aEE STAUCTURAL imme ANALYSIS Second EditionMatrix Structural Analysis Second Edition William McGuire Professor of Civil Engineering, Emeritus Comell University Richard H. Gallagher Late Professor and President Clarkson Oniversity Ronald D. Ziemian “Associate Professor of Civil Engineering Bucknell University John Wiley & Sons, Inc. New York / Chichestr/ Weinkelm Brisbane /Singapore/ TorontoTA 6G M25 200 © werunmeany 208 050347968 ACQUISITIONS EDITOR Weyne Anderson MARKETING MANAOER Katherine epbzn PRODUCTION EDITOR Ken Santor CoviR DESIGNER Tyas Koga TELUSTRATION COORDINATOR — Sigmund MalinowskiEugene Aictio ELECTRONIGILLUSTRATIONS Precision Grephics Inc. “This book was set in 10152 Times Tem by UG ! GGS Information Services, Ine, and printed tnt bound by Hanlton Printing Compan, Inc. The cover was printed by Phoenix Color Corporation. ‘isto i pied on ase pope 6 “The paper in this book was manufactured by a mill whose forest management programs sin sostained yield horvesting ofits Gamberlands. Sustained yield harvesting principles ares thatthe numbers of trees cut each year does not exceed the einount of new growth Copycght © 200 Job Wiley & Sons, Tne. A sight reserved ‘No par ofthis publication maybe reproduced, stored in oseval este or transite ae Bray mean, elecons mectaaesh photocopying. record, senting uted under Setions 107 or 108 of the 1976 United States Coonan Ach nthont ier te prio ites permission of the Publisher, seb rRatie trough peyincnt of the sporopste percopy fe tothe Copyright ‘Gearance Center, 222 Rosewood Drive, Danvers, MA 01923, (508) 750840, fx (Gag) 7504470 Requests fo te Publier for permiston shot be eddresed to the eemions Department, Jon Wiley & Sons, Inc, 605 Third Aveave, Now York, NY 0158-032, (212) 830-013, fox (212) 880-608, Mai [email protected]. “Library of Congress Caaloging-In-Publlcation Dare MeOuit, Willie, 1920- ‘Matrix stvctral soalyss oy Wiliams McGuire, Richard Gallagher, Ropsld D Zemian. —2ad 0 Pe. Iaslades dex. ISBN 0-471. 129186 (oth: alk paper) 1, Structural analysis (Eaginecrng)—Matrc methods 1. Galisgher, Rcbud H. I Ziemian, Ronald D. 1H. Tide, TAGIZADS 2000 ear. sasss7 cr Printed io the United States of Ameria 3098765432Preface to the Second Edition ‘The first edition of this book was written 20 years ago, but our aims remain the same: to place proper emphasis on the methods of matrix structural analysis used in cutrent practice and to provide the groundwork for forthcoming practice and allied, more advanced subject matter. The present edition accounts for changes in practice that have taken place in the intervening years, and it incorporates advances in the art of analysis we regard as suitable for application now and of increasing importance to practitioners in the years just ahead ‘Among the major changes from the first edition is the addition of chapters on the nonlinear analysis of framed structures, treatment of the solution of nonlinear prob- lems, and packaging of a compact disk containing the computer program, MASTAN2. ‘The major reductions include the deletion of Chapter 7 on the flexibility method, Section 10.7 on the transfer matrix method, and Chapter 12 on the finite clement ‘method. The entire text and illustrative examples have also been edited extensively, ‘and several chapters have been repositioned. These changes deserve some explanation. ‘The reduction in coverage of the flexibility method was presaged by a comment in the preface to the first edition: “It is hardly used in practice.” Although the method is advantageous in certain circumstances and we believed it might receive greater atten: tion if cast in a moce efficient form, such has not been the case. Therefore, the de- velopment of flexibility method equations of global analysis has been deleted. But extensive coverage of the flexibility approach to the determination of clement force- displacement relationships has been retained because of the value that it has in many tances. Elimination of the global flexibility method formulation also led to recon sideration of the chapter oa the equations of statics and kinematies (Chapter 6 of the first edition) which was in large measure preparation for the identification of redundant forces in the flexibility method. That material has been deleted also. The other dele- tions—-material oa the transfer matrix method and the general finite clement concept— ‘wore made in the interest in retaining a reasonably wieldy volume. “The additions represent some departures from the spirit of the frst edition. In that edition’s preface it was noted that, in concentrating on the direct stiffness method, we were dealing with an accredited, widely practiced procedure, whercas an earlier gen- eration of texts frequently dealt with subject matter that was not fully crystallized. Nonlinear structural analysis has a long history and itis the subject of a number of books and some highly sophisticated computer programs, but it can hardly be called “erystallized.”" Many of its features are ad hoc—in the good sense of the term—and are in need of reduction to more routine procedures through further research and ‘generalization, Others may always require the expertise of specialists in analysis, as, ‘opposed to the design engineer who has many additional concerns. Nevertheless, itis believed to be timely, and indeed necessary, to include an introduction to nonlinear analysis in a text on tho analysis of framed structures, particularly those of the civil engineering variety. Techniques for its practical application are at hand, and there is, a growing awareness of their place in design. ‘Similarly in the fist edition we decided to exclude computer programs, citing in its preface their rapid obsolescence and the widespread availability of commercial pro- grams such as STRUDL. Advances ia all areas of computer technology aver the pastMi Preface to the Second Badition twenty years have changed tht. Particularly relevant to our subject isthe capacity for packaging structural analysis programs of significant power in an interactive graphic~ PSpposted medium suitable for running on readily available personal computers The computer is Bota substitite for mastery ofthe subject matter, but in the ability forthe {eer to control the terms of the analysis and to picture the results that itmow provides, itean be an invaluable eid to understanding the theory and appreciating the physical gnificance of the outcome. We believe these pedagogical advantages outweigh the / navitabilty of obsolescence and are therefore packaging with every copy of the text | MASTAN2!, an interactive graphies program with provisions for geometric aad ma- terial nonlinear analysis, as well 8& conventional linear elastic anlysis. MASTAN2 is based on MATLAB", a premicr software package for numerie computing and data { analysis. “Also, whereas we used SI units exclusively inthe fst edition, in this one many of the examples and problems-particulary