FEA CH 1-6 PDF

Contents PREFACE e xiii INTRODUCTION oe 1.1 Brief Review of Finite Element History 2 1.2 Some Sample Applications 5 References 10 MATRIX ALGEBRA AND LINEAR EQUATIONS 12° 21 Matrices 12 2.2 Row and Column Matrices 12 2.3 Addition and Subtraction of Matrices 13 2.4 Scalar Multipliers 13 2.5 Matrix Multiplication 14 2.6 Fortran Statements for Multiplying and Printing Matrices 16 2.7 Transpose of Matrix 17 2.8 Special Matrices 19 2.9 Matrix Partition (submatrices) 21 2.10 Orthogonal Matrix (restricted to square matrix) 21 2.11 Coordinate Transformation Matrix 22 2.12 Determinant (restricted to square matrix) 26 2.13. Matrix Inversion (restricted to square matrix) 29evevslim 4 Biel Lil gisiee nn urs SUL COUTILeE 4 Promemk 44 BASIS STRUCTURAL peunenis Uringpdr ot Superpenition dee Wine lone kya Load Syste aT Muawvell-nent weripescal Tieoeus ae Doutgy Thiercine $1 Meteraace: G5 Frublens tet TRUS RAR FI EMERNTE Ab ennee, 0 47 Siifluee.: Byuution< Tera Lrake far (lemanrin est Cine dimarer 74 aq u Euuutivuie for ae Prise lat Isbemenr ¢ieuat | Phuruciy ina Pwe Dlrmangieaal Blase 25 ad Papuatiouce Tur at Pras. thir Tieavent Cleiestat, 48 ctisbeluidee anak Sevieelerr Pre tsevevate teed bay Bui ‘truce 37 Borne Mawle Treatunceks uf Bauuuples al Fin inmae imensdiomul Suey KL j Hines Wquatians 87 450 Cumetuding Remuk. pai Reloveruucs 1M Pugbleins 104 A A BEAM ANID HLSME FTLAME Fi eMente Bian Bu Uriberna Wiemann 1100 BB Manus Atbiiruriy tra Pie Pirianruel Plane (nies neu) Le 3300 Anmilesth . minnre LDS aa fed Luke TAContents, 5.5 Application of Plane Frame Elements 148 5.6 Concluding Remarks 170 References 170 Problems 171 — NONUNIFORM AND CURVED BEAM FINITE ELEMENTS 176 6.1 Nonuniform Straight Beam Element 177 6.2 Application to Tapered Beam Example 179 63 ircularly Curved Beam Finite Elements 184 6.4 Cubic-Cubic Circularly Curved Beam Finite Element 187 6.5 Application to Curved Beam Example 191 6.6 Concluding Remarks 194 References 195 Problems 196 FREE VIBRATION OF TRUSS BAR, BEAM, AND PLANE FRAME _~ FINITE ELEMENTS 198 — 7.1 One-D.O.F. Spring-Mass System 198 7.2. Axial Vibration of Two-D.O.F. Truss Bar Elements 199 7.3. Axial Vibration of Four-D.O.F. Truss Bar Elements 205 7.4 — Flexural Vibration of Beam Elements 209 7.5 Axial-Flexural Vibration of Frame Elements 215 7.6 Axial-Flexural Coupling Effect in Frame Vibration 224 7.7 Reduction of Large Eigenvalue Problem 225 7.8 Concluding Remarks 228 References 229 Problems 229 BUCKLING AND LARGE DEFLECTION OF COLUMN AND PLANE FRAME FINITE ELEMENTS 231 8.1 Governing Differential Equations fora Beam Element with the Effect of Axial Force 232 8.2 Formulation of a Uniform Beam Finite Element with Con- stant Axial Force 234 8.3 Consistent Incremental Stiffness Matrix for Distributed Axial Force 237 8.4 Buckling of Column on Elastic Foundation 241 8.5 Reduction Method for Buckling Problem 243 8.6 Buckling of Plane Frame 24410 " PLANE 91 9.2 93 9.4 9.5 96 Contents Large Deflection of Beams and Plane Frames 247 Concluding Remarks 248 References 250 Problems 251 STRESS AND PLANE STRAIN FINITE ELEMENTS 254 Two-Dimensional Elastic Equations 255 Bight-D.O.F. Rectangular Plane Stress and Plane Strain Finite Element 262 SicD.O.F. Triangular Plane Stress and Plane Strain Finite lement 272 Higher-Order Rectangular Elements 288 Higher-Order Triangular Elements 294 Comparison of Results 302 References 307 Problems 309 AXISYMMETRIC AND GENERAL SOLID FINITE ELEMENTS. a3 10.1 Axisymmetric Solid Element under Symmetric Loads 414 10.2 Axisymmetric Solid Element under Nonaxisymmetric Loads 322 10.3 Computer Programs and Sample Analysis 325 10.4 Tetrahedral Elements 329 10.5 Rectangular Hexahedronal Elements 340 10.6 Concluding Remarks 347 References 348 Problems 350 ATION AND CURVED ISOPARAMETRIC ELEMENTS = 11.1 Numerical Integration 355 i Curved Isoparametrie Elements 375 ug Concluding Remarks 392 References 393 Problems 394Contents 12 13 xi PLATE ELEMENTS IN BENDING 399 12.1 Theory of Plates in Bending 400 12.2. Selection of Displacement Functions and Patch Test 404 12.3. Rectangular Elements 413 12.4 Triangular Elements 439 12.5 Quadrilateral Elements 460 12.6 Hybrid Stress Plate Elements 462 12.7 Mixed Plate Bending Elements 466 12.8 Numerical Results 466 12.9 Concluding Remarks 469 References 470 Problems 476 SPECIAL-PURPOSE FINITE ELEMENT PROGRAMS 481 13.1 Static Analysis of Plane Truss and Plane Frame Struc- tures 482 13.2. Free-Vibration Analysis of Plane Truss and Plane Frame Structures 493 13.3. Six-D.O.F. Triangular Plane Stress and Plane Strain Element 504 13.4 Sixteen-D.O.F. Rectangular Plate =~ Element in Bending 519 INDEX 538Preface This book is based on the classroom materials developed by the author during a period of fifteen years. The materials are for two three-credit semester courses (or three one-quarter courses), one at the junior or senior level and the other at the senior or graduate level. This book is designed to include the following features: 1. It is simple and self-contained. To understand the materials, the reader needs only to have the background of a sophomore-level strength of materials course. Basic references have been well interpreted and digested in the text. Minimal instructional aid or discussion is needed. 2. The level of explanation for each subject goes deeper than that found in most texts. This is a book that undergraduates will feel comfortable reading. Ample illustrative examples, figures, and problems are given. Although it is perfectly suitable as a research reference, it can comfortably be used for a sequence of two undergraduate courses. 3. Throughout the text, examples with figures and numbers are given gen- erously. Physical interpretation and practical application are emphasized. Abstract mathematics and sudden interruptions of explanations are avoided. 4. This book can easily be understood by a practicing engineer who gradu- ated from college a long time ago and has become gradually less familiar with mathematics