure 4.11 A schematic of the beam in Example 4.4
2.5 m
and the rotations at nodes 2 and 3. Also, compute the reaction forces and moment at nodes 1 and 2. Note that this problem is statically indeterminate. We will use two elements to represent this problem. The stiffness matrices of the elements are computed from Eq. (4.43):
[K](e)
12 EI 6L = 3 D 12 L 6L
6L 4L2 6L 2L2
12 6L 12 6L
6L 2L2 T 6L 4L2
Substituting appropriate values for element (1), we have
[K](1) =
9
200 * 10 * 1.186 * 10 53
4
12 6(5) D 12 6(5)
6(5) 4(5)2 6(5) 2(5)2
12 6(5) 12 6(5)
6(5) 2(5)2 T 6(5) 4(5)2
For convenience, the nodal degrees of freedom are shown alongside the stiffness matrices. For element (1), we have
[K](1)
2277120 5692800 = D 2277120 5692800
5692800 18976000 5692800 9488000
2277120 5692800 2277120 5692800
5692800 U11 9488000 U12 T 5692800 U21 18976000 U22
Computing the stiffness matrix for element (2), we have
[K](2)
12 200 * 109 * 1.186 * 104 6(2.5) = D 12 (2.5)3 6(2.5)
6(2.5) 4(2.5)2 6(2.5) 2(2.5)2
12 6(2.5) 12 6(2.5)
6(2.5) 2(2.5)2 T 6(2.5) 4(2.5)2
Showing the nodal degrees of freedom alongside the stiffness matrix for element (2), we have
[K](2)
M04_MOAV4303_04_GE_C04.INDD 246
18216960 22771200 = D 18216960 22771200
22771200 37952000 22771200 18976000
18216960 22771200 18216960 22771200
22771200 U21 18976000 U22 T 22771200 U31 37952000 U32
27/11/14 9:43 AM
Section 4.3 Finite Element Formulation of Beams 247
Assembling [K](1) and [K](2) to obtain the global stiffness matrix yields
[K](G)
2277120 5692800 2277120 = F 5692800 0 0
5692800 18976000 5692800 9488000 0 0
2277120 5692800 20494080 17078400 18216960 22771200
5692800 9488000 17078400 56928000 22771200 18976000
0 0 18216960 22771200 18216960 22771200
0 0 22771200 V 18976000 22771200 37952000
Referring to Table 4.2, we can compute the load matrix for elements (1) and (2). The respective load matrices are wL 25 * 103 * 5 2 2 2 wL 25 * 103 * 52 62500 12 12 52083 = h x = h x = d t wL 25 * 103 * 5 62500 2 2 52083 2 3 2 wL 25 * 10 * 5 12 12 
5 F6(1)
wL 25 * 103 * 2.5 2 2 2 wL 25 * 103 * 2.52 31250 12 12 13021 = h x = h x = d t wL 25 * 103 * 2.5 31250 2 2 13021 2 3 2 wL 25 * 10 * 2.5 12 12 
5 F6(2)
Combining the two load matrices to obtain the global load matrix, we obtain
5 F6(G)
M04_MOAV4303_04_GE_C04.INDD 247
62500 62500 52083 52083 62500  31250 93750 = f v = f v 52083  13021 39062 31250 31250 13021 13021
27/11/14 9:43 AM
248 Chapter 4 Axial Members, Beams, and Frames
Applying the boundary conditions U11 = U12 = 0 at node 1 and the boundary condition U21 = 0 at node 2, we have 1 0 0 0 0 0 U11 0 0 1 0 0 0 0 U12 0 0 0 1 0 0 0 U21 0 F V f v=f v 5692800 9488000 17078400 56928000 22771200 18976000 U22 39062 0 0 18216960 22771200 18216960 22771200 U31 31250 U32 0 0 22771200 18976000 22771200 37952000 13021 Considering the applied boundary conditions, we reduce the global stiffness matrix and the load matrix to 56928000 C 22771200 18976000
22771200 18216960 22771200
18976000 U22 39062 22771200 S c U31 s = c 31250 s 37952000 U32 13021
Solving the three equations simultaneously results in the unknown nodal values. The displacement result is [U]T = [0
0
0
0.0013723(rad)
0.0085772(m)
0.004117(rad)]
We can compute the nodal reaction forces and moments from the relationship
5 R 6 = [K]5 U6  5 F6 (4.50)
where 5 R 6 is the reaction matrix. Substituting for the appropriate values in Eq. (4.50), we have R1 2277120 5692800 2277120 5692800 M1 5692800 18976000 5692800 9488000 R2 2277120 5692800 20494080 17078400 f v = F M2 5692800 9488000 17078400 56928000 R3 0 0 18216960 22771200 M3 0 0 22771200 18976000
0 0 0 0 18216960 22771200 V * 22771200 18976000 18216960 22771200 22771200 37952000 0 0 0 f vf 0.0013723 0.0085772 0.0041170
M04_MOAV4303_04_GE_C04.INDD 248
62500 52083 93750 v 39062 31250 13021
27/11/14 9:43 AM
Section 4.3 Finite Element Formulation of Beams 249
Performing the matrix operation results in the following reaction forces and moments at each node: R1 54687(N) M1 39062(N # m) R 132814(N) f 2v = f v M2 0 R3 0 M3 0 Note that by calculating the reaction matrix using the nodal displacement matrix, we can check the validity of our results. There is a reaction force and a reaction moment at node 1; there is a reaction force at node 2; there is no reaction moment at node 2, as expected; and there are no reaction forces or moments at node 3, as expected. The accuracy of the results is discussed further in Section 4.7.
Example 4.4 Revisited We will now show how to use Excel to set up and solve Example 4.4. 1. In cell A1 type Example 4.4, and in cells A3, A4, and A5 type E =, I =, and w = as shown. After inputting the value of E in cell B3, select B3 and in the “Name Box” type E and hit the Return key. Similarly, after inputting the values of I and w in cells B4 and B5 select B4 and B5 and in the corresponding “Name Box” type I and W. Next, create the table shown with the element and node numbers, Length, I, and E as shown.
2. Next compute [K1] and [K2] as shown and name them Kelement1 and Kelement2. To create [K1], select the range H16: K19 and type =(E*I>Length1^3)*C16:F19 and while holding down the Ctrl and Shift keys, press the Return key. Create [K2] in a similar way.
M04_MOAV4303_04_GE_C04.INDD 249
27/11/14 9:43 AM
250 Chapter 4 Axial Members, Beams, and Frames
3. Now create the {F1} and {F2} elements and name them Felement1 and Felement2. To create {F1}, select cell D26 and type = − w*Length1/2, and in cell D27 type = − w*Length1^2/12, and so on.
4. Next, create the [A1] and [A2] matrices and name them Aelement1 and Aelement2 as shown. If you have forgotten what the A matrices represent, see Section 2.5, Equation (2.9).
5. We now create the stiffness matrix for each element (with their proper positions in the global matrix) and name them K1G and K2G. For example, to create [K]1G, select B41:G46 and type = MMULT(TRANSPOSE(Aelement1),MMULT(Kelement1,Aelement1)) and while holding down the Ctrl and Shift keys, press the Return key. In a similar way, create [K]2G, as shown.
M04_MOAV4303_04_GE_C04.INDD 250
27/11/14 9:43 AM
Section 4.3 Finite Element Formulation of Beams 251
6. The {F}1G and {F}2G matrices are computed next. To create {F}1G, select the range B55:B60 and type = MMULT(TRANSPOSE(Aelement1),Felement1) and while holding down the Ctrl and Shift keys, press the Return key. Name this range F1G and in a similar way create {F}2G and name it F2G.
7. Create the final global stiffness and load matrices. Select the range B62:G67 and type =K1G+K2G and while holding down the Ctrl and Shift keys, hit the Return key. Name this range KG. In a similar way, create the global load matrix and name it FG.
M04_MOAV4303_04_GE_C04.INDD 251
27/11/14 9:43 AM
252 Chapter 4 Axial Members, Beams, and Frames
8. Apply the boundary conditions. Copy the appropriate portion of the KG matrix and paste it in the range C76:E78 as values only. Name the range KwithappliedBC. Similarly, create the corresponding load matrix in the range C80:C82 and name it FwithappliedBC.
9. Select the range C84:C86 and type = MMULT(MINVERSE(KwithappliedBC),FwithappliedBC) and while holding down the Ctrl and Shift keys, press the Return key. Moreover, as shown, copy the values of {U partial} and add the boundary conditions U11 = 0, U12 = 0, and U21 = 0 into cells C88:C93. Name this matrix, UG.
10. Calculate the reaction forces and moments. Select the range C95:C100 and type = (MMULT(KG,UG)FG) and while pressing the Ctrl and Shift keys, press the Return key.
M04_MOAV4303_04_GE_C04.INDD 252
27/11/14 9:43 AM
Section 4.3 Finite Element Formulation of Beams 253
The complete Excel sheet is shown next.
M04_MOAV4303_04_GE_C04.INDD 253
27/11/14 9:43 AM
254 Chapter 4 Axial Members, Beams, and Frames
4.4 Finite Element Formulation of Frames Frames represent structural members that may be rigidly connected with welded joints or bolted joints. For such structures, in addition to rotation and lateral displacement, we also need to be concerned about axial deformations. Here, we focus on plane frames. The frame element, shown in Figure 4.12, consists of two nodes. At each node, there are three degrees of freedom: a longitudinal displacement, a lateral displacement, and a rotation. Referring to Figure 4.12, note that ui1 represents the longitudinal displacement and ui2 and ui3 represent the lateral displacement and the rotation at node i, respectively. In the same manner, uj1, uj2, and uj3 represent the longitudinal displacement, the lateral displacement, and the rotation at node j, respectively. In general, two frames of reference are required to describe frame elements: a global coordinate system and a local frame of reference. We choose a fixed global coordinate system (X, Y) for several uses: (1) to represent the location of each joint (node) and to keep track of the orientation of each element using angles such as u; (2) to apply the constraints and the loads in terms of their respective global components; and (3) to represent the solution. We also need a local, or elemental, coordinate system to describe the axialload behavior of an element. The relationship between the local coordinate system (x, y) and the global coordinate system (X, Y) is shown in Figure 4.12. Because there are three degrees of freedom associated with each node, the stiffness matrix for the frame element will be a 6 * 6 matrix. The local degrees of freedom are related to the global degrees of freedom through the transformation matrix, according to the relationship
M04_MOAV4303_04_GE_C04.INDD 254
[u] = [T][U] (4.51)
27/11/14 9:43 AM
Section 4.4 Finite Element Formulation of Frames 255 uj2
Y
uj1
uj3
X
j
ui2
y
x
ui1 u
ui3 i
Figure 4.12 A frame element.
where the transformation matrix is cos u sin u 0 [T] = F 0 0 0
sin u cos u 0 0 0 0
0 0 1 0 0 0
0 0 0 cos u sin u 0
0 0 0 sin u cos u 0
0 0 0 V (4.52) 0 0 1
In the previous section, we developed the stiffness matrix attributed to bending for a beam element. This matrix accounts for lateral displacements and rotations at each node and is
[K](e) xy
ui1 ui2 0 0 0 12 0 6L El = 3 F 0 L 0 0 12 0 6L
ui3 uj1 0 0 6L 0 4L2 0 0 0 6L 0 2L2 0
uj2 0 12 6L 0 12 6L
uj3 0 6L 2L2 V 0 6L 4L2
ui1 ui2 ui3 uj1 uj2 uj3
(4.53)
To represent the contribution of each term to nodal degrees of freedom, the degrees of freedom are shown above and alongside the stiffness matrix in Eq. (4.53). In Section 4.1 we derived the stiffness matrix for members under axial loading as
[K](e) axial
M04_MOAV4303_04_GE_C04.INDD 255
ui1 AE L 0 0 = H AE L 0 0
ui2 ui3 0
0
0 0
0 0
0
0
0 0
0 0
uj1 uj2 uj3 AE 0 0 ui1 L 0 0 0 ui2 0 0 0 ui3 X AE 0 0 uj1 L 0 0 0 uj2 0 0 0 uj3
(4.54)
27/11/14 9:43 AM
256 Chapter 4 Axial Members, Beams, and Frames
Adding Eqs. (4.53) and (4.54) results in the stiffness matrix for a frame element with respect to local coordinate system x, y
º
AE L 0 0
[K](e) xy = ∞
∫
0
0
12EI L3 6EI L2
6EI L2 4EI L
AE L 0 0
0 
12EI L3 6EI L2
0 6EI L2 2EI L


AE L 0 0
AE L 0 0
0
0
12EI L3 6EI  2 L
6EI L2 2EI L
0
0

12EI L3 6EI  2 L
≠ ∞ (4.55)
6EI L2 ª 4EI L

Note that we need to represent Eq. (4.55) with respect to the global coordinate system. To perform this task, we must substitute for the local displacements in terms of the global displacements in the strain energy equation, using the transformation matrix and performing the minimization. (See Problem 4.13.) These steps result in the relationship [K](e) = [T]T[K](e) xy [T] (4.56)
where [K](e) is the stiffness matrix for a frame element expressed in the global coordinate system X, Y. Next, we will demonstrate finite element modeling of frames with another example. EXAMPLE 4.5 Consider the overhang frame shown in Figure 4.13. The frame is made of steel, with E = 30 * 106 lb/in2. The crosssectional areas and the second moment of areas for the two members are shown in Figure 4.13. The frame is fixed as shown in the figure, and we are interested in determining the deformation of the frame under the given distributed load. We model the problem using two elements. For element (1), the relationship between the local and the global coordinate systems is shown in Figure 4.14. Similarly, the relationship between the coordinate systems for element (2) is shown in Figure 4.15. Note that for this problem, the boundary conditions are U11 = U12 = U13 = U31 = U32 = U33 = 0. For element (1), the local and the global frames of reference are aligned in the same direction; therefore, the stiffness matrix for element (1) can be computed from Eq. (4.55) resulting in
M04_MOAV4303_04_GE_C04.INDD 256
27/11/14 9:43 AM
Section 4.4 Finite Element Formulation of Frames 257 800 lb/ft
A = 7.65 in2 I = 204 in4
9 ft in2
A = 7.65 I = 204 in4
10 ft Figure 4.13 An overhang frame supporting a distributed load. u12 y Y
u13 X
u22
x
u23
u11
1
u21 2
(1) Figure 4.14 The configuration of element (1).
[K](1)
1912.5 0 0 = 103 F 1912.5 0 0
0 42.5 2550 0 42.5 2550
0 2550 204000 0 2550 102000
1912.5 0 0 1912.5 0 0
0 42.5 2550 0 42.5 2550
0 2550 102000 V 0 2550 204000
For element (2), the stiffness matrix represented with respect to the local coordinate system is
[K](2) xy
M04_MOAV4303_04_GE_C04.INDD 257
2125 0 0 = 103 F 2125 0 0
0 58.299 3148.148 0 58.299 3148.148
0 3148.148 226666 0 3148.148 113333
2125 0 0 2125 0 0
0 58.299 3148.148 0 58.299 3148.148
0 3148.148 113333 V 0 3148.148 226666
27/11/14 9:43 AM
258 Chapter 4 Axial Members, Beams, and Frames Y
u23 2
X
y
u22
x u21
(2)
u33 u32
3
u31
Figure 4.15 The configuration of element (2).
For element (2), the transformation matrix is cos(270) sin(270) 0 [T] = F 0 0 0 0 1 0 [T] = F 0 0 0
1 0 0 0 0 0
0 0 1 0 0 0
sin(270) cos(270) 0 0 0 0 0 0 0 0 1 0
0 0 0 1 0 0
0 0 1 0 0 0
0 0 0 cos(270) sin(270) 0
0 0 0 sin(270) cos(270) 0
0 0 0 V 0 0 1
0 0 0 V 0 0 1
The transpose of the transformation matrix is 0 1 0 [T]T = F 0 0 0
M04_MOAV4303_04_GE_C04.INDD 258
1 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
27/11/14 9:43 AM
Section 4.4 Finite Element Formulation of Frames 259
Substituting for [T]T, [K](2) xy , and [T] into Eq. (4.56), we have
[K](2)
0 1 0 = 103 F 0 0 0
1 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 2125 0 0 2125 0 0 0 0 58.299 3148.148 0 58.299 3148.148 0 0 3148.148 226666 0 3148.148 113333 V F V 0 2125 0 0 2125 0 0 0 0 58.299 3148.148 0 58.299 3148.148 1 0 3148.148 113333 0 3148.148 226666 0 1 0 F 0 0 0
1 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
and performing the matrix operation, we obtain
[K](2)
58.299 0 3148.148 = 103 F 58.299 0 3148.148
0 2125 0 0 2125 0
3148.148 0 226666 3148.148 0 113333
58.299 0 3148.148 58.299 0 3148.148
0 2125 0 0 2125 0
3148.148 0 113333 V 3148.1480 0 226666
Constructing the global stiffness matrix by assembling [K](1) and [K](2), we have
[K](G)
1912.5 0 0 1912.5 = 103 I 0 0 0 0 0
0 42.5 2550 0 42.5 2550 0 0 0
0 2550 204000 0 2550 102000 0 0 0
1912.5 0 0 1912.5 + 58.299 0 0 + 3148.148 58.299 0 3148.148
0 42.5 2550 0 42.5 + 2125 2550 0 2125 0
0 0 0 58.299 0 3148.148 58.299 0 3148.148
M04_MOAV4303_04_GE_C04.INDD 259
0 2550 102000 0 + 3148.148 2550 204000 + 226666 3148.148 0 113333
0 0 0 0 2125 0 0 2125 0
0 0 0 3148.148 0 Y 113333 3148.1480 0 226666
27/11/14 9:43 AM
260 Chapter 4 Axial Members, Beams, and Frames
The load matrix is
5 F6(1)
0 0 wL 800 * 10 2 2 0 wL2 800 * 102 * 12 4000 12 12 80000 = i v y = i y= f 0 0 0 wL 800 * 10 4000 2 2 80000 2 2 wL 800 * 10 * 12 12 12
In the load matrix, the force terms have the units of lb, whereas the moment terms have the units of lb # in. Application of the boundary conditions (U11 = U12 = U13 = U31 = U32 = U33 = 0) reduces the 9 * 9 global stiffness matrix to the following 3 * 3 matrix: 1970.799 10 C 0 3148.148 3
0 2167.5 2550
3148.148 U21 0 2550 S c U22 s = c 4000 s 430666 U23 80000
Solving these equations simultaneously results in the following displacement matrix: [U]T = [0
0
0
0.0002845(in)
0.0016359(in)
0.00017815(rad)
0
0
0]
This problem will be revisited later in the chapter and solved with ANSYS. 4.5 ThreeDimensional Beam Element ANSYS’s threedimensional beam elements are suited for situations wherein the beam may be subjected to loads that can create tension, compression, bending about different axes, and twisting (torsion). At each node, there are six degrees of freedom, displacements in X, Y, and Zdirections, and rotation about X, Y, and Zaxes. A seventh degree of freedom (warping magnitude) is optional. Therefore, absent a seventh degree the elemental matrix for a threedimensional beam element is a 12 * 12 matrix. ANSYS’s threedimensional elastic beam element is shown in Figure 4.16. The element input data include node locations, the crosssectional properties, and the material properties. Note that BEAM188 element is defined by two or three nodes. For user control of the element orientation about the element xaxis use the third node option. The third node (K), if used, defines a plane (with I and J) containing the element x and z axes (as shown in Figure 4.16). Also see Example 4.6. The input data for BEAM188 is summarized below: Nodes I, J, K (K orientation node is optional)
M04_MOAV4303_04_GE_C04.INDD 260
27/11/14 9:43 AM
Section 4.5 ThreeDimensional Beam Element 261 z 1
4
K
2
3
I
y Y x X J
Z
5
Figure 4.16 BEAM188 element, the threedimensional elastic beam element used by ANSYS.
Degrees of Freedom UX, UY, UZ (displacements in X, Y, and Zdirections) ROTX (rotation about Xaxis), ROTY (rotation about Yaxis), ROTZ (rotation about Zaxis) Section Properties The section properties may be inputted directly or calculated by ANSYS. They include: Area of section; Moment of inertia about Y and Z axes; Product of inertia; Warping constant; Torsional constant; Y and Z coordinates of centroid; and shear deflection constant. Material Properties EX (modulus of elasticity), ALPX (Poisson’s ratio), DENS (density), GXY (shear modulus), DAMP (damping) Surface Loads Pressures face 1 (I  J) ( Z normal direction) face 2 (I  J) ( Y normal direction) face 3 (I  J) ( +X tangential direction) face 4 (I) ( +X axial direction) face 5 (J) ( X axial direction) (use negative value for opposite loading) Stresses As you will see in Example 4.6, to review stresses in beams, you must first copy these results into element tables, and then you can list them or plot them. These items are obtained using item label and sequence numbers. BEAM188 output includes stress values—examples of these stresses are given in Table 4.3.
M04_MOAV4303_04_GE_C04.INDD 261
27/11/14 9:43 AM
262 Chapter 4 Axial Members, Beams, and Frames
Table 4.3 Examples of stresses computed by ANSYS SDIR
Axial direct stress
SBYT
Bending stress on the element + Y side of the beam
SBYB
Bending stress on the element  Y side of the beam
SBZT
Bending stress on the element + Z side of the beam
SBZB
Bending stress on the element  Z side of the beam
Table 4.4 Item and sequence numbers for the BEAM188 element Name
Item
I
J
SDIR
SMIC
31
36
SBYT
SMIC
32
37
SBYB
SMIC
33
38
SBZT
SMIC
34
39
SBZB
SMIC
35
40
Once you decide which stress values you want to look at, you can read them into a table using item labels and sequence numbers. Examples of the item labels and sequence numbers for BEAM188 are summarized in Table 4.4. See Example 4.6 for details on how to read stress values into a table for a beam element.
4.6 An Example Using ANSYS ANSYS offers two beam elements that can be used to model structural problems. BEAM188 is a 3D element with tension, compression, and bending capabilities. The element has six degrees of freedom at each node: translation in the x and ydirections and rotation about the zaxis and rotation about X, Y, and Zaxes. The element input data include node locations, crosssectional properties, and the material properties. Output data include nodal displacements and additional elemental output. Examples of elemental output are given in Table 4.3. BEAM189 is a quadratic threenode element in 3D. It is suited for analyzing slender to stubby/thick beam structures. Example 4.6 Consider the cantilever beam shown in the accompanying figure. The beam is made of an aluminum alloy, with E = 10 * 106 lb/in2. The crosssectional area and the applied loads are also shown in the figure. We will use ANSYS’s Beam188 to solve this problem and compare the results with those of beam theory.
M04_MOAV4303_04_GE_C04.INDD 262
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 263 Y X
z
Z PY = 50 lb
PZ = 75 lb 4 in y
4 in 36 in
y
PX = 100 lb 2 in
k 3
i
i
1
2
(beam orientation node)
z 2 in
To solve this problem using ANSY, first, we need to make sure that the beam crosssection is oriented in the way we want. Note that the beam element’s crosssection axes (as shown in the accompanying figure) are oriented differently from those of the global axes. Pay close attention to the orientation of the element’s crosssection axes, y and z, and the orientation of the global Y and Z axes. When analyzing beam problems, using ANSYS, it is good practice to make use of the orientation node k to define the orientation of a beam element. The orientation node k defines a plane (with nodes i and j) containing the element x and zaxes (see Figure 4.16). If you do not define an orientation node, the default orientation of the element yaxis is automatically calculated to be parallel to the global XY plane. For the case where the element is parallel to the global Zaxis, the element yaxis is oriented parallel to the global Yaxis. Enter the ANSY program by using the Launcher. Type Beam (or a file name of your choice) in the Jobname entry field of the ANSY Launcher and Pick Run to start GUI. utility menu: File S Change Title . . .
M04_MOAV4303_04_GE_C04.INDD 263
27/11/14 9:43 AM
264 Chapter 4 Axial Members, Beams, and Frames
main menu: Preprocessor S Element Type S Add>Edit>Delete
Next, click on the Options . . . button and set K1, K2, . . . . , K15 options, as shown.
M04_MOAV4303_04_GE_C04.INDD 264
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 265
Assign the modulus of elasticity by using the following commands: main menu: Preprocessor S Material Models S Structural S Linear S Elastic S Isotropic
M04_MOAV4303_04_GE_C04.INDD 265
27/11/14 9:43 AM
266 Chapter 4 Axial Members, Beams, and Frames
Close the “Define Material Model Behavior” window. main menu: Preprocessor S Sections S Beam S Common Sections
A list of predefined shapes
Number of cells along width, default = 2 Number of cells along height, default = 2
You may view the section properties by issuing the following commands: main menu: Preprocessor S Sections S List Sections
M04_MOAV4303_04_GE_C04.INDD 266
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 267
ANSYS Toolbar: SAVE_DB main menu: Preprocessor S Modeling S Create S Nodes S In Active CS
M04_MOAV4303_04_GE_C04.INDD 267
27/11/14 9:43 AM
268 Chapter 4 Axial Members, Beams, and Frames
main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes [pick node 1] [pick node 2] [pick node 3] [anywhere in the ANSYS graphics window] OK utility menu: Plot S Elements Toolbar: SAVE_DB Apply boundary conditions with the following commands: main menu: Solution S Define Loads S Apply S Structural S Displacement S On Nodes [pick node 1]
[apply anywhere in the ANSYS graphics window]
M04_MOAV4303_04_GE_C04.INDD 268
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 269
main menu: Solution S Define Loads S Apply S Structural S Force/Moment S On Nodes [pick node 2]
[apply anywhere in the ANSYS graphics window]
M04_MOAV4303_04_GE_C04.INDD 269
27/11/14 9:43 AM
270 Chapter 4 Axial Members, Beams, and Frames
main menu: Solution S Solve S Current LS OK Close (the solution is done!) window. Close (the STAT Command) window. main menu: General Postproc S List Results S Nodal Solution
M04_MOAV4303_04_GE_C04.INDD 270
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 271
In a similar way, list Rotation vector sum.
main menu: General Postproc S List Results S Reaction Solution
M04_MOAV4303_04_GE_C04.INDD 271
27/11/14 9:43 AM
272 Chapter 4 Axial Members, Beams, and Frames
main menu: General Postproc S Element Table S Define Table
M04_MOAV4303_04_GE_C04.INDD 272
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 273
M04_MOAV4303_04_GE_C04.INDD 273
27/11/14 9:43 AM
274 Chapter 4 Axial Members, Beams, and Frames
main menu: General Postproc S Element Table S List Element Table
M04_MOAV4303_04_GE_C04.INDD 274
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 275
The results of ANSYS analysis and those from the beam theory are summarized in the accompanying table. As you can see the results are in good agreement. Results using beam theory (see Table 4.1) vX =
PXL AE
=
(100)(36) (8)(10 * 106)
= 0.0000045 in
ANSYS results 0.0000045 in
(vY) max =
 (50)(36)3  PYL3 = =  0.00729 in 3EI 3(10 * 106)(10.67)
 0.00735 in
(vZ) max =
(75)(36)3 PZL3 = = 0.0437 in 3EI 3(10 * 106)(2.67)
0.0445 in
sxx  axial =
PX 100 lb = = 12.5 2 A 8 in
12.5
(szz  bending) max = (syy  bending) max =
Mzc I Myc I
= =
(50)(36)(2) 10.67 (75)(36)(1) 2.67
lb in2
337.5
lb in2
1012
= 337.4 = 1011
lb in2 lb in2
lb in2
(uZ) max =
 (50)(36)2  PYL2 = =  0.0003036 rad 2EI 2(10 * 106)(10.67)
 0.0003037 rad
(uY) max =
(75)(36)2  PZL2 = =  0.00182 rad 2EI 2(10 * 106)(2.67)
 0.00182 rad
a Fx = 0; 100 + Rx = 0; Rx =  100 lb; Mx = 0 a Fy = 0;
 50 + Ry = 0; Ry = 50 lb; My = (75 lb)(36 in) = 2700 lb # in
a Fz = 0; 75 + Rz = 0; Rz =  75 lb; Mz = (50 lb)(36 in) = 1800 lb # in
M04_MOAV4303_04_GE_C04.INDD 275
Rx =  100 lb; Mx = 0 Ry = 50 lb; My = 2700 lb # in Rz =  75 lb; Mz = 1800 lb # in
27/11/14 9:43 AM
276 Chapter 4 Axial Members, Beams, and Frames
Example 4.5 Revisited Let us consider the overhang frame again in order to solve this problem using ANSYS. Recall that the frame is made of steel with E = 30 * 106 lb/in2. The respective crosssectional areas and the second moments of areas for the two members are shown in Figure 4.13 (repeated in Figure 4.17 for your convenience). The members have a depth of 12.22 in. The frame is fixed as shown in the figure. We are interested in determining the deflections and the rotation of the frame under the given distributed load. In this example, we show you how you may setup beam and frame problems using the user defined section, and without using the third node (K) option. 800 lb/ft
A = 7.65 in2 I = 204 in4
9 ft A = 7.65 in2 I = 204 in4
10 ft
Figure 4.17 An overhang frame supporting a distributed load.
Enter the ANSYS program by using the Launcher. Type Frame2D (or a file name of your choice) in the Jobname entry field of the ANSYS Product Launcher and Pick Run to start GUI. Create a title for the problem. This title will appear on ANSYS display windows to provide a simple way to identify the displays. Use the following command sequences: utility menu: File S Change Title c
M04_MOAV4303_04_GE_C04.INDD 276
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 277
main menu: Preprocessor S Element Type S Add/Edit/Delete
M04_MOAV4303_04_GE_C04.INDD 277
27/11/14 9:43 AM
278 Chapter 4 Axial Members, Beams, and Frames
Next click on Options . . . button and set K1, K2, . . ., K15 option as shown.
Assign the modulus of elasticity by using the following commands: main menu: Preprocessor S Material Props S Material Models S Structural S Linear S Elastic S Isotropic
M04_MOAV4303_04_GE_C04.INDD 278
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 279
Close the “Define Material Model Behavior” window. main menu: Preprocessor S Sections S Beam S Common Sections
data
ANSYS Toolbar: SAVE_DB Set up the graphics area (i.e., work plane, zoom, etc.) with the following commands: utility menu: Workplane S WP Settings . . .
M04_MOAV4303_04_GE_C04.INDD 279
27/11/14 9:43 AM
280 Chapter 4 Axial Members, Beams, and Frames
utility menu: Workplane S Display Working Plane Bring the workplane to view by the command utility menu: PlotCtrls S Pan, Zoom, Rotate . . . Click on the small circle until you bring the workplane to view. Then create the nodes and elements: main menu: Preprocessor S Modeling S Create S Nodes S On Working Plane [WP = 0,108]
[WP = 120,108]
[WP = 120,0]
OK main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes [pick node 1]
[pick node 2]
[apply anywhere in the ANSYS graphics window]
[pick node 2]
[pick node 3]
M04_MOAV4303_04_GE_C04.INDD 280
27/11/14 9:43 AM
Section 4.6 An Example Using ANSYS 281
[anywhere in the ANSYS graphics window] OK
utility menu: Plot S Elements
Toolbar: SAVE_DB
Apply boundary conditions with the following commands:
main menu: Solution S Define Loads S Apply S Structural S Displacement S On Nodes [pick node 1]
[pick node 3]
[anywhere in the ANSYS graphics window]
main menu: Solution S Define Loads S Apply S Structural S Pressure S On Beams
M04_MOAV4303_04_GE_C04.INDD 281
27/11/14 9:43 AM
282 Chapter 4 Axial Members, Beams, and Frames
[pick element 1]
[anywhere in the ANSYS graphics window]
To see the applied distributed load and boundary conditions, use the following commands: utility menu: Plot Ctrls S Symbols c utility menu: Plot S Elements ANSYS Toolbar: SAVE_DB
M04_MOAV4303_04_GE_C04.INDD 282
27/11/14 9:44 AM
Section 4.6 An Example Using ANSYS 283
Solve the problem: main menu: Solution S Solve S Current LS OK
M04_MOAV4303_04_GE_C04.INDD 283
27/11/14 9:44 AM
284 Chapter 4 Axial Members, Beams, and Frames
Close (the solution is done!) window. Close (the/STAT Command) window. Begin the postprocessing phase and plot the deformed shape with the following commands: main menu: General Postproc S Plot Results S Deformed Shape
M04_MOAV4303_04_GE_C04.INDD 284
27/11/14 9:44 AM
Section 4.6 An Example Using ANSYS 285
List the nodal displacements with the following commands: main menu: General Postproc S List Results S Nodal Solution
Also list Rotation vector sum.
M04_MOAV4303_04_GE_C04.INDD 285
27/11/14 9:44 AM
286 Chapter 4 Axial Members, Beams, and Frames
List the reactions with the following commands: main menu: General Postproc S List Results S Reaction Solution
M04_MOAV4303_04_GE_C04.INDD 286
27/11/14 9:44 AM
Section 4.7 Verification of Results 287
Exit ANSYS and save everything: ANSYS Toolbar: QUIT
4.7 Verification of Results Refer to Example 4.2. One way of checking for the validity of our FEA findings of Example 4.2 is to arbitrarily cut a section through the column and apply the static equilibrium conditions. As an example, consider cutting a section through the column containing element (2), as shown in the accompanying illustration. 160,000 lb
(2)
160,000 lb
The average normal stress in that section of the column is finternal 160,000 s(2) = = = 4030 lb/in2 A 39.7 In a similar way, the average stress in element (4) can be checked by finternal 60,000 s(4) = = = 1511 lb/in2 A 39.7 The stresses computed in this manner are identical to the results obtained earlier using the energy method.
M04_MOAV4303_04_GE_C04.INDD 287
27/11/14 9:44 AM
288 Chapter 4 Axial Members, Beams, and Frames
It is always necessary to compute the reaction forces and moments for beam and frame problems. The nodal reaction forces and moments can be computed from the relationship 5 R 6 = [K]5 U6  5 F6 We computed the reaction matrix for Example 4.4, repeated here: R1 54687(N) M1 39062(N ~ m) R2 132814(N) f v = f v M2 0 R3 0 M3 0 Earlier, we discussed how to check the validity of results qualitatively. It was mentioned that the results indicated that there is a reaction force and a reaction moment at node 1; there is a reaction force at node 2; there is no reaction moment at node 2, as expected; and there are no reaction forces or moments at node 3, as expected for the given problem. Let us also perform a quantitative check on the accuracy of the results. We can use the computed reaction forces and moments against the external loading to check for static equilibrium (see Figure 4.18): +
c ΣFY = 0
13,2814 + 54,687  (25,000) (7.5) = 1 ≈ 0
and + ΣMnode 2 = 0 ⤿
39,062  54,687(5) + (25,000) (7.5) (1.25) = 2 ≈ 0
Similarly, in reference to Example 4.5, we find that the reaction results generated using ANSYS are shown in Figure 4.19. Checking for static equilibrium, we find that + ΣFX = 0 S
c ΣFY = 0 + ΣMnode 1 = 0 ⤿
534.40  534.40 = 0 4516.7 + 3483.3  (800) (10) = 0 101,470 + 18,259 + 3483.3(10) (12)  (534.40) (9) (12) (800) (10) (5) (12) ≈ 0 25,000 N/m
1
2
3
39,062 N · m 5m
54,687 N
2.5 m
132,814 N
Figure 4.18 The freebody diagram for Example 4.4.
M04_MOAV4303_04_GE_C04.INDD 288
27/11/14 9:44 AM
Chapter 4 Summary 289 800 lb/ft
534.40 lb
1 101,470 lb · in 4516.7 lb 9 ft
534.40 lb 18,259 lb · in 10 ft 3483.3 lb Figure 4.19 The freebody diagram for Example 4.5.
These simple problems illustrate the importance of checking for equilibrium conditions when verifying results. Summary At this point you should 1. know how to formulate stiffness matrix for a member under axial loading. 2. know that it is wise to use simple analytical solutions rather than finite element modeling for a simple problem whenever appropriate. Use finite element modeling only when it is necessary to do so. 3. know that the stiffness matrix for a beam element with two degrees of freedom at each node (the vertical displacement and rotation) is
[K](e)
12 EI 6L = 3D L 12 6L
6L 4L2 6L 2L2
12 6L 12 6L
6L 2L2 T 6L 4L2
4. know how to compute the load matrix for a beam element by consulting Table 4.2 for equivalent nodal forces.
M04_MOAV4303_04_GE_C04.INDD 289
27/11/14 9:44 AM
290 Chapter 4 Axial Members, Beams, and Frames
5. know that the stiffness matrix for a frame element (with local and global coordinate systems aligned) consisting of two nodes with three degrees of freedom at each node (axial displacement, lateral displacement and rotation) is AE L 0 0
[K](e) =
I AE 
0
0
12EI L3 6EI L2
6EI L2 4EI L
0
L 0 0