those of a nonlinear natuce that have strong, Jimplications for design-—are ia Imperial units. We do this because the SI system has rer teen fully asimilated in practice, and we believe that appreciation of the physical Significance ofthe results is essential to their understanding, ‘us aims forthe present edition are explained further inthe following discussion of ‘the motivation and salient features ofthe respective chapters: “Three purposes are served by Chapter i, the introduction, iss, we present # concise sketch of the development ofthe subject. One intent of this history is to emphasize thet computerized structural analysis methods ate merely one part of a continuing process that extends back more than 150 years, Second, the role that computerized TTructural analysis bas played in the design of existing structures is outined. Finally, the computer capabilities themselves are tied to the programs written for structural fnalyis, Particular altention is given to the development and use of interactive com: puter graphics. ‘Chapters 2-5 represent closely allied subject matter related primarily to the direet stiffness method. Chapter 2 serves to define terminology, coordinate systems, and the host fandamtental notions of structural behavior, but itelo contain two developments ot great generality. The first i the basic character of elemental relations in the form of aiffnets and fesbiity and theistransformabilty from one to another and even to Shernative formats. The second is the fundamental idea of direct stifiness analysis, described here by means ofthe simplest structural element. "A more forial treatment is given to direct stiffness analysis in Chapter 3 and, con- sequently itis possible (o examine more closely the implications for large-scale prac- tical computation. The later include considerations such a8 the characteristics ofthe igebraie equations that are to bé formed and solved. Ia Chapters 4 and 5 the emain- ing tools needed forthe linear elastic stifiess analysis of complete frames are estab- ished, The stifiness matrix ofa rather general space-frame clement is formulated and | then applied, in illustrative examples, toa variety of specic situations “Two comments on Chapters 2-5 are appropriate, The frst is that, except for some ‘editing ofthe text and modification of the examples and problems, they are identical to the same chapters of the first edition. The second is that this intial development of the subject features the basic physical conditions of structural analysis: equilibriam, compatibility, and the mechanical properties of the material. Reliance on work and cnergy concepts is kept to a minimum. Our feeling is that it is preferable to help the aa name MASTAN® iso tony for Matix Scat Ay, 2 Bion. a developed ye | a MAN aposon hws lees oan exang commer sector ae Seweat ace rors Sea ae re wademuk of he MatWors lac 9 Ape Hi Dee Natick MAOI709098 |Fa Preface tothe Second Bilton student become attuned to the overall approach to computerized structural anal this way, before proceeding to the more powerful and versatile—but more mathe- ‘matical—virtual work-based methods. ‘We do introduce the formulation of matrix structural analysis on the basis of virtual work concepls in Chapters 6 and 7. A treatment of this type is necessary to give ‘maximum scope to various aspects of practical design analysis of frameworks, such as tapored members and distributed loads, as well as to the study of nonlinear analysis and the finite element approach. The theoretical groundwork of the virtual work prin- ciple is Iaid in Chapter 6. Both virtual displacement and virtual force concepts are covered, but far greater attention is given to the former because oftheir role in stiffness formulations. Chapter 7 examines the implementation of the virtual work principle in linear elastic matrix structural analysis. ‘Chapters 8-10 comprise an introduction to the nonlinear matrix analysis of frame- works. The approach is one of using selected structural systems and types of material as the basis for explaining how theory can be translated into procedures useful in the design office. In the interest of focusing on sousces and end results in these chapters, wo defer to Chapter 12 the treatment of methods and details of nonlinear equation solution and to appendices the coverage of some aspects of the basic structural me- chanics and behavior An overview of sources of structural nonlinearity, levels of analysis used in dealing ‘with them, and a system for casting the required equations in matrix stiffness method forms presented in Chapter & In the course of doing this the notions of the geometric stiffness matrix and a plastic reduction matrix are introduced. Also, the range of con- ditions to be considered is illustrated by the solution of some elementary problems by traditional methods. In Chapter 9, the groundwork prepared in the previous chapter is used as the base for the development and application of a matrix method for the nonlinear elastic (geometric nonlinear) analysis of both planar and space frames. The calculation of the elastic critical load of ideal systems ig also trcated. Chapter 10 em- ploys the same base, plus the concepts of plastic hinge theory, for the material and geometric analysis of planar systems. Included is the determination of the inelastic critical load of ideal systems by an adaptation of tangent modulus theory. ‘Some of the more popular methods of solution of linear algebraic equations are stud- ied in Chapter 11, and the same is done for nonlinear equations in Chapter 12. Equation solving is more in the realm of applied mathematics, numerical analysis, or computer science than in structural engineering. Nevertheless, since primary responsibility for the entire analysis generally belongs to structural engineers, they should have more than superficial knowledge of the important aspects of equation solving, Operations peculiar to the handling of equations in MASTAN2 are explained in detail. Otherwise, ‘these chapters are introductions to the subject that summarize useful methods, men- tion pitfalls that may be encountered, and give guidance to more specialized sources. In the interest of developing basic stiffness methods of analysis inthe earlier chap- ters, we put off consideration of some procedures essential to the efficient and realistic solution of practical design analysis problems, Chapter 13 is devoted to the exposition of certain of these practices, including condensation of analysis equations prior to solution by elimination of specified unknowns, substructuring, the imposition of con- straints, connection stiffness and finite joint size, the exploitation of symmetry and antisymmetry, and procedures for the economical reanalysis of structures when i an iterative design sequence. Attention is aso given to various types of coordinate systems that are alternative to global coordinates, such as the local coordinate systems that