You are on page 1of 375
FINITE ELEMENT PROCEDURES | IN ENGINEERING ANALYSIS | Adee. BATHE PRENTICE-HALL CIVIL ENGINEERING AND ENGINEERING MECHARICS SERIES FINITE ELEMENT PROCEDURES IN ENGINEERING ANALYSIS Prenricr-HALL, INC., Englewood Cliffs, New Jersey 07632 tary of Coe Cetin i Pilon Dat "intone pocfere in caletng sai a iri enc ad inn Editor pro production supervision and Interior design by: James M.Chege Manufacturing buyer: Joyce Levatino ‘Ant production: Bruce Kenslaar Allright rire. No part oF tis book tray be reproduced in any fot ey {ny means thou persion in from the publisher, os Printed inthe United States of America 20:19 18.17 16 15 1413.12 mcr Hat manasa ne, Lond ENTICE;HALL OF AUSTRALIA BY. Len Shey camera LOF CANADA, Lt0,, Toronto RENTICE-HALL OF Iota Puavate Limited, New Delhi PRENTICE-HALL OF JAPAN, INC., Tokyo pay PRENTICE-HaLL oF Sourieast Asta Pte. Lt, Singapore Wurreseatt Booxs Loatep, Welligion, New Zealand CONTENTS Preface xii MATRICES AND LINEAR ALGEBRA 1 Elementary Concepts of Matrices 2 11 Introduction 2 1.2 Introduction to Matrices 2 13 Special Matrices 4 1.4 Matrix Equality, Addition, and Multiplication byaScalar 5 1.5 Multiplication of Matrices 7 16 The Inverse Matrix 1 1.7 Partitioning of Matrices 13 18 ‘The Trace and Determinant of a Matrix, 15 References 18 Vectors, Matrices and Tensors 19 21 Introduction 19 22. Vector Spaces, Subspaces, and the Span ofa Matrix 19 23. Matrix Representation of Linear Transformation 25 24 Change of Basis 29 2.5. Matrix Representation of Variational Formulation 31 2.6 The Definition of Tensors 37 2.7. The Bigenproblem Av = do, with A Being a Symmetric Matrix 43 28 The Rayleigh Quotient and the Mimimax Characterization of Eigenvalues 53 29 Vector and Matrix Norms 58 References 62 PART Il THE FINITE ELEMENT METHOD 65 3 Some Basic Concepts of Engineering Analysis 66 3.1 Introduction 665 3.2. Analysis of Discrete Systems 67 321 Steady-state problems 67 322 Propagation problems 16 4.23. Eigenvalue problems 79 3.24 On the nature of solutions 86 3.3. Analysis of Continuous Systems 97 431 Differential formulation 94 33.2 Variational formulation 96 53.3 Weighted residual methods: Ritz method 3.4 Imposition of Constraints 1/0 References 1/3 4 Formulation of the Finite Element Method— Linear Analysis in Solid and Structural Mechanics 114 4.1 Introduction 114 4.2 Formulation of the Displacement-Based Finite Element Method 15 421 General derivation of finite element ‘equilibrium equations — 120 422 Imposition of boundary conditions. 139 423 Generalized coordinate models for specific problems 143 424° Lumping of structure properties and loads 42.5 Convergence of analysis results 165 426 Caleulation of stresses 177 102 162 43, Incompatible, Mixed and Hybrid Finite Element Models, Finite Difference Methods 178 43.1 Incompatible models 178 43.2. Mixed and hybrid models 182 43.4 Finite difference differential and energy methods References 19 5 Formulation and Calculation of lsoparametric Finite Element Matrices 194 Sl Introduction 94 5.2 Isoparametric Derivation of Bar Element Stiffness Matrix 195 5.3. Formulation of Continuum Elements 196 S31 Rectangular elements 197 53.2 Triangular elements 220 533 Blemens matrices in global coordinate systems Contents. 2a 187 54. Formulation of Structural Elements 234 SAL Beam elements 24 SA2 Plate and shell elements 251 5.5 Convergence Considerations 260 S51 Continuum elements 260 55.2 Suructural elements 263 5.6 Associate Element Families 265 5.6.1 Use of different interpolations 265 5.6.2 Addition of incompatible modes 266 5.7 Numerical Integration 268 5.7.1 Interpolation using a polynomial 269 57.2 The Newton—Cates formulas (one-dimensional integration) 270 5.7.3 The Gauss formulas (one-dimensional integration) 214 5.7.4 Integration in to and three dimensions 277 5,8. Practical Considerations in Isoparametric Element Calculations 278 58.1 Use of numerical integration 219 ‘S82 Caleulation of stresses. 288 583 Some modeling considerations 290 59 A Computer Program Implementation of Isoparamettic Finite Elements 294 References 297 Finite Element Nonlinear Analysis in Solid and Structural Mechanics 301 6.1 Introduction to Nonlinear Analysis 301 6.2 Formulation of the Continuum Mechanics Incremental Equations of Motion 3/4 621 The basic problem = 315 6.22 Swese and strain tensors 318 623 Total and updated Lagrangian formulations, ‘materiaily-nonlinear-only analysis 335. 