0
12EI L3 6EI L2

AE L 0 0
AE L
6EI L2 2EI L

0 0
0
0
12EI L3 6EI  2 L
6EI L2 2EI L
0
0

12EI L3 6EI  2 L
Y
6EI L2 4EI L

Note that for members that are not horizontal, the local degrees of freedom are related to the global degrees of freedom through the transformation matrix, according to the relationship 5 u 6 = [T]5 U6
where the transformation matrix is cos u sin u 0 [T] = F 0 0 0
sin u cos u 0 0 0 0
0 0 1 0 0 0
0 0 0 cos u sin u 0
0 0 0 sin u cos u 0
0 0 0 V 0 0 1
6. know how to compute the stiffness matrix for a frame element with an arbitrary orientation with respect to the global coordinate system using the relationship. [K](e) = [T]T[K](e) xy [T] 7. know how to compute the load matrix for a frame element by consulting Table 4.2 for equivalent nodal forces. References ANSYS User’s Manual: Procedures, Vol. I, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Commands, Vol. II, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Elements, Vol. III, Swanson Analysis Systems, Inc.
M04_MOAV4303_04_GE_C04.INDD 290
27/11/14 9:44 AM
Chapter 4 Problems 291 Beer, P., and Johnston, E. R., Mechanics of Materials, 2nd ed., New York, McGrawHill, 1992. Hibbeler, R. C., Mechanics of Materials, 2nd ed., New York, Macmillan, 1994. Segrlind, L., Applied Finite Element Analysis, 2nd ed., New York, John Wiley and Sons, 1984.
Problems 1. Determine the deflections of point D and point F and the axial stress in each member of the system shown in the accompanying figure. (E = 29 * 106 ksi.) 5"
Ac = 1 in2 D
5"
5000 lb
Ac = 0.5 in2
B
10"
F
10"
2. Consider a fourstory building with steel columns similar to the one presented in Example 4.2. The column is subjected to the loading shown in the accompanying figure. Assuming axial loading, (a) determine vertical displacements of the column at various floor column connection points and (b) determine the stresses in each portion of the column. (E = 29 * 106 lb/in2, A = 59.1 in2.) 45,000 lb 5
35,000 lb
(4)
15 ft
4
35,000 lb
(3)
15 ft
3
35,000 lb
(2)
15 ft
2
(1)
15 ft
1
M04_MOAV4303_04_GE_C04.INDD 291
27/11/14 9:44 AM
292 Chapter 4 Axial Members, Beams, and Frames
3. Determine the deflection of point D and the axial stress in each member in the system shown in the accompanying figure. (E = 10.6 * 103 ksi.)
Ac = 1 in2
Ac = 0.75 in2 5000 lb
Ac = 0.5 in2
D
C
2500 lb
12"
9"
8"
4. A 20fttall post is used to support advertisement signs at various locations along its height, as shown in the accompanying figure. The post is made of structural steel with a modulus of elasticity of E = 29 * 106 lb/in2. Not considering wind loading on the signs, (a) determine displacements of the post at the points of load application and (b) determine stresses in the post. 100 lb
5 ft
5 ft
10 ft
150 lb
200 lb
Ac = 1 in2
Ac = 2 in2
Ac = 3 in2
5. Determine the deflections of point D and point F in the system in the accompanying figure. Also compute the axial force and stress in each member. (E = 29 * 106 ksi.)
M04_MOAV4303_04_GE_C04.INDD 292
27/11/14 9:44 AM
Chapter 4 Problems 293
Ac = 0.2 in2
800 lb Ac = 0.15 in2
1000 lb D
Ac = 0.3 in2 6" Ac = 0.2
FF Ac = 1 in22 Ac = 1 in
1000 lb 6"
in2
800 lb 4"
6. Determine the deflections of point D and point F in the system in the accompanying figure. Also compute the axial force and stress in each member. E = 101 GPa Ac = 2 cm2
15 cm E = 69 GPa Ac = 5 cm2
25 kN D
F E = 73 GPa Ac = 5 cm2
E = 101 GPa
30 cm 50 cm
40 cm
50 cm
7. The beam shown in the accompanying figure is a wideflange W18 * 35, with a crosssectional
area of 10.3 in2 and a depth of 17.7 in. The second moment of area is 510 in4. The beam is subjected to a uniformly distributed load of 2000 lb/ft. The modulus of elasticity of the beam is E = 29 * 106 lb/in2. Using manual calculations, determine the vertical displacement at node 3 and the rotations at nodes 2 and 3. Also, compute the reaction forces at nodes 1 and 2 and reaction moment at node 1. 2,000 lb/ft
1
2
14 ft
3
7 ft
8. The beam shown in the accompanying figure is a wideflange W16 * 31 with a crosssectional
area of 9.12 in2 and a depth of 15.88 in. The second moment of area is 375 in4. The beam is subjected to a uniformly distributed load of 2000 lb/ft and a point load of 800 lb. The modulus of elasticity of the beam is E = 29 * 106 lb/in2. Using manual calculations, determine the
M04_MOAV4303_04_GE_C04.INDD 293
27/11/14 9:44 AM
294 Chapter 4 Axial Members, Beams, and Frames vertical displacement at node 3 and the rotations at nodes 2 and 3. Also, compute the reaction forces at nodes 1 and 2 and reaction moment at node 1. 2,000 lb/ft
1
2
10 ft
3
2.5 ft
2.5 ft
800 lb
9. The lamp frame shown in the accompanying figure has hollow, square cross sections and is made of steel, with E = 29 * 106 lb/in2. Using hand calculations, determine the endpoint deflection of the cross member where the lamp is attached. 6 ft
35 lb 15 ft
6 in
4 in
Main
Cross member
1/4 in thick
3/16 in thick
10. A park picnictable top is supported by two identical metal frames; one such frame is shown in the accompanying figure. The frames are embedded in the ground and have hollow, circular crosssectional areas. The tabletop is designed to support a distributed load of 300 lb/ft2. Using ANSYS, size the cross section of the frame to support the load safely. The modulus of elasticity is E = 29 * 106 lb/in2.
M04_MOAV4303_04_GE_C04.INDD 294
27/11/14 9:44 AM
Chapter 4 Problems 295 4 ft
28 in
32 in All members have the same hollow circular cross section.
11. The frame shown in the accompanying figure is used to support a load of 3000 lb. The main
vertical section of the frame has an annular cross section with an area of 8.63 in2 and a polar radius of gyration of 2.75 in. The outer diameter of the main tubular section is 6 in. All other members also have annular cross sections with respective areas of 2.24 in2 and polar radii of gyration of 1.91 in. The outer diameter of these members is 4 in. Using ANSYS, determine the deflections at the points where the load is applied. The frame is made of steel, with a modulus elasticity of E = 29 * 106 lb/in2. 12 ft
3 ft
1500 lb
455
1500 lb 15 ft
M04_MOAV4303_04_GE_C04.INDD 295
27/11/14 9:44 AM
296 Chapter 4 Axial Members, Beams, and Frames
12. Verify the equivalent nodal loading for a beam element subjected to a triangular load, as shown in the accompanying figure. w
L
3 wL 20
7 wL 20
wL2 30
wL2 20
13. Referring to the section in this chapter discussing the frame elements, show that the stiffness matrix represented with respect to the global coordinate system is related to the stiffness matrix described with respect to the frame’s local coordinate system, according to the relationship [K](e) = [T]T[K](e) xy [T]
14. The frame shown in the accompanying figure is used to support a load of 500 lb/ft. Using ANSYS, size the cross sections of each member if standardsize steel square tubing is to be used. Use three different sizes. The deflection of the centerpoint is to be kept under 0.05 in. 500 lb/ft
5 ft 10 ft
15 ft
M04_MOAV4303_04_GE_C04.INDD 296
27/11/14 9:44 AM
Chapter 4 Problems 297
15. The frame shown in the accompanying figure is used to support the load given in the figure. Using ANSYS, size the members if standard sizes of steel Ibeams are to be used.
700 lb/ft 12 ft
500 lb/ft
12 ft
15 ft
15 ft
16. Verify the equivalent nodal loading for a beam element subjected to the load shown in the accompanying figure. P
L — 2
P — 2 L — 2
M=
P — 2
PL 8
M=
PL 8
17. Use a oneelement model and calculate the deflection and slope at the endpoint of the beam shown in the accompanying figure. Compare your results to deflection and slope values given in Table 4.1. w0
E, I
X L
M04_MOAV4303_04_GE_C04.INDD 297
27/11/14 9:44 AM
298 Chapter 4 Axial Members, Beams, and Frames
18. Use a twoelement model and solve Example 4.3. Compare your results to the deflections and the end slope of the single element model.
19. Use a oneelement model and calculate the deflection and slope at the endpoint of the beam shown in the accompanying figure. Compare your results to the deflection and slope values given in Table 4.1. P
E, I
X L
20. Use a oneelement model and calculate the deflection and the slope at the midpoint of the beam shown in the accompanying figure. Compare your results to the deflection and slope values given in Table 4.1. w
E, I
L
21. The beam shown in the accompanying figure is a wideflange W18 * 35, with a cross
sectional area of 10.3 in2 and a depth of 17.7 in. The second moment of area is 510 in4. The beam is subjected to a point load of 2500 lb. The modulus of elasticity of the beam E = 29 * 106 lb/in2. Use a twoelement model and calculate the deflection of the midpoint of the beam. Compare your results to exact values. 2500 lb
7.5 ft
7.5 ft
22. The frame shown in the accompanying figure is used to support a load of 1000 lb. Using ANSYS, size the cross sections of each member if standard sizes of steel Ibeams are used.
M04_MOAV4303_04_GE_C04.INDD 298
27/11/14 9:44 AM
Chapter 4 Problems 299 4 ft
5 ft
4 ft
1000 lb
8 ft
23. The frame shown in the accompanying figure is used to support the indicated load. Using ANSYS, size the cross sections of each member if standard sizes of steel Ibeams are used. 2500 lb/ft
15 ft
3000 lb/ft
3000 lb/ft
20 ft
20 ft
15 ft
M04_MOAV4303_04_GE_C04.INDD 299
27/11/14 9:44 AM
300 Chapter 4 Axial Members, Beams, and Frames
24. The frame shown in the accompanying figure is used to support the indicated load. Using ANSYS, size the cross sections of each member if standard sizes of steel Ibeams are used.
10
/m N K
K N /m
10
455
5 KN
455
5m
40 KN/m 5 KN
5m
5m
M04_MOAV4303_04_GE_C04.INDD 300
27/11/14 9:44 AM
Chapter 4 Problems 301
25. The beams shown in the accompanying figure are used to support the indicated load. Using ANSYS, size the cross sections of each beam if standard sizes of steel Ibeams are used. 10 ft A
Z
B Y
5
ft
X
C 500 lb
1000 lb
26. Show that [D]{U} = {U}T[D]T. 27. Starting with Equations (4.39) and (4.41), show that the relationship given by Eqs. (4.42) is true.
28. Starting with Equation (4.42) and using the results given by Eq. (4.36a) perform the integrations to obtain the results given in the first column of the stiffness matrix, Equation (4.43).
29. Starting with Equation (4.42) and using the results given by Eq. (4.36a) perform the integrations to obtain the results given in the fourth row of the stiffness matrix, Equation (4.43).
M04_MOAV4303_04_GE_C04.INDD 301
27/11/14 9:44 AM
302 Chapter 4 Axial Members, Beams, and Frames
30. Design Project Size the members of the bridge shown in the accompanying figure for a case in which traffic is backed up with a total of four trucks equally spaced on the bridge. A typical truck has a payload weight of 64,000 lb and a cab weight of 8000 lb. As a starting point, you may use one cross section for all beam elements. You may also assume one cross section for all truss members. The roadbed weighs 1500 lb/ft and is supported by Ibeams. Use standard steel Ibeam sizes. Design your own truss configuration. In your analysis, you may assume that the concrete column does not deflect significantly. Write a brief report discussing how you came up with the final design. 200 ft 100 ft 25 ft
25 ft
10 ft
Ibeam
Design your own truss 58 ft
M04_MOAV4303_04_GE_C04.INDD 302
Concrete
27/11/14 9:44 AM
C h a p t e r
5
OneDimensional Elements
The objectives of this chapter are to introduce the concepts of onedimensional elements and shape functions and their properties in more detail. The idea of local and natural coordinate systems is also presented here. In addition, onedimensional elements used by ANSYS are discussed. These are the main topics discussed in Chapter 5: 5.1 Linear Elements 5.2 Quadratic Elements 5.3 Cubic Elements 5.4 Global, Local, and Natural Coordinates 5.5 Isoparametric Elements 5.6 Numerical Integration: Gauss–Legendre Quadrature 5.7 Examples of OneDimensional Elements in ANSYS 5.1 Linear Elements The heat transfer example in this section is employed to introduce the basic ideas of onedimensional elements and shape functions. Fins are commonly used in a variety of engineering applications to enhance cooling. Common examples include a motorcycle engine head, a lawn mower engine head, extended surfaces (heat sinks) used in electronic equipment, and finnedtube heat exchangers. A straight fin of a uniform cross section is shown in Figure 5.1, along with a typical temperature distribution along the fin. As a first approximation, let us divide the fin into three elements and four nodes. The actual temperature distribution may be approximated by a combination of linear functions, as shown in Figure 5.1. To better approximate the actual temperature gradient near the base of the fin in our finite element model, we have placed the nodes closer to each other in that region. It should be clear that we can improve the accuracy of our approximation by increasing the number of elements as well. However, for now, let us be content with the threeelement model and focus our attention on a typical element,
M05_MOAV4303_04_GE_C05.INDD 303
27/11/14 9:46 AM
304 Chapter 5 OneDimensional Elements T Actual temperature profile Tbase Approximate temperature profile
(2)
Base
(1) 1
Tfluid
(3)
X
3
2
4
Figure 5.1 Temperature distribution for a fin of uniform cross section.
L
T
Ti
Tj i
(e)
j Xj
Xi /
X Figure 5.2 Linear approximation of temperature distribution for an element.
as shown in Figure 5.2. The temperature distribution along the element may be interpolated (or approximated) using a linear function, as depicted in Figure 5.2. The forthcoming derivation of the shape functions are similar to the one we showed in Chapter 4, Section 4.1. As a review and for the sake of continuity and convenience, the steps to derive the shape functions are presented here again. The linear temperature distribution for a typical element may be expressed as T (e) = c1 + c2 X (5.1)
The element’s end conditions are given by the nodal temperatures Ti and Tj, according to the conditions
M05_MOAV4303_04_GE_C05.INDD 304
T = Ti at X = Xi (5.2) T = Tj at X = Xj
27/11/14 9:46 AM
Section 5.1 Linear Elements 305
Substitution of nodal values into Eq. (5.1) results in two equations and two unknowns: Ti = c1 + c2Xi (5.3) Tj = c1 + c2Xj
Solving for the unknowns c1 and c2, we get
c1 =
c2 =
TiXj  TjXi Xj  Xi Tj  Ti
(5.4) (5.5)
Xj  Xi
The element’s temperature distribution in terms of its nodal values is TiXj  TjXi
T (e) =
Xj  Xi
+
Tj  Ti Xj  Xi
X (5.6)
Grouping the Ti terms together and the Tj terms together, we obtain T (e) = ¢
Xj  X Xj  Xi
≤Ti + ¢
X  Xi ≤T (5.7) Xj  Xi j
We now define the shape functions, Si and Sj, according to the equations
Si =
Sj =
Xj  X Xj  Xi
=
Xj  X /
(5.8)
X  Xi X  Xi = (5.9) Xj  Xi /
where / is the length of the element. Thus, the temperature distribution of an element in terms of the shape functions can be written as T (e) = SiTi + SjTj (5.10)
Equation (5.10) can also be expressed in matrix form as T (e) = [Si
Sj] b
Ti r (5.11) Tj
As you recall, for the structural example in Chapter 4, the deflection u(e) for a typical column element is represented by u(e) = [Si
Sj] b
ui r (5.12) uj
where ui and uj represent the deflections of nodes i and j of an arbitrary element (e). It should be clear by now that we can represent the spatial variation of any unknown
M05_MOAV4303_04_GE_C05.INDD 305
27/11/14 9:46 AM
306 Chapter 5 OneDimensional Elements
variable over a given element by using shape functions and the corresponding nodal values. Thus, in general, we can write
Ψi r (5.13) Ψj where Ψi and Ψj represent the nodal values of the unknown variable, such as temperature, deflection, or velocity. Ψ(e) = [Si
Sj] b
Properties of Shape Functions The shape functions possess unique properties that are important for us to understand because they simplify the evaluation of certain integrals when we are deriving the conductance or stiffness matrices. One of the inherent properties of a shape function is that it has a value of unity at its corresponding node and has a value of zero at the adjacent node. Let us demonstrate this property by evaluating the shape functions at X = Xi and X = Xj. Evaluating Si at X = Xi and X = Xj, we get S i 0 X = Xi =
Xj  X /
2 X = Xi
=
Xj  Xi /
= 1 and Si 0 X = Xj =
Xj  X /
2 X = Xj
=
Xj  Xj /
= 0 (5.14)
Evaluating Sj at X = Xi and X = Xj, we obtain
S j 0 X = Xi =
Xj  Xi X  Xi Xi  Xi X  Xi 2 2 = = 0 and Sj 0 X = Xj = = = 1 (5.15) / / / / X = Xi X = Xj
This property is also illustrated in Figure 5.3. Another important property associated with shape functions is that the shape functions add up to a value of unity. That is, Xj  X X  Xi Si + Sj = + = 1 (5.16) Xj  Xi Xj  Xi It can also be readily shown that for linear shape functions, the sum of the derivatives with respect to X is zero. That is, d Xj  X d X  Xi 1 1 ¢ ≤ + ¢ ≤ = + = 0 (5.17) dX Xj  Xi dX Xj  Xi Xj  Xi Xj  Xi Si
Sj
1
1
Xi
Xj
X
Xi
Xj
X
Figure 5.3 Linear shape functions.
M05_MOAV4303_04_GE_C05.INDD 306
27/11/14 9:46 AM
Section 5.2 Quadratic Elements 307
EXAMPLE 5.1 We have used linear onedimensional elements to approximate the temperature distribution along a fin. The nodal temperatures and their corresponding positions are shown in Figure 5.4. What is the temperature of the fin at (a) X = 4 cm and (b) X = 8 cm? In Chapter 6, we will discuss in detail the analysis of onedimensional fin problems, including the computation of nodal temperatures. However, for now, using the given nodal temperatures, we can proceed to answer both parts of the question: a. The temperature of the fin at X = 4 cm is represented by element (2); (2) T (2) = S(2) 2 T2 + S3 T3 =
T =
X3  X X  X2 T2 + T3 / /
5  4 4  2 (41) + (34) = 36.35C 3 3
b. The temperature of the fin at X = 8 cm is represented by element (3); (3) T (3) = S(3) 3 T3 + S4 T4 =
T =
X  X3 X4  X T3 + T4 / /
10  8 8  5 (34) + (20) = 25.6°C 5 5 (2)
(3)
For this example, note the difference between S3 and S3 .
Tfluid = 185C
X
Tbase = 505C 1
(1)
2
2 cm
(2) 3 cm
3
(3)
4
T1 50 T 41 d 2t = d t °C T3 34 T4 20
5 cm
Figure 5.4 The nodal temperatures and their corresponding positions along the fin in Example 5.1.
5.2 Quadratic Elements We can increase the accuracy of our finite element findings either by increasing the number of linear elements used in the analysis or by using higher order interpolation functions. For example, we can employ a quadratic function to represent the spatial variation of an unknown variable. Using a quadratic function instead of a linear function requires
M05_MOAV4303_04_GE_C05.INDD 307
27/11/14 9:46 AM
308 Chapter 5 OneDimensional Elements
that we use three nodes to define an element. We need three nodes to define an element because in order to fit a quadratic function, we need three points. The third point can be created by placing a node, such as node k, in the middle of an element, as shown in Figure 5.5. Referring to the previous example of a fin, using quadratic approximation, the temperature distribution for a typical element can be represented by T (e) = c1 + c2X + c3X2 (5.18)
and the nodal values are
T = Ti at X = Xi (5.19)
T = Tk at X = Xk
T = Tj at X = Xj Three equations and three unknowns are created upon substitution of the nodal values into Eq. (5.18):
Ti = c1 + c2Xi + c3X2i (5.20)
Tk = c1 + c2Xk + c3X2k
Tj = c1 + c2Xj + c3X2j Solving for c1, c2, and c3 and rearranging terms leads to the element’s temperature distribution in terms of the nodal values and the shape functions: T (e) = SiTi + SjTj + SkTk (5.21)
In matrix form, the above expression is T (e) = [Si
Sj
Ti Sk] c Tj s (5.22) Tk
T
Ti Tk Tj i
k
j
Xi
Xk
Xj //2
/
M05_MOAV4303_04_GE_C05.INDD 308
X Figure 5.5 Quadratic approximation of the temperature distribution for an element.
27/11/14 9:46 AM
Section 5.3 Cubic Elements 309
where the shape functions are 2 (X  Xj) (X  Xk) (5.23) /2 2 Sj = 2 (X  Xi) (X  Xk) / 4 Sk = 2 (X  Xi) (X  Xj) /
Si =
In general, for a given element the variation of any parameter Ψ in terms of its nodal values may be written as (e)
Ψ
= [Si
Sj
Ψi Sk] c Ψj s (5.24) Ψk
It is important to note here that the quadratic shape functions possess properties similar to those of the linear shape functions; that is, (1) a shape function has a value of unity at its corresponding node and a value of zero at the other adjacent node, and (2) if we sum up the shape functions, we will again come up with a value of unity. The main difference between linear shape functions and quadratic shape functions is in their derivatives. The derivatives of the quadratic shape functions with respect to X are not constant. 5.3 Cubic Elements The quadratic interpolation functions offer good results in finite element formulations. However, if additional accuracy is needed, we can resort to even higher order interpolation functions, such as thirdorder polynomials. Thus, we can use cubic functions to represent the spatial variation of a given variable. Using a cubic function instead of a quadratic function requires that we use four nodes to define an element. We need four nodes to define an element because in order to fit a thirdorder polynomial, we need four points. The element is divided into three equal lengths. The placement of the four nodes is depicted in Figure 5.6. Referring to the previous example of a fin, using cubic approximation, the temperature distribution for a typical element can be represented by T (e) = c1 + c2 X + c3 X2 + c4 X3 (5.25)
and the nodal values are
M05_MOAV4303_04_GE_C05.INDD 309
T T T T
= = = =
Ti Tk Tm Tj
at at at at
X X X X
= = = =
Xi (5.26) Xk Xm Xj
27/11/14 9:46 AM
310 Chapter 5 OneDimensional Elements T
Ti
Tk Tm Tj
Xi
Xk //3
Xm //3
X
Xj //3
Figure 5.6 Cubic approximation of the temperature distribution for an element.
/
Four equations and four unknowns are created upon substitution of the nodal values into Eq. (5.25). Solving for c1, c2, c3, and c4 and rearranging terms leads to the element’s temperature distribution in terms of the nodal values and the shape functions: T (e) = SiTi + SjTj + SkTk + SmTm (5.27)
In matrix form, the above expression is
T (e) = [Si
Sj
Sk
Ti T Sm] d j t (5.28) Tk Tm
where the shape functions are 9 (X  Xj)(X  Xk)(X  Xm) (5.29) 2/3 9 Sj = 3 (X  Xi)(X  Xk)(X  Xm) 2/ 27 Sk = 3 (X  Xi)(X  Xj)(X  Xm) 2/ 27 Sm =  3 (X  Xi)(X  Xj)(X  Xk) 2/ Si = 
It is worth noting that when the order of the interpolating function increases, it is necessary to employ Lagrange interpolation functions instead of taking the above approach to obtain the shape functions. The main advantage the Lagrange method offers is that using it, we do not have to solve a set of equations simultaneously to obtain the unknown coefficients of the interpolating function. Instead, we represent the shape functions in terms of the products of three linear functions. For cubic interpolating
M05_MOAV4303_04_GE_C05.INDD 310
27/11/14 9:46 AM
Section 5.3 Cubic Elements 311
functions, the shape function associated with each node can be represented in terms of the product of three linear functions. For a given node—for example, i—we select the functions such that their product will produce a value of zero at other nodes—namely, j, k, and m—and a value of unity at the given node, i. Moreover, the product of the functions must produce linear and nonlinear terms similar to the ones given by a general thirdorder polynomial function. To demonstrate this method, let us consider node i, with the global coordinate Xi. First, the functions must be selected such that when evaluated at nodes j, k, and m, the outcome is a value of zero. We select Si = a1(X  Xj)(X  Xk)(X  Xm) (5.30)
which satisfies the above condition. That is, if you substitute for X = Xj, or X = Xk, or X = Xm, the value of Si is zero. We then evaluate a1 such that when the shape function Si is evaluated at node i(X = Xi), it will produce a value of unity: / 2/ 1 = a1(Xi  Xj)(Xi  Xk)(Xi  Xm) = a1( /) ¢  ≤ ¢  ≤ 3 3 Solving for a1, we get a1 = 
9 2/3
and substituting into Eq. (5.30), we have Si = 
9 (X  Xj)(X  Xk)(X  Xm) 2/3
The other shape functions are obtained in a similar fashion. Keeping in mind the explanation offered above, we can generate shape functions of an (N  1)order polynomial directly from the Lagrange polynomial formula: N X  XM omitting (X  XK) (X  X1)(X  X2) g (X  XN) SK = q = (5.31) (XK  X1)(XK  X2) g (XK  XN) M = 1 XK  XM omitting (XK  XK)
Note that in order to accommodate any order polynomial representation in Eq. (5.31) numeral values are assigned to the nodes and the subscripts of the shape functions. In general, using a cubic interpolation function, the variation of any parameter Ψ in terms of its nodal values may be written as
Ψ(e) = [Si
Sj
Sk
Ψi Ψj Sm ] d t Ψk Ψm
Once again, note that the cubic shape functions possess properties similar to those of the linear and the quadratic shape functions; that is, (1) a shape function has a value of unity at its corresponding node and a value of zero at the other adjacent node, and (2) if we sum up the shape functions, we will come up with a value of unity. However, note that taking the spatial derivative of cubic shape functions will produce quadratic results.
M05_MOAV4303_04_GE_C05.INDD 311
27/11/14 9:46 AM
312 Chapter 5 OneDimensional Elements
EXAMPLE 5.2 Using Lagrange interpolation functions, generate the quadratic shape functions directly from the Lagrange polynomial formula, Eq. (5.31): N (X  XM) omitting (X  XK) SK = q (X K  XM) omitting (XK  XK) M=1
For quadratic shape functions, N 1 = 2 and K = 1, 2, 3. Refer to Figure 5.5 and note that subscripts 1, 2, and 3 correspond to nodes i, k, and j respectively. Also, it is important to distinguish the difference between lowercase k, denoting a specific node, and uppercase K, a variable subscript, denoting various nodes. For node i or K = 1, Si = S1 =
(X  X2)(X  X3) (X  X2)(X  X3) 2 = = 2 (X  X2)(X  X3) (X1  X2)(X1  X3) / / ¢  ≤( /) 2
For node k or K = 2, Sk = S2 =
(X  X1)(X  X3) (X  X1)(X  X3) 4 = = 2 (X  X1)(X  X3) (X2  X1)(X2  X3) / / / ¢ ≤¢  ≤ 2 2
For node j or K = 3, Sj = S3 =
(X  X1)(X  X2) (X  X1)(X  X2) 2 = = 2 (X  X1)(X  X2) (X3  X1)(X3  X2) / / (/) ¢ ≤ 2
The results are identical to shape functions given by Eq. (5.23). 5.4 Global, Local, and Natural Coordinates Most often, in finite element modeling, it is convenient to use several frames of reference, as we briefly discussed in Chapters 3 and 4. We need a global coordinate system to represent the location of each node, orientation of each element, and to apply boundary conditions and loads (in terms of their respective global components). Moreover, the solution, such as nodal displacements, is generally represented with respect to the global directions. On the other hand, we need to employ local and natural coordinates because they offer certain advantages when we construct the geometry or compute integrals. The advantage becomes apparent particularly when the integrals contain products of shape functions. For onedimensional elements, the relationship between a global coordinate X and a local coordinate x is given by X = Xi + x, as shown in Figure 5.7.
M05_MOAV4303_04_GE_C05.INDD 312
27/11/14 9:46 AM
Section 5.4 Global, Local, and Natural Coordinates 313
Node i
Global
X
Local
x
Node j
Xi
Xj /
Figure 5.7 The relationship between a global coordinate X and a local coordinate x.
Substituting for X in terms of the local coordinate x in Eqs. (5.8) and (5.9), we get
Si =
Xj  X
=
Xj  (Xi + x)
= 1 
/ / (Xi + x)  Xi X  Xi x Sj = = = / / /
x (5.32) / (5.33)
where the local coordinate x varies from 0 to /; that is 0 … x … /. OneDimensional Linear Natural Coordinates Natural coordinates are basically local coordinates in a dimensionless form. It is often necessary to use numerical methods to evaluate integrals for the purpose of calculating elemental stiffness or conductance matrices. Natural coordinates offer the convenience of having 1 and 1 for the limits of integration. For example, if we let j =
2x  1 /
where x is the local coordinate, then we can specify the coordinates of node i as 1 and node j by 1. This relationship is shown in Figure 5.8. We can obtain the natural linear shape functions by substituting for x in terms of j into Eqs. (5.32) and (5.33). This substitution yields 1 (1  j) (5.34) 2 1 Sj = (1 + j) (5.35) 2
Si =
Natural linear shape functions possess the same properties as linear shape functions; that is, a shape function has a value of unity at its corresponding node and has a value of j
j = 1 Node i
j=1
Local x
Node j
Xi
Xj /
M05_MOAV4303_04_GE_C05.INDD 313
Figure 5.8 The relationship between the local coordinate x and the natural coordinate j.
27/11/14 9:46 AM
314 Chapter 5 OneDimensional Elements
zero at the adjacent node in a given element. As an example, the temperature distribution over an element of a onedimensional fin may expressed by T (e) = SiTi + SjTj =
1 1 (1  j)Ti + (1 + j)Tj (5.36) 2 2
It is clear that at j = 1, T = Ti and at j = 1, T = Tj. 5.5 Isoparametric Elements By now, it should be clear that we can represent other variables, such as the displacement u, in terms of the natural shape functions Si and Sj according to the equation u(e) = Siui + Sjuj =
1 1 (1  j)ui + (1 + j)uj (5.36a) 2 2
Also note that the transformation from the global coordinate X(Xi … X … Xj) or the local coordinate x(0 … x … /) to j can be made using the same shape functions Si and Sj. That is, X = SiXi + SjXj =
1 1 (1  j)Xi + (1 + j)Xj (5.36b) 2 2
or x = Si xi + Sj xj =
1 1 (1  j)xi + (1 + j)xj 2 2
Comparing the relationships given by Eqs. (5.36), (5.36a), and (5.36b), we note that we have used a single set of parameters (such as Si, Sj) to define the unknown variables u, T, and so on, and we used the same parameters (Si, Sj) to express the geometry. Finite element formulation that makes use of this idea is commonly referred to as isoparametric (iso meaning the same or uniform) formulation, and an element expressed in such a manner is called an isoparametric element. We discuss isoparametric formulation further in Chapters 7 and 10. Example 5.3 Determine the temperature of the fin in Example 5.1 at the global location X = 8 cm using local coordinates. Also determine the temperature of the fin at the global location X = 7.5 cm using natural coordinates. Tfluid = 185C Tbase = 505C 1
(1) 2 cm
M05_MOAV4303_04_GE_C05.INDD 314
2
(2) 3 cm
3
(3) 5 cm
4
T1 50 T2 41 d t = d t °C T3 34 T4 20
27/11/14 9:46 AM
Section 5.5 Isoparametric Elements 315
a. Using local coordinates, we find that the temperature of the fin at X = 8 cm is represented by element (3) according to the equation (3) T (3) = S(3) 3 T3 + S4 T4 = ¢ 1 
x x ≤T + T4 / 3 /
Note that element (3) has a length of 5 cm, and the location of a point 8 cm from the base is represented by the local coordinate x = 3: T = ¢1 
3 3 ≤(34) + (20) = 25.65C 5 5
b. Using natural coordinates, we find that the temperature of the fin at X = 7.5 cm is represented by element (3) according to the equation T (3) = S(3) T3 + S(3) T4 = 3 4
1 1 (1  j)T3 + (1 + j)T4 2 2
Because the point with the global coordinate X = 7.5 cm is located in the middle of element (3), the natural coordinate of this point is given by j = 0: T(3) =
1 1 (1  0)(34) + (1 + 0)(20) = 275C 2 2
OneDimensional Natural Quadratic and Cubic Shape Functions The natural onedimensional quadratic and cubic shape functions can be obtained in a way similar to the method discussed in the previous section. The quadratic natural shape functions are 1 Si =  j(1  j) (5.37) 2 1 Sj = j(1 + j) (5.38) 2 Sk = (1 + j)(1  j) (5.39)
The natural onedimensional cubic shape functions are
M05_MOAV4303_04_GE_C05.INDD 315
1 (1  j)(3j + 1)(3j  1) (5.40) 16 1 Sj = (1 + j)(3j + 1)(3j  1) (5.41) 16 9 Sk = (1 + j)(j  1)(3j  1) (5.42) 16 9 Sm = (1 + j)(1  j)(3j + 1) (5.43) 16 Si =
27/11/14 9:46 AM
316 Chapter 5 OneDimensional Elements
For the sake of convenience, the results of Sections 5.1 to 5.4 are summarized in Table 5.1. Make sure to distinguish the differences among presentations of the shape functions using global, local, and natural coordinates. Example 5.4 Xj
Evaluate the integral
LXi
S2j dX using (a) global coordinates and (b) local coordinates.
a. Using global coordinates, we obtain Xj
LXi
Xj
LXi
S2j dX =
¢
Xj X  Xi 2 1 / ≤ dX = 2 (X  Xi)3 ` = / 3 3/ Xi
b. Using local coordinates, we obtain Xj
LXi
S2j dX =
/
x 2 x3 / / ¢ ≤ dx = 2 ` = 3 3/ 0 L0 /
This simple example demonstrates that local coordinates offer a simple way to evaluate integrals containing products of shape functions. 5.6 Numerical Integration: Gauss–Legendre Quadrature As we discussed earlier, natural coordinates are basically local coordinates in a dimensionless form. Moreover, most finite element programs perform element numerical integration by Gaussian quadratures, and as the limit of integration, they use an interval from 1 to 1. This approach is taken because when the function being integrated is known, the Gauss–Legendre formulae offer a more efficient way of evaluating an integral as compared to other numerical integration methods such as the trapezoidal method. Whereas the trapezoidal method or Simpson’s method can be used to evaluate integrals dealing with discrete data (see Problem 24), the Gauss–Legendre method is based on the evaluation of a known function at nonuniformally spaced points to compute the integral. The twopoint Gauss–Legendre formula is developed next in this section. The basic goal behind the Gauss–Legendre formulae is to represent an integral in terms of the sum of the product of certain weighting coefficients and the value of the function at some selected points. So, we begin with
I =
L a
f(x)dx = a wi f(xi) (5.44)
b
n
i=1
Next, we must ask (1) How do we determine the value of the weighting coefficients, represented by the wi >s? (2) Where do we evaluate the function, or in other words, how do we select these points (xi)? We begin by changing the limits of integration from a to b to 1 to 1 with the introduction of the variable l such that x = c0 + c1l
M05_MOAV4303_04_GE_C05.INDD 316
27/11/14 9:46 AM
317
M05_MOAV4303_04_GE_C05.INDD 317
27/11/14 9:46 AM
Cubic
Quadratic
Linear
Interpolation function
27 (X  Xi)(X  Xj)(X  Xk) 2/3
Sm =
Sk =
27 (X  Xi)(X  Xj)(X  Xm) 2/3
Sk =
Sm = 
Sj =
9 (X  Xi)(X  Xk)(X  Xm) 2/3
Si =
Sj =
9 (X  Xj)(X  Xk)(X  Xm) 2/3
9 x x x ¢ ≤ ¢3¢ ≤  1≤ ¢1  ¢ ≤ ≤ 2 / / /
9 x x x ¢ ≤ ¢2  3¢ ≤ ≤ ¢1  ¢ ≤ ≤ 2 / / /
1 x x x ¢ ≤ ¢2  3¢ ≤ ≤ ¢1  3¢ ≤ ≤ 2 / / /
1 x x x ¢1  ≤ ¢2  3¢ ≤ ≤ ¢1  3¢ ≤ ≤ 2 / / /
x x Sk = 4 ¢ ≤ ¢ 1  ¢ ≤ ≤ / /
4 (X  Xi)(X  Xj) /2
Sk =
Si = 
x x Sj = ¢ ≤ ¢ 2 ¢ ≤  1 ≤ / /
2 (X  Xi)(X  Xk) /2
x x  1≤ ¢2¢ ≤  1≤ / /
Sj =
x /
x /
Si = ¢
Sj =
Si = 1 
2 (X  Xj)(X  Xk) /2
/
/ X  Xi
Xj  X
In terms of local coordinate x 0 … x … /
Si =
Sj =
Si =
In terms of global coordinate X Xi … X … Xj
Table 5.1 Onedimensional shape functions
1 j(1 + j) 2
Sm =
Sk =
Sj =
Si =
9 (1 + j)(1  j)(3j + 1) 16
9 (1 + j)(j  1)(3j  1) 16
1 (1 + j)(3j + 1)(3j  1) 16
1 (1  j)(3j + 1)(3j  1) 16
Sk = (1  j)(1 + j)
Sj =
1 Si =  j(1  j) 2
1 (1  j) 2 1 Sj = (1 + j) 2 Si =
In terms of natural coordinate j 1 … j … 1
318 Chapter 5 OneDimensional Elements
Matching the limits, we get a = c0 + c1( 1) b = c0 + c1(1) and solving for c0 and c1, we have c0 =
(b + a) 2
c1 =
(b  a) 2
and
Therefore,
x =
(b + a) (b  a) + l (5.45) 2 2
and
dx =
(b  a) dl (5.46) 2
Thus, using Eqs. (5.45) and (5.46), we find that any integral in the form of Eq. (5.44) can be expressed in terms of an integral with its limits at 1 and 1: 1
I =
L1
f(l) dl = a wi f(li) (5.47) n
i=1
The twopoint Gauss–Legendre formulation requires the determination of two weighting factors w1 and w2 and two sampling points l1 and l2 to evaluate the function at these points. Because there are four unknowns, four equations are created using Legendre polynomials (1, l, l2, l3) as follows: 1
w1 f(l1) + w2 f(l2) =
L1
1 dl = 2
1
w1 f(l1) + w2 f(l2) =
L1
l dl = 0
1
w1 f(l1) + w2 f(l2) =
L1
l2 dl =
2 3
1
w1 f(l1) + w2 f(l2) =
L1
l3 dl = 0
The above equations lead to the equations w1(1) + w2(1) = 2 w1(l1) + w2(l2) = 0
M05_MOAV4303_04_GE_C05.INDD 318
27/11/14 9:46 AM
Section 5.6 Numerical Integration: Gauss–Legendre Quadrature 319
Table 5.2 Weighting factors and sampling points for Gauss–Legendre formulae Points (n)
Weighting factors (wi)
Sampling points (li)
2
w1 = 1.00000000 w2 = 1.00000000
l1 =  0.577350269 l2 = 0.577350269
3
w1 = 0.55555556 w2 = 0.88888889 w3 = 0.55555556
l1 =  0.774596669 l2 = 0 l3 = 0.774596669
4
w1 w2 w3 w4
= = = =
0.3478548 0.6521452 0.6521452 0.3478548
l1 l2 l3 l4
=  0.861136312 =  0.339981044 = 0.339981044 = 0.861136312
5
w1 w2 w3 w4 w5
= = = = =
0.2369269 0.4786287 0.5688889 0.4786287 0.2369269
l1 l2 l3 l4 l5
=  0.906179846 =  0.538469310 = 0 = 0.538469310 = 0.906179846
2 3 3 3 w1(l1) + w2(l2) = 0 w1(l1)2 + w2(l2)2 =
Solving for w1, w2, l1, and l2, we have w1 = w2 = 1, l1 = 0.577350269, and l2 = 0.577350269. The weighting factors and the 2, 3, 4, and 5 sampling points for Gauss–Legendre formulae are given in Table 5.2. Note that as the number of sampling points increases, so does the accuracy of the calculations. As you will see in Chapter 7, we can readily extend the Gauss–Legendre quadrature formulation to two or three dimensional problems. EXAMPLE 5.5 Evaluate the integral I =
L2
6
(x2 + 5x + 3) dx using the Gauss–Legendre twopoint
sampling formula. This integral is simple and can be evaluated analytically, leading to the solution I = 161.333333333. The purpose of this example is to demonstrate the Gauss–Legendre procedure. We begin by changing the variable x to l by using Eq. (5.45). So, we obtain x =
(b + a) (b  a) (6 + 2) (6  2) + l = + l = 4 + 2l 2 2 2 2
and dx =
M05_MOAV4303_04_GE_C05.INDD 319
(b  a) (6  2) dl = dl = 2 dl 2 2
27/11/14 9:46 AM
320 Chapter 5 OneDimensional Elements
Thus, the integral I can be expressed in terms of l: f(l)
I =
L2
6
y
u
f(x) 1
(x2 + 5x + 3)dx =
L1
(2)[(4 + 2l)2 + 5(4 + 2l) + 3)]dl
Using the Gauss–Legendre twopoint formula and Table 5.2, we compute the value of the integral I from I ≅ w1 f(l1) + w2 f(l2) From Table 5.2, we find that w1 = w2 = 1, and evaluating f(l) at l1 = 0.577350269 and l2 = 0.577350269, we obtain f(l1) = (2)[[4 + 2( 0.577350269)]2 + 5(4 + 2( 0.577350269) + 3)] = 50.6444526769 f(l2) = (2)[[4 + 2(0.577350269)]2 + 5(4 + 2(0.577350269) + 3)] = 110.688880653 I = (1)(50.6444526769) + (1)110.688880653 = 161.33333333 EXAMPLE 5.6 Xj
Evaluate the integral
LXi
S2j dX in Example 5.4 using the Gauss–Legendre twopoint
formula. Recall from Eq. (5.35) that Sj = 12(1 + j) and by differentiating the relationship 2x between the local coordinate x and the natural coordinate j (i.e., j =  11 / 2 dj = dx) we find dx = 2/ dj. Also note that for this problem, j = l. So, / Xj
I =
LXi
S2j dX =
Xj
LXi
¢
/ 2 X  Xi 2 x 2 / 1 1 ≤ dX = ¢ ≤ dx = J (1 + j) R dj / 2 L1 2 L0 /
Using the Gauss–Legendre twopoint formula and Table 5.2, we compute the value of the integral I from I ≅ w1 f(l1) + w2 f(l2) From Table 5.2, we find that w1 = w2 = 1, and evaluating f(l) at l1 = 0.577350269 and l2 = 0.577350269, we obtain 2
2
2
2
f(j1) =
/ 1 / 1 J (1 + j1) R = J (1  0.577350269) R = 0.022329099389/ 2 2 2 2
f(j2) =
/ 1 / 1 J (1 + j2) R = J (1 + 0.577350269) R = 0.31100423389/ 2 2 2 2
I = (1)(0.022329099389/) + (1)(0.31100423389/) = 0.333333333/ Note that the above result is identical to the results of Example 5.4.
M05_MOAV4303_04_GE_C05.INDD 320
27/11/14 9:46 AM
Chapter 5 Problems 321
5.7 Examples of OneDimensional Elements in ANSYS ANSYS offers uniaxial link elements that may be used to represent onedimensional problems. These link elements include LINK31, LINK33, and LINK34. The LINK33 element is a uniaxial heat conduction element. It allows for the transfer of heat between its two nodes via conduction mode. The nodal degree of freedom associated with this element is temperature. The element is defined by its two nodes, crosssectional area, and material properties such as thermal conductivity. The LINK34 element is a uniaxial convection link that allows for heat transfer between its nodes by convection. This element is defined by its two nodes, a convective surface area, and a convective heat transfer (film) coefficient. The LINK31 element can be used to model radiation heat transfer between two points in space. The element is defined by its two nodes, a radiation surface area, a geometric shape factor, emissivity, and the Stefan–Boltzman constant. In Chapter 6, we will use LINK33 and LINK34 to solve a onedimensional heatconduction problem. Summary At this point you should 1. have a good understanding of the linear onedimensional elements and shape functions, their properties, and their limitations. 2. have a good understanding of the quadratic and cubic onedimensional elements and shape functions, their properties, and their advantages over linear elements. 3. know why it is important to use local and natural coordinate systems. 4. know what is meant by isoparametric element and formulation. 5. have a good understanding of Gauss–Legendre quadrature. 6. know examples of onedimensional elements in ANSYS. References ANSYS User’s Manual: Elements, Vol. III, Swanson Analysis Systems, Inc. Chandrupatla, T., and Belegundu, A., Introduction to Finite Elements in Engineering, Prentice Hall, 1991. Incropera, F. P., and DeWitt, D. P., Fundamentals of Heat and Mass Transfer, 2nd ed., New York, John Wiley and Sons, 1985. Segrlind, L., Applied Finite Element Analysis, 2nd ed., New York John Wiley and Sons, 1984.
Problems 1. We have used linear onedimensional elements to approximate the temperature distribution along a fin. The nodal temperatures and their corresponding positions are shown in the accompanying figure. (a) What is the temperature of the fin at X = 8 cm? (b) Evaluate the heat loss from the fin using the relationship Q = kA where k = 200 W/m # K and A = 15 mm2.
M05_MOAV4303_04_GE_C05.INDD 321
dT 0 dX X = 0
27/11/14 9:46 AM
322 Chapter 5 OneDimensional Elements
Tfluid = 185C
T1 100 T2 800 d t = d t °C T3 64 T4 41
Tbase = 1005C 1
(1)
2
2 cm
2. Evaluate the integral (b) local coordinates.
3
(2) 3 cm
(3)
4
5 cm
Xj
LX
S2i dX for a linear shape function using (a) global coordinates and i
3. Starting with the equations Ti = c1 + c2Xi + c3X i2 Tk = c1 + c2Xk + c3X k2 Tj = c1 + c2Xj + c3X j2 solve for c1, c2, and c3, and rearrange terms to verify the shape functions given by Si =
2 (X  Xj)(X  Xk) /2
Sj =
2 (X  Xi)(X  Xk) /2
Sk =
4 (X  Xi)(X  Xj) /2
4. For Problem 3, use the Lagrange functions to derive the quadratic shape functions by the method discussed in Section 5.3.
5. Derive the expressions for quadratic shape functions in terms of the local coordinates and compare your results to the results given in Table 5.1.
6. Verify the results given for onedimensional quadratic natural shape functions in Table 5.1 by showing that (1) a shape function has a value of unity at its corresponding node and a value of zero at the other nodes, and (2) if we sum up the shape functions, we will come up with a value of unity.
7. Verify the results given for the local cubic shape functions in Table 5.1 by showing that (1) a shape function has a value of unity at its corresponding node and a value of zero at the other nodes and (2) if we sum up the shape functions, we will come up with a value of unity.
8. Verify the results given for the natural cubic shape functions in Table 5.1 by showing that (1) a shape function has a value of unity at its corresponding node and a value of zero at the other nodes and (2) if we sum up the shape functions, we will come up with a value of unity.
M05_MOAV4303_04_GE_C05.INDD 322
27/11/14 9:46 AM
Chapter 5 Problems 323
9. Obtain expressions for the spatial derivatives of the quadratic and cubic shape functions. 10. As previously explained, we can increase the accuracy of our finite element findings either by increasing the number of elements used in the analysis to represent a problem or by using a higher order approximation. Derive the local cubic shape functions. Xj
Si dX for a quadratic shape function using (a) global coordinates, LXi (b) natural coordinates, and (c) local coordinates. 12. Assume that the deflection of a cantilever beam was approximated with linear onedimensional elements. The nodal deflections and their corresponding positions are shown in the accompanying figure. (a) What is the deflection of the beam at X = 2.5 ft? (b) Evaluate the slope at the endpoint.
11. Evaluate the integral
100 lb/ft x EI = 2 * 106 lb · ft2 y 1
2
3
1 ft
4
2 ft
5
1 ft
1 ft
y1 0 y2 0.003275 e y3 u = e 0.022275 u in y4 0.034400 y5 0.046875
13. We have used linear onedimensional elements to approximate the temperature distribution inside a metal plate. A heating element is embedded within a plate. The nodal temperatures and their corresponding positions are shown in the accompanying figure. What is the temperature of the plate at X = 28 mm? Assume that (a) linear elements were used in obtaining nodal temperatures and (b) quadratic elements were used.
I 8 cm R 10 mm
1 I
M05_MOAV4303_04_GE_C05.INDD 323
2 X
3
4
5
6
7
8
9
T1 120 T2 119 T3 116 T4 111 i T5 y = i 104 y °C T6 95 T7 84 T8 71 T9 56
27/11/14 9:46 AM
324 Chapter 5 OneDimensional Elements
14. Quadratic elements are used to approximate the temperature distribution in a straight fin. The nodal temperatures and their corresponding positions are shown in the accompanying figure. What is the temperature of the fin at X = 8 cm? T1 100 74 T2 T3 56 g T4 w = g 44 w °C 36 T5 31 T6 T7 28
Tbase = 1005C
2
1 2 cm
4
3 2 cm
6
5
2 cm
2 cm
2 cm
7 2 cm
15. Develop the shape functions for a linear element, shown in the accompanying figure, using the local coordinate x whose origin lies at the onefourth point of the element. x i
j //3 / Xi
Xj
16. Using the natural coordinate system shown in the accompanying figure, develop the natural shape functions for a linear element. j i
j
Xi
Xj
/
17. In the accompanying figure, the deflection of nodes 2 and 3 are 0.02 mm and 0.025 mm, respectively. What are the deflections at point A and point B, provided that linear elements were used in the analysis? 6 cm
3 cm
1
12 cm
M05_MOAV4303_04_GE_C05.INDD 324
3
2 A
4
B
6 cm
8 cm
27/11/14 9:47 AM
Chapter 5 Problems 325
18. Consider the steel column depicted in the accompanying figure. Under the assumption of axial loading, and using linear elements, we determined that the downward vertical displacements of the column at various floor–column connection points are u1 0 u2 0.03283 e u3 u = e 0.05784 u in u4 0.07504 u5 0.08442 Using local shape functions, determine the deflections of points A and B, located in the middle of elements (3) and (4) respectively. 30,000 lb 5
25,000 lb
(4) B 4
15 ft
25,000 lb
(3) A 3
15 ft
25,000 lb
(2)
15 ft
2
(1)
15 ft
1
19. Determine the deflection of points A and B on the column in Problem 18 using natural coordinates. 20. A 20fttall post is used to support advertisement signs at various locations along its height, as shown in the accompanying figure. The post is made of structural steel with a modulus of elasticity of E = 29 * 106 lb/in2. Not considering wind loading on the signs and using linear elements, we determined that the downward deflections of the post at the points of load application are u1 0 u2 6.2069 * 104 d t = d t in u3 8.7931 * 104 4 u4 10.8621 * 10
M05_MOAV4303_04_GE_C05.INDD 325
27/11/14 9:47 AM
326 Chapter 5 OneDimensional Elements Determine the deflection of point A, located at the midpoint of the middle member, using (a) global shape functions, (b) local shape functions, and (c) natural shape functions. 100 lb
4 5 ft
150 lb
Ac = 1 in2
3 A 200 lb
5 ft
Ac = 2 in2
2
10 ft
Ac = 3 in2
1
21. Evaluate the integral in Problem 11 using Gauss–Legendre twopoint formula. 22. Evaluate the given integral analytically and using Gauss–Legendre formula. L1
5
(x3 + 6x2 + 10)dx
23. Evaluate the given integral analytically and using Gauss–Legendre two and threepoint formulae. 8
(3x4 + x2  7x + 10)dx L2
24. In Section 5.6 we mentioned that when the function being integrated is known, the Gauss– Legndre formulae offer a more efficient way of evaluating an integral as compared to the trapezoidal method. The trapezoidal approximation of an integral deals with discrete data at uniformly spaced intervals and is computed from La
M05_MOAV4303_04_GE_C05.INDD 326
b
1 1 f (x)dx ≈ h¢ y0 + y1 + y2 + g + yn  2 + yn  1 + yn ≤ 2 2
27/11/14 9:47 AM
Chapter 5 Problems 327 The above equation is known as the trapezoidal rule, and h represents the spacing among data points, y0, y1, y2, c, and yn. For Problem 23, use h = 1 and generate 11 data points: y0 at x = 2, y1 at x = 1, c, and y10 at x = 8. Use the generated data and the trapezoidal rule to approximate the integral. Compare this result to the results of Problem 23. y = f(x)
y1
y0 yn h
yn  1
h x
a
b
25. Derive and plot the spatial derivatives for linear, quadratic, and cubic elements. Discuss the differences.
26. Temperature distribution along a section of a material is given by T0 = 80°C, T1 = 70°C,
T2 = 62°C, T3 = 55°C with X0 = 0, X1 = 2 cm, X2 = 4 cm, X3 = 6 cm. Use single linear, quadratic, and cubic elements to approximate the given temperature distribution. Plot the actual data points and compare them to the linear, quadratic, and cubic approximations. Also, estimate the spatial derivative of the given data and compare it to the derivatives of the linear, quadratic, and cubic representations. 27. Use the isoparametric formulation to express the following information: T1 = 100°C, T2 = 605C with the corresponding global coordinates X1 = 2 cm, X2 = 5 cm. Show the transformation equations between the global coordinate, the local coordinate, and the natural coordinate. 28. Use the isoparametric formulation to express the following deflection information: U1 = 0.01 cm, U2 = 0.02 cm with the corresponding global coordinates X1 = 2 cm, X2 = 12 cm. Show the transformation equations between the global coordinate, the local coordinate, and the natural coordinate.
M05_MOAV4303_04_GE_C05.INDD 327
27/11/14 9:47 AM
C h a p t e r
6
Analysis of OneDimensional Problems
The main objective of this chapter is to introduce the analysis of onedimensional problems. Most often, a physical problem is not truly onedimensional in nature; however, as a starting point, we may model the behavior of a system using onedimensional approximation. This approach can usually provide some basic insight into a more complex problem. If necessary, as a next step we can always analyze the problem using a two or threedimensional approach. This chapter first presents the onedimensional Galerkin formulation used for heat transfer problems. This presentation is followed by an example demonstrating analysis of a onedimensional fluid mechanics problem. The main topics discussed in Chapter 6 are 6.1 Heat Transfer Problems 6.2 A Fluid Mechanics Problem 6.3 An Example Using ANSYS 6.4 Verification of Results
6.1 Heat Transfer Problems Recall that in Chapter 1 we discussed the basic steps involved in any finite element analysis; to refresh your memory, these steps are repeated here. Preprocessing Phase 1. Create and discretize the solution domain into finite elements; that is, subdivide the problem into nodes and elements. 2. Assume shape functions to represent the behavior of an element; that is, assume an approximate continuous function to represent the solution for a element. The onedimensional linear and quadratic shape functions were discussed in Chapter 5.
M06_MOAV4303_04_GE_C06.INDD 328
27/11/14 9:50 AM
Section 6.1 Heat Transfer Problems 329
3. Develop equations for an element. This step is the main focus of the current chapter. We will use the Galerkin approach to formulate elemental descriptions. In Chapter 4, we used the minimum potential energy theorem to generate finite element models for members under axial loading and for beam elements. 4. Assemble the elements to represent the entire problem. Construct the global stiffness or conductance matrix. 5. Apply boundary conditions and loading. Solution Phase 6. Solve a set of linear algebraic equations simultaneously to obtain nodal results, such as the temperature at different nodes or displacements. Postprocessing Phase 7. Obtain other important information. We may be interested in determining the heat loss or stress in each element. We now focus our attention on step 3 of the preprocessing phase. We formulate the conductance and the thermal load matrices for a typical onedimensional fin element. We considered a straight fin of a uniform cross section in Chapter 5. For the sake of convenience, the fin is shown again in Figure 6.1. The fin is modeled using three elements and four nodes. The temperature distribution along the element is interpolated using linear functions. The actual and the approximate piecewise linear temperature distribution along the fin are shown in Figure 6.1. We will concentrate on a typical element belonging to the fin and formulate the conductance matrix and the thermal load matrix for such an element. Onedimensional heat transfer in a straight fin is governed by the following heat equation, as given in any introductory text on heat transfer:
kA
d2 T  hpT + hpTf = 0 (6.1) dX 2
T
Tbase
(2)
Base
(1) 1
M06_MOAV4303_04_GE_C06.INDD 329
2
Tfluid
(3) 3
X 4
Figure 6.1 The actual and approximate temperature distribution for a fin of uniform cross section.
27/11/14 9:51 AM
330 Chapter 6 Analysis of OneDimensional Problems dqconvection
qx
Tf , h
qx + dx dx
Tb
A (crosssectional area)
W
x
L We start by applying the energy balance to a differential element qx = qx + dx + dqconvection qx = qx +
dqx dx
dx + dqconvection
Next we use Fourier’s law qx =  kA
dT dx
and use Newton’s law of cooling, dqconvection = h(dAs)(T  Tf) 0 =
dqx dx
dx + dqconvection =
d dT ¢  kA ≤ dx + h(dAs)(T  Tf) dx dx
Writing dAs (differential surface area) in terms of the perimeter of the fin and dx and simplifying, we are left with  kA
d2T + hp(T  Tf) = 0 dx2
Figure 6.2 Derivation of the heat equation for a fin.
Equation (6.1) is derived by applying the conservation of energy to a differential section of a fin, as shown in Figure 6.2. The heat transfer in the fin is accomplished by conduction in the longitudinal direction (xdirection) and convection to the surrounding fluid. In Eq. (6.1), k is the thermal conductivity, and A denotes the crosssectional area of the fin. The convective heat transfer coefficient is represented by h, the perimeter of the fin is denoted by p, and Tf is the temperature of the surrounding fluid. Equation (6.1) is subjected to a set of boundary conditions. First, the temperature of the base is generally known; that is,
M06_MOAV4303_04_GE_C06.INDD 330
T(0) = Tb (6.2)
27/11/14 9:51 AM
Section 6.1 Heat Transfer Problems 331
The other boundary condition deals with the heat loss at the tip of the fin. In general, there are three possibilities. One possibility is that the tip is long enough so that the temperature of the tip is equal to the temperature of the surrounding fluid temperature. This situation is represented by the condition
T(L) = Tf (6.3) The situation in which the heat loss from the tip of the fin may be neglected is represented by the condition kA
dT ` = 0 (6.4) dX X = L
If the heat loss from the tip of the fin should be included in the analysis, then we have the condition kA
dT ` = hA(TL  Tf) (6.5) dX X = L
Equation (6.5) is obtained by applying the energy balance to the crosssectional area of the tip. Equation (6.5) simply states that the heat conducted to the tip’s surface is convected away by the surrounding fluid. Therefore, we can use one of the boundary conditions given by Eqs. (6.3)–(6.5) and the base temperature to model an actual problem. Before we proceed with the formulation of the conductance matrix and the thermal load matrix for a typical element, let us emphasize the following points: (1) The governing differential equation of the fin represents the balance of energy at any point along the fin and thus governs the balance of energy at all nodes of a finite element model as well, and (2) the exact solution of the governing differential equation (if possible) subject to two appropriate boundary conditions renders the detailed temperature distribution along the fin, and the finite element solution represents an approximation of this solution. We now focus on a typical element and proceed with the formulation of the conductance matrix, recalling that the temperature distribution for a typical element may be approximated using linear shape functions, as discussed in Chapter 5. That is, T (e) = [Si
Sj] b
Ti r (6.6) Tj
where the shape functions are given by
Si =
Xj  X /
and Sj =
X  Xi (6.7) /
In order to make this derivation as general as possible and applicable to other type of problems with the same form of differential equations, let c1 = kA, c2 = hp, c3 = hpTf, and Ψ = T. Thus, Eq. (6.1) can be written as
M06_MOAV4303_04_GE_C06.INDD 331
c1
d2 Ψ + c2Ψ + c3 = 0 (6.8) dX2
27/11/14 9:51 AM
332 Chapter 6 Analysis of OneDimensional Problems
Recall from our introductory discussion of weighted residual methods in Chapter 1 that when we substitute an approximate solution into the governing differential equation, the approximate solution does not satisfy the differential equation exactly, and thus, an error, or a residual, is produced. Also recall that the Galerkin formulation requires the error to be orthogonal to some weighting functions. Furthermore, the weighting functions are chosen to be members of the approximate solution. Here we will use the shape functions as the weighting functions because they are members of the approximate solution. We can obtain the residual equations in one of two ways: using a nodal approach or an elemental approach. Consider three consecutive nodes, i, j, and k, belonging to two adjacent elements (e), and (e + 1), as shown in Figure 6.3. Elements (e) and (e + 1) both contribute to error at node j. Realizing this fact, we can write the residual equation for node j as Xj
+ 1) Rj = R(e) + R(e = j j
+
LXi X k
LXj
S(e) Jc1 j
(e) d2 Ψ + c Ψ + c R dX 2 3 dX2 2
Ψ + c2Ψ + c3 R dX2
d + 1) S(e Jc1 j
(e + 1)
dX = 0 (6.9)
Pay close attention to the subscripts denoting node numbers and superscripts referring to element numbers while following the forthcoming derivation. Writing the residual or the error equations for each node of a finite element model leads to a set of equations of the form R1 0 R2 0 e R3 u = e 0 u (6.10) # 0 Rn 0
where R2 = R(1) + R(2) , R3 = R(2) + R(3) , and so on. 2 2 3 3 Let us now look at the contribution of each element to the nodal residual equations more closely. Expanding Eq. (6.9) for a number of nodes (1, 2, 3, 4, and so on), we get
R1 =
(e) i
M06_MOAV4303_04_GE_C06.INDD 332
X2
LX1
(1)
S1 Jc1
(1) d2 Ψ + c Ψ + c R dX = 0 (6.11) 2 3 dX2
(e + 1) j
k
Figure 6.3 Elements (e) and (e + 1) and their respective nodes.
27/11/14 9:51 AM
Section 6.1 Heat Transfer Problems 333
(2) R2 = R(1) 2 + R2 =
X2
LX1
S(1) 2 Jc1
(1) d2 Ψ + c Ψ + c R dX 2 3 dX2
element 2 contribution
$++++++1++++%++++++++1++&
X3
+
LX2
S(2) 2 Jc1
(2) d2 Ψ + c Ψ + c R dX = 0 2 3 dX 2
(6.12)
element 2 contribution
$++++++1++++%++++++++1++& (3) R3 = R(2) 3 + R3 =
X3
LX2
S(2) 3 Jc1
(2) d2 Ψ + c2Ψ + c3 R dX 2 dX
element 3 contribution
$++++++1++++%++++++++1++&
X4
+
LX3
S(3) 3 Jc1
(3) d2 Ψ + c2Ψ + c3 R dX = 0 2 dX
(6.13)
element 3 contribution
$++++++1++++%++++++++1++& (4) R4 = R(3) 4 + R4 =
X4
LX3
S(3) 4 Jc1
(3) d2 Ψ + c Ψ + c R dX 2 3 dX2
element 4 contribution
$++++++1++++%++++++++1++&
X5
+
LX4
S(4) 4 Jc1
(4) d2 Ψ + c2Ψ + c3 R dX = 0 2 dX
(6.14)
and so on. Note that element (2) contributes to Eqs. (6.12) and (6.13), and element (3) contributes to Eqs. (6.13) and (6.14), and so on. In general, an arbitrary element (e) having nodes i and j contributes to the residual equations in the following manner: Xj
R(e) = i
R(e) = j
Si(e) Jc1
(e) d2 Ψ + c2Ψ + c3 R dX (6.15) 2 dX
Sj(e) Jc1
(e) d2 Ψ + c Ψ + c R dX (6.16) 2 3 dX 2
LXi
Xj
LXi
This approach leads to the elemental formulation. It is important to note that we have set up the residual equations for an arbitrary element (e). Moreover, the elemental matrices obtained in this manner are then assembled to present the entire problem, and the residuals equations are set equal to zero. Evaluation of the integrals given by Eqs. (6.15) and (6.16) will result in the elemental formulation. But first, we will manipulate the secondorder terms into
M06_MOAV4303_04_GE_C06.INDD 333
27/11/14 9:51 AM
334 Chapter 6 Analysis of OneDimensional Problems
firstorder terms. This manipulation is accomplished by using the chain rule in the following manner: dSi dΨ dΨ d2 Ψ d ¢ Si ≤ = Si + dX dX dX dX dX2
Si
(6.17)
dSi dΨ d2 Ψ d dΨ = ¢ Si ≤ (6.18) 2 dX dX dX dX dX
Substituting Eq. (6.18) into Eq. (6.15), we obtain R(e) i
=
X j
LXi
dSi dΨ d dΨ ¢S ≤ ≤ + Si(c2Ψ + c3) ≤ dX (6.19) dX i dX dX dX
¢ c1 ¢
We eventually need to follow the same procedure for Eq. (6.16) as well, but for now let us focus only on one of the residual equations. There are four terms in Eq. (6.19) that need to be evaluated: R(e) = i
X j
LXi +
c1 ¢
Xj dSi dΨ d dΨ ¢ Si ≤ ≤ dX + c1 ¢ ≤ dX dX dX dX dX LXi
Xj
LXi
Si(c2Ψ) dX +
Xj
LXi
Si c3 dX
(6.20)
Considering and evaluating the first term, we have
Xj
LXi
c1 ¢
d dΨ dΨ dΨ dΨ ¢ Si ≤ ≤ dX = c1Si ` c1Si ` = c1 ` (6.21) dX dX dX X = Xj dX X = Xi dX X = Xi
It is important to realize that in order for us to obtain the result given by Eq. (6.21), Si is zero at X = Xj and Si = 1 at X = Xi. The second integral in Eq. (6.20) is evaluated as Xj
LXi
c1 ¢ 
dSi dΨ c1 ≤ dX =  (Ψi  Ψj) (6.22) dX dX /
To obtain the results given by Eq. (6.22), first substitute for Si = Ψ = SiΨi + SjΨj =
Xj  X /
Xj  X /
and
X  Xi Ψj and then proceed with evaluation of /
Ψi +
integral. Evaluation of the third and the fourth integrals in Eq. (6.20) yields
M06_MOAV4303_04_GE_C06.INDD 334
Xj
LXi
Si(c2Ψ) dX = Xj
LXi
c2 / c2 / Ψi + Ψ (6.23) 3 6 j
Sic3 dX = c3
/ 2
(6.24)
27/11/14 9:51 AM
Section 6.1 Heat Transfer Problems 335
In exactly the same manner, we can evaluate the second residual equation for node j, as given by Eq. (6.16). This evaluation results in the following equations: Xj
LXi
d dΨ dΨ 2 dΨ 2 dΨ 2 ¢S ≤ ≤ dX = c1Sj c S = c1 (6.25) dX j dX dX X = X 1 j dX X = Xi dX X = Xj
c1 ¢
j
Xj
LXi
c1 ¢ 
dSj dΨ dX dX
Xj
LXi
S j (c2 Ψ) dX = Xj
LXi
c1 ( Ψi + Ψj) /
(6.26)
c2 / c2 / Ψi + Ψ 6 3 j
(6.27)
≤ dX = 
Sj c3 dX = c3
/ 2
(6.28)
It should be clear by now that evaluation of Eqs. (6.15) and (6.16) results in two sets of linear equations, as given by: dΨ 2 c1 R dX X = Xi 1 b ir = d t c Rj dΨ 2 / 1 c1 dX X = Xj c1
+
c2 / 2 J 6 1
1 Ψ d b i r 1 Ψj
c3 / 1 1 Ψ R b ir + b r 2 Ψj 2 1
(6.29)
Note that from here on, for the sake of simplicity of presentation of conductance and load matrices, we will set the residual of an element equal to zero. However, as we mentioned earlier, it is important to realize that the residuals are set equal to zero after all the elements have been assembled. Therefore, we rewrite the Eq. (6.29) as dΨ 2 c1 dX X = Xi 1 d t + c dΨ 2 / 1 c1 dX X = Xj c1
c2/ 2 1 Ψi db r + J 1 Ψj 6 1
c3/ 1 1 Ψ R b ir = b r (6.30) 2 Ψj 2 1
Combining the unknown nodal parameters, we obtain c1 d
c1 where
dΨ 2 dX X = X
i
dΨ 2 dX X = X
j
(e) t + 5 [K](e) c1 + [K]c2 6 b
[K](e) c1 =
M06_MOAV4303_04_GE_C06.INDD 335
c1 1 J / 1
Ψi r = 5 F6 (e) (6.31) Ψj
1 R 1
27/11/14 9:51 AM
336 Chapter 6 Analysis of OneDimensional Problems
is the elemental conductance for a heat transfer problem (or, it could represent the stiffness for solid mechanics problems), due to the c1 coefficient, c2 / 2 J 6 1
[K](e) c2 =
1 R 2
is the elemental conductance (or, for a solid mechanics problem, the stiffness) due to the c2 coefficient, and 5 F6 (e) =
c3 / 1 b r 2 1
is the load matrix for a given element. The terms c1 d
dΨ 2 dX X = Xi
dΨ 2 c1 dX X = Xj
t
contribute to both the conductance (or, for a solidmechanics problem, the stiffness) matrix and the load matrix. They need to be evaluated for specific boundary conditions. We shall undertake this task shortly. However, let us first write down the conductance matrix for a typical onedimensional fin in terms of its parameters. The conductance matrices are given by [K](e) c1 =
c1 1 J / 1
1 kA 1 R = J 1 / 1
1 R (6.32) 1
and [K](e) c2 =
c2 / 2 J 6 1
hp/ 2 1 R = J 2 6 1
1 R (6.33) 2
In general, the elemental conductance matrix may consist of three terms: The [K](e) c1 term is due to conduction heat transfer along the fin (through the crosssectional area); the [K](e) c2 term represents the heat loss through the top, bottom, and side surfaces (periphery) of an element of a fin; and, depending on the boundary condition of the tip, an additional elemental conductance matrix [K](e) B.C. can exist. For the very last element containing the tip surface, and referring to the boundary condition given by Eq. (6.5), the heat loss through the tip surface can be evaluated as dΨ 2 dT 2 kA X = Xi dX dX X = Xi 0 d t = d t = b r dΨ 2 dT hA(T j  Tf) c1 2 kA dX X = Xj dX X = Xj c1
M06_MOAV4303_04_GE_C06.INDD 336
(6.34)
27/11/14 9:51 AM
Section 6.1 Heat Transfer Problems 337
By rearranging and simplifying, we have kA d
kA
dT 2 dX X = Xi dT 2 dX X = Xj
t = b
0 0 r = J hA(Tj  Tf) 0
(e)
[K]B.C. = J
0 0
0 T 0 R b ir  b r (6.35) hA Tj hATf
0 R (6.36) hA
0 r belongs to the right side of Eq. (6.31) with the thermal load matrix. hATf It shows the contribution of the boundary condition of the tip to the load matrix:
The term  b
5F6 (e) B.C. = b
0 r (6.37) hATf
To summarize, the conductance matrix for all elements, excluding the last element, is given by [K](e) = b
kA 1 J / 1
hp / 2 1 R + J 1 6 1
1 R r (6.38) 2
If the heat loss through the tip of the fin must be accounted for, the conductance matrix for the very last element must be computed from the equation [K](e) = b
hp / 2 1 R + J 1 6 1
kA 1 J / 1
1 0 R + J 2 0
0 R r (6.39) hA
The thermal load matrix for all elements, excluding the last element, is given by 5F6 (e) =
hp /Tf 1 b r (6.40) 2 1
If the heat loss through the tip of the fin must be included in the analysis, the thermal load matrix for the very last element must be computed from the relation 5F6 (e) =
hp /Tf 1 0 b r + b r (6.41) 2 1 hATf
The next set of examples demonstrates the assembly of elements to present the entire problem and the treatment of other boundary conditions. Example 6.1 A Fin Problem Aluminum fins of a rectangular profile, shown in Figure 6.4, are used to remove heat from a surface whose temperature is 100°C. The temperature of the ambient air is 20°C.
M06_MOAV4303_04_GE_C06.INDD 337
27/11/14 9:51 AM
338 Chapter 6 Analysis of OneDimensional Problems
Tfluid = 205C, h = 30 W/m2 · K Tbase = 1005C 1
(1) 20 mm
2
3
(2) 20 mm
4
(3) 20 mm
(4)
5
20 mm
Figure 6.4 Finite element model of a straight fin.
The thermal conductivity of aluminum is 168 W/m # K (W/m # °C). The natural convective heat transfer coefficient associated with the surrounding air is 30 W/m2 # K (W/m2 # °C). The fins are 80 mm long, 5 mm wide, and 1 mm thick. (a) Determine the temperature distribution along the fin using the finite element model shown in Figure 6.4. (b) Compute the heat loss per fin. We will solve this problem using two boundary conditions for the tip. First, let us include the heat transfer from the tip’s surface in the analysis. For elements (1), (2), and (3) in the situation, the conductance and thermal load matrices are given by [K](e) = b 5F6 (e) =
kA 1 J / 1
hp/ 2 1 R + J 1 6 1
1 Rr 2
hp /Tf 1 b r 2 1
Substituting for the properties, we obtain [K](e) = b 5F6 (e) =
(168) (5 * 1 * 10  6) 20 * 10
3
J
1 1
1 30 * 12 * 20 * 10  6 2 R + J 1 6 1
1 Rr 2
30 * 12 * 20 * 10  6 * 20 1 0.072 b r = b r 2 1 0.072
The conductance matrix for elements (1), (2), and (3) is [K](1) = [K](2) = [K](3) = J
M06_MOAV4303_04_GE_C06.INDD 338
0.0444 0.0408
0.0408 W R 0.0444 °C
27/11/14 9:51 AM
Section 6.1 Heat Transfer Problems 339
and the thermalload matrix for elements (1), (2), and (3) is 5F6 (1) = 5F6 (2) = 5F6 (3) = b
0.072 rW 0.072
Including the boundary condition of the tip, the conductance and load matrices for element (4) are obtained in the following manner: [K](e) = b
kA 1 J / 1
hp / 2 1 R + J 1 6 1
1 0 R + J 2 0
[K](4) = J
0.0444 0.0408
0.0408 0 R + J 0.0444 0
0 0.0444 R = J (30 * 5 * 1 * 10  6) 0.0408
5F6 (e) =
0 Rr hA 0.0408 W R 0.04455 °C
hp /Tf 1 0 b r + b r 2 1 hATf
5F6 (4) = b
0.072 0 0.072 r + b r = b rW 6 0.072 (30 * 5 * 1 * 10 * 20) 0.075
Assembly of the elements leads to the global conductance matrix [K](G) and the global load matrix 5F6(G):
[K](G)
5F6(G)
0.0444 0.0408 = E 0 0 0
0.0408 0.0444 + 0.0444 0.0408 0 0
0 0.0408 0.0444 + 0.0444 0.0408 0
0 0 0.0408 0.0444 + 0.0444 0.0408
0 0 0 U 0.0408 0.04455
0.072 0.072 + 0.072 = e 0.072 + 0.072 u 0.072 + 0.072 0.075
Applying the base boundary condition T1 = 100°C, we find that the final set of linear equations becomes 1 0.0408 E 0 0 0
M06_MOAV4303_04_GE_C06.INDD 339
0 0.0888 0.0408 0 0
0 0.0408 0.0888 0.0408 0
0 0 0.0408 0.0888 0.0408
0 T1 100 0 T2 0.144 0 U e T3 u = e 0.144 u 0.0408 T4 0.144 0.04455 T5 0.075
27/11/14 9:51 AM
340 Chapter 6 Analysis of OneDimensional Problems
We can obtain the nodal temperatures from the solution of the above equation. The nodal solutions are T1 100 T2 75.03 e T3 u = e 59.79 u °C T4 51.56 T5 48.90 Note that the nodal temperatures are given in °C and not in °K. Because the crosssectional area of the given fin is relatively small, we could have neglected the heat loss from the tip. Under this assumption, the elemental conductance and forcing matrices for all elements are given by: [K](1) = [K](2) = [K](3) = [K](4) = J
0.0444 0.0408
5F6(1) = 5F6(2) = 5F6(3) = 5F6(4) = b
0.0408 W R 0.0444 °C
0.072 rW 0.072
Assembly of the elements leads to the global conductance matrix [K](G) and the global load matrix 5F6(G): [K](G)
5F6(G)
0.0444 0.0408 = E 0 0 0
0.0408 0.0444 + 0.0444 0.0408 0 0
0 0.0408 0.0444 + 0.0444 0.0408 0
0 0 0.0408 0.0444 + 0.0444 0.0408
0 0 0 U 0.0408 0.0444
0.072 0.072 + 0.072 = e 0.072 + 0.072 u 0.072 + 0.072 0.072
Applying the base boundary condition T1 = 100°C, we find that the final set of linear equations becomes 1 0.0408 E 0 0 0
M06_MOAV4303_04_GE_C06.INDD 340
0 0.0888 0.0408 0 0
0 0.0408 0.0888 0.0408 0
0 0 0.0408 0.0888 0.0408
0 T1 100 0 T2 0.144 0 U e T3 u = e 0.144 u 0.0408 T4 0.144 0.0444 T5 0.072
27/11/14 9:51 AM
Section 6.1 Heat Transfer Problems 341
which has approximately the same solution as that calculated previously: T1 100 T2 75.08 e T3 u = e 59.89 u °C T4 51.74 T5 49.19 Compared to the previous results, the nodal temperatures are slightly higher because we neglected the heat loss through the end surface of the tip. The total heat loss Q from the fin can be determined by summing the heat loss through individual elements:
Qtotal = ΣQ(e)
Q(e) =
=
(6.42)
Xj
LXi
hp(T  Tf) dX
Xj
LXi
hp((SiTi + SjTj)  Tf) dX = hp / ¢ ¢
Ti + Tj 2
≤  Tf ≤ (6.43)
Applying the temperature results to Eqs. (6.42) and (6.43), we have Qtotal = Q(1) + Q(2) + Q(3) + Q(4) Ti + Tj Q(1) = hp / ¢ ¢ ≤  Tf ≤ = 30 * 12 * 20 2 75.08 + 59.89 Q(2) = 30 * 12 * 20 * 10  6 ¢ ¢ ≤ 2 59.89 + 51.74 Q(3) = 30 * 12 * 20 * 10  6 ¢ ¢ ≤ 2 51.74 + 49.19 Q(4) = 30 * 12 * 20 * 10  6 ¢ ¢ ≤ 2 Qtotal = 1.3051 W
* 10  6 ¢ ¢
100 + 75.08 ≤  20≤ = 0.4862 W 2
 20≤ = 0.3418 W  20≤ = 0.2578 W  20≤ = 0.2193 W
Example 6.1 Revisited Using Example 6.1, we will show how to use Excel to set up and solve a onedimensional heat transfer problem. 1. In cell A1 type Example 6.1, and in cells A3, A4, A5, A6, A7, and A8 type L=, P=, A=, k=, h =, and Tf= , respectively, as shown. After inputting the value of L in cell B3, select B3, and in the “Name Box” type L and hit the Return key. Similarly, after inputting the values of P, A, k, h, and Tf in cells B4, B5, B6, B7, and B8 select each cell and in the corresponding “Name Box” type P, A, k, h, and Tf, respectively. Make sure to hit the Return key after you name each variable. Also, create the table shown.
M06_MOAV4303_04_GE_C06.INDD 341
27/11/14 9:51 AM
342 Chapter 6 Analysis of OneDimensional Problems
2. Create the [K] and {F} matrices as shown. For example, select cell E19 and type = (A*k , L) + (h*P*L , 6)*2 . As another example, select cells H22:H23 and type = h*P*L*Tf , 2
and while holding down the Ctrl and Shift keys, hit the Return key.
3. Next, create [A1], [A2], [A3], and [A4] matrices and name them Aelement1, Aelement2, Aelment3, and Aelement4, as shown. See Section 2.5, Equation (2.9), if you have forgotten what these matrices represent. First create [A1] and then to create [A2], [A3], and [A4], copy [A1], rows 25 through 27 into rows 29 through 31; 33 through 35; and 37 through 39 and modify them accordingly. The nodal temperatures T1, T2, T3, T4, T5, and Ti and Tj are shown alongside the [A1], [A2], [A3], and [A4] matrices to aid us observe the contribution of node to its neighboring elements.
M06_MOAV4303_04_GE_C06.INDD 342
27/11/14 9:51 AM
Section 6.1 Heat Transfer Problems 343
4. Next, create the conductance matrix for each element (with their proper positions in the global matrix) and name them K1G, K2G, K3G, and K4G. For example, to create [K]1G, select B41:F45 and type = MMULT(TRANSPOSE(Aelement1),MMULT(Kelement,Aelement1)) and while holding down the Ctrl and Shift keys hit the Return key. In a similar way, create [K]2G, [K]3G, and [K]4G as shown.
5. The final global matrix is created next. Select the range B65:F69 and type = K1G + K2G + K3G + K4G and while holding down the Ctrl and Shift keys hit, the Return key.
6. Create the load matrix as shown.
M06_MOAV4303_04_GE_C06.INDD 343
27/11/14 9:51 AM
344 Chapter 6 Analysis of OneDimensional Problems
7. Apply the boundary conditions. Copy the appropriate portion of the KG matrix and paste it in the range C77:G81 as values only and modify it as shown. Name the range KwithappliedBC. Similarly, create the corresponding load matrix in the range C83:C87 and name it FwithappliedBC.
8. Select the range C89:C93 and type = MMULT(MINVERSE(KwithappliedBC),FwithappliedBC) and while holding down the Ctrl and Shift keys, hit the Return key.
The complete Excel sheet is shown next.
M06_MOAV4303_04_GE_C06.INDD 344
27/11/14 9:51 AM
Section 6.1 Heat Transfer Problems 345
Example 6.2 A Composite Wall Problem A wall of an industrial oven consists of three different materials, as depicted in Figure 6.5. The first layer is composed of 5 cm of insulating cement with a clay binder that has a thermal conductivity of 0.08 W/m # K. The second layer is made from 15 cm of 6ply asbestos board with a thermal conductivity of 0.074 W/m # K (W/m # °C). The exterior consists of 10cm common brick with a thermal conductivity of 0.72 W/m2 # K (W/m # °C). The inside wall temperature of the oven is 200°C, and the outside air is 30°C with a convection coefficient of 40 W/m2 # K (W/m2 # °C). Determine the temperature distribution along the composite wall.
M06_MOAV4303_04_GE_C06.INDD 345
27/11/14 9:51 AM
346 Chapter 6 Analysis of OneDimensional Problems
Tsurface = 2005C
Tfluid = 305C, h = 40 W/m2 · K k = 0.08 W/m · K
k = 0.074 W/m · K
k = 0.72 W/m · K
15 cm
10 cm
5 cm
X
T1 = 2005C
(1) 1
(2)
(3) 3
2
4
Figure 6.5 A composite wall of an industrial oven.
This heat conduction problem is governed by the equation
kA
d2 T = 0 (6.44) dX2
dT 0 X = 30 cm = and is subjected to the boundary conditions T1 = 200°C and kA dX hA(T4  Tf). For this example, we compare Eq. (6.44) to Eq. (6.8), finding that c1 = kA, c2 = 0, c3 = 0, and Ψ = T. Thus, for element (1), we have
[K](1) =
kA 1 J / 1
1 0.08 * 1 1 R = J 1 0.05 1
1 1.6 R = J 1 1.6
1.6 W R 1.6 °C
0 5F6 (1) = b r W 0 For element (2), we have [K](2) =
kA 1 J / 1
1 0.074 * 1 1 R = J 1 0.15 1
1 0.493 R = J 1 0.493
0.493 W R 0.493 °C
0 5 F6(2) = b r W 0
M06_MOAV4303_04_GE_C06.INDD 346
27/11/14 9:51 AM
Section 6.2 A Fluid Mechanics Problem 347
For element (3), including the boundary condition at node 4, we have [K](3) =
1 0 R + J 1 0
kA 1 J / 1
= J 5 F6(3) = b
7.2 7.2
0 0.72 * 1 1 R = J hA 0.1 1
1 0 R + J 1 0
0 R (40 * 1)
7.2 W R 47.2 °C
0 0 0 r = b r = b rW hATf (40 * 1 * 30) 1200
Assembling elements, we obtain 1.6 1.6 [K](G) = D 0 0
1.6 1.6 + 0.493 0.493 0
0 0.493 0.493 + 7.2 7.2
0 0 T 7.2 47.2
0 0 5F6(G) = d t 0 1200 Applying the boundary condition at the inside furnace wall, we get 1 1.6 D 0 0
0 2.093 0.493 0
0 0.493 7.693 7.2
0 T1 200 0 T 0 Td 2t = d t 7.2 T3 0 47.2 T4 1200
and solving the set of linear equations, we have the following results: T1 200 T 162.3 d 2t = d t °C T3 39.9 T4 31.5 Note that this type of heat conduction problem can be solved just as easily using fundamental concepts of heat transfer without resorting to finite element formulation. The point of this exercise was to demonstrate the steps involved in finite element analysis using a simple problem. 6.2 A Fluid Mechanics Problem Example 6.3 A Fluid Mechanics Problem In a chemical processing plant, aqueous glycerin solution flows in a narrow channel, as shown in Figure 6.6. The pressure drop along the channel is continuously monitored. The upper wall of the channel is maintained at 50°C, while the lower wall is kept at 20°C.
M06_MOAV4303_04_GE_C06.INDD 347
27/11/14 9:51 AM
348 Chapter 6 Analysis of OneDimensional Problems 505C 7 6 u
5
h
4 3
y 2 x
1 205C
Figure 6.6 Laminar flow of aqueous glycerin solution through a channel.
The variation of viscosity and density of the glycerin with the temperature is given in Table 6.1. For a relatively low flow, the pressure drop along the channel is measured to be 120 Pa/m. The channel is 3 m long, 9 cm high, and 40 cm wide. Determine the velocity profile and the mass flow rate of the fluid through the channel. The laminar flow of a fluid with a constant viscosity inside a channel is governed by the balance between the net shear forces and the net pressure forces acting on a parcel of fluid. The equation of motion is
m
dp d2 u = 0 (6.45) dx dy2
subject to the boundary conditions u(0) = 0 and u(h) = 0. Here, u represents fluid dp velocity, m is the dynamic viscosity of the fluid, and dx is the pressure drop in the direction of the flow. For this problem, when comparing Eq. (6.45) to Eq. (6.8), we find that dp c1 = m, c2 = 0, c3 =  dx , and Ψ = u.
Table 6.1 Properties of aqueous glycerin solution as a function of temperature
M06_MOAV4303_04_GE_C06.INDD 348
Temperature (°C)
Viscosity (kg/m # s)
Density (kg/m3)
20
0.90
1255
25
0.65
1253
30
0.40
1250
35
0.28
1247
40
0.20
1243
45
0.12
1238
50
0.10
1233
27/11/14 9:51 AM
Section 6.2 A Fluid Mechanics Problem 349
Here, the viscosity of the aqueous glycerin solution varies with the height of the channel. We will use an average value of viscosity over each element when computing the elemental resistance matrices. The average values of viscosity and density associated with each element are given in Table 6.2. Using the properties from Table 6.2, we can compute the elemental flow resistance matrices as kg 51.67 R 2# 51.67 m s
[K](1) =
m 1 J / 1
1 0.775 1 R = J 2 1 1 1.5 * 10
1 51.67 R = J 1 51.67
[K](2) =
m 1 J / 1
1 0.525 1 R = J 2 1 1 1.5 * 10
1 35 R = J 1 35
[K](3) =
m 1 J / 1
1 0.340 1 R = J 2 1 1 1.5 * 10
1 22.67 R = J 1 22.67
[K](4) =
m 1 J / 1
1 0.240 1 R = J 2 1 1 1.5 * 10
1 16 R = J 1 16
[K](5) =
m 1 J / 1
1 0.160 1 R = J 2 1 1 1.5 * 10
1 10.67 R = J 1 10.67
[K](6) =
m 1 J / 1
1 0.110 1 R = J 2 1 1 1.5 * 10
1 7.33 R = J 1 7.33
kg 35 R 2# 35 m s kg 22.67 R 2# 22.67 m s kg 16 R 2# 16 m s kg 10.67 R 2# 10.67 m s kg 7.33 R 2# 7.33 m s
Since [K] represents resistance to flow, we have opted to use the term elemental flowresistance matrix instead of elemental stiffness matrix. Because the flow is fully develdp oped, dx is constant; thus, the forcing matrix has the same value for all elements: 5F6(1) = 5F6(2) = c = 5F6(5) = 5F6(6) =