are convenient for sloping supports. Tn Appendix A, principles of structural mechanics and geometry fundamental to solving the problem of equilibrium of a deformed structure are reviewed and thenPreface to the Second Faition employed inthe derivation of terms that were incomporated without proof in Be 8°2° cori atifness matrices of Chapter 9. There are several reasons for placing this ma. ween an appendix rather than inthe main body of the text. Throughoul the text Wo ter ap madd matrix notaGion, Concise treatment of the appended materia} atts ute of some elementary concepts of tensors and india notation, Wk ols rehltes the use of some less abvious definitions of sess and strain, Heh a8 second aise streaes and Green-Lageangesiins, and an excursion into the analyst avnite rotations, These changes in style and alteration of the analytical base aro n° Chutnuce but we believe their introduction in the carliex chapters would bave bes a8 aoa anyintrsion on the main agrument. Nevertheless, although iis materi} thet rane vental to an initial study, it is fundamental to a true understanding of the Prosesses of nonlisear analysis and to their extension beyond the Fins explored in the main text. ‘che problem of nonuniform torsion in three-dimensional frameworks &2Iso At drened in ths appendix but for a somewhat diferent reason. Practically, is key, dress inthe resistance of many systems to the destabilizing effects of combined daar and flexure. Failure to treat the subject would be a serious omission, Bt Con, Sp tis «presented in an appendix rather than i the main body ofthe fext borause tony aove of a cutique rather than the presentation of methods, something that was done in Chapter 12. ascents outined here represent subject matter on the scale of two three credit: hour ooe-semester courses, with considerable selectivity of coverage and latitude in wont goment available to the instuctr. For oxample, the frst seven chapters plus Mocted portions of Chapters 11 and 13 might serve for an introductory union” oF seer evel couse. At the same level, but for classes of students with @ previous, senor Gatrcdection tothe subject, the first ive chapters could serve for review and are amoat of understanding of basic concepts, with formal iastruetion stating with Te tdnal work approach (Chapters 6 and 7). The second level course, atthe elestive the vir guroductory graduate level, might feature the virtual work approach co both Fane of nonlinear elastic analysis, material nonlinear analysis, and stability studies (Chapters 6-10, Chapter 12, additional portions of Chapter 13, and reference (¢ Ne speakers), The program MASTANZ.can be a valuable adjunct at ll levels for deh apPWation, for review of the text's mumerous examples, for problem solving, as the career analysis and dosign studies, and in support of individual programming. An Mtactive application is to have students develop their own programs, ethes a8 olor oe to the analysis toutines in MASTAN2 or as additions to the programs existing Our co-author, Professor R. H. Gallagher, died in September 3997, while this edition was in preparation, The idea forthe original book was his, and be was active ia the planning ofthis edition and in the revision of soveral chapters, His contsibations ar¢ aie to the work. The remaining authors prepared the material on nonlinear analysis Jand we bear sole respoasibility for any of its shortcomings. William McGuire Ronald D. ZiemianAcknowledgments We are indebted to a number of people for their help in the preparation ofthis edition, Our work would not be complete without an acknowledgment of this debt and a par- ticular offering of thanks to the following: ‘To Professors J. F. Abel and G. G, Deierlein for using drafts as classroom notes and for their wise advice throughout. To Professors F. C. Filippou, J. F. Hajjar, EB. M. Lui, and G. H. Paulino for their thorough, constructive reviews of an early draft and to Professor K. M. Mosalam for the benefit of his experience in using it in class. ‘To the men whose Comell graduate research contributed to the development of ‘material inthe book: J. L. Gross, CI. Pesquera, M. Gattass, J.G. Orbison, Y.B. Yang, S. 1 Hilmy, J. L. Castatier, 8. N, Sutharshana, D. W. White, C. S. Chen, C. Chrysos- tomou, and M. R. Attala, ‘To Ms. N. Bulock and her colleagues at the MathWorks, Inc. and Professor J. Maneval of Bucknell University for their MATLAB support in developing /MASTAN2, To Ms. P, Welzel for her secretarial assistance in preparing the manuscript for publication.Contents Chapter 1 Chapter 2 Chapter 3 Chapter 4 Symbols xvii Introduction 1 11 A Brief History of Structural Analysis 2 1.2 Computer Programs 3 1.2.1 Computational Flow and General Purpose Programs 4 122 The Program MASTAN2 6 References 8 Definitions and Concepts 9 2A Degrees of Freedom 9 22 Coordinate Systems and Conditions of Analysis 11 23. Structure Téealization 14 24 Axial Force Blement—Force-Displacement Relationships 16 ‘24.1 Element Stiffness Equations 16 242. Blement Flexibility Equations 18 25 Axial Force Element—Global Stiffness Equations 19 2.6 Examples 20 27° Problems 26 Formation of the Global Analysis Equations 31. 34 Direct Stiffness Method—The Basic Equations 31 32. Direct Stiffness Method—The General Procedure 39 33. Some Features of the Stiffness Equations 46 34 Indeterminacy 47 35 Problems 49 References 55 Stiffness Analysis of Frames—I 56 44 Stress-Strain Relationships 56 42 Work and Enorgy 58 43° Reciprocity 60 44° Flexibility Stifiness Transformations 62 44.1 Stiffness to Flexibility Transformation 62 442. Flexibility to Stitiness Transformation 63si Contents Chapter 5 Chapter 6 45. ‘The Framework Element Stiffness Matrix 65 46 45. Axial Force Member 67 452 Pure Torsional Member 67 453. Beam Bent About Its z Axis 68 454 Beam Bent About Its y Axis. 72 455 Complete Element Stiffness Matrix 73 ‘A Commentary on Deformations and Displacement Variables 74 461 Neglected Deformations 74 462 Displacement Variables 75 47 Bxamples 76 48. Problems 88 References. 