6.3 Isoparametric Finite Element Discretization 347 6.3.1 Truss and cable elements 345 63.2 Two-dimensional axisymmetric, plane stain ‘and plane stress elements" 351 6.33 Three-dimensional solid elements 358 63:4 Beam elements 358 6.35 Plate and shell elements 371 6.4 Use of Constitutive Relations 379 64.1 Elastic material behavior 380 64.2 Inelastic material behavior with special reference to elasto-plasticity and creep 386 6.5. Some Practical Considerations 396 65.1 The general approach to nonlinear analysis 397 65.2 Collapse and buckling analyses 398 653 Element distortions 399 654 Order of numerical integration 401 References 405 Contents 7 Finite Element Analysis of Heat Transfer, Field Problems, and Fluid Flow 407 7.1 Introduction 407 7.2 Heat Transfer Analysis 407, 7.24 Governing classical heat transfer equations 7.2.2 Step-by-step incremental equations 410 7.2.3. Finite element discretization of heat transfer equations 412 7.3 Analysis of Field Problems 4/8 73.1 Seepage 418 73.2 Incompressible inviscid flow 419 73.3 Torsion 420 74 Analysis of Viscous Incompressible Fluid Flow 74d Velocity-pressure formulation 425 742 Penalty method formation 427 References 428 408 PART ill SOLUTION OF FINITE ELEMENT EQUILIBRIUM EQUATIONS 431 8 Solution of Equilibrium Equations in Static Analysis 432 8.1 Introduction 432 8.2 Direct Solutions Using Algorithms Based ‘on Gauss Elimination 434 $821 Introduction to Gauss elimination 434 $2.2 The Gauss elimination olution 439 823 Computer implementation of Gauss elimination — The active column solution 441 23 8.24 Cholesky factorization, static condensation, substructures, ‘nd frontal solution 449. 8.25 Solution of equations with symmetric nonpostine definite ‘coefficient matrices 458 83. Direct Solutions Using Orthogonal Mattices 83.1 The Givens factorization 471 83.2 The Householder factorization 475 84° The Gauss-Seidel Iterative Solution 478 85 Solution Errors 487 86 Solution of Nonlinear Equations 489 86.1 Newton-Raphson schemes 490 86.2 The BFGS method 492 84.5 Convergence eriteria 494 References 495 Contents, 70 10 Solution of Equilibrium Equations in Dynamic Analysis 499 9.1 Introduction 499 9.2. Direct Integration Methods 500 O21 The cena ifeene metod S22 Te Hone method 306 323 Te Wihon Ometiod 508 934 The Newnork metiod —31 | 423 Neca ofr nme eprtrs S18 ode Superpostion I 98 INT chee base to moda erred Aiplcenents 2. ns th mpg ected 319 5533 taba wth dig ced 52 94 Analy of Dict integration Methods $22 Qa” Dect gration pronation ‘disper 942 subi aeyae” 530 943 deca anys Al 944 Sone patent consideration 9.5. Solution of Nonlinear Equations in Dyeamic Anis 47 at Explanation 58 332 ima meranon 50 353 Soin wig mete siprposion 2 96 Solution of Nowstwtuat Problems 333 Retarenss S94 516 Preliminaries to the Solution of Eigenproblems 587 10.1 Introduction $87 ; 10.2 Fundamental Fats Used nthe Sluion or Eigenysems S52 roprs ofthe demos 360 1035 Thechraaesut pao of te eeprom cg int and oft cated conan! prem sh 1023 shine” 50 1023 Tngoraton of the gerald elepeten 9c ing ward form 73 103 Approximate Solution Tshigues 579 131 Si contactor SB M32 Raylegh fiona 56 1033 Component most gmihess 98 ins Lancome’ 8? 104 Solution Eros 398 elena oo Ccomrents 11 Solution Methods for Eigenproblems 607 1d Introduction 607 11.2 Vector Iteration Methods 609 1121 Inverse iteration 610 1122 Forward weration 616 i 1123 Shifting in vector iteration 619 ] M24 Rayleigh quotient iteration 623 INDEX 125° Matrix deflation and Gram—Schmidt orthosonalization 626 11.26 Some practical conserations concerning vector iterations 628 11.3 Transformation Methods 630 M131 The Jacobi method 631 112.2 The generalized Jacobi method 637 133 The Houscholder-QReinserse iteration soltion 64S 114 Polynomial teration Techniques 656 1141 Explicit polynomial iteration 656 1142 Implicit polynomial iteration 657 11.5 Methods Based on the Sturm Sequence Property 651, References 663, 12. Solution of Large Eigenproblems 666 t 12.1 Introduction 666 ‘ 122 The Determinant Search Method 667 1221 Preliminary considerations 668 1222 The solution algorithm 669 122.3 Some remarks on the implementation of the determinant search method 671 123° The Subspace Iteration Method 672 123. Preliminary considerations 615. j 123.2 Subspace eration 618 123.3, Starting iteration vectors 680 1234 Convergence 683 123.5 Implementation ofthe subspace iteration method 685. 