dp /  ( 120)(1.5 * 10 2) 1 dx 1 0.9 N b r= b r = b r 2 2 1 2 1 0.9 m
Table 6.2 Properties of each element Element
M06_MOAV4303_04_GE_C06.INDD 349
Average Viscosity (kg/m # s)
Average Density (kg/m3)
1
0.775
1254
2
0.525
1252
3
0.34
1249
4
0.24
1245
5
0.16
1241
6
0.11
1236
27/11/14 9:51 AM
350 Chapter 6 Analysis of OneDimensional Problems
The negative value associated with the pressure drop represents the decreasing nature of the pressure along the direction of flow in the channel. The global resistance matrix is obtained by assembling the elemental resistance matrices:
[K](G)
51.67 51.67 0 0 0 0 0 51.67 51.67 + 35 35 0 0 0 0 0 35 35 + 22.67 22.67 0 0 0 = G 0 0 22.67 22.67 + 16 16 0 0 W 0 0 0 16 16 + 10.67 10.67 0 0 0 0 0 10.67 10.67 + 7.33 7.33 0 0 0 0 0 7.33 7.33
and the global forcing matrix is
5 F6(G)
0.9 0.9 = g 0.9 0.9 0.9
0.9 + 0.9 + 0.9 + 0.9 w + 0.9 + 0.9 0.9
Applying the noslip boundary conditions at the walls leads to the matrix 1 51.67 0 G 0 0 0 0
0 86.67 35 0 0 0 0
0 35 57.67 22.67 0 0 0
0 0 22.67 38.67 16 0 0
0 0 0 16 26.67 10.67 0
0 0 0 0 10.67 18 0
0 u1 0 0 u2 1.8 0 u3 1.8 0 W g u4 w = g 1.8 w 0 u5 1.8 7.33 u6 1.8 1 u7 0
The solution provides the fluid velocities at each node: u1 0 u2 0.1233 u3 0.2538 g u4 w = g 0.3760 w m/s u5 0.4366 u6 0.3588 u7 0
M06_MOAV4303_04_GE_C06.INDD 350
27/11/14 9:51 AM
Section 6.3 An Example Using ANSYS 351
The mass flow rate through the channel can be determined from # # mtotal = Σm (e) # m (e) =
yj
Lyi
ruW dy =
yj
Lyi
rW(Siui + Sjuj) dy = rW/ ¢
(6.46) ui + uj 2
≤ (6.47)
In Eq. (6.47), W represents the width of the channel and r is density of the fluid. The elemental and total mass flow rates are given by ui + uj 0 + 0.1233 # ≤ = 1254 * 0.4 * 1.5 * 10  2 * = 0.4638 kg/s m (1) = rW/ ¢ 2 2 0.1233 + 0.2538 # m (2) = 1252 * 0.4 * 1.5 * 10  2 * = 1.4164 kg/s 2 0.2538 + 0.3760 # m (3) = 1249 * 0.4 * 1.5 * 10  2 * = 2.3598 kg/s 2 0.3760 + 0.4366 # m (4) = 1245 * 0.4 * 1.5 * 10  2 * = 3.0350 kg/s 2 0.4366 + 0.3588 # m (5) = 1241 * 0.4 * 1.5 * 10  2 * = 2.9612 kg/s 2 0.3588 + 0 # m (6) = 1236 * 0.4 * 1.5 * 10  2 * = 1.3304 kg/s 2 # mtotal = 11.566 kg/s 6.3 An Example Using ANSYS Example 6.4 Revisited A wall of an industrial oven consists of three different materials, as shown in Figure 6.4, repeated here as Figure 6.7. The first layer is composed of 5 cm of insulating cement with a clay binder that has a thermal conductivity of 0.08 W/m # K. The second layer is made from 15 cm of 6ply asbestos board with a thermal conductivity of 0.074 W/m # K. The exterior consists of 10cm common brick with a thermal conductivity of 0.72 W/m2 # K. The inside wall temperature of the oven is 200°C, and the outside air is 30°C with a convection coefficient of 40 W/m2 # K. Determine the temperature distribution along the composite wall. The following steps demonstrate how to create onedimensional conduction problems with convective boundary conditions in ANSYS. This task includes choosing appropriate element types, assigning attributes, applying boundary conditions, and obtaining results. To solve this problem using ANSYS, we employ the following steps: Enter the ANSYS program by using the Launcher. Type HeatTran (or a file name of your choice) in the Jobname entry field of the dialog box, and pick Run to start the GUI.
M06_MOAV4303_04_GE_C06.INDD 351
27/11/14 9:51 AM
M06_MOAV4303_04_GE_C06.INDD 352
T1 = 2005C 1
(1)
5 cm
2
X
k = 0.08 W/m · K
Tsurface = 2005C
3
(3)
10 cm
k = 0.72 W/m · K
Figure 6.7 A composite wall of an industrial oven.
(2)
15 cm
k = 0.074 W/m · K
4
Tfluid = 305C, h = 40 W/m2 · K
352 Chapter 6 Analysis of OneDimensional Problems
27/11/14 9:51 AM
Section 6.3 An Example Using ANSYS 353
Create a title for the problem. This title will appear on ANSYS display windows to provide a simple way of identifying the displays: utility menu: File S Change Title c
Define the element type and material properties: main menu: Preprocessor S Element Type S Add/Edit/Delete
M06_MOAV4303_04_GE_C06.INDD 353
27/11/14 9:51 AM
354 Chapter 6 Analysis of OneDimensional Problems
Assign the crosssectional area of the wall. main menu: Preprocessor S Real Constants S Add/Edit/Delete
Assign the thermal conductivity values. main menu: Preprocessor S Material Props S Material Models S Thermal S Conductivity S Isotropic
M06_MOAV4303_04_GE_C06.INDD 354
27/11/14 9:51 AM
Section 6.3 An Example Using ANSYS 355
From the Define Material Model Behavior window: Material S New Model N
Now doubleclick on Isotropic.
M06_MOAV4303_04_GE_C06.INDD 355
27/11/14 9:51 AM
356 Chapter 6 Analysis of OneDimensional Problems
From the Define Material Model Behavior window: Material S New Model N
Assign the thermal conductivity of the third layer by doubleclicking on Isotropic again.
M06_MOAV4303_04_GE_C06.INDD 356
27/11/14 9:51 AM
Section 6.3 An Example Using ANSYS 357
From the Define Material Model Behavior window: Material S New Model N
Assign the heat transfer coefficient by doubleclicking on Convection or Film Coef.
ANSYS Toolbar: SAVE_DB Set up the graphics area (i.e., workplane, zoom, etc.): utility menu: Workplane S WP Settings c
M06_MOAV4303_04_GE_C06.INDD 357
27/11/14 9:51 AM
358 Chapter 6 Analysis of OneDimensional Problems
Toggle on the workplane by the following sequence: utility menu: Workplane S Display Working Plane Bring the workplane to view using the following sequence: utility menu: PlotCtrls S Pan, Zoom, Rotate c Create nodes by picking points on the workplane: main menu: Preprocessor S Modeling S Create S Nodes S On Working Plane On the workplane, pick the location of nodes and apply them: [WP = 0,0]
M06_MOAV4303_04_GE_C06.INDD 358
27/11/14 9:51 AM
Section 6.3 An Example Using ANSYS 359
[WP = 0.05,0]
[WP = 0.2,0]
[WP = 0.3,0]
Create the node for the convection element: [WP = 0.3,0]
OK You may want to turn off the workplane now and turn on node numbering: utility menu: Workplane S Display Working plane utility menu: PlotCtrls S Numbering c
You may want to list nodes at this point in order to check your work: utility menu: List S Nodes c
M06_MOAV4303_04_GE_C06.INDD 359
27/11/14 9:51 AM
360 Chapter 6 Analysis of OneDimensional Problems
Close ANSYS Toolbar: SAVE_DB Define elements by picking nodes: main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes [node 1 and then node 2]
[use the middle button anywhere in the ANSYS graphics window to apply.] OK
M06_MOAV4303_04_GE_C06.INDD 360
27/11/14 9:51 AM
Section 6.3 An Example Using ANSYS 361
Assign the thermal conductivity of the second layer (element), and then connect the nodes to define the element: main menu: Preprocessor S Modeling S Create S Elements S Element Attributes
main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes [node 2 and then node 3]
[anywhere in the ANSYS graphics window]
OK
Assign the thermal conductivity of third layer (element), and then connect the nodes to define the element: main menu: Preprocessor S Modeling S Create S Elements S Element Attributes
M06_MOAV4303_04_GE_C06.INDD 361
27/11/14 9:51 AM
362 Chapter 6 Analysis of OneDimensional Problems
main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes
[node 3 and then node 4*]
[anywhere in the ANSYS graphics window]
OK
Create the convection link: main menu: Preprocessor S Modeling S Create S Elements S Element Attributes
*
Press the OK key of the MultipleEntities window and proceed.
M06_MOAV4303_04_GE_C06.INDD 362
27/11/14 9:51 AM
Section 6.3 An Example Using ANSYS 363
main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes
[node 4†and then 5‡] OK ANSYS Toolbar: SAVE_DB
Apply boundary conditions:
main menu: Solution S Define Loads S Apply S Thermal S Temperature S On Nodes [node 1]
[anywhere in the ANSYS graphics window]
†
Press the OK key of the MultipleEntities window. Press the Next key of the MultipleEntities window, and then OK.
‡
M06_MOAV4303_04_GE_C06.INDD 363
27/11/14 9:51 AM
364 Chapter 6 Analysis of OneDimensional Problems
main menu: Solution S Define Loads S Apply S Thermal S Temperature S On Nodes
[node 5*]
[anywhere in the ANSYS graphics window]
ANSYS Toolbar: SAVE_DB *
Press the Next key and then the OK key of the MultipleEntities window and proceed.
M06_MOAV4303_04_GE_C06.INDD 364
27/11/14 9:51 AM
Section 6.3 An Example Using ANSYS 365
Solve the problem: main menu: Solution S Solve S Current LS OK Close (the solution is done!) window. Close (the/STAT Command) window. For the postprocessing phase, obtain information such as nodal temperatures: main menu: General Postproc S List Results S Nodal Solution
M06_MOAV4303_04_GE_C06.INDD 365
27/11/14 9:51 AM
366 Chapter 6 Analysis of OneDimensional Problems
Close Exit ANSYS and save everything. Toolbar: QUIT
6.4 Verification of Results There are various ways to verify your findings. Consider the nodal temperatures of Example 6.2, as computed by ANSYS and displayed in Table 6.3. In general, for a heat transfer problem under steadystate conditions, conservation of energy applied to a control volume surrounding an arbitrary node must be satisfied. Are the energies flowing into and out of a node balanced out? Let us use Example 6.2 to demonstrate this important concept. The heat loss through each layer of the composite wall must be equal. Furthermore, heat loss from the last layer must equal the heat removed by the surrounding air. So, Q(1) = Q(2) = Q(3) = Q(4) Q(1) = kA
∆T 200  162.3 = (0.08) (1) ¢ ≤ = 60 W / 0.05
Q(2) = (0.074) (1) ¢ Q(3) = (0.72) (1) ¢
162.3  39.9 ≤ = 60 W 0.15
39.9  31.5 ≤ = 60 W 0.1
Table 6.3 Nodal temperatures Node Number
M06_MOAV4303_04_GE_C06.INDD 366
Temperature (°C)
1
200
2
162.3
3
39.9
4
31.5
5
30
27/11/14 9:51 AM
Chapter 6 References 367
and the heat removal by the fluid is given by Q(4) = hA∆T = (40) (1) (31.5  30) = 60 W For thermal elements, ANSYS provides information such as heat flow through each element. Therefore, we could have compared those values with the one we calculated above. Another check on the validity of your results could have come from examining the slopes of temperatures in each layer. The first layer has a temperature slope of 754°C/m. For the second layer, the slope of the temperature is 816°C/m. This layer consists of a material with relatively low thermal conductivity and, therefore, a relatively large temperature drop. The slope of the temperature in the exterior wall is 84°C/m. Because the exterior wall is made of a material with relatively high thermal conductivity, we expect the temperature drop through this layer not to be as significant as the other layers. Now consider the fin problem in Example 6.1. For this problem, recall that all elements have the same length. We determined the temperature distribution and the heat loss from each element. Comparing heat loss results, it is important to realize that element (1) has the highest value because the greatest thermal potential exists at the base of the fin, and as the temperature of the fin drops, so does the rate of heat loss for each element. This outcome is certainly consistent with the results we obtained previously. These simple problems illustrate the importance of checking for equilibrium conditions when verifying results.
Summary At this point you should: 1. know how to formulate conductance or resistance matrices, and be able to for mulate load matrices for various onedimensional problems. 2. know how to apply appropriate boundary conditions. 3. have a good understanding of the Galerkin and energy formulations of one dimensional problems. 4. know how to verify your results. References ANSYS User’s Manual: Procedures, Vol. I, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Commands, Vol. II, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Elements, Vol. III, Swanson Analysis Systems, Inc. Incropera, F., and Dewitt, D., Fundamentals of Heat and Mass Transfer, 2nd ed., New York, John Wiley and Sons, 1985. Glycerin Producers’ Association, “Physical Properties of Glycerin And Its Solutions,” New York. Segrlind, L., Applied Finite Element Analysis, 2nd ed., New York, John Wiley and Sons, 1984.
M06_MOAV4303_04_GE_C06.INDD 367
27/11/14 9:51 AM
368 Chapter 6 Analysis of OneDimensional Problems
Problems 1. Aluminum fins, similar to the ones in Example 6.1, with rectangular profiles are used to remove heat from a surface whose temperature is 1505C. The temperature of the ambient air is 20°C. The thermal conductivity of aluminum is 168 W/m # K. The natural convective coefficient associated with the surrounding air is 35 W/m2 # K. The fins are 150 mm long, 5 mm wide, and 1 mm thick. (a) Determine the temperature distribution along a fin using five equally spaced elements. (b) Approximate the total heat loss for an array of 50 fins.
2. For the aluminum fins in Problem 1, determine the temperature of a point on the fin 50 mm from the base. Also compute the fraction of the total heat that is lost through this section of the fin. 3. A pin fin, or spine, is a fin with a circular cross section. An array of aluminum pin fins are used to remove heat from a surface whose temperature is 120°C. The temperature of the ambient air is 25°C. The thermal conductivity of aluminum is 168 W/m # K. The natural convective coefficient associated with the surrounding air is 30 W/m2 # K. The fins are 100 mm long and have respective diameters of 4 mm. (a) Determine the temperature distribution along a fin using five equally spaced elements. (b) Approximate the total heat loss for an array of 100 fins. 4. A rectangular aluminum fin is used to remove heat from a surface whose temperature is 80°C. The temperature of the ambient air varies between 18°C and 25°C. The thermal conductivity of aluminum is 168 W/m # K. The natural convective coefficient associated with the surrounding air is 25 W/m2 # K. The fin is 100 mm long, 5 mm wide, and 1 mm thick. (a) Determine the temperature distribution along a fin using five equally spaced elements for both ambient conditions. (b) Approximate the total heat loss for an array of 50 fins for each ambient condition. (c) The exact temperature distribution and heat loss for a fin with a negligible heat loss at its tip is given by the following hyperbolic functions T(x)  Tf Tb  Tf
cosh J =
hp (L  x) R A kAc
cosh J
hp (L) R A kAc
Q = 2hpkAc atanhJ
hp (L) R ≤(Tb  Tf) A kAc
Compare your finite element results to the exact results.
M06_MOAV4303_04_GE_C06.INDD 368
27/11/14 9:51 AM
Chapter 6 Problems 369 Xj
SihpTf dX for a situation in which the heat transfer coefficient h LX i varies linearly over a given element. 6. The front window of a car is defogged by supplying warm air at 90°F to its inner surface. The glass has a thermal conductivity of k = 0.8 W/m # °C with a thickness of approximately 1/4 in. With the supply fan running at moderate speed, the heat transfer coefficient associated with the air is 50 W/m2 # K. The outside air is at a temperature of 20°F with an associated heat transfer coefficient of 100 W/m2 # K. Determine (a) the temperatures of the inner and outer surfaces of the glass and (b) the heat loss through the glass if the area of the glass is approximately 15 ft2. 7. A wall of an industrial oven consists of three different materials, as shown in the accompanying figure. The first layer is composed of 10 cm of insulating cement with a thermal conductivity of 0.12 W/m # K. The second layer is made from 20 cm of 8ply asbestos board with a thermal conductivity of 0.071 W/m # K. The exterior consists of 12cm cement mortar with a thermal conductivity of 0.72 W/m2 # K. The inside wall temperature of the oven is 250°C, and the outside air is at a temperature of 35°C with a convection coefficient of 45 W/m2 # K. Determine the temperature distribution along the composite wall.
5. Evaluate the integral
Tsurface = 2505C
Tfluid = 355C, h = 45 W/m2 · K k = 0.12 W/m · K
k = 0.071 W/m · K
10 cm
k = 0.72 W/m · K
20 cm
12 cm
X
T1 = 2505C
(1) 1
(2)
(3) 3
2
4
8. Replace the temperature boundary condition of the inside wall of the oven in Problem 7 with
air temperature of 4055C and an associated convection coefficient of 100 W/m2 # K. Show the contribution of this boundary condition to the conductance matrix and the forcing matrix of element (1). Also determine the temperature distribution along the composite wall.
9. The equation for the heat diffusion of a onedimensional system with heat generation in a Cartesian coordinate system is k
M06_MOAV4303_04_GE_C06.INDD 369
0 2T # + q = 0 0X 2
27/11/14 9:52 AM
370 Chapter 6 Analysis of OneDimensional Problems # The rate of thermal energy generation q represents the conversion of energy from electrical, chemical, nuclear, or electromagnetic forms to thermal energy within the volume of a given # system. Derive the contribution of q to the load matrix. Consider a strip of heating elements embedded within the rear glass of a car producing a uniform heat generation at a rate of approximately 8000 W/m3. The glass has a thermal conductivity of k = 0.8 W/m # °C with a thickness of approximately 6 mm. The heat transfer coefficient associated with the 20°C air inside the back of the car is approximately 20 W/m2 # K. The outside air is at a temperature of  5°C with an associated heat transfer coefficient of 50 W/m2 # K. Determine the temperatures of the inner and outer surfaces of the glass.
10. Verify the evaluation of the integral given by Eq. (6.23): Xj
LX
Si(c2Ψ) dX =
i
c2 / c2 / Ψi + Ψ 3 6 j
11. Verify the evaluation of the integral given by Eq. (6.26): Xj
LX
c1 ¢ 
i
dSj dΨ c1 ≤ dX =  ( Ψi + Ψj) dX dX /
12. The deformation of an axial element of length / due to the change in its temperature is given by
dT = a/∆T where dT is the change in the length of the element, a is the thermal expansion coefficient of the material, and ∆T represents the temperature change. Formulate the contribution of thermal strains to the strain energy of an element. Also formulate the stiffness and the load matrices for such an element. 13. You are to size fins of a rectangular cross section to remove a total of 200 W from a [email protected] surface whose temperature is to be kept at 80°C. The temperature of the surrounding air is 25°C, and you may assume that the natural convection coefficient value is 25 W/m2 # K. Because of restrictions on the amount of space, the fins cannot be extended more than 100 mm from the hot surface. You may select from the following materials: aluminum, copper, or steel. In a brief report, explain how you came up with your final design.
M06_MOAV4303_04_GE_C06.INDD 370
27/11/14 9:52 AM
C h a p t e r
7
TwoDimensional Elements
The objective of this chapter is to introduce the concept of twodimensional shape functions, along with twodimensional elements and their properties. Natural coordinates associated with quadrilateral and triangular elements are also presented. We derive the shape functions for rectangular elements, quadratic quadrilateral elements, and triangular elements. Examples of twodimensional thermal and structural elements in ANSYS are also presented. The main topics discussed in Chapter 7 include the following:
7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8
Rectangular Elements Quadratic Quadrilateral Elements Linear Triangular Elements Quadratic Triangular Elements Axisymmetric Elements Isoparametric Elements TwoDimensional Integrals: GaussLegendre Quadrature Examples of TwoDimensional Elements in ANSYS
7.1 Rectangular Elements In Chapter 6, we studied the analysis of onedimensional problems. We investigated heat transfer in a straight fin. We used onedimensional linear shape functions to approximate temperature distributions along elements and formulated the conductance matrix and the thermal load matrix. The resulting systems of equations, once solved, yielded the nodal temperatures. In this chapter, we lay the groundwork for the analysis of twodimensional problems by first studying twodimensional shape functions and elements. To aid us in presenting this material, let us consider the straight fin shown in Figure 7.1. The dimensions of the fin and thermal boundary conditions are such that we cannot accurately approximate temperature distribution along the fin by a onedimensional function. The temperature varies in both the Xdirection and the Ydirection.
M07_MOAV4303_04_GE_C07.INDD 371
27/11/14 9:55 AM
372 Chapter 7 TwoDimensional Elements T Tn Ti
Tm Tj
Y
n
m j
X
Base
i
Figure 7.1 Using rectangular elements to describe a twodimensional temperature distribution.
At this point, it is important to understand that the onedimensional solutions are approximated by line segments, whereas the twodimensional solutions are represented by plane segments. This point is illustrated in Figure 7.1. A closeup look at a typical rectangular element and its nodal values is shown in Figure 7.2. It is clear from examining Figure 7.2 that the temperature distribution over the element is a function of both X and Ycoordinates. We can approximate the temperature distribution for an arbitrary rectangular element by T (e) = b1 + b2x + b3y + b4xy (7.1)
x, y Local coordinate system X, Y Global coordinate system
Tn T Ti
Tj
Tm
y m
n
w
Y i
j
x
/ X
Figure 7.2 A typical rectangular element.
M07_MOAV4303_04_GE_C07.INDD 372
27/11/14 9:55 AM
Section 7.1 Rectangular Elements 373
Note that there are four unknowns (b1, b2, b3, b4) in Eq. (7.1), because a rectangular element is defined by four nodes: i, j, m, n. Also note that the function varies linearly along the edges of the element, and it becomes nonlinear inside the element (see also Problem 31). Elements with these types of characteristics are commonly referred to as bilinear elements. The procedure for deriving twodimensional shape functions is essentially the same as that for onedimensional elements. To obtain b1, b2, b3, and b4, we use the local coordinates x and y. Considering nodal temperatures, we must satisfy the following conditions:
T = Ti at x = 0
and y = 0
(7.2)
T = Tj at x = /
and y = 0
T = Tm at x = /
and y = w
T = Tn at x = 0
and y = w
Applying the nodal conditions given by Eq. (7.2) to Eq. (7.1) and solving for b1, b2, b3, and b4, we have
b1 = Ti
b3 =
b2 =
1 (T  Ti) / j
1 1 (T  Ti) b4 = (T  Tj + Tm  Tn) (7.3) w n /w i
Substituting expressions given for b1, b2, b3, and b4 into Eq. (7.1) and regrouping para meters will result in the temperature distribution for a typical element in terms of shape functions:
T (e) = [Si
Sj
Sm
Ti Tj Sn] d t (7.4) Tm Tn
The shape functions in the above expression are y x ≤ ¢ 1  ≤ (7.5) w /
Si = ¢ 1 
Sj =
y x ¢1  ≤ w /
Sm =
xy /w
Sn =
y x ¢1  ≤ w /
M07_MOAV4303_04_GE_C07.INDD 373
27/11/14 9:55 AM
374 Chapter 7 TwoDimensional Elements Uny
Umy Unx
Umx
m
n
Uiy
Ujy Uix
Ujx
j
i u(e) = SiUix + SjUjx + SmUmx + SnUnx v(e) = SiUiy + SjUjy + SmUmy + SnUny
Figure 7.3 A rectangular element used in formulating planestress problems.
It should be clear that we can use these shape functions to represent the variation of any unknown variable Ψ over a rectangular region in terms of its nodal values Ψi, Ψj, Ψm, and Ψn. Thus, in general, we can write
Ψ(e) = [Si
Sj
Sm
Ψi Ψ Sn] d j t (7.6) Ψm Ψn
For example, Ψ could represent a solid element displacement in a certain direction as shown in Figure 7.3. Natural Coordinates As was discussed in Chapter 5, natural coordinates are basically local coordinates in a dimensionless form. Moreover, as you may recall, most finite element programs perform element numerical integration by Gaussian quadratures, and as the limits of integration, they use an interval from 1 to 1. The origin of the local coordinate system x, y used earlier coincides with the natural coordinates j = 1 and h = 1, as shown in Figure 7.4. 2y 2x If we let j =  1 and h =  1, then the shape functions in terms of the w / natural coordinates j and h are
M07_MOAV4303_04_GE_C07.INDD 374
1 (1  j)(1  h) (7.7) 4 1 Sj = (1 + j)(1  h) 4 1 Sm = (1 + j)(1 + h) 4 1 Sn = (1  j)(1 + h) 4 Si =
27/11/14 9:55 AM
Section 7.2 Quadratic Quadrilateral Elements 375 y h
(x = 0, y = w) (j = 1, h = 1)
(x = /, y = w) (j = 1, h = 1) n
w
m
j
(0, 0)
j
i (j = 1, h = 1) (x = 0, y = 0)
x
(j = 1, h = 1) (x = /, y = 0) /
Figure 7.4 Natural coordinates used to describe a quadrilateral element.
These shape functions have the same general basic properties as their onedimensional counterparts. For example, Si has a value of 1 when evaluated at the coordinates of node i, but has a value of zero at all other nodes. Alternatively, we could have obtained the expressions given in Eq. (7.7) by using a product of linear functions similar to the Lagrange functions as explained in Chapter 5, Section 5.3. For example, for node i, we select the functions such that their product will produce a value of zero at other nodes—namely, j, m, and n—and a value of unity at the given node i. Along the j–m side (j = 1) and n–m side (h = 1), so if we choose the product of functions (1  j)(1  h) then the product will produce a value of zero along j–m side and n–m side. We then evaluate a1, an unknown coefficient, such that when the shape function Si is evaluated at node i(j = 1) and (h = 1), it will produce a value of unity. That is 1 = a1(1  j)(1  h) = a1(1( 1))(1( 1)) 1 a1 =
1 4
7.2 Quadratic Quadrilateral Elements The eightnode quadratic quadrilateral element is basically a higher order version of the twodimensional fournode quadrilateral element. This type of element is better suited for modeling problems with curved boundaries. A typical eightnode quadratic element is shown in Figure 7.5. When compared to the linear elements, for the same number of elements, quadratic elements offer better nodal results. In terms of the natural coordinates j and h, the eightnode quadratic element has the general form of
M07_MOAV4303_04_GE_C07.INDD 375
Ψ(e) = b1 + b2j + b3h + b4jh + b5j2 + b6h2 + b7j2h + b8jh2 (7.8)
27/11/14 9:55 AM
376 Chapter 7 TwoDimensional Elements n
o m
(1, 1)
n
p
o
m
(1, 1)
h p
/
i
k j
(1, 1)
/
j
i
k
j
(1, 1)
Figure 7.5 Eightnode quadratic quadrilateral element.
To solve for b1, b2, b3, c, b8, we must first apply the nodal conditions and create eight equations from which we can solve for these coefficients. Instead of using this laborious and difficult method, we will follow an alternative approach, which is demonstrated next. In general, the shape function associated with each node can be represented in terms of the product of two functions F1 and F2:
S = F1(j, h)F2(j, h) (7.9) For a given node, we select the first function F1 such that it will produce a value of zero when evaluated along the sides of the element that the given node does not contact. Moreover, the second function F2 is selected such that when multiplied by F1, it will produce a value of unity at the given node and a value of zero at other neighboring nodes. The product of functions F1 and F2 must also produce linear and nonlinear terms similar to the ones given by Eq. (7.8). To demonstrate this method, let us consider corner node m, with natural coordinates j = 1 and h = 1. First, F1 must be selected such that along the ijside (h = 1) and inside (j = 1), the function will produce a value of zero. We select F1(j, h) = (1 + j)(1 + h) which satisfies the condition. We then select F2(j, h) = c1 + c2j + c3h The coefficients in F2 should be selected such that when F2 multiplied by F1, they will produce a value of unity at the given node m, and a value of zero at the adjacent neighboring nodes / and o. Evaluating Sm at node m should give Sm = 1 for j = 1 and h = 1;
M07_MOAV4303_04_GE_C07.INDD 376
27/11/14 9:55 AM
Section 7.2 Quadratic Quadrilateral Elements 377
evaluating Sm at node / should give Sm = 0 for j = 1 and h = 0; and evaluating Sm at node o should give Sm = 0 for j = 0 and h = 1. Applying these conditions to Eq. (7.9), we obtain
F1(j, h) F2(j, h) $++++%++++& $+++11++%++++11+&
1 = (1 + 1)(1 + 1)(c1 + c2(1) + c3(1)) = 4c1 + 4c2 + 4c3 0 = (1 + 1)(1 + 0)(c1 + c2(1) + c3(0)) = 2c1 + 2c2 0 = (1 + 0)(1 + 1)(c1 + c2(0) + c3(1)) = 2c1 + 2c3 which results in c1 =  14, c2 = 14, and c3 = 14 with Sm = (1 + j)(1 + h)(  14 + 14 j + 14 h). The shape functions associated with the other corner nodes are determined in a similar fashion. The corner node shape functions are 1 Si =  (1  j)(1  h)(1 4 1 Sj = (1 + j)(1  h)( 1 4 1 Sm = (1 + j)(1 + h)( 1 4 1 Sn =  (1  j)(1 + h)(1 4
+ j + h) (7.10) + j  h) + j + h) + j  h)
Let us turn our attention to the shape functions for the middle nodes. As an e xample, we will develop the shape function associated with node o. First, F1 must be selected such that along the ijside (h = 1), inside (j = 1), and jmside (j = 1), the function will produce a value of zero. We select F1(j, h) = (1  j)(1 + h)(1 + j) Note that the product of the terms given by F1 will produce linear and nonlinear terms, as required by Eq. (7.8). Therefore, the second function F2 must be a constant; otherwise, the product of F1 and F2 will produce thirdorder polynomial terms, which we certainly do not want! So, F2(j, h) = c1 Applying the nodal condition So = 1 for j = 0 and h = 1 leads to
F1(j, h) $+++11+++%++++11++&
1 = (1  0)(1 + 1)(1 + 0)
M07_MOAV4303_04_GE_C07.INDD 377
F2(j, h) $%&
c1 = 2c1
27/11/14 9:55 AM
378 Chapter 7 TwoDimensional Elements
resulting in c1 = 12 with So = 12 (1  j)(1 + h)(1 + j) = So = 12 (1 + h)(1  j2). Using a similar procedure, we can obtain the shape functions for the midpoint nodes k, /, and p. Thus, the midpoint shape functions are 1 (1 2 1 S/ = (1 2 1 So = (1 2 1 Sp = (1 2
Sk =
 h)(1  j2)
(7.11)
+ j)(1  h2) + h)(1  j2)  j)(1  h2)
EXAMPLE 7.1 We have used twodimensional rectangular elements to model the stress distribution in a thin plate. The nodal stresses for an element belonging to the plate are shown in Figure 7.6. What is the value of stress at the center of this element? The stress distribution for the element is
s(e) = [Si
Sj
Sm
si s Sn] d j t sm sn
2345 lb/in2 Stress 2220 lb/in2
1925 lb/in2
1845 lb/in2
y m
n
0.15 in.
Y j
i
x
0.25 in. X
M07_MOAV4303_04_GE_C07.INDD 378
Figure 7.6 Nodal stresses for Example 7.1.
27/11/14 9:55 AM
Section 7.2 Quadratic Quadrilateral Elements 379
where si, sj, sm, and sn are stresses at nodes i, j, m, and n respectively, and the shape functions are given by Eq. (7.5): Si = ¢ 1 Sj =
y y x x ≤ ¢1  ≤ = ¢1 ≤ ¢1 ≤ w / 0.25 0.15
y y x x ¢1  ≤ = ¢1 ≤ w / 0.25 0.15
Sm =
xy xy = /w (0.25)(0.15)
Sn =
y y x x ¢1  ≤ = ¢1 ≤ w / 0.15 0.25
For the given element, the stress distribution in terms of the local coordinates x and y is given by
s(e)
Si si Sj sj Sm sm $++++%++++& $%& $+++%+++& $%& $++%++& $%& y y xy x x ≤ ¢1 ≤ (2220) + ¢1 ≤ (1925) + (1845) = ¢1 0.25 0.15 0.25 0.15 (0.25)(0.15) Sn sn $+++%+++& $%& y x + ¢1 ≤ (2345) 0.15 0.25 We can compute the stress at any point within this element from the aforementioned equation. Here, we are interested in the value of the stress at the midpoint. Substituting x = 0.125 and y = 0.075 into the equation, we have s(0.125,0.075) = 555 + 481 + 461 + 586 = 2083 lb/in2 Note that we could have solved this problem using natural coordinates. This approach may be easier because the point of interest is located at the center of the element j = 0 and h = 0. The quadrilateral natural shape functions are given by Eq. (7.7): Si = Sj = Sm = Sn = s(0.125,0.075) =
M07_MOAV4303_04_GE_C07.INDD 379
1 (1  j)(1  h) = 4 1 (1 + j)(1  h) = 4 1 (1 + j)(1 + h) = 4 1 (1  j)(1 + h) = 4 1 1 (2220) + (1925) 4 4
1 1 (1  0)(1  0) = 4 4 1 1 (1 + 0)(1  0) = 4 4 1 1 (1 + 0)(1 + 0) = 4 4 1 1 (1  0)(1 + 0) = 4 4 1 1 + (1845) + (2345) = 2083 lb/in2 4 4
27/11/14 9:55 AM
380 Chapter 7 TwoDimensional Elements
Thus, the stress at the midpoint of the rectangular element is the average of the nodal stresses. EXAMPLE 7.2 Confirm the expression given for the quadratic quadrilateral shape function Sn. Referring to the procedure discussed in Section 7.2, we can represent Sn by Sn = F1(j, h)F2(j, h) For the shape function Sn, F1 should be selected such that it will have a value of zero along the ijside (h = 1) and the jmside (j = 1) So, we choose F1(j, h) = (1  j)(1 + h) Furthermore, F2 is given by F2(j, h) = c1 + c2j + c3h and the coefficients c1, c2, and c3 are determined by applying the following conditions: Sn = 1 for j = 1 and h = 1 Sn = 0 for j = 0
and h = 1
Sn = 0 for j = 1 and h = 0 Recall from our discussion in the previous section, the coefficients in F2 are selected such that when F2 is multiplied by F1, they will produce a value of unity at node n, and a value of zero at adjacent neighboring nodes o and p. After applying these conditions we get 1 = 4c1  4c2 + 4c3 0 = 2c1 + 2c3 0 = 2c1  2c2 resulting in c1 =  14, c2 =  14, and c3 = 14, which is identical to the expression previously given for Sn. That is, 1 Sn =  (1  j)(1 + h)(1 + j  h) 4 7.3 Linear Triangular Elements A major disadvantage associated with using bilinear rectangular elements is that they do not conform to a curved boundary very well. In contrast, triangular elements, shown describing a twodimensional temperature distribution in Figure 7.7, are better suited to approximate curved boundaries. A triangular element is defined by three nodes, as shown in Figure 7.8. Therefore, we can represent the variation of a dependent variable, such as temperature, over the triangular region by
M07_MOAV4303_04_GE_C07.INDD 380
T (e) = a1 + a2X + a3Y (7.12)
27/11/14 9:55 AM
Section 7.3 Linear Triangular Elements 381 T
Tk
Tj
Ti
Y
k
j
X
Base
i
Figure 7.7 Using triangular elements to describe a twodimensional temperature distribution.
T Tk Tj
Ti (Xj ,Yj) Y (Xk ,Yk)
(Xi ,Yi) X
Figure 7.8 A triangular element.
Considering the nodal temperatures as shown in Figure 7.8, we must satisfy the following conditions:
T = Ti at X = Xi
and Y = Yi (7.13)
T = Tj at X = Xj
and Y = Yj
T = Tk at X = Xk and Y = Yk
M07_MOAV4303_04_GE_C07.INDD 381
27/11/14 9:55 AM
382 Chapter 7 TwoDimensional Elements
Substituting nodal values into Eq. (7.12), we have Ti = a1 + a2Xi + a3Yi (7.14) Tj = a1 + a2Xj + a3Yj Tk = a1 + a2Xk + a3Yk
Solving for a1, a2, and a3, we obtain 1 [(XjYk  XkYj)Ti + (XkYi  XiYk)Tj + (XiYj  XjYi)Tk] (7.15) 2A 1 [(Yj  Yk)Ti + (Yk  Yi)Tj + (Yi  Yj)Tk] a2 = 2A 1 a3 = [(Xk  Xj)Ti + (Xi  Xk)Tj + (Xj  Xi)Tk] 2A
a1 =
where A is the area of the triangular element and is computed from the equation 2A = Xi(Yj  Yk) + Xj(Yk  Yi) + Xk(Yi  Yj) (7.16)
See Example 7.3 for how Eq. (7.16) is derived. Substituting for a1, a2, and a3 into Eq. (7.12) and grouping Ti, Tj, and Tk terms yields T (e) = [Si
Sj
Ti Sk] c Tj s (7.17) Tk
where the shape functions Si, Sj, and Sk are 1 (a + biX + diY) (7.18) 2A i 1 Sj = (a + bjX + djY) 2A j 1 Sk = (a + bkX + dkY) 2A k
Si =
and
ai = XjYk  XkYj bi = Yj  Yk di = Xk  Xj aj = XkYi  XiYk bj = Yk  Yi dj = Xi  Xk ak = XiYj  XjYi bk = Yi  Yj dk = Xj  Xi Again, keep in mind that triangular shape functions have some basic properties, like other shape functions defined previously. For example, Si has a value of unity when evaluated at the coordinates of node i and has a value of zero at all other nodes. Or, as another example, the sum of the shape functions has a value of unity. That property is demonstrated by the equation
M07_MOAV4303_04_GE_C07.INDD 382
Si + Sj + Sk = 1 (7.19)
27/11/14 9:55 AM
Section 7.3 Linear Triangular Elements 383 h =1
j
Q
h
h =0
0 j =
j
A1 P
k Y
A2
A3
N
j =1
l
=1
M
l
=0
l
i
X Figure 7.9 Natural (area) coordinates for a triangular element.
Natural (Area) Coordinates for Triangular Elements Consider point P with coordinates (X, Y) inside the triangular region. Connecting this point to nodes, i, j, and k results in dividing the area of the triangle into three smaller areas A1, A2, and A3, as shown in Figure 7.9. Let us now perform an experiment. We move point P from its inside position to coincide with point Q along the kjedge of the element. In the process, the value of area A1 becomes zero. Moving point P to coincide with node i stretches A1 to fill in the entire area A of the element. Based on the results of our experiment, we can define a natural, or area, coordinate j as the ratio of A1 to the area A of the element so that its values vary from 0 to 1. Similarly, moving point P from its inside position to coincide with point M, along the kiedge, results in A2 = 0. Moving point P to coincide with node j stretches A2 such that it fills the entire area of the element; that is, A2 = A. We can define another area coordinate h as the ratio of A2 to A, and its magnitude varies from 0 to 1. Formally, for a triangular element, the natural (area) coordinates j, h, and l are defined by A1 (7.20) A A2 h = ` A A3 l = A
j =
It is important to realize that only two of the natural coordinates are linearly independent, because A3 A1 A2 A + + = = 1 = j + h + l A A A A
M07_MOAV4303_04_GE_C07.INDD 383
27/11/14 9:55 AM
384 Chapter 7 TwoDimensional Elements
For example, the [email protected] can be defined in terms of j and h by l = 1  j  h (7.21)
We also can show that the triangular natural (area) coordinates are exactly identical to the shape functions Si, Sj, and Sk. That is,
j = Si (7.22)
h = Sj
l = Sk As an example, consider j, which is the ratio of A1 to A: 1 [(XjYk  XkYj) + X(Yj  Yk) + Y(Xk  Xj)] A1 2 j = = (7.23) A 1 [X (Y  Yk) + Xj(Yk  Yi) + Xk(Yi  Yj)] 2 i j
Comparison of Eq. (7.23) to Eq. (7.18)* shows that j and Si are identical. Equation (7.23) was derived by describing triangular areas A1 and A in terms of the coordinates of their vertexes and using Eq. (7.16). Note that the coordinates of point P is designated by X and Y, because point P could lie anywhere within the area A.
EXAMPLE 7.3 Verify that the area of a triangular element can be computed from Equation (7.16). As shown in the accompanying diagram, the area of triangle ABD is equal to one half of the area of the parallelogram ABCD—with AB and AD as its sides. Y
C ,Y ) D (X k k
k
j e3
e2
A
i
B (Xj ,Yj)
(Xi ,Yi)
e1
X
*
Substitute for A, ai, bi, di in terms of nodal coordinates.
M07_MOAV4303_04_GE_C07.INDD 384
27/11/14 9:55 AM
Section 7.4 Quadratic Triangular Elements 385
S S The area of the parallelogram ABCD is equal to the magnitude of AB * AD ; that is, S S 2A = A B * AD where
u u S AB = (Xj  Xi)e1 + (Yj  Yi)e2 u u S AD = (Xk  Xi)e1 + (Yk  Yi)e2
u
u
u
and e1, e2, and e3 are unit vectors in the shown directions. Carrying out the crossprodS S uct operation in terms of the components of vectors AB and AD , u u u u S S 2A = AB * AD = [(Xj  Xi)e1 + (Yj  Yi)e2] * [(Xk  Xi)e1 + (Yk  Yi)e2] u
u
u
u
u
u
u
u
Noting that e1 * e1 = 0, e1 * e2 = e3, and e2 * e1 = e3, u
u
2A = (Xj  Xi)(Yk  Yi)e3  (Yj  Yi)(Xk  Xi)e3 and simplifying and regrouping terms, we show that the relationship given by Eq. (7.16) is true. That is, 2A = Xi(Yj  Yk) + Xj(Yk  Yi) + Xk(Yi  Yj) 7.4 Quadratic Triangular Elements The spatial variation of a dependent variable, such as temperature, over a region may be approximated more accurately by a quadratic function, such as T (e) = a1 + a2X + a3Y + a4X 2 + a5XY + a6Y 2 (7.24)
By now, you should understand how to develop shape functions. Therefore, the shape functions for a quadratic triangular element, shown in Figure 7.10, are given below without proof. The shape functions in terms of natural coordinates are
Si = j(2j  1)
(7.25)
Sj = h(2h  1)
Sk = l(2l  1) = 1  3(j + h) + 2(j + h)2
S/ = 4jh
Sm = 4hl = 4h(1  j  h)
Sn = 4jl = 4j(1  j  h)
EXAMPLE 7.4 We have used twodimensional triangular elements to model the temperature distribution in a fin. The nodal temperatures and their corresponding positions for an element are shown in Figure 7.11. (a) What is the value of temperature at X = 2.15 cm and
M07_MOAV4303_04_GE_C07.INDD 385
27/11/14 9:55 AM
386 Chapter 7 TwoDimensional Elements h= 0
l=1
k h= 1 – 2
n
m h=
l=1 – 2
1
i
j 0 j=
j=
1
/ 1– j= 2
l=0
k
m
n
i j
/
Figure 7.10 A quadratic triangular element.
T
775C 675C
695C
Y
j
(2.40, 1.65)
k (1.5, 1.0) i (2.25, 0.75) X
Figure 7.11 Nodal temperatures and coordinates for the element in Example 7.4.
M07_MOAV4303_04_GE_C07.INDD 386
27/11/14 9:55 AM
Section 7.4 Quadratic Triangular Elements 387
Y = 1.1 cm? (b) Determine the components of temperature gradients for this element. (c) Determine the location of 70°C and 75°C isotherms. a. The temperature distribution inside the element is T (e) = [Si
Sj
Ti Sk]c Tj s Tk
where the shape functions Si, Sj, and Sk are Si =
1 (a + biX + diY) 2A i
Sj =
1 (a + bjX + djY) 2A j
Sk =
1 (a + bkX + dkY) 2A k
and ai = XjYk  XkYj = (2.4)(1.0)  (1.5)(1.65) = 0.075 aj = XkYi  XiYk = (1.5)(0.75)  (2.25)(1.0) = 1.125 ak = XiYj  XjYi = (2.25)(1.65)  (2.40)(0.75) = 1.9125 bi = Yj  Yk = 1.65  1.0 = 0.65 bj = Yk  Yi = 1.0  0.75 = 0.25 bk = Yi  Yj = 0.75  1.65 = 0.9 di = Xk  Xj = 1.50  2.40 = 0.9 dj = Xi  Xk = 2.25  1.5 = 0.75 dk = Xj  Xi = 2.40  2.25 = 0.15 and 2A = Xi(Yj  Yk) + Xj(Yk  Yi) + Xk(Yi  Yj) 2A = 2.25(1.65  1.0) + 2.40(1.0  0.75) + 1.5(0.75  1.65) = 0.7125
M07_MOAV4303_04_GE_C07.INDD 387
Si =
1 1 (a + biX + diY) = ( 0.075 + 0.65X  0.9Y) 2A i 0.7125
Sj =
1 1 (aj + bjX + djY) = ( 1.125 + 0.25X + 0.75Y) 2A 0.7125
Sk =
1 1 (ak + bkX + dkY) = (1.9125  0.9X + 0.15Y) 2A 0.7125
27/11/14 9:55 AM
388 Chapter 7 TwoDimensional Elements
The temperature distribution for the element is T =
69 67 ( 0.075 + 0.65X  0.9Y) + ( 1.125 + 0.25X + 0.75Y) + 0.7125 0.7125 77 (1.9125  0.9X + 0.15Y) 0.7125
After simplifying, we have T = 93.632  10.808X  0.421Y Substituting for coordinates of the point X = 2.15 and Y = 1.1 leads to T = 69.935C. b. In general, the gradient components of a dependent variable Ψ(e) are computed from
0Ψ(e) 0 = [S Ψ + SjΨj + SkΨk] 0X 0X i i
0Ψ(e) 0 = [S Ψ + SjΨj + SkΨk] 0Y 0Y i i 0Ψ(e) 0X 1 bi d t = J 0Ψ(e) 2A di 0Y
bj dj
Ψi bk R c Ψj s (7.26) dk Ψk
It should be clear from examining Eq. (7.26) that the gradients have constant values. This property is always true for linear triangular elements. The temperature gradients are computed from 0T (e) 0X 1 bi d (e) t = J 0T 2A di 0Y
bj dj
Ti bk 1 0.65 R c Tj s = J dk 0.7125 0.9 Tk
0.25 0.75
69 0.9 10.808 R c 67 s = b r 0.15 0.421 77
Note that differentiation of the simplified temperature equation (T = 93.632  10.808X  0.421Y) directly would have resulted in exactly the same values. c. The location of 70°C and 75°C isotherms can be determined from the fact that over a triangular element, temperature varies linearly in both X and Ydirections. Thus, we can use linear interpolation to calculate coordinates of isotherms. First, let us focus on the 70°C constant temperature line. This isotherm will intersect the 77°C 69°[email protected] according to the relations 77  70 1.5  X 1.0  Y = = 77  69 1.5  2.25 1.0  0.75
M07_MOAV4303_04_GE_C07.INDD 388
27/11/14 9:55 AM
Section 7.5 Axisymmetric Elements 389 T
775C 755C
705C
675C
695C
Y (1.5, 1.0) (1.69, 0.94)
(2.40, 1.65)
(2. 13 ,1 .45 )
(1.68, 1.13)
(2.16, 0.78) (2.25, 0.75) X Figure 7.12 The isotherms of the element in Example 7.4.
which results in the coordinates X = 2.16 cm and Y = 0.78 cm. The 70°C isotherm also intersects the 77°C 67°[email protected]: 77  70 1.5  X 1.0  Y = = 77  67 1.5  2.4 1.0  1.65 These relations result in the coordinates X = 2.13 cm and Y = 1.45 cm. Similarly, the location of the 75°C isotherm is determined using the 77°C 69°[email protected]: 77  75 1.5  X 1.0  Y = = 77  69 1.5  2.25 1.0  0.75 which results in the coordinates X = 1.69 and Y = 0.94. Finally, along the 77°C 67°[email protected], we have: 77  75 1.5  X 1.0  Y = = 77  67 1.5  2.4 1.0  1.65 These equations result in the coordinates X = 1.68 and Y = 1.13. The isotherms and their corresponding locations are shown in Figure 7.12. 7.5 Axisymmetric Elements There are special classes of threedimensional problems whose geometry and loading are symmetrical about an axis, such as the zaxis, as shown in Figure 7.13. These threedimensional problems may be analyzed using twodimensional axisymmetric elements. In this section, we discuss the triangular and rectangular axisymmetric elements, and in Chapters 9 and 10 will show finite element formulations employing these types of elements.
M07_MOAV4303_04_GE_C07.INDD 389
27/11/14 9:55 AM
390 Chapter 7 TwoDimensional Elements z
z z n i
m
k
j
i
j
r
r
n
m k i
j
i
j
r
Figure 7.13 Examples of axisymmetric elements.
Axisymmetric Triangular Elements In Section 7.3, we developed the shape functions for a linear triangular element. Recall that the variation of any unknown variable Ψ over a triangular region in terms of its nodal values Ψi, Ψj, Ψk and the shape functions may be represented by Ψ(e) = [Si
Sj
Ψi Sk]c Ψj s Ψk
where 1 (a + biX + diY) 2A i 1 Sj = (a + bjX + djY) 2A j 1 Sk = (a + bkX + dkY) 2A k Si =
We now express the above shape functions in terms of r and zcoordinates—coordinates that are typically used for axisymmetric triangular elements. A typical axisymmetric triangular element and its coordinates is shown in Figure 7.14.
M07_MOAV4303_04_GE_C07.INDD 390
27/11/14 9:55 AM
Section 7.5 Axisymmetric Elements 391 z k
j
Zk i Zj Zi
r Ri Rk Rj Figure 7.14 An axisymmetric triangular element.
Substituting for the spatial coordinates X and Y in terms of r and zcoordinates and the nodal coordinates Xi, Yi, Xj, Yj and Xk, Yk in terms of Ri, Zi, Rj, Zj and Rk, Zk results in the following set of shape functions 1 (a + bir + diz) (7.27) 2A i 1 Sj = (a + bjr + djz) 2A j 1 Sk = (a + bkr + dkz) 2A k
Si =
where
ai = RjZk  RkZj
bi = Zj  Zk
di = Rk  Rj
aj = RkZi  RiZk
bj = Zk  Zi
dj = Ri  Rk
ak = RiZj  RjZi
bk = Zi  Zj
dk = Rj  Ri
Axisymmetric Rectangular Elements In Section 7.1 we discussed the rectangular element and derived the following shape functions.
M07_MOAV4303_04_GE_C07.INDD 391
27/11/14 9:55 AM
392 Chapter 7 TwoDimensional Elements z y
n
m
w i
Zn
j
x
Zi r Ri Rj Figure 7.15 An axisymmetric rectangular element.
Si = ¢ 1 
y x ≤ ¢1  ≤ w /
xy /w
Sm =
Sj =
y x ¢1  ≤ w /
Sn =
y x ¢1  ≤ w /
Let’s now consider the axisymmetric rectangular element shown in Figure 7.15. The relationship between the local coordinates x and y and the axisymmetric coordinates r and z is shown in Figure 7.15. Substituting for x in terms of r and for y in terms of z, and making use of the following relationships among the nodal coordinates, we get r = Ri + x
or
x = r  Ri
z = Zi + y
or
y = z  Zi
and x
y
b
b
/  (r  Ri) w  (z  Zi) y r  Ri z  Zi x Si = ¢1  ≤ ¢1  ≤ = £ 1 ≥£1 ≥ = £ ≥£ ≥ (7.28) / w / w / w
Realizing that / = Rj  Ri and w = Zn  Zi, we can simplify Eq. (7.28) to obtain the shape function Si in terms of r and z coordinates in the following manner:
M07_MOAV4303_04_GE_C07.INDD 392
27/11/14 9:55 AM
Section 7.5 Axisymmetric Elements 393
Si = £
/ $+%+& Rj  Ri  (r  Ri)
w $+%+& Rj  r Zn  z Zn  Zi  (z  Zi) ≥£ ≥ = ¢ ≤¢ ≤ w w /
/
Similarly, we can obtain the other shape functions. Thus, the shape functions for an axisymmetric rectangular element are given by Rj  r
Si = ¢
Sj = ¢
r  Ri Zn  z ≤¢ ≤ w /
Sm = ¢
r  Ri z  Zi ≤¢ ≤ w /
Sn = ¢
/
Rj  r /
≤¢
Zn  z ≤ (7.29) w
≤¢
z  Zi ≤ w
We discuss the application of these elements in solving heat transfer and solid mechanics problems in Chapters 9 and 10. EXAMPLE 7.5 We have used axisymmetric rectangular elements to model temperature distribution in a hollow cylinder. The values of nodal temperature for an element belonging to the cylinder are shown in Figure 7.16. What is the value of the temperature at r = 1.2 cm and z = 1.4 cm? z
50 C
47 C
n
48 C
2 cm
m
i
j 44 C
1 cm r 0.5 cm 2 cm Figure 7.16 Nodal temperatures and coordinates for the element in Example 7.5.
M07_MOAV4303_04_GE_C07.INDD 393
27/11/14 9:55 AM
394 Chapter 7 TwoDimensional Elements
The temperature distribution for the element is given by T (e) = [Si
Sj
Sm
Ti T Sn]d j t Tm Tn
where Ti, Tj, Tm, and Tn are the values of temperature at nodes i, j, m, and n respectively. Substituting for the coordinates of the point in the shape functions, we get Si = ¢
Rj  r /
≤¢
Zn  z 2  1.2 2  1.4 ≤ = ¢ ≤¢ ≤ = 0.32 w 1.5 1
Sj = ¢
r  Ri Zn  z 1.2  0.5 2  1.4 ≤¢ ≤ = ¢ ≤¢ ≤ = 0.28 w / 1.5 1
Sm = ¢
r  Ri z  Zi 1.2  0.5 1.4  1 ≤¢ ≤ = ¢ ≤¢ ≤ = 0.19 w / 1.5 1
Sn = ¢
Rj  r /
≤¢
z  Zi 2  1.2 1.4  1 ≤ = ¢ ≤¢ ≤ = 0.21 w 1.5 1
The temperature at the given point is T = (0.32)(48) + (0.28)(44) + (0.19)(47) + (0.21)(50) = 47.11°C 7.6 Isoparametric Elements As we discussed in Chapter 5, Section 5.5, when we use a single set of parameters (a set of shape functions) to define the unknown variables u, v, T, and so on, and use the same parameters (the same shape functions) to express the geometry, we are using an isoparametric formulation. An element expressed in such a manner is called an isoparametric element. Let us turn our attention to the quadrilateral element shown in Figure 7.17. Let us also consider a solid mechanics problem, in which a body undergoes a deformation. Using a quadrilateral element, the displacement field within an element belonging to this solid body can be expressed in terms of its nodal values as
u(e) = Si Uix + Sj Ujx + Sm Umx + Sn Unx (7.30)
v(e) = Si Uiy + Sj Ujy + Sm Umy + Sn Uny We can write the relations given by Eq. (7.30) in matrix form:
M07_MOAV4303_04_GE_C07.INDD 394
u S b r = J i v 0
0 Si
Sj 0
0 Sj
Sm 0
0 Sm
Sn 0
Uix Uiy Ujx 0 U R h jy x (7.31) Sn Umx Umy Unx Uny
27/11/14 9:55 AM
Section 7.6 Isoparametric Elements 395 h Uny
Umy Unx
(1, 1)
(1, 1)
Umx
m
n A(j, h)
j Uiy
(1, 1)
Ujy Uix j (1, 1)
i
Ujx
Figure 7.17 A quadrilateral element used in formulating planestress problems.
Note that using isoparametric formulation, we can use the same shape functions to describe the position of any point, such as A, within the element by the equations
x = Si xi + Sj xj + Smxm + Snxn (7.32)
y = Siyi + Sjyj + Smym + Snyn As you will see in Chapter 10, the displacement field is related to the components 0v 0v 0u of strains (exx = 0u 0x , eyy = 0y , and gxy = 0y + 0x ) and, subsequently, to the nodal displacements using shape functions. In deriving the elemental stiffness matrix from strain energy, we need to take the derivatives of the components of the displacement field with respect to the x and ycoordinates, which in turn means taking the derivatives of the appropriate shape functions with respect to x and y. At this point, keep in mind that the shape functions are expressed in terms of j and h Eq. (7.7). Thus, in general, it is necessary to establish relationships that allow the derivatives of a function f(x,y) to be taken with respect to x and y and to express them in terms of derivatives of the function f(x,y) with respect to j and h. This point will become clear soon. Using the chain rule, we can write
0f(x,y) 0f(x,y) 0x 0f(x,y) 0y = + 0j 0x 0j 0y 0j
0f(x,y) 0f(x,y) 0x 0f(x,y) 0y = + (7.33) 0h 0x 0h 0y 0h Expressing Eq. (7.33) in matrix form, we have
M07_MOAV4303_04_GE_C07.INDD 395
[J] $++%++& (7.34) 0f(x,y) 0f(x,y) 0x 0y 0j 0j 0j 0x d t = D T d t 0f(x,y) 0f(x,y) 0x 0y 0h 0h 0h 0y
27/11/14 9:55 AM
396 Chapter 7 TwoDimensional Elements
where the J matrix is referred to as the Jacobian of the coordinate transformation. The relationships of Eq. (7.34) can be also presented as 0f(x,y) 0f(x,y) 0x 0j d t = [J]  1 d t (7.35) 0f(x,y) 0f(x,y) 0y 0h
For a quadrilateral element, the J matrix can be evaluated using Eq. (7.32) and (7.7). This evaluation is left as an exercise for you; see problem 7.24. We will discuss the derivation of the element stiffness matrix using the isoparametric formulation in Chapter 10. For the sake of convenience, the results of Sections 7.1 to 7.6 are summarized in Table 7.1. Table 7.1 Twodimensional shape functions Linear Rectangular
y x ≤ ¢1  ≤ / w y x Sj = ¢ 1  ≤ / w xy Sm = /w y x Sn = ¢ 1  ≤ w / Si = ¢ 1 
y h (x = 0, y = w) (j = 1, h = 1)
(x = /, y = w) (j = 1, h = 1) m
n
w
j
(0, 0) j
i (j = 1, h = 1) (x = 0, y = 0)
Si =
1 (1  j)(1  h) 4
Sj =
1 (1 + j)(1  h) 4
x
(j = 1, h = 1) (x = /, y = 0) /
Sm =
1 (1 + j)(1 + h) 4
Sn =
1 (1  j)(1 + h) 4
1 Si =  (1  j)(1  h)(1 + j + h) 4
Quadratic Quadrilateral
Sj = n
Sm = o
m
(1, 1)
n
p
o
m
(1, 1)
h /
i
k j
M07_MOAV4303_04_GE_C07.INDD 396
p
(1, 1)
/
j
i
k
j
1 (1 + j)(1  h)(  1 + j  h) 4 1 (1 + j)(1 + h)(  1 + j + h) 4
1 Sn =  (1  j)(1 + h)(1 + j  h) 4 Sk =
1 (1  h)(1  j2) 2
S/ =
1 (1 + j)(1  h2) 2
So =
1 (1 + h)(1  j2) 2
Sp =
1 (1  j)(1  h2) 2
(1, 1)
27/11/14 9:55 AM
Section 7.7 TwoDimensional Integrals: Gauss–Legendre Quadrature 397
Table 7.1 Continued 1 (a + biX + diY) 2A i 1 (a + bjX + djY) Sj = 2A j 1 Sk = (a + bkX + dkY) 2A k ai = XjYk  XkYj bi = Yj  Yk
di = Xk  Xj
aj = XkYi  XiYk
bj = Yk  Yi
dj = Xi  Xk
ak = XiYj  XjYi
bk = Yi  Yj
dk = Xj  Xi
Si =
Linear Triangular h =1 Q A1 P A2 A3
h
h =0 k
Y
l
=1
M
0 j =
j N
j
j =1
l
=0
l
i X
Quadratic Triangular
Si = j(2j  1) h=
Sj = h(2h  1)
0
n
m
Sl = 4jh Sm = 4hl = 4h(1  j  h)
h=
1 l=– 2
Sk = l(2l  1) = 1  3(j + h) + 2(j + h)2
k h= 1 2–
l=1
Sn = 4jl = 4j(1  j  h)
0
j j=
/
1 j=
i
1– j= 2
1
l=0
k
m
n
i /
j
7.7 TwoDimensional Integrals: Gauss–Legendre Quadrature As we discussed in Chapter 5, most finite element programs perform numerical integration for elements by Gaussian quadratures, and as the limits of integration, they use an interval from 1 to 1. We now extend the Gauss–Legendre quadrature formulation to twodimensional problems as follows: 1
I =
1
L 1 L 1
1
f(j, h)dj dh ≅
L 1
J a wi f(ji, h) R dh ≅ a a wiwj f(ji, hj) (7.36) n
i=1
n
n
i=1 j=1
The relationships of Eq. (7.36) should be selfevident. Recall that the weighting factors and the sampling points are given in Chapter 5, Table 5.2. EXAMPLE 7.6 To demonstrate the steps involved in Gauss–Legendre quadrature computation, let us consider evaluating the integral 2
I =
M07_MOAV4303_04_GE_C07.INDD 397
L0 L0
2
(3y2 + 2x)dx dy
27/11/14 9:55 AM
398 Chapter 7 TwoDimensional Elements
As you know, the given integral can be evaluated analytically as 2
I = =
L0 L0 L0
2 2
(3y + 2x)dx dy =
2 2
2
[(3y x + x
)]20 dy
=
L0
L0
2
J
L0
2
(3y2 + 2x) dx R dy
2
(6y2 + 4) dy = 24
We now evaluate the integral using Gauss–Legendre quadrature. We begin by changing y and xvariables into j and h, using the relationships of Eq. (5.45): x = 1 + j
and
dx = dj
y = 1 + h
and
dy = dh
Thus, the integral I can be expressed by 2
I =
L0 L0
2
1
2
(3y + 2x)dx dy =
1
L 1 L 1
[3(1 + h)2 + 2(1 + j)]dj dh
Using the twopoint sampling formula, we have I ≅ a a wiwj f(ji, hj) n
n
i=1 j=1
I ≅ a a wiwj[3(1 + hj)2 + 2(1 + ji)] 2
2
i=1 j=1
To evaluate the summation, we start with i = 1, while changing j from 1 to 2, and we repeat the process with i = 2, while changing j from 1 to 2: I ≅ [(1)(1)[3(1 + ( 0.577350269))2 + 2(1 + ( 0.577350269))] + (1)(1)[3(1 + (0.577350269))2 + 2(1 + ( 0.577350269))]] + [(1)(1)[3(1 + ( 0.577350269))2 + 2(1 + (0.577350269))] + (1)(1)[3(1 + (0.577350269))2 + 2(1 + (0.577350269))]] = 24.000000000 7.8 Examples of TwoDimensional Elements in ANSYS ANSYS offers many twodimensional elements that are based on linear and quadratic quadrilateral and triangular shape functions. We will discuss the formulation of twodimensional thermal and solidstructural problems in detail in Chapters 9 and 10. For now, consider some examples of twodimensional structuralsolid and thermalsolid elements. Plane35 is a sixnode triangular thermal solid element. The element has one degree of freedom at each node, the temperature. Convection and heat fluxes may be input as surface loads at the element faces. The output data for this element include nodal temperatures and element data, such as thermal gradients and thermal fluxes.
M07_MOAV4303_04_GE_C07.INDD 398
27/11/14 9:55 AM
Chapter 7 References 399
Plane77 is an eightnode quadrilateral element used in modeling twodimensional heat conduction problems. It is basically a higher order version of the two dimensional, fournode quadrilateral element PLANE55. This element is more capable of modeling problems with curved boundaries. At each node, the element has a single degree of freedom, the temperature. Output data include nodal temperatures and element data, such as thermal gradient and thermal flux components. Plane182 is a fournode element used for 2D modeling of solid structures. The element can be used as either a plane element (plane stress, plane strain, or generalized plane strain) or an axisymmetric element. Plane183 can be used as a sixnode (triangular) or eightnode (quadrilateral) structural solid element. It has quadratic displacement behavior and is well suited for modeling irregular shapes. The element is defined by either six or eight nodes having two degrees of freedom at each node: translations in the nodal x and y directions. The element may be used as a plane element (plane stress, plane strain, and generalized plane strain) or as an axisymmetric element. The KEYOPT(1) is set to 1 for triangular shaped elements. Finally, it may be worth noting that although you generally achieve better results and greater accuracy with higher order elements, these elements require more computational time. This time requirement is because numerical integration of elemental matrices is more involved. Summary At this point you should 1. have a good understanding of the linear twodimensional rectangular and triangular shape functions and of elements, along with their properties and limitations. 2. have a good understanding of the quadratic twodimensional triangular and quadrilateral elements, as well as shape functions, along with their properties and their advantages over linear elements. 3. know why it is important to use natural coordinate systems. 4. know what is meant by axisymmetric element. 5. know what is meant by isoparametric element and formulation. 6. know how to use Gauss–Legendre quadrature to evaluate twodimensional integrals. 7. know examples of twodimensional elements in ANSYS. References ANSYS User’s Manual: Elements, Vol. III, Swanson Analysis Systems, Inc. Chandrupatla, T., and Belegundu, A., Introduction to Finite Elements in Engineering, Englewood Cliffs, NJ, Prentice Hall, 1991. CRC Standard Mathematical Tables, 25th ed., Boca Raton, FL, CRC Press, 1979. Segrlind, L., Applied Finite Element Analysis, 2nd ed., New York, John Wiley and Sons, 1984.
M07_MOAV4303_04_GE_C07.INDD 399
27/11/14 9:55 AM
400 Chapter 7 TwoDimensional Elements
Problems 1. We have used twodimensional rectangular elements to model temperature distribution in a thin plate. The values of nodal temperatures for an element belonging to such a plate are given in the accompanying figure. Using local shape functions, what is the temperature at the center of this element?
1125F Temperature 1085F 1025F 1055F y m
n
0.15 in
Y i
j
x
0.3 in X
2. Determine the temperature at the center of the element in Problem 1 using natural shape functions.
3. For a rectangular element, derive the x and ycomponents of the gradients of a dependent variable Ψ.
4. Determine the components of temperature gradients at the midpoint of the element in
Problem 1. Knowing that the element has a thermal conductivity of k = 100 Btu/hr # ft # 5F, compute the x and ycomponents of the heat flux. 5. Compute the location of the 1045F and 1065F isotherms for the element in Problem 1. Also, plot these isotherms. 6. Twodimensional triangular elements have been used to determine the stress distribution in a machine part. The nodal stresses and their corresponding positions for a triangular element are shown in the accompanying figure. What is the value of stress at x = 2.2 cm and y = 1.2 cm?
7. Plot the 8.1 GPa and 7.88 GPa stress contour lines for an element of the machine part in Problem 6.
8. For a quadratic quadrilateral element, confirm the expressions given for the shape functions Si and Sj.
M07_MOAV4303_04_GE_C07.INDD 400
27/11/14 9:55 AM
Chapter 7 Problems 401 Stress
8.27 GPa 7.58 GPa 7.96 GPa
Y
j (2.40, 1.65)
k (1.5, 1.0) i (2.25, 0.75) X
9. For a quadratic quadrilateral element, confirm the expressions given for the shape functions Sk and S/.
10. For triangular elements, the integral that includes products of area coordinates may be evaluated using the factorial relationship shown below: LA
jahblc dA =
a!b!c! 2A (a + b + c + 2)!
Using the above relationship, evaluate the integral
L
(S2i + SjSk) dA
A
11. Show that the area A of a triangular element can be computed from the determinant of 1 31 1
Xi Xj Xk
Yi Yj 3 = 2A Yk
gral 1 [S]T hT dA arises; h is the heat transfer coefficient, and T represents the temperature. A Using a linear triangular element, evaluate the aforementioned integral, provided that temperature variation is given by
12. In the formulation of twodimensional heat transfer problems, the need to evaluate the inte
T (e) = [Si
Sj
Ti Sk] c Tj s Tk
and h is a constant. Also, note that for triangular elements, the integral that includes products of area coordinates may be evaluated using the factorial relationship shown below: LA
M07_MOAV4303_04_GE_C07.INDD 401
jahblc dA =
a!b!c! 2A (a + b + c + 2)!
27/11/14 9:55 AM
402 Chapter 7 TwoDimensional Elements
13. In the formulation of twodimensional heat transfer problems, the need to evaluate the integral L A
k¢
0[S]T 0T ≤ dA 0X 0X
arises. Using a bilinear rectangular element, evaluate the aforementioned integral, provided temperature is given by
T (e) = [Si
Sj
Sm
Ti T Sn] d j t Tm Tn
and k is the thermal conductivity of the element and is a constant.
14. Look up the expressions for the ninenode quadratic quadrilateral element (Lagrangian element). Discuss its properties and compare it to the eightnode quadratic quadrilateral element. What is the basic difference between the Lagrangian element and its eightnode quadratic quadrilateral counterpart? 15. For triangular elements, show that the area coordinate h = Sj and the area coordinate l = Sk.
16. Verify the results given for natural quadrilateral shape functions in Eq. (7.7) by showing that (1) a shape function has a value of unity at its corresponding node and a value of zero at the other nodes and (2) if we sum up the shape functions, we will come up with a value of unity.
17. Verify the results given for natural quadratic triangular shape functions in Eq. (7.25) by showing that a shape function has a value of unity at its corresponding node and a value of zero at the other nodes.
18. For plane stress problems, using triangular elements, we can represent the displacements u and y using a linear triangular element similar to the one shown in the accompanying figure. Uky k
Ukx
Uiy Ujy
Uix i
Ujx j
The displacement variables, in terms of linear triangular shape functions and the nodal displacements, are
M07_MOAV4303_04_GE_C07.INDD 402
27/11/14 9:55 AM
Chapter 7 Problems 403 u = Si Uix + Sj Ujx + Sk Ukx v = Si Uiy + Sj Ujy + Sk Uky Moreover, for plane stress situations, the strain displacement relationships are exx =
0u 0x
eyy =
0v 0y
gxy =
0u 0v + 0y 0x
Show that for a triangular element, strain components are related to the nodal displacements according to the relation
exx bi 1 c eyy s = C0 2A gxy di
0 di bi
bj 0 dj
0 dj bj
Uix Uiy 0 U dk S f jx v Uiy bk Ukx Uky
bk 0 dk
19. Consider point Q along the kjside of the triangular element shown in the accompanying figure. Connecting this point to node i results in dividing the area of the triangle into two smaller areas A2 and A3, as shown. h=1 j=0 s
L jk
h=0
j
Q
Y k
A2
A3
j=1
i X
Along the kjedge, the natural, or area, coordinate j has a value of zero. Show that along the kjedge, the other natural (area) coordinates h and l reduce to onedimensional natural coordinates that can be expressed in terms of the local coordinate s according to the equations
M07_MOAV4303_04_GE_C07.INDD 403
h =
A2 s = 1 A Lj  k
l =
A3 s = A Lj  k
27/11/14 9:55 AM
404 Chapter 7 TwoDimensional Elements
20. For the element in Problem 19, derive the simplified area coordinates along the ij and kiedges using the onedimensional coordinate s.
21. As you will see in Chapters 9 and 10, we need to evaluate integrals along the edges of a triangular element to develop the load matrix in terms of surface loads or derivative boundary conditions. Referring to Problem 19 and making use of the relations L0
1
(x)m  1(1  x)n  1dx =
Γ(m)Γ(n) Γ(m + n)
Γ(n) = (n  1)! and Γ(m) = (m  1)! show that L
L0
1
a
¢1 
b
s s s a!b! ≤ ¢ ≤ d¢ ≤ = L L L L (a + b + 1)!
and L0
Lk  j
(h)a(l)bds = Lj  k
L0
a
1
¢1 
b
s a!b! s s ≤ ¢ ≤ d¢ ≤ = L Lj  k Lj  k Lj  k (a + b + 1)! j  k
22. Consider a triangular element subjected to a distributed load along its kiedge, as shown in the accompanying figure. j
k
j
k
px
py
i
i
y
x
Using the minimum total potentialenergy method, the differentiation of the work done by these distributed loads with respect to the nodal displacements gives the load matrix, which is computed from
where
5F 6(e) = Si 0 S [S]T = F j 0 Sk 0
M07_MOAV4303_04_GE_C07.INDD 404
0 Si 0 V Sj 0 Sk
[S]T 5p6dA, LA
and
5p6 = b
px r py
27/11/14 9:55 AM
Chapter 7 Problems 405 Realizing that along the kiedge, Sj = 0, evaluate the load matrix for a situation in which the load is applied along the kiedge. Use the results of Problem 21 to help you. Note, in this problem, A is equal to the product of the element thickness and the edge length.
23. For the element in Problem 22, evaluate the load matrices for a situation in which the distributed load is acting along the ijedge and the jkedge.
24. For a quadrilateral element, evaluate the Jacobian matrix J and its inverse J  1 using Eqs. (7.32) and (7.7).
25. Verify the shape function Sj, Sm, and Sn given in Eq. (7. 29) for the axisymmetric rectangular element.
26. We have used axisymmetric triangular elements to model temperature distribution in a system. The values of nodal temperature for an element belonging to the system are shown in the accompanying figure. What is the value of the temperature at r = 1.8 cm and z = 2 cm? z 74 C k
j
2.2 cm 67 C
i 70 C 1.7 cm 1.6 cm
r 1.5 cm 1.8 cm 2 cm
27. We have used axisymmetric rectangular elements to model temperature distribution in a system. The values of nodal temperature for an element belonging to the system are shown in the accompanying figure. What is the value of the temperature at r = 2.1 cm and z = 1.4 cm? 28. For an axisymmetric rectangular element, derive the r and zcomponents of the gradients of dependent variable Ψ.
29. Evaluate the given integral analytically and using Gauss–Legendre formula. 3
L0 L0
M07_MOAV4303_04_GE_C07.INDD 405
6
(5y3 + 2x2 + 5) dx dy
27/11/14 9:56 AM
406 Chapter 7 TwoDimensional Elements z
60 C
56 C
n
57 C
1.5 cm
m
i
j 53 C
1 cm r 1 cm 2.5 cm
30. Evaluate the given integral analytically and using Gauss–Legendre formula. 5
L2 L 2
10
(y3 + 3y2 + 5y + 2x2 + x + 10) dx dy
31. Show that for the rectangular element shown in Figure 7.2 and represented by Eq. (7.1), the temperature distribution varies linearly along the edges of the element, and it becomes nonlinear inside the element. (Hint: realize that along the ij edge, y = 0, along in, x = 0, along jm, x = /, and along the nm edge y = w.)
M07_MOAV4303_04_GE_C07.INDD 406
27/11/14 9:56 AM
C h a p t e r
8
More ANSYS*
The main objective of this chapter is to introduce the essential capabilities and the organization of the ANSYS program. The basic steps in creating and analyzing a model with ANSYS are discussed here, along with an example used to demonstrate these steps at the end of the chapter. The main topics discussed in Chapter 8 include the following: 8.1 ANSYS Program 8.2 ANSYS Database and Files 8.3 Creating a Finite Element Model with ANSYS: Preprocessing 8.4 hMethod Versus pMethod 8.5 Applying Boundary Conditions, Loads, and the Solution 8.6 Results of Your Finite Element Model: Postprocessing 8.7 Selection Options 8.8 Graphics Capabilities 8.9 ErrorEstimation Procedures 8.10 An Example Problem 8.1 Ansys Program The ANSYS program has two basic levels: the Begin level and the Processor level. When you first enter ANSYS, you are at the Begin level. From the Begin level, you can enter one of the ANSYS processors, as shown in Figure 8.1. A processor is a collection of functions and routines to serve specific purposes. You can clear the database or change a file assignment from the Begin level. There are three processors that are used most frequently: (1) the preprocessor (PREP7), (2) the processor (SOLUTION), and (3) the general postprocessor (POST1). The preprocessor (PREP7) contains the commands needed to build a model:
• define element types and options • define element real constants *Materials were adapted with permission from ANSYS documents.
M08_MOAV4303_04_GE_C08.INDD 407
27/11/14 9:59 AM
408 Chapter 8 More ANSYS
Begin level
PREP7
SOLUTION
POST1
POST26
OPT
Etc.
Processor level
Figure 8.1 Organization of ANSYS program.
• • • •
define material properties create model geometry define meshing controls mesh the object created
The solution processor (SOLUTION) has the commands that allow you to apply boundary conditions and loads. For example, for structural problems, you can define displacement boundary conditions and forces, or for heat transfer problems, you can define boundary temperatures or convective surfaces. Once all the information is made available to the solution processor (SOLUTION), it solves for the nodal solutions. The general postprocessor (POST1) contains the commands that allow you to list and display results of an analysis:
• • • •
read results data from results file read element results data plot results list results
There are other processors that allow you to perform additional tasks. For example, the timehistory postprocessor (POST26) contains the commands that allow you to review results over time in a transient analysis at a certain point in the model. The design optimization processor (OPT) allows the user to perform a design o ptimization analysis. 8.2 ANSYS Database and Files The previous section explained how the ANSYS program is organized. This section discusses the ANSYS database. ANSYS writes and reads many files during a typical analysis. The information you input when modeling a problem (e.g., element type, material property, dimensions, geometry, etc.) is stored as input data. During the solution phase, ANSYS computes various results, such as displacements, temperatures,
M08_MOAV4303_04_GE_C08.INDD 408
27/11/14 9:59 AM
Section 8.2 ANSYS Database and Files 409
stresses, etc. This information is stored as results data. The input data and the results data are stored in the ANSYS database. The database can be accessed from anywhere in the ANSYS program. The database resides in the memory until the user saves the database to a database file Jobname.DB. Jobname is a name that the user specifies upon entering the ANSYS program; this feature will be explained in more detail later. The database can be saved and resumed at any time. When you issue the RESUME_DB command, the database is read into the memory from the database file that was saved most recently. In other words, the database becomes what you saved most recently. When you are uncertain about the next step you should take in your analysis, or if you want to test something, you should issue the SAVE_DB command before proceeding with your test. That way, if you are unhappy with the results of your test, you can issue the RESUME_DB command, which will allow you to go back to the place in your analysis where you started testing. The SAVE_DB, RESUME_DB, and QUIT commands are located in the ANSYS toolbar. In addition, the “Clear & Start New” option, located on the utility menu, allows the user to clear the database. This option is useful when you want to start a new, but do not wish to leave and reenter ANSYS. When you are ready to exit the ANSYS program, you will be given four options: (1) Save all model data; (2) Save all model data and solution data; (3) Save all model data, solution data, and postprocessing data; or (4) Save nothing. As previously explained, ANSYS writes and reads many files during a typical analysis. The files are of the form of Jobname.Ext. Recall that Jobname is a name you specify when you enter the ANSYS program at the beginning of an analysis. The default jobname is file. Files are also given unique extensions to identify their content. Typical files include the following:
• The log file (Jobname.LOG): This file is opened when ANSYS is first entered. Every command you issue in ANSYS is copied to the log file. Jobname.LOG is closed when you exit ANSYS. Jobname.LOG can be used to recover from a system crash or a serious user error by reading in the file with the /INPUT command. • The error file (Jobname.ERR): This file is opened when you first enter ANSYS. Every warning and error message given by ANSYS is captured by this file. If Jobname.ERR already exists when you begin a new ANSYS session, all new warnings and error messages will be appended to the bottom of this file. • The database file (Jobname.DB): This file is one of the most important ANSYS files because it contains all of your input data and possibly some results. The model portion of the database is automatically saved when you exit the ANSYS program. • The output file (Jobname.OUT): This file is opened when you first enter ANSYS. Jobname.OUT is available if you are using the GUI; otherwise, your computer monitor is your output file. Jobname.OUT captures responses given by ANSYS to every command executed by the user. It also records warning and error messages and some results. If you change the Jobname while in a given ANSYS session, the output file name is not changed to the new Jobname.
M08_MOAV4303_04_GE_C08.INDD 409
27/11/14 9:59 AM
410 Chapter 8 More ANSYS
Other ANSYS files include the structural analysis results file (Jobname.RST); the t hermal results file (Jobname.RTH); the magnetic results file (Jobname.RMG); the graphics file (Jobname.GRPH); and the element matrices file (Jobname.EMAT). 8.3 Creating a Finite Element Model with ANSYS: Preprocessing The preprocessor (PREP7) contains the commands needed to create a finite element model: 1. define element types and options 2. define element real constants if required for the chosen element type 3. define material properties 4. create model geometry 5. define meshing controls 6. mesh the object created
1. Define element types and options. ANSYS provides 150 various elements to be used to analyze different problems. Selecting the correct element type is a very important part of the analysis process. A good understanding of finite element theory will benefit you the most in this respect, helping you choose the correct element for your analysis. In ANSYS, each element type is identified by a category name followed by a number. For example, twodimensional solid elements have the category name PLANE. Furthermore, PLANE182 is a fournode quadrilateral element used to model structural solid problems. The element is defined by four nodes having two degrees of freedom at each node, translation in the x and ydirections. PLANE183 is an eightnode (four corner points and four midside nodes) quadrilateral element used to model twodimensional structural solid problems. It is a higher order version of the twodimensional, fournode quadrilateral element type, PLANE182. Therefore, the PLANE183 element type offers more accuracy when modeling problems with curved boundaries. At each node, there are two degrees of freedom, translation in the x and ydirections. Many of the elements used by ANSYS have options that allow you to specify additional information for your analysis. These options are known in ANSYS as keyoptions (KEOPTs). For example, for PLANE183, with KEOPT (3) you can choose plane stress, axisymmetric, plane strain, or plane stress with the thickness analysis option. You can define element types and options by choosing main menu: Preprocessor S Element Type S Add/Edit/Delete A
You will see the Element Types dialog box next, shown in Figure 8.2. List: A list of currently defined element types will be shown here. If you have not defined any elements yet, then you need to use the Add button to add an element.
M08_MOAV4303_04_GE_C08.INDD 410
27/11/14 9:59 AM
Section 8.3 Creating a Finite Element Model with ANSYS: Preprocessing 411
A
B Figure 8.2 Element Types dialog box.
Figure 8.3 Library of Element Types dialog box.
B
The Library of Element Types dialog box will appear next (see Figure 8.3). Then you choose the type of element you desire from the Library. Action Buttons: The purpose of the Add button is to add an element, as we just discussed. The Delete button deletes the selected (highlighted) element type. The Options button opens the element type options dialog box. You can then choose one of the desired element options for a selected element. For example, if you had selected the element PLANE183 with KEOPT (3) you could choose plane stress, axisymmetric, plane strain, or plane stress with the thickness analysis option, as shown in Figure 8.4.
2. Define element real constants. Element real constants are quantities that are specific to a particular element. For example, a truss element requires crosssectional area. It is important to realize that real constants vary from one element type to another; furthermore,
M08_MOAV4303_04_GE_C08.INDD 411
27/11/14 9:59 AM
412 Chapter 8 More ANSYS
Figure 8.4 The element type options dialog box.
not all elements require real constants. Real constants can be defined by the command main menu: Preprocessor S Real Constants S Add/Edit/Delete You will then see the Real Constants dialog box, as shown in Figure 8.5. List: A list of currently defined real constants will be shown here. If you have not defined any real constants at this point, you need to use the Add button to add real constants. An example of a dialog box for a PLANE182 element’s real constants is shown in Figure 8.6. Action Buttons: The purpose of the Add button has already been explained. The B Delete Button deletes the selected (highlighted) real constants. The Edit button opens a new dialog box that allows you to change the values of existing real constants. 3. Define material properties. At this point, you define the physical properties of your material. For example, for solid structural problems, you may need to define the modulus of elasticity, Poisson’s ratio, or the density of the material, whereas for thermal problems, you A
A
B Figure 8.5 Real Constants dialog box.
M08_MOAV4303_04_GE_C08.INDD 412
27/11/14 9:59 AM
Section 8.3 Creating a Finite Element Model with ANSYS: Preprocessing 413
Figure 8.6 An example of the dialog box for a PLANE 82 (with options) element’s real constants.
may need to define thermal conductivity, specific heat, or the density of the material. You can define material properties by the command main menu: Preprocessor S Material Props S Material Models You will then see the Define Material Model Behavior dialog box, as shown in Figure 8.7. The next dialog box allows you to define the appropriate properties for your analysis, as shown in Figure 8.8. You can use multiple materials in your model if the object you are analyzing is made of different materials. From the Define Material Model Behavior dialog box, click on Material button and then choose New Model.
4. Create model geometry. There are two approaches to constructing a finite element model’s geometry: (1) direct (manual) generation and (2) the solidmodeling approach. Direct generation, or manual generation, is a simple method by which you specify the location of nodes and manually define which nodes make up an element. This approach is generally applied to simple problems that can be modeled with line elements, such as links, beams, and pipes, or if the object is made of simple geometry, such as rectangles. This approach is illustrated in Figure 8.9. Refer back to the truss problem of Example 3.1 in Chapter 3 to refresh your memory about the manual approach, if necessary.
M08_MOAV4303_04_GE_C08.INDD 413
27/11/14 9:59 AM
414 Chapter 8 More ANSYS
Figure 8.7 Material Model dialog box.
Figure 8.8a Isotropic structural material properties dialog box.
With the solidmodeling approach, you use simple primitives (simple geometric shapes), such as rectangles, circles, polygons, blocks, cylinders, and spheres, to construct the model. Boolean operations are then used to combine the primitives. Examples of boolean operations include addition, subtraction, and intersection. You then specify the desired element size and shape, and ANSYS will automatically generate all the nodes and the elements. This approach is depicted in Figure 8.10.
M08_MOAV4303_04_GE_C08.INDD 414
27/11/14 9:59 AM
Section 8.3 Creating a Finite Element Model with ANSYS: Preprocessing 415
Figure 8.8b Linear Isotropic Material Properties dialog box.
500 lb 3 ft
500 lb 3 ft
3 ft
500 lb (3)
500 lb (6)
4
3 (2)
(4)
5 (5)
(1) 1
2
Figure 8.9 A truss problem: First nodes 1–5 are created, then nodes are connected to form elements (1)–(6).
Figure 8.10 An example of the solidmodeling approach.
M08_MOAV4303_04_GE_C08.INDD 415
27/11/14 9:59 AM
416 Chapter 8 More ANSYS
Solid Modeling Direct Generation
Figure 8.11 The Create dialog box.
The construction of the model in ANSYS begins when you choose the Create option, as shown in Figure 8.11. You choose this option with the command sequence main menu: Preprocessor S Modeling S Create When you create entities such as keypoints, lines, areas, or volumes, they are automatically numbered by the ANSYS program. You use keypoints to define the vertices of an object. Lines are used to represent the edges of an object. Areas are used to represent twodimensional solid objects. They are also used to d efine the surfaces of threedimensional objects. When using primitives to build a m odel, you need to pay special notice to the hierarchy of the entities. Volumes are bounded by areas, areas are bounded by lines, and lines are bounded by keypoints. Therefore, volumes are considered to be the highest entity, and the keypoints are the lowest entity in solid modeling hierarchy. Remembering this concept is particularly important if you need to delete a primitive. For example, when you define one rectangle, ANSYS automatically creates nine entities: four keypoints, four lines, and one area. The relationship among keypoints, lines, and areas is depicted in Figure 8.12. K4
L4
L3
K3
L2
A1
K1
M08_MOAV4303_04_GE_C08.INDD 416
L1
K2
Figure 8.12 The relationship among the keypoints, lines, and areas.
27/11/14 9:59 AM
Section 8.3 Creating a Finite Element Model with ANSYS: Preprocessing 417
By picking By typing in x, y coordinates Figure 8.13 The Rectangle menu.
Area primitives and volume primitives are grouped under the Areas and Volumes categories in the Create menu. Now let us consider the Rectangle and the Circle menus, because they are commonly used to build twodimensional models. The Rectangle menu offers three methods for defining a rectangle, as shown in Figure 8.13. The command for accessing the Rectangle menu is main menu: Preprocessor S Modeling S Create S Rectangle The Circle menu offers several methods for defining a solid circle or annulus, as shown in Figure 8.14. The Partial Annulus option is limited to circular areas spanning 180° or less. In order to create a partial circle that spans more than 180°, you need to use the By Dimensions option. An example of creating a partial annulus spanning from u = 45° to u = 315° is shown in Figure 8.15. Note that you can create a solid circle by setting [email protected] = 0. The Working Plane (WP) In ANSYS, you will use a working plane (WP) to create and orient the geometry of the object you are planning to model. All primitives and other modeling entities are
By picking By typing in radii and angles Figure 8.14 The Circle menu.
M08_MOAV4303_04_GE_C08.INDD 417
27/11/14 9:59 AM
418 Chapter 8 More ANSYS
RAD1 +u
RAD2
Figure 8.15 An example of creating a partial annulus spanning from u = 45° to u = 315°.
defined with respect to this plane. The working plane is basically an infinite plane with a twodimensional coordinate system. The dimensions of the geometric shapes are defined with respect to the WP. By default, the working plane is a Cartesian plane. You can change the coordinate system to a polar system, if so desired. Other attributes of the working plane may be set by opening the WP settings dialog submenu, as shown in Figure 8.16. To access this dialog box, issue the following sequence of commands: utility menu: Work Plane S WP Settings c A
B C
Coordinate System: Choose the workingplane coordinate system you want to use. You locate or define points in terms of X and Ycoordinates when using the Cartesian coordinate system. You can also locate or define points with respect to a polar coordinate system using R and [email protected] Display Options: This section is where you turn on the grid or grid and triad. The triad appears in the center (0,0 coordinates) of your working plane. Snap Options: These options control the locations of points that are picked. When activated, these options allow you to pick locations nearest to the snap point. For example, in a Cartesian working plane, Snap Incr controls the X and Yincrements within the spacing grid. If you have set a spacing of 1.0 and a snap increment of 0.5, then within the X,Y grid you can pick coordinates with 0.5 increments. For example, you cannot pick the coordinates 1.25 or 1.75.
M08_MOAV4303_04_GE_C08.INDD 418
27/11/14 9:59 AM
Section 8.3 Creating a Finite Element Model with ANSYS: Preprocessing 419
A
B C
D
Figure 8.16 The WP settings dialog box.
D
Grid Control Spacing: This number defines the spacing between the grid lines. Minimum: This number is the minimum Xlocation at which you want the grid to be displayed with respect to the Cartesian coordinate system. Maximum: This number is the maximum Xlocation at which you want the grid to be displayed with respect to the Cartesian coordinate system. Radius: This number is the outside radius that you want the grid to be displayed with respect to the polar coordinate system. Tolerance: This number is the amount that an entity can be off of the current working plane and still be considered as on the plane. The working plane is always active and, by default, not displayed. To display the working plane, you need to issue the following command: utility menu: Work Plane S Display Working Plane You can move the WP origin to a different location on the working plane. This feature is useful when you are defining primitives at a location other than the global location. You can move the WP origin by choosing the commands utility menu: Work Plane S Offset WP to S XYZ Locations +
M08_MOAV4303_04_GE_C08.INDD 419
27/11/14 9:59 AM
420 Chapter 8 More ANSYS
A
B
C
D
E
Figure 8.17 The dialog box for offsetting the WP.
You can relocate the working plane by offsetting it from its current location, as shown in Figure 8.17. To do so, issue the command utility menu: Work Plane S Offset WP by Increments c A
B
C
Offset buttons: Picking these buttons will cause an immediate offset of your working plane in the direction shown on the buttons. The amount of offset is controlled by the Offset slider and the SnapIncr value on the WP setting dialog box. Offset Slider: This number controls the amount of offset that occurs with each pick of the offset buttons. If the slider is set to 1, the offset will be one times the SnapIncr value on the WP setting dialog box. Offset Dialog Input: This feature allows you to input the exact X, Y, and Z offset values for the working plane. For instance, typing 1,2,2 into this field and pressing
M08_MOAV4303_04_GE_C08.INDD 420
27/11/14 9:59 AM
Section 8.3 Creating a Finite Element Model with ANSYS: Preprocessing 421
E
the Apply or OK button will move the working plane one unit in the positive Xdirection and two units each in the positive Y and Zdirections. Location Status: This section displays the current location of the working plane in global Cartesian coordinates. This status is updated each time the working plane is translated. You can also relocate the working plane by aligning it with specified keypoints, nodes, coordinate locations, and so on, as shown in Figure 8.18. To align the working plane, issue the command utility menu: Work Plane S Align WP with
Plotting Model Entities You can plot various entities, such as keypoints, lines, areas, volumes, nodes, and elements, using the Plot menu. From the utility menu, you can issue one of the following commands to plot: utility menu: Plot S Keypoints utility menu: Plot S Lines utility menu: Plot S Areas utility menu: Plot S Volumes utility menu: Plot S Nodes utility menu: Plot S Elements The Plot Numbering Controls menu, shown in Figure 8.19, contains a useful graphics option that allows you to turn on keypoint numbers, line numbers, area numbers, and so on to check your model. To access this option, use the command utility menu: PlotCtrls S Numbering c You may need to replot to see the effects of the numbering command you issue. 5. Define meshing controls. The next step in creating a finite element model is dividing the geometry into nodes and elements. This process is called meshing. The ANSYS program can automatically generate the nodes and elements, provided that you specify the element attributes and the element size: Keypoints Nodes XYZ Locations Plane Normal to Line Active Coord Sys Specified Coord Sys... Global Cartesian
M08_MOAV4303_04_GE_C08.INDD 421
+ + + +
Figure 8.18 Working planerelocation using the Align command.
27/11/14 9:59 AM
422 Chapter 8 More ANSYS
Figure 8.19 The Plot Numbering Controls dialog box.
1. The element attributes include element type(s), real constants, and material properties. 2. The element size controls the fineness of the mesh. The smaller the element size, the finer the mesh. The simplest way to define the element size is by defining a global element size. For example, if you specify an element edge length of 0.1 units, then ANSYS will generate a mesh in which no element edge is larger than 0.1 units. Another way to control the mesh size is by specifying the number of element divisions along a boundary line. The Global Element Sizes dialog box is shown in Figure 8.20. To access this dialog box, issue the following commands:
main menu: Preprocessor S Meshing S Size Cntrls S Manual Size S Global S Size
6. Mesh the object. You should get into the habit of saving the database before you initiate meshing. This way, if you are not happy with the mesh generated, you can resume the database and change the element size and remesh the model. To initiate meshing, invoke the commands main menu: Preprocessor S Meshing S Mesh S Areas S Free
M08_MOAV4303_04_GE_C08.INDD 422
27/11/14 9:59 AM
Section 8.3 Creating a Finite Element Model with ANSYS: Preprocessing 423
Figure 8.20 The Global Element Size dialog box.
Once a picking menu appears, you can pick individual areas or use the Pick All button to select all areas for meshing. Upon selection of the desired areas, Pick the Apply or OK buttons to mesh. The meshing process can take some time, depending on the model complexity and the speed of your computer. During the meshing process, ANSYS periodically writes a meshing status to the output window. Therefore, it is useful to bring the output window to the front to see the meshing status messages. Free meshing uses either mixedarea element shapes or alltriangular area elements, whereas the mapped meshing option uses all quadrilateral area elements and all hexahedral (brick) volume elements. Mapped area mesh requirements include three or four sides, equal numbers of elements on opposite sides, and even numbers of elements for threesided areas. If you want to mesh an area that is bounded by more than four lines, you can use the concatenate command to combine some of the lines to reduce the total number of lines. Concatenation is usually the last step you take before you start meshing the model. To concatenate, issue the following series of commands: main menu: Preprocessor S Meshing S Concatenate S Lines or Areas Modifying Your Meshed Model If you want to modify your model, you must keep in mind certain rules enforced by ANSYS: 1. Meshed lines, areas, or volumes may not be deleted or moved. 2. You can delete the nodes and the elements with the meshing Clear command. Also, areas contained in volumes may not be deleted or changed. Lines contained in areas may not be deleted. Lines can be combined or divided into smaller segments with line operation commands. Keypoints contained in lines may not be deleted. You start the clearing process by issuing the commands main menu: Preprocessor S Meshing S Clear
M08_MOAV4303_04_GE_C08.INDD 423
27/11/14 9:59 AM
424 Chapter 8 More ANSYS
The clearing process will delete nodes and elements associated with a selected model entity. Then you can use deleting operations to remove all entities associated with an entity. The “ c and below” options delete all lower entities associated with the specified entity, as well as the entity itself. For example, deleting “Area and below” will automatically remove the area, the lines, and the keypoints associated with the area. 8.4 hMethod Versus pMethod For solid structural problems, to obtain displacements, stresses, or strain, ANSYS offers two solution methods: hmethod and pmethod. The hmethod makes use of elements that are based on shape functions that are typically quadratic. We have already pointed out that the elements based on quadratic shape functions are better than linear elements but not as good as higher order elements, such as cubic. We have also explained that the element size affects the accuracy of your results as well. Whereas the hmethod makes use of quadratic elements, the pmethod uses higher order polynomial shape functions to define elements. Because the hmethod uses quadratic elements, mesh refinement may be necessary if you desire very accurate results. A simple way to determine if the element size is fine enough to produce good results is by solving the problem with a certain number of elements and then comparing its results to the results of a model with twice as many elements. If you detect substantial difference between the results of the two models, then the mesh refinement is necessary. You may have to repeat the process by refining the mesh until you can’t detect a substantial difference between the models. As mentioned above, the pmethod—which can be used only for linear structural static problems—makes use of polynomial shape functions that are of higher orders than quadratic ones. The user can specify a degree of accuracy, and the pmethod manipulates the order of the polynomial or the plevel to better fit the degree of difficulty associated with the boundaries of the given problem and its behavior to the applied loads. A problem is solved at a given plevel, and then the order of the polynomial is increased and the problem is solved again. The results of the iterations are then compared to a set of convergence criteria that the user has specified. In general, a higher plevel approximation leads to better results. One of the main benefits of using the pmethod is that in order to obtain good results, you don’t need to manually manipulate the size of elements by creating finer meshes. Depending on the problem, a coarse mesh can provide reasonable results. Finally, as you will learn in Section 8.9, ANSYS offers errorestimation procedures that calculate the level of solution errors due to meshing employed. It is worth noting that the pmethod adaptive refinement procedure offers error estimates that are more precise than those of the hmethod and can be calculated locally at a point or globally. 8.5 Applying Boundary Conditions, Loads, and the Solution The next step of finite element analysis involves applying appropriate boundary conditions and the proper loading. There are two ways to apply the boundary conditions and loading to your model in ANSYS. You can either apply the conditions to the solid model
M08_MOAV4303_04_GE_C08.INDD 424
27/11/14 9:59 AM
Section 8.5 Applying Boundary Conditions, Loads, and the Solution 425
(keypoints, lines, and areas), or the conditions can be directly imposed on the nodes and elements. The first approach may be preferable because should you decide to change the meshing, you will not need to reapply the boundary conditions and the loads to the new finite element model. It is important to note that if you decide to apply the conditions to keypoints, lines, or areas during the solution phase, ANSYS automatically transfers the information to nodes. The solution processor (SOLUTION) has the commands that allow you to apply boundary conditions and loads. It includes the following options: for structural problems: displacements, forces, distributed loads (pressures), temperatures for thermal expansion, gravity for thermal problems: temperatures, heat transfer rates, convection surfaces, internal heat generation for fluid flow problems: velocities, pressures, temperatures for electrical problems: voltages, currents for magnetic problems: potentials, magnetic flux, current density Degrees of Freedom (DOF) Constraints In order to constrain a model with fixed (zero displacements) boundary conditions, you need to choose the command sequence main menu: Solution S Define Loads S Apply S Structural S Displacement You can specify the given condition on the keypoints, lines, areas, or nodes. For example, if you choose to constrain certain keypoints, then you need to invoke the commands
main menu: Solution S Define Loads S Apply S Structural S Displacement S On Keypoints
A picking menu will appear. You then pick the keypoints to be constrained and press the OK button. An example of a dialog box for applying displacement constraints on keypoints is shown in Figure 8.21. The KEXPND field in the dialog box of Figure 8.21 is used to expand the constraint specification to all nodes between the keypoints, as shown in Figure 8.22. Once you have applied the constraints, you may want to display the constraint symbols graphically. To turn on the boundary condition symbols, open the Symbols dialog box, as shown in Figure 8.23, by choosing the commands utility menu: PlotCtrls S Symbols c Line or Surface Loads In order to specify distributed loads on a line or surface of a model, you need to issue the following commands:
main menu: Solution S Define Loads S Apply S Structural S Pressures S On Lines or On Areas
M08_MOAV4303_04_GE_C08.INDD 425
27/11/14 9:59 AM
426 Chapter 8 More ANSYS
(a)
(b) Figure 8.21 The dialog box for applying displacements on keypoints.
A picking menu will appear. You then pick the line(s) or surfaces that require a pressure load and press the OK button. An example of a dialog box for applying pressure loads on line(s) is shown in Figure 8.24. For uniformly distributed loads, you need to specify only VALI. For a linear distribution, you need to specify both VALI and VALJ, as shown in Figure 8.25. It is important to note that in ANSYS, a positive VALI represents pressure into the surface.
M08_MOAV4303_04_GE_C08.INDD 426
27/11/14 9:59 AM
Section 8.6 Results of Your Finite Element Model: Postprocessing 427 KEXPND = NO Transfer to K1
K2
model Nodes
KEXPND = YES Transfer to K1
K2
model
Figure 8.22 KEXPND options.
Obtain a Solution Once you have created the model and have applied the boundary conditions and appropriate loads, then you need to instruct ANSYS to solve the set of equations generated by your model. But first save the database. To initiate the solution, pick the commands main menu: Solution S Solve S Current LS The next section is about reviewing the results of your analysis. 8.6 Results of Your Finite Element Model: Postprocessing There are two postprocessors available for review of your results: (1) POST1 and (2) POST26. The general postprocessor (POST1) contains the commands that allow you to list and display results of an analysis:
• • • •
Deformed shape displays and contour displays Tabular listings of the results data of the analysis Calculations for the results data and path operations Error estimations
You can read results data from the results file by using one of the choices from the dialog box shown in Figure 8.26. This dialog box may be accessed via the following command: main menu: General Postproc For example, if you are interested in viewing the deformed shape of a structure under a given loading, you choose the Plot Deformed Shape dialog box, as shown in Figure 8.27. To access this dialog box, issue the following sequence of commands: main menu: General Postproc S Plot Results S Deformed Shape
M08_MOAV4303_04_GE_C08.INDD 427
27/11/14 9:59 AM
428 Chapter 8 More ANSYS
Figure 8.23 The Symbols dialog box.
You can also use contour displays to see the distribution of certain variables, such as a component of stress or temperature over the entire model. For example, issue the following command to access the dialog box shown in Figure 8.28.
main menu: General Postproc S Plot Results S Contour Plot S Nodal Solution
M08_MOAV4303_04_GE_C08.INDD 428
27/11/14 9:59 AM
Section 8.6 Results of Your Finite Element Model: Postprocessing 429
Figure 8.24 The dialog box for applying pressure loads on lines.
100
100
VALI = 100
200 100 VALI = 100 VALJ = 200
200 VALI = 200 VALJ = 100
100 Figure 8.25 An example illustrating how to apply uniform and nonuniform loads.
As already mentioned, you can list the results in a tabular form as well. For example, to list the reaction forces, you issue the following command, which gives you a dialog box similar to the one shown in Figure 8.29: main menu: General Postproc S List Results S Reaction Solu Select the component(s) of your choice and press the OK button.
M08_MOAV4303_04_GE_C08.INDD 429
27/11/14 9:59 AM
430 Chapter 8 More ANSYS
Figure 8.26 The General Postprocessing dialog box.
Figure 8.27 The Plot Deformed Shape dialog Box.
The timehistory postprocessor (POST26) contains the commands that allow you to review results over time in a transient analysis. These commands will not be discussed here, but you may consult the ANSYS online help for further information about how to use the timehistory postprocessor. Once you have finished reviewing the results and wish to exit the ANSYS program, choose the Quit button from the ANSYS toolbar and pick the option you want. Press the OK button.
M08_MOAV4303_04_GE_C08.INDD 430
27/11/14 9:59 AM
Section 8.6 Results of Your Finite Element Model: Postprocessing 431
Figure 8.28 The Contour Nodal Solution Data dialog box.
Figure 8.29 The List Reaction Solution dialog box.
M08_MOAV4303_04_GE_C08.INDD 431
27/11/14 9:59 AM
432 Chapter 8 More ANSYS
If, for any reason, you need to come back to modify a model, first launch ANSYS, and then type the filename in the Initial Jobname entry field of the interactive dialog box. Then press the Run button. From the file menu, choose Resume Jobname.DB. Now you have complete access to your model. You can plot keypoints, nodes, elements, and the like to make certain that you have chosen the right problem. 8.7 Selection Options The ANSYS program uses a database to store all of the data that you define during an analysis. ANSYS also offers the user the capability to select information about only a portion of the model, such as certain nodes, elements, lines, areas, and volumes for further processing. You can select functions anywhere within ANSYS. To start selecting, issue the following command to bring up the dialogue box shown in Figure 8.30. utility menu: Select S Entities . . . The various selection commands and their respective uses are as follows: Select: Reselect: Also Select: Unselect: Select All: Select None: Invert:
To select a subset of active items from the full set. To select again from the currently selected subset. To add a different subset to the current subset. To deactivate a portion of the current subset. To restore the full set. To deactivate the full set (opposite of the Select All command). To switch between the active and inactive portions of the set.
Entity: Nodes, Elements, Volumes, Areas, etc. Criterion: By Location, Attached To, etc. Location Settings:
Select Reselect Also Select Unselect Invert
Figure 8.30 The Select Entities dialog box.
M08_MOAV4303_04_GE_C08.INDD 432
27/11/14 9:59 AM
Section 8.8 Graphics Capabilities 433
The select dialog box can be used to select or unselect entities of your solid or finite element model. You can make selections based on the location of your entities in space, or you can select entities that are attached to other selected entities, such as nodes that are attached to selected elements. Be aware, however, that you must reactivate all entities before solving your model. Unselected entities will not be included in a solution. For example, if you select a subset of nodes on which to apply constraints, you should reactivate all nodes before solving. ANSYS allows the user to activate all entities with one simple operation by the command utility menu: Select S Everything You can also select a set of related entities in a hierarchical fashion. For example, given a subset of areas, you can select (a) all lines defining the areas, (b) all keypoints defining those lines, (c) all elements belonging to the areas, and so on. To select in this fashion, use the command utility menu: Select S Everything Below ANSYS also provides the capability to group some selected entities into a component. You can group one type of entity—such as nodes, elements, keypoints, or lines— into a component to be identified by a userdefined name (up to eight characters long). 8.8 Graphics Capabilities Good graphics are especially important for visualizing and understanding a problem being analyzed. The ANSYS program provides numerous features that allow you to enhance the visual information presented to you. Some examples of the graphics capabilities of ANSYS include deformed shapes, result contours, sectional views, and animation. Consult the ANSYS procedure manual for additional information about more than 100 different graphics functions available to the user. Up to five ANSYS windows can be opened simultaneously within one graphics window. You can display different information in different windows. ANSYS windows are defined in screen coordinates (1 to +1 in the xdirection and 1 to +1 in the ydirection). By default, ANSYS directs all graphics information to one window (window 1). In order to define additional windows, you need to access the windowlayout dialog box, as shown in Figure 8.31. To do so, issue the following commands: utility menu: PlotCtrls S Window Controls S Window Layout . . . There are three important concepts that you need to know with respect to window layout: (1) focus point, (2) distance, and (3) viewpoint. The focus point, with coordinates XF, YF, ZF, is the point on the model that appears at the center of the window. By changing the coordinates of the focus point, you can make a different point on the model appear at the center of the window. Distance determines the magnification of an image. As the distance approaches infinity, the image becomes a point on the screen. As the distance is decreased, the image size increases until the image fills the window. Viewpoint determines the direction from which the object is viewed. A vector is established from the viewpoint to the origin of the display coordinate. The line of sight is parallel to this vector and is directed at the focus point.
M08_MOAV4303_04_GE_C08.INDD 433
27/11/14 9:59 AM
434 Chapter 8 More ANSYS
Figure 8.31 The Window Layout dialog box.
Next, the Pan–Zoom–Rotate dialog box allows you to change viewing directions, zoom in and out, or rotate your model. You can access this dialog box, shown in Figure 8.32 by the following commands: utility menu: PlotCtrls S Pan, Zoom, Rotate . . . The various commands within the Pan–Zoom–Rotate dialog box and their respective functions are: Zoom: Pick the center and the corner of the zoom rectangle. Box Zoom: Pick the two corners of the zoom rectangle. Win Zoom: Same as Box Zoom, except the zoom rectangle has the same proportions as the window. • Zoom out. ● Zoom in. Dynamic Mode:
Allows you to pan, zoom, and rotate the image dynamically.
Pan model in X and Ydirections.
Move the mouse right and left to rotate the model about the Zaxis of the screen. Move the mouse up and down to zoom in and out. Move the mouse right and left to rotate the model about the Yaxis of the screen. Move the mouse up and down to rotate the model about the Xaxis of the screen. Fit: Changes the graphics specifications such that the image fits the window exactly. Reset: Resets the graphics specifications to their default values.
M08_MOAV4303_04_GE_C08.INDD 434
27/11/14 9:59 AM
Section 8.9 ErrorEstimation Procedures 435
Viewing Direction
Zooming Options
Pan/Zoom Buttons
Rotate Buttons (screen coordinates)
Dynamic Pan–Zoom–Rotate
Figure 8.32 The Pan–Zoom–Rotate dialog box.
In Section 8.10, an example problem will demonstrate the basic steps in creating and analyzing a model with ANSYS. 8.9 ErrorEstimation Procedures In the previous chapters we discussed how to use fundamental principles, such as statics equilibrium conditions or the conservation of energy, to check for the validity of results. We have also noted that when economically feasible or practical, the experimental verification of a finite element model is the best way to check for the validity of results. Moreover, it has been pointed out that the element size affects the accuracy of your results. Now, consider how you know whether the element sizes associated with a meshed model are fine enough to produce good results. A simple way to find out is to first model a problem with a certain number of elements and then compare its results to the results of a model that you create with twice as many elements. In other words, double the number of original elements and compare the results of the analysis. If no significant difference between the results of the two meshes is detected, then the
M08_MOAV4303_04_GE_C08.INDD 435
27/11/14 9:59 AM
436 Chapter 8 More ANSYS
meshing is adequate. If substantially different results are obtained, then further mesh refinement might be necessary. The ANSYS program offers errorestimation procedures that calculate the level of solution error due to mesh discretization. Error calculations used by ANSYS are based on discontinuity of stresses (or heat fluxes) along interelemental boundaries. Because neighboring elements share common nodes, the difference between the nodal stresses calculated for each element results in a discontinuous stress solution from element to element. The degree of discontinuity is based on both the mesh discretization and the severity of the stress gradient. Therefore, it is the difference in stresses from element to element that forms the basis for error calculation in ANSYS. Error calculations in ANSYS are presented in three different forms: (1) the elementalenergy error (SERR for structural problems and TERR for thermal problems), which measures the error in each element based on the differences between averaged and unaveraged nodal stress or thermal flux values; (2) the percent error in energy norm (SEPC for structural problems and TEPC for thermal problems), which is a global measure of error energy in the model that is based on the sum of the elementalerror energies; and (3) the nodalcomponent value deviation (SDSG for structural problems and TDSG for thermal problems), which measures the local error quantity for each element and is determined by computing the difference between the averaged and unaveraged values of stress or heat flux components for an element. To display error distributors, use the following commands: main menu: General Postproc S Plot Results S Contour Plot S Element Solu You can select and plot the elementalenergy error to observe the higherror regions where mesh refinement may be necessary. You can also plot SDSG (or TDSG) to identify and quantify the region of maximum discretization errors by using the following command: main menu: General Postproc S Element Table S Define Table The elementalenergy error or the nodalcomponent deviations can be listed as well by using the following command: main menu: General Postproc S Element Table S List Element Table Note that ANSYS stresscontour plots and listings give the upper and the lower error bounds based on SDSG or TDSG calculations. The estimatederror bound of plotted stresses is denoted by SMXB or SMNB labels in the graphicsstatus area. To make the task of mesh evaluation and refinement simpler, ANSYS offers adaptive meshing, which is a process that automatically evaluates meshdiscretization error and performs mesh refinement to reduce the error. The adaptive meshing performed by the ADAPT program of ANSYS will perform the following tasks: (1) it will generate an initial mesh and solve the model; (2) based on error calculations, it will determine if mesh refinement is needed; (3) if mesh refinement is necessary, it will automatically refine the mesh and solve the new model; and (4) it will refine the mesh until a loop limit or an acceptable error limit has been reached. Note that to begin the first run of the
M08_MOAV4303_04_GE_C08.INDD 436
27/11/14 9:59 AM
Section 8.10 An Example Problem 437
adaptivemeshing program, you need to create the initial model by defining the element type, material property, and so on.
8.10 An Example Problem Consider one of the many steel brackets (E = 29 * 106 lb/in2, v = 0.3) used to support bookshelves. The dimensions of the bracket are shown in Figure 8.33. The bracket is loaded uniformly along its top surface, and it is fixed along its left edge. Under the given loading and the constraints, plot the deformed shape; also determine the principal stresses and the von Mises stresses in the bracket. The following steps demonstrate how to solve this problem using ANSYS: Enter the ANSYS program by using the Launcher. Type Bracket (or a file name of your choice) in the Jobname entry field of the dialog box.
12 in
R = 0.25 in
6 in
R = 4 in
R = 0.5 in
2 in
Thickness = 0.125 in
10 lb/in
Figure 8.33 A schematic of the steel bracket in the example problem.
M08_MOAV4303_04_GE_C08.INDD 437
27/11/14 9:59 AM
438 Chapter 8 More ANSYS
Pick Run to start the GUI. Create a title for the problem. This title will appear on ANSYS display windows to provide a simple way of identifying the displays. To create a title, issue the command utility menu: File S Change Title c
Define the element type and material properties: main menu: Preprocessor S Element Type S Add/Edit/Delete
M08_MOAV4303_04_GE_C08.INDD 438
27/11/14 9:59 AM
Section 8.10 An Example Problem 439
M08_MOAV4303_04_GE_C08.INDD 439
27/11/14 9:59 AM
440 Chapter 8 More ANSYS
Assign the thickness of the bracket: main menu: Preprocessor S Real Constants S Add/Edit/Delete
M08_MOAV4303_04_GE_C08.INDD 440
27/11/14 9:59 AM
Section 8.10 An Example Problem 441
Assign the modulus of elasticity and the Poisson’sratio values:
main menu: Preprocessor S Material Props S Material Models S Structural S Linear S Elastic S Isotropic
ANSYS Toolbar: SAVE_DB Set up the graphics area—that is, the work plane, zoom, and so on:
M08_MOAV4303_04_GE_C08.INDD 441
27/11/14 9:59 AM
442 Chapter 8 More ANSYS
utility menu: Workplane S Wp Settings. . .
Toggle on the workplane by the command sequence utility menu: Workplane S Display Working Plane Bring the workplane to view by the command sequence utility menu: PlotCtrls S Pan, Zoom, Rotate . . .
M08_MOAV4303_04_GE_C08.INDD 442
27/11/14 9:59 AM
Section 8.10 An Example Problem 443
Click on the small circle and the arrows until you bring the workplane to view, and then create the geometry:
main menu: Preprocessor S Modeling S Create S Areas S Rectangle S By 2 Corners
a) On the workplane, pick the location of the corners of Areas 1 and 2, as shown in Figure 8.34, and apply:
A1
A2
A3
A4 Figure 8.34 The areas making up the bracket.
M08_MOAV4303_04_GE_C08.INDD 443
27/11/14 9:59 AM
444 Chapter 8 More ANSYS
[At WP = 0,12 in the upper left corner of the workplane, press the left button] [First, expand the rubber band down 2.0 and right 4.0 and, then, press the left button] [WP = 4,12] [Expand the rubber band down 2.0 and right 7.0]
OK
b) Create circle A3 by the commands main menu: Preprocessor S Modeling S Create S Areas S Circle S Solid Circle [WP = 11,11] [Expand the rubber band to a radius of 1.0]
OK
c) Create quartercircle A4 by the command main menu: Preprocessor S Modeling S Create S Areas S Circle S Partial Annulus Type in the following values in the given fields: [WPX = 0] [WPY = 10] [Rad1 = 0] [Theta1 = 0] [Rad2 = 4] [Theta2 = 90] OK d) Before creating the fillet, join the keypoints of Areas 1, 2, and 4 by the commands main menu: Preprocessor S Modeling S Operate S Booleans S Glue S Areas Pick Areas 1, 2, and 4. OK
M08_MOAV4303_04_GE_C08.INDD 444
27/11/14 9:59 AM
Section 8.10 An Example Problem 445
e) Create the fillet by the commands main menu: Preprocessor S Modeling S Create S Lines S Line Fillet [Pick the bottom edge of rectangular Area 2]
[Pick the curved edge of quartercircle Area 4] APPLY
Then, issue the command utility menu: PlotCtrls S Pan, Zoom, Rotate . . . Use the Box Zoom button to zoom about the fillet region, and issue the command utility menu: Plot S Lines f) Create an area for the fillet with the commands main menu: Preprocessor S Modeling S Create S Areas S Arbitrary S By Lines Pick the fillet line and the two intersecting smaller lines. OK g) Add the areas together with the commands
main menu: Preprocessor S Modeling S Operate S Booleans S Add S Areas Click on the Pick All button and issue the command utility menu: PlotCtrls S Pan, Zoom, Rotate . . . Click on the Fit button and then Close.
M08_MOAV4303_04_GE_C08.INDD 445
27/11/14 9:59 AM
446 Chapter 8 More ANSYS
h) Create the area of the small hole, but first change the Snap Incr value in the WP Settings dialog box to 0.25. OK Then issue the commands main menu: Preprocessor S Modeling S Create S Areas S Circle S Solid Circle [WP = 11,11]
Expand the rubber band to a radius of 0.25] OK i) Subtract the area of the small hole with the commands main menu: Preprocessor S Modeling S Operate S Booleans S Subtract S Areas [Pick the bracket area]
[anywhere in the ANSYS graphics area, Apply]
[Pick the small circular area (r = 0.25)]
[anywhere in the ANSYS graphics area, Apply]
OK
Now you can toggle off the workplane grids with the command utility menu: Workplane S Display Working Plane ANSYS Toolbar: SAVE_DB You are now ready to mesh the area of the bracket to create elements and nodes. Issue the commands main menu: Preprocessor S Meshing S Size Cntrls S Manual Size S Global S Size
M08_MOAV4303_04_GE_C08.INDD 446
27/11/14 9:59 AM
Section 8.10 An Example Problem 447
ANSYS Toolbar: SAVE_DB main menu: Preprocessor S Meshing S Mesh S Areas S Free Click on the Pick All button. Apply boundary conditions:
main menu: Solution S Define Loads S Apply S Structural S Displacement S On Keypoints
Pick the three keypoints: (1) upper left corner of Area 1, (2) two inches below the keypoint you just picked (i.e., the upper left corner of Area 4), and (3) the lower left corner of Area 4. OK
M08_MOAV4303_04_GE_C08.INDD 447
27/11/14 9:59 AM
448 Chapter 8 More ANSYS
main menu: Solution S Define Loads S Apply S Structural S Pressure S On Lines Pick the upper two horizontal lines associated with Area 1 and Area 2 (on the upper edge of the bracket). OK
Solve the problem: main menu: Solution S Solve S Current LS OK Close (the solution is done!) window. Close (the/STAT Command) window if it appears. For the postprocessing phase, first plot the deformed shape by using the commands main menu: General Postproc S Plot Results S Deformed Shape
M08_MOAV4303_04_GE_C08.INDD 448
27/11/14 9:59 AM
Section 8.10 An Example Problem 449
Plot the von Mises stresses with the commands main menu: General Postproc S Plot Results S Contour Plot S Nodal Solu
M08_MOAV4303_04_GE_C08.INDD 449
27/11/14 9:59 AM
450 Chapter 8 More ANSYS
M08_MOAV4303_04_GE_C08.INDD 450
27/11/14 9:59 AM
Chapter 8 Summary 451
Repeat the previous step and pick the principal stresses to be plotted. Then, exit ANSYS and save everything: Toolbar: QUIT
Summary At this point you should know 1. the basic organization of the ANSYS program. There are three processors that you will use frequently: (1) the preprocessor (PREP7), (2) the processor (SOLUTION), and (3) the general postprocessor (POST1). 2. the commands the preprocessor (PREP7) contains that you need to use to build a model:
• • • • • •
define element types and options define element real constants define material properties create model geometry define meshing controls mesh the object created
3. the commands the Solution processor (SOLUTION) has that allow you to apply boundary conditions and loads. The solution processor also solves for the nodal solutions and calculates other elemental information. 4. the commands the general postprocessor (POST1) contains that allow you to list and display results of an analysis:
• • • •
M08_MOAV4303_04_GE_C08.INDD 451
read results data from results file read element results data plot results list results
27/11/14 9:59 AM
452 Chapter 8 More ANSYS
5. that ANSYS writes and reads many files during a typical analysis. 6. that ANSYS also offers the user the capability to select information about a portion of the model, such as certain nodes, elements, lines, areas, and volumes, for further processing. 7. that the ANSYS program provides numerous features that allow you to enhance the visual information presented to you. Some examples of the graphics capabilities of ANSYS are deformed shapes, result contours, sectional views, and animation. References ANSYS Manual: Introduction to ANSYS, Vol. I, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Procedures, Vol. I, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Commands, Vol. II, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Elements, Vol. III, Swanson Analysis Systems, Inc.
M08_MOAV4303_04_GE_C08.INDD 452
27/11/14 9:59 AM
C h a p t e r
9
Analysis of TwoDimensional Heat Transfer Problems
The main objective of this chapter is to introduce you to the analysis of twodimensional heat transfer problems. General conduction problems and the treatment of various boundary conditions are discussed here. The main topics of Chapter 9 include the following: 9.1 General Conduction Problems 9.2 Formulation with Rectangular Elements 9.3 Formulation with Triangular Elements 9.4 Axisymmetric Formulation of ThreeDimensional Problems 9.5 Unsteady Heat Transfer 9.6 Conduction Elements Used by ANSYS 9.7 Examples Using ANSYS 9.8 Verification of Results 9.1 General Conduction Problems In this chapter, we are concerned with determining how temperatures may vary with position in a medium as a result of either thermal conditions applied at the boundaries of the medium or heat generation within the medium. We are also interested in determining the heat flux at various points in a system, including its boundaries. Knowledge of temperature and heat flux fields is important in many engineering applications, including, for example, the cooling of electronic equipment, the design of thermalfluid systems, and material and manufacturing processes. Knowledge of temperature distributions is also useful in determining thermal stresses and corresponding deflections in machine and structural elements. There are three modes of heat transfer: conduction, convection, and radiation. Conduction refers to that mode of heat transfer that occurs
M09_MOAV4303_04_GE_C09.INDD 453
27/11/14 10:03 AM
454 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
when there exists a temperature gradient in a medium. The energy is transported from the hightemperature region to the lowtemperature region by molecular activities. Using a twodimensional Cartesian frame of reference, we know that the rate of heat transfer by conduction is given by Fourier’s Law: 0T (9.1) 0X 0T qY = kA (9.2) 0Y
qX = kA
qX and qY are the X and the Ycomponents of the heat transfer rate, k is the thermal 0T 0T conductivity of the medium, A is the crosssectional area of the medium, and 0X and 0Y are the temperature gradients. Fourier’s Law may also be expressed in terms of heat transfer rates per unit area as 0T (9.3) 0X 0T qY> = k (9.4) 0Y
qX> = k
q
q
where qX> = AX and qY> = AY are called heat fluxes in the Xdirection and the Ydirection, respectively. It is important to realize that the direction of the total heat flow is always perpendicular to the isotherms (constant temperature lines or surfaces). This relationship is depicted in Figure 9.1. Convective heat transfer occurs when a fluid in motion comes into contact with a solid surface whose temperature differs from the moving fluid. The overall heat transfer rate between the fluid and the surface is governed by Newton’s Law of Cooling, which is q = hA(Ts  Tf) (9.5)
where h is the heat transfer coefficient, Ts is the solid surface temperature, and Tf represents the temperature of the moving fluid. The value of the heat transfer coefficient for
qY
q = qX + qY
Isotherm
qX
Y
X
Figure 9.1 The heat flux vector is always normal to the isotherms.
M09_MOAV4303_04_GE_C09.INDD 454
27/11/14 10:03 AM
Section 9.1 General Conduction Problems 455
a particular situation is determined from experimental correlations that are available in many books about heat transfer. All matters emit thermal radiation. This rule is true as long as the body in question is at a finite temperature (expressed in Kelvin or Rankine Scale). Simply stated, the amount of energy emitted by a surface is given by the equation q″ = esTs 4 (9.6)
where q″ represents the rate of thermal energy per unit area emitted by the surface, e is the emissivity of the surface 0 6 e 6 1, and s is the Stefan–Boltzman constant (s = 5.67 * 108 W/m2 # K4). It is important to note here that unlike conduction and convection modes, heat transfer by radiation can occur in a vacuum, and because all objects emit thermal radiation, it is the net energy exchange among the bodies that is of interest to us. The three modes of heat transfer are depicted in Figure 9.2. In Chapter 1, it was explained that engineering problems are mathematical models of physical situations. Moreover, many of these mathematical models are differential equations that are derived by applying the fundamental laws and principles of nature to a system or a control volume. In heat transfer problems, these governing equations represent the balance of mass, momentum, and energy for a medium. Chapter 1 stated that when possible, the exact solutions of the governing differential equations should be sought because the exact solutions render the detailed behavior of a system. However, for many practical engineering problems, it is impossible to obtain exact solutions to the governing equations because either the geometry is too complex or the boundary conditions are too complicated. The principle of the conservation of energy plays a significant role in the analysis of heat transfer problems. Consequently, you need to understand this principle fully in Surrounding Temp, Tsur
T1
q" q" Ts , e
T2
Thermal radiation emitted by a surface Conduction through a solid object
q"
Moving fluid at Tf
Solid surface Ts 7Tf
Convective heat transfer from a surface
Figure 9.2 Various modes of heat transfer.
M09_MOAV4303_04_GE_C09.INDD 455
27/11/14 10:03 AM
456 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
order to model a physical problem correctly. The principle of the conservation of energy states the following: The rate at which thermal and/or mechanical energy enters a system through its boundaries, minus the rate at which the energy leaves the system through its boundaries, plus the rate of energy generation within the volume of the system, must equal the rate at which energy is stored within the volume of the system. This statement is represented by Figure 9.3 and the equation # # # # Ein  Eout + Egeneration = Estored (9.7) # # Ein and Eout represent the amount of energy crossing into and out of the surfaces of a # system. The thermal energy generation rate Egeneration represents the rate of the conversion of energy from electrical, chemical, nuclear, or electromagnetic forms to thermal energy within the volume of the system. An example of such conversion is the electric current running through a solid conductor. On the other hand, the energy storage term represents the increase or decrease in the amount of thermal internal energy within the volume of the system due to transient processes. It is important to understand the contribution of each term to the overall energy balance of a system in order to model an actual situation properly. A good understanding of the principle of the conservation of energy will also help verify the results of a model. This chapter focuses on the conduction mode of heat transfer with possible convective or radiative boundary conditions. For now, we will focus on steadystate twodimensional conduction problems. Applying the principle of the conservation of energy to a system represented in a Cartesian coordinate system results in the following heat diffusion equation:
kX
02T 02T # + kY 2 + q = 0 (9.8) 2 0X 0Y . Eout
. . Egenerated, Estored
. Ein
M09_MOAV4303_04_GE_C09.INDD 456
Figure 9.3 The principle of the conservation of energy.
27/11/14 10:03 AM
Section 9.1 General Conduction Problems 457
# The derivation of Eq. (9.8) is shown in Figure 9.4. In Eq. (9.8), q represents the heat generation per unit volume, within a volume having a unit depth. There are several boundary conditions that occur in conduction problems:
1. A situation wherein heat loss or gain through a surface may be neglected. This situation, shown in Figure 9.5, is commonly referred to as an adiabatic surface qY + dY
qX
dY
qX + dX
dX
qY First, we begin by applying the principle of the conservation of energy to a small region (differential volume) in a medium: # # # # Ein  Eout + Egeneration = Estored 0T # qX + qY  (qX + dX + qY + dY) + q dXdY(1) = rc dXdY 0t qX + qY  ¢ qX +
0qX 0X
dX + qY +
0qY 0Y
0T # dY≤ + q dXdY = rc dXdY 0t
Simplifying, we get 
0qX 0X
dX 
0qY 0Y
0T # dY + q dXdY = rc dXdY 0t
Making use of Fourier’s Law, we have qX =  kXA
0T 0T =  kX dY(1) 0X 0X
q Y =  k YA
0T 0T =  kY dX(1) 0Y 0Y