92 Stiffness Analysis of Frame—II 93 5.1 Coordinate Transformations 93 52 53 5.11 ‘Transformation Matrices 94 5.12 Transformation of Degrees of Freedom 98 513 ‘Transformations and Energy 98 5.14 Rectangular Transformation Matrices 100 Loads Between Nodal Points 108 Conditions 120 53.1 Initial Stcain Problems 122 532. Thermal Strain Problems 125 SA. Problems 130 ‘Virtual Work Principles 137 64 62 63 64 65 Principle of Virtual Displacements—Rigid Bodies 138 Principle of Virtual Displacements—Deformable Bodies 142 Virtual Displacement Analysis Procedure and Detailed Expressions 144 63.1 General Procedure 144 632 Internal Virtual Work 145 633. External Virtual Work 148 Construction of Analytical Solutions by the Principal of Virtual Displacements 150 64.1 Exact Solutions 150 642 Approximate Solutions and the Significance ofthe Chosen Virtual Displacements 153 Examples 154 Principle of Virtual Forces 158 65.1 Equations of Equilibrium 158 65.2 Characteristics of Virtual Force Systems 161Chapter 7 Chapter 8 Contents xit 653 Formulation of the Virtual Forces Principle 162 654 Construction of Analytical Solutions by the Virtual Forces Principle 164 66 Problems 170 References 173, Virtual Work Principles in Framework Analysis 174 7A Description of the Displaced State of Elements 175, 7.4 Definition of the Shape Function Mode of Desc 712 Formulation of Shape Funetions 176 713 Characteristics of Shape Functions 180 7.2. Virtual Displacoments in the Formulation of Elerient Stiffness Equations 181 72. Construction of Expressions for Real and Virtual Displacements 181 72.2 Virtual Displacements Formula for an Element Stiffness Matrix 182 723 Application to Standard Axial, Torsional, and Flexural Elements 183, 73 Nonuniform Elements 184 74° Nonuniform Torsion 186 741 Au Blemeat Stiffness Matrix 187 742 Application and Examples 190 75. Loads Between Nodal Points and Initial Strain Effect—A General ‘Approach 194 76 Virtual Forces in the Formulation of Blement Force-Displacement Equations 199 7.64 Construction of Blement Equations by the Principle of Virtual Forces 199 762. Further Applications—Shearing Deformations and Curved Elements 203 4.7 Problems 209 References 214 175 Nonlinear Analysis of Frames—An Introduction 216 841 Nonlinear Behavior, Analysis, and Design 216 841.1 Sources of Nonlinearity 217 81.2 Levels of Analysis 218 813 Examples from Established Theory 220 81.4 A Commentary on Stability 230 82 A Matrix Approach 234 83. Tho Equations of Analysis and Their Solution 235, 83.1 Equation Solution—The Options 236 832 A Fundamental Problem 237 84 Problems 239 References 241dv Contents Chapter 9 Chapter 10 Chapter 11. ‘Geometric Nonlinear and Elastic Critical Load Analysis 242 91 Geomettie Stifiness Matrices for Planar Elements 243 9.411 Axial Force Member 243, 9412 Combined Bending and Axial Force 245 91.3 Examples of Plane Structure Analysis 246 92 Combined Torsion and Axial Foree 254 93. Three Dimensional Geometric Nonlinear Analysis—An Overview 257 94 Bxamples of Three Dimensional Structure Analysis 259 95 Problems 263 References 268 Material Nonlinear Analysis 269 40.1 Nonlinear Material Behavior 269 40.11 Plasticity Theory 270 10.12 Plastic Analysis 272 401.3 Further Considerations 273 102 A Plastic Hinge Method For Ductile Frames 274 1021 ‘The Yield Surface and a Plastic Reduction Matrix 274 40.22 Defiition of the Yield Surface 277 403 Inelastic Critical Load Theory 278 104 Examples 279 105 ‘The Yield Surface Concept—A Brief Survey of Further Applications 290 105. Spread of Plasticity 290 1052 Multiple Yield Surfaces 292 1053. Reinforced Concrete Members 293 106 Problems 294 References 299 Solution of Linear Algebraic Equations 301 41.1 ‘The Basic Choice—Direct Inversion versus Elimination or Iteration 302 112 Direct Elimination Methods 302 4121 Gauss Elimination 304 11.22. Cholesky Method 307 11.3 Iterative Methods 310 113. Gauss-Seidel and Jacobi Iterations 310 1132. Conjugate Gradient Method 312 114 Sparseness and Bandedness 316 ALS Frontal Solvers 317 116 Bross 322 116.1 ‘The Condition Number 326 1162. Estimate of Condition Number 327Chapter 12 Chapter 13 7 11.63 Error Estimates and Preconditioning 329 11.64 Detecting, Controlling, and Correcting Error 333 Problems 325 References 337 Solution of Nonlinear Equilibrium Equations 339 124 222 123 4 ms 126 127 2s ns Incremental Analysis 339 Incremental Single-Step Methods 340 122.1 Buler Method 341 1222 Second-Order Runge-Kutta Methods 341 Incremental-Iterative Methods 345 1231 Load Control Method 347 1232 Displacement Control Methods 348 42.33 Work Control Method 349 1234 Constant Arc Length Methods 349 123.5 Modified Iterative Technique 349 123.6 Convergence Criteria 350 ‘Automatic Load Incrementation 351 3241 Change in Stiffness 351 1242 Number of Iterations 352 Element Result Celeulations 352 125.1 Updating Deformed Geometry 352 32.52 Foree Recovery 352 Plastic Hinge Constraints 355 ‘Limit Point and Post-Limit Analysis 358 Citieal Load Analysis—An Higenproblem 360 128 Reduction to Standard Form 361 1282 Polynomial Expansion 363, 1283 Power Method 364 1284 Inverse Iteration 367 Problems 370 References 373, Special Analysis Procedures 375 Bt 132 133 4 BS 136 Condensation 376 Substructuring 377 Constraints 385 Joint Coordinates 388 Connections and Joints 393 1351 Flexible Connections 393 1352 Finite Joint Size 398 Symmetry and Antisymmetry 403, Contents xvxvi Contents 13.7 Reanalysis Techniques 408 138 Problems 413 References 418 Nonlinear Analysis—A Further Look 419 "AL Virtual Displacement Principles in Lagrangian Formulations 420 ‘A? An Updated Lagrangian Formulation and lis Linearization 425 'A3. Application to the Framework Element 426 ‘A's Fite Rotations and Equilibrium in the Deformed Configuration 431 ‘A441 ‘Tho Finite Rotation Equation 422 ‘$42 Application of the Finite Rotation Equation to the Framework Element 433 AS Nomuniform Torsion 437 References 441 Appendix A. ‘On Rigid Body Motion and Natural Deformation 444 BA The Nature of the Problem 444 382. Distinguishing Between Rigid Body Motion and Natural Deformation 445 3. Citique of Fores Recovery Methods 447 B31. The Natural Deformation Approach 447 B32 The Elementary Approach 48 References 450 Appendix B AuthorIndex 451 Subject Index 453Symbols {In mattix structural analysis, many physical quantities and mathematical operations must be represented symbolically, Preparation of the equations of analysis in a form suitable for computer solution requires that all symbols used be defined in a rigid fashion amenable to numerical interpretation. On the other hand, the development of these equations, with stress upon their physical significance, is often accomplished through the use of simple, les formal symbols-symbols that vary with the principle under discussion and have a clear physical connotation in the case at hand. In the interest of generality and uniformity, we shall use some basic symbols to denote certain {quantities throughout the text, But the precise interpretation of any of these symbols tust be obtained from the local contoxt in which it is used, and in which it will be explained. ‘Ta general, we use the letter P to designate applied direct forces and P,, to designate applied moments. Similarly, R and R,, are used for ditect and moment reactions. At their cnds and within elements, we generally denote direct forces by F and moments by M. All ofthese symbols may carry clarifying subscripts and superscripts to indicate direction, point of application, or member to which the symbol applies. The symbols may appear in either single component or vector fora. Tho symbols tv, and w will designate translational displacements in the x, y, and z directions and @,, 8, and 8, rotational displacements about these axes. Generally the letter kwill refer toa stiffness quantity and d to a flexibility quantity. Matrices are denoted by a boldface letter within