123.6 Some remarks on the subspace iteration method 634 References 695 APPENDIX Implementation of the Finite Element Method 697 A.l Introduction 697 A.2 Computer Program Organization for Calculation of Structure Matrices 699 A21” Nodal point and element information read-in 699 A22 Calculation of element sifness, mas, ‘and equivalent nodal loads” 702 A223 Acsombloge of sinucture matrices TOL x Contents A3 Aa Calculation of Element Stresses 706 Example Program STAP 706 AA Data Input 10 computer program STAP AAD Listing of program STAP 114 References 725 708 re contents PREFACE are employed to represent ent regions of planes theedinensional, ator stel teat nthe eal evelopment Plane strain, axisymmetcic, heat anf seep, ld flow and According to three bs hich " sc areas in which knowledge is a ided into three parts. In the first See ae ee a ea leh are presented, Many alee nae ‘ng of the finite clement procedures presented later The second part of the by method and the man 00k comprises the formulation ofthe finite element rate ad the sumer prcedres wed to evaluate te tome meee tthe complete clement assemblage. Starting from. clement 7 Donlinear analyses of solids and structures eaten field and fivid mechanics problems are considered. Since the early use of finite Clement analysis, a large number of finite element formulations have been Geveloped and published. The objective in this book is not to summarize all the finite element formulations and models available, but rather to establish the basic and general principles and describe some finite clement procedures that are effective at present. In the last part of the book, methods for the efficient solution of the finite element equilibrium equations in static and dynamic analyses are presented. This phase of finite element analysis usually comprises most of the computer effort land, consequently, deserves a great deal of attention. It should be noted that ‘eause the numerical procedures employed in this phase largely determine the cost of analysis, they also determine whether an analysis can or cannot be per formed in practice. ‘Throughout the presentation, the aim is to establish the finite element pro- cedures using relatively simple and physical concepts, even at the expense of Josing some mathematical rigorousness, The principles and procedures are illustrated using about 250 worked-out examples, which are an integral part of the presentation, Some short computer programs are also included to demon strate the finite element procedures in a compact manner My primary aim in writing this book is to provide a tool for teaching finite clement procedures to upper-level undergraduate and graduate students in engineering. As such I am using the book at M.LT. Although the topic is finite clement analysis procedures, many of the numerical techniques are quite general and could be employed effectively in almost any method of analysis. In addi- tion, engineers using finite element computer programs should find much valu- able information in the text ‘A very difficult aspect of writing a book is to give references that appro- priately acknowledge the achievements of the various researchers in the fiel. ‘This is particularly difficult in this book, because the field of finite element analysis has been expanding very rapidly. I would like to extend my sincere ‘apologies to all those researchers whose work I have not referenced to a suff cient degree in this book. “The endeavor to write this book is a result of the excitement and challenge that I have experienced in working in the field of finite element analysis. 