0 0T 0 0T 0T # ¢  kX dY ≤ dX ¢  kY dX ≤dY + q dXdY = rc dXdY 0X 0X 0Y 0Y 0t
r and c are the density and specific heat of the medium, and t represents time. For a steadystate situation, temperature does not change with time, and consequently, the righthand side is zero. After simplifying, we obtain kX
02T 02T # + kY 2 + q = 0 0X2 0Y
Figure 9.4 The derivation of the equation of heat conduction under steadystate conditions.
M09_MOAV4303_04_GE_C09.INDD 457
27/11/14 10:03 AM
458 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Perfectly insulated surface
Y X
Figure 9.5 An adiabatic, or perfectly insulated, surface.
or a perfectly insulated surface. In conduction problems, symmetrical lines also represent adiabatic lines. This type of boundary condition is represented by 0T 2 = 0 (9.9) 0X (X = 0, Y)
2. A situation for which a constant heat flux is applied at a surface. This boundary condition, shown in Figure 9.6, is represented by the equation k
0T 2 = q0> (9.10) 0X X = 0
3. A situation for which cooling or heating is taking place at a surface due to convection processes. This situation, shown in Figure 9.7, is represented by the equation k
q"0
0T 2 = h[T(0, y)  Tf] (9.11) 0X (X = 0, Y)
Y X
Figure 9.6 A constant heat flux applied at a surface.
M09_MOAV4303_04_GE_C09.INDD 458
27/11/14 10:03 AM
Section 9.1 General Conduction Problems 459
Y
Tf , h
X
Figure 9.7 Convention processes causing cooling or heating to take place at a surface.
4. A situation wherein heating or cooling is taking place at a surface due to net radiation exchange with the surroundings. The expression for this condition will depend on the view factors and the emissivity of the surfaces involved. 5. A situation in which conditions 3 and 4 both exist simultaneously. 6. Constant surfacetemperature conditions occur when a fluid in contact with a solid surface experiences phase change, as shown in Figure 9.8. Examples include condensation or evaporation of a fluid at constant pressure. This condition is represented by
T(0, Y) = T0 (9.12)
The modeling of actual situations with these boundary conditions will be discussed and illustrated with examples after we consider finite element formulations of twodimensional heat conduction problems.
T0
Y X
Figure 9.8 Constant surfacetemperature conditions occur due to phase change of a fluid in contact with a solid surface.
M09_MOAV4303_04_GE_C09.INDD 459
27/11/14 10:03 AM
460 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
9.2 Formulation with Rectangular Elements Twodimensional bilinear rectangular elements were covered in detail in Chapter 7. Recall that for problems with straight boundaries, linear rectangular shape functions offer simple means to approximate the spatial variation of a dependent variable, such as temperature. For convenience, the expression for a rectangular element in terms of its nodal temperatures and shape functions is repeated here (also see Figure 9.9). The expression is
T (e) = [Si
Sj
Ti T Sn] d j t (9.13) Tm Tn
Sm
where the shape functions Si, Sj, Sm, and Sn are given by
Si = ¢ 1 
Sj =
y x ≤ ¢ 1  ≤ (9.14) w /
y x ¢1  ≤ w /
Sm =
xy /w
Sn =
y x ¢1  ≤ w /
We now apply the Galerkin approach to the heat diffusion equation, Eq. (9.8) expressed in local coordinates x, y, yielding four residual equations:
Tn T Ti
Tj
Tm
y m
n
w
Y j
i
x
/ X
M09_MOAV4303_04_GE_C09.INDD 460
Figure 9.9 A typical rectangular element.
27/11/14 10:03 AM
Section 9.2 Formulation with Rectangular Elements 461
L A
Si ¢ kx
02T 02T # + k + q ≤ dA (9.15) y 2 2 0x 0y
R(e) = j
L A
Sj ¢ kx
02T 02T # + ky 2 + q ≤ dA 2 0x 0y
R(e) m =
L A
Sm ¢ kx
R(e) n =
L A
Sn ¢ kx
R(e) = i
02T 02T # + ky 2 + q ≤ dA 2 0x 0y
02T 02T # + k + q ≤ dA y 0x2 0y2
Note that from here on, for the sake of simplicity of presentation of conductance and load matrices, we will set the residual of an element equal to zero. However, as we mentioned earlier, it is important to realize that the residuals are set equal to zero after all the elements have been assembled. We can rewrite the four equations given by (9.15) in a compact matrix form as L A
[S]T ¢ kx
02T 02T # + ky 2 + q ≤ dA = 0 (9.16) 2 0x 0y
where the transpose of the shape functions is given by the following matrix: Si S [S]T = d j t (9.17) Sm Sn
Equation (9.16) consists of three main integrals:
02T 02T # T ≤ dA + [S] ¢ k ≤ dA + [S]Tq dA = 0 (9.18) y 2 2 0x 0y L L L A A A # Let C1 = kx, C2 = ky, and C3 = q so that we can later apply the results of the forthcoming derivation to other types of problems with similar forms of governing differential equations. As will be demonstrated later in Chapters 10 and 12, we will use the general findings of this chapter to analyze the torsion of solid members and ideal fluid flow problems. So making respective substitutions, we have [S]T ¢ kx
L A
[S]T ¢ C1
02T 02T T ≤dA + [S] ¢ C ≤dA + [S]TC3 dA = 0 (9.19) 2 2 0x2 0y L L A A
Evaluation of the integrals given by Eq. (9.19) will result in the elemental formulation. We first manipulate the secondorder terms into firstorder terms by using the chain rule in the following manner:
M09_MOAV4303_04_GE_C09.INDD 461
0[S]T 0T 0 0T 02T ¢ [S]T ≤ = [S]T 2 + (9.20) 0x 0x 0x 0x 0x
27/11/14 10:03 AM
462 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Rearranging Eq. (9.20), we have [S]T
0[S]T 0T 02T 0 0T = ¢ [S]T ≤ (9.21) 2 0x 0x 0x 0x 0x
Applying the results given by Eq. (9.21) to the first and the second terms in Eq. (9.19), we obtain
L A
[S]T ¢ C1
0[S]T 0T 02T 0 T 0T ≤ dA = C ¢ [S] ≤ dA C ¢ ≤ dA (9.22) 1 1 0x 0x 0x 0x 0x2 L L A A
L A
[S]T ¢ C2
0[S]T 0T 02T 0 0T ≤ dA = C2 ¢ [S]T ≤ dA C2 ¢ ≤ dA (9.23) 2 0y 0y 0y 0y 0y L L A A
Using Green’s theorem, we can write the terms L A
C1
0 0T ¢ [S]T ≤ dA 0x 0x
L A
C2
0 0T ¢ [S]T ≤ dA 0y 0y
and
in terms of integrals around the element boundary. We will come back to these terms later. For now, let us consider the 
L A
C1 ¢
0[S]T 0T ≤ dA 0x 0x
term in Eq. (9.22). This term can easily be evaluated. Evaluating the derivatives for a rectangular element, we obtain 0T 0 = [S 0x 0x i
Also evaluating
Sj
Sm
Ti Ti Tj 1 T Sn] d t = [( w + y) (w  y) y y] d j t (9.24) Tm /w Tm Tn Tn
0[S]T we have 0x Si w + y 0[S]T 0 Sj 1 w  y = d t = d t (9.25) 0x 0x Sm /w y Sn y
Substituting the results of Eqs. (9.24) and (9.25) into the term 
M09_MOAV4303_04_GE_C09.INDD 462
0[S]T 0T C1 ¢ ≤ dA 0x 0x L A
27/11/14 10:03 AM
Section 9.2 Formulation with Rectangular Elements 463
we have w + y Ti 0[S]T 0T 1 w  y T ≤dA = C1 d t [( w + y) (w  y) y  y] d j t dA  C1 ¢ 2 0x 0x y Tm (/w) L L A A y Tn (9.26) Integrating yields w + y Ti 1 w  y T C1 d t [( w + y) (w  y) y  y] d j t dA 2 y T (/w) L m A y Tn 2 C1w 2 = D 6/ 1 1
2 2 1 1
1 1 2 2
1 Ti 1 T T d j t (9.27) 2 Tm 2 Tn
To get the results of Eq. (9.27), we carry out the integration for each term in the 4 * 4 matrix. For example, by integrating the expression in the first row and first column, we get /
/
w
1 1 w3 w2 ( w + y)2 dy dx = ¢ w 2w +  2w ≤ dx 2 2 3 2 (/w) L0 L0 L0 (/w) /
=
1 w3 1 w 3/ 1w dx = = 2 2 3/ (/w) L0 3 (/w) 3
or the integration of the expression in the first row and third column yields /
/
w
1 1 w2 w3 ( w + y)y dy dx = ¢w + ≤ dx 2 2 2 3 (/w) L0 L0 L0 (/w) /
=
1 w3 1 w 3/ 1w dy = = 2 2 6 6/ (/w) L0 (/w) 6
In the same manner, we can evaluate the term 
L A
C2 ¢
0[S]T 0T ≤ dA 0y 0y
in Eq. (9.23) in the ydirection:
0T 0 = [S 0y 0y i
M09_MOAV4303_04_GE_C09.INDD 463
Sj
Sm
Ti Ti T 1 T Sn] d j t = [( / + x) x x (/  x)] d j t (9.28) Tm /w Tm Tn Tn
27/11/14 10:03 AM
464 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
And evaluating
0[S]T we have 0y Si / + x 0[S] 0 S 1 x = d j t = d t (9.29) 0x 0y Sm /w x Sn /  x T
Substituting the results of Eqs. (9.28) and (9.29) into the term 
L A
C2 ¢
0[S]T 0T ≤ dA 0y 0y
we have 0[S]T 0T 1  C2 ¢ ≤ dA = C2 d 0y 0y (/w)2 L L A A
/ + x Ti x T t [( / + x) x x (/  x)] d j t dA x Tm /  x Tn (9.30)
Evaluation of the integral yields / + x Ti 1 x T C2 d t [( / + x) x x (/  x)] d j t dA = 2 x T (/w) L m A /  x Tn
2 C2 / 1 D 6w 1 2
1 2 2 1
Next, we will evaluate the thermal load term
1 2 2 1 L A
2 Ti 1 T T d j t (9.31) 1 Tm 2 Tn [S]TC3 dA:
Si 1 A C S 1 3 [S]TC3 dA = C3 d j t dA = d t (9.32) S 4 1 L LA m A Sn 1
We now return to the terms L A
M09_MOAV4303_04_GE_C09.INDD 464
C1
0 0T ¢ [S]T ≤ dA 0x 0x
27/11/14 10:03 AM
Section 9.2 Formulation with Rectangular Elements 465
and L A
C2
0 0T ¢ [S]T ≤ dA 0y 0y
As mentioned earlier, we can use Green’s theorem to rewrite these area integrals in terms of line integrals around the element boundary:
L A
C1
0 0T 0T ¢ [S]T ≤ dA = C1[S]T cos u dt (9.33) 0x 0x 0x L t
L A
C2
0 0T 0T ¢ [S]T ≤ dA = C2[S]T sin u dt (9.34) 0y 0y 0y L t
t represents the element boundary, and u measures the angle to the unit normal. A Review of Green’s Theorem Before we proceed with the evaluation of Eqs. (9.33) and (9.34), let’s briefly review Green’s theorem, which is given by the following equation: O
Region
¢
0g 0f ≤dxdy = 0x 0y
L
fdx + gdy (9.35)
Contour
In the relationship given by Eq. (9.35), f (x, y) and g (x, y) are continuous functions and have continuous partial derivatives. Next, using simple area examples, we demonstrate how Green’s theorem is applied. We can establish a relationship between the area of a region bounded by its contour t and a line integral around it by substituting for f = 0 and g = x or by letting f = y and g = 0 in Eq. (9.35). This approach leads to the following relationships:
O
A =
L
dxdy =
Region
xdy (9.36)
Contour,t
or
A =
O
L
dxdy = 
Region
ydx (9.37)
Contour,t
We can also combine Eqs. (9.36) and (9.37), which results in yet another relationship between area and line integrals:
2A = 2
O
Region
M09_MOAV4303_04_GE_C09.INDD 465
dxdy =
L
(xdy  ydx) (9.38)
Contour,t
27/11/14 10:03 AM
466 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Let’s now use the relationship given by Eq. (9.38) to calculate the area of a circle having a radius R. As shown in the accompanying figure, for a circle
R
y
u
x
x2 + y2 = R2 x = R cos u
and
dx = R sin u d u
y = R sin u
and
dy = R cos u d u
Substituting these relationships into Eq. (9.38), we have y
dx
d
dy
c
x
d
LContour, t
(xdy  ydx) =
c
2A =
2p
L0 ((R cos u)(R cos ud u)  (R sin u)( R sin ud u))
and simplifying, we get 2A =
L0
2p
R2(cos2 u + sin2 u)du =
L0
2p
R2 du = 2pR2
or A = pR2 As another example, let’s calculate the area of the rectangle shown in the accompanying diagram.
y
y=b 06x6a
b 3 x=0 06y6b
4
x=a 06y6b 2
1
M09_MOAV4303_04_GE_C09.INDD 466
y=0 06x6a
x a
27/11/14 10:03 AM
Section 9.2 Formulation with Rectangular Elements 467
Using Eq. (9.36), we have
A =
Lt1
xdy +
Lt2
xdy +
Lt3
xdy +
Lt4
xdy (9.39)
Note that along t1, x varies from 0 to a and y is zero (dy = 0), and along t2, x = a and y varies from 0 to b. Along t3, x varies from a to 0 and y is constant (dy = 0), and along t4, x = 0 and y varies from b to 0. Then, substituting for these relationships in Eq. (9.39), we get A = 0 +
L0
b
ady + 0 + 0 = ab
Let us now return to Eqs. (9.33) and (9.34), which contribute to the derivative boundary conditions. To understand what is meant by derivative boundary conditions, consider an element with a convection boundary condition, as shown in Figure 9.10. Neglecting radiation, the application of the conservation of energy in the xdirection to the jm edge requires that the energy that reaches the jm edge through conduction must be equal to the energy being convected away (by the fluid adjacent to the jm edge). So, k
0T = h(T  Tf) (9.40) 0x
Substituting the righthand side of Eq. (9.40) into Eq. (9.33), we get L t
C1[S]T
0T 0T cos u dt = k[S]T cos u dt =  h[S]T(T  Tf) cos u dt (9.41) 0x 0x L L t t
n
m Tf , h
qconduction
i
M09_MOAV4303_04_GE_C09.INDD 467
qconvection
j
Figure 9.10 A rectangular element with a convective boundary condition.
27/11/14 10:03 AM
468 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
The integral given by Eq. (9.41) has two terms:

L t
h[S]T(T  Tf) cos u dt =  h[S]TT cos u dt + h[S]TTf cos u dt (9.42) Lt L t
h[S]TT cos u dt and h[S]TT sin u dt, for convective boundary condiLt Lt tions along different edges of the rectangular element, contribute to conductance matrices:
The terms
[K](e)
2 h/ij 1 = D 6 0 0
[K](e)
0 h/mn 0 = D 6 0 0
1 2 0 0
0 0 0 0 0 0 0 0
0 0 T (9.43) 0 0 0 0 2 1
0 0 T (9.45) 1 2
[K](e)
0 h/jm 0 = D 6 0 0
0 2 1 0
0 1 2 0
0 0 T (9.44) 0 0
[K](e)
2 h/ni 0 = D 6 0 1
0 0 0 0
0 0 0 0
1 0 T (9.46) 0 2
Referring to Figure 9.9, note that in the above matrices, /ij = /mn = / and /jm = /in = w. To get the results of Eqs. (9.43) through (9.46), we need to carry out the integration for each term in the 4 * 4 matrix. T
i
Si S h[S]TTdt = h d j t Lt Lt Sm Sn
S2i SS = h D i j SiSm L t SiSn
M09_MOAV4303_04_GE_C09.INDD 468
[Si
SjSi S2j SjSm SjSn
Sj
Sm
Sm Si Sm Sj S2m Sm Sn
Sn]
Ti T d j t dt Tm Tn
SnSi Ti SnSj T T d j t dt (9.47) SnSm Tm S2n Tn
27/11/14 10:03 AM
Section 9.2 Formulation with Rectangular Elements 469
For example, along the i–j edge, where Sm = 0 and Sn = 0, the contribution of Eq. (9.47) to the conductance matrix becomes S2i SS h[S]T Tdt = h D i j 0 L L t t 0
SjSi S2j 0 0
0 0 0 0
0 Ti 0 T T d j t dt (9.48) 0 Tm 0 Tn
Expressing the shape functions Si and Sj in terms of natural coordinates: Si =
1 (1  j)(1  h) 4
Sj =
1 (1 + j)(1  h) 4
and noting j =
dj =
2x  1 /
2 / dx or dx = dj / 2
and j varies from 1 to 1, Eq. (9.48) then becomes Lt
h[S]TTdt =
2 1 ¢ (1  j)(1  h) ≤ 4
F¢ 1 (1  j)(1  h)(1 + j)(1  h) ≤ 16 2 L 0 1 0 1
h/ij
M09_MOAV4303_04_GE_C09.INDD 469
¢
1 (1  j)(1  h)(1 + j)(1  h) ≤ 16 2 1 ¢ (1 + j)(1  h) ≤ 4 0 0
0
0
0
0V
0 0
0 0
Ti T d j t dj (9.49) Tm Tn
27/11/14 10:03 AM
470 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Making use of the fact that along i–j edge h = 1, and integrating Eq. (9.49), we get
h/ij
1 ¢ (1  j) ≤ 2
1
2
h[S]T Tdt = F 1 2 L ¢ 4 (1  j)(1 + j) ≤ L t 1 0 0 2 h/ij 1 = D 6 0 0
1 2 0 0
0 0 0 0
1 ¢ (1  j)(1 + j) ≤ 4 2 1 ¢ (1 + j) ≤ 2 0 0
0 0 0 0
0
Ti Tj 0 V dj d T t m T 0 n 0
0 Ti 0 T Td j t 0 Tm 0 Tn
where 2
1
1 j3 1 1 2 ¢ (1  j) ≤ dj = Jj +  j2 R = 2 4 3 3 L1 1
and 1
j3 1 1 1 1 ¢ (1  j)(1 + j) ≤ dj = Jj R = 4 4 3 3 L1 1 1
2 1 j3 1 1 2 ¢ (1 + j) ≤ dj = Jj + + j2 R = 4 3 3 L1 2 1
Similarly, we can obtain the results given by Eqs. (9.44) through (9.46). h[S]TTf cos u dt and h[S]TTf sin u dt contribute to the elemental Lt Lt thermal load matrix. Evaluating these integrals along the edges of the rectangular element, we obtain The terms
5 F6(e) =
5 F6(e) =
M09_MOAV4303_04_GE_C09.INDD 470
hTf /ij 2
1 1 d t (9.50) 0 0
hTf /mn 2
0 0 d t (9.52) 1 1
5 F6(e) =
5 F6(e) =
hTf /jm 2
hTf /ni 2
0 1 d t (9.51) 1 0 1 0 d t (9.53) 0 1
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 471
Let us summarize what we have done so far. The conductance matrix for a bilinear rectangular element is given by:
[K](e)
2 kxw 2 = D 6/ 1 1
2 2 1 1
1 1 2 2
1 2 ky / 1 1 T + D 2 6w 1 2 2
1 2 2 1
1 2 2 1
2 1 T 1 2
Note that the elemental conductance matrix is composed of (1) a conduction component in the xdirection; (2) a conduction component in the ydirection; and (3) a possible heat transfer term by convection around the edge of a given element, as given by Eqs. (9.43) to (9.46). The thermal load matrix for an element could have two components: (1) a component due to possible heat generation within a given element, and (2) a component due to possible convection heat transfer along an element’s edge(s), as given by Eqs. (9.50) to (9.53). The contribution of the heat generation to the element’s thermalload matrix is given by
5 F6(e)
1 # qA 1 = d t 4 1 1
It is worth noting that in situations in which constant heatflux boundary conditions occur along the edges of a rectangular element, the elemental load matrix is given by (see Problem 5)
5 F6(e) =
5 F6(e)
1 1 d t 2 0 0
qo>/ij
0 qo>/mn 0 = d t 2 1 1
5 F6(e) =
5 F6(e)
qo>/jm 2
0 1 d t 1 0
1 qo>/ni 0 = d t 2 0 1
The next step involves assembling elemental matrices to form the global matrices and solving the set of equations [K] 5 T6 = 5 F6 to obtain the nodal temperatures. We will demonstrate this step in Example 9.1. For now, let us turn our attention to the derivation of the elemental conductance and load matrices for a triangular element. 9.3 Formulation with Triangular Elements As we discussed in Chapter 7, a major disadvantage associated with using rectangular elements is that they do not conform to curved boundaries. In contrast, triangular elements are better suited to approximate curved boundaries. For the sake of convenience, a triangular element is shown in Figure 9.11.
M09_MOAV4303_04_GE_C09.INDD 471
27/11/14 10:03 AM
472 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems T
Tk Tj
Ti (Xj ,Yj) Y (Xk ,Yk)
(Xi ,Yi) X
Figure 9.11 A triangular element.
Recall that a triangular element is defined by three nodes and that we represent the variation of a dependent variable, such as temperature, over a triangular region using shape functions and the corresponding nodal temperatures by the equation
T
(e)
= [Si
Sj
Ti Sk] c Tj s (9.54) Tk
where the shape functions Si, Sj, and Sk are 1 (a + biX + diY ) 2A i 1 Sj = (a + bjX + djY ) 2A j 1 Sk = (a + bkX + dkY ) 2A k Si =
A is the area of the element and is computed from the equation 2A = Xi(Yj  Yk) + Xj(Yk  Yi) + Xk(Yi  Yj) Also,
ai = XjYk  XkYj bi = Yj  Yk di = Xk  Xj (9.55)
aj = XkYi  XiYk bj = Yk  Yi dj = Xi  Xk
ak = XiYj  XjYi bk = Yi  Yj dk = Xj  Xi
M09_MOAV4303_04_GE_C09.INDD 472
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 473
Employing the Galerkin approach, the three residual equations for a triangular element, in matrix form, are given by LA
[S]T ¢ kX
02T 02T # + kY 2 + q ≤ dA = 0 (9.56) 2 0X 0Y
where Si [S]T = c Sj s Sm We now proceed with steps similar to the ones we followed to formulate the conductance and thermal load matrices for rectangular elements. First, we rewrite the secondderivative expressions in terms of the firstderivative expressions using the chain rule. Evaluating the integral 
L A
C1 ¢
0[S]T 0T ≤ dA 0X 0X
for a triangular element, we obtain Si bi 0[S]T 0 1 = c Sj s = c bj s (9.57) 0X 0X 2A Sk bk
0T 0 = [S 0X 0X i
Ti 1 Sk] c Tj s = [b 2A i Tk
Sj
Ti bk] c Tj s (9.58) Tk
bj
Substituting for the derivatives, we get bi 0[S]T 0T 1  C1 ¢ ≤ dA = C1 c bj s [bi 2 0X 0X L LA 4A b A k
bj
Ti bk]c Tj s dA (9.59) Tk
bi bj b 2j bj bk
bi bk Ti bj bk S c Tj s (9.60) b 2k Tk
and integrating, we are left with
bi 1 C1 c bj s [bi 2 L A 4A bk
bj
Ti b 2i C1 bk] c Tj s dA = C bi bj 4A Tk bi bk
In the same manner, we can evaluate the term 
M09_MOAV4303_04_GE_C09.INDD 473
L A
C2 ¢
0[S]T 0T ≤ dA 0Y 0Y
27/11/14 10:03 AM
474 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
as Si di 0[S]T 0 1 = c Sj s = c dj s 0Y 0Y 2A Sk dk
0T 0 = [S 0Y 0Y i
Sj
(9.61)
Ti 1 Sk]c Tj s = [d 2A i Tk
Ti dk] c Tj s (9.62) Tk
dj
Substituting for the derivatives and integrating, we have
di 1 c dj s [di C2 2 L A 4A dk
Ti d 2i C2 dk] c Tj s dA = C di dj 4A Tk di dk
dj
di dj d 2j dj dk
di dk Ti dj dk S c Tj s (9.63) d 2k Tk
For a triangular element, the thermal load matrix due to the heat generation term C3 is LA
[S]TC3 dA = C3
Si 1 C3 A c Sj s dA = c 1 s (9.64) 3 LA S 1 k
Lt Lt condition along the edges of the triangular element results in the equations Evaluating the terms
2 C1 6 0
1 2 0
0 0 S (9.65) 0
2 h/ki = C0 6 1
0 0 0
1 0 S (9.67) 2
[K](e) =
(e)
[K]
h[S]TT cos u dt and
h/ij
h[S]TT sin u dt for a convective boundary
[K](e) =
0 C0 6 0
0 2 1
h/jk
0 1 S (9.66) 2
Note that in the above matrices, /ij, /jk, and /ki represent the respective lengths of the h[S]TTf cos u dt and h[S]TTf sin u dt Lt Lt contribute to the elemental thermal loads. Evaluating these integrals along the edges of the triangular element yields three sides of the triangular element. The terms
5 F6(e) = (e)
5 F6
M09_MOAV4303_04_GE_C09.INDD 474
=
hTf /ij 2 hTf /ki 2
1 c 1 s (9.68) 0 1 c 0 s (9.70) 1
5 F6(e) =
hTf /jk 2
0 c 1 s (9.69) 1
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 475
Let us summarize the triangular formulation. The conductance matrix for a triangular element is (e)
[K]
b 2i kX = C bi bj 4A bi bk
bi bk d 2i kY bj bk S + C di dj 4A 2 bk di dk
bi bj b 2j bj bk
di dj d 2j dj dk
di dk dj dk S d 2k
Note once again that the elemental conductance matrix for a triangular element is composed of (1) a conduction component in the Xdirection; (2) a conduction component in the Ydirection; and (3) a possible heat loss term by convection from the edge(s) of a given element, as given by Eq. (9.65) to (9.67). The thermal load matrix for a triangular element could have two components: (1) a component resulting from a possible heatgeneration term within a given element, and (2) a component due to possible convection heat loss from the element’s edge(s), as given by Eq. (9.68) to (9.70). The contribution of the heat generation to the element’s load matrix is (e)
5 F6
1 # qA = c1s 3 1
The development of constant heat flux boundary conditions for triangular elements is left as an exercise. (See Problem 6.) Next, we use an example to demonstrate how to assemble the elemental information to obtain the global conductance matrix and the global load matrix. Example 9.1 Consider a small industrial chimney constructed from concrete with a thermal conductivity value of k = 1.4 W/m # K, as shown in Figure 9.12. The inside surface temperature of the chimney is assumed to be uniform at 100°C. The exterior surface is exposed to the surrounding air, which is at 30°C, with a corresponding natural convection heat transfer coefficient of h = 20 W/m2 # K. Determine the temperature distribution within the concrete under steadystate conditions. We can make use of the symmetry of the problem, as shown in Figure 9.12, and only analyze a section of chimney containing 1/8 of the area. The selected section of the chimney is divided into nine nodes with five elements. Elements (1), (2), and (3) are squares, while elements (4) and (5) are triangular elements. Consult Table 9.1 while following the solution. Table 9.1 The relationship between the elements and their corresponding nodes
M09_MOAV4303_04_GE_C09.INDD 475
Element
i
j
m or k
n
(1) (2) (3) (4) (5)
1 3 4 2 5
2 4 5 5 9
4 7 8 4 8
3 6 7
27/11/14 10:03 AM
476 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
T = 1005C
Tsurr = 305C 20 cm
h = 20W/m2 · K
60 cm
Y 7
6 (2) 3
8 (5)
(3) 4
5 (4)
(1) 1
9
X
2
Figure 9.12 A schematic of the chimney in Example 9.1.
The conductance matrix due to conduction in a rectangular element is given by
[K](e)
2 kw 2 = D 6/ 1 1
2 2 1 1
1 1 2 2
1 2 1 k/ 1 T + D 2 6w 1 2 2
1 2 2 1
1 2 2 1
2 1 T 1 2
Elements (1), (2), and (3) all have the same dimensions; therefore,
[K](1) = [K](2) = [K](3)
M09_MOAV4303_04_GE_C09.INDD 476
2 (1.4)(0.1) 2 = D 6(0.1) 1 1 2 (1.4)(0.1) 1 + D 6(0.1) 1 2
2 2 1 1 1 2 2 1
1 1 2 2 1 2 2 1
1 1 T 2 2 2 1 T 1 2
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 477
To help with assembly of the elements later, the corresponding node numbers for each element are shown on the top and the side of each matrix: 1(i) 2( j) 4(m) 3(n) 0.933 0.233 0.466 0.233 1 0.233 0.933 0.233 0.466 2 [K](1) = D T 0.466 0.233 0.933 0.233 4 0.233 0.466 0.233 0.933 3
[K](2)
[K](3)
3(i) 0.933 0.233 = D 0.466 0.233
4( j) 0.233 0.933 0.233 0.466
7(m) 0.466 0.233 0.933 0.233
6(n) 0.233 3 0.466 4 T 0.233 7 0.933 6
4(i) 0.933 0.233 = D 0.466 0.233
5( j) 0.233 0.933 0.233 0.466
8(m) 0.466 0.233 0.933 0.233
7(n) 0.233 4 0.466 5 T 0.233 8 0.933 7
For triangular elements (4) and (5), the conductance matrix is (e)
[K]
b 2i k = C bi bj 4A bi bk
bi bj b 2j bj bk
bi bk d 2i k bj bk S + C di dj 4A 2 bk di dk
di dj d 2j dj dk
di dk dj dk S d 2k
where the [email protected] and [email protected] are given by the relations of Eq. (9.55). Because the [email protected] and [email protected] are calculated from the difference of the coordinates of the involved nodes, it does not matter where we place the origin of the coordinate system X, Y. Evaluating the coefficients for element (4), we have bi = Yj  Yk = 0.1  0.1 = 0
di = Xk  Xj = 0  0.1 = 0.1
bj = Yk  Yi = 0.1  0 = 0.1
dj = Xi  Xk = 0  0 = 0
bk = Yi  Yj = 0  0.1 = 0.1
dk = Xj  Xi = 0.1  0 = 0.1
Evaluating the coefficients for element (5) renders the same results because the difference between the coordinates of its nodes is identical to that of element (4). Therefore, elements (4) and (5) will both have the following conductance matrix: (4)
[K]
(5)
= [K]
0 1.4 = C0 4(0.005) 0
0 (0.1)2 (0.1)( 0.1)
0 (0.1)( 0.1) S ( 0.1)2 1.4 + C 4(0.005)
M09_MOAV4303_04_GE_C09.INDD 477
( 0.1)2 0 ( 0.1)(0.1)
0 0 0
( 0.1)(0.1) 0 S 2 (0.1)
27/11/14 10:03 AM
478 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Showing the corresponding node numbers on the top and the side of each respective conductance matrix for elements (4) and (5), we obtain 2(i) 0.7 = C 0 0.7
5( j) 0 0.7 0.7
4(k) 0.7 2 0.7 S 5 1.4 4
5(i) 0.7 = C 0 0.7
9( j) 0 0.7 0.7
8(k) 0.7 5 0.7 S 9 1.4 8
[K](4)
[K](5)
As explained earlier, the convective boundary condition contributes to both the conductance matrix and the load matrix. The convective boundary condition contributes to the conductance matrices of elements (2) and (3) according to the relationship
[K](e)
0 h/mn 0 = D 6 0 0
[K](2) = [K](3)
0 0 0 0
0 0 2 1
0 i 0 j T 1 m 2 n
0 (20)(0.1) 0 = D 6 0 0
0 0 0 0
0 0 2 1
0 0 0 0 T = D 1 0 2 0
0 0 0 0
0 0 0.666 0.333
0 0 T 0.333 0.666
Including the nodal information, the conductance matrices for elements (2) and (3) are
M09_MOAV4303_04_GE_C09.INDD 478
[K](2)
3 0 = 0 D 0 0
4 0 0 0 0
7 0 0 0.666 0.333
6 0 3 0 4 T 0.333 7 0.666 6
[K](3)
4 0 = 0 D 0 0
5 0 0 0 0
8 0 0 0.666 0.333
7 0 4 0 5 T 0.333 8 0.666 7
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 479
Heat loss by convection also occurs along jk edge of element (5); thus,
(e)
[K]
(e)
[K]
[K](5)
=
0 C0 6 0
h/jk
0 2 1
0 (20)(0.1) = C0 6 0 5 0 = C0 0
9 0 0.666 0.333
0 i 1S j 2 k 0 2 1
0 0 1S = C0 2 0
0 0.666 0.333
0 0.333 S 0.666
8 0 5 0.333 S 9 0.666 8
The convective boundary condition contributes to the thermal load matrices for elements (2) and (3) along their mn edge according to the relationship
5 F6(e) =
hTf /mn 2
0 0 0 (20)(30)(0.1) 0 0 0 d t = d t = d t 1 2 1 30 1 1 30
Including the nodal information, we have
5 F6(2)
0 3 0 4 = d t 30 7 30 6
5 F6(3)
0 4 0 5 = d t 30 8 30 7
The convective boundary condition contributes to the load matrix for element (5) along its jk edge according to the matrix 5 F6(e) =
M09_MOAV4303_04_GE_C09.INDD 479
hTf /jk 2
0 0 0 (20)(30)(0.1) c1s = c 1 s = c 30 s 2 1 1 30
27/11/14 10:03 AM
480 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Again, including the nodal information, we have 0 5 5 F6(5) = c 30 s 9 30 8 Next, we need to assemble all of the elemental matrices. Using the nodal information presented next to each element, the global conductance matrix becomes 1 2 3 4 5 6 7 8 9
[K](G)
0.933  0.233  0.233  0.233 1.633  0.466  0.233  0.466 1.866  0.466  0.933  0.466 = I 0 0 0 0 0  0.233 0 0  0.466 0 0 0 0 0 0
 0.466 0 0 0 0 0  0.933 0 0 0 0 0  0.466 0  0.233  0.466 0 0 4.199  0.933  0.466  0.466  0.466 0  0.933 2.333 0  0.466  0.933 0 Y  0.466 0 1.599 0.1 0 0  0.466  0.466 0.1 3.198 0.1 0  0.466  0.933 0 0.1 3.665  0.367 0 0 0 0  0.367 1.366
1 2 3 4 5 6 7 8 9
Applying the constant temperature boundary condition at nodes 1 and 2 results in the global matrix
[K](G)
1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0  0.233  0.466 1.866  0.466 0  0.233  0.466 0 0  0.466  0.933  0.466 4.199  0.933  0.466  0.466  0.466 0 = I 0 0 0  0.933 2.333 0  0.466  0.933 0 Y 0 0  0.233  0.466 0 1.599 0.1 0 0 0 0  0.466  0.466  0.466 0.1 3.198 0.1 0 0 0 0  0.466  0.933 0 0.1 3.665  0.367 0 0 0 0 0 0 0  0.367 1.366 Assembling the thermal load matrix, we have
5 F6(G)
M09_MOAV4303_04_GE_C09.INDD 480
0 0 0 0 = i 0 y 30 30 + 30 30 + 30 30
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 481
and applying the constant temperature boundary condition at nodes 1 and 2 leads to the following final form of the thermal load matrix:
5 F6(G)
100 100 0 0 = i 0 y 30 60 60 30
The final set of nodal equations is given by 1 0 0.233 0.466 I 0 0 0 0 0
0 1 0.466 0.933 0 0 0 0 0
0 0 1.866 0.466 0 0.233 0.466 0 0
0 0 0.466 4.199 0.933 0.466 0.466 0.466 0
0 0 0 0.933 2.333 0 0.466 0.933 0
0 0 0.233 0.466 0 1.599 0.1 0 0
0 0 0.466 0.466 0.466 0.1 3.198 0.1 0
0 0 0 0.466 0.933 0 0.1 3.665 0.367
0 0 0 0 0 Y 0 0 0.367 1.366
T1 100 T2 100 T3 0 T4 0 * i T5 y = i 0 y T6 30 T7 60 T8 60 T9 30 Solving the set of linear equations simultaneously leads to the following nodal solution: [T]T = [100
100
70.83
67.02
51.56
45.88
43.67
40.10
32.73]°C
To check for the accuracy of the results, first note that nodal temperatures are within the imposed boundary temperatures. Moreover, all temperatures at the outer edge are slightly above 30°C, with node 9 having the smallest value. This condition makes physical sense because node 9 is the outermost cornerpoint. As another check on the validity of the results, we can make sure that the conservation of energy, as applied to a control volume surrounding an arbitrary node, is satisfied. Are the energies flowing
M09_MOAV4303_04_GE_C09.INDD 481
27/11/14 10:03 AM
482 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems T6 = 45.88 6
3
4
T4 = 67.02
Figure 9.13 Applying the principle of energy balance to node 3 to check the validity of our results.
1 T1 = 100
into and out of a node balanced out? As an example, let us consider node 3. Figure 9.13 shows the control volume surrounding node 3 used to apply the conservation of energy principle. We start with the equation aq = 0 and using Fourier’s law we have k(0.1) ¢
67.02  T3 45.88  T3 100  T3 ≤ + k(0.05) ¢ ≤ + k(0.05) ¢ ≤ = 0 0.1 0.1 0.1
Solving for T3, we find that T3 = 69.98°C. This value is reasonably close to the value of 70.83°C, particularly considering the coarseness of the element sizes. We will discuss the verification of results further with another example problem solved using ANSYS. Example 9.1 Revisited We will now show how to use Excel to set up and solve Example 9.1. 1. In cell A1 type Example 9.1, and in cells A3, A4, and A5 type k =, h =, and Tsurr = as shown. After inputting the value of k in cell B3, select B3 and in the “Name Box” type k and hit the Return key. Similarly, after inputting the values of h and Tsurr in cells B4 and B5, select B4 and B5, and in the corresponding “Name Box” type h and Tsurr, respectively, and hit the Return key after each entry.
M09_MOAV4303_04_GE_C09.INDD 482
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 483
2. Create the table shown with element and node numbers and the x, ycoordinates of the nodes. Also, create the width and the length of the rectangular elements. Select cell J9 and name it W. Similarly, select cell J10 and name it L. In cell J11 type = W*L and name this cell, Arectangle.
3. Complete the table by inputting the element, node numbers, and the coordinates of the triangular elements. In cell F25, type = [email protected]; in cell F26 type = [email protected]; in cell F27 type = [email protected]; and name them deltai, deltaj, and deltak, respectively. Similarly, in cells G25 through G27 create betai, betaj, and betak. Note the contents of G25, G26, and G27 are [email protected]; [email protected]; [email protected] Input the magnitude of the triangular element in cell J26 and name it Atriangle. In a similar way, input and name the information for element 5.
4. Compute the [K] matrices due to conduction for rectangular elements, [K1], [K2], and [K3] as shown and name them Kelement1. Note, they are equal that is [K1] = [K2] = [K3].
M09_MOAV4303_04_GE_C09.INDD 483
27/11/14 10:03 AM
484 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
5. Next, compute the [K] matrices due to conduction for triangular elements, that is, [K4] and [K5] as shown and name them Kelement4. Note, that [K4] = [K5].
6. Create the [K] matrices due to convection. Select the range I49:L52 and type = (h*L , 6)*D49:G52 and while holding down the Ctrl and Shift keys hit the Return key. Name the range I49:L52, K2duetoconvection. In a similar way, compute the [K5] matrix due to convection. Select the range G54:I56 and type = (h*L , 6)*C54:E56 and while holding down the Ctrl and Shift keys, hit the Return key. Name the range G54:I56, K5duetoconvection.
7. Create the {F2}, {F3}, and {F5} matrices as shown. Name the ranges E59:E62 and K59:K61, Felement2 and Felement5, respectively.
M09_MOAV4303_04_GE_C09.INDD 484
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 485
8. Next, create [A1], [A2], [A3], [A4], and [A5] matrices and name them Aelement1, Aelement2, Aelment3, Aelement4, Aelement5, respectively. If you have forgotten what each [A] matrix represents, see Section 2.5, Eq. (2.9).
9. Now, create the stiffness matrix for each element (with their proper positions in the global matrix) and name them K1G, K2G, K3G, K4G, and K5G. For example, to create [K]1G, select B87:J95 and type =MMULT(TRANSPOSE(Aelement1),MMULT(Kelement1,Aelement1)) and while holding down the Ctrl and Shift keys, hit the Return key. In a similar way, create [K]2G, [K]3G, [K]4G, and [K]5G as shown. Make sure to include the K due to convection to appropriate [K]G matrices. For example, as shown, for [K]5G, in the range B127:J135, you need to type =MMULT(TRANSPOSE(Aelement5),MMULT(Kelement5+K5duetoconvect ion,Aelement5))
M09_MOAV4303_04_GE_C09.INDD 485
27/11/14 10:03 AM
486 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
10. Next, create the load matrices and name them F2G, F3G, and F5G. For example, to create 5 F62G, in the range B137:B145 type =MMULT(TRANSPOSE(Aelement2),Felement2) and while holding down the Ctrl and Shift keys, hit the Return key.
M09_MOAV4303_04_GE_C09.INDD 486
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 487
11. The final global matrix is created next. Select the range B147:J155 and type =K1G+K2G+K3G+K4G+K5G and while holding down the Ctrl and Shift keys, hit the Return key.
12. Create the global load matrix as shown.
13. Apply the boundary conditions. Copy the appropriate portion of the KG matrix and paste it in the range C167:K175 as values only. Name the range KwithappliedBC. Similarly, create the corresponding load matrix in the range C177:C185 and name it FwithappliedBC.
14. Select the range C187:C195 and type =MMULT(MINVERSE(KwithappliedBC),FwithappliedBC) and while holding down the Ctrl and Shift keys hit the Return key.
M09_MOAV4303_04_GE_C09.INDD 487
27/11/14 10:03 AM
488 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
The complete Excel sheet is shown next.
M09_MOAV4303_04_GE_C09.INDD 488
27/11/14 10:03 AM
Section 9.3 Formulation with Triangular Elements 489
M09_MOAV4303_04_GE_C09.INDD 489
27/11/14 10:04 AM
490 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
9.4 Axisymmetric Formulation of ThreeDimensional Problems As we explained in Chapter 7, Section 7.5, there is a special class of threedimensional problems whose geometry and loading are symmetrical about an axis, such as a zaxis, as shown in Figure 7.12. These threedimensional problems may be analyzed using twodimensional axisymmetric elements. We discussed the formulation of axisymmetric elements in Section 7.5; in this section, we discuss the finite element formulation of axisymmetric conduction problems using triangular axisymmetric elements.
M09_MOAV4303_04_GE_C09.INDD 490
27/11/14 10:04 AM
Section 9.4 Axisymmetric Formulation of ThreeDimensional Problems 491
We begin our finite element formulation by applying the principle of the conservation of energy to a differential volume represented in a cylindrical coordinate system. The resulting heat conduction equation is given by
0T 1 0 0T 0 0T 0T 1 0 # ¢ k r ≤ + 2 ¢ ku ≤ + ¢k ≤ + q = rc (9.71) r 0r r 0r 0u 0z z 0z 0t r 0u # In Eq. (9.71), kr, ku, kz represent thermal conductivities in r, [email protected], and zdirections, q is the heat generation per unit volume, and c represents the specific heat of the material. The derivation of Eq. (9.71) is shown in Figure 9.14. For steadystate problems, the righthand side of Eq. (9.71) vanishes. Moreover, for axisymmetric situations, there is no variation of temperature in the [email protected], and assuming thermal conductivity is constant (kr = ku = kz = k), Eq. (9.71) reduces to k 0 0T 02T # ¢ r ≤ + k 2 + q = 0 (9.72) r 0r 0r 0z
The Galerkin residuals for an arbitrary triangular element become 5 R(e) 6 =
where
L V
[S]T ¢
k 0 0T 02T # ¢ r ≤ + k 2 + q ≤ dV (9.73) r 0r 0r 0z
Si = Si [S]T = c Sj s Sk
and Sj =
Sk =
1 (a + bir + diz) 2A i 1 (a + bjr + djz) and 2A j 1 (a + bkr + dkz) 2A k
ai = RjZk  RkZj
bi = Zj  Zk
di = Rk  Rj
aj = RkZi  RiZk
bj = Zk  Zi
dj = Ri  Rk
ak = RiZj  RjZi
bk = Zi  Zj
dk = Rj  Ri
As marked below, Eq. (9.73) has three main parts
5 R(e) 6 =
LV
[S]T ¢
Part three
d
Part two
f
h
Part one
k 0 0T 02T # ¢ r ≤ ≤dV + [S]T ¢ k 2 ≤dV + [S]Tq dV (9.74) r 0r 0r 0z LV LV
Using the chain rule, we can rearrange the terms in part one in the following manner:
M09_MOAV4303_04_GE_C09.INDD 491
k 0 0T k 0[S]T 0T k 0 0T ¢ [S]Tr ≤ = r + [S]T ¢ r ≤ (9.75) r 0r r 0r 0r r 0r 0r 0r
27/11/14 10:04 AM
492 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Z r Y u
X
qz + dz rdu
qr
qu + du dz
qu
qr + dr
dr qz
Applying the conservation of energy to a differential cylindrical segment, # # # # Ein  Eout + Egenerated = Estored The r, [email protected], and zcomponents of Fourier’s law in cylindrical coordinates are Ar
0T 0T =  krrdudz 0r 0r e
qr =  krAr
e
Au
ku
ku drdz 0T 0T qu =  Au = r 0u r 0u Ar
0T 0T =  kz rdudr 0z 0z e
qz =  kzAz
0qu
0r
e
0qr
qr +
qu + du
qz + dz
e
(qr + qu + qz) 
0qr
e
e
qr + dr
•
dV
e
(qr + qu + qz)  (qr + dr + qu + du
dV
0T # + qz + dz) + q drdzrdu = rc drdzrdu 0t
dr + qu +
0qu 0u
du + qz +
0qz 0z
0qz
µ
dz + q# drdzrdu = rc drdzrdu 0T 0t
0T dr du dz + q•drdzrdu = rc drdzrdu 0r 0u 0z 0t Substituting for qr, qu, and qz in the above equation 