the symbols [} (for a square or rectangular matrix), { } (for a column vector), | (for a row vector), and [| for a Giagonal matrix. Matrix operations are represented by [ J" as the inverse [ J"as the tuanspore, | [as a determinant, and ||] as a norm, ‘As a forther guide, the following isa ist ofthe principal symbols used in the text ‘As indicated above, most of these may contain clarifying or modifying subscript, for supplementary marks (overbars hat symbols”, etc.) that will be defined ia con- text. The same applies to the individual components of matrices and vectors listed below. A Area B Bimoment Cy Warping constant fa lement flexibility matrix DP) Global flexibility matrix, diagonal matrix E,{E] Elastic modulus, matrix of elastic constants e ‘Normal strain e Infinitesimal strain tensor F ‘Normal or shearing force 1) ‘Vector of element nodal forces wileill Ssiabols (¢) (6 HW AYE Shear modialus Gradiont vector Constaint equation coeficients, Hialt-bandbwidth Momeat of inertia Polar moment of inertia Jdentity matrix St. Vensmnt torsion constant Spring stifiness, rotational spring stifiness ‘Flement stifiness matrix Global stiffness matrix Length Lower triangular matrix Direction cosines Ratio of beading moment to plastic moment Bending or twisting moment Plastic moment Vector of clement shape functions [Number of degrees of freedom, number of nodes, [Null matrix and vector Vector of global nodal forces Axial force ‘Axial yield or squash load Ratio of axial force to squash load, number of elements ‘Vector of element force distribution functions Distributed load intensity ‘Vector of reaction forces, unbalanced load vector Residual vector Ratius of gyration ‘Current siffaess parameter Basis vector “Temperature change from stress-free state, torsional moment ‘Cauchy stress tensor Second Fiola-Kirchoff stress tensor Force transformation matrix Strain energy and complementary strain energy Upper triangular matrix Displacement components Potential energy of applied lords Volume ‘Work and complementary work Cartesian coordinates Plastic section modulusSymbols xix Greek Symbols @ (Coofficient of thermal expansion, factor defied in context @B,8 Direction angles B Rate of twist, relaxation factor []. 7] Transformation matrix 7 Shear strain fa) ‘Vector of nodal point displacements 4 Displacement 8 ‘Virtual quantity, variation e Green Lagrange strain tensor fe Absolute percent relative error 8 Angular displacement « ‘Curvature, condition number a Eigenvalue, load ratio, plastic strain factor, effective length factor » Poisson’s ratio £ ‘Nondimensional coordinate 1 ‘Total potential energy 7 3416. e Radius of curvature r Shear stress, fraction of load increment © ‘Normal stress ° ‘Yield stress my igenvecior ® ‘Yield surface function [a Static equilibrium matrix 4 Angle of measure £ ‘Acceptable percent error tolerance o igenvelue In addition to the above literal and matrix symbolism, we shal! use the following graphic symbols wherever itis desired to indicate or to stress some particular charac- teristic of a force or structure, > +> _ Direct force and moment components Me wee Direct force and moment resuftants p> > Reactive force and moment components io c Alternative representation of momentsax Symbols inal Pinned joint Rigid joint Pinned connection Roller support Pinned support Elastic support Fixed supportChapter 1 Introduction One of the responsibilities of the structural design engincer is to devise arrangements and proportions of members that can withstand, economically and efficiently, the con- ditions anticipated during the lifetime of a structure. A central aspect of this function is the calculation of the disteibution of forces within the structure and the displaced state of the system. Our abjective isto deseribe moder methods for performing these calculations in the particular case of framed structures. ‘The number of structures that are actually simple frameworks represents only a part of those whose idealization in the form of a framework is acceptable for the purposes of analysis. Buildings of various ‘types, portions of aerospace and ship structures, and radio telescopes and the like can often be idealized as frameworks. In design, both serviceability limit states and strength limitstates should be considered. A serviceability limit state is one in which the structure would become unfit for normal, service because of excessive deformation or vibration, or problems of maintenance or durability. Generally, linear elastic analysis gives a good picture of the actual response. to service loads. A strength limit state is one in which the structure would become unsafe. Except for sudden brittle or fatigue crack propagation, the attainment of a strength limit is generally the result of increasingly nonlinear elastic or inelastic re- sponse that culminates in structural instability. Most often, the internal load distribu- tion at a strength limit state is calculated by a linear method. In proportioning mem- bers, the analytical results are modified in an empirical or judgmental way to account for the offects of noalinearity. Whereas this practice is common, the availability and application of practicable mothods of nonlinear analysis is steadily increasing, One of the aims of this book is to provide an introduction to typical methods of this type. Fundamentally, the behavior of all types of structures—frameworks, plates, shells, or solids—is described by means of differential equations. In practice, the writing of differential equations for framed structures is rarely necessary. It has been long estab- lished that such structures may be treated as assemblages of one-dimensional members. Exact or approximate solutions to the ordinary differential equations for each member are well-known, These solutions can be cast in the form of relationships between the forces and the displacements at the ends of the member. Proper combinations of these relationships with the equations of equilibrium and compatibility at the joints and supports yields a system of algebraic equations that describes the behavior of the structure. ‘Structures consisting of two-or three-dimensional components—plates, membranes, shells, solids—are more complicated in that rarely do exact solutions exist for the applicable partial differential equations. One approach to obtaining practical, numer- ical solutions is the finite element method. The basic concept of the method is that a 12 Chapter 1 Introduction continuum (the total structure) can be modeled analytically by its subdivision into ‘egions (the finite elements), in each of which the behavior is deseribed by a set of assumed functions representing the stresses or displacements in that region. This pes- nls the problem formulation to be altered to one of the establishment of a system of algebraic equations. . “The practical, numerical solution of problems in structural analysis thus is seen to involve the formation and solutions of systems—sometimes very large systems-—of algebraic equations. Als, it should be fairly clear that a member ofa framed structure js simply one example of 2 more broadly defined family of finite elements. ‘Viewed in this way, structural analysis may be broken down into five parts: L. Basie mechanics. ‘The fundamental relationships of stress and strain, compat bility, and equilibrium. +2, Finite element mechanics. ‘The exact or approximate solution of the differential equations of the element 4, Equation formulation. The establishment of the governing algebraic equ of the system. 