1 am ‘much indebted to Prof. E. L. Wilson of the University of California, Berkeley, —ay “finite-element-father"—for his active support of my early research work fn finite clement methods and his continued interest in my work. Since 1975 the Mechanical Engineering Department of MIT. has provided me with & stimulating environment for my teaching, research and writing this book, for which I am very grateful. [also would like to extend very special thanks to Lee W. Ho, who while my teaching assistant at M.LT. has been very helpful, and to Todnna Buttlar for her outstanding job in typing the manuscript. Thank you also to James Chege, the editor at Prentice-Hall for his helpful collaboration Finally, truly unbounded thanks are due to Zorka, my wife, who with her love and patience supported me in writing this book K.J, Bathe Prerace oe | MATRICES | AND | LINEAR ALGEBRA I ELEMENTARY CONCEPTS OF MATRICES 1.1 INTRODUCTION me copay sites lgcbra that are important in finite element analysis, a tae are presented in Chapter 2. “eement 1.2 INTRODUCTION To MaTRICES The efsivenss of sing mates in practical acai i ea Py coming he solution oasct finer smetanens eet Sy-dt yo 48) 46m 455+ ge] X14 4645 — dre =0 an = 45 + Su 20 2 where the unknowns are xy, Xs, i, and x4. Using matrix notation, this set of equations is written as $4 1 apny yo 46-4 afl ft ae) 1-4 6 4x5] ~] 0 o 1-4 sie! Lol 4 where itis noted that, rather logically, the coefficients of the unknowns (5, f ther in one array; the left-hand side unknowns (xy. x ,,and x.) and the right-hand-side known quantities are each grouped together in additional arrays. Although written differently, the relation (1,2) still reads, etc.) are grouped to) the same way’as (1-1), However, using matrix symbols to identify the arrays in (1.2), we can now write the set of simultaneous equations as Ax=b a3 where A is the matrix of the coefficients in the set of linear equations, x is th matrix of unknowns, and b is the matrix of kaown quantities; Le., ps4. oF ra ° ae x ' A= ; x=] ho ® as “ | 14 6 4 % ° o 1-45 te ol ‘The Formal definition of a matrix, as given below, seems now apparent. A matrix is an array of ordered numbers. A general matsix consists Definition fn numbers arranged in ms rows and 9 columns, giving the following array A [ 7 as) We say that this matrix has order m x m (mm by n). When we have only one row (m= I) or one column (n — 1), we call A also a vector. Matrices are represented in this book using boldface letters, and usually uppercase letters when they ate not vectors, On the other hand, vectors can be upper case or lower case boldface. We therefore see that the following are matrices Li} le cal for 22 1 “6 2 32 6 where the first and the last matrices are also column and row vectors, respectively ‘A typical clement in the ith row and jth column of Ais identified as a, €.2., in the first matrix in (1.4), a,, = 5 and ay, = —4. Considering the elements 2, in (1), we note that the subscript i runs from 1 to m and the subscript j runs from | ton. A comma between subscripts will be used when there is any risk of confusion, €.8, dys», jm OF to denote differentiation (see Chapter 6). In general, the utility of matrices in practice arises from the fact that we can Identify and sanipulate an array of many numbers by use of a single symbol see.1.2 ItmagucrioN to MATRICES 3 Relationships among large sets of numbers can be stated in a clear and compact way. Furthermore, if the relationships involve properly defined known and unknown quantities, such as in the solution of a set of simultaneous equations, it is possible to solve for the unknown quantities in a systematic manner. This is where the computer is of so much help, because it can calculate and manipulate the elements of matrices at very high speeds. However, we as the analysts are required to clearly define the variables in a problem, to prescribe the calculation procedures for the evaluation of the elements in the matrices involved, and to design the solution procedures 1.3 SPECIAL MATRICES Whenever the elements of a matrix obey a certain law, we can consider the matrix to be of special form. A real matrix is a matrix whose elements are all real. A complex matrix has elements that may be complex, We shall deal only with real matrices. In addition, the matrix will often be symmetric, whieh is the property defined next, Definition: The \ranspose of the om xm matrix A, written as AP, is obtained by interchanging the rows and colunms