kudrdz 0T 0 0T 0 0 0T 0T ¢  krrdudz ≤dr ¢≤du ¢  kzrdudr ≤dz + q•drdzrdu = rc drdzrdu 0r 0r 0u r 0u 0z 0z 0t
and simplifying the differential volume terms, we have 1 0 0T 1 0 0T 0 0T 0T # ¢ k r ≤ + 2 ¢ ku ≤ + ¢k ≤ + q = rc r 0r r 0r 0u 0z z 0z 0t r 0u Figure 9.14 The derivation of equation of heat conduction in cylindrical coordinates under steadystate conditions.
M09_MOAV4303_04_GE_C09.INDD 492
27/11/14 10:04 AM
Section 9.4 Axisymmetric Formulation of ThreeDimensional Problems 493
Now, using the relationship given by Eq. (9.75), the part one in Eq. (9.74) may be expressed as [S]T ¢
0[S]T 0T k 0 0T k 0 0T ¢r ≤ ≤ = ¢ [S]Tr ≤  k (9.76) r 0r 0r r 0r 0r 0r 0r
Similarly, using the chain rule and rearranging terms in part two of Eq. (9.74), we have
k
0[S]T 0T 0 0T 02T ¢ [S]T ≤ = k[S]T 2 + k (9.77) 0z 0z 0z 0z 0z
and k[S]T
0[S]T 0T 02T 0 0T = k ¢ [S]T ≤  k (9.78) 2 0z 0z 0z 0z 0z
Applying the results given by Eqs. (9.76) and (9.78), the contributions of part one and part two to the residual matrix is g
Part one
L V
[S]T ¢
0[S]T 0T k 0 0T k 0 0T ¢ r ≤ ≤dV = ¢ [S]Tr ≤dV k dV (9.79) r 0r 0r r 0r 0r 0r 0r L L V V g
Part two
L V
[S]T ¢ k
0[S]T 0T 02T 0 T 0T ≤dV = ¢ [S] ≤dV dV 0z 0z 0z 0z2 L L V 0z V
(9.80)
Similar to what we did in Section 9.2, using Green’s theorem, we can rewrite the volume k 0 0T 0 0T ¢ [S]Tr ≤dV and ¢ [S]T ≤dV in terms of area integrals around integrals, r 0r 0r 0z 0z L L V V the element surface boundaries. We will come back to these terms later; for now let us 0[S]T 0T consider the  k dV term in Eq. (9.79). This term can easily be evaluated. 0r 0r L V Evaluating the derivative for a triangular element, we obtain 0T 0 = [Si 0r 0r
Also, evaluating
Sj
Ti 1 Sk]c Tj s = [b 2A i Tk
bj
Ti bk]c Tj s (9.81) Tk
0[S]T , we have 0r Si bi 0[S]T 0 1 = c Sj s = c bj s (9.82) 0r 0r 2A Sk bk
Substituting the results of Eqs. (9.81) and (9.82) into the term integrating, we have
M09_MOAV4303_04_GE_C09.INDD 493
0[S]T 0T dV and 0r 0r L V k
27/11/14 10:04 AM
494 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
bi b2i 0[S]T 0T 1 k  k dV = k c bj s [bi bj bk]dV = C bibj 2 0r 0r 4A2 L L V V 4A b bb k
i k
bibj b2j bjbk
bibk bjbk S V(9.83) b2k
Next, we use the Pappus–Guldinus theorem to compute the volume of the element that is generated by revolving the area element about the zaxis. You may recall from your Statics class that the Pappus–Guldinus theorem is used to calculate the volume of a body that is generated by revolving a sectional area about an axis similar to the one shown in Figure 9.15. V = 2p rA (9.84)
In Equation (9.84), 2p r represents the distance traveled by the centroid of the sectional area. See Example 9.2 to refresh your memory about how the Pappus–Guldinus theorem
z A=
pR2 2
r=
c
r
4R 3p
V = 2p r A =
4 p R3 3
R
(a) Generating volume of a sphere 1 bh 2 b r = 3
A=
h
r
V = 2p r A = c
p b2h 3
b
(b) Generating volume of a cone
M09_MOAV4303_04_GE_C09.INDD 494
Figure 9.15 The Pappus–Guldinus theorem—A volume generated by revolving an area about a fixed axis.
27/11/14 10:04 AM
Section 9.4 Axisymmetric Formulation of ThreeDimensional Problems 495
is used to calculate a volume of an object. Substituting in Eq. (9.83) and simplifying, we get b2i 0[S]T 0T prk k dV = C bibj 0r 0r 2A L V bibk
bibj b2j bjbk
bibk bjbk S (9.85) b2k
As we discussed in Chapter 7, we can use shape functions to describe the position of any point within an element. Therefore, the radial position of the centroid of the element for which Si = Sj = Sk = 1/3 can be expressed by r = SiRi + SjRj + SkRk =
Ri + Rj + Rk 3
(9.86)
Equation (9.86) can be used to calculate r. In a manner shown previously, we can also 0[S]T 0T evaluate the term dV L V 0z 0z 0T 0 = [Si 0z 0z
Ti 1 Sk]c Tj s = [d 2A i Tk
Sj
dj
Ti dk]c Tj s (9.87) Tk
Si di 0[S]T 0 1 = c Sj s = c dj s (9.88) 0z 0z 2A Sk dk
Substituting for the derivatives and integrating, we have di 0[S]T 0T 1 k dV = k c dj s [di 2 0z 0z L LV 4A d V
d 2i k dk]dV = C didj 4A2 didk
dj
k
didj d 2j djdk
didk djdk S V (9.89) d 2k
And using the Pappus–Guldinus theorem, substituting for V = 2p rA in Eq. (9.89), we have d 2i 0[S]T 0T k k dV = C didj 0z 0z 4A2 L V dd
i k
didj d 2j djdk
didk d 2i p rk djdk S 2p rA = C didj 2A 2 dk didk
didj d 2j djdk
didk djdk S (9.90) d 2k
The thermal load due to the heat generation term is evaluated by first substituting for Ri + Rj + Rk dV = 2p rdA and r = , then integrating the resulting expression 3 Si Si # [S] q dV = q [S] dV = q c Sj s dV = q 2p r c Sj s dA (9.91) L L L L V V V V Sk Sk T
#
#
T
#
and substituting for Si, Sj, and Sk and integrating, we get
M09_MOAV4303_04_GE_C09.INDD 495
27/11/14 10:04 AM
496 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
L V
# # [S]Tq dV = q 2p r
We now return to the terms
# Si 2 2pq A c Sj s dA = C1 12 L V Sk 1
1 2 1
1 Ri 1 S c Rj s (9.92) 2 Rk
k 0 0T 0 0T ¢ [S]Tr ≤dV and ¢ [S]T ≤dV. r 0r 0r 0z 0z L LV V
As mentioned earlier, we can use Green’s theorem to rewrite these volume integrals in terms of area integrals around the boundaries similar to the approach we discussed in Section 9.2. For convective boundary conditions along the areas generated by revolving the edges of the triangular element, the above terms contribute to both the conductance matrix and the thermal load matrix in the following manner: Along i9j, [K](e) =
3Ri + Rj C Ri + Rj 12 0
2p/ij
Ri + Rj Ri + 3Rj 0
0 0 S (9.93) 0
and 5f6
(e)
Along j9k, [K]
=
(e)
0 C0 12 0
2p/jk
=
2phTf /ij 6
0 3Rj + Rk Rj + Rk
2Ri + Rj c Ri + 2Rj s (9.94) 0 0 Rj + Rk S (9.95) Rj + 3Rk
and 5 f6 (e) =
Along k9i, [K](e) =
2p/ki C 12
2phTf/jk 6
3Ri + Rk 0 Ri + R k
0 0 0
0 c 2Rj + Rk s (9.96) Rj + 2Rk Ri + Rk 0 S (9.97) Ri + 3Rk
and (e)
5 f6
=
2phTf /ki 6
2Ri + Rk c 0 s (9.98) Ri + 2Rk
Let us now summarize the axisymmetric triangular formulation. The conductance matrix for an axisymmetric triangular element is (e)
[K]
M09_MOAV4303_04_GE_C09.INDD 496
b2i p rk = C bibj 2A bibk
bibj b2j bjbk
bibk d 2i p rk bjbk S + C didj 2A b2k didk
didj d 2j djdk
didk djdk S d 2k
27/11/14 10:04 AM
Section 9.5 Unsteady Heat Transfer 497
Note that the elemental conductance matrix for an axisymmetric triangular element is composed of (1) a conduction component in the radial direction; (2) a conduction component in the zdirection; and (3) a possible heat loss or gain term by convection from the edge(s) of a given element as given by Eqs. (9.93), (9.95), and (9.97). The thermal load matrix for an axisymmetric triangular element could have two components: (1) a component resulting from a possible heat generation term within a given element, Eq. (9.92), and (2) a component due to possible convection heat loss or gain from the element’s edge(s) as given by Eqs. (9.94), (9.96), and (9.98). The development of constant heat flux boundary conditions for an axisymmetric element is left as an exercise. (See Problem 25.) Example 9.2 Use the Pappus–Guldinus theorem to calculate the volume of the solid portion of the annulus generated by revolving the rectangle about the zaxis, as shown in the accompanying figure. z
3 in
r = 4 in
7 in
The centroid of the revolving rectangle is located 4 in from the zaxis. V = 2p rA = 2p(4 in)(2 in)(7 in) = 112p in3 Alternatively, the volume of the solid portion of the annulus can be calculated from V = pR22h  pR21h = p((5 in)2  (3 in)2)(7 in) = 112p in3 As you can see, the results are in agreement. 9.5 Unsteady Heat Transfer In this section, we are concerned with determining how temperatures may vary with position and time as a result of applied or existing thermal conditions. First, we review some important concepts dealing with transient heat transfer problems. A good understanding of these concepts will better assist you to model a physical problem using ANSYS. We begin our review with problems for which the spatial variation of temperature is negligible during the transient process and consequently, the variation of
M09_MOAV4303_04_GE_C09.INDD 497
27/11/14 10:04 AM
498 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems h = heat transfer coefficient,
W m2 · K
r = density of the sphere,
Tf = fluid Temperature, 5C
c = specific heat of sphere, Steel Pellet
kg m3
J kg · K
k = thermal conductivity of sphere,
W m·K
Ti = initial temperature of sphere, 5C
Figure 9.16 Cooling of a small spherical pellet.
temperature with time is only important. Consider the cooling of a small spherical steel pellet as shown in Figure 9.16. The thermophysical variables affecting the solution are shown in Figure 9.16. Because of the relatively small size of the pellet, the temperature variation with position inside the sphere is negligible. We can obtain an expression for variation of tem# # # # perature with time starting with Eq. (9.7), Ein  Eout + Egen = Estored, and realizing that # # for this problem Ein = 0, Eg = 0. Moreover, convection heat transfer occurs between # the sphere and the surrounding fluid according to Eq. (9.5), Eout = hA(T  Tf). The # Estored term represents the decrease in the thermal energy of the sphere due to the cooldT # # # # ing process and is given by Estored = rcV . Substituting for Ein = 0, Eg = 0, Eout = dt dT # hA(T  Tf), and Estored = rcV in Eq. (9.7), we get dt hA(T  Tf) = rcV
dT (9.99) dt
Assuming a relatively large cooling reservoir, it is reasonable to expect the temperature of fluid not to change with time. Thus, in Equation (9.99) we have assumed that Tf is constant. Equation (9.99) is a first order differential equation and to solve it, we let Θ = T  Tf, then we separate variables and integrate, which results in hAΘ = rcV
t Θ rcV rcV Θi dΘ dΘ 1 = dt 1 t = ln (9.100) dt hA LΘi Θ hA Θ L0
Thus, the time that it takes for a sphere to reach a certain temperature T is computed from
t =
rcV Ti  Tf ln (9.101) hA T  Tf
And if we were interested in knowing what the temperature of a sphere was at a given time, we could rearrange Eq. (9.101) in the following manner
M09_MOAV4303_04_GE_C09.INDD 498
T  Tf Ti  Tf
= exp ¢ 
hA t≤ (9.102) rcV
27/11/14 10:04 AM
Section 9.5 Unsteady Heat Transfer 499
For a transient problem, we may also be interested in knowing the amount of heat transferred to the fluid (or removed from a solid object) up to a certain time t. This information is obtained from
Q =
L 0
t
hA(T  Tf)dt (9.103)
Substituting for T  Tf from Eq. (9.102) in Eq. (9.103) and integrating it, we get Q = rcV(Ti  Tf) J1  exp ¢ 
hA t≤ R (9.104) rcV
When dealing with the analysis of transient heat transfer problems, there are two dimensionless quantities, the Biot and Fourier numbers, that are very useful. The Biot number provides a measure of the thermal resistance within the solid, being cooled (or heated in some problems), relative to the thermal resistance offered by the cooling (heating) fluid. The Biot number is defined according to
Bi =
hLc (9.105) ksolid
where Lc is a characteristic length, and is typically defined as the ratio of the volume of the object to its exposed surface areas. Small Biot numbers (typically, Bi 6 0.1) imply that thermal resistance within the solid object is negligible and consequently, the temperature distribution is approximately uniform at any instant inside the solid object. Another important variable is the Fourier number, which is a dimensionless time parameter. It provides a measure of the rate of conduction within the solid relative to the rate of the thermal storage. The Fourier number is defined as
Fo =
at (9.106) L2c
k . The thermal diffurc sivity value represents a measure of how well the material conducts heat in comparison to storing the heat. Thus, materials with low thermal diffusivity are better in storing thermal energy than they are in conducting it. The method which we described in this section to obtain solutions to problems wherein spatial variation of temperature at a given time step is neglected is called the Lumped Capacitance method. For problems in which Bi 6 0.1, lumped capacitance method renders accurate results.
In Eq. (9.106), a is called thermal diffusivity and is equal to a =
M09_MOAV4303_04_GE_C09.INDD 499
27/11/14 10:04 AM
500 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Exact Solutions We now consider problems for which spatial and temporal variations of temperature must be considered. Onedimensional transient heat transfer, described in a Cartesian coordinate system, is governed by 02T 1 0T = (9.107) 2 a 0t 0x
The solution of the governing differential heat equation requires application of two boundary conditions and one initial condition. Infinite series solutions exist for simple geometries with convective boundary conditions and uniform initial temperatures. These solutions provide temperature distribution as a function of position x inside the medium and time t. The infinite series solutions are also available for long cylinders and spheres. For heat transfer problems wherein Fo 7 0.2, the one term approximate solution provides accurate results and are presented in the form of charts called Heisler charts. For simple threedimensional problems, the solution may be represented in terms of the product of one the dimensional solutions. For an indepth review of exact solutions, you are encouraged to study a good text on the heat transfer. For example of such a text, see Incropera and DeWitt (1996). Example 9.3 In an annealing process, thin steel plates (k = 40 W/m.K, r = 7800 kg/m3, c = 400 J/kg.K) are heated to temperatures of 900°C and then cooled in an environment with temperature of 35°C and h = 25 W/m2.K. Each plate has a thickness of 5 cm. We are interested in determining how long it would take for a plate to reach a temperature of 50°C and what the temperature of the plate is after one hour. First, we need to calculate the Biot number to see if the lumped capacitance method is applicable. The characteristic length for this problem is equal to half of the plate thickness. Recall that characteristic length is defined as the ratio of volume of the object to its exposed surface areas. Thus Lc =
(area)(thickness) V 0.05 m = = = 0.025 m Aexposed 2(area) 2
Bi =
(25 W/m2.K)(0.025 m) hLc = = 0.015 ksolid 40 W/m.k
Because the Biot number is less than 0.1, the lumped capacitance method is applicable. We use Eq. (9.101) to determine the time that it takes for a plate to reach a temperature of 50°C. t =
(7800 kg/m3)(400 J/kg.K)(0.025 m) 900  35 rcV Ti  Tf ln = ln = 12650 s = 3.5 hr hA T  Tf (25 W/m.K) 50  35
M09_MOAV4303_04_GE_C09.INDD 500
27/11/14 10:04 AM
Section 9.5 Unsteady Heat Transfer 501
We use Eq. (9.102) to determine the temperature of a plate after one hour. T  Tf Ti  Tf
= exp ¢ 
(25 W/m.K)(3600 s) hA T  35 t≤ = = exp ¢ ≤ rcV 900  35 (7800 kg/m3)(400 J/kg.K)(0.025 m) 1 T = 304.4°C
Finite Difference Approach There are many practical heat transfer problems for which we cannot obtain exact solutions, consequently we resort to numerical approximations. As we explained in Chapter 1, there are two common classes of numerical methods: (1) finite difference methods and (2) finite element methods. To better understand the finite element formulation of transient problems, let us first review two common finite difference procedures, the explicit method and Implicit method. The first step of any finite difference scheme is discretization. We divide the medium of interest into a number of subregions and nodes as shown in Figure 9.17. The governing differential heat equation (e.g., Eq. (9.107)) is then written for each node, and the derivatives in the governing equations are replaced by difference equations. Using the explicit method, the temperature of an arbitrary node n at time step p + 1, Tnp + 1 is determined from the knowledge of temperatures of node n and its neighboring nodes n  1 and n + 1 at the previous time step p. Starting with expressing Eq. (9.107) in a finite difference form, Tnp 1  2Tnp + Tnp+ 1 1 Tnp + 1  Tnp = a ∆t (∆x)2 and simplifying, Tnp + 1 =
a∆t a∆t (Tnp 1 + Tnp+ 1) + J1  2¢ ≤ R Tnp (∆x)2 (∆x)2
we get Tnp + 1 = Fo(Tnp 1 + Tnp+ 1) + [1  2Fo]Tnp (9.108)
Tn 
Tn
1
x
Tn +
1
x
Figure 9.17 Discretization of the medium into smaller subregions and nodes.
M09_MOAV4303_04_GE_C09.INDD 501
27/11/14 10:04 AM
502 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
where
Fo =
a∆t (9.109) (∆x)2
We start at time t = 0 corresponding to time step p = 0 and use the initial temperature values in Eq. (9.108) to determine the temperature of node n at the next time step corresponding to time t = ∆t. We then proceed by using the newly calculated temperatures as values for time step p and compute the temperatures at next time step p + 1. The march in time continues until the nodal temperature distribution corresponding to the desired time is obtained. As the name implies, explicit method makes explicit use of the knowledge of the temperature at previous time steps to compute the nodal values at the next time step. Although the explicit procedure is simple, there is a restriction on the size of time step that must be followed. Failure to follow the restriction will lead to unstable solutions that will not make physical sense. In Eq. (9.108) the term (1  2Fo) must be positive, that is 1  2Fo Ú 0 or Fo … 12 to ensure stable solutions. To better understand this restriction, consider a situation for which temperature of node n at time step p is, say, T pn = 60°C, and the temperatures of neighboring nodes n  1 and n + 1 at time step p are T pn  1 = T pn + 1 = 80°C. If we were to use a Fourier value greater than 0.5, say 0.6, then using Equation (9.108), we would get Tnp + 1 = Fo(Tnp 1 + Tnp+ 1) + [1  2Fo]Tnp = 0.6(80 + 80) + [1  2(0.6)](60) = 84°C As you can see, the resulting temperature is not physically meaningful, because the temperature of node n at time step p + 1 cannot exceed the temperature of its neighboring nodes. This violates the second law of thermodynamics. Therefore, for onedimensional problems, the stability criterion is given by Fo …
1 (9.110) 2
For twodimensional problems, it can be shown that the stability criterion becomes Fo …
1 (9.111) 4
Implicit Method In order to overcome the time step restriction imposed by stability requirement, we can resort to implicit method. Using the implicit method, the temperature of an arbitrary node n at time step p + 1, T pn + 1 is determined from its value at the previous time p and the value of temperatures at its neighboring nodes n  1 and n + 1 at time step p + 1. The heat equation, Eq. (9.107) for an arbitrary node n is expressed by Tnp+11  2Tnp + 1 + Tnp++11 1 Tnp + 1  Tnp = a ∆t (∆x)2 and when simplified it results in
M09_MOAV4303_04_GE_C09.INDD 502
Tnp + 1  Tnp = Fo(Tnp+11  2Tnp + 1 + Tnp++11) (9.112)
27/11/14 10:04 AM
Section 9.5 Unsteady Heat Transfer 503
Separating the unknown temperatures at time step p + 1 from the known temperature at time step p in Eq. (9.112), we get Tnp + 1  Fo(Tnp+11  2Tnp + 1 + Tnp++11) = Tnp (9.113)
The implicit method leads to a set of linear equations that are solved simultaneously to obtain the nodal temperatures at time step p + 1. Example 9.4 demonstrate the steps involved in formulating and obtaining solutions using explicit and implicit schemes.
Example 9.4 Consider the long thin plate shown in Figure 9.18, which is initially at a uniform temperature of T = 250°C. The plate is made of Silicon carbide with the following properties: k = 510 W/m. K, r = 3100 kg/m3, and c = 655 J/kg.K. The plate is suddenly subjected to moving cold water with a very high convective heat transfer coefficient approximating a constant surface temperature of 0°C. We are interested in determining the variation of temperature within the plate.
T(x, 0) = 2505C
05C
05C
0T 0x
=0 x=0
T6 = 0
1 cm 1
2
3
4
5
6
x
10 cm Figure 9.18 A schematic of the plate in Example 9.4.
M09_MOAV4303_04_GE_C09.INDD 503
27/11/14 10:04 AM
504 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Explicit Solution We can make use of the symmetry of the problem, as shown in Figure 9.18, and only analyze half of the plate. The selected section of the plate is divided into six nodes as shown in Figure 9.18. The thermal diffusivity of the plate is a =
k 510 W/m.K = = 2.5 * 104 m2/s rc (3100 kg/m3)(655 J/kg.K)
To obtain meaningful results, we set Fo =
1 2
and solve for the appropriate time step
(2.5 * 10  4 m2/s)∆t a∆t 1 = = 1 ∆t = 0.2 s 2 (∆x)2 (0.01 m)2
Fo =
The nodal equations are: For node 1:
0T 2 = 0 0x x = 0
p+1
1 T1
p+1
= T2
Substituting for Fo = 0.5 in Eq. (9.108), for nodes 2, 3, 4, and 5, we have Tnp + 1 = 0.5(Tnp 1 + Tnp+ 1) p+1
= 0°C and for node 6: T6 Starting with the initial condition and marching out in time, we obtain the results shown in Table 9.2. Implicit Solution The boundary condition of node 1 requires that p+1 p+1 For node 1: T 1 = T 2 We can use Eq. (9.113) to formulate the expressions for nodes 2, 3, 4, and 5: p+1 T6
Tnp++11 + 4Tnp + 1  Tnp+11 = 2Tmp
and for node 6: = 0°C Expressing the nodal equations in a matrix form, we have 1 1 0 F 0 0 0
M09_MOAV4303_04_GE_C09.INDD 504
1 4 1 0 0 0
0 1 4 1 0 0
0 0 1 4 1 0
0 0 0 1 4 0
0 T1P + 1 0 0 T2P + 1 2T2P 0 T P+1 2T P V f 3P + 1 v = f 3P v 0 T4 2T4 P+1 1 T5 2T5P P+1 1 T6 0
27/11/14 10:04 AM
Section 9.5 Unsteady Heat Transfer 505
Table 9.2 Temperatures for Example 9.4, using the explicit method Temperature (ºC) p
n = 1 x = 0
t (sec)
2
3
4
5
6
0.01
0.02
0.03
0.04
0.05
0
0
250
250
250
250
250
1
0.2
250
250
250
250
250
0
2
0.4
250
250
250
250
125
0
125
3
0.6
250
250
250
187.5
4
0.8
250
250
218.7
187.5
93.7
250
0 0
5
1
234.3
234.3
218.7
156.2
93.7
0
6
1.2
226.5
226.5
195.2
156.2
78.1
0
7
1.4
210.8
210.8
191.3
136.6
78.1
0
8
1.6
201.1
201.1
173.7
134.7
68.3
0
1.8
187.4
187.4
167.9
121
67.3
0
2
177.6
177.6
154.2
117.6
60.5
0
9 10 11
2.2
165.9
165.9
147.6
107.3
58.8
0
12
2.4
156.7
156.7
136.6
103.2
53.6
0
13
2.6
146.6
146.6
129.9
95.1
51.6
0
14
2.8
138.2
138.2
120.8
90.7
47.5
0
15
3
129.5
129.5
114.4
84.1
45.3
0
Initially the plate is at a uniform temperature of 250°C. Using the initial conditions in the right hand side of the above matrix, we get 1 1 0 F 0 0 0
1 4 1 0 0 0
0 1 4 1 0 0
0 0 1 4 1 0
0 0 0 1 4 0
0 T11 0 0 T21 500 0 T1 500 V f 31 v = f v 0 T4 500 1 T51 500 1 T61 0
Solving the set of linear equations we obtain the nodal temperature values corresponding to t = 0.2 s [T1] = [248.36 248.36 245.09 232.02 183.00 0] Using the above results we can now solve for nodal values at t = 0.4 s 1 1 0 F 0 0 0
1 4 1 0 0 0
0 1 4 1 0 0
0 0 1 4 1 0
0 0 0 1 4 0
0 T11 0 0 T21 2 * 248.36 0 T1 2 * 245.09 V f 31 v = f v 0 T4 2 * 232.02 1 T51 2 * 183.00 1 T61 0
[T2] = [244.38 244.38 236.44 211.19 144.29 0]
M09_MOAV4303_04_GE_C09.INDD 505
27/11/14 10:04 AM
506 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Using similar steps, we can march in time to obtain solutions at any desired time. Example 9.4 demonstrated the difference between the explicit and implicit schemes. Next, we will look at the finite element formulation of transient heat transfer problems. Finite Element Approach In Sections 9.2 and 9.3 we derived the conductance and thermal load matrices for steady state heat problems. The finite element formulation of these problems leads to the following general form: [K] 5 T6 = 5 F6 (9.114)
[conductance matrix] 5 temperature matrix6 = 5 thermal load matrix6
For transient problems, we must also account for the thermal energy storage term, which leads to # [C] 5 T 6 + [K] 5 T6 = 5 F6 (9.115)
[heat storage matrix] + [conductance matrix] 5 temperature matrix6 =
5 thermal load matrix6
To obtain the solution, at discrete points in time, to the system of equations given by Eq. (9.115), we use a time integration procedure. The time step used in this procedure plays an important role in the accuracy of results. If we select a time step that is too small, then spurious oscillation may occur in the temperature solutions leading to meaningless results. On the other hand, if the time step is too large, then the temperature gradients h∆x cannot be accurately computed. We can use the Biot number aBi = b and the ksolid a∆t Fourier number aFo = b to arrive at a reasonable time step. ∆x and ∆t represent (∆x)2 the mean element width and the time step, respectively. For problems for which the Bi 6 1, the time step size may be estimated by setting the Fourier number equal to a scaling value b whose magnitude varies between 0.1 and 0.5 as shown below. a∆t Fo = = b (∆x)2 and solving for ∆t, we get
∆t = b
(∆x)2 a
where 0.1 … b … 0.5 (9.116)
If for a problem Bi 7 1, then the time step may be estimated from the product of the Fourier and Biot numbers in the following manner: (Fo)(Bi) = J
M09_MOAV4303_04_GE_C09.INDD 506
a∆t h∆x RJ R = b ksolid (∆x)2
27/11/14 10:04 AM
Section 9.6 Conduction Elements Used by ANSYS 507
and solving for ∆t, we get ∆t = b
(∆x)ksolid (∆x)rc = b ha h
where 0.1 … b … 0.5 (9.117)
ANSYS uses the following generalized Euler scheme for time integration: # # 5 T p + I 6 = 5 T p 6 + (1  u)∆t5 T p 6 + u∆t5 T p + 1 6 (9.118)
In Eq. (9.118), u is called the Euler parameter. Temperature solutions are obtained for time step p + 1 from the knowledge of temperature values at previous time step p, starting at time t = 0, corresponding to the time step p = 0, when the initial temperatures are known. For implicit scheme, which is unconditionally stable, the value of u is 1 1 limited to … u … 1. When u = , the integration scheme is commonly referred to as 2 2 Crank–Nicolson. It provides accurate results for most transient heat transfer problems. When u = 1, the integration technique is called Backward Euler and is the default setting in ANSYS. Substituting Eq. (9.118) into Eq. (9.115), gives
d
Equivalent F matrix
i
Equivalent K matrix
1 1  u # p (9.119) 1 5 F6 + [C] ¢ 5Tp6 + 5T 6 ≤ p+1 ¢ [C] + [K] ≤ 5 T 6 = u∆t 1 u∆t
System of equations represented by Eq. (9.119) are solved to obtain nodal temperatures at discrete points in time. Now, a few words about how the thermal loads should be applied. Thermal loads can be applied either suddenly as a step function or increased (or decreased) over a period of time as a ramp function as shown in Figure 9.19. Note that when using ANSYS, the value of the step load is typically applied at the first time substep. In Section 9.7 we will use ANSYS to solve a transient heat transfer problem. 9.6 Conduction Elements Used by ANSYS ANSYS offers many twodimensional thermalsolid elements that are based on linear and quadratic quadrilateral and triangular shape functions: PLANE35 is a sixnode triangular thermalsolid element. The element has one degree of freedom at each node—namely, temperature. Convection and heat fluxes may be input as surface loads at the element’s faces. The output data for this element include nodal temperatures and other data, such as thermal gradients and thermal fluxes. This element is compatible with the eightnode PLANE77 element. PLANE55 is a fournode quadrilateral element used in modeling twodimensional conduction heat transfer problems. The element has a single degree of freedom,
M09_MOAV4303_04_GE_C09.INDD 507
27/11/14 10:04 AM
Value of step load is applied at the first time subset
Thermal Load
Thermal Load
508 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Value of thermal load increases with time
Time
Time
(a) Step load
(b) Ramp load
Figure 9.19 Thermal load options for a transient problem.
which is temperature. Convection or heat fluxes may be input at the element’s faces. Output data include nodal temperatures and element data, such as thermal gradient and thermal flux components. PLANE77 is an eightnode quadrilateral element used in modeling twodimensional heat conduction problems. It is basically a higher order version of the twodimensional fournode quadrilateral PLANE55 element. This element is better capable of modeling problems with curved boundaries. At each node, the element has a single degree of freedom, which is temperature. Output data include nodal temperatures and element data, such as thermal gradient and thermal flux components. Keep in mind that although you generally achieve more accuracy of results with higher order elements, they require more computational time because numerical integration of elemental matrices is more involved. 9.7 Examples Using ANSYS Steady State Example Consider a small chimney constructed from two different materials. The inner layer is constructed from concrete with a thermal conductivity k = 0.07 Btu/hr # in # °F. The outer layer of the chimney is constructed from bricks with a thermal conductivity value k = 0.04 Btu/hr # in # °F. The temperature of the hot gases on the inside surface of the chimney is assumed to be 140°F, with a convection heat transfer coefficient of 0.037 Btu/hr # in2 # °F. The outside surface is exposed to the surrounding air, which is at 10°F, with a corresponding convection heat transfer coefficient h = 0.012
M09_MOAV4303_04_GE_C09.INDD 508
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 509
1 in
1 in
12 in
26 in
12 in Concrete
26 in
Figure 9.20 A schematic of the chimney in the example problem of Section 9.7.
Btu/hr # in2 # °F. The dimensions of the chimney are shown in Figure 9.20. Determine the temperature distribution within the concrete and within the brick layers under steadystate conditions. Also, plot the heat fluxes through each layer. The following steps demonstrate how to choose the appropriate element type, create the geometry of the problem, apply boundary conditions, and obtain nodal results for this problem using ANSYS. Enter the ANSYS program by using the Launcher. Type Chimney (or a file name of your choice) in the Jobname entry field of the dialog box. Pick Run to start the GUI. Create a title for the problem. This title will appear on ANSYS display windows to provide a simple way of identifying the displays. To create a title, issue the following command: utility menu: File S Change Title. . .
M09_MOAV4303_04_GE_C09.INDD 509
27/11/14 10:04 AM
510 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Define the element type and material properties with the commands main menu: Preprocessor S Element Type S Add/Edit/Delete
From the Library of Element Types, under Thermal Mass, choose Solid, then choose Quad 4node 55:
M09_MOAV4303_04_GE_C09.INDD 510
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 511
Assign the thermal conductivity values for concrete and brick. First, assign the value for concrete with the commands main menu: Preprocessor S Material Props S Material Models S Thermal S Conductivity S Isotropic
From the Define Material Model behavior window: Material S New Model. . .
M09_MOAV4303_04_GE_C09.INDD 511
27/11/14 10:04 AM
512 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Double click on Isotropic and assign thermal Conductivity for brick.
ANSYS Toolbar: SAVE_DB Set up the graphics area (i.e., the workplane, zoom, etc.) with the commands utility menu: Workplane S WP Settings. . .
M09_MOAV4303_04_GE_C09.INDD 512
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 513
Toggle on the workplane by the issuing the command utility menu: Workplane S Display Working Plane Bring the workplane to view with the command utility menu: PlotCtrls S Pan, Zoom, Rotate. . . Click on the small circle until you bring the workplane to view. Then create the brick section of the chimney by issuing the commands main menu: Preprocessor S Modeling S Create S Areas S Rectangle S By 2 Corners On the workplane, pick the respective locations of the corners of areas and apply: [WP = 0,0 lower left corner of the workplane] [Expand the rubber band up 26.0 and right 26.0] [WP = 6,6] [Expand the rubber band up 14.0 and right 14.0] OK To create the brick area of the chimney, subtract the two areas you have created with the commands main menu: Preprocessor S Modeling S Operate S Booleans S Subtract S Areas [Pick area 1] [Use the middle button anywhere in the ANSYS graphics window to apply] [Pick area 2] [Anywhere in the ANSYS graphics window] OK
M09_MOAV4303_04_GE_C09.INDD 513
27/11/14 10:04 AM
514 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Next, create the area of concrete by issuing the following commands: main menu: Preprocessor S Modeling S Create S Areas S Rectangles S By 2 Corners On the workplane, pick the respective locations of the corners of areas and apply: [WP = 6,6] [Expand the rubber band up 14.0 and right 14.0] This is our area number 4. [WP = 7,7] [Expand the rubber band up 12.0 and right 12.0] Apply. This is our area number 5 OK Next, subtract the two inside areas with the commands main menu: Preprocessor S Modeling S Operate S Booleans S Subtract S Areas [Pick the area number 4] [Use the middle button anywhere in the ANSYS graphics window to apply] [Pick the area number 5] [Anywhere in the ANSYS graphics window] OK To check your work thus far, plot the areas. First, toggle off the workplane and turn on area numbering by the commands utility menu: Workplane S Display Working Plane utility menu: PlotCtrls S Numbering . . .
M09_MOAV4303_04_GE_C09.INDD 514
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 515
utility menu: Plot S Areas ANSYS Toolbar: SAVE_DB We now want to mesh the areas to create elements and nodes. But first, we need to specify the element sizes. So issue the commands main menu: Preprocessor S Meshing S Size Cntrls S Manualsize S Global S Size
M09_MOAV4303_04_GE_C09.INDD 515
27/11/14 10:04 AM
516 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Next, glue areas to merge keypoints with the commands main menu: Preprocessor S Modeling S Operate S Booleans S Glue S Areas Select Pick All to glue the areas. We also need to specify material attributes for the concrete and the brick areas before we proceed with meshing. So, issue the commands main menu: Preprocessor S Meshing S Mesh Attributes S Picked Areas [Pick the concrete area] [Anywhere in the ANSYS graphics window to apply]
M09_MOAV4303_04_GE_C09.INDD 516
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 517
main menu: Preprocessor S Meshing S Mesh Attributes S Picked Areas [Pick the brick area] [Anywhere in the ANSYS graphics window to apply]
ANSYS Toolbar: SAVE_DB We can proceed with meshing now. So, issue the following commands: main menu: Preprocessor S Meshing S Mesh S Areas S Free Select Pick All and proceed. Then issue the command utility menu: PlotCtrls S Numbering . . .
M09_MOAV4303_04_GE_C09.INDD 517
27/11/14 10:04 AM
518 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Apply boundary conditions using the command main menu: Solution S Define Loads S Apply S Thermal S Convection S On lines
M09_MOAV4303_04_GE_C09.INDD 518
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 519
Pick the convective lines of the concrete, and press the OK button to specify the convection coefficient and the temperature:
main menu: Solution S Define Loads S Apply S Thermal S Convection S On lines Pick the exterior lines of the brick layer, and press the OK button to specify the convection coefficient and the temperature.
M09_MOAV4303_04_GE_C09.INDD 519
27/11/14 10:04 AM
520 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
To see the applied convective boundary conditions, issue the command utility menu: PlotCtrls S Symbols . . .
M09_MOAV4303_04_GE_C09.INDD 520
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 521
utility menu: Plot S Lines ANSYS Toolbar: SAVE_DB Now, solve the problem with the following commands: main menu: Solution S Solve S Current LS OK
M09_MOAV4303_04_GE_C09.INDD 521
27/11/14 10:04 AM
522 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Close (the solution is done!) window. Close (the/STAT Command) window. Begin the postprocessing phase. First obtain information, such as nodal temperatures and heat fluxes with the command main menu: General Postproc S Plot Results S Contour Plot S Nodal Solu
M09_MOAV4303_04_GE_C09.INDD 522
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 523
The contour plot of the temperature distribution is shown in Figure 9.21.
Figure 9.21 Temperature contour plot.
Now use the following command to plot the heat flow vectors (the plot is shown in Figure 9.22): main menu: General Postproc S Plot Results S Vector Plot S Predefined
M09_MOAV4303_04_GE_C09.INDD 523
27/11/14 10:04 AM
524 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Figure 9.22 Plot of the heat flow vectors.
M09_MOAV4303_04_GE_C09.INDD 524
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 525
Next, issue the following commands:
utility menu: Plot S Areas main menu: General Postproc S Path Operations S Define Path S On Working Plane
Pick the two points along the line marked as A–A, as shown in Figure 9.23, and press the OK button.
A
A B
B
Concrete Brick
M09_MOAV4303_04_GE_C09.INDD 525
Figure 9.23 Defining the path for path operation.
27/11/14 10:04 AM
526 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Then, issue the commands main menu: General Postproc S Path Operations S Map onto Path
M09_MOAV4303_04_GE_C09.INDD 526
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 527
main menu: General Postproc S Path Operations S Plot Path Item S On Graph
Figure 9.24 The variation of temperature gradients along path A–A.
M09_MOAV4303_04_GE_C09.INDD 527
27/11/14 10:04 AM
528 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Figure 9.25 The variation of temperature gradients along path B–B.
In a similar fashion, plot the variation of temperature gradients along path BB.
Finally, exit ANSYS and save everything: ANSYS Toolbar: Quit
M09_MOAV4303_04_GE_C09.INDD 528
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 529
Transient Example In this problem we will use ANSYS to study the transient response of a fin. Aluminum fins (k = 170 W/m.K, r = 2800 kg/m3, c = 870 J/kg.K) are commonly used to dissipate heat from various devices. An example of a section of a fin is shown in Figure 9.26. The fin is initially at a uniform temperature of 28°C. Assume that shortly after the device is turned on, the temperature of the base of the fin is suddenly increased to 90°C. The temperature of the surrounding air is 28°C with a corresponding heat transfer coefficient of h = 30 W/m2.K. T(0) = 285C
905C 25 mm Tf = 285C 105 mm
h = 30 30 mm
W m2 · K
10 mm
10 mm
Figure 9.26 A schematic of the fin in the example problem.
Enter ANSYS program by using the Launcher. Type TranFin (or a file name of your choice) in the Jobname entry field of the dialog box. Pick Run to start the Graphic User Interface (GUI). Create a title for the problem. utility menu: File S Change Title . . .
M09_MOAV4303_04_GE_C09.INDD 529
27/11/14 10:04 AM
530 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
main menu: Preprocessor S Element Type S Add/Edit/Delete
M09_MOAV4303_04_GE_C09.INDD 530
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 531
main menu: Preprocessor S Material Props S Material Models S Thermal S Conductivity S Isotropic
main menu: Preprocessor S Material Props S Material Models S Thermal S Specific Heat
M09_MOAV4303_04_GE_C09.INDD 531
27/11/14 10:04 AM
532 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
main menu: Preprocessor S Material Props S Material Models S Thermal S Density
ANSYS Toolbar: SAVE_DB
main menu: Preprocessor S Modeling S Create S Areas S Rectangle S By 2 Corners
Apply
M09_MOAV4303_04_GE_C09.INDD 532
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 533
Apply
Apply
M09_MOAV4303_04_GE_C09.INDD 533
27/11/14 10:04 AM
534 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
OK main menu: Preprocessor S Modeling S Operate S Booleans S Add S Areas Pick All main menu: Preprocessor S Meshing S Size Cntrls S Smart Size S Basic
M09_MOAV4303_04_GE_C09.INDD 534
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 535
main menu: Preprocessor S Meshing S Mesh S Areas S Free Pick All main menu: Solution S Analysis Type S New Analysis
OK
main menu: Solution S Define Loads S Settings S Uniform Temp
main menu: Solution S Define Loads S Apply S Thermal S Convection S On Lines
M09_MOAV4303_04_GE_C09.INDD 535
27/11/14 10:04 AM
536 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Pick the edges of the fin which are exposed to the convective environment, and Apply.
main menu: Solution S Define Loads S Apply S Thermal S Temperature S On Lines
Pick the line representing the base of the fin.
OK
M09_MOAV4303_04_GE_C09.INDD 536
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 537
Next, we will set a time duration of 300 seconds and a time step size of 1 second. We will also use ANSYS’s automatic time stepping capabilities to modify the time step size as needed.
main menu: Solution S Load Step Opts S Time/Frequenc S [email protected] Step
Now, we set the file write frequency to every subset.
M09_MOAV4303_04_GE_C09.INDD 537
main menu: Solution S Load Step Opts S Output Ctrls S DB/Results File
27/11/14 10:04 AM
538 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
main menu: Solution S Solve S Current LS Close (the solution is done!) window. Close (the /STAT Command) window. Next, we define a variable called corner_ pt to store the temperature of the corner node with the following location X = 0.01 m, Y = 0.0475 m, Z = 0. You can choose some other node if you wish or you can directly pick the node of interest when asked in the Time History Postprocessing.
M09_MOAV4303_04_GE_C09.INDD 538
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 539
utility menu: Parameters S Scalar Parameters . . .
Press the Accept button.
Close
M09_MOAV4303_04_GE_C09.INDD 539
27/11/14 10:04 AM
540 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
main menu: TimeHist Postproc
Add data
Press the Add Data button, the green plus button. Next, doubleclick on Nodal Solution, then DOF Solution, and Temperature.
M09_MOAV4303_04_GE_C09.INDD 540
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 541
Type Corner_Temp or a name of your choice in the Variable Name field as shown.
M09_MOAV4303_04_GE_C09.INDD 541
27/11/14 10:04 AM
542 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
Type corner_ pt in the field shown. Recall corner_point corresponds to the node with the location X = 0.01 m, Y = 0.0475 m, Z = 0. Press the Enter key and then the OK button. This is the step where you could pick another node by manual picking instead of using the defined parameter corner_ pt.
Close main menu: TimeHist Postproc S Graph Variables Type 2 in the NVAR1 1st variable to graph field.
M09_MOAV4303_04_GE_C09.INDD 542
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 543
OK
M09_MOAV4303_04_GE_C09.INDD 543
27/11/14 10:04 AM
544 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
You can now see how the temperature of the corner_ pt changes with time. The next few steps will show how to animate the temperature as it changes with time. main menu: General Postproc S Read Results S First Set Utility menu: Plot Ctrls S Numbering . . .
M09_MOAV4303_04_GE_C09.INDD 544
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 545
Utility menu: Plot Ctrls S Style S Contours S Non_uniform Contours . . .
Enter 28 in V1, 50 in V2, and 90 in V3 fields to specify the upper bounds of the first, second, and third contours. Now you can animate the results by issuing the following command: utility menu: Plot Ctrls S Animate S Over Time . . .
M09_MOAV4303_04_GE_C09.INDD 545
27/11/14 10:04 AM
546 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
M09_MOAV4303_04_GE_C09.INDD 546
27/11/14 10:04 AM
Section 9.7 Examples Using ANSYS 547
Then you can exit ANSYS.
M09_MOAV4303_04_GE_C09.INDD 547
27/11/14 10:04 AM
548 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
9.8 Verification of Results First, let us discuss some simple yet powerful ways to verify your results visually. For symmetrical problems, you should always identify lines of symmetry created by geometrical and thermal conditions. Lines of symmetry are always adiabatic lines, meaning that no heat flows in the directions perpendicular to these lines. Because no heat flows in the directions perpendicular to lines of symmetry, they constitute heat flow lines. In other words, heat flows parallel to these lines. Consider the variation of the tempera0T 0T ture gradients 0X and 0Y and their vector sum along path A–A, as shown in Figure 9.25. Note that path A–A (in Figure 9.23) is a line of symmetry and, therefore, constitutes 0T an adiabatic line. Because of this fact, the magnitude of 0X is zero along path A–A and 0T equals vector sum as shown in Figure 9.25. Comparing the variation of the tempera0Y 0T 0T ture gradients 0X and 0Y and their vector sum along path B–B, as shown in Figure 9.26, 0T 0T renders the conclusion that the magnitude of 0Y is now zero and 0X equals vector sum. Another important visual inspection of the results requires that the isotherms (lines of constant temperatures) always be perpendicular to the adiabatic lines, or lines of symmetry. You can see this orthogonal relationship in the temperature contour plot of the chimney, as shown in Figure 9.21. We can also perform a quantitative check on the validity of the results. For example, the conservation of energy applied to a control volume surrounding an arbitrary node must be satisfied. Are the energies flowing into and out of a node balanced out? This approach was demonstrated earlier with Example 9.1. Summary At this point you should 1. understand the fundamental concepts of the three modes of heat transfer. You should also know the various types of boundary conditions that could occur in a conduction problem. 2. know how the conductance matrices and the load matrices for twodimensional conduction problems were obtained. The conductance matrix for a bilinear rectangular element is
[K](e)
2 2 1 1
2 kx w  2 = D 6/  1 1
1 1 2 2
1 2 / k 1 1 y T + D 2 6w  1 2 2
1 2 2 1
1 2 2 1
2 1 T 1 2
Heat loss by convection around the edge of a rectangular element can also contribute to the conductance matrix
[K](e)
M09_MOAV4303_04_GE_C09.INDD 548
2 h/ ij 1 = D 6 0 0
1 2 0 0
0 0 0 0
0 0 T 0 0
[K](e)
0 h/ jm 0 = D 6 0 0
0 2 1 0
0 1 2 0
0 0 T 0 0
27/11/14 10:04 AM
Chapter 9 Summary 549
[K](e)
0 h/mn 0 = D 6 0 0
0 0 0 0
0 0 2 1
0 0 T 1 2
[K](e)
2 h/ni 0 = D 6 0 1
0 0 0 0
0 0 0 0
1 0 T 0 2
The load vector for a rectangular element could have many components. It could have a component due to a possible heat generation term within a given element: 1 qA 1 = d t 4 1 1 #
5 F6 (e)
It could also have a possible convection heat loss term(s) along the edge(s):
5 F6 (e) =
5 F6 (e) =
hTf / ij 2
1 1 d t 0 0
hTf / mn 2
0 0 d t 1 1
5 F6 (e) =
5 F6 (e) =
hTf / jm 2
hTf / ni 2
0 1 d t 1 0 1 0 d t 0 1
The conductance matrix for a triangular element is
(e)
[K]
b2i kx = C bibj 4A bibk
bibj b2j bjbk
bibk d 2i ky bjbk S + C didj 4A b2k didk
didj d 2j djdk
didk djdk S d 2k
Heat loss by convection around the edge of a triangular element can also contribute to the conductance matrix according to the equations
[K](e) =
[K](e) =
M09_MOAV4303_04_GE_C09.INDD 549
2 C1 6 0
1 2 0
0 0S 0
2 h/ki C0 6 1
0 0 0
1 0S 2
h/ij
[K](e) =
0 C0 6 0
h/jk
0 2 1
0 1S 2
27/11/14 10:04 AM
550 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
The load matrix for a triangular element could have many components. It could have a component due to a possible heat generation term within a given e lement: 5 F6 (e) =
1 # qA c1s 3 1
Also, it could have a possible convection heat loss term(s) along the edge(s): 5 F6(e) = [F](e) =
hTf / ij 2 hTf / ki 2
1 c1s 0
[F](e) =
hTf / jk 2
0 c1s 1
1 c0s 1
3. understand the contribution of convective boundary conditions to the conductance matrix and the forcing matrix. 4. how the conductance and load matrices were obtained for axisymmetric problems using triangular elements. 5. be familiar with formulation of transient heat transfer problems. 6. always find ways to verify your results.
References ANSYS User’s Manual: Procedures, Vol. I, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Commands, Vol. II, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Elements, Vol. III, Swanson Analysis Systems, Inc. Incropera, F., and Dewitt, D., Fundamentals of Heat and Mass Transfer, 2nd ed., New York, John Wiley and Sons, 1985. Segrlind, L., Applied Finite Element Analysis, 2nd ed., New York, John Wiley and Sons, 1984.
Problems 1. Construct the conductance matrices for the elements shown in the accompanying figure. Also, assemble the elements to obtain the global conductance matrix. The properties and the boundary conditions for each element are shown in the figure.
M09_MOAV4303_04_GE_C09.INDD 550
27/11/14 10:04 AM
Chapter 9 Problems 551
Tf = 155C, h = 30 W/m2 · K 1.75 cm 5
4
1.5 cm
k = 160 W/m · K
Tf = 305C, h = 20 W/m2 · K
(2) Insulated
3
2 k = 180 W/m · K (1) m
1
2c
Tf = 205C, h = 10 W/m2 · K
2. Construct the load matrix for each element in Problem 1. Also, assemble the elemental load matrices to construct the global load matrix.
3. Construct the conductance matrices shown in the accompanying figure. Also, assemble the elements to obtain the global conductance matrix. The properties and the boundary conditions for each element are shown in the figure. 1.75 cm Insulated 5
4
Tf = 155C, h = 30 W/m2 · K
1.5 cm
k = 160 W/m · K (2) 3
2 k = 180 W/m · K (1)
Tf = 305C, h = 20 W/m2 · K 1
m
2c
Tf = 205C, h = 10 W/m2 · K
4. Construct the load matrix for each element in Problem 3. Also, assemble the elemental load matrices to construct the global load matrix.
M09_MOAV4303_04_GE_C09.INDD 551
27/11/14 10:04 AM
552 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
5. Show that for a constant heat flux boundary condition qo>, evaluation of the terms
[S]Tqo> cos u dt and [S]Tqo> sin u dt along the edges of the rectangular element results in Lt Lt the elemental load matrices
5F 6 (e) =
5 F 6 (e)
1 1 d t 2 0 0
qo>/ij
5F 6 (e) =
0 qo>/mn 0 = d t 1 2 1
5F 6 (e)
qo>/jm 2
0 1 d t 1 0
1 qo>/ni 0 = d t 0 2 1
6. Evaluate the constant heat flux boundary condition in Problem 5 for a triangular element. 7. Using the results of Problem 5, construct the load matrix for each element shown in the accompanying figure. Also, assemble the elemental matrices to construct the global load matrix. The boundary conditions are shown in the figure. Note that if all elements are made from the same material, then heat flow becomes onedimensional. Insulated surface
10
11
12
7
8
9 Tf = 105C, h = 80 W/m2 · K
q"= 100 W/cm
4
5
6 1 cm
2
1
3
Insulated surface
1.5 cm
8. In the Galerkin formulation of twodimensional fins, the convection heat loss from the [S]ThTdA. The term contributes LA to the elemental conductance matrix. Show that for a bilinear rectangular element, the integral yields periphery of the extended surface gives rise to the term
[S]ThTdA = [S]Th[Si LA LA
M09_MOAV4303_04_GE_C09.INDD 552
Sj
Sm
Ti 4 Tj hA 2 Sn] d t dA = D Tm 36 1 Tn 2
2 4 2 1
1 2 4 2
2 Ti 1 T Td j t 2 Tm 4 Tn
27/11/14 10:04 AM
Chapter 9 Problems 553
9. Evaluate the integral in Problem 8 for a triangular element. Show that LA
[S]ThTdA =
LA
[S]Th[Si
Sj
Ti 2 hA Sk] c Tj sdA = C1 12 Tk 1
1 2 1
1 Ti 1 S c Tj s Tk 2
10. Consider a small rectangular aluminum plate with dimensions of 20 cm * 10 cm and a ther
mal conductivity value of k = 168 W/m # K, as shown in the accompanying figure. The plate is exposed to the boundary conditions shown in the figure. Using manual calculations, determine the temperature distribution within the plate, under steadystate conditions. (Hint: Because there are two axes of symmetry, you should model only a quarter of the plate.)
Tf = 205C, h = 80 W/m2 · K
Ts = 805C
Ts = 805C
Tf = 205C, h = 80 W/m2 · K
11. Aluminum fins with triangular profiles, shown in the accompanying figure, are used to remove heat from a surface whose temperature is 150°C. The temperature of the surrounding air is 20°C. The natural heat transfer coefficient associated with the surrounding air is 30 W/m2 # K. The thermal conductivity of aluminum is k = 168 W/m # K. Using manual calculations, determine the temperature distribution along a fin. Approximate the heat loss for one such fin.
M09_MOAV4303_04_GE_C09.INDD 553
27/11/14 10:04 AM
554 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
2
4
20 mm
6 7 1
3
5
90 mm
12. For the fin in Problem 11, use ANSYS to determine the temperature distribution within the fin. What is the overall heat loss through the fin? Compare these results to the results of your manual calculations.
13. Aluminum fins with parabolic profiles, shown in the accompanying figure, are used to remove heat from a surface whose temperature is 120°C. The temperature of the surrounding air is 20°C. The natural heat transfer coefficient associated with the surrounding air is 25 W/m2 # K. The thermal conductivity of aluminum is k = 168 W/m # K. Using ANSYS, determine the temperature distribution along a fin. Approximate the heat loss for one such fin. y 20 mm
5 mm
80 mm
14. Using ANSYS, determine the temperature distribution in the window assembly shown in the accompanying figure. During the winter months, the inside air temperature is kept at 68°F, with a corresponding heat transfer coefficient of h = 1.46 Btu/hr # ft2 # °F. Assume an outside air temperature of 10°F and a corresponding heat transfer coefficient of h = 6 Btu/hr # ft2 # °F. What is the overall heat loss through the window assembly?
M09_MOAV4303_04_GE_C09.INDD 554
27/11/14 10:04 AM
Chapter 9 Problems 555 1 in
Aluminum
3/4 in
1/4 in
Glass
Tf = 685F, h = 1.46 Btu/hr · ft2 · F
7 in
36 in
Tf = 105F, h = 6 Btu/hr · ft2 · F
3 in
1 in
24 in
Concrete
11.75 in
15. Aluminum fins (k = 170 W/m # K) are commonly used to dissipate heat from electronic devices. An example of such a fin is shown in the accompanying figure. Using ANSYS, determine the temperature distribution within the fin. The base of the fin experiences a constant flux of q′ = 1000 W/m. A fan forces air over the surfaces of the fin. The temperature of the surrounding air is 20°C with a corresponding heat transfer coefficient of h = 40 W/m2 # K. 2 mm
2 mm 1 mm 1 mm
6 mm
4 mm
6 mm
1 mm 1 mm
1 mm
q' = 1000 W/m
M09_MOAV4303_04_GE_C09.INDD 555
27/11/14 10:04 AM
556 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
16. Hot water flows through pipes that are embedded in a concrete slab. A section of the slab is shown in the accompanying figure. The temperature of the water inside the pipe is 50°C, with a corresponding heat transfer coefficient of 200 W/m2 # K. With the conditions shown at the surface, use ANSYS to determine the temperature of the surface. Assuming that the heat transfer coefficient associated with the hotwater flow remains constant, find the water temperature at which the surface freezes. Neglect the thermal resistance through the pipe walls. Tf = 55C, h = 34 W/m2 · K
D = 1.7 cm 0.3 m
0.15 m
7.5 cm 20 cm
Assume to be perfectly insulated 1.5 m
17. Consider the heat transfer through a basement wall with the dimensions given in the accompanying figure. The wall is constructed from concrete and has a thermal conductivity of k = 1.0 Btu/hr # ft # °F. The nearby ground has an average thermal conductivity of k = 0.85 Btu/hr # ft # °F. Using ANSYS, determine the temperature distribution within the wall and the heat loss from the wall. The inside air is kept at 68°F with a corresponding heat transfer coefficient of h = 1.46 Btu/hr # ft2 # °F. Assume an outside air temperature of 15°F, and a corresponding heat transfer coefficient of h = 6 Btu/hr # ft2 # °F. Assume that at about four feet away from the wall, the horizontal component of the heat transfer in the soil becomes negligible.
8 ft
Tinside = 685F, h = 1.46 Btu/hr · ft2 · F
6 in
M09_MOAV4303_04_GE_C09.INDD 556
Concrete wall
3 ft
Toutside = 155F, h = 6 Btu/hr · ft2 · F
Soil
4 ft
27/11/14 10:04 AM
Chapter 9 Problems 557
18. We would like to include the heat transfer rates through an uninsulated basement floor in our model in Problem 17. Considering the heat transfer model shown in the accompanying figure, determine the temperature distributions in the wall, the floor, and the soil, and the heat loss from the floor and the wall. As shown in the figure, assume that at about four feet away from the wall and the floor, the horizontal and the vertical components of the heat transfer in the soil become negligible.
3'
Toutside = 155F, h = 6 Btu/hr · ft2 · F
Tinside = 685F, h = 1.46 Btu/hr · ft2 · F
6" Concrete wall
8'
6" Concrete floor
4'
Soil
12'
4'
19. In order to enhance heat transfer rates, the inside surface of a tube is extended to form longitudinal fins, as shown in the accompanying figure. Determine the temperature distribution inside the tube wall, given the following data:
r1
H
r2 t
M09_MOAV4303_04_GE_C09.INDD 557
27/11/14 10:04 AM
558 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems r1 = 2 in r2 = 214 in t = 34 in H = 34 in
k = 400 W/m # K Tinside = 80°C hinside = 150 W/m2 # K Toutside = 15°C houtside = 30 W/m2 # K
20. Consider the concentrictube heat exchanger shown in the accompanying figure. A mixture of aqueous ethylene glycol solution arriving from a solar collector is passing through the inner tube. Water flows through the annulus as shown in the figure. The average temperature of the water at the section shown is 15°C, with a corresponding heat transfer coefficient of h = 200 W/m2 # K. The average temperature of the ethylene glycol mixture is 48°C, with an associated heat transfer coefficient of h = 150 W/m2 # K. In order to enhance the heat transfer rates between the fluids, the outside surface of the inner tube is extended to form longitudinal fins, as shown in the figure. Determine the temperature distribution inside the heat exchanger’s walls, assuming that the outside of the heat exchanger is perfectly insulated. Also, determine the heat transfer rate between the fluids. Insulation Water
5 mm
Ethylene glycol solution 17.5 mm
12.5 mm
5 mm
M09_MOAV4303_04_GE_C09.INDD 558
50 mm
27/11/14 10:04 AM
Chapter 9 Problems 559
21. Use Green’s theorem Eq. (9.36) to calculate the area of the triangle shown in the accompanying figure. y
2
3 a
x
b 1
22. Calculate the conductance matrix for the axisymmetric triangular element shown. The element belongs to a solid body made of aluminum alloy with a thermal conductivity k = 200 W/m # K. z Tf = 155 C W h = 25 2 m .k
k (5, 8) Coordinates in mm . q = 6 * 103 W/mm3
i (2, 2)
j (9, 3)
r
23. Construct the load matrix for Problem 22. 24. Construct the load matrix for each axisymmetric triangular element shown. Also, assemble the elemental load matrices to construct the global load matrix. Tf = 155 C, h = 30 W/m2 . K 1.75 cm 4
k = 160 W/m . K
5
(3)
1.5 cm
(2) Insulated
k = 160 W/m . K
2
Tf = 305 C, h = 30 W/m2 . K
3
k = 180 W/m . K (1) 1
m
2c
Tf = 205 C, h = 30 W/m2 . K
M09_MOAV4303_04_GE_C09.INDD 559
27/11/14 10:04 AM
560 Chapter 9 Analysis of TwoDimensional Heat Transfer Problems
25. Evaluate the constant heatflux boundary condition for an axisymmetric triangular element. T
26. Evaluate the integral 27. Evaluate the integral 
k 0[S] 0T dV for an axisymmetric rectangular element. r 0r 0r L V 0[S]T 0T
LV 0z 0z
dV for an axisymmetric rectangular element.
28. Under certain conditions, heat transfer rate q in a twodimensional system may be computed using the equation q = kS(T1  T2), where k is the thermal conductivity of the medium, S is the shape factor, and T1 and T2 are temperatures of the surfaces, as shown in accompanying diagrams. Using ANSYS plot the temperature distribution in the soil for each of the situations shown. Also, using ANSYS calculate the heattransfer rate for each case and compare it to the shape factor solution. First situation deals with an isothermal sphere that is buried in soil with k = 0.5 W/m # K, z = 10 m, D = 1 m, T1 = 300°C, T2 = 27°C, and the corresponding shape factor is given by S =
2pD . D 1 4z T2
z
T1
Soil D
Second situation deals with an isothermal horizontal pipe is buried in soil with k = 0.5 W/m # K, z = 2 m, D = 0.5 m, L = 50 m, T1 = 100°C, T2 = 5°C and the corre2pL . sponding shape factor is given by S = 4z lna b D T2
Soil
z
T1
D
L
M09_MOAV4303_04_GE_C09.INDD 560
27/11/14 10:04 AM
Chapter 9 Problems 561
29. For the fin in Problem 15, assume that the fin is initially at the surrounding temperature of 20°C and the base of the fin experiences a sudden flux of q′ = 1000 W/m. Study the transient response of the fin. Compare the steady state results to the transient response for long after the sudden flux is applied.
30. For the slab in Problem 16, assume that the slab is initially at the temperature of surrounding  5°C. Suddenly hot water at a temperature of 50°C with a corresponding heat transfer coefficient of 200 W/m2 # K is introduced inside the pipes. Study the transient response of the system. How long will it take for the surface of the slap to reach a temperature of 2°C.
31. Design Problem At some ski resorts, in order to keep ice from forming on the surface of uphill roads leading to condominiums, hot water is pumped through pipes that are embedded beneath the surface of the road. You are to design a hydronic system to perform such a task. Choose your favorite ski resort and look up its design conditions, such as the ambient air temperature, and soil temperature. The system that you construct may consist of a series of tubes, pumps, a hotwater heater, valves, fittings, and so on. Basic information sought includes the type of pipes, their sizes, the spacing between the tubes, the configuration of the piping system, and the distance below the surface the pipes should be embedded. If time allows, you may also size the pump and the hotwater heater.
M09_MOAV4303_04_GE_C09.INDD 561
27/11/14 10:04 AM
C h a p t e r
1 0
Analysis of TwoDimensional Solid Mechanics Problems
The objective of this chapter is to introduce you to the analysis of twodimensional solid mechanics problems. Structural members and machine components are generally subject to a push–pull, bending, or twisting type of loading. The components of common structures and machines normally include beams, columns, plates, and other members that can be modeled using twodimensional approximations. Axial members, beams, and frames were discussed in Chapter 4. The main topics discussed in Chapter 10 include the following: 10.1 Torsion of Members with Arbitrary CrossSection Shape 10.2 PlaneStress Formulation 10.3 Isoparametric Formulation: Using a Quadrilateral Element 10.4 Axisymmetric Formulation 10.5 Basic Failure Theories 10.6 Examples Using ANSYS 10.7 Verification of Results
10.1 Torsion of Members with Arbitrary CrossSection Shape There are still many practicing engineers who generate finite element models for problems for which there exist simple analytical solutions. For example, you should not be too quick to use the finite element method to solve simple torsional problems. This type of problem includes torsion of members with circular or rectangular cross sections. Let us briefly review the analytical solutions that are available for torsional problems. When studying the mechanics of materials, you were introduced to the torsion of long, straight members with circular cross sections. A problem is considered to be a torsional
M10_MOAV4303_04_GE_C10.INDD 562
27/11/14 10:07 AM
Section 10.1 Torsion of Members with Arbitrary CrossSection Shape 563
T
T
Figure 10.1 Torsion of a shaft.
problem when the applied moment or torque twists the member about its longitudinal axis, as shown in Figure 10.1. Over the elastic limit, the shear stress (t) distribution within a member with a circular cross section, such as a shaft or a tube, is given by the equation
t =
Tr (10.1) J
where T is the applied torque, r is the radial distance measured from the center of the shaft to a desired point in the cross section, and J represents the polar moment of inertia of the crosssectional area. It should be clear from examination of Eq. (10.1) that the maximum shear stress occurs at the outer surface of the shaft, where r is equal to the radius of the shaft. Also, recall that the angle of twist (u) caused by the applied torque can be determined from the equation
TL (10.2) JG
u =
in which L is the length of the member and G is the shear modulus (modulus of rigidity) of the material. Furthermore, there are analytical solutions that can be applied to torsion of members with rectangular crosssectional areas.* When a torque is applied to a straight bar with a rectangular crosssectional area, within the elastic region of the material, the maximum shearing stress and an angle of twist caused by the torque are given by
tmax =
T (10.3) c1wh2
u =
TL (10.4) c2Gwh3
L is the length of the bar and w and h are the larger and smaller sides of the crosssection, respectively. (See Figure 10.2.) The values of coefficients c1 and c2 (given in Table 10.1) are dependent on the aspect ratio of the cross section. As the aspect ratio approaches large numbers (w/h S ∞), c1 = c2 = 0.3333. This relationship is demonstrated in Table 10.1. *See Timoshenko and Goadier (1970) for more detail.
M10_MOAV4303_04_GE_C10.INDD 563
27/11/14 10:07 AM
564 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems h
w
L
Figure 10.2 A straight rectangular bar in torsion.
The maximum shear stress and the angle of twist for crosssectional geometries with high aspect ratios (w/h 7 10) are given by:
tmax =
T (10.5) 0.333 wh2
u =
TL (10.6) 0.333 Gwh3
These types of members are commonly referred to as thinwall members. Examples of some thinwall members are shown in Figure 10.3. Therefore, if you come across a problem that fits these categories, solve it using the torsional formulae. Do not spend a great deal of time generating a finite element model. Finite Element Formulation of Torsional Problems Fung (1965) discusses the elastic torsional behavior of noncircular shafts in detail. There are two basic theories: (1) St. Venant’s formulation and (2) the Prandtl formulation. Table 10.1 c1 and c2 values for a bar with a rectangular cross section w/h
M10_MOAV4303_04_GE_C10.INDD 564
c1
c2
1.0
0.208
0.141
1.2
0.219
0.166
1.5
0.231
0.196
2.0
0.246
0.229
2.5
0.258
0.249
3.0
0.267
0.263
4.0
0.282
0.281
5.0
0.291
0.291
10.0
0.312
0.312
∞
0.333
0.333
27/11/14 10:07 AM
Section 10.1 Torsion of Members with Arbitrary CrossSection Shape 565
Figure 10.3 Examples of thinwall members.
Here, we will use the Prandtl formulation. The governing differential equation for the elastic torsion of a shaft in terms of the stress function f is 02f
0x2
+
02f 0y2
+ 2Gu = 0 (10.7)
where G is the shear modulus of elasticity of the bar and u represents the angle of twist per unit length. The shear stress components are related to the stress function f according to the equations 0f (10.8) 0y
tzx =
tzy = 
0f (10.9) 0x
Note that with Prandtl’s formulation, the applied torque does not directly appear in the governing equation. Instead, the applied torque is related to the stress function and is
T = 2
f dA (10.10) L A
In Eq. (10.10), A represents the crosssectional area of the shaft. Comparing the differential equation governing the torsional behavior of a member, Eq. (10.7), to the heat diffusion equation, Eq. (9.8), we note that both of these equations have the same form. Therefore, we can apply the results of Section 9.2 and Section 9.3 to torsional problems. However, when comparing the differential equations for torsional problems, we let c1 = 1 and c2 = 1, c3 = 2Gu. The stiffness matrix for a rectangular element
M10_MOAV4303_04_GE_C10.INDD 565
27/11/14 10:07 AM
566 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems fn f fi fm fj y m
n
w
Y j
i
x
/ X
Figure 10.4 Nodal values of the stress function for a rectangular element.
then becomes
[K](e)
2 w 2 = D 6/ 1 1
2 2 1 1
1 1 2 2
1 2 1 / 1 T + D 2 6w 1 2 2
1 2 2 1
1 2 2 1
2 1 T (10.11) 1 2
where w and / are the length and the width, respectively, of the rectangular element, as shown in Figure 10.4. The load matrix for an element is
{F}(e)
1 2GuA 1 = d t (10.12) 4 1 1
and for triangular elements, shown in Figure 10.5, the stiffness and load matrices are (e)
[K]
b2i 1 = C bibj 4A bibk
F(e) =
bibj b2j bjbk
bibk d 2i 1 bjbk S + C didj 4A b2k didk
didj d 2j djdk
1 2GuA c1s 3 1
didk djdk S (10.13) d 2k (10.14)
where the area A of the triangular element and the a, b, and [email protected] are given by 2A ai aj ak
= = = =
Xi(Yj  Yk) + Xj(Yk  Yi) + Xk(Yi  Yj) XjYk  XkYj bi = Yj  Yk di = Xk  Xj XkYi  XiYk bj = Yk  Yi dj = Xi  Xk XiYj  XjYi bk = Yi  Yj dk = Xj  Xi
Next, we will consider an example problem dealing with torsion of a steel bar.
M10_MOAV4303_04_GE_C10.INDD 566
27/11/14 10:07 AM
Section 10.1 Torsion of Members with Arbitrary CrossSection Shape 567 f
fk fj
fi (Xj, Yj) Y (Xk, Yk)
(Xi, Yi) X
Figure 10.5 Nodal values of the stress function for a triangular element.
Example 10.1 Consider the torsion of a steel bar (G = 11 * 103 ksi) having a rectangular cross section, as shown in the accompanying figure. Assume that u = 0.0005 rad/in. Using the finite element procedure discussed above, we are interested in determining the shear stress distribution within the bar. Steel bar G = 11 * 103 ksi P
0.5
in 1 in
1 in
P
Before we proceed with the solution, note that the point of this exercise is to demonstrate the finite element steps for a torsional problem. As we mentioned earlier, this problem has a simple analytical solution which we will present when we revisit this problem later in this chapter.
M10_MOAV4303_04_GE_C10.INDD 567
27/11/14 10:07 AM
568 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Y 6
(3) 4
(2)
(1) 1
5
2
X
3
Figure 10.6 A schematic of the bar in Example 10.1.
We can make use of the symmetry of the problem, as shown in Figure 10.6, and only analyze a section of the shaft containing 1/8 of the area. The selected section of the shaft is divided into six nodes and three elements (a crude model). Elements (1) and (3) are triangular, while element (2) is rectangular. Consult Table 10.2 while following the solution. For triangular elements (1) and (3), the stiffness matrix is (1)
[K]
(3)
= [K]
b2i 1 = C bibj 4A bibk
bibk d 2i 1 bjbk S + Cdd 4A i j b2k didk
bibj b2j bjbk
didj d 2j djdk
didk djdk S d 2k
Evaluating the [email protected] and [email protected] for element (1), we get bi = Yj  Yk = 0  0.25 = 0.25 di = Xk  Xj = 0.125  0.125 = 0 bj = Yk  Yi = 0.25  0 = 0.25
dj = Xi  Xk = 0  0.125 = 0.125
bk = Yi  Yj = 0  0 = 0
dk = Xj  Xi = 0.125  0 = 0.125
Table 10.2 The relationship between elements and their corresponding nodes and their [coordinates] Element
i
j
m or k
(1)
1 [0, 0]
2 [0.125, 0]
(2)
2 [0.125, 0]
3 [0.25, 0]
5 [0.25, 0.25]
(3)
4 [0.125, 0.25]
5 [0.25, 0.25]
6 [0.25, 0.5]
M10_MOAV4303_04_GE_C10.INDD 568
n
4 [0.125, 0.25] 4 [0.125, 0.25]
27/11/14 10:07 AM
Section 10.1 Torsion of Members with Arbitrary CrossSection Shape 569
Note that evaluating the [email protected] and [email protected] for element (3) renders the same values because the difference between the coordinates of its nodes is identical to that of element (1). Therefore, elements (1) and (3) will both have the following stiffness matrix: [K](1) = [K](3) =
0.0625 1 C 0.0625 4(0.015625) 0 +
0.0625 0.0625 0
0 0S 0
0 1 C0 4(0.015625) 0
0 0.015625 0.015625
0 0.015625 S 0.015625
To help with assembly of the elements later, the corresponding node numbers for the elements are shown on the top and the side of the stiffness matrices. The stiffness matrices for elements (1) and (3), are
[K](1)
[K](3)
1(i) 1 = C 1 0
2( j) 1 1.25 0.25
4(k) 0 1 0.25 S 2 0.25 4
4(i) 1 = C 1 0
5( j) 1 1.25 0.25
6(k) 0 4 0.25 S 5 0.25 6
The stiffness matrix for element (2) is
[K](2)
2 w 2 = D 6/ 1 1
2 2 1 1
1 1 2 2
1 2 1 / 1 T + D 6w 1 2 2 2
[K](2)
2 0.25 2 = D 6(0.125) 1 1
2 2 1 1
1 1 2 2
1 2 2 1
1 2 2 1
1 2 1 0.125 1 T + D 2 6(0.25) 1 2 2
2 1 T 1 2 1 2 2 1
1 2 2 1
2 1 T 1 2
The stiffness matrix for element (2) with the corresponding node numbers is
(2)
[K]
2(i) 0.83333333 = 0.58333333 D 0.41666667 0.16666667
3( j) 0.58333333 0.83333333 0.16666667 0.41666667
5(m) 0.4166666 0.16666667 0.83333333 0.58333333
4(n) 0.16666667 0.41666667 T 0.58333333 0.83333333
2 3 5 4
The load matrices for the triangular elements (1) and (3) are computed from
M10_MOAV4303_04_GE_C10.INDD 569
27/11/14 10:07 AM
570 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
5 F6(e) =
1 57.29166667 1 57.29166667 4 2GuA c 1 s leading to: 5 F6(1) = c 57.29166667 s 2 and 5 F6(3) = c 57.29166667 s 5 3 1 57.29166667 4 57.29166667 6
The load matrix for the rectangular element (2) including the nodal information is
5 F6(2)
1 85.9375 2GuA 1 85.9375 = d t = d t 4 1 85.9375 1 85.9375
2 3 5 4
Using the nodal information presented next to each element, the global stiffness matrix becomes
[K](G)
1 1 1 = 0 F 0 0 0
2 1 2.08333333 0.58333333 0.08333333 0.41666666 0
3 0 0.58333333 0.83333333 0.41666667 0.16666667 0
4 0 0.08333333 0.41666667 2.08333333 1.58333333 0
5 6 0 0 0.41666666 0 0.16666667 0 V 1.58333333 0 2.08333333 0.25 0.25 0.25
1 2 3 4 5 6
Assembling the load matrices, we get
5 F6(G)
57.3 57.3 57.3 + 85.9 143.2 85.9 85.93 = f v = f v 57.3 + 57.3 + 85.9 200.5 57.3 + 85.9 143.2 57.3 57.3
Applying the boundary condition of f = 0 to nodes 3, 5, and 6, we get the following 3 * 3 matrix: 1 C 1 0
1 2.083 0.083
0 f1 57.3 0.083 S c f 2 s = c 143.2 s 2.083 f3 200.5
Solving the set of linear equations simultaneously leads to the following nodal solution: [f]T = [250
193
0
104
0
0]
We can compute the shear stress components (see Figure 10.7) from Eqs. (10.8) and (10.9) in the following manner. For elements (1) and (3) tZX =
0f 0 0 = [Sif i + Sjf j + Skf k] = [S 0Y 0Y 0Y i
M10_MOAV4303_04_GE_C10.INDD 570
Sj
fi 1 Sk]c f j s = [d 2A i fk
dj
fi dk]c f j s fk
27/11/14 10:07 AM
Section 10.1 Torsion of Members with Arbitrary CrossSection Shape 571
Y
X tZY tZX
Z
tZY = 
Figure 10.7 The directions of shear stress components.
0f 0 0 =  [Sif i + Sjf j + Skf k] =  [Si 0X 0X 0X
Sj
fi Sk]c f j s fk
1 [b 2A i
bj
fi bk]c f j s fk
= 
The shear stress components for element (1) are (1) tZX
(1) tZY
= [di
= [bi
fi 1 dk]c f j s = [0 0.03125 fk
dj
bj
0.125
fi 1 bk]c f j s = [0.25 0.03125 fk
250 0.125] c 193 s = 356 lb/in2 104 0.25
250 0] c 193 s = 456 lb/in2 104
Similarly, we can compute the shear stress components for element (3). (3) tZX
1 = [0 0.03125
(3) tZY
1 = [0.25 0.03125
0.125
104 0.125] c 0 s = 0 lb/in2 0
0.25
104 0]c 0 s = 832 lb/in2 0
Because the spatial derivatives of linear triangular elements are constants, note that the computed shear stress distribution over the region represented by the triangular element is constant. This is one of the drawbacks of using linear triangular elements. We could have divided the selected section of the shaft into many more elements or used
M10_MOAV4303_04_GE_C10.INDD 571
27/11/14 10:07 AM
572 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
higher order elements, as discussed in Chapter 7, to come up with better results. The shear stress components over the rectangular element is obtained in the same manner.
tZX
0f 0 0 = = [Sif i + Sjf j + Smf m + Snf n] = [S 0Y 0Y 0Y i
tZX
1 = [( / + x) /w
x
x
Sj
Sm
fi f Sn]d j t fm fn
fi f (/  x)] d j t fm fn
Similarly,
tZY
0f 1 = =  [( w + y) 0X /w
(w  y)
y
fi f y] d j t fm fn
Note that for the bilinear rectangular element, the shear stress components vary with position and can be computed for a specific location within the element. It is left as an exercise for you to substitute for the local coordinates of a point within the boundaries of element (2) to obtain the shear stress components. We will revisit this problem later and solve it using ANSYS. Example 10.1 Revisited We will now show how to use Excel to set up and solve Example 10.1. 1. In cell A1 type Example 10.1, and in cells A3 and A4 type G= and 𝚯= as shown. After inputting the value of G in cell B3, select B3 and in the “Name Box” type G and hit the Return key. Similarly, after inputting the value of Θ in cell B4, select B4 and in the “Name Box” type Theta and hit the Return key.
2. Create the table shown with the element and node numbers and the coordinates of the nodes. In cell F8, type [email protected]; in cell F9 type [email protected]; in cell F10 type [email protected]; and name them deltai, deltaj, and deltak, respectively. Similarly, in cells G8 through G10 create betai, betaj, and betak. Note the contents of G8, G9, and G10 are [email protected]; [email protected]; [email protected]
M10_MOAV4303_04_GE_C10.INDD 572
27/11/14 10:07 AM
Section 10.1 Torsion of Members with Arbitrary CrossSection Shape 573
In cell I9 and J9 type Atriangle and its value. Name the content of J9, Atriangle. Similarly, create W and L and the corresponding values and name them W and L (cell J13 and J14 W and L). Calculate the area of the rectangle in cell J15 using =W*L and name it Arectangle.
3. Compute the [K1] and [K3] matrices, using the delta, beta, and area values. Name the range G22:I24, Kelment1 and Kelement3.
4. In a similar way create Kelement2.
M10_MOAV4303_04_GE_C10.INDD 573
27/11/14 10:07 AM
574 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
5. Create the {F1}, {F3}, and {F2} matrices as shown.
6. Next, create [A1], [A3], and [A2] matrices and name them Aelement1, Aelement3, Aelment2, as shown. If you have forgotten what A matrices represent, see Section 2.5, Eq. (2.9).
7. Next, create the stiffness matrix for each element (with their proper positions in the global matrix) and name them K1G, K3G, and K2G. See Eq. (2.9). For example, to create [K]1G, select B54:G59 and type = MMULT(TRANSPOSE(Aelement1),MMULT(Kelement1,Aelement1)) and while holding down the Ctrl and Shift keys, hit the Return key. In a similar way, create [K]3Gand [K]2G as shown.
M10_MOAV4303_04_GE_C10.INDD 574
27/11/14 10:07 AM
Section 10.1 Torsion of Members with Arbitrary CrossSection Shape 575
8. Create the load matrices as shown.
9. The final global matrix is created next. Select the range B82:G87 and type =K1G+K2G+K3G and while holding down the Ctrl and Shift keys, hit the Return key. Name the range B82:G87, KG.
10. Create the load matrix as shown and name it FG.
11. Apply the boundary conditions. Copy the appropriate portion of the KG matrix and paste it in the range C96:E98 as values only. Name the range KwithappliedBC. Similarly, create the corresponding load matrix in the range C100:C102 and name it FwithappliedBC.
M10_MOAV4303_04_GE_C10.INDD 575
27/11/14 10:07 AM
576 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
12. Select the range C104:C106 and type =MMULT(MINVERSE(KwithappliedBC),FwithappliedBC) and while holding down the Ctrl and Shift keys hit the Return key.
The complete Excel sheet is shown next.
M10_MOAV4303_04_GE_C10.INDD 576
27/11/14 10:07 AM
Section 10.1 Torsion of Members with Arbitrary CrossSection Shape 577
M10_MOAV4303_04_GE_C10.INDD 577
27/11/14 10:07 AM
578 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
10.2 PlaneStress Formulation We begin by reviewing some of the fundamental concepts dealing with the elastic behavior of materials. Consider an infinitesimally small cube volume surrounding a point within a material. An enlarged version of this volume is shown in Figure 10.8. The faces Y sYY
tYX tYZ tZY
tXY tZX
sZZ
Z
M10_MOAV4303_04_GE_C10.INDD 578
sXX
tXZ X
Figure 10.8 The components of stress at a point.
27/11/14 10:07 AM
Section 10.2 PlaneStress Formulation 579
of the cube are oriented in the directions of (X, Y, Z) coordinate system.* The application of external forces creates internal forces and, subsequently, stresses within the material. The state of stress at a point can be defined in terms of the nine components on the positive faces and their counterparts on the negative surfaces, as shown in the figure. However, recall that because of equilibrium requirements, only six independent stress components are needed to characterize the general state of stress at a point. Thus the general state of stress at a point is defined by: [S]T = [sXX
sYY
sZZ
tXY
tYZ
tXZ] (10.15)
where sXX, sYY, and sZZ are the normal stresses and tXY, tYZ, and tXZ are the shear stress components, and they provide a measure of the intensity of the internal forces acting over areas of the cube faces. In many practical problems, we come across situations where there are no forces acting in the Zdirection and, consequently, no internal forces acting on the Z faces. This situation is commonly referred to as a planestress situation, as shown in Figure 10.9. For a planestress situation, the state of stress reduces to three components: [S]T = [sXX
sYY
tXY] (10.16)
Y sYY tYX tXY sXX
sXX tXY
X
tYX
sYY Z
sYY tYX tXY sXX
sXX tXY tYX sYY Figure 10.9 Plane state of stress. *Note that throughout this section X, Y, Z and x, y, z coordinate systems are aligned.
M10_MOAV4303_04_GE_C10.INDD 579
27/11/14 10:08 AM
580 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
We have just considered how an applied force can create stresses within a body. As you know, the applied force will also cause a body to undergo deformation, or change in its shape. We can use a displacement vector to measure the changes that occur in the posiS tion of a point within a body. The displacement vector d can be written in terms of its Cartesian components as S S S S d = u(x, y, z) i + y(x, y, z) j + w(x, y, z) k where the i, j, and k components of the displacement vector represent the difference in the coordinates of the displacement of the point from its original position (x, y, z) to a new position (x′, y′, z′) caused by loading, as given by the equations u(x, y, z) = x′  x y(x, y, z) = y′  y w(x, y, z) = z′  z To better measure the size and shape changes that occur locally within the material, we define normal and shear strains. The state of strain at a point is, therefore, characterized by six independent components: [E]T = [exx
eyy
ezz
gxy
gyz
gxz] (10.17)
exx, eyy, and ezz are the normal strains, and gxy, gyz, and gxz are the shearstrain components. These components provide information about the size and shape changes that occur locally in a given material due to loading. The situation in which no displacements occur in the zdirection is known as a planestrain situation. As you may recall from your study of the mechanics of materials, there exists a relationship between the strain and the displacement. These relationships are
exx =
gxy
0u 0x
eyy =
0u 0v = + 0y 0x
gyz
0y 0y
ezz =
0w 0z
0v 0w = + 0z 0y
gxz
0u 0w = + 0z 0x
(10.18)
Over the elastic region of a material, there also exists a relationship between the state of stresses and strains, according to the generalized Hooke’s law. These relationships are
exx =
1 [s  n(syy + szz)] E xx
(10.19)
eyy =
1 [s  n(sxx + szz)] E yy
ezz =
1 [s  n(sxx + syy)] E zz
gxy =
1 1 1 t gyz = t gzx = t G xy G yz G zx
M10_MOAV4303_04_GE_C10.INDD 580
27/11/14 10:08 AM
Section 10.2 PlaneStress Formulation 581
where E is the modulus of elasticity (Young’s modulus), v is Poisson’s ratio, and G is the shear modulus of elasticity (modulus of rigidity). For a planestress situation, the generalized Hooke’s law reduces to 1 sxx E n c syy s = D 2 1  n txy 0
0 exx 0 T c eyy s (10.20) 1  n gxy 2
n 1 0
or, in a compact matrix form, 5 S 6 = [v]5 E 6 (10.21)
where
[S]T = [sxx
[N] =
syy
E D 1  n2
txy]
1 n
n 1
0
0
0 0 T 1  n 2
exx {E} = c eyy s gxy For a planestrain situation, the generalized Hooke’s law becomes sxx E c syy s = D (1 + n)(1  2n) txy
1  n n
n 1  n
0
0
0 0
exx T c eyy s (10.22) 1 gxy  n 2
Furthermore, for plane stress situations, the strain–displacement relationship becomes
exx =
0u 0x
eyy =
0y 0y
gxy =
0u 0y + (10.23) 0y 0x
We have discussed throughout this text that the minimum total potential energy approach is very commonly used to generate finite element models in solid mechanics. External loads applied to a body will cause the body to deform. During the deformation, the work done by the external forces is stored in the material in the form of elastic energy, which is called strain energy. For a solid material under biaxial loading, the strain energy Λ is
M10_MOAV4303_04_GE_C10.INDD 581
Λ(e) =
1 (sxx exx + syy eyy + txygxy) dV (10.24) 2L V
27/11/14 10:08 AM
582 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Or, in a compact matrix form, Λ(e) =
1 [S]T 5 E6 dV (10.25) 2L V
Substituting for stresses in terms of strains using Hooke’s law, Eq. (10.25) can be written as Λ(e) =
1 1 1 ([v]{E})T{E} = {E}T[v]T{E} = 5 E6T[v]5 E6 dV (10.26) 2 LV 2L 2L V V