4, Equation solution. Computational methods and algorithms. ‘5, Solution Interpretation. The presentation of results in a form useful in design. book deals chiefly with parts 3, 4, and 5. Specifically, it is on matrix siructural analysis. Ths isthe approach to these parts that currently seems to be most suitable for automation of the equation formulation process and for taking advantage of the powerful capabilities of the computer in solving large-order systems of equations, An inderstanding of basicstructural mechanics and basiematrix algebra is presumed, Only that segment of the finite element method relating to framed structures will be in~ luded, other aspects being left to texts specializing in the fundamentals of finite ele- Jnent mechanics and procedures (eg, Refs. 1-1 and 1.2). Computational methods and algorithms wil be discussed in Chapters 11 and 12, but more comprehensive coverage can be found in books on numerical analysis. (eg, Ref. 13). "An appreciation of the approach to structural analysis we are taking requires some understanding of the history of this and related subjects. The following bricf review may help. 14 A BRIEF HISTORY OF STRUCTURAL ANALYSIS immediately preceded by the great accomplishments of the school of French elasticians, such as Navier and St. Venant, the period from 1850 to 1875 is a logical starting point for our review. The concept of framework analysis emerged dur ing this peviod, through the efforts of Maxwell (Ref. 1.4), Casigiano (Ref. 1.5), and ‘Mohr (Ref. 1.6), among others. At the same time, the concepts of matrices were being, introduced and defined by Sylvester, Hamilton, and Cayley (Ref. 1.7). Thése concepts are the foundations of matrix structural analysis, which did not take form until nearly £0 years later. ‘An excellent chronicle of developments in structural mechanics in the period 1875 10 1920's found in Timoshenko's History of Strength of Materials (Ref. 1.8). Very litle progress wes made in this period in the development of theory and the analytical {echiniques subsidiary to matrix structural analysis. To a great extent, this was due to prectical imitations on the solvability of algebraic equations with more than a few Gntnowns, For the structures of primary interest in that period—trusses and frames—an analysis approach based on member forces as unknowns was almost uni- vversally employed.12 Computer Programs 3 ‘Around 1920, asa result of the efforts of Maney (Ref. 19) in the United States and Ostenfeld (Ref. 1.10) in Denmark, the basic ideas of a truss and framework analysis, approach based on displacement parameters as unknowns took form. These ideas represent the forerunners of matrix structural analysis concepts in vogue today.Severe limitations on the size of the problems that could be handled by either force or dis- placement unknowns nevertheless prevailed until 1932 when Hardy Cross introduced ‘the method of moment distribution (Ref. 1.11). This method made feasible the solution of structural analysis problems that were an order of magnitude more complex than ‘the most sophisticated problems treatable by the prior approaches. Moment distribu- tion became the staple of structural frame analysis practice for the next 25 years. Today it remains of value as a ready method for solving small problems and as an aid in visualizing bebavior, but it is primitive in power compared to computer methods and thas been superseded by them in the solution of large problems. ‘Association of the mathematical concepts of matrix theory with the engineering concepts of structural analysis appeared in the 1930s, chiefly in the work of Frazer, Duncan, and Collar (Ref. 1.12). The liaison developed erratically through the 1940s, Dat there was no motivation for a firm union until digital computers were born in the carly 1950s, Individuals who foresaw the impact of computers on both theory and practice then undertook tho codification of the well-established framework analysis procedures in the format best suited to the computer, the matrix format. Two note Worthy developments were the publications of Arayris and Kelsey (Ref. 1.13) and ‘Tumer, Clough, Martin, and Topp (Ref. 11d). These papers wedded the concepts of framework analysis aud coritinuum analysis and cast the resulting procedures in a ‘matrix format, They were a strong influence on developments in subsequent years. An {important factor in the acceptance of computerized structural analysis in the civil en gineering profession has been the series of ASCE conferences on electronic compu- tation, especially the fist three (Refs. 1.15-1.17). ‘The desirability of accounting for nonlinear elastic and inelastic behavior was rec- ‘ognized from the earliest days of the development of structural analysis, and significant contributions were made in the nineteenth century. By 1888 a practical theory for including the effect of elastic displacements in the analysis of suspension bridges had been evolved by Melan (Ref. 1.18), and in 1899 Goodman presented a clear expla- ition of the plastic hinge concept (Ref. 1.19). In the early years of the twentieth century, Steinman published an English translation of Melan’s deflection theory (Ref. 1.20) and, as reviewed in Reference 1.21, contributions to plastic analysis and design ‘were made by Kazinezy, Meier-Liebnitz, and others. But it was not until the latter half of this century that reasonably general, practicable methods for including both zeo- ‘metric and material nonlinearity in analysis became a reality. “Among the seminal events of this period was Turner, Dill, Martin, and Melosh’ 1960 publication of a geometric stiffness matrix for the uniform axial force member (Ref, 1.22). On the material behavior side, a dramatic thrust occurred between the end of World War If and 1965, Most of the significant developments of that period are reported in References 1.21, 1.23, and 1.24, Since then, there has been a steadily in creasing flow of contributions to the theory and application of nonlinear methods of ‘analysis and design. Most of them have been incremental, but their cumulative effect has been profound. 