in A. If A = AT, it follows that the mumber of ‘rows and columns in A are equal and that yj = ay. Because m= 0, we say that Ais @ square matrix of order n, and because 34) = ay we say thar A is a symmettic matrix. Note that symmetry implies thar A is square but not vice versa; it a square matric need not be symmetric. A symmetric matrix of order 4 was, for example, the coefficient matrix A in(1.2). We can verify that AT — A by simply checking that a, = a, for i,j = | a Another special matrix is the identity (or unit) matrix Ty, which is a square ‘matrix of order n with only zero elements except for its diagonal entries, which are unity. For example, the identity matrix of order 3 is feo oy o10 an loot In practical calculations the order of an identity matrix is often implied and the subscript is not written, In analogy with the identity matrix, we also use identity (or unit) vectors of order n, defined as e,, where the subscript Findicates that the vector is the ith column of an identity matrix. We shall work abundantly with symmetric banded matrices. Bandedness means that all elements beyond the bandwidth of the matrix are zero. Because Ais symmetric, we can state this condition as ay 0 for j> it ms as) where 2m, +11 the bandwidth of A. As an example, the following matrix is symmetric banded matrix of order 5, The half-bandwidth m is 2 4 ELewenrany CONCEPTS oF Marnices cnae.1 32100 23410 a-lrasot as o1674 00143 Ifthe halF-bandwidth of a matrix is zer0, we have nonzero elements only on she diagonal of the matrix and denote it as a diagonal matrix. For example, the identity matrix is a diagonal matrix. In computer calculations with mate the elements of the matrices in high-speed storage. An obvious way of storing the elements of a matrix A of order m x1 is simply to dimension in the FORTRAN program an array A(M, N), where Mm and N 1 and store ceach matrix element ,, in the storage location A(LI). However, in many ealeu- fations we store in this way unnecessarily many zero elements of A, which are never needed in the calculations. Also, if A is symmetsic, we should probably take advantage of it and store only the upper half of the matrix, including the diagonal elements. In general, only a restricted number of high-speed storage Jocations will be available, and iis necessary to use dn effective storage scheme in order to be able t0 take into high-speed core the maximum matrix size possible. If the matrix is too large to be contained in high-speed storage, the solution process involves reading and writing on secondary storage, which can add signifi cantly to the solution cost. Fortunately, in finite element analysis, the system matrices are symmetsic and banded. Therefore, with an effective storage scheme, rather large-order matrices can be kept in high-speed core. Let us denote by A(I) the Ith element in the one-dimensional storage array A. ‘A diagonal matrix of order n would simply be stored as shown in Fig. 1.1(a): AW eu; Pate teee (1.10) Consider a banded matrix as shown in Fig, 1.1(b). We will see later that zero elements within the “skyline” of the matrix may be changed to nonzero clements in the solution process; for example, ays may be a zero element but becomes nonzero during the solution process (see Section 8.2.3). Therefore, we allocate storage locations to zero elements within the skyline but do not need to store zero elements that are outside the skyline. The storage scheme that will be used in the finite element solution process is indicated in Fig. [.1 and is explained further in the Appendix. s, we nced to use a scheme of storing 1.4 MATRIX EQUALITY, ADDITION, AND MULTIPLICATION BY A SCALAR We have defined matrices to be ordered arrays of numbers and identified them by single symbols. In order to be able to deal with them in a similar way as we deal with ordinary numbers, its necessary to define rules corresponding to those ‘which govern equality, addition, subtraction, multiplication, and division of tordinary numbers, We shall simply state the matrix rules and not provide mot vation for them. The rationale for these rules will appear later, a it will turn out Sec. 1.4 MATRIX EQUALIY, ADDITION, AND MULTIFICATION BY A SCALAR 5

You might also like