Note in arriving at the final form of integral, we have used Eq. (2.12) and [y]T = [y]. We are now ready to look at finite element formulation of plane stress problems using triangular elements. We can represent the displacements u and v using a linear triangular element similar to the one shown in Figure 10.10. The displacement variable, in terms of linear triangular shape functions and the nodal displacements, is
u = SiUix + SjUjx + SkUkx (10.27)
y = SiUiy + SjUjy + SkUky We can write the relations given by Eq. (10.27) in a matrix form:
0 Si
u S b r = J i v 0
Sj 0
0 Sj
Sk 0
Uix Uiy 0 U R f jx v (10.28) Sk Ujy Ukx Uky
Uky k
Ukx
Uiy
i
Uix
Ujy
j
M10_MOAV4303_04_GE_C10.INDD 582
Ujx
Figure 10.10 A triangular element used in formulating plane stress problems.
27/11/14 10:08 AM
Section 10.2 PlaneStress Formulation 583
The next step involves relating the strains to the displacement field and, subsequently, relating the strains to the nodal displacements using shape functions. Referring to the strain–displacement relations as given by Eq. (10.23), we need to take the derivatives of the components of the displacement field with respect to the x and y coordinates, which in turn means taking the derivatives of the appropriate shape functions with respect to x and y. Performing these operations results in the following relations: 0u 0 1 = (S U + SjUjx + SkUkx) = [b U + bjUjx + bkUkx] (10.29) 0x 0x i ix 2A i ix 0y 0 1 = = (SiUiy + SjUjy + SkUky) = [d U + djUjy + dkUky] 0y 0y 2A i iy 0u 0y 1 = + = [d U + biUiy + djUjx + bjUjy + dkUkx + bkUky] 0y 0x 2A i ix
exx = eyy gxy
Representing the relations of Eq. (10.29) in a matrix form, we have
exx bi 1 c eyy s = C0 2A gxy di
0 di bi
bj 0 dj
0 dj bj
bk 0 dk
Uix Uiy 0 U dk S f jx v (10.30) Ujy bk Ukx Uky
and in a compact matrix form, Eq. (10.30) becomes 5 E 6 = [B] 5 U6 (10.31)
where
exx 5 E6 = c eyy s gxy
[B] =
bi 1 C0 2A di
0 di bi
bj 0 dj
0 dj bj
bk 0 dk
0 dk S bk
Uix Uiy U 5 U6 = f jx v Ujy Ukx Uky
Substituting into the strain energy equation for the strain components in terms of the displacements, we obtain Λ(e) =
1 1 5 E 6T[v]5 E 6 dV = [U]T[B]T[v][B][U] dV (10.32) 2L 2 L V V
Differentiating with respect to the nodal displacements, we obtain
M10_MOAV4303_04_GE_C10.INDD 583
0Λ(e) 0 1 = ¢ [U]T[B]T[v][B][U] dV ≤ 0Uk 0Uk 2 L V
for k = 1, 2, c 6 (10.33)
27/11/14 10:08 AM
584 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Evaluation of Eq. (10.33) results in the expression [K](e) 5U6. The expression for the stiffness matrix is thus [K](e) =
[B]T[v][B] dV = V[B]T[v][B] (10.34) LV
Here, V is the volume of the element and is the product of the area of the element and its thickness. Example 10.2 will show how Eq. (10.34) is used to evaluate the stiffness matrix for a twodimensional triangular plane stress element. Load Matrix To obtain the load matrix for a twodimensional plane stress element, we must first compute the work done by the external forces, such as distributed loads or point loads. The work done by a concentrated load Q is the product of the load component and the corresponding displacement component. We can represent the work done by concentrated loads in a compact matrix form as W (e) = 5 U6T 5 Q 6 (10.35)
A distributed load with px and py components does work according to the relationship W(e) =
L A
(upx + `py) dA (10.36)
where u and y are the displacements in the x and y directions, respectively, and A represents the surface over which the distributed load components are acting. The magnitude of the surface A is the product of the element thickness t and the length of the edge over which the distributed load is applied. Using triangular elements to represent the displacements, we find that the work done by distributed loads becomes W(e) =
where
L A
5 U6T[S]T 5 p 6 dA (10.37)
5p6 = b
px r py
The next step in evaluating the load matrix involves the minimization process. In the case of the concentrated load, differentiation of Eq. (10.35) with respect to nodal displacements yields the components of the loads:
M10_MOAV4303_04_GE_C10.INDD 584
5 F6(e)
Qix Qiy Q = f jx v (10.38) Qjy Qkx Qky
27/11/14 10:08 AM
Section 10.2 PlaneStress Formulation 585
The differentiation of the work done by the distributed load with respect to the nodal displacements gives the load matrix 5 F6(e) =
where
[S]T 5 p 6 dA (10.39) L A
Si 0 S j [S]T = F 0 Sk 0
0 Si 0 V Sj 0 Sk
Consider an element subjected to a distributed load along its kiedge, as shown in Figure 10.11. Evaluating Eq. (10.39) along the kiedge and realizing that along the kiedge, Sj = 0, we have
5 F6(e)
0 Si Si 0 0 px 0 V dA = t F Sj b p y r L/ki 0 0 Sk Sk 0
Si 0 Sj = F 0 L A Sk 0
0 px Si py tLik 0 px 0 V b r d/ = f v (10.40) 0 py 2 0 0 px Sk py
Note that the effect of the distributed load in Figure 10.11 along the kiedge is represented by two equal nodal forces at i and k, with each force having x and y components. In a similar fashion, we can formulate the load matrix for a distributed load acting along
k
j
k
px
j
py
i
i
y
x
M10_MOAV4303_04_GE_C10.INDD 585
Figure 10.11 A distributed load acting over the kiedge of a triangular element.
27/11/14 10:08 AM
586 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
other sides of the triangular element. Evaluation of the integral in Eq. (10.39) along the ijedge and the jkedge results in
5 F6(e)
px py tLij px = f v 2 py 0 0
5 F6(e)
0 0 tLjk px = f v (10.41) 2 py px py
It is worth noting that, generally speaking, linear triangular elements do not offer as accurate results as do the higher order elements. The purpose of the above derivation was to demonstrate the general steps involved in obtaining the elemental stiffness and load matrices. Next, we will derive the stiffness matrix for a quadrilateral element using isoparametric formulation. 10.3 Isoparametric Formulation: Using a Quadrilateral Element As we discussed in Chapters 5 and 7, when we use a single set of parameters (a set of shape functions) to define the unknown variables u, v, T, and so on, and to express the position of any point within the element, we are using isoparametric formulation. An element expressed in such manner is called an isoparametric element. We will now turn our attention to the quadrilateral element previously shown as Figure 7.16 (repeated here for convenience). Using a quadrilateral element, we can express the displacement field within an element by Eq. (7.30):
u = SiUix + SjUjx + SmUmx + SnUnx
` = SiUiy + SjUjy + SmUmy + SnUny
(7.30)
h Uny
Umy Unx
(1, 1)
(1, 1)
Umx
m
n A(j, h)
j Uiy
(1, 1)
Ujy Uix
i
M10_MOAV4303_04_GE_C10.INDD 586
j (1, 1)
Ujx
Figure 7.16 A quadrilateral element used in formulating plane stress problems.
27/11/14 10:08 AM
Section 10.3 Isoparametric Formulation: Using a Quadrilateral Element 587
We can write the relations given by Eq. (7.30) in matrix form, given previously in Eq. (7.31):
u S b r = J i y 0
0 Si
Sj 0
0 Sj
0 Sm
Sm 0
Sn 0
Uix Uiy Ujx 0 U R h jy x Sn Umx Umy Unx Uny
(7.31)
Note that using isoparametric formulation, we can use the same shape functions to describe the position of any point within the element by the relationships in Eq. (7.32):
x = Si xi + Sj xj + Sm xm + Sn xn
(7.32)
y = Siyi + Sjyj + Smym + Snyn 0y The displacement field is related to the components of strains (exx = 0u 0x , eyy = 0y , 0u 0y and gxy = 0y + 0x ) and subsequently to the nodal displacements through shape functions. In Chapter 7, we also showed that using the Jacobian of the coordinate transformation, we can write the following, previously presented as Eq. (7.34):
[J]
e
0f(x, y) 0x 0j 0j d t = D 0f(x, y) 0x 0h 0h
0f(x, y) 0y 0j 0x Td t 0f(x, y) 0y 0h 0y
(7.34)
The relationship of Eq. (7.34) was also presented as the following, previously shown as Eq. (7.35):
M10_MOAV4303_04_GE_C10.INDD 587
0f(x, y) 0f(x, y) 0x 0j d t = [J]1 d t 0f(x, y) 0f(x, y) 0y 0h
(7.35)
27/11/14 10:08 AM
588 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
For a quadrilateral element, the J matrix can be evaluated using Eqs. (7.32) and (7.7): 0x 0j [J] = D 0x 0h
0y 0 [S x + Sj xj + Sm xm + Sn xn] 0j 0j i i T = D 0y 0 [S x + Sj xj + Sm xm + Sn xn] 0h 0h i i
0 [S y + Sjyj + Smym + Snyn] 0j i i T 0 [S y + Sjyj + Smym + Snyn] 0h i i (10.42)
[J] =
1 [  (1  h)xi + (1  h)xj + (1 + h)xm  (1 + h)xn] J 4 [  (1  j)xi  (1 + j)xj + (1 + j)xm + (1  j)xn] [  (1  h)yi + (1  h)yj + (1 + h)ym  (1 + h)yn] J R = J 11 [  (1  j)yi  (1 + j)yj + (1 + j)ym + (1  j)yn] J21
J12 R (10.43) J22
Also recall that the inverse of a twodimensional square matrix is given by [J]1 =
J11J22
1 J J 22  J12J21 J21
J12 1 J R = J 22 J11 det J J21
J12 R (10.44) J11
We can now proceed with the formulation of the stiffness matrix. The strain energy of an element is Λ(e) =
1 1 5 E 6 T[v]5 E 6 dV = (te) 5 E 6 T[v]5 E 6 dA (10.45) 2L 2 L V A
where te is the thickness of the element. Recall the strain–displacement relationships in matrix form
exx 5 E6 = c eyy s = f gxy
0u 0x 0y 0y
v (10.46)
0u 0y + 0y 0x
Evaluating the derivatives, we obtain
0u 0x 1 J d t = J 22 0u det J J21 0y
0u J12 0j Rd t (10.47) J11 0u 0h
0y 0x 1 J d t = J 22 0y det J J21 0y
0y J12 0j Rd t (10.48) J11 0y 0h
and
M10_MOAV4303_04_GE_C10.INDD 588
27/11/14 10:08 AM
Section 10.3 Isoparametric Formulation: Using a Quadrilateral Element 589
Combining Eqs. (10.46), (10.47), and (10.48) into a single relationship, we have
0y 0u + 0y 0x
[A]
i
5 E6 = f
0u 0x 0y 0y
v =
J22 1 C 0 det J J21
J12 0 J11
0 J21 J22
0u 0j 0u 0 0h J11 S g 0y w (10.49) J12 0j 0y 0h
Note how we defined the [A] matrix, to be used later. Using Eq. (7.30), we can perform the following evaluation: [ D]
y
0u 0j 0u (1 0h 1 (1 g 0y w = D 4 0j 0y 0h
 h) 0 (1  h) 0 (1  j) 0 (1 + j) 0 (1 0 (1  h) 0 (1  h) 0 (1  j) 0 (1 + j)
+ h) 0 (1 + h) 0 + j) 0 (1  j) 0 T 0 (1 + h) 0 (1 + h) 0 (1 + j) 0 (1  j)
e
5 U6
Uix Uiy Ujx U h jy x ( 10.50) Umx Umy Unx Uny
We can express the relationship in Eq. (10.50) in a compact matrix form as 5 E 6 = [A][D]5 U6 (10.51)
Next, we need to transform the dA term (dA = dx dy) in the strain energy integral into a product of natural coordinates. This transformation is achieved in the following manner:
Λ(e)
M10_MOAV4303_04_GE_C10.INDD 589
e
dA 1 1 1 1 = (te) 5 E 6T[v]5 E 6 dA = (te) 5 E 6 T[v]5 E 6 det Jdjdh (10.52) 2 2 L L L 1 1 A
27/11/14 10:08 AM
590 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Substituting for the strain matrix {E} and the properties of the material matrix [v] into Eq. (10.52) and differentiating the strain energy of the element with respect to its nodal displacements, we find that the expression for the element stiffness matrix becomes: 1
[K](e) = te
1
L1 L1
[[A][D]]T[v][A][D] det Jdj dh (10.53)
Note that the resulting stiffness matrix is an 8 * 8 matrix. Furthermore, as discussed in Chapter 7, the integral of Eq. (10.53) is to be evaluated numerically, using the Gauss– Legendre formula. Example 10.2 A twodimensional triangular plane stress element made of steel, with modulus of elasticity E = 200 GPa and Poisson’s ratio y = 0.32, is shown in Figure 10.12. The element is 3 mm thick, and the coordinates of nodes i, j, and k are given in centimeters in Figure 10.12. Determine the stiffness and load matrices under the given conditions. The element stiffness matrix is [K](e) = V[B]T[v][B]
where
V = tA [B] =
[v] =
bi 1 C0 2A di E D 1  n2
0 di bi
0 dj bj
bj 0 dj
1 n
n 1
0
0
bk 0 dk
0 dk S bk
0 0 T 1  n 2
Y P = 500 N 505 j
(1.5, 1.0) k px = 12 MPa
(2.40, 1.65) i (2.25, 0.75) X
Figure 10.12 The loading and nodal coordinates for the element in Example 10.2.
M10_MOAV4303_04_GE_C10.INDD 590
27/11/14 10:08 AM
Section 10.3 Isoparametric Formulation: Using a Quadrilateral Element 591
Thus, bi = Yj  Yk = 1.65  1.0 = 0.65
di = Xk  Xj = 1.50  2.40 = 0.9
bj = Yk  Yi = 1.0  0.75 = 0.25
dj = Xi  Xk = 2.25  1.5 = 0.75
bk = Yi  Yj = 0.75  1.65 = 0.9 dk = Xj  Xi = 2.40  2.25 = 0.15 and 2A = Xi(Yj  Yk) + Xj(Yk  Yi) + Xk(Yi  Yj) 2A = 2.25(1.65  1.0) + 2.40(1.0  0.75) + 1.5(0.75  1.65) = 0.7125 Substituting appropriate values into the above matrices, we have 0.65 1 C 0 0.7125 0.9
0 0.9 0.65
0.25 0 0.75
0.65 0 1 0.25 [B]T = F 0.7125 0 0.9 0
0 0.9 0 0.75 0 0.15
0.9 0.65 0.75 V 0.25 0.15 0.9
[B] =
N 1 cm2 0.32 D 1  (0.32)2 0
200 * 105 [v] =
0.32 1 0
0 0.75 0.25
0.9 0 0.15
0 0.15 S 0.9
0 22281640 0 T = C 7130125 1  0.32 0 2
7130125 22281640 0
0 0 S 7575758
Carrying out the matrix operations results in the element stiffness matrix:
(e)
[K]
=
0.7125 (0.3) ¢ ≤ 2 (0.7125)2
0.65 0 0.25 F 0 0.9 0
0 0.9 0 0.75 0 0.15
0.9 0.65 22281640 0.75 V C 7130125 0.25 0 0.15 0.9 0.65 C 0 0.9
M10_MOAV4303_04_GE_C10.INDD 591
0 0.9 0.65
7130125 22281640 0
0.25 0 0.75
0 0.75 0.25
0 0 S 7575758
0.9 0 0.15
0 0.15 S 0.9
27/11/14 10:08 AM
592 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Simplifying, we obtain
[K](e)
3273759 1811146 314288 = F 372924 2959471 1438221
1811146 4473449 439769 2907167 1371376 1566282
314288 439769 1190309 580495 876020 1020265
372924 2907167 580495 2738296 953420 168871
2959471 1371376 876020 953420 3835491 417957
1438221 1566282 1020265 V (N/cm) 168871 417957 1397411
The load matrix due to the distributed load is
5 F6(e)
px 1200 142 py 0 0 (0.3) 2(2.25  1.5)2 + (0.75  1.0)2 tLik 0 0 0 = f v = f v = f v 2 0 2 0 0 px 1200 142 py 0 0
The load matrix due to the concentrated load is
5 F6(e)
0 0 0 0 0 0 Qjx 500 cos (50) 321 = f v = f v = f v Qjy 500 sin (50) 383 0 0 0 0 0 0
The complete load matrix for the element is
5 F6(e)
M10_MOAV4303_04_GE_C10.INDD 592
142 0 321 = f v(N) 383 142 0
27/11/14 10:08 AM
Section 10.4 Axisymmetric Formulation 593
10.4 Axisymmetric Formulation In this section we briefly discuss the formulation of a stiffness matrix using axisymmetric triangular elements. The steps that we are about to take are similar to the steps shown in Section 10.2, where we expressed the state of the stress at a point using a Cartesian coordinate system. However, as you saw in Chapter 9, Section 9.4, axisymmetric formulations require the use of a cylindrical coordinate system. Moreover, because geometry and loading are symmetrical about the zaxis, the state of stress and strain at a point are defined by only the following components:
[s]T = [srr
szz
trz
suu] (10.54)
[E]T = [err
ezz
grz
euu] (10.55)
The relationship between the strains and the displacements are given by
err =
0u 0r
ezz =
0w 0z
0u 0w + 0z 0r
grz =
euu =
u (10.56) r
See Figure 10.13 to help you visualize the deformations. The relationship between the stress and the strain is given by generalized Hooke’s law according to 1 srr n E(1  n) s 1  n d zz t = H trz (1 + n)(1  2n) 0 suu n 1  n
n 1  n
0
1
0
0
1  2n 2(1  n)
0
n 1  n
0
1
rdu szz
r du z
tzr
srr
r
0u u + 0z dz
0u u + 0r dr
dz
trz
du
err e X d zz t (10.57) grz euu
(r +u)du u
z
n 1  n n 1  n
dr
dz
w u dr
w+ r
0w dr 0r
Figure 10.13 The component of the stress and displacements in a cylindrical coordinate system.
M10_MOAV4303_04_GE_C10.INDD 593
27/11/14 10:08 AM
594 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
We can express Eq. (10.57) in a compact matrix form by 5 S 6 = [N]5 E 6 (10.58)
The next step involves writing the displacement variables u and w in terms of axisymmetric triangular shape functions. The nodal displacements of the axisymmetric triangular element are shown in Figure 10.14.
b
u S r = J i w 0
0 Si
Sj 0
0 Sj
Sk 0
Uir Uiz 0 U R f jr v (10.59) Sk Ujz Ukr Ukz
The expressions for the shape functions are given in Eq. (7.27). As a next step we need to evaluate the strains from the strain displacement relationships.
err =
0u 0 = [SiUir + SjUjr + SkUkr] (10.60) 0r 0r
ezz =
0w 0 = [S U + SjUjz + SkUkz] (10.61) 0z 0z i iz
grz =
0u 0w 0 0 + = [SiUir + SjUjr + SkUkr] + [SiUiz + SjUjz + SkUkz](10.62) 0z 0r 0z 0r
euu =
SiUir + SjUjr + SkUkr u = (10.63) r r
As we did in Section 10.2, after taking the derivatives of the components of the displacement fields we can express the relationship between the strain components and the nodal displacements in a compact matrix form by
{E} = [B]5U6 (10.64) Ukz k
Ukr
Uiz i
Ujz
Uir j
M10_MOAV4303_04_GE_C10.INDD 594
Ujr
Figure 10.14 The nodal displacements of an axisymmetric triangular element.
27/11/14 10:08 AM
Section 10.5 Basic Failure Theories 595
We then substitute for the strain matrix (in terms of displacements) in the strain energy of the axisymmetric element that eventually leads to the stiffness matrix, which is computed from [K](e) = 2p 1 [B]T[n][B]rdA (10.65)
Note that the resulting matrix is a 6 * 6 matrix. 10.5 Basic Failure Theories One of the goals of most structural solid analyses is to check for failure. The prediction of failure is quite complex in nature; consequently, many investigators have been studying this topic. This section presents a brief overview of some failure theories. For an indepth review of failure theories, you are encouraged to study a good text on the mechanics of materials or on machine design. (For a good example of such a text, see Shigley and Mischke, 1989.) Using ANSYS, you can calculate the distribution of the stress components sx, sy, and txy, and the principal stresses s1, and s2 within the material. But how would you decide whether or not the solid part you are analyzing will permanently deform or fail under the applied loading? You may recall from your previous study of the mechanics of materials that to compensate for what we do not know about the exact behavior of a material and/or to account for future loading for which we may have not accounted, but to which someone may subject the part, we introduce a Factor of Safety (F.S.), which is defined as
F.S. =
Pmax Pallowable
(10.66)
where Pmax is the load that can cause failure. For certain situations, it is also customary to define the factor of safety in terms of the ratio of maximum stress that causes failure to the allowable stresses if the applied loads are linearly related to the stresses. But how do we apply the knowledge of stress distributions in a material to predict failure? Let us begin by reviewing how the principal stresses and maximum shear stresses are computed. The inplane principal stresses at a point are determined from the values of sxx, syy, and txy at that point using the equation
s1,2 =
sx + sy 2
{
C¢
sx  sy 2
2
2 ≤ + txy (10.67)
The maximum inplane shear stress at the point is determined from the relationship
tmax =
C¢
sx  sy 2
2
2 ≤ + txy (10.68)
There are a number of failure criteria, including the maximumnormalstress theory, the maximumshearstress theory, and the distortionenergy theory. The distortionenergy theory, often called the von Mises–Hencky theory, is one of the most commonly used
M10_MOAV4303_04_GE_C10.INDD 595
27/11/14 10:08 AM
596 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
criteria to predict failure of ductile materials. This theory is used to define the start of yielding. For design purposes, the von Mises stress sv is calculated according to the equation sy = 2s21  s1s2 + s22 (10.69)
A safe design is one that keeps the von Mises stresses in the material below the yield strength of the material. The relationship among the von Mises stress, the yield strength, and the factor of safety is
sy =
SY (10.70) F.S.
where SY is the yield strength of the material, obtained from a tension test. Most brittle materials have a tendency to fail abruptly without any yielding. For a brittle material under planestress conditions, the maximumnormalstress theory states that the material will fail if any point within the material experiences principal stresses exceeding the ultimate normal strength of the material. This idea is represented by the equations s1 = Sultimate
s2 = Sultimate (10.71)
where Sultimate is the ultimate strength of the material, obtained from a tension test. The maximumnormalstress theory may not produce reasonable predictions for materials with different tension and compression properties; in such structures, consider using the Mohr failure criteria instead. 10.6 Examples Using Ansys ANSYS offers a number of elements that can be used to model twodimensional solidstructural problems. Some of these elements were introduced in Chapter 7. The twodimensional solidstructural elements in ANSYS include PLANE182 and PLANE183. Plane182 is a fournode element used for 2D modeling of solid structures. The element can be used as either a plane element (plane stress, plane strain or generalized plane strain) or an axisymmetric element. Plane183 can be used as a sixnode (triangular) or eightnode (quadrilateral) structural solid element. It has quadratic displacement behavior and is well suited for modeling irregular shapes. The element is defined by either six or eight nodes having two degrees of freedom at each node: translations in the nodal x and y directions. The element may be used as a plane element (plane stress, plane strain and generalized plane strain) or as an axisymmetric element. The KEYOPT(1) is set to 1 for triangular shaped element. As the theory in Section 10.1 suggested, because of similarities between the governing equation of torsional problems and heat transfer, in addition to the elements listed above, you can use thermal solid elements (e.g., PLANE35, a sixnode
M10_MOAV4303_04_GE_C10.INDD 596
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 597
triangular element; PLANE55, a fournode quadrilateral element; or PLANE77, an eightnode quadrilateral element) to model torsional problems. However, when using the solid thermal elements, make sure that the appropriate values are supplied to the property fields and the boundary conditions. Example 10.1 (revisited) demonstrates this point. Example 10.2 Revisited Consider Example 10.1, which deals with the torsion of a steel bar (G = 11 * 103 ksi) having a rectangular cross section, as shown in the accompanying figure. Assuming that u = 0.0005 rad/in, and using ANSYS, we are interested in determining the location and magnitude of the maximum shear stress. We will then compare the solution generated with ANSYS to the exact solution for a straight rectangular bar, as discussed in Section 10.1. Steel bar G = 11 * 103 ksi P
0.5
in 1 in
1 in
P
Enter the ANSYS program by using the Launcher. Type Torsion (or a file name of your choice) in the Jobname entry field of the dialog box. Pick Run to start the GUI. Create a title for the problem. utility menu: File S Change Title . . .
M10_MOAV4303_04_GE_C10.INDD 597
27/11/14 10:08 AM
598 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
main menu: Preprocessor S Element Type S Add/Edit/Delete
main menu: Preprocessor S Material Props S Material Models S Thermal S Conductivity S Isotropic
M10_MOAV4303_04_GE_C10.INDD 598
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 599
ANSYS Toolbar: SAVE_DB main menu: Preprocessor S Modeling S Create S Areas S Rectangle S by 2 Corners
M10_MOAV4303_04_GE_C10.INDD 599
27/11/14 10:08 AM
600 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
main menu: Preprocessor S Meshing S Size Cntrls S
Smart Size S Basic
main menu: Preprocessor S Meshing S Mesh S Areas S Free Pick All main menu: Solution S Define Loads S Apply S Thermal S Temperature S On Line Pick the edges of the rectangle and apply a zero constant temperature boundary condition on all four edges.
main menu: Solution S Define Loads S Apply S Thermal S Heat Generat S On Area
M10_MOAV4303_04_GE_C10.INDD 600
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 601
Pick the area of the rectangle and apply a constant value of 2Gu = 11000.
main menu: Solution S Solve S Current LS OK Close (the solution is done!) window. Close (the /STAT Command) window. main menu: General Postproc S Plot Results S Contour Plot S Nodal Solu
M10_MOAV4303_04_GE_C10.INDD 601
27/11/14 10:08 AM
602 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
The solution is then given in the following figure:
By calculating the aspect ratio of the cross section
w 1.0 in = = 2.0 and consulting h 0.5 in
Table 10.1, we find that c1 = 0.246 and c2 = 0.229. Substituting for G, w, h, u, c1, and c2 in Eqs. (10.3) and (10.4), we get u = tmax =
T(1 in) TL = 0.0005 rad/in = 1 T = 157.5 lb.in 3 c2Gwh 0.229(11 * 106 lb/in2)(1 in)(0.5 in)3 T 157.5 lb.in = = 2560 lb/in2 2 c1wh 0.246(1 in)(0.5 in)2
When comparing 2560 lb/in2 to the FEA results of 2558 lb/in2, you see that we could have saved lots of time by calculating the maximum shear stress using the analytical solution and avoided generating a finite element model.
M10_MOAV4303_04_GE_C10.INDD 602
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 603
Example 10.3 The bicycle wrench shown in Figure 10.15 is made of steel with a modulus of elasticity N E = 200 GPa a200 * 105 b and a Poisson’s ratio n = 0.3. The wrench is 3 mm (0.3 cm) cm2 thick. Determine the von Mises stresses under the given distributed load and boundary conditions. R = 1.25 cm
The sides of the middle hexagon are 9 mm long. 1.5 cm
3 cm
1 cm
The sides of the corner hexagons are 7 mm long.
1 cm
88 N/cm
Fixed all the way around this hexagon. Figure 10.15 A schematic for the bicycle wrench in Example 10.3.
The following steps demonstrate how to (1) create the geometry of the problem, (2) choose the appropriate element type, (3) apply boundary conditions, and (4) obtain nodal results: Enter the ANSYS program by using the Launcher. Type Bikewh (or a file name of your choice) in the Jobname entry field of the dialog box. Pick Run to start the GUI. Create a title for the problem. This title will appear on ANSYS display windows to provide a simple way to identify the displays. So, issue the command utility menu: File S Change Title . . .
M10_MOAV4303_04_GE_C10.INDD 603
27/11/14 10:08 AM
604 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Define the element type and material properties with the following commands: main menu: Preprocessor S Element Type S Add/Edit/Delete
M10_MOAV4303_04_GE_C10.INDD 604
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 605
Close Assign the thickness of the wrench with the following commands: main menu: Preprocessor S Real Constants S Add/Edit/Delete
M10_MOAV4303_04_GE_C10.INDD 605
27/11/14 10:08 AM
606 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Assign the modulus of elasticity and Poisson’s ratio values by using the following commands: main menu: Preprocessor S Material Props S Material Models S Structural S Linear S Elastic S Isotropic
M10_MOAV4303_04_GE_C10.INDD 606
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 607
ANSYS Toolbar: SAVE_DB Set up the graphics area (i.e., workplane, zoom, etc.) with the following commands: utility menu: Workplane S Wp Settings . . .
Toggle on the workplane by using the command utility menu: Workplane S Display Working Plane Bring the workplane to view by using the command utility menu: PlotCtrls S Pan, Zoom, Rotate . . . Click on the small circle until you bring the work plane to view. Then, create the geometry with the following commands: main menu: Preprocessor S Modeling S Create S Areas S Rectangles S By 2 Corners On the workplane, create the two rectangles: Use the mouse buttons as shown below, or type the values in the appropriate fields. [WP = 2.25, 0.5] Expand the rubber band up 1.5 and right 3.0] or
M10_MOAV4303_04_GE_C10.INDD 607
27/11/14 10:08 AM
608 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
[WP = 7.25, 0.5] [Expand the rubber band up 1.5 and right 3.0] OK Create the circles with the following commands: main menu: Preprocessor S Modeling S Create S Areas S Circle S Solid Circle [WP = 1.25,1.25] [Expand the rubber band to a radius of 1.25] or
M10_MOAV4303_04_GE_C10.INDD 608
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 609
[WP = 6.25,1.25] [Expand the rubber band to a radius of 1.25] [WP = 11.25, 1.25] [Expand the rubber band to a radius of 1.25] OK Add the areas together with the commands: main menu: Preprocessor S Modeling S Operate S Booleans S Add S Areas Click on the Pick All button, and then create the hexagons. First, change the Snap Incr in the WP Settings dialog box to 0.1 with the command
M10_MOAV4303_04_GE_C10.INDD 609
27/11/14 10:08 AM
610 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
utility menu: PlotCtrls S Pan, Zoom, Rotate . . . Click on the Box Zoom, and put a box zoom around the left circle, then using the following commands create the hexagon: main menu: Preprocessor S Modeling S Create S Areas S Polygon S Hexagon Use the mouse buttons as shown below, or type the values in the appropriate fields: [1.25, 1.25] [Expand the hexagon to WP Rad = 0.7, Ang = 120] or
M10_MOAV4303_04_GE_C10.INDD 610
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 611
Then, issue the command utility menu: PlotCtrls S Pan, Zoom, Rotate . . . Click on the Fit button. Then, click on the Box Zoom, and put a box zoom around the center circle. Use the mouse buttons as shown below, or type the values in the appropriate fields: [6.25, 1.25] [Expand the hexagon to WP Rad = 0.9, Ang = 120] or
M10_MOAV4303_04_GE_C10.INDD 611
27/11/14 10:08 AM
612 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
utility menu: PlotCtrls S Pan, Zoom, Rotate . . . Click on the Fit button. Then, click on the Box Zoom, and put a box zoom around the rightend circle. Use the mouse buttons as shown below, or type the values in appropriate fields: [11.25, 1.25] [Expand the hexagon to WP Rad = 0.7, Ang = 120] or
ANSYS Toolbar: SAVE_DB Subtract the areas of the hexagons to create the driver holes: main menu: Preprocessor S Modeling S Operate S Booleans S Subtract S Areas [Pick the solid area of the wrench] [Apply anywhere in the ANSYS graphics area]
M10_MOAV4303_04_GE_C10.INDD 612
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 613
[Pick the left hexagon area]
[Pick the center hexagon area]
[Pick the right hexagon area]
[Apply anywhere in the ANSYS graphics area] OK Now you can toggle off the workplane grids with the following command: utility menu: Workplane S Display Working Plane ANSYS Toolbar: SAVE_DB You are now ready to mesh the area of the bracket to create elements and nodes. So, issue the following commands: main menu: Preprocessor S Meshing S Size Cntrls S Manual Size S Global S Size
ANSYS Toolbar: SAVE_DB main menu: Preprocessor S Meshing S Mesh S Areas S Free
M10_MOAV4303_04_GE_C10.INDD 613
27/11/14 10:08 AM
614 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Click on the Pick All button. OK Apply the boundary conditions and the load: main menu: Solution S Define Loads S Apply S Structural S Displacements S On Keypoints Pick the six corner keypoints of the left hexagon: OK
main menu: Solution S Define Loads S Apply S Structural S Pressure S On Lines Pick the appropriate horizontal line, as shown in the problem statement: OK
M10_MOAV4303_04_GE_C10.INDD 614
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 615
Solve the problem: main menu: Solution S Solve S Current LS OK Close (the solution is done!) window. Close (the STAT Command) window. Begin the postprocessing phase and plot the deformed shape with the following commands: main menu: General Postproc S Plot Results S Deformed Shape
M10_MOAV4303_04_GE_C10.INDD 615
27/11/14 10:08 AM
616 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Plot the von Mises stresses with the following commands: main menu: General Postproc S Plot Results S Contour Plot S Nodal Solu
M10_MOAV4303_04_GE_C10.INDD 616
27/11/14 10:08 AM
Section 10.6 Examples Using Ansys 617
M10_MOAV4303_04_GE_C10.INDD 617
27/11/14 10:08 AM
618 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
Exit ANSYS and save everything: ANSYS Toolbar: QUIT
10.7 Verification of Results Now we turn our attention to Example 10.3. There are several ways to check the validity of the results of this problem. You can print the reaction forces and check the value of their sum against the applied force. Are statics equilibrium conditions satisfied? Using the path operations of ANSYS, you can also cut an arbitrary section through the wrench and visually assess the x and ycomponents of the local stresses and shear stresses along the section. You can integrate the stress information along the path to obtain the internal forces and compare their values to the applied force. Are statics equilibrium conditions satisfied? These questions are left to you to confirm.
Summary At this point you should 1. know that it is wise to use simple analytical solutions rather than finite element modeling for a simple problem whenever appropriate. Use finite element modeling only when it is necessary to do so. Simple analytical solutions are particularly appropriate when you are solving basic torsional problems. 2. know that the stiffness matrix for torsional problems is similar to the conductance matrix obtained for twodimensional conduction problems. The stiffness matrix and the load matrix using a rectangular element are
[K](e)
M10_MOAV4303_04_GE_C10.INDD 618
2 w 2 = D 6/  1 1
2 2 1 1
1 1 2 2
1 2 1 / 1 T + D 2 6w  1 2 2
1 2 2 1
1 2 2 1
2 1 T 1 2
27/11/14 10:08 AM
Chapter 10 Summary 619
5 F6(e)
1 2GuA 1 = d t 4 1 1
For triangular elements, the stiffness and load matrices are, respectively
(e)
b2i 1 = C bibj 4A bibk
(e)
1 2GuA = c1s 3 1
[K]
5 F6
bibj b2j bjbk
bibk d 2i 1 bjbk S + Cdd 4A i j b2k di dk
di dj d 2j dj dk
di dk dj dk S d 2k
3. know that the stiffness matrix for a planestress triangular element is [K](e) = V[B]T[N][B] where V = tA bi 1 [B] = C0 2A di
0 di bi
bj 0 dj
0 dj bj
bk 0 dk
0 dk S bk
1 E n [v] = D 1  n2 0
n 1 0
0 0 T 1  n 2
and bi = Yj  Yk
di = Xk  Xj
bj = Yk  Yi
dj = Xi  Xk
bk = Yi  Yj
dk = Xj  Xi
2A = Xi(Yj  Yk) + Xj(Yk  Yi) + Xk(Yi  Yj) 4. know that the load matrix due to a distributed load along the element’s edges is
5 F6(e)
M10_MOAV4303_04_GE_C10.INDD 619
px py tLij px = f v 2 py 0 0
5 F6(e)
0 0 tLjk px = f v 2 py px py
5 F6(e)
px py tLik 0 = f v 2 0 px py
27/11/14 10:08 AM
620 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
5. understand how an element’s stiffness matrix is obtained through the isopara metric formulation. 6. understand how an element’s stiffness matrix is obtained through axisymmetric formulation. References ANSYS User’s Manual: Procedures, Vol. I, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Commands, Vol. II, Swanson Analysis Systems, Inc. ANSYS User’s Manual: Elements, Vol. III, Swanson Analysis Systems, Inc. Beer, P., and Johnston, E. R., Mechanics of Materials, 2nd ed., New York, McGrawHill, 1992. Fung, Y. C., Foundations of Solid Mechanics, Englewood Cliffs, NJ, PrenticeHall, 1965. Hibbeler, R. C., Mechanics of Materials, 2nd ed., New York, Macmillan, 1994. Segrlind, L., Applied Finite Element Analysis, 2nd ed., New York, John Wiley and Sons, 1984. Shigley, J. E., and Mischke, C. R., Mechanical Engineering Design, 5th ed., New York, McGrawHill, 1989. Timoshenko, S. P., and Goodier J. N., Theory of Elasticity, 3rd ed., New York, McGrawHill, 1970.
Problems 1. Using ANSYS, verify the stressconcentration chart for a flat bar with a circular hole under axial loading. Refer to a textbook on the mechanics of materials or textbook on machine design for the appropriate chart. Recall that the stressconcentration factor k is defined as k =
smax savg
and for this case, its value changes from approximately 3.0 to 2.0, depending on the size of the hole. Use ANSYS’s selection options to list the value of smax at point A or B.
A P
P
B
2. Consider one of the many steel brackets (E = 29 * 106 lb/in2, n = 0.3) used to support
bookshelves. The thickness of the bracket is 1/8 in. The dimensions of the bracket are shown in the accompanying figure. The bracket is loaded uniformly along its top surface, and it is fixed along its left edge. Under the given loading and the constraints, plot the deformed shape; also, determine the von Mises stresses in the bracket.
M10_MOAV4303_04_GE_C10.INDD 620
27/11/14 10:08 AM
Chapter 10 Problems 621 12 in
R 0.25
6 in
2 in
6 in Fillet radius 0.5 in
1 in
15 lb/in
3. A [email protected]@thick plate supports a load of 200 lb, as shown in the accompanying figure. The plate is
made of steel, with E = 29 * 106 lb/in2 and n = 0.3. Using ANSYS, determine the principal stresses in the plate. When modeling, distribute the load over part of the bottom portion of the hole.
M10_MOAV4303_04_GE_C10.INDD 621
27/11/14 10:08 AM
622 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
5.0 in
Thickness = 1/8 in
5.0 in
Radius = 0.5 in
Radius = 1.0 in 1.0 in 100 lb
4. Elements (1) and (2) are subjected to the distributed loads shown in the accompanying figure. Replace the distributed loads by equivalent loads at nodes 3, 4, and 5. y
200 100 N/cm
3
4
(1)
2.0 cm
1
5
(x = 2.5, y = 2.0)
(2)
2
x
1.5 cm
M10_MOAV4303_04_GE_C10.INDD 622
27/11/14 10:08 AM
Chapter 10 Problems 623
5. Using a steel sample similar to the one shown in the accompanying figure, perform a numerical tension test over the elastic region of the material. Plot the stress–strain diagram over the elastic region. Use ANSYS’s selection options to list the stress and strain values at the midsection of the steel sample. 13/16 in
1/8 in 5 in midsection 9/16 in
1 in
6. Example 1.4 (revisited). A steel plate is subjected to an axial load, as shown in the accompa
nying figure. The plate is 1/16 in thick, and it has a modulus of elasticity E = 29 * 106 lb/in2. Recall that we approximated the deflections and average stresses along the plate using the concept of onedimensional direct formulation. Using ANSYS, determine the deflection and the x and ycomponents of the stress distributions in the plate. Also, determine the location of the maximumstressconcentration regions. Plot the variation of the xcomponent of the stress at sections A–A, B–B, and C–C. Compare the results of the directformulation model to the results obtained from ANSYS. Furthermore, recall that for the given problem, it was mentioned that the way in which you apply the external load to your finite element model will influence the stressdistribution results. Experiment with applying the load over an increasingly large loadcontact surface area. Discuss your results. C
B
A
2" 1" — 8 B
1"
800 lb 2"
C 1"
M10_MOAV4303_04_GE_C10.INDD 623
A 4"
2"
27/11/14 10:08 AM
624 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
7. Consider a plate with a variable cross section supporting a load of 1500 lb, as shown in the accompanying figure. Using ANSYS, determine the deflection and the x and ycomponents of the stress distribution in the plate. The plate is made of a material with a modulus of elasticity E = 10.6 * 103 ksi. In Problem 24 of Chapter 1, you were asked to analyze this problem using simple direct formulation. Compare the results of your directformulation model to the results obtained from ANSYS. Experiment with applying the load over an increasingly large loadcontact surface area. Discuss your results. 4 in
Thickness = 0.125 in 2.5 in
1" — 8
10 in
1 in — 2
5 in
2 in
1500 lb
M10_MOAV4303_04_GE_C10.INDD 624
27/11/14 10:08 AM
Chapter 10 Problems 625
8. A thin steel plate with the profile given in the accompanying figure is subjected to an axial load. Using ANSYS, determine the deflection and the x and ycomponents of the stress distributions in the plate. The plate has a thickness of 0.125 in and a modulus of elasticity of E = 28 * 103 ksi. In Problem 4 of Chapter 1, you were asked to analyze this problem using simple direct formulation. Compare the results of your directformulation model to the results obtained from ANSYS. Experiment with applying the load over an increasingly larger loadcontact surface area. Discuss your results. 4 in
Holes equally spaced 2 in
1 in — 8 1 in — 2
12 in
1 in — 2
1 in — 2
2 in
500 lb
M10_MOAV4303_04_GE_C10.INDD 625
27/11/14 10:08 AM
626 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
9. Consider the torsion of a steel bar (G = 11 * 103 ksi and n = 0.3) having an equilateral
triangular cross section, as shown in the accompanying figure. Assuming that u = 0.005 rad/in and using ANSYS, determine the location(s) and magnitude of the maximum shear stress. Compare the solution generated with ANSYS to the exact solution obtained from the equation GLu tmax = # . 2 31
L = 1.2 in
10. Consider the torsion of a steel wideflange member (W4 * 13 and G = 11 * 103 ksi)
with dimensions shown in the accompanying figure. Assuming u = 0.0005 rad/in and using ANSYS, plot the shear stress distributions. Could you have solved this problem using the thinwall member assumption and, thus, avoid resorting to a finite element model? 0.345 in
4.16 in 0.28 in
4.06 in
11. Consider the torsion of a steel bar (G = 11 * 103 ksi) having a square cross section, as shown
in the accompanying figure. Assuming that u = 0.0005 rad/in and using ANSYS, determine the location(s) and magnitude of the maximum shear stress. Compare the solution generated
with ANSYS to the exact solution obtained from the equation tmax =
M10_MOAV4303_04_GE_C10.INDD 626
Ghu . 1.6
27/11/14 10:08 AM
Chapter 10 Problems 627
h = 1.2 in
h
12. Consider the torsion of a steel bar (G = 11 * 103 ksi) having a polygon cross section, as
shown in the accompanying figure. Assuming that u = 0.0005 rad/in and using ANSYS, determine the location(s) and magnitude of the maximum shear stress. Compare the solution GLu
generated with ANSYS to the exact solution obtained from the equation tmax = 0.9 .
L = 1.2 in
13. Consider the torsion of a steel bar (G = 11 * 103 ksi) having an elliptical cross section,
as shown in the accompanying figure. Assuming that u = 0.0005 rad/in and using ANSYS, determine the location(s) and magnitude of the maximum shear stress. Compare the solution
2 generated with ANSYS to the exact solution obtained from the equation tmax = Gbh u .
b2 + h2
h = 1 in
b = 0.5 in
14. Consider the torsion of a steel bar (G = 11 * 103 ksi) having a hollow, circular cross section, as shown in the accompanying figure. Assuming that u = 0.0005 rad/in and using ANSYS, determine the location(s) and magnitude of the maximum shear stress. Compare the solution GDu
generated with ANSYS to the exact solution obtained from the equation tmax = 2 . D = 1.5 in
M10_MOAV4303_04_GE_C10.INDD 627
27/11/14 10:08 AM
628 Chapter 10 Analysis of TwoDimensional Solid Mechanics Problems
15. Design Project The purpose of this project is twofold: (1) to provide a basis for the application of soliddesign principles using finite element methods and (2) to foster competitiveness among students. Each student is to design and construct a structural model from a 38 * 6 * 6 in sheet of plexiglas material that adheres to the specifications and rules given later in this problem and that is capable of competing in three areas: (1) maximum failure load per model weight, (2) predication of failure load using ANSYS, and (3) workmanship. A sketch of a possible model is shown in the accompanying figure. Each end of the model will have a diameter hole (eye) of d 7 1/2″ drilled through it perpendicular to the axis of loading, for which pins can be inserted and the model loaded in tension. The dimension a must also be set such that a 7 1″. The distance between the eyes will be / 7 2″. The maximum thickness of the member in the region of the eye will be t 6 3/8″. This requirement will ensure that the model fits into the loading attachment. A dimension of b 6 1″ from the center of the eyes to the outer edge in the direction of loading must be maintained so that the loading attachment can be used. The maximum width is limited to w 6 6″, and the maximum height is limited to h 6 6″. Any configuration may be used. Two sheets of 38 * 6 * 6 in Plexiglas) will be provided. You can use one sheet to experiment and one sheet for your final design. Write a brief report discussing the evolution of your final design. P
t
Holes parallel to and at an angle of 905 to the axis of loading
Section A–A
d b
A
A /
h
b a
a
P w
M10_MOAV4303_04_GE_C10.INDD 628
27/11/14 10:08 AM
C h a p t e r
1 1
Dynamic Problems
The main objective of this chapter is to introduce analysis of dynamic systems. A dynamic system is defined as a system that has mass and components, or parts, that are capable of relative motion. Examples of dynamic systems include structures such as buildings, bridges, water towers, planes, and machine components. In most engineering applications mechanical vibration is undesirable. However, there are systems, such as shakers, mixers, and vibrators, that are intentionally designed to vibrate. Before discussing finite element formulations of dynamic problems, we review dynamics of particles and rigid bodies, and examine basic concepts dealing with vibration of mechanical and structural systems. A good understanding of the fundamental concepts of dynamics and vibration is necessary for an accurate finite element modeling of actual physical situations. After we establish the basic foundation, we consider finite element formulation of an axial member, a beam, and frame elements. The main topics discussed in Chapter 11 include the following: 11.1 Review of Dynamics 11.2 Review of Vibration of Mechanical and Structural Systems 11.3 Lagrange’s Equations 11.4 Finite Element Formulation of Axial Members 11.5 Finite Element Formulation of Beams and Frames 11.6 Examples Using ANSYS 11.1 Review of Dynamics The subject of dynamics is traditionally divided into two broad areas: kinematics and kinetics. Kinematics deals with space and time relationships—in other words, the geometric aspect of motion. Kinematics is the study of variables such as distance traveled by an object, its speed, and its acceleration. The fundamental dimensions associated with all of these variables are length and time. When studying kinematics, we are not concerned with causes of motion but focus instead on the motion itself. Study of kinetics, on the
M11_MOAV4303_04_GE_C11.INDD 629
27/11/14 10:12 AM
630 Chapter 11 Dynamic Problems at time t Particle
O
at time t + ∆t
∆x
x
Figure 11.1 Rectilinear motion of a particle.
other hand, deals with the relationship between forces and moments, and the resulting motion. Kinematics of a Particle Let us first define a particle and explain when it is appropriate to model a problem using a particle. For an object to be considered as a particle, all forces acting on the object must act at the same point and create no rotation. Moreover, an object is modeled as a particle when its size does not play a significant role in the way it behaves. In general, the motion of a particle is described by its position, instantaneous velocity, and the instantaneous acceleration. The motion of an object along a straight path is called rectilinear motion, which is the simplest form of motion. The kinematical relationships for an object moving along a straight line is given by Eqs. (11.1) through (11.3). In these equations, x represents the position of the object, t is time, v is velocity, and a represents the acceleration of the particle. dx (11.1) dt dv a = (11.2) dt vdv = adx (11.3)
v =
The position and displacement of an object moving along a straight path are shown in Figure 11.1. Plane Curvilinear Motion When a particle moves along a curved path, its motion can be described using a number of coordinate systems. As shown in Figure 11.2, Y
P y
r j i x
M11_MOAV4303_04_GE_C11.INDD 630
X
Figure 11.2 Rectangular components of motion of a particle.
27/11/14 10:12 AM
Section 11.1 Review of Dynamics 631 u
u
using a rectangular coordinate system (x, y), we can describe the position r, velocity v, u and acceleration a of the particle with u
u
u
r = xi + yj
u
u
u
u
u
u
v = vx i + vy j
a = ax i + ay j
(11.4) dy and vy = (11.5) dt
dx where vx = dt where ax =
dvx dt
and ay =
dvy dt
(11.6)
In Eqs. (11.4) through (11.6) x and y are rectangular components of the position vector; vx, vy, ax, and ay are the Cartesian components of the velocity and acceleration vectors. Normal and Tangential Coordinates The plane motion of a particle can also be described using normal and tangential unit vectors, as shown in Figure 11.3. It isu imporu tant to note that unlike the previously employed Cartesianu unit uvectors i and j , whose directions are fixed in space, the directions of unit vectorsu et andu en change as they move along with the particle. The changes in the direction of et and en, as the particle moves from position 1 to position 2, are shown in Figure 11.3. The velocity and acceleration of u u a particle in terms of unit vectors et and en are u
u
(11.7)
v = vet u
u
2
v r
u
a = anen + at et where an =
and at =
dv (11.8) dt
Polar Coordinates A polar coordinate (or radial and transverse) system offers yet another way of describing motion of an object along a curved path. To locate the object, two pieces of information are used: a radial distance r in a direction specified by u u the unit vector er and the angular coordinate u in the eu direction, as shown in Figure 11.4. The position, velocity, and acceleration of the object are given by u
u
(11.9)
r = rer
v = vrer + vu eu where vr =
a = arer + at et where ar =
u
u
u
u
u
u
dr dt
and vu = r
du dt
(11.10)
d2 r du 2 d2 u dr du  r ¢ ≤ and au = r 2 + 2¢ ≤ ¢ ≤ (11.11) 2 dt dt dt dt dt
y
r en
P
en y
1 et
M11_MOAV4303_04_GE_C11.INDD 631
2
et
Figure 11.3 Normal and tangential components of motion of a particle.
27/11/14 10:12 AM
632 Chapter 11 Dynamic Problems er
eu
Y
P
r
u
X
Figure 11.4 Polar components of a curvilinear motion.
u
u
Note similar to normal and tangential unit vectors, the directions of eu and er change as they move along with the particle. Relative Motion The relationship between the position of two particles moving along different paths is shown in Figure 11.5. When studying Figure 11.5, note that the X,Y coordinate system is fixed, and an observer located at its origin u(point O) measures the u absolute motion of the particles A and B. Therefore, vectors rA and rB represent the absoluteu positions of particle A and B with respect to the observer. On the other hand, vector rB/A represents the position of particle B with respect to A. The relationship among these vectors is given by u
u
u
rB = rA + rB/A (11.12)
Taking the time derivative of Eq. (11.12), we get u
u
u
vB = vA + vB/A (11.13)
u
u
where vB and vA are absolute velocities of particle A and B measured by the observer u at O, and vB/A is the velocity of particle B relative to A (measured by someone moving along with A). By taking the time derivative of Eq. (11.13), we can obtain the relationship between the absolute accelerations of particle A and B and acceleration of particle B relative to A.
Y Pa th
of
A
/A
th Pa
of
B
rA
A
rB
rB
O
M11_MOAV4303_04_GE_C11.INDD 632
B X
Figure 11.5 Relationship between two particles moving along different paths.
27/11/14 10:12 AM
Section 11.1 Review of Dynamics 633
Kinetics of a Particle The equations of motion for a particle of constant mass m under the influence of forces F1, F2, F3, c, Fn is governed by Newton’s second law according to
a Fi = ma (11.14) n
u
u
i=1
We can describe the equations of motion using either a rectangular, normal and tangential, or polar coordinate system. Using a rectangular coordinate system, Eq. (11.14) is expressed by a Fx = max (11.15) a Fy = may (11.16)
In a normal and tangential coordinate system, Eq. (11.14) becomes
a Fn = man (11.17) a Ft = mat (11.18)
And in a polar coordinate system, we have
a Fr = mar (11.19) a Fu = mau (11.20)
where the respective components of accelerations are given in the previous sections. It is important to point out that the only way to account correctly for every force acting on a particle is to draw a freebody diagram. A freebody diagram represents the interaction of the particle with its surroundings. To draw a freebody diagram, as the name implies, you must free the body from its surroundings and show all the interaction of the body with its surroundings with forces of appropriate magnitudes and directions (see Figure 11.6). Newton’s second law of motion is a vectorial equation that relates forces acting on the object to its mass and acceleration. If the solution to a problem requires position and velocity information, then kinematical relationships are used to obtain such information from the knowledge of the object’s acceleration.
Sign
Fwind
Wind Sign
W = mg
Rx
A A
M11_MOAV4303_04_GE_C11.INDD 633
M
Ry
Figure 11.6 A freebody diagram of the object shown. Note the object shown can not be considered as a particle. The sign is merely used to demonstrate the concept of freebody diagram.
27/11/14 10:12 AM
634 Chapter 11 Dynamic Problems
Work and Energy Principle Whereas Newton’s second law is a vectorial equation, the work–energy principle is a scalar relationship. The work–energy principle relates the work done by forces over a distance to the object’s mass and speed. The work–energy principle becomes very useful for situations for which we are interested in determining the change in speed of an object due to applied forces. The work done by a force moving an object from position 1 to position 2, as shown in Figure 11.7, is defined by
W12 =
L
u
u
F.dr (11.21)
The work–energy principle simply states that net work done by all forces acting on an object will bring about a change in its kinetic energy according to L
u
u
F.dr =
1 1 mv22  mv21 (11.22) 2 2
1 1 mv22 and mv21 represent the kinetic energy of the object and correspond to 2 2 its 2 and 1 positions. When using the work–energy principle, you must pay close attention to two important things: (1) work is done only when a force undergoes a displacement; (2) if the tangential component of the force (the component that moves the object) and displacement have the same direction, then the work done is considered positive, and if they have opposite directions, work done is a negative quantity.
The terms
Linear Impulse and Linear Momentum For problems wherein the time history of forces is known (i.e., how forces act on an object over a time period), the impulse and momentum approach may be used to determine the resulting changes in the velocities of the object. Newton’s second law can be rearranged and integrated over time in the following manner: u u dmv a F = dt 1 a Fdt = dmv (11.23) u
u
Lt1
t2
a Fdt = mv2  mv1 (11.24) u
u
u
Y
Fn F
Ft
dr
2
1
X
M11_MOAV4303_04_GE_C11.INDD 634
Figure 11.7 Work done by a force moving an object.
27/11/14 10:12 AM
Section 11.1 Review of Dynamics 635
Equation (11.24) may be expressed in any of the previously mentioned coordinate systems. For example, using a Cartesian coordinate system, Eq. (11.24) becomes
Lt1
t2
Lt1
t2
a Fxdt = m(vx)2  m(vx)1 (11.25) a Fydt = m(vy)2  m(vy)1 (11.26)
Next, we will use Example 11.1 to demonstrate some of the ideas we have discussed so far. Example 11.1 A small sphere having mass m is released from position 1, as shown in Figure 11.8. We are interested in determining the velocity of the sphere as a function of u. We solve this problem using both Newton’s law and the work–energy principle. Using Newton’s second law and referring to the freebody diagram shown in the accompanying figure, applying Eq. (11.18), we have a Ft = mat mg cos u = mat 1 at = g cos u
Now, using Eq. (11.3), the kinematical relationship between velocity and acceleration, and noting that ds = Ldu, vdv = at ds L0
v
vdv =
L0
u
g cos uLdu
1
v = 22Lg sin u
We now solve the problem using the work–energy principle. The force due to the tension in the string does not perform any work because it acts normal to the path of the sphere. Relating the work done by the weight of the sphere W12 = mgL sin u to the change in its kinetic energy, we get 1 mgL sin u = mv2  0 1 v = 22Lg sin u 2 You realize that it is easier to solve this problem using the work–energy principle, because we were interested in determining the speed of the sphere and not its acceleration. L 1 u
s
T man u
mg
M11_MOAV4303_04_GE_C11.INDD 635
en et
mat
Figure 11.8 The sphere of Example 11.1.
27/11/14 10:12 AM
636 Chapter 11 Dynamic Problems
Kinematics of a Rigid Body In this section we discuss kinematics of a rigid body. Unlike a particle model, a rigid body is an object whose size affects its dynamic behavior and forces can act anywhere on the body. Moreover, as the name implies, a rigid body is considered rigid—no deformation due to application of forces. A rigid body is an idealization of an actual situation in which the magnitude of motion caused by forces and moments is much larger than internal displacements. Motion of a rigid body may be classified as pure translation, pure rotation, or a combination of translation and rotation called general plane motion. Translation of a Rigid Body When a rigid body undergoes a pure translational motion, all constituent particles move with the same velocity and acceleration. As shown in Figure 11.9, points A and B have the same velocity. Rotation of a Rigid Body about a Fixed Axis When a rigid body rotates about a fixed axis, the particles of the rigid body follow circular paths, as shown in Figure 11.10. The relationships between the velocity of a point A such as vA and its acceleration and the angular velocity and acceleration of the rigid body, v and a, are given by
(11.27)
vA = rAv
an = rA v2 =
at = rAa vA
A
B
v2A rA
= vAv (11.28) (11.29)
Path of A
A
vB = vA
Path of B
B
Figure 11.9 A pure translation of a rigid body.
yA
A
v
a rA
yA
an A
at
Figure 11.10 Pure rotation motion.
M11_MOAV4303_04_GE_C11.INDD 636
27/11/14 10:12 AM
Section 11.1 Review of Dynamics 637
In general, the velocity and the acceleration components of a point (such as A) located u u on a rigid body inu terms of the position vector rA, the angular velocity v, and the angular acceleration a are given by u
u
u
vA = v * rA
u
(11.30)
u
u
an = v * (v * rA) (11.31)
at = a * rA
u
u
(11.32)
General Plane Motion The motion of a rigid body undergoing simultaneous rotation and translation is commonly referred to as general plane motion. The relationship between the velocities of two points A and B at an instant is given by u
u
u
vA = vB + vA/B (11.33)
u
u
where vA and vB are the absolute velocities of points A and B and vA/B represents the velocity of point A relative to point B. The magnitude of vA/B is expressed by vA/B = rA/B v, and its direction is normal to the position vector rB/A, as shown in Figure 11.11. The acceleration of points A and B are related according to u
u
u
u
u
u
aA = aB + aA/B = aB + (aA/B)n + (aA/B)t (11.34)
General plane motion Pure translation
A
Pure rotation C
B
vB
=
B
rA/B
vB
A
vB
+
B
vA
A
A B
vA/B
vB vA = vB + vA/B vA/B
vA
Figure 11.11 The relationship between the velocities of two points belonging to a rigid body undergoing a general plane motion.
M11_MOAV4303_04_GE_C11.INDD 637
27/11/14 10:12 AM
638 Chapter 11 Dynamic Problems
where u
u
u
u
(aA/B)n = v * (v * rA/B) (11.35)
(aA/B)t = a * rA/B
u
u
u
(11.36)
and the magnitudes of the normal and tangential components of the acceleration of A relative to B are given by v2A/B = rA/B v2 (11.37) rA/B
(aA/B)n =
(aA/B)t = rA/B a
(11.38)
The directions of these components are shown in Figure 11.12. Kinetics of a Rigid Body The study of kinetics of a rigid body includes forces and moments that create the motion of the rigid body. Rectilinear Translation The rectilinear translation of a rigid body under the influence of forces F1, F2, F3, c, Fn is governed by Newton’s second law according to
a Fx = m(aG)x (11.39) a Fy = m(aG)y (11.40)
aB
aB B
B
aB =
(a
+
A
A
B
)n
B A/
r A/B
A aA (aA/B)t (aA/B)n aB
(aA/B)t
aA u
Figure 11.12 The directions of the normal and tangential components of aA/B.
M11_MOAV4303_04_GE_C11.INDD 638
27/11/14 10:12 AM
Section 11.1 Review of Dynamics 639 u
Although we have related the sum of forces to the acceleration of the mass center aG, it is important to realize that for a rigid body undergoing translation, all constituent particles have the same velocity and acceleration. Moreover, since the body is not rotating, the sum of the moments of the forces about mass center G must be zero.
⤽ MG = 0 (11.41) a However, if we were to take the sum of the moment about another point, such as O, then the sum of the moments about that point is not zero, because the inertia forces m(aG)x and m(aG)y create moments about that point. That is,
⤽ MO = m(aG)x d1  m(aG)y d2 (11.42) a The freebody and inertia diagrams for a rigid body undergoing pure translational motion are shown in Figure 11.13.
Rotation about a Fixed Axis The rotation of a rigid body is governed by the following equations: 2 a Fn = mrG/O v (11.43)
a Ft = mrG/O a (11.44) ⤽ M O = IO a (11.45) a
In Eq. (11.45), Io is the mass moment of inertia of the body about point O, as shown in Figure 11.14. Whereas mass provides a measure of resistance to translational motion, the mass moment of inertia represents the amount of the body’s inherent resistance to rotational motion. For a special case when the rigid body is rotating about its mass center G, the equations of motion become a Fx = 0 a Fy = 0
(11.46) (11.47)
⤽ MG = IG a (11.48) a
(maG)y
F1
O
O F2
G
=
d1
G
(maG)x
d2 F3 Figure 11.13 The kinetics of a rigid body under translational motion.
M11_MOAV4303_04_GE_C11.INDD 639
27/11/14 10:12 AM
640 Chapter 11 Dynamic Problems
2
en
v
O
F2
m
rG/O
rG
/O
F1 O
et =
G
G IGa mrG/Oa
F3 Figure 11.14 The rotational motion of a rigid body.
General Plane Motion For a rigid body that is translating and rotating simultaneously, the equations of motion become a Fx = m(aG)x (11.49) a Fy = m(aG)y (11.50)
⤽ MG = IG a (11.51) a or as shown in Figure 11.15, the sum of the moments about another point such as O must include the moments created by inertia forces. ⤽ MO = IG a + m(aG)xd1  m(aG)yd2 (11.52) a
Work–Energy Relations The work–energy principle relates the work done by forces and moments to the change in the kinetic energy of the rigid body as shown in
(maG)y
F1
O =
F2
d1
G
(maG)x
IGa d2 F3
Figure 11.15 General plane motion of a rigid body.
M11_MOAV4303_04_GE_C11.INDD 640
27/11/14 10:12 AM
F
Section 11.1 Review of Dynamics 641
F
1
F
1
1 2 2 IG v
M = F1 d
d
1 mv2 G 2
Figure 11.16 Work–energy principle for a rigid body.
Figure 11.16. The forces and moments acting on a rigid body do work according to
W12 =
L
u
u
F.dr (11.53)
Mdu (11.54) L where dr and du are the infinitesimal translation and rotational displacements. Because the rigid body can translate and rotate, its kinetic energy has two parts: a translational and a rotational part. The rigid body’s translational part of kinetic energy is given by W12 =
T =
1 mv2G (11.55) 2
The rotational kinetic energy of a rigid body about it mass center G is given by
T =
1 I v2 (11.56) 2 G
The total kinetic energy of a rigid body undergoing a general plane motion is then given by
T =
1 1 mv2G + IG v2 (11.57) 2 2
The rotational kinetic energy of a rigid body about an arbitrary point O is T =
1 I v2 2 O
Impulse and Momentum For problems in which time history of forces and moments are known, the impulse and momentum approach is used to determine the changes in the velocities of the rigid body. The linear impulse and momentum equations for a rigid body are
M11_MOAV4303_04_GE_C11.INDD 641
Lt1 Lt1
t2
t2
a Fdt = m(vG)2  m(vG)1 (11.58a) u
u
u
a MG dt = IG(v)2  IG(v)1 (11.58b)
27/11/14 10:12 AM
642 Chapter 11 Dynamic Problems
In Eq. (11.58b), a MG represents the sum of all moments about mass center G, and IG is the mass moment of inertia of the rigid body about its mass center G. Next, we use Example 11.2 to show how the concepts discussed previously are applied to formulate the equations of motion for a rigid body.
Example 11.2 In this example we derive the equations of motion for the system shown in the accompanying figure. When studying this example, keep in mind that the rod can rotate and translate. L/ 2
L/ 2 G
k2
k1
Freebody diagram Inertia diagram
L/ 2 L/ 2
k1(x  L u ) 2
mx u
G
..
k2(x + L u ) 2 IG u
x
..
G
The freebody diagram of the rod is shown above. This is a snapshot of the rod at a point in time as it oscillates. One way of arriving at the freebody diagram shown is by imagining that you pull the rod straight up and turn it counterclockwise and then let it go. To keep track of the motion of the rod, we will use x and x∙∙ to locate translation of the mass center and use u and u ∙∙ to measure the rotational motion. Note x is measured from the static equilibrium position and as a result the weight of the rod and the spring forces in static position cancel each other out. Applying Eqs. (11.50) and (11.51), we have ∙∙ a Fx = m(aG)x = mx L L k1 ¢ x u≤  k2 ¢ x + u≤ = mx∙∙ 2 2
and simplifying, we get the equation of translational motion mx∙∙ + (k1 + k2)x  (k1  k2)
M11_MOAV4303_04_GE_C11.INDD 642
L u = 0 2
27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 643
The equation for rotational component is obtained from + ∙∙ ⤻ a M G = IG a = IG u
k1 ¢ x 
L L L L u ≤  k2 ¢ x + u≤ = IGu ∙∙ 2 2 2 2
and simplifying we get the equation for rotational motion, which is given by IGu ∙∙  (k1  k2)
L L 2 x + (k1 + k2) ¢ ≤ u = 0 2 2
Now that we have reviewed dynamics of particles and rigid bodies, we are ready to review the basic definitions, fundamental principles, and governing equations in vibrations. 11.2 Review of Vibration of Mechanical and Structural Systems A dynamic system is defined as a system that has mass and components (parts) that are capable of relative motion. A dynamic system has the following properties: Because of the mass of the system and the changes in the velocity of the system, the kinetic energy of the system can increase or decrease with time. Elastic members of the system are capable of storing elastic energy. The materials making up the system have damping characteristics that could convert a portion of the work or energy input (into the system) into heat. Work or energy enters the system through excitation of support or direct application of forces. Examples of dynamic systems are given in Table 11.1. Degrees of Freedom The degree of freedom is defined as the number of spatial coordinates required to completely describe the motion of a system—in other words, the number of coordinates needed to keep track of all components or lumped masses making up the system. For example, the fourstory building shown in Table 11.1 requires four spatial coordinates to locate the temporal position of each floor mass. As another example we may use a model with three degrees of freedom to study the dynamic behavior of an airplane as shown in Table 11.1. Simple Harmonic Motion Consider the simple single degree of freedom system consisting of a linear spring and mass, as shown in Figure 11.17.
M11_MOAV4303_04_GE_C11.INDD 643
27/11/14 10:12 AM
644 Chapter 11 Dynamic Problems
Table 11.1 Examples of dynamic systems Actual System
Possible Dynamic Model
A water tower Water tower
m
x1
Stiffness and damping
A fourstory building Stiffness and damping
m4
m3 m2
m1
An airplane z2
m2
Stiffness and damping z1
m1
x4
x3 x2 x1
z3
m3
The freebody diagram for the static equilibrium is shown in Figure 11.17(b). As you can see, at static equilibrium, the weight of the mass is supported by the spring force, kdstatic = W. In the forthcoming analysis, k represents the stiffness of the spring in N/mm (or lb/in), and dstatic is the static deflection in mm (or in). To obtain the governing equation of motion, we displace the mass and then release it, as shown in Figure 11.17(c).
M11_MOAV4303_04_GE_C11.INDD 644
27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 645 [email protected] static k
m Static equilibrium
m
W = mg (b)
(a)
Static equilibrium y0
y
m
(c)
k( @static + y)
m
m
W
.. my (d)
Figure 11.17 A simple single degree of freedom system: (a) the system, (b) the freebody diagram of the system in static equilibrium, (c) displacing mass by y0, and (d) the freebody diagram for the oscillating mass.
We then apply Newton’s second law, which leads to
∙∙ a Fy = my (11.59) kdstatic  ky + W = my∙∙ (11.60)
We previously showed that W = kdstatic; therefore, Eq. (11.60) reduces to my∙∙ + ky = 0 (11.61)
Referring to Figure 11.17(c), you are reminded that y is measured from the static equilibrium position of the system. It is customary to write Eq. (11.61) in the following manner: y∙∙ + v2n y = 0 (11.62)
where
v2n =
k = undamped natural circular frequency of the system (rad/s) (11.63) m
In order to solve the governing differential equation of motion (Eq. 11.62), we need to first define the initial conditions. Since Eq. (11.62) is a secondorder differential equation, two initial conditions are required. Let’s assume that at time t = 0, we pull the mass down by a distance of y0 and then release it without giving the mass any initial velocity. Then, for time t = 0, we can write the following initial conditions: y = y0 (or written alternatively as y(0) = y0) (11.64)
and
M11_MOAV4303_04_GE_C11.INDD 645
y∙ = 0 (or y∙(0) = 0) (11.65)
27/11/14 10:12 AM
646 Chapter 11 Dynamic Problems
For this simple degree of freedom system, the general solution to the governing differential equation is then given by y(t) = c1 sin vn t + c2 cos vn t (11.66)
You should know from your studies in differential equation class how to obtain Eq. (11.66). Moreover, you should recall that the solution to a differential equation must satisfy the differential equation. In other words, if we were to substitute the solution, Eq. (11.66), back into the governing differential equation of motion, Eq. (11.62), the outcome must be zero. To demonstrate this point, we substitute the solution back into the governing differential equation: y
y ∙∙
g
i
( c1v2n sin vn t  c2v2n cos vn t) + v2n (c1 sin vn t + c2 cos vn t) = 0 0 = 0 Q.E.D.
Applying the initial condition y(0) = y0, we have y0 = c1 sin (0) + c2 cos (0) which leads to c2 = y0, and applying y∙(0) = 0 leads to y∙ = c1vn cos vn t  c2 vn sin vn t 0 = c1 vn cos(0)  c2vn sin(0) c1 = 0 After substituting for c1 and c2 in Eq. (11.66), the solution representing the location of the mass (from the static equilibrium position) as a function of time is given by
y(t) = y0 cos vn t (11.67) and the velocity and acceleration of the mass are given by dy = y0 vn sin vn t (11.68) dt d 2y y∙∙(t) = 2 = y0 v2n cos vn t (11.69) dt y∙(t) =
In order to look at the harmonic behavior of the system, we have plotted the position of the mass in Figure 11.18. Using Figure 11.18, we can now define period and frequency for a dynamic system. As shown in Figure 11.18, period T is defined as the time that it takes for the mass m to complete one cycle, typically measured in seconds. On the other hand, frequency f is defined as the number of cycles per second and is expressed in Hertz. The relationship between frequency and period is given by
f =
1 (11.70) T
This is a good place to point out the difference between the circular frequency v, which is expressed in radians per second (rad/s), and frequency f, which is expressed in cycles
M11_MOAV4303_04_GE_C11.INDD 646
27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 647
k Time
m Amplitude, y0
y
Period, T
y Figure 11.18 The behavior of a simple single degree of freedom system.
per second (hz). The relationship between v and f is given by
v = 2pf v¢
(11.71)
cycles radians 2p radians ≤ = ¢ ≤f¢ ≤ second cycle second
Example 11.3 Consider the simple degree of freedom system shown in the accompanying figure. We are interested in determining the position and the velocity of the 4 kg mass as a function of time. The spring has a stiffness value of 39.5 N/cm. To start the vibration, we will pull down the mass by 2 cm and then release it with no initial velocity.
k = 39.5 N/cm Mass y
The natural circular frequency is calculated from Eq. (11.63), vn =
k (39.5 N/cm)(100 cm/1m) = 31.42 rad/s = A Am 4 kg
and the frequency and period are
vn 31.42 = = 5 hz 2p 2p 1 1 T = = = 0.2 s f 5 f =
M11_MOAV4303_04_GE_C11.INDD 647
27/11/14 10:12 AM
648 Chapter 11 Dynamic Problems
It takes 0.2 seconds for the mass to complete one cycle. When designing for dynamic systems, the knowledge of magnitude of forces transmitted to the support, or the foundation, is important to ensure the integrity of the foundation. For Example 11.3, the magnitude of the force transmitted to the support is determined from R(t) = ky + W = ky0 cos vn t + W Substituting for W = mg = (4 kg)(9.81 m/s2) = 39.2 N, we get R(t) = (39.5)(2)cos(31.42t) + 39.2 The maximum reaction force occurs when the spring is stretched the maximum amount and whenever the value of cos(31.42t) becomes 1. Rmax = (39.5)(2) + 39.2 = 118.2 N Note that for the given system, the value of Rmax depends on the initial displacement. Also note that the support reaction should be equal to the weight of the system when y = 0. Forced Vibration of a Single Degree of Freedom System The vibration of a structural system may be induced by a number of sources such as wind, earthquake, or an unbalanced rotating machinery housed on a floor of a building. These sources of excitations may create forces that are sinusoidal, sudden, random, or that vary with time in a certain manner. We can learn a lot about the behavior of a system by considering basic excitation functions, such as a sinusoidal, step, or a ramp function. Moreover, to simplify presentation and since for most structures damping is relatively small, we will not consider it in our forthcoming presentations. Let us now consider a spring mass system subjected to a sinusoidal force, as shown in Figure 11.19. The motion of the system is governed by my∙∙ + ky = F0 sin vt (11.72)
Dividing both sides of Eq. (11.72) by mass m, y∙∙ +
and by letting v2n =
F0 k y = sin vt (11.73) m m
k , we have m y∙∙ + v2n y =
F0 sin vt (11.74) m
k
m y F (t ) = F0 sin vt
M11_MOAV4303_04_GE_C11.INDD 648
Figure 11.19 A spring mass system subjected to a harmonic forcing function.
27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 649
The general solution to Eq. (11.74) has two parts: a homogenous solution and a particular solution. As shown previously, the homogenous solution yh is given by yh(t) = A sin vn t + B cos vn t (11.75)
For the particular solution yp, we assume a function of the form
yp(t) = Y0 sin vt (11.76) and differentiating it to get
yp∙∙,
we get
yp∙ (t) = Y0 v cos vt (11.77)
yp∙∙(t) = Y0 v2sin vt Substituting for yp∙∙ and yp into Eq. (11.74) Y0 v2 sin vt + v2n Y0 sin vt =
F0 sin vt m
and solving for Y0, we have F0 m F0 F0 2 vn m k Y0 = = = (11.78) v2 + v2n v2 + v2n v 2 1 a b vn v2n Note that upon arriving at the final expression for Y0, we divided the numerator and the denominator by v2n, and we substituted for v2n = mk . The forced behavior of the spring mass system to the harmonic forcing function is then given by Forced Response
g
y(t) = A cos vn t + B sin vn t +
F0 k
e
Natural Response
v 2 1  ¢ ≤ vn
sin vt
(11.79)
It is important to point out that the natural response of the system will eventually die out because every system has some inherent damping. Consequently, in the upcoming discussions we will focus only on the forced response. To shed more light on the response of a single degree of freedom system to a sinusoidal force, using Eq. (11.78), we have plotted the ratio of the amplitude of the forced response Y0 to the static deflection F0 v k (that is caused if the force is applied statically) as a function of frequency ratio vn . The response is shown in Figure 11.20. It is clear from examining Figure 11.20 that as the frequency ratio approaches unity, the amplitude of forced oscillation becomes very large. This condition, known as resonance, is highly undesirable.
M11_MOAV4303_04_GE_C11.INDD 649
27/11/14 10:12 AM
650 Chapter 11 Dynamic Problems 4
3
2
F0 k
Y0
1
0
1
2
3
4
v vn
1
2
3 Figure 11.20 Amplitude of the forced behavior as a function of the frequency ratio.
4
Forced Vibration Caused by an Unbalanced Rotating Mass As we mentioned earlier, vibration of a system may be induced by an unbalanced rotating mass within a machine. In machines the vibration occurs when the center of mass of the rotating component does not coincide with its center of rotation. The vibration of an unbalanced machine may be modeled by a single degree of freedom system that is excited by a sinusoidal forcing function. The sinusoidal forcing function results from the vertical component of the normal acceleration of unbalanced mass, as shown in Figure 11.21. For these situations, the forcing function is represented by F(t) = m0e v2 sin vt (11.80)
where m0 is the unbalanced mass, e is eccentricity, and v is the angular speed of the rotating component. Comparing Eq. (11.80) to Eq. (11.72), we note that F0 = m0e v2 and substituting for F0 in Eq. (11.79), and considering only the forced response we get Y0
b
M11_MOAV4303_04_GE_C11.INDD 650
y(t) =
m0ev2 k v 2 1  ¢ ≤ vn
sin vt
(11.81)
27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 651
m0ev2 cos vt 2
ev m0
v
ev2 sin
m0
e
u
m0
vt
u =vt
k 2
k 2
Figure 11.21 An unbalanced rotating machine.
and rewriting the amplitude of the vibration Y0, by substituting k = mv2n, and regrouping, m0 ev2 v2 m0 ev2 mv2n v2n Y0 m k Y0 = = 1 (11.82) = m0 e v 2 v 2 v 2 1  a b 1  a b 1  a b vn vn vn
m0 ev2 From examining the numerator in Eq. (11.82), we note that by increasing the m mass of the system m, we can reduce the amplitude of vibration. You may have seen turbine or large pumps mounted on heavy concrete blocks to reduce the amplitude of unexpected vibration. Moreover, we can show that as v V 1 vn
1
Y0 m = 0 (11.83) m0 e
and v W 1 vn
1
Y0 m = 1 (11.84) m0 e
Y0 m v and have as a function of frequency ratio vn m0 e shown it in Figure 11.22. The behavior of the system as a function of frequency ratio is selfevident when examining Figure 11.22. The derivation of the response of a single degree of freedom mass spring system to a suddenly applied force, a ramp function, or a suddenly applied force that decays with time is left as an exercise. See Problems 4 and 5 at the end of this chapter.
Using Eq. (11.82), we have plotted
M11_MOAV4303_04_GE_C11.INDD 651
27/11/14 10:12 AM
652 Chapter 11 Dynamic Problems 4
3
2
Y0 m m0e
1
0
1
2
3
4
v vn
1 2 3 Y0m Figure 11.22 A plot of as a function m0e v of frequency ratio . vn
4
Forces Transmitted to Foundation As we mentioned earlier, when designing for dynamic systems, the knowledge of the forces transmitted to the support or the foundation is important to ensure the integrity of the foundation. The relationship between the vibration of the mass and the forces transmitted to the foundation is depicted in Figure 11.23. As shown in Figure 11.23, forces are transmitted to the foundation through the springs. The magnitude of these forces, which varies with time, is given by F0 F0 k F(t) = ky(t) = sin vt = sin vt (11.85) 2 v v 2 1  a b 1  a b vn vn k
However, for most engineering applications, we are interested in determining maximum magnitude of forces transmitted to the foundation, which occurs when sin vt has a value of 1 in Eq. (11.85), and is given by
M11_MOAV4303_04_GE_C11.INDD 652
Fmax =
F0 (11.86) v 2 1  a b vn
27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 653 F(t ) = F0 sin vt
F0 sin vt
Mass
Mass
k 2
ky
k 2
ky
Foundation
Foundation
Figure 11.23 Forces transmitted to a foundation.
In the field of vibration, it is customary to define a transmission ratio or transmissibility TR as the ratio of Fmax to the static magnitude F0: F0 v 2 1  a b vn Fmax 1 2 = 5 5 = 3 3 (11.87) TR = 2 F0 F0 v 2 1  a b vn
Figure 11.24 shows the relationship between transmissibility and the frequency ratio, as given by Eq. (11.87). It is clear from examining Figure 11.24, as the frequency ratio approaches unity, the transmissibility approaches very large values, which could have catastrophic consequences. Furthermore, Figure 11.24 shows that in order to keep transmissibility TR
TR 7 1
TR 6 1
1
1
√2
2
3
4
v vn
Figure 11.24 Transmissibility as a function of frequency ratio.
M11_MOAV4303_04_GE_C11.INDD 653
27/11/14 10:12 AM
654 Chapter 11 Dynamic Problems
low, machines should operate at frequencies that are much greater than their natural frequencies. Support Excitation In this section we review support excitation. A spring mass system that is being excited by its support is shown in Figure 11.25. Using Newton’s second law, we can formulate the governing equation of motion of the oscillating mass as follows: k(y1  y2) = my1∙∙ (11.88)
Note that the magnitude of the spring force depends on the relative position of the mass with respect to its support. Separating the excitation term from the response, we have my1∙∙ + ky1 = ky2 (11.89)
Dividing both sides of Eq. (11.89) by m, we get y1∙∙ + v2n y1 = v2n y2 = v2n Y2 sin vt (11.90)
We can use the solution we obtained earlier to the differential equation given by F0 Eq. (11.74). By comparing Eq. (11.90) to Eq. (11.74), and noting that v2nY2 = , the m solution to Eq. (11.90) then becomes
y1(t) =
v 1  ¢ ≤ vn
f
Amplitude of mass
v2nY2 v2n 2
sin vt =
Y2
v 2 1  ¢ ≤ vn
sin vt
(11.91)
To get a physical feeling for what the solution represents, let’s turn our attention to the amplitude of mass, as given by Eq. (11.91). If we replace the spring with a stiff rod and excite the support, as shown in Figure 11.26(a), we note that because of the rod’s large k .. my 1 y1 Mass k 2
Mass k 2
y2 = Y2 sin vt
Mass
k (y1  y2)
Support
Figure 11.25 Support excitation.
M11_MOAV4303_04_GE_C11.INDD 654
27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 655 Support
Support
Steel Rod (very large k)
Soft Spring
Mass
Mass
(a)
(b)
Figure 11.26 Experiments demonstrating the relationship between the amplitude of support excitation and the amplitude of the mass.
value and consequently large vn, the frequency ratio would be very small, vvn V 1, and consequently the amplitude of oscillation of mass would equal Y2. As you would expect, the mass will move with the same amplitude as the support. Now, if we replace the rod with a soft spring having a small k value (thus a small vn), the frequency ratio becomes v vn W 1. Substituting a large value for frequency ratio in Eq. (11.91) will show that the mass will vibrate with very small amplitude and will appear nearly stationary. Multiple Degrees of Freedom In the previous sections we considered the natural and forced behavior of a single degree of freedom system. Next, we will demonstrate some of the important characteristics of a multiple degrees of freedom system using a simple two degrees of freedom system. Consider the two degrees of freedom system shown in Figure 11.27. We are interested in determining the natural frequencies of the system shown. We begin by formulating the governing equations of motion for each mass. To write the equations of motion, we will initiate the free vibration by displacing the masses such that x2 7 x1. Using the freebody diagrams shown, the equations of motion are m1x1∙∙ + 2kx1  kx2 = 0 (11.92) m2x2∙∙  kx1 + 2kx2 = 0 (11.93)
or, in a matrix form, J
m1 0
0 x∙∙ 2k R b 1∙∙ r + J m2 x2 k
k x 0 R b 1r = b r 2k x2 0
Note that Eqs. (11.92) and (11.93) are secondorder homogenous differential equations. Also note that these equations are coupled because both x1 and x2 appear in each equation. This type of system is called elastically coupled and may be represented in the general matrix form by
M11_MOAV4303_04_GE_C11.INDD 655
[M]5 x∙∙ 6 + [K] 5 x6 = 0 (11.94)
27/11/14 10:12 AM
656 Chapter 11 Dynamic Problems Assuming x2 7 x1 k
kx1
m1
..  kx1 + k (x2  x1) = m1 x 1
m1 x1
k (x2  x1)
k
k (x2  x1)
m2
m2 x2
..  k (x2  x1)  kx2 = m2 x 2
kx2
k
Figure 11.27 A schematic diagram of an elastic system with two degrees of freedom.
where [M] and [K] are the mass and the stiffness matrices respectively. We can simplify Eqs. (11.92) and (11.93) by dividing both sides of each equation by the values of the respective masses:
x1∙∙ +
2k k x x = 0 (11.95) m1 1 m1 2
x2∙∙ 
k 2k x1 + x = 0 (11.96) m2 m2 2
Using matrix notation, we premultiply the matrix form of the equations of motion by the inverse of the mass matrix [M]1, which leads to 5 x∙∙ 6 + [M]1[K] 5 x6 = 0 (11.97)
As a next step, we assume a harmonic solution of the form x1(t) = X1 sin (vt + f) and x2(t) = X2 sin (vt + f) [or in matrix form, 5 x6 = 5 X 6 sin (vt + f)] and substitute the assumed solutions into the differential equations of motion, Eqs. (11.95) and (11.96), to create a set of linear algebraic equations. This step leads to v2X1 sin(vt + f) +
2k k X sin(vt + f) X sin(vt + f) = 0 m1 1 m1 2
v2X2 sin(vt + f) 
k 2k X sin(vt + f) + X sin(vt + f) = 0 m2 1 m2 2
After simplifying the sin(vt + f) terms, we get
M11_MOAV4303_04_GE_C11.INDD 656
2k m X 1 v2 b 1 r + D X2 k m2
k m1 X 0 T b 1 r = b r (11.98) 2k X2 0 m2