1.2 COMPUTER PROGRAMS ‘When the structural engincer leaves the university and eaters the design office, it is probable that he or she will encounter computational hardware and software of con-4 Chapter 1 Introduction siderable sophistication. The available programs may require only the simplest data in ease pang about elutis for the problems at hand. Wit such tools available, it \ Sreciuca for tho stident of engineering to question the uty or even the necessity, a rah my appear to be a study of already-programmed procedures. The answer to {ine question i the responsible ofthe design engineer. The decisions that must \ tends by tis individual, extending from the analytical description ofthe structure, tease ihe evaluation ofthe correctoess of the namerical analysis, and through to the aterpretation ofthe outptt, are not amenable to computerization, The structural wee er met undetstan the principles of analysis. Computerization has relieved the Fae the rote operations in design analysis, but it cannot relieve the engineer of ‘esponsibility for the designed siructre : Paloughout this book, therefor, istrative numerical examples and problems are | deed, best of them are simple The aim is to avoid obscuring the point or principle cir study jaa cloud of numerical computations, But in the process the advantage of matrix notation and the matrix approach may not become immediately apparent i cat clar thatthe algebraic equations for these elementary structures could have eee tiled ina less forma, and perhaps more physically obvious, way. Such i not eee rier ost structures eccountered in practice, however. They ate larger, more Revoicated, and generally require more extensive calculations. Iti the efficient anal- See etsuch ystems that we are mainly concerned wit, Although adopting the matrix a acl may nt infcenee the way We think about structural behavior, one of is 1 serratus ie that mentioned earlier: It provides the best format presently known for analyzing stracures by computer 1.2.4 Computational Klow and General Purpose Programs “The equations ofthe matrix/faite element approach are of a form so generally appl \ cable that it is possible in theory to waite a single computer program that wil lve an ! imost limitless variety of problems in structural mechanics. Many commercially avail- able general-purpose programs attempt to obtain this objective, altzough usually om @ vestrieted scale. The advantage of general-purpose programs is not merely this capa- bility but the unity afforded in the instruction of prospective users, input and output data interpretation procedures, and documentation, "The four components in the flowchart of Figure 1.1 are common to virtually all Hl general-purpose, finite element analysis programs. As» minimum, the input phase ' Should require of the user no information beyond that relating to the material of onstruction, geometric description of the nite element representation (including sup- 1 port conditions), and the conditions of loading, The more sophisticated general-pur ' pose programs facilitate this input process through such features as prestored material Droperty schediles and graphical displays of the finite clement idealization so that Errors in input may be detected prior to performance of calculations. Ta the phase comprised of the library of finite elements is the coded formulative process for the individual elements. Most general-purpose programs contain a variety , Prelements, samples of which are illustrated in Figure 1.2, The only element we are Concerned with here is the framework element. Texts such as References 1.1 and 12 Geeeribo the use of other elements in analytically modeling a continuum. The element brary phase of the general-purpose program receives the stored input data and es tablishes the clement algebraic relationships by application of the relevant coded for~ Iulative process. This phase ofthe program also includes all operations necessary 10 position the element algebraic relationships for connection to the neighboring ele- 1 peents as well as the connection process itself. The latter operations produce the al- ' ¢gebraic equations that characterize the response of structure.12 Computer Programs 5 Tague ‘Defirition of phys modet tc com ak oading fod boundary conditions iy acon] ‘Genertion ofmatheatieat odes for sructual elements 3 Sohtion od applied loadings Construction and caltion of Spatheanties model for «1 Strutural ester Cre ouput Display of predicted isplscemente and reas Figure 14 Structural analysis computational flow. ‘The solution phase operates on the equations of the problem formed in the prior phase, In the case of a linear static analysis program this may mean no more than the single solution of a set of linear algebraic equations for a known right-hand side, Non- linear static analysis normally requires a series of incremental linear analyses. Solutions for dynamic response may soquire very extensive coinputations over a time-history of applied loads. In still other cases, particularly where the number of unknowns is very large, it may be advantageous to divide the total siructure into several substructures. ‘A finite element idealization of each substructure is made and analyzed, and the results, ‘are properly recombined into a solution for the full structure. Included in the solution phase ate the back-substitution operations needed to obtain all the desired aspects of the solution, © @ @ igure 12 Sample finite elements (2) Framework member, (6) Plane stress, (© Flat plate bending. (4) Sold element, (e) Curved thin shel.6 Chapter'1 Introduction ‘The ouput phase presents the analyst with a numerical or pictorial record of the solution upon which engineers can base decisions regarding the proportioning of the Structure and other design questions. The numerical record is commonly a printed list dl forces or stresses at points or within the finite elements, of displacements ofspeciied points, and of other desired information, Pictorially, data such as moment diagrams, Tontour plots of the principal stresses, or plots of the displaced state ofthe structore thay be displayed oa the compuler sereen and printed as permanent records, “The phases of the general-purpose program have one feature in common: the mod- fagpedts. Insofar as possible, the procedures treated in this larity of their component ext are formulated in a manner and form consistent with those found in widely dis- ihibuted general-purpose programs. But they should also be of use to the student in preparing more linited or special-purpose programs. Often, programs fora particular Purpose ean be designed to treat that problem more eliciently than the general purpose programs can. 12.2 The Program MASTAN2 ‘A CD containing a structural analysis computer program entitled MASTAN? is bun ‘hed with this textbook. As shown in Figures 13 and 1.4, MASTANZis an interactive Graphics program that provides preprocessing, analysis, and postprocessing eapa- filities, Preprocessing options include definition of structural geometry, support con Gitions, applied loads, and element properties. The analysis routines provide the tier the opportunity to perform fist- or second-order elestic or inelastic analyses St two: oF three-dimensional frames and trusses subjected to stati loads. Post- iguee 13 Preprocessing capabilites of MASTAN2.12 Computer Programs 7 os g & os i | 1stOsdeeBlstic UT Land order Best 3 os —BstOnder nettle a A baat Onder Inet = 1 a rr Lateral Displacement (Node 7) (Response cores Figue 14 Posiprocessing capabilities of MASTAN2. processing capabilities include the interpretation of structural behavior through de- formation and force diagrams, printed output, and facilities for plotting response ‘in many ways, MASTAN? is similar to today's commercially available software in functionality. The number of pre- and post-processing options, however, has been limited to minimize the amount of time needed for a user to become proficient with, jt. Many of the theoretical and numerical formulations prescnted in the context of the following chapters provide @ basis for the program's linear and nonlincar analysis, routines. In this regard, the reader is strongly encouraged to use this software as.a tool for demonstration, reviewing examples, solving problems, and perhaps performing, ‘analysis and design studies. MASTAN? has been written in modular format, There fore, the reader is also provided the opportunity fo develop and implement additional or alternative analysis routines directly within the program. Finally, it should be noted ‘that MASTAN2 will execute on any computing platform where MATLAB isavailable,Chapter 1 Introduction EFERENCES LAR. H. Gallagher, Finite Element Analysis: Fundamentals, Prentice-Hall, Engle- wood Cliffs, NJ. 1975 1.2 K-43. Bathe, Finite Elemem Procedures, Prentice-Hall, Englewood Cliffs, NJ.. 1996. 13. G.H.Golub and C. F. Van Loan, Metrix Computations, Third Edition, The Johns Hopkins University Press, Baltimore, MD., 1996, 14 J.C. Maxwell, “On the Calculations of the Equilibrium and Stiffness of Frames,” Phil. Mag. (4), 27.294 (1864), LS A. Castigiiano, Theorie de I'Equilibre des Systomes Elastiques, Turin, 1879 (En- zglish translation, Dover Publications, New York, 1966). 1.6, Mohr, “Beitrag zur Theorie der Holz-und Eisen Konsiruktioner Architekien und Ingenieur Verienes zu Hannover, 1868. 17 A. Cayley, “A Memoir on the Theory of Mutrices,” Phil, Trans, 148, 17-37 * Zeit des (357). 8S. Timosh: of Strength of rials, McGraw-Hill, New York, 1953. 19 G. B. May » 1, University of Minnesota, Minne- apolis, 1915. 110 A. Ostenfeld, Die Deformaionsmesdode, Springer-Verlag OG, Berlin, 1926, L1H. Cross, “Analysis of Continuous Frames by Distributing Fixed-End Moment Trans. ASCE, 96, 1~10 (1932). L12 R. A. Frazer, W. J. Duncan, and A. R. Collar, Elementary Matrices and Some Applications to. Dynamics and Differential Equations, Cambridge University Press, London, 1938. 1.43 J. H. Argytis, and S. Kelsey, Energy Theorems and Structural Analysis, Butter \worth Scientific Publications, London, 1960. 1.14 M. J. Turner, R. Clough, H. Martin, and L. Topp, “Stiffness and Deflection Anal: ysis of Complex Structures,” J. Aero. Sci, (23), 9, 805-823 (Sept. 1956). 145 Proceedings of the Conference on Electronic Computation. Held in Kansas City, Mo., 1958, ASCE, N.Y. 116 Proceedings of the 2nd Conference on Electronic Computaiion. Held in Pittsburgh, Pa., 1960, ASCE, N.Y. LIT Proceedings of the Third Conference on Electronic Computation. Held in Boulder, CO,, 1963. Proc. ASCE, J. of the Struct. Div, 89, No. ST4 (Aug. 1963). 148 J, Melan, Theorie der eisemen Bogenbricken und derHingebriicken, Handbuch der Ingenicurwissenschafien, Leipzig, 1888, 1.49 J. Goodman, Mechanics Applied to Engineering, Longmans-Green, London, 1899, 1.20 J, Melan, Theory of Arches and Suspension Bridges, translation by D. B. Stein man, M. C. Clark Pub., Chicago, 1913 121 Plastic Design in Steel, A Guide and Commentary, Manual No. 41, 2nd edition ASCE, New York, 1971. 1.22 M, J. Tumer, E, H. Dill, H.C. Martin, and R. J. Melosh, “Large Deflections of ‘Structures Subjected to Heating and External Loads” J. Aero/Space Sciences, 27, February, 1960. 1.23 J. F. Baker, M. R. Horne, and J. Heyman, The Steel Skeleton, Vol. 2, Cambridge University Press, Cambridge, 1956. 1.24 Plastic Design of Multi-Story Stee! Frames, Vols. 1 and 2, Department of Civil Engineering, Lehigh University, Bethlehem, PA, 1965,Chapter 2 Definitions and Concepts ‘This and the following three chapters contain the fundamentals of the displacement ‘method of matrix struetural analysis, Particularly useful components of the flexibility method are also treated. The terms displacement method and flexibility method refer to general approaches to analysis. In their simplest forms, both approaches can be reduced to sets of formal rules and procedures, However, a particular form of the displacement approach—the direct stiffness method—is dominant in general structural analysis and is therefore emphasized here. Use of the procedures developed in these chapters can give one a certain facility in analyzing simple structures, but further study is needed to appreciate the scope and power of these methods and their uader- lying principles of structural behavior. To promote such understanding, we later explore some of the most important ramifications of the fundamentals developed in Chapters 2-5. We begin the present chapter with an explanation of the concept of degrees of frocdom and then follow with a description of the principal coordinate systems and sign conventions to be used throughout the book. [dealization-of framework structures for the purpose of analysis is examined next. We then proceed to the definition of influence coefficients: relationships between forces and displacements. Globel stiffness ‘equations for axial foree members are developed ané their use is illustrated in simple examples. 2.1 DEGREES OF FREEDOM In this chapter we are concerned with the overall behavior ofthe elements ofa struc ture as defined by the displacements of the structure's joints under the action of forces applied atthe joints The study of the relative displacements of points within individual members—the sirains—and the distribution of forces per unit area—the siresses—ie the subject of later chapters. Displacement components required for the definition of the behavior of typical structures are illustrated in Figure 2.1 Ia the pin-jointed plane truss (Figure 21a). the members are stretched or compressed by the applied load. The net effect isthe dis~ placed structure shown, to exaggerated scale, by the soli lines. Except at supports ox joints that are otherwise constrained, the movement ofeach joint can be described by {Wo translarional dsplecement components, such as w and v in the figure, As we'll demonstrate later, onee the « and » components ofall joints fave been determined, the entite response—the reactions and member forces 2s well as the resultant displ rents—can be defined. For a pin-jointed space truss, three components, such a3 i ¢, and w in Figure 2.1b, are required at each joint.