27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 657
or, in a general matrix form, v2 5 X 6 + [M]1[K] 5 X 6 = 0 (11.99)
Note that 5 x6 = b
x1(t) r represents the position of each mass as the function of time, x2(t)
X1 r matrix denotes the amplitudes of each oscillating mass, and f is the X2 phase angle. Equation (11.98) may be written as the 5 X 6 = b
1 v2 J 0
1 m1 0 X1 Rc s + D 1 X2 0
0 1 m2
TJ
2k k
k X R b 1 r = 0 (11.100) 2k X2
or by 2k m1 DD k m2
k m1 1 T  v2 C 2k 0 m2

0 X 0 S T b 1 r = b r (11.101) 1 X2 0
Simplifying Eq. (11.101) further, we have v2 + D
k m2
2k m1

k m1
2k v2 + m2
Tb
X1 r = 0 (11.102) X2
Problems with governing equations of the type (11.99) or (11.102) have nontrivial solutions only when the determinant of the coefficient matrix is zero. Let’s assign some numerical values to the above example problem and proceed with the solution. Let m1 = m2 = 0.1 kg and k = 100 N/m. Forming the determinant of the coefficient matrix and setting it equal to zero, we get 2
2 v + 2000 1000
1000 2 = 0 (11.103) v2 + 2000
( v2 + 2000)(  v2 + 2000)  ( 1000)( 1000) = 0 (11.104)
Simplifying Eq. (11.104), we have v4  4000v2 + 3,000,000 = 0 (11.105)
Equation (11.105) is called the characteristic equation, and its roots are the natural frequencies of the system.
M11_MOAV4303_04_GE_C11.INDD 657
27/11/14 10:12 AM
658 Chapter 11 Dynamic Problems
v21 = l1 = 1000 (rad/s)2 v22 = l2 = 3000 (rad/s)2
and and
v1 = 31.62 rad/s v2 = 54.77 rad/s
Once the v2 values are known, they can be substituted back into Eq. (11.102) to solve for the relationship between X1 and X2. The relationship between the amplitudes of mass oscillating at natural frequencies is called natural modes. We can use either relationship (rows) in Eq. (11.102). ( v2 + 2000) X1  1000X2 = 0
and substituting for v21 = 1000 S
( 1000 + 2000)X1  1000 X2 = 0
X2 = 1 X1
Or, using the second row, 1000 X1 + ( v2 + 2000)X2 = 0
and substituting for v21 = 1000
1000X1 + ( 1000 + 2000)X2 = 0 S
X2 = 1 X1
As expected, the results are identical. The second mode is obtained in a similar manner by substituting for v22 = 3000 in Eq. (11.102). ( v2 + 2000) X1  1000X2 = 0
and substituting for v22 = 3000
( 3000 + 2000)X1  1000X2 = 0 S
X2 = 1 X1
It is important to note again that the solution of the eigenvalue problems leads to establishing a relationship among the unknowns, not specific values. To shed more light on what the above natural frequencies and modes represent, consider the following experiment. Pull down both mass one and two by, say, 1 inch (X1 = X2 = 1), and then release the system. Under these initial conditions, the system will oscillate at the first natural frequency (v1 = 31.62 rad/s). However, if you were to give the system the following initial conditions: pull mass one up by 1 inch and pull mass two down by 1 inch (X2 = X1 = 1), then release them, the system would oscillate at the second natural frequency (v2 = 54.77 rad/s). Any other initial conditions will result in the system oscillating such that both natural frequencies affect its behavior. Equations of Motion for Forced Vibration of Multiple Degrees of Freedom In the previous section we showed that the general form of equations of motion for free vibration of multiple degrees of freedom is
[M]5 x∙∙ 6 + [K]5 x6 = 0 (11.106)
With damping [C] included, the equations of motion for free vibration of multiple degrees of freedom becomes
M11_MOAV4303_04_GE_C11.INDD 658
[M]5 x∙∙ 6 + [C]5 x∙ 6 + [K] 5 x6 = 0 (11.107)
27/11/14 10:12 AM
Section 11.2 Review of Vibration of Mechanical and Structural Systems 659
The natural response of a system with a multiple degrees of freedom can also be obtained using modal analysis. Modal analysis involves the uncoupling of the differential equations of motion using what is called principal coordinates. The basic idea is to represent the motion of each part (mass) by a single coordinate that makes no reference to any other coordinate. Once the equations of motion are uncoupled, then each independent equation is treated as a single degree of freedom system. To better understand the idea of principal coordinates, consider the two degrees of freedom system shown in Figure 11.27 and assume that m1 = m2 = m, then the equations of motion become 2k k x x = 0 (11.95b) m 1 m 2 k 2k x2∙∙ x + x = 0 (11.96b) m 1 m 2
x1∙∙ +
Substitute in Eqs. (11.95) and (11.96) m1 = m2 = m to arrive at Eqs. (11.95b) and (11.96b). Now consider first adding Eqs. (11.95b) and (11.96b) and then subtracting Eq. (11.96b) from Eq. (11.95b). These operations result in x1∙∙ + x2∙∙ +
k (x + x2) = 0 m 1
1 p1∙∙ +
k p = 0 m 1
x1∙∙  x2∙∙ +
3k (x  x2) = 0 m 1
1 p2∙∙ +
3k p = 0 m 2
where p1 = x1 + x2 and p2 = x1  x2. As you can see by using the principal coordinates p1 and p2, we were able to decouple the equations of motion. The natural frek 3k quencies of the system are v1 = and v2 = . It is left as an exercise for you to Am Am verify these results by determining the roots of characteristics equation. Although the decoupling of equations of motion are much more involved than what is shown here, the above example demonstrates the basic idea of principal coordinates and decoupling. The modal analysis is also used in determining the natural and the forced response of multiple degrees of freedom systems with damping. The matrix form of equations of motion for multiple degrees of freedom subjected to forces is given by [M]5 x∙∙ 6 + [K]5 x6 = 5 F6 (11.108)
where 5 F6 is the force matrix. And with damping, we have the following relationship: [M]5 x∙∙ 6 + [C]5 x∙ 6 + [K]5 x6 = 5 F6 (11.109)
Up to this point, we have explained how to approximate the behavior of systems with distributed mass and elastic properties by discrete models that consist of lumped masses and equivalent (or bulk) stiffness, and a finite number of degrees of freedom. Moreover, as you have seen, these models are represented by ordinary differential equations, subject to initial conditions, whose solutions render the natural or the forced responses of the system.
M11_MOAV4303_04_GE_C11.INDD 659
27/11/14 10:12 AM
660 Chapter 11 Dynamic Problems
Because rods and beams play significant role in many engineering applications, we will discuss their finite element formulations in detail in Sections 11.4 and 11.5. Rods and beams are continuous systems which theoretically possess infinite numbers of degrees of freedom and natural frequencies. However, for most practical problems, only the first few natural frequencies are important. In general, the governing equations of motions of continuous systems are partial differential equations whose exact solutions require both boundary and initial conditions. Except for a few simple problems, the solutions are very complex and hard to find. Therefore, we resort to numerical approximation of discrete models to solve many practical problems. We will discuss the finite element formulation of bars, beams, and frames using Lagrange’s equations in the forthcoming sections.
11.3 Lagrange’s Equations In the previous sections we derived the governing equations of motions of vibrating systems using Newton’s second law. In this section we introduce another approach, which uses Lagrange’s equations to formulate equations of motions. Lagrange’s equations are given by 0T 0Λ d 0T ¢ ∙≤ + = Qi (i = 1, 2, 3, . . , n) (11.110) dt 0q i 0qi 0qi
where
t = time T = kinetic energy of the system qi = coordinate system qi∙ = time derivate of the coordinate system representing velocity Λ = potential energy of the system Qi = nonconservative forces or moments We will use Example 11.4 to demonstrate how to use Lagrange’s equations to formulate equations of motion for a dynamic system.
Example 11.4 Use Lagrange’s equations to formulate the governing equations of motion for the systems shown in Figure 11.28. The spring mass system shown in Figure 11.28(a) is a single degree of freedom system that requires only one coordinate q to describe its behavior. To apply Eq. (11.110), we must first express the kinetic and potential energies of the system in terms of the coordinate q and its derivative q ∙. Note that for this problem, we let q = x. The kinetic
M11_MOAV4303_04_GE_C11.INDD 660
27/11/14 10:12 AM
Section 11.3 Lagrange’s Equations 661
k
k1
mass (a)
m1 x
x1 k2 m2 x2 (b)
Figure 11.28 Spring mass systems of Example 11.4.
and potential energies of the system are 2 1 mx∙ 2 1 Λ = kx2 2
T =
Next, as required by Lagrange’s equation, we differentiate the kinetic energy term with respect to q ∙, or in our case, x ∙ : 2 0T 0T 0 1 1 = ∙ = ∙ ¢ mx∙ ≤ = (2) ¢ ≤mx∙ = mx∙ 0q∙ 0x 0x 2 2
Then, taking the time derivative of
0T , we get 0x∙
d 0T d ¢ ∙≤ = (mx∙) = mx∙∙ dt 0x dt Because T is a function of x ∙ and not x, the in Eq. (11.110), we have
0T 0Λ = 0. Evaluating the potential term, , 0x 0q
0Λ 0Λ 0 1 2 = = ¢ kx ≤ = kx 0q 0x 0x 2 Finally, substituting for each term in Eq. (11.110), we get mx ∙∙
kx
5
5
c
0
0
5
0T 0Λ d 0T ¢ ≤ + = Qi dt 0qi∙ 0qi 0qi
M11_MOAV4303_04_GE_C11.INDD 661
27/11/14 10:12 AM
662 Chapter 11 Dynamic Problems
As expected, the governing differential equation of motion is given by mx∙∙ + kx = 0 The system shown in Figure 11.28(b) has two degrees of freedom; consequently, we need two coordinates, x1 and x2, to formulate the kinetic and potential energies. 1 1 ∙2 ∙2 m1x1 + m2x2 2 2 1 1 Λ = k1x21 + k2(x2  x1)2 2 2 T =
Taking the derivatives of each term, as required by Eq. (11.110), we get 0T = m1x1∙ 0x1∙
and
d 0T ¢ ≤ = m1x1∙∙ dt 0x1∙
0T = m2x2∙ 0x2∙
and
d 0T ¢ ≤ = m2x2∙∙ dt 0x2∙
0T 0T = = 0 0x1 0x2 0Λ = k1x1 + k2(x1  x2) 0x1 0Λ = k2(x2  x1) 0x2 Substituting each term in Lagrange’s equation, Eq. (11.110), we have m1x1∙∙ + (k1 + k2)x1  k2x2 = 0 m2x2∙∙  k2x1 + k2x2 = 0 or we can express the equations of motion in a matrix form by J
m1 0
0 x∙∙ k + k2 R b 1∙∙ r + J 1 m2 x2 k2
k2 x 0 R b 1r = b r k2 x2 0
11.4 Finite Element Formulation of Axial Members In this section, using Lagrange’s equations, we first formulate the mass matrix for an axial member and then use it to obtain the natural frequencies of an axial member. Recall that the displacement of an axial member may be expressed using onedimensional shape functions Si and Sj, as shown below: u = SiUi + SjUj (11.111)
The shape functions in terms of the local coordinate x shown in Figure 11.29 are given by
Si = 1 
Sj =
M11_MOAV4303_04_GE_C11.INDD 662
x L
x (11.112) L (11.113)
27/11/14 10:12 AM
Section 11.4 Finite Element Formulation of Axial Members 663 Uj
Ui x i
j L
Figure 11.29 An axial member.
It is important to note that whereas the displacement function for a static problem depends on coordinate x only, for a dynamic problem the displacement function is a function of x and time t; that is, u = u(x, t). The total kinetic energy of the member is the sum of the kinetic energies of its constituent particles (or smaller chunks). L
T =
g ∙2 u dx (11.114) L0 2
In Eq. (11.114), u∙ represents the velocity of the particles along the member, and g is mass per unit length. The velocity of the member can be expressed in terms of its nodal velocities Ui∙ and Uj∙ and is given by u∙ = SiUi∙ + SjUj∙ (11.115)
Substituting Eq. (11.115) into Eq. (11.114), we have
T =
g L (SiUi∙ + SjUj∙)2dx (11.116) 2 L0
And taking the derivatives as required by Lagrange’s equation, Eq. (11.110), we get
g L 0T 2Si(SiUi∙ + SjUj∙)dx ∙ = 0Ui 2 L0
(11.117)
g L 0T 2Sj(SiUi∙ + SjUj∙)dx ∙ = 0Uj 2 L0
(11.118)
L
L
d 0T ¢ ≤ = gJ Si2Ui∙∙dx + Si SjUj∙∙dx R (11.119) dt 0Ui∙ L0 L0
d 0T ¢ ≤ = gJ SiSjUi∙∙dx + Sj2Uj∙∙dx R (11.120) dt 0Uj∙ L0 L0
L
L
Note that Si and Sj are functions of x alone, whereas Ui∙∙ and Uj∙∙ represent accelerations at nodes i and j respectively, which are functions of time. Evaluating the integrals of Eqs. (11.119) and (11.120), we get
M11_MOAV4303_04_GE_C11.INDD 663
g
L0
L
Si2dx = g
L0
L
¢1 
gL x 2 ≤ dx = (11.121) L 3
27/11/14 10:12 AM
664 Chapter 11 Dynamic Problems
g
L0 g
L
SiSj dx = g
L0
L
Sj2dx = g
L0
L
L0
L
¢1 ¢
gL x x ≤ ¢ ≤dx = (11.122) L L 6
gL x 2 ≤ dx = L 3
(11.123)
Substituting the results of the integrals given by Eqs. (11.121) through (11.123) into Eqs. (11.119) and (11.120), leads to [M]{ü}. Then for an axial member, the mass matrix becomes [M](e) =
gL 2 J 6 1
1 R (11.124) 2
We derived the stiffness matrix for an axial element in Chapter 4, Section 4.1, which is given by [K](e) =
AE 1 J L 1
1 R 1
Next, we will use Example 11.5 to demonstrate how to use the results of this section.
Example 11.5 Consider the 30cm long aluminum rod shown in Figure 11.30. The rod has a modulus of elasticity E = 70 GPa and density r = 2700 kg/m3 (g = 5.4 kg/m). The rod is fixed at one end, as shown in the figure, and we are interested in approximating the natural frequencies of the rod using the threeelement model shown. The mass matrix for each element is computed from Eq. (11.124) [M](1) = [M](2) = [M](3) =
gL 2 J 6 1
(5.4)(0.1) 2 1 R = J 2 6 1
1 0.18 R = J 2 0.09
0.09 R 0.18
E = 70 GPa, r = 2700 kg/m3 A = 20 cm2
30 cm
U1 1
U2 2
10 cm
M11_MOAV4303_04_GE_C11.INDD 664
U3 3
10 cm
U4 4
10 cm
Figure 11.30 The aluminum rod of Example 11.5.
27/11/14 10:12 AM
Section 11.4 Finite Element Formulation of Axial Members 665
and the elemental stiffness matrix is [K](1) = [K](2) = [K](3) =
1 R = 1
AE 1 J L 1
(20 * 104)(70 * 109) 1 J 0.1 1
1 1 R = 1.4 * 109 J 1 1
1 R 1
Assembling the mass and the stiffness matrices,
[M](G)
0.18 0.09 = D 0 0
0.09 0.36 0.09 0
[K](G)
1.4 1.4 = 109 D 0 0
0 0.09 0.36 0.09 1.4 2.8 1.4 0
0 0 T 0.09 0.18 0 1.4 2.8 1.4
0 0 T 1.4 1.4
Applying the boundary condition—because node 1 is fixed—the first rows and columns of mass and stiffness matrices are eliminated. Application of the boundary condition at node 1 reduces the mass and the stiffness matrices to [M]
(G)
0.36 = C 0.09 0
(G)
2.8 = 10 C 1.4 0
[K]
0.09 0.36 0.09
9
0 0.09 S 0.18 1.4 2.8 1.4
0 1.4 S 1.4
As previously discussed in the section dealing with the natural vibration of multiple degrees of freedom systems, to get the natural frequencies of the system, we need to solve [M]1[K]5 X 6 = v2 5 X 6 or, in the case of the rod, [M]1[K] 5 U 6 = v2 5 U 6. Computing the inverse of the mass matrix, we get 2.9915 [M]1 = C 0.8547 0.4274
0.8547 3.4188 1.7094
0.4274 1.7094 S 6.4103
And calculating 0.9573 [M]1[K] = 1010 C 0.7179 0.3590
M11_MOAV4303_04_GE_C11.INDD 665
0.7179 1.3162 1.4359
0.1795 0.7179 S 1.1368
27/11/14 10:12 AM
666 Chapter 11 Dynamic Problems
and solving for the eigenvlaues, we obtain the natural frequencies: v1 = 1.5999 * 105 rad/s, v2 = 0.8819 * 105 rad/s, v3 = 0.2697 * 105 rad/s. Example 11.5 Revisited Using Example 11.5, we will show how to use Excel to set up and solve a dynamic problem. 1. In cell A1 type Example 11.5, and in cells A3, A4, A5, and A6 type A=, E=, L=, and G= , respectively, as shown. After inputting the value of A in cell B3, select B3 and in the “Name Box” type A and hit the Return key. Similarly, after inputting values of E, L, and g in cells B4, B5, and B6, select each cell and in each “Name Box” type E, L, and Gamma, respectively. Make sure to hit the Return key after you name each variable. Also, create the table shown.
2. Create the [M] and [K] matrices as shown. For example, select cell C16 and type =(Gamma*L , 6)*2. As another example, select cell C22 and type =(A*E , L)*1. Also, select the range C16:D17 and name it Melement1. Similarly, select the range C22:D23 and name it Kelement1.
3. Next, create [A1], [A2], and [A3] matrices and name them Aelement1, Aelement2, and Aelment3, as shown. See Section 2.5, Eq. (2.9). if you have forgotten what these matrices represent. First create [A1] and then to create [A2], and [A3], copy [A1], rows 25 through 27 into rows 29 through 31; 33 through 35; and modify them accordingly. The nodal temperatures U1, U2, U3, and U4, and Ui and Uj are shown
M11_MOAV4303_04_GE_C11.INDD 666
27/11/14 10:12 AM
Section 11.4 Finite Element Formulation of Axial Members 667
alongside the [A1], [A2], and [A3] matrices to aid us observe the contribution of node to its neighboring elements.
4. Next, create the mass matrix for each element (with their proper positions in the global matrix) and name them M1G, M2G, and M3G. For example, to create [M]1G, select B37:E40 and type =MMULT(TRANSPOSE(Aelement1),MMULT(Melement1,Aelement1)) and while holding down the Ctrl and Shift keys, hit the Return key. In a similar way, create [M]2G, and [M]3G, as shown. The final global mass matrix is created next. Select the range B52:E55 and type =M1G+M2G+M3G and while holding down the Ctrl and Shift keys, hit the Return key.
5. Create the stiffness matrix for each element (with their proper positions in the global matrix) and name them K1G, K2G, and K3G. For example, to create [K]1G, select B57:E60 and type =MMULT(TRANSPOSE(Aelement1),MMULT(Kelement1,Aelement1))
M11_MOAV4303_04_GE_C11.INDD 667
27/11/14 10:12 AM
668 Chapter 11 Dynamic Problems
and while holding down the Ctrl and Shift keys, hit the Return key. In a similar way, create [K]2G, and [K]3G, as shown. The final global stiffness matrix is created next. Select the range B72:E75 and type =K1G+K2G+K3G and while holding down the Ctrl and Shift keys, hit the Return key.
6. Apply the boundary conditions. Copy the appropriate portion of the MG matrix and paste it in the range C77:E79 as values only and modify it as shown Name the range MwithappliedBC. Similarly, create the corresponding stiffness matrix in the range C81:E83 and name it KwithappliedBC.
7. Next, we will create the [M]1 and [M]1[K] matrices. Select the range C85:E87 and type =MINVERSE(MwithappliedBC) and while holding down the Ctrl and Shift keys hit the Return key. Name this range InversofM. Then select the range C89:E91 and type =MMULT(InverseofM,KwithappliedBC) and while holding down the Ctrl and Shift keys, hit the Return key. Name this range Mminus1K.
M11_MOAV4303_04_GE_C11.INDD 668
27/11/14 10:12 AM
Section 11.4 Finite Element Formulation of Axial Members 669
8. Next, we create the identity matrix and name it I.
9. Input an initial guess (e.g. 2e10) in cell D98 and name it Omegasquared. In cell C100 type =MDETERM(Mminus1KOmegasquared*I) Also, in cell B102 type =SQRT(Omegasquared).
We now use the Goal Seek . . . function of Excel to calculate the eigenvalues. Choose the Data tab and the WhatIf Analysis and then Goal Seek. . .
Your initial guess will be changed to the new value. The complete Excel sheet is shown next.
M11_MOAV4303_04_GE_C11.INDD 669
27/11/14 10:12 AM
670 Chapter 11 Dynamic Problems
M11_MOAV4303_04_GE_C11.INDD 670
27/11/14 10:13 AM
Section 11.5 Finite Element Formulation of Beams and Frames 671
11.5 Finite Element Formulation of Beams and Frames In this section, using Lagrange’s equations, we will formulate the vibration of a beam and a frame element. Recall from Chapter 4, Section 4.2 that the deflection of a beam element may be represented using shape functions Si1, Si2, Sj1, Sj2 and the nodal displacements Ui1, Ui2, Uj1, Uj2 by v = Si1Ui1 + Si2Ui2 + Sj1Uj1 + Sj2Uj2 where the shape functions are given by 3x2 + L2 2x2 Si2 = x + L 3x2 2x3 Sj1 = 2  3 L L x2 x3 Sj2 =  + 2 L L Si1 = 1 
2x3 L3 x3 L2
For the sake of presentation continuity, the beam element shown previously in Figure 4.8 is repeated here. Ui1
Uj1
y Y
Uj2
Ui2 x
X
i
j L
Figure 4.8 A beam element.
The kinetic energy of a beam element is determined by adding the kinetic energy of its constituent particles according to L
T =
g ∙2 v dx (11.125) L0 2
In Eq. (11.125), v ∙ represents the velocity distribution within the beam and is a function of time and position. We can represent the velocity of the beam in terms of shape
M11_MOAV4303_04_GE_C11.INDD 671
27/11/14 10:13 AM
672 Chapter 11 Dynamic Problems
functions and the lateral and rotational velocities of nodes i and j according to v ∙ = Si1Ui1∙ + Si2Ui2∙ + Sj1Uj1∙ + Sj2Uj2∙ (11.126)
Substituting for the velocity distribution in the kinetic energy equation, we have L
T =
g ∙2 g L v dx = (Si1Ui1∙ + Si2Ui2∙ + Sj1Uj1∙ + Sj2Uj2∙ )2dx (11.127) 2 2 L0 L0
Although v ∙ is a function of time and position, note that the nodal velocities are only a function of time. The shape functions account for the spatial variations in Eqs. (11.26) and (11.27). Evaluating the derivative terms as required by Lagrange’s equations, we get
g L 0T = 2Si1 (Si1Ui1∙ + Si2Ui2∙ + Sj1Uj1∙ + Sj2Uj2∙ )dx (11.128) 0Ui1∙ 2 L0
g L 0T = 2Si2 (Si1Ui1∙ + Si2Ui2∙ + Sj1Uj1∙ + Sj2Uj2∙ )dx (11.129) 0Ui2∙ 2 L0
g L 0T = 2Sj1(Si1Ui1∙ + Si2Ui2∙ + Sj1Uj1∙ + Sj2Uj2∙ )dx (11.130) 0Uj1∙ 2 L0
g L 0T = 2Sj2 (Si1Ui1∙ + Si2Ui2∙ + Sj1Uj1∙ + Sj2Uj2∙ )dx (11.131) 0Uj2∙ 2 L0 and evaluating the
d 0T ¢ ≤ terms, we have dt 0qi L
d 0T ¢ ≤ = gJ Si1 (Si1Ui1∙∙ + Si2Ui2∙∙ + Sj1Uj1∙∙ + Sj2Uj2∙∙) dx R (11.132) dt 0Ui1∙ L0
d 0T ¢ ≤ = gJ Si2 (Si1Ui1∙∙ + Si2Ui2∙∙ + Sj1Uj1∙∙ + Sj2Uj2∙∙) dx R (11.133) dt 0Ui2∙ L0
d 0T ¢ ≤ = gJ Sj1 (Si1Ui1∙∙ + Si2Ui2∙∙ + Sj1Uj1∙∙ + Sj2Uj2∙∙) dx R (11.134) dt 0Uj1∙ L0
d 0T ∙∙ ∙∙ ∙∙ ∙∙ ¢ ∙ ≤ = gJ Sj2 (Si1Ui1 + Si2Ui2 + Sj1Uj1 + Sj2Uj2) dx R (11.135) dt 0Uj2 L0
L
L
L
Like nodal velocities, note that the lateral and rotational accelerations of nodes i and j Ui1∙∙, Ui2∙∙, Uj1∙∙, Uj1∙∙ are independent of coordinate x and are only functions of time. This realization allows us to pull the nodal accelerations out of the integrals in Eqs. (11.132) through (11.135), and only integrate the products of the shape functions. It is important to point out that when integrating Eqs. (11.132) to (11.135), we need not evaluate all
M11_MOAV4303_04_GE_C11.INDD 672
27/11/14 10:13 AM
Section 11.5 Finite Element Formulation of Beams and Frames 673
16 integrals. Some of these integrals are identical. The integrals that must be evaluated follow: g
L0
L0
g L
Si1Si2 dx = g
g
g
L0
L
L0
L
L0
L
S2i1 dx
¢1 
g
11gL2 3x 2x3 2x2 x3 11 + ≤ ¢ x + ≤ dx = = mL (11.137) 2 3 2 L 210 210 L L L
L
Si1Sj2 dx = g g L0
13gL 3x2 2x3 2 13 + 3 ≤ dx = = m (11.136) 2 35 35 L L
¢1 
2
Si1Sj1 dx = g
L0
= g
L
L0
L0
L0
L
¢1 
L
¢1 
L
S2j2 dx
L
Sj2Sj1 dx = g
L0
= g L
¢
9gL 3x2 2x3 3x2 2x3 + 3 ≤ ¢ 2  3 ≤ dx = (11.138) 2 70 L L L L
13gL2 3x2 2x3 x2 x3 + ≤ ¢ + ≤ dx = (11.139) L 420 L2 L3 L2 L0
L
¢
gL3 x2 x3 2 + 2 ≤ dx = (11.140) L 105 L
11gL2 x2 x3 3x2 2x3 + 2 ≤ ¢ 2  3 ≤ dx = (11.141) L 210 L L L
Incorporating the results of the integrations leads to [M]{V∙∙ }, then the mass matrix for a beam element is given by
[M](e)
156 gL 22L = D 420 54 13L
22L 4L2 13L 3L2
54 13L 156 22L
13L 3L2 T (11.142) 22L 4L2
We now turn our attention to the stiffness matrix. Recall from Section 4.2 that the stiffness matrix for a beam element is
[K](e)
12 EI 6L = 3D L 12 6L
6L 4L2 6L 2L2
12 6L 12 6L
6L 2L2 T (11.143) 6L 4L2
We discuss the finite element formulation of the mass matrix for a frame element next, and then demonstrate finite element modeling of oscillating frames with Example 11.7. Frame Element You may recall from our discussion of frames in Chapter 4 that frames represent structural members that may be rigidly connected with welded joints or bolted joints. For such structures, in addition to rotation and lateral displacement, we also need to be concerned about axial deformations. For the sake of presentation continuity, the frame element, Figure 4.12, is repeated here.
M11_MOAV4303_04_GE_C11.INDD 673
27/11/14 10:13 AM
674 Chapter 11 Dynamic Problems uj2
uj1
Y uj3 X
ui2
y
j
ui1
x
u
ui3 i
Figure 4.12 A frame element.
In the previous section we developed the mass matrix for a beam element. This matrix accounts for lateral displacements and rotations at each node and is
[M](e)
0 0 gL 0 = F 420 0 0 0
0 156 22L 0 54 13L
0 22L 4L2 0 13L 3L2
0 0 0 0 0 0
0 54 13L 0 156 22L
0 13L 3L2 V (11.144) 0 22L 4L2
The mass matrix for a member under axial movement was developed in Section 11.4 and is given by
[M](e)
2 0 gL 0 = F 6 1 0 0
0 0 0 0 0 0
0 0 0 0 0 0
1 0 0 2 0 0
0 0 0 0 0 0
0 140 0 0 gL 0 0 V = F 420 70 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
70 0 0 140 0 0
0 0 0 0 0 0
0 0 0 V (11.145) 0 0 0
Adding Eqs. (11.144) and (11.145) results in the mass matrix for a frame element:
M11_MOAV4303_04_GE_C11.INDD 674
[M](e)
140 0 gL 0 = F 420 70 0 0
0 156 22L 0 54 13L
0 22L 4L2 0 13L 3L2
70 0 0 140 0 0
0 54 13L 0 156 22L
0 13L 3L2 V (11.146) 0 22L 4L2
27/11/14 10:13 AM
Section 11.5 Finite Element Formulation of Beams and Frames 675
In Section 4.2, we derived the stiffness matrix for the frame element, which is given by AE L 0 0
[K](e) xy
=
I AE 
0
0
12EI L3 6EI L2
6EI L2 4EI L
0
0
L 0 0

12EI L3 6EI L2

AE L
0
0
12EI L3 6EI  2 L
6EI L2 2EI L
0
0

0 0 AE L
6EI L2 2EI L

12EI L3 6EI  2 L
0 0
Y
(11.147)
6EI L2 4EI L

In Chapter 4 we also discussed the role and the importance of using local and global coordinate systems to formulate and analyze finite element models. You may recall that the local degrees of freedom are related to the global degrees of freedom through the transformation matrix, according to the relationship 5 u 6 = [T] 5 U 6 (11.148)
where the transformation matrix is cos u sin u 0 [T] = F 0 0 0
sin u cos u 0 0 0 0
0 0 1 0 0 0
0 0 0 cos u sin u 0
0 0 0 sin u cos u 0
0 0 0 V (11.149) 0 0 1
The equations of motion in the element’s local coordinate system is given by ∙∙ (e) (e) [M](e) xy 5 u 6 + [K]xy 5 u 6 = 5 f6 (11.150)
We can now make use of the relationships between local and global displacements and accelerations 5 u 6 = [T]5 U 6 and 5 u∙∙ 6 = [T] 5 U ∙∙ 6 and the local and global description of forces 5 f6 = [T]5 F6 and substitute for 5 u 6 , 5 u∙∙ 6 , 5 f6 in Eq. (11.150). These substitutions result in 5 u6
5f 6 c
∙∙
c
c
5 u∙∙6
(e) (e) [M](e) xy [T]5 U 6 + [K]xy [T]5U6 = [T]5 F6
(11.151)
Premultiplying Eq. (11.151) by [T]1,
M11_MOAV4303_04_GE_C11.INDD 675
∙∙ 1 (e) 1 (e) [T]1[M](e) xy [T]5 U 6 + [T] [K]xy [T]5 U 6 = [T] [T]5 F6 (11.152)
27/11/14 10:13 AM
676 Chapter 11 Dynamic Problems
It can be readily shown that for the transformation matrix, [T]1 = [T]T (see Example 11.6). Using this relationship and simplifying Eq. (11.152), we get [M]
(e)
[K]
(e)
d
d
[T]T[M]xy(e)[T] 5 U ∙∙ 6 + [T]T[K]xy(e)[T] 5 U 6 = 5 F6 (e) (11.153)
Finally, the equations of motion in terms of global coordinates become
[M](e) 5 U ∙∙ 6 + [K](e) 5 U 6 = 5 F6 (e) (11.154)
where
[M](e) = [T]T[M](e) xy [T] (11.155)
and
[K](e) = [T]T[K](e) xy [T] (11.156)
We will demonstrate how to use these equations to determine the natural frequencies of beams and frames with Example 11.7.
Example 11.6 In this example we show that [T]1 = [T]T. We begin by showing that [T]T[T] = [I] cos u sin u 0 G 0 0 0
sin u cos u 0 0 0 0
0 0 1 0 0 0
0 0 0 cos u sin u 0
0 0 0 sin u cos u 0
0 cos u 0 sin u 0 0 WG 0 0 0 0 1 0
sin u cos u 0 0 0 0
0 0 1 0 0 0
1 0 0 G 0 0 0
0 0 0 1 0 0
0 0 0 W 0 0 1
0 1 0 0 0 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 cos u sin u 0
0 0 0 sin u cos u 0
0 0 0 W = 0 0 1
Note that the relationship cos2 u + sin2 u = 1 is used to simplify the results. Because [T]1[T] = [I] and we just showed that [T]T[T] = [I], then [T]1 = [T]T must be true.
M11_MOAV4303_04_GE_C11.INDD 676
27/11/14 10:13 AM
Section 11.5 Finite Element Formulation of Beams and Frames 677
Example 11.7 Consider the frame shown in Figure 11.31. The frame is made of steel, with E = 30 * 106 lb/in2. The crosssectional areas and second moment of areas for the members are shown in the figure. The frame is fixed as shown, and we are interested in determining the natural frequencies using the threeelement model shown. Members (1) and (3) are W12 * 26 steel beams, while member (2) is a W16 * 26 steel beam. The mass per unit length is g =
26 lb = 0.0056 lb.s2/in2 (12 in)(32.2 ft/s2)(12 in/ft)
For each element, the relationship between the local and global coordinate systems is shown in Figure 11.32. The stiffness values to be used in the stiffness matrix for elements (1) and (3) are (7.65 in2)(30 * 106 lb/in2) AE = = 1,275,000 lb/in L (15 ft)(12 in/ft) (12)(30 * 106 lb/in2)(204 in4) 12EI = = 12,592 lb/in L3 ((15 ft)(12 in/ft))3 (6)(30 * 106 lb/in2)(204 in4) 6EI = = 1,133,333 lb L2 ((15 ft)(12 in/ft))2 (2)(30 * 106 lb/in2)(204 in4) 2EI = = 68,000,000 lb.in L (15 ft)(12 in/ft) (4)(30 * 106 lb/in2)(204 in4) 4EI = = 136,000,000 lb.in L (15 ft)(12 in/ft) 20 ft
I = 204 in4 A = 7.65 in2 depth = 12.22 in
I = 301 in4 A = 7.68 in2 depth = 15.69 in
2
15 ft 1 3
Figure 11.31 The frame of Example 11.7.
M11_MOAV4303_04_GE_C11.INDD 677
27/11/14 10:13 AM
678 Chapter 11 Dynamic Problems Y
y (2)
x
X
2
3 Y
2
3
X
y x (3)
x
(1)
y
Y X
1
4
Figure 11.32 The configuration of elements (1), (2), and (3).
The local stiffness matrices for elements (1) and (3) are AE L 0 0
(3) [K](1) xy = [K]xy =
I AE 
0
M11_MOAV4303_04_GE_C11.INDD 678
0 12.592 1133.333 0 12.592 1133.333
0
12EI L3 6EI L2
6EI L2 4EI L
0
0
L 0
1275 0 0 = 103 F 1275 0 0
0

12EI L3 6EI L2
0 1133.333 136000 0 1133.333 68000
6EI L2 2EI L

1275 0 0 1275 0 0

AE L 0 0
AE L 0 0
0
0
12EI L3 6EI  2 L
6EI L2 2EI L
0
0

12EI L3 6EI  2 L
0 12.592 1133.333 0 12.592 1133.333
Y
6EI L2 4EI L

0 1133.333 68000 V 0 1133.333 136000
27/11/14 10:13 AM
Section 11.5 Finite Element Formulation of Beams and Frames 679
The local mass matrix for elements (1) and (3) is
(3) [M](1) xy = [M]xy
=
140 0 gL 0 = F 420 70 0 0
0 156 22L 0 54 13L
0 22L 4L2 0 13L 3L2
70 0 0 140 0 0
0 54 13L 0 156 22L
0 13L 3L2 V 0 22L 4L2
(0.0056 lb.s2/in2)(15 ft)(12 in/ft) * 420
140 0 0 F 70 0 0
0 156 (22)(15)(12) 0 54 (13)(15)(12)
(3) [M](1) xy = [M]xy
0 (22)(15)(12) (4)((15)(12))2 0 (13)(15)(12) (3)((15)(12))2 140 0 0 = 0.0024F 70 0 0
70 0 0 140 0 0 0 156 3960 0 54 2340
0 54 (13)(15)(12) 0 156 (22)(15)(12) 0 3960 129600 0 2340 97200
0 (13)(15)(12)  (3)((15)(12))2 V 0 (22)(15)(12) (4)((15)(12))2
70 0 0 140 0 0
0 54 2340 0 156 3960
0 2340 97200 V 0 3960 129600
For element (1), the transformation matrix and its transpose are cos(90) sin(90) 0 [T] = F 0 0 0
sin(90) cos(90) 0 0 0 0
0 0 1 0 0 0
0 0 0 cos(90) sin(90) 0 0 1 0 [T]T = F 0 0 0
M11_MOAV4303_04_GE_C11.INDD 679
0 0 0 sin(90) cos(90) 0 1 0 0 0 0 0
0 0 1 0 0 0
0 0 0 1 0 0 V = F 0 0 0 0 1 0 0 0 0 0 1 0
0 0 0 1 0 0
1 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
0 0 0 V 0 0 1
27/11/14 10:13 AM
680 Chapter 11 Dynamic Problems
For element (3), the transformation matrix and its transpose are cos(270) sin(270) 0 sin(270) cos(270) 0 0 0 1 [T] = F 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 V = F V cos(270) sin(270) 0 0 0 0 0 1 0 sin(270) cos(270) 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 1
0 1 0 [T]T = F 0 0 0
1 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
Substituting for [T]T, [K](1) xy , and [T] into Eq. (11.156), we have:
[K](1)
0 1 0 = 103 F 0 0 0
1 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 1275 0 0 1275 0 0 0 0 12.592 1133.333 0 12.592 1133.333 0 0 1133.333 136000 0  1133.333 68000 VF V 0 1275 0 0 1275 0 0 0 0  12.592 1133.333 0 12.592  1133.333 1 0 1133.333 68000 0  1133.333 136000 0 1 0 F 0 0 0
[K](1)
12.592 0 1133.33 = 103 F 12.592 0 1133.333
M11_MOAV4303_04_GE_C11.INDD 680
0 1275 0 0 1275 0
1133.33 0 136000 133.333 0 68000
12.592 0 1133.333 12.59 0 1133.33
1 0 0 0 0 0
0 0 1 0 0 0
0 1275 0 0 1275 0
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
1133.333 0 68000 V 1133.333 0 136000
27/11/14 10:13 AM
Section 11.5 Finite Element Formulation of Beams and Frames 681
Substituting for [T]T, [M](1) xy , and [T] into Eq. (11.155), we have
[M](1)
0 1 0 = 0.0024F 0 0 0 140 0 0 F 70 0 0
1 0 0 0 0 0
0 0 1 0 0 0
0 156 3960 0 54 2340
[M](1)
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
0 3960 129600 0 2340 97200
70 0 0 140 0 0
156 0 3960 = 0.0024F 54 0 2340
0 54 2340 0 156 3960
0 140 0 0 70 0
0 0 2340 1 97200 0 VF 0 0 3960 0 129600 0
3960 0 129600 2340 0 97200
54 0 2340 156 0 3960
1 0 0 0 0 0
0 70 0 0 140 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
2340 0 97200 V 3960 0 129600
Similarly, the stiffness matrix and the mass matrix for element (3) are
[K](3)
0 1 0 = 103 F 0 0 0
1 0 0 0 0 0
0 0 0 0 1 0 0 0 0 1 0 0
0 0 0 1 0 0
0 1275 0 0 1275 0 0 0 12.592 1133.333 0  12.592 0 0 1133.333 136000 0 1133.333 VF 0  1275 0 0 1275 0 0 0  12.592  1133.333 0 12.592 1 0 1133.333 68000 0  1133.333 0 1 0 F 0 0 0
[K](3)
M11_MOAV4303_04_GE_C11.INDD 681
12.592 0 1133.33 = 103 F 12.592 0 1133.333
0 1275 0 0 1275 0
1133.33 0 136000 133.333 0 68000
12.592 0 1133.333 12.59 0 1133.33
1 0 0 0 0 0
0 1275 0 0 1275 0
0 0 1 0 0 0
0 1133.333 68000 V 0  1133.333 136000 0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
1133.333 0 68000 V 1133.333 0 136000
27/11/14 10:13 AM
682 Chapter 11 Dynamic Problems
[M](3)
0 1 0 = 0.0024F 0 0 0
140 0 0 F 70 0 0
0 156 3960 0 54 2340
[M](3)
1 0 0 0 0 0
0 0 1 0 0 0
0 3960 129600 0 2340 97200
0 0 0 0 1 0
0 0 0 1 0 0
70 0 0 140 0 0
156 0 3960 = 0.0024F 54 0 2340
0 0 0 V 0 0 1 0 54 2340 0 156 3960
0 140 0 0 70 0
0 0 2340 1 97200 0 VF 0 0 3960 0 129600 0
3960 0 129600 2340 0 97200
54 0 2340 156 0 3960
1 0 0 0 0 0 0 70 0 0 140 0
0 0 1 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 V 0 0 1
2340 0 97200 V 39600 0 129600
The stiffness values for element (2) are (7.68 in2)(30 * 106 lb/in2) AE = = 960,000 lb/in L (20 ft)(12 in/ft) (12)(30 * 106 lb/in2)(301 in4) 12EI = = 7838 lb/in L3 ((20 ft)(12 in/ft))3 (6)(30 * 106 lb/in2)(301 in4) 6EI = = 940,625 lb L2 ((20 ft)(12 in/ft))2 (2)(30 * 106 lb/in2)(301 in4) 2EI = = 75,250,000 lb.in L (20 ft)(12 in/ft) (4)(30 * 106 lb/in2)(301 in4) 4EI = = 150,500,000 lb.in L (20 ft)(12 in/ft) For element (2), the local and the global frames of reference are aligned in the same direction; therefore, the stiffness matrix is
M11_MOAV4303_04_GE_C11.INDD 682
27/11/14 10:13 AM
Section 11.5 Finite Element Formulation of Beams and Frames 683
AE L 0 0
(2)
[K]
=
I AE 
0
12EI L3 6EI L2
6EI L2 4EI L
0
0
L 0

0

0
12EI L3 6EI L2
960 0 0 103 F 960 0 0
0
6EI L2 2EI L
0
0
12EI L3 6EI  2 L
6EI L2 2EI L
0
0
12EI L3 6EI  2 L
0 0
0 940.625 150500 0 940.625 75250
0 
AE L

0 7.838 940.625 0 7.838 940.625
AE L
960 0 0 960 0 0
Y
=
6EI L2 4EI L

0 7.838 940.625 0 7.838 940.625
0 940.625 75250 V 0 940.625 150500
The mass matrix for element (2) is
[M](2)
140 0 gL 0 = F 420 70 0 0
140 0 0 156 0 (22)(20)(12) F 70 0 0 54 0  (13)(20)(12)
0 22L 4L2 0 13L  3L2
70 0 0 140 0 0
0 54 13L 0 156  22L
0 70 0 (22)(20)(12) 0 54 (4)((20)(12))2 0 (13)(20)(12) 0 140 0 (13)(20)(12) 0 156  (3)((20)(12))2 0  (22)(20)(12)
[M](2)
M11_MOAV4303_04_GE_C11.INDD 683
0 156 22L 0 54  13L
140 0 0 = 0.0032F 70 0 0
0 156 5280 0 54 3120
0  13L (0.0056 lb.s2/in2)(20 ft)(12 in/ft)  3L2 V = 420 0  22L 4L2 0  (13)(20)(12)  (3)((20)(12))2 V 0  (22)(20)(12) (4)((20)(12))2
0 5280 230400 0 3120 172800
70 0 0 140 0 0
0 54 3120 0 156 5280
0 3120 172800 V 0 5280 230400
27/11/14 10:13 AM
684 Chapter 11 Dynamic Problems
Next, we will construct the global stiffness and mass matrices.
[K](G)
[M](G)
12.59 0  1133.333  12.59 0 3  1133.333 = 10 I 0 0 0 0 0 0
0.37 0  9.50 0.13 0 5.62 = I 0 0 0 0 0 0
0 0.34 0 0 0.17 0 0 0 0 0 0 0
0 1275 0 0  1275 0 0 0 0 0 0 0
 9.50 0 311.04  5.62 0  233.28 0 0 0 0 0 0
 1133.333 0 136000 113.333 0 68000 0 0 0 0 0 0
0.13 0  5.62 0.82 0 9.50 0.22 0 0 0 0 0
 12.59 0 1133.333 972.59 0 1133.333  960 0 0 0 0 0
0  1275 0 0 1282.84 940.63 0  7.84 940.63 0 0 0
 1133.333 0 68000 1133.33 940.63 286500 0  940.63 75250 0 0 0
0 0 0  960 0 0 972.59 0 1133.33  12.59 0 1133.33
0 0 0 0  7.84  940.63 0 1282.84  940.63 0  1275 0
0 0 0 0 940.63 75250 1133.33  940.63 286500  1133.33 0 68000
0 0.17 0 0 0.84 16.90 0 0.17  9.98 0 0 0
5.62 0  233.28 9.50 16.90 1048.32 0 9.98  552.96 0 0 0
0 0 0 0.22 0 0 0.82 0 9.50 0.13 0  5.62
0 0 0 0 0.17 9.98 0 0.84  16.90 0 0.17 0
0 0 0 0 0 0  12.59 0  1133.33 12.59 0  1133.33 0 0 0 0  9.98  552.96 9.50  16.90 1048.32 5.62 0  233.28
0 0 0 0 0 0 0  1275 0 0 1275 0 0 0 0 0 0 0 0.13 0 5.62 0.37 0  9.50
0 0 0 0 0 0 Y 1133.333 0 68000  1133.333 0 136000
0 0 0 0 0 0 0 0.17 0 0 0.34 0
0 0 0 0 0 0 Y  5.62 0  233.28  9.50 0 311.04
Applying the boundary conditions, the global stiffness and mass matrices reduce to
[K](G)
M11_MOAV4303_04_GE_C11.INDD 684
972.59 0 1133.33 = 103 F 960 0 0
0 1282.84 940.63 0 7.84 940.63
1133.33 940.63 286500 0 940.63 75250
960 0 0 972.59 0 1133.33
0 7.84 940.63 0 1282.84 940.63
0 940.63 75250 V 1133.33 940.63 286500
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 685
[M](G)
0.82 0 9.50 = F 0.22 0 0
0 0.84 16.90 0 0.17 9.98
9.50 16.90 1048.32 0 9.98 552.96
0.22 0 0 0.82 0 9.50
0 0.17 9.98 0 0.84 16.90
0 9.98 552.96 V 9.50 16.90 1048.32
And solving [M]1[K]5 U 6 = v2 5 U 6 for the eigenvalues, we get
v1 = 95 rad/s v2 = 355 rad/s v3 = 893 rad/s v4 = 1460 rad/s v5 = 1570 rad/s v6 = 2100 rad/s
11.6 Examples Using ANSYS In this section, ANSYS is used to solve two problems. First we revisit Example 11.7 and then consider the natural oscillation of a straight member with a rectangular crosssection. As the formulations in the previous sections showed, the stiffness matrix for oscillating rods, beam, and frame elements are the same as those for static problems. However, you must include density values in the material models for calculation of mass matrices. Therefore, when modeling dynamic problems, the preprocessing phase including element selection is identical to that of static problems. It is during the solution phase that the correct dynamic analysis type must be selected. Example 11.7 Revisited Consider the frame from Example 11.7 shown in Figure 11.31. The frame is made of steel, with E = 30 * 106lb/in2. The crosssectional areas and second moment of areas for the members are shown in the figure. The frame is fixed as shown, and we are interested in determining the natural frequencies using the threeelement model shown. Members (1) and (3) are W12 * 26 steel beams, while member (2) is a W16 * 26 steel beam. 20 ft
I = 204 in4 A = 7.65 in2 depth = 12.22 in
I = 301 in4 A = 7.68 in2 depth = 15.69 in
2
15 ft 1 3
M11_MOAV4303_04_GE_C11.INDD 685
27/11/14 10:13 AM
686 Chapter 11 Dynamic Problems
Enter ANSYS program by using the Launcher. Type Osciframe (or a file name of your choice) in the Jobname entry field of the dialog box. Pick Run to start the Graphic User Interface (GUI). Create a title for the problem. utility menu: File S ChangeTitle . . .
main menu: Preprocessor S Element Type S Add/Edit/Delete
OK
M11_MOAV4303_04_GE_C11.INDD 686
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 687
main menu: Preprocessor S Sections S Beams S Common Sections
M11_MOAV4303_04_GE_C11.INDD 687
27/11/14 10:13 AM
688 Chapter 11 Dynamic Problems
Close main menu: Preprocessor S Material Props S Material Models S Structural S Linear S Elastic S Isotropic
OK main menu: Preprocessor S Material Props S Material Models S Structural S Density
Note, here, density is equal to mass per unit volume.
OK
M11_MOAV4303_04_GE_C11.INDD 688
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 689
ANSYS Toolbar: SAVE_DB main menu: Preprocessor S Modeling S Create S Nodes S In Active CS
Apply
Apply
M11_MOAV4303_04_GE_C11.INDD 689
27/11/14 10:13 AM
690 Chapter 11 Dynamic Problems
Apply
OK main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes Pick Nodes 1 and 2 and Apply, then pick Nodes 4 and 3. OK main menu: Preprocessor S Modeling S Create S Elements S Elem Attributes
OK main menu: Preprocessor S Modeling S Create S Elements S Auto Numbered S Thru Nodes
M11_MOAV4303_04_GE_C11.INDD 690
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 691
Pick nodes 2 and 3 main menu: Solution S Define Loads S Apply S Structural S Displacement S On Nodes Pick nodes 1 and 4
main menu: Solution S Analysis Type S New Analysis
main menu: Solution S Analysis Type S Analysis Options As we mentioned previously, we use modal analysis to obtain the natural frequencies and the mode shapes of a vibrating system. Moreover, you should recall that when a natural frequency of a system matches the excitation frequency, resonance will occur.
M11_MOAV4303_04_GE_C11.INDD 691
27/11/14 10:13 AM
692 Chapter 11 Dynamic Problems
You must perform a new analysis each time you change boundary conditions. After you select the Modal in the New Analysis window, you need to choose a mode extraction method. ANSYS offers the following extraction methods: Block Lanczos (default), Subspace, Powerdynamics, Reduced, Unsymmetric, Damped, and QR Damped. You use the Block Lanczos method to solve large symmetrical eingenvalue problems. It uses sparse matrix solver and has a fast convergence rate. You can also use the Subspace method to solve large symmetrical problems. This method offers several solution controls to manage the iteration process. The Powerdynamics method is used for models with over 100,000 degrees of freedom and makes use of lumped mass approximation. As the name implies the Reduced method uses reduced system matrices to extract frequencies. It is not as accurate as the Subspace method but it has a faster convergence rate. Typically, you use the Unsymmetric method for problems dealing with fluidstructure interactions whose matrices are unsymmetrical. The Damped method is used for problems for which damping must be included in the model. The QR Damped method has a faster convergence rate than the Damped method. It uses the reduced modal damped matrix to calculate frequencies. If you choose the Reduced, Unsymmetic, or the Damped methods, then you must specify the number of modes to expand—the reduced solution is expanded to include the full degrees of freedom. In ANSYS, expansion also means writing the mode shapes to result files. When using the Reduced method, you need to define the master degrees of freedom—you choose at least twice as many master degrees of freedom as there are number of modes of interest. Additional modal analysis options include specifying a frequency range for mode extraction. For most cases you don’t need to specify the range and use the ANSYS’s default setting. When using the Reduced method you need to specify the number of reduced modes to be listed. Finally, if you are planning to perform spectrum analysis, then you need to normalize the mode shapes with respect to mass—the default setting.
M11_MOAV4303_04_GE_C11.INDD 692
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 693
main menu: Solution S Solve S Current LS Close (the solution is done!) window. Close (the/STAT Command) window. main menu: General Postproc S Results Summary
Note frequency values are given in hertz.
M11_MOAV4303_04_GE_C11.INDD 693
27/11/14 10:13 AM
694 Chapter 11 Dynamic Problems
Example 11.8 In this example problem we use ANSYS to study the natural oscillation of an aluminum strip with a rectangular cross section. The strip is 3 cm wide, 0.5 mm thick, and 10 cm long. It has a density of 2800 kg/m3, a modulus of elasticity of E = 73 GPa, and a Poisson’s ratio of 0.33. We assume the strip to be fixed at one end. Enter the ANSYS program by using the Launcher. Type Oscistrip (or a file name of your choice) in the Jobname entry field of the dialog box. Pick Run to start the GUI. Create a title for the problem. utility menu: File S Change Title . . . Type in the title of your choice main menu: Preprocessor S Element Type S Add/Edit/Delete
Apply
OK
M11_MOAV4303_04_GE_C11.INDD 694
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 695
Close main menu: Preprocessor S Material Props S Material Models S Structural S Linear S Elastic S Isotropic
OK
M11_MOAV4303_04_GE_C11.INDD 695
27/11/14 10:13 AM
696 Chapter 11 Dynamic Problems
OK ANSYS Toolbar: SAVE_DB main menu: Preprocessor S Modeling S Create S Areas S Rectangle S By 2 Corners
OK main menu: Preprocessor S Meshing S Size Cntrls S Smart Size S Basic
M11_MOAV4303_04_GE_C11.INDD 696
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 697
OK main menu: Preprocessor S Meshing S Mesh S Areas S Free First click on Pick All, then press OK. If you get a warning, it is ok; proceed. The purpose of this example is to show how to animate different modes of oscillation (using a coarse mesh). main menu: Preprocessor S Modeling S Operate S Extrude S Elem Ext Opts Choose Element type number 2 SOLID185. Also in the Element sizing options for extrusion, VAL1 No. Elem divs field type the value 10.
OK main menu: Preprocessor S Modeling S Operate S Extrude S Areas S By XYZ Offset Pick All
M11_MOAV4303_04_GE_C11.INDD 697
27/11/14 10:13 AM
698 Chapter 11 Dynamic Problems
OK utility menu: Plot Ctrls S Pan, Zoom, Rotate . . . Choose ISO utility menu: Select S Entities
Select Elements, By Attributes, pick Elem type num, type 1 in the Min, Max, Inc field, and pick Unselect. Apply
Then select the nodes to apply the zero displacement boundary conditions. Select Nodes, By Location, pick Z coordinates, type 0 in the Min, Max, and pick From Full set.
M11_MOAV4303_04_GE_C11.INDD 698
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 699
Apply main menu: Solution S Define Loads S Apply S Structural S Displacement S On Nodes Pick All
OK
M11_MOAV4303_04_GE_C11.INDD 699
27/11/14 10:13 AM
700 Chapter 11 Dynamic Problems
utility menu: Select S Everything . . . main menu: Solution S Analysis Type S New Analysis
OK main menu: Solution S Analysis Type S Analysis Options
M11_MOAV4303_04_GE_C11.INDD 700
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 701
OK Use the help menu to obtain information on Block Lanczos method.
main menu: Solution S Solve S Current LS Close (the solution is done!) window. Close (the /STAT Command) window. Close
M11_MOAV4303_04_GE_C11.INDD 701
27/11/14 10:13 AM
702 Chapter 11 Dynamic Problems
main menu: General Postproc S Read Results S First Set utility menu: PlotCtrls S Animate S Mode Shape
OK
M11_MOAV4303_04_GE_C11.INDD 702
27/11/14 10:13 AM
Section 11.6 Examples Using ANSYS 703
Look at the next mode by issuing the commands main menu: General Postproc S Read Results S Next Set utility menu: PlotCtrls S Animate S Mode Shape Exit from ANSYS.
M11_MOAV4303_04_GE_C11.INDD 703
27/11/14 10:13 AM
704 Chapter 11 Dynamic Problems
Summary At this point you should 1. have a good grasp of the fundamental concepts and governing equations of motions for a particle, a rigid body, and a dynamic system. 2. have a good understanding of fundamental definitions and concepts of vibration of mechanical and structural systems. 3. understand the formulation and natural vibration behavior for a system with single and multiple degrees of freedom. 4. understand the formulation and forced vibration behavior for a system with single and multiple degrees of freedom system. 5. understand the finite element formulation of axial members, beams, and frames. 6. know how to use ANSYS to solve some dynamic problems. References Beer, F. P., and Johnston, E. R., Vector Mechanics for Engineers, 5th ed., New York, McGrawHill, 1988. Steidel, R., An Introduction to Mechanical Vibrations, 3rd ed., New York, John Wiley and Sons, 1971. Timoshenko, S., Young, D. H., and Weaver, W., Vibration Problems in Engineering, 4th ed., New York, John Wiley and Sons, 1974.
Problems 1. A simple dynamic system is modeled by a single degree of freedom with m = 10 kg and
kequivalent = 100 N/cm. Calculate the frequency and period of oscillation. Also, determine the maximum velocity and acceleration of the system. (y(0) = 5 mm) 2. The system described in Problem 1 is subjected to a sinusoidal forcing function F(t) = 20 sin(20t) in Newtons. Calculate the amplitude, maximum velocity, and acceleration of the system.
3. Derive the equations of motion for the system shown in the accompanying figure. What is the natural frequency of the system?
L/2 k1
C
L/2 2k1
M11_MOAV4303_04_GE_C11.INDD 704
27/11/14 10:13 AM
Chapter 11 Problems 705
4. Derive the response of a single degree of freedom mass spring system to a suddenly applied force F0, as shown in the accompanying figure. Plot the response. Compare the response of the system to a situation where the force F0 is applied as a ramp function. Also compare the response of the system to a suddenly applied force to a situtation where the force F0 is applied statically. F(t)
F (t) F0
m y k
t
5. Derive the response of a single degree of freedom mass spring system to a suddenly applied
force F0 that decays with time according to F(t) = F0e  c1t, as shown in the accompanying figure. The value of c1 defines the rate of decay. Plot the response of the system for different values of c1. Compare the response of the system to a situation where the force F0 is applied statically. F (t) F(t) F0
m y
F (t) = F0 ec1t
k
t
6. Derive the response of the system shown in the accompanying figure to the excitation shown.
k2
y2 =Y0 sinvt
m k1 2
M11_MOAV4303_04_GE_C11.INDD 705
k1 2
y1
27/11/14 10:13 AM
706 Chapter 11 Dynamic Problems
7. Using both Newton’s second law and Lagrange’s equations, formulate the equations of motion for the system shown in the accompanying figure.
k1
L
L
L k2
8. Using both Newton’s second law and Lagrange’s equations, formulate the equations of motion for the system shown in the accompanying figure.
L1
L2 G k2
k1
9. Using both Newton’s second law and Lagrange’s equations, formulate the equations of motion for the system shown in the accompanying figure.
x3
m3 k3
k3 m2
k2
x2 k2
m1 k1
M11_MOAV4303_04_GE_C11.INDD 706
x1 k1
27/11/14 10:13 AM
Chapter 11 Problems 707
10. Determine the first two natural frequencies of the axial members shown in the accompanying figure. Members are made from structural steel with a modulus of elasticity of E = 29 * 106 lb/in2 and a mass density of 15.2 slugs/ft3.
Ac = 0.75 in2 12–
Ac = 0.5 in2 9–
Ac = 1 in2 12–
11. Determine the first two natural frequencies of the axial member shown in the accompanying figure. Member is made from aluminum alloy with a modulus of elasticity of E = 10 * 106 lb/in2 and a mass density of 5.4 slugs/ft3.
Ac = 0.5 in2 5 ft
12. Determine the first three natural frequencies of the post shown. The post is made of structural steel with a modulus of elasticity of E = 29 * 106 lb/in2 and a mass density of 15.2 slugs/ft3. Consider only the axial oscillation.
5 ft
Ac = 0.75 in2
5 ft
Ac = 2.15 in2
10 ft
Ac = 2.95 in2
13. The cantilevered beam shown in the accompanying figure is a W18 * 50. Determine its first two natural frequencies.
15 ft
M11_MOAV4303_04_GE_C11.INDD 707
27/11/14 10:13 AM
708 Chapter 11 Dynamic Problems
14. The cantilevered beam shown in the accompanying figure is a W16 * 31. Determine its first three natural frequencies.
15 ft
7 ft
15. The simply supported beam shown in the accompanying figure is a W6 * 15. Determine its first two natural frequencies.
20 ft
16. Determine the first two natural frequencies of the simply supported beam with the rectangular cross section shown in the accompanying figure.
15 cm
r = 700 kg/m3, E = 10 GPa 7 cm 5m
17. Consider the overhang frame shown in the accompanying figure. The crosssectional areas and second moment of areas for each W12 * 26 member are shown in the figure. Determine the first three natural frequencies of the system. 10 ft
A = 7.65 in2 I = 204 in4
M11_MOAV4303_04_GE_C11.INDD 708
9 ft
27/11/14 10:13 AM
Chapter 11 Problems 709
18. Consider the overhang frame shown in the accompanying figure. The members of the frame are W5 * 16. Determine the first three natural frequencies of the system.
12 ft
20 ft
7 ft
19. Resolve Example 11.7 for the case where all members of the frame are W12 * 26. 20. Using ANSYS, determine the first three natural frequencies of the frame shown in the accompanying figure. 10 ft
W4 * 13
30 ft W5 * 16
M11_MOAV4303_04_GE_C11.INDD 709
27/11/14 10:13 AM
710 Chapter 11 Dynamic Problems
21. Using ANSYS, determine the first three natural frequencies of the frame shown in the accompanying figure.
W8 * 58
W8 * 58 455
20 ft
455
W10 * 112
W10 * 112
15 ft
22. Resolve Example 11.8 for the case where thickness of the strip is 1 mm, and 2 mm. Compare results and discuss your findings.
M11_MOAV4303_04_GE_C11.INDD 710
27/11/14 10:13 AM
C h a p t e r 1 2
Analysis of Fluid Mechanics Problems
The main objective of this chapter is to introduce you to the analysis of fluid mechanics problems. First, we will discuss the direct formulation of pipenetwork problems. Then, we consider finite element formulation of ideal fluid behavior (inviscid flow). Finally, we briefly look at the flow of fluid through porous media and finite element formulation of underground seepage flows. The main topics discussed in Chapter 12 include the following: 12.1 Direct Formulation of Flow Through Pipes 12.2 Ideal Fluid Flow 12.3 Groundwater Flow 12.4 Examples Using ANSYS 12.5 Verification of Results 12.1 Direct Formulation of Flow Through Pipes We begin by reviewing fundamental concepts of fluid flow through pipes. The internal flow through a conduit may be classified as laminar or turbulent flow. In laminar flow situations, a thin layer of dye injected into a pipe will show as a straight line. No mixing of fluid layers will be visible. This situation does not hold for turbulent flow, in which the bulk mixing of adjacent fluid layers will occur. Laminar and turbulent flow are depicted in Figure 12.1. Laminar flow typically occurs when the Reynolds number of the flowing fluid is less than 2100. The Reynolds number is defined as
Re =
rVD (12.1) m
where r and m are the density and the dynamic viscosity of the fluid respectively. V represents the average fluid velocity, and D represents the diameter of the pipe. The flow is
M12_MOAV4303_04_GE_C12.INDD 711
27/11/14 10:15 AM
712 Chapter 12 Analysis of Fluid Mechanics Problems Dye
(a) Laminar flow in a pipe Dye
(b) Turbulent flow in a pipe
Figure 12.1 Laminar and turbulent flows.
said to be in a transition region when the Reynolds number is typically between 2100 and 4000. The behavior of the fluid flow is unpredictable in the transition region. The flow is generally considered to be turbulent when the Reynolds number is greater than 4000. The conservation of mass for a steady flow requires that the mass flow rate at any section of the pipe remains constant according to the equation m1∙ = m2∙ = r1V1A1 = r2V2A2 = constant (12.2)
Again, r is the density of the fluid, V is the average fluid velocity at a section, and A represents the crosssectional area of the flow as shown in Figure 12.2. Section 2 Section 1
r 1 V1 A1
Flow
r 2 V2 A2
Figure 12.2 Flow of fluid through a conduit with variable cross section.
M12_MOAV4303_04_GE_C12.INDD 712
27/11/14 10:15 AM
Section 12.1 Direct Formulation of Flow Through Pipes 713
For an incompressible f low—a flow situation where the density of the fluid remains constant—the volumetric flow rate Q through a conduit at any section of the conduit is also constant:
Q1 = Q2 = V1A1 = V2A2 (12.3) For a fully developed laminar flow, there exists a relationship between the volumetric flow rate and the pressure drop P1  P2 along a pipe of length L. This relationship is given by
Q =
pD4 P1  P2 ¢ ≤ (12.4) 128m L
The pressure drop for a turbulent flow is commonly expressed in terms of head loss, which is defined as
Hloss =
P1  P2 L V2 (12.5) = f rg D 2g
where f is the friction factor, which depends on the surface roughness of the pipe and the Reynolds number. For turbulent flows, we can also obtain a relationship between the volumetric flow rate and the pressure drop by substituting for V in terms of the flow rate in Eq. (12.5) and rearranging terms: Q2 =
1 p2D5 P1  P2 ¢ ≤ (12.6) f 8r L
When we compare turbulent flow to laminar flow, we note that for turbulent flow, the relationship between the flow rate and pressure drop is nonlinear. Pipes in Series For flow of a fluid through a piping network consisting of a series of pipes with respective diameters D1, D2, D3, c, as shown in Figure 12.3, the conservation of mass (continuity equation) requires that under s teadystate conditions, the mass flow rate through each pipe be the same: m1∙ = m2∙ = m3∙ = c = constant (12.7)
D1
Flow
D2
D3
Figure 12.3 Pipes in series.
M12_MOAV4303_04_GE_C12.INDD 713
27/11/14 10:15 AM
714 Chapter 12 Analysis of Fluid Mechanics Problems
Moreover, for an incompressible flow, the volumetric flow rate through each pipe that is part of a piping network in series is constant. That is, Q1 = Q2 = Q3 = c = constant (12.8)
Expressing the flow rates in terms of the average fluid velocity in each pipe, we obtain V1D21 = V2D22 = V3D23 = c = constant (12.9)
For pipes in series, the total pressure drop through a network is determined from the sum of the pressure drops in each pipe: ∆Ptotal = ∆P1 + ∆P2 + ∆P3 + c (12.10)
Pipes in Parallel
For flow of a fluid through a piping network consisting of pipes in parallel arrangement, as shown in Figure 12.4, the conservation of mass (continuity equation) requires that ∙ mtotal = m1∙ + m2∙ (12.11)
Moreover, for an incompressible flow, Qtotal = Q1 + Q2 (12.12)
For pipes in parallel configuration, the pressure drop in each parallel branch is the same, and is related according to ∆Ptotal = ∆P1 = ∆P2 (12.13)
Finite Element Formulation
Consider an incompressible laminar flow of a viscous fluid through a network of piping systems, as shown in Figure 12.5. We start by subdividing the problem into nodes and elements. This example may be represented by a model that has four nodes and four elements. m•1
•
m2
Figure 12.4 Pipes in parallel.
Q Qi + 1
Qi
i
(e)
i+1
Figure 12.5 A network problem: an incompressible laminar flow of a viscous fluid through a network of piping systems.
M12_MOAV4303_04_GE_C12.INDD 714
27/11/14 10:15 AM
Section 12.1 Direct Formulation of Flow Through Pipes 715
The behavior of the fluid flow inside a pipe section is modeled by an element with two nodes. The elemental description is given by the relationship between the flow rate and the pressure drop as given by Eq. (12.4), such that
Q =
pD4 Pi  Pi + 1 ¢ ≤ = C(Pi  Pi + 1) (12.14) 128m L
where the flowresistance coefficient C is given by
C =
pD4 (12.15) 128Lm
Because there are two nodes associated with each element, we need to create two equations for each element. These equations must involve nodal pressure and the element’s flow resistance. Consider the flow rates Qi and Qi + 1 and the nodal pressures Pi and Pi + 1 of an element, which are related according to the equations = C(Pi  Pi + 1)
Qi
Qi + 1 = C(Pi + 1  Pi) (12.16) The equations given by (12.16) were formulated such that the conservation of mass is satisfied as well. The sum of Qi and Qi + 1 is zero, which implies that under steadystate element conditions, what flows into a given element must flow out. Equations (12.16) can be expressed in matrix form by
b
Qi C r = J Qi + 1 C
pD4 C P 128Lm Rb i r = E C Pi + 1 pD4 128Lm
pD4 128Lm P U b i r (12.17) 4 pD Pi + 1 128Lm

The element’s flowresistance matrix is then given by
[R](e)
pD4 128Lm = E pD4 128Lm
pD4 128Lm U (12.18) pD4 128Lm

Applying the elemental description given by Eq. (12.17) to all elements and assembling them will lead to the formation of the global flow matrix, the flowresistance matrix, and the pressure matrix. Example 12.1 Oil with dynamic viscosity of m = 0.3N ∙ s/m2 and density of r = 900 kg/m3 flows through the piping network shown in Figure 12.6. The 2–4–5 branch was added in parallel to the 2–3–5 branch to allow for the flexibility of performing maintenance on one
M12_MOAV4303_04_GE_C12.INDD 715
27/11/14 10:15 AM
716 Chapter 12 Analysis of Fluid Mechanics Problems 1 L = 70.71 m D = 10 cm (1) L = 50.99 m D = 7.5 cm
2
3
(2) L = 50 m D = 7.5 cm
(3)
L = 53.85 m D = 5 cm
(4)
4
(5)
L = 60 m D = 10 cm
5
L = 70.71m D = 5 cm
6
(6)
Figure 12.6 The piping network of Example 12.1.
branch while the oil flows through the other branch. The dimensions of the piping system are shown in Figure 12.6. Determine the pressure distribution in the system if both branches are on line. The flow rate at node 1 is 5 * 104 m3/s. The pressure at node 1 is 39182 Pa (g) and the pressure at node 6 is 3665 Pa (g). For the given conditions, the flow is laminar throughout the system. How does the flow divide in each branch? The elemental flow resistance is given by Eq. (12.18) as
[R](e)
pD4 128Lm = D pD4 128Lm
pD4 128Lm T pD4 128Lm

We model the given network using six elements and six nodes. Evaluating the respective resistance matrices for elements (1)–(6), we obtain [R](1) = 109 J
115.70 115.70
[R](3) = 109 J
51.77 51.77
[R](5) = 109 J
7.23 7.23
M12_MOAV4303_04_GE_C12.INDD 716
115.70 1 R 115.70 2 51.77 2 R 51.77 4 7.23 4 R 7.23 5
[R](2) = 109 J
50.76 50.76
[R](4) = 109 J
9.50 9.50
[R](6) = 109 J
136.35 136.35
50.76 2 R 50.76 3 9.50 3 R 9.50 5 136.35 5 R 136.35 6
27/11/14 10:15 AM
Section 12.1 Direct Formulation of Flow Through Pipes 717
Note that in order to aid us in assembling the elemental resistance matrices into the global resistance matrix, the corresponding nodes are shown alongside of each element’s resistance matrix. So, we have 115.7 115.7 0 0 0 115.7 115.7 + 50.76 + 51.77 50.76 51.77 0 0 50.76 50.76 + 9.50 0 9.50 109 F 0 51.77 0 51.77 + 7.23 7.23 0 0 9.50 7.23 9.50 + 7.23 + 136.35 0 0 0 0 136.35
0 1 0 2 0 3 V 0 4 136.35 5 136.35 6
Applying the boundary conditions P1 = 39182 and P2 = 3665, we obtain 1 115.7 0 F 0 0 0
0 218.23 50.76 51.77 0 0
0 50.76 60.26 0 9.50 0
0 51.77 0 59.0 7.23 0
0 0 9.50 7.23 153.08 0
0 P1 39182 0 P2 0 0 P3 0 Vf v = f v 0 P4 0 136.35 P5 0 1 P6 3665
Solving the systems of equations simultaneously results in the nodal pressure values: [P]T = [39182
34860
29366
30588
2
3665]Pa
The flow rate in each branch is determined from Eq. (12.14): Q
=
pD4 Pi  Pi + 1 ¢ ≤ = C(Pi  Pi + 1) 128m L
Q(2) = 50.76 * 109 (34860  29366) = 2.79 * 104 m3/s Q(3) = 51.77 * 109 (34860  30588) = 2.21 * 104 m3/s Q(4) = 9.50 * 109 (29366  2) = 2.79 * 104 m3/s Q(5) = 7.23 * 109(30588  2) = 2.21 * 104 m3/s The verification of these results is discussed in Section 12.5.
M12_MOAV4303_04_GE_C12.INDD 717
27/11/14 10:16 AM
718 Chapter 12 Analysis of Fluid Mechanics Problems
Example 12.1 Revisited We will now show how to use Excel to set up and solve Example 12.1. 1. In cell A1 type Example 12.1, and in cell A3 type M= as shown. After inputting the value of m in cell B3, select B3 and in the “Name Box” type Mu and hit the Return key. Next, create the table shown with element and node numbers, length, diameter, and the fluid viscosity for each member. In cells D6:D11 input the length for each element and name them Length1, Length2, Length3, Length4, Length5, and Length6, respectively. Also, name the values in E6:E11, Diameter1, Diameter2, Diameter3, Diameter4, Diameter5, and Diameter6.
2. Next compute the constant portion of the flow resistance coefficient, that is the value, p>128m, for the [R1] matrix, as shown and name the selected range Celement1. In a similar way create the constant terms for [R2], [R3], [R4], [R5], and [R6] and name the selected ranges Celement2, Celement3, Celement4, Celement5, Celement6.
M12_MOAV4303_04_GE_C12.INDD 718
27/11/14 10:16 AM
Section 12.1 Direct Formulation of Flow Through Pipes 719
3. Next, create the [A1] matrix and name it Aelement1, as shown. If you have forgotten what the A matrices represent, see Section 2.5, Eq. (2.9). The nodal pressures P1, P2, P3, P4, P5, P6 and Pi andPj are shown alongside the [A1] matrix to aid us observe the contribution of node to its neighboring elements. Similarly, create the [A2], [A3], [A4], [A5], and [A6] matrices and name them Aelement2, Aelement3, Aelment4, Aelement5, Aelement6, as shown.
4. We now create the [R]1G, [R]2G, [R]3G, [R]4G, [R]5G, and [R]6G (with their proper positions in the global matrix) and name them K1G, K2G, K3G, K4G, K5G, and K6G (note they cannot be named R1G, R2G, etc., since they are not valid Excel names). For example, to create [R]1G, select B36:G37 and type = MMULT(TRANSPOSE(Aelement1),MMULT(((Diameter1^4>Length1)* Celement1),Aelement1)) and while holding down the Ctrl and Shift keys, hit the Return key. In a similar way, create [R]2G, [R]3G, [R]4G, [R]5G, and [R]6G as shown.
M12_MOAV4303_04_GE_C12.INDD 719
27/11/14 10:16 AM
720 Chapter 12 Analysis of Fluid Mechanics Problems
5. The final global matrix is created next. Select the range B101:G106 and type = K1G+K2G+K3G+K4G+K5G+K6G and while holding down the Ctrl and Shift keys hit the Return key. Name the range B101:G106, RG. Apply the boundary conditions. Copy the appropriate portion
M12_MOAV4303_04_GE_C12.INDD 720
27/11/14 10:16 AM
Section 12.1 Direct Formulation of Flow Through Pipes 721
of the RG matrix and paste it in the range C109:H114 as values only. Name the range RwithappliedBC. Similarly, create the corresponding load matrix in the range C116:C121 and name it FwithappliedBC.
6. Select the range C123:C128 and type = MMULT(MINVERSE(RwithappliedBC),FwithappliedBC) and while holding down the Ctrl and Shift keys hit the Return key.
M12_MOAV4303_04_GE_C12.INDD 721
27/11/14 10:16 AM
722 Chapter 12 Analysis of Fluid Mechanics Problems
The complete Excel sheet is shown next.
M12_MOAV4303_04_GE_C12.INDD 722
27/11/14 10:16 AM
Section 12.2 Ideal Fluid Flow 723
12.2 Ideal Fluid Flow All fluids have some viscosity; however, in certain flow situations it may be reasonable to neglect the effects of viscosity and the corresponding shear stresses. The assumption may be made as a first approximation to simplify the behavior of real fluids with relatively small viscosity. Also, in many external viscous flow situations, we can divide the flow into two regions: (1) a thin layer close to a solid boundary—called the boundary layer region—where the effects of viscosity are important and (2) a region outside the boundary layer where the viscous effects are negligible, in which the fluid is considered to be inviscid. This concept is demonstrated for the case of the flow of air over an airfoil
M12_MOAV4303_04_GE_C12.INDD 723
27/11/14 10:16 AM
724 Chapter 12 Analysis of Fluid Mechanics Problems Freestream velocity Inviscid region
Boundarylayer region An airfoil Boundarylayer region Inviscid region Figure 12.7 The flow of air over an airfoil.
in Figure 12.7. For inviscid flow situations, the only forces considered are those resulting from pressure and the inertial forces acting on a fluid element. Before discussing finite element formulation of ideal fluid problems, let us review some fundamental information. For a twodimensional flow field, the fluid velocity is S
u
u
V = nx i + ny j (12.19)
where nx and ny are the x and ycomponents of the fluid’s velocity vector, respectively. The conservation of mass (continuity equation) for a t wodimensional incompressible fluid can be expressed in the differential form in terms of fluid’s velocity components as 0ny 0nx + = 0 (12.20) 0x 0y
The derivation of Equation (12.20) is shown in Figure 12.8. The Stream Function and Streamlines For a steady flow, a streamline represents the trajectory of a fluid particle. The streamline is a line that is tangent to the velocity of a fluid particle. Streamlines provide a means for visualizing the flow patterns. The stream function c(x, y) is defined such that it will satisfy the continuity equation Eq. (12.20) according to the following relationships:
nx =
0c 0y
and ny = 
0c (12.21) 0x
Note that upon substitution of Eq. (12.21) into Eq. (12.20), the conservation of mass is satisfied. Along a line of constant c(x, y), we have
M12_MOAV4303_04_GE_C12.INDD 724
dc = 0 =
0c 0c dx + dy = ny dx + nx dy (12.22) 0x 0y
27/11/14 10:16 AM
Section 12.2 Ideal Fluid Flow 725
(rv)y + dy
(rv)x
(rv)x + dx
dy dx
(rv)y First, we begin by applying the conservation of mass to a small region (differential control volume) in the flow field. dmcontrol volume ∙ ∙ min  mout = dt rnx dy  ¢ r + ¢r +
0r 0y
0r 0x
0nx
dx ≤ ¢ nx +
dy ≤ ¢ ny +
0ny 0y
0x
dx ≤dy + rny dx 
dy ≤dx =
0r 0t
dxdy
Simplifying, we get  ¢ nx 
0r 0x
0rnx 0x
dx + r
dxdy 
0nx 0x
0rny 0y
dx ≤dy  ¢ ny dxdy =
0r 0t
0r 0y
dy + r
0ny 0y
dy ≤dx =
0r 0t
dxdy
dxdy
Canceling out the dxdy terms and assuming incompressible fluid (r = constant), 0nx 0x
+
0ny 0y
= 0
Figure 12.8 The derivation of continuity equation for an incompressible fluid.
or ny dy = (12.23) nx dx
Equation (12.23) can be used to determine the stream function for a specific flow. Moreover, Eq. (12.23) gives the relationship between the slope at any point along a streamline and the fluid velocity components. This relationship is shown in Figure 12.9. To shed more light on the physical meaning of Eq. (12.23), consider Figure 12.10 in
M12_MOAV4303_04_GE_C12.INDD 725
27/11/14 10:16 AM
726 Chapter 12 Analysis of Fluid Mechanics Problems
vy
V vx
vy
V vx
Figure 12.9 The relationship between the slope at any point along a streamline and the fluid velocity components.
Streamline
which the flow of a fluid around a sharp corner is shown. For this flow situation, the velocity field is represented by S
u
u
V = cxi  cyj
To obtain the expression for the stream function, we make use of Eq. (12.23): ny nx
=
dy cy = cx dx
Integrating, we have dy dx = y L L x Evaluating the integral results in the stream function, which is given by xy = constant or c = xy y
C (x1, y2)
B (x2, y2) c2 A (x2, y1)
c1 x
M12_MOAV4303_04_GE_C12.INDD 726
Figure 12.10 The flow of fluid around a sharp corner.
27/11/14 10:16 AM
Section 12.2 Ideal Fluid Flow 727
To visualize the trajectory of fluid particles we can plot the streamlines by assigning various values to c as shown in Figure 12.10. Note that the individual assigned values of streamlines are not important; it is the difference between their values that is important. The difference between the values of two streamlines provides a measure of volumetric flow rate between the streamlines. To demonstrate this idea, let us refer back to Figure 12.10. Along, the A–B section, we can write y2 y2 c2 0c Q nx dy = dy = dc = c2  c1 (12.24) = w Ly1 0y Lc1 Ly1
Similarly, along the B–C section, we have x2 x2 c2 0c Q ny dx = dx = dc = c2  c1 (12.25) = w Lx1 0x Lc1 Lx1
Therefore, the difference between the values of the streamlines represents the volume tric flow rate per unit width w. The Irrotational Flow, Potential Function, and Potential Lines As mentioned earlier, there are many flow situations for which the effects of viscosity may be neglected. Moreover at low speeds, the fluid elements within inviscid flow situations may have an angular velocity of zero (no rotation). These types of flow situations are referred to as irrotational flows. A twodimensional flow is considered to be irrotational when 0ny
0x

0nx = 0 (12.26) 0y
We can also define a potential function f such that the spatial gradients of the potential function are equal to the components of the velocity field:
nx =
0f 0x
ny =
0f (12.27) 0y
Along a line of constant potential function, we have
df = 0 =
0f 0f dx + dy = nx dx + ny dy = 0 (12.28) 0x 0y dy nx =  (12.29) ny dx
By comparing Eqs. (12.29) and (12.23), we can see that the streamlines and the velocity potential lines are orthogonal to each other. It is clear that the potential function
M12_MOAV4303_04_GE_C12.INDD 727
27/11/14 10:16 AM
728 Chapter 12 Analysis of Fluid Mechanics Problems
complements the stream function. Using the relationships in Eq. (12.27) to substitute for nx and ny in the continuity equation Eq. (12.20), we have 02f
0x2
+
02f 0y2
= 0 (12.30)
Using the definitions of stream functions as given by Eq. (12.21) and substituting for nx and ny in Eq. (12.26), we have 02c
0x2
+
02c 0y2
= 0 (12.31)
Equations (12.30) and (12.31), which are forms of Laplace’s equation, govern the motion of an ideal irrotational flow. Typically, for potential flow situations, the boundary conditions are the known freestream velocities, and at the solid surface boundary, the fluid cannot have velocity normal to the surface. The latter condition is given by the equation 0f = 0 (12.32) 0n
Here, n represents a direction normal to the solid surface. Comparing the differential equation governing the irrotational flow behavior of an inviscid fluid, Eq. (12.30), to the heat diffusion equation, Eq. (9.8), we note that both of these equations have the same form; therefore, we can apply the results of Sections 9.2 and 9.3 to the potential flow problems. However, when comparing the differential equations for irrotational flow problems, we let C1 = 1, C2 = 1, and C3 = 0. Later in this chapter, we will use ANSYS to analyze the flow of an ideal fluid around a cylinder. Now, let us briefly discuss the analysis of viscous flows. As mentioned earlier, all real fluids have viscosity. The analysis of a complex viscous flow is generally performed by solving the governing equations of motion for a specific boundary condition using the finite differencing approach. However, in recent years, we have made some advances in the finite element formulation of viscous fluid flow problems. Bathe (1996) discusses a Galerkin procedure for the analysis of the t wodimensional laminar flow of an incompressible fluid. For more details on the formulation of viscous laminar flows, also see the theory volume of ANSYS documents.
Example 12.2 Consider Figure 12.11, in which a simple uniform flow (constant velocity) in the horizontal direction is shown. As you would expect, for such flow, the fluid particles follow straight horizontal lines and thus the streamlines should be straight parallel lines. We are interested in using the theory discussed in the previous sections to obtain the equations
M12_MOAV4303_04_GE_C12.INDD 728
27/11/14 10:16 AM
Section 12.3 Groundwater Flow 729 y
y f1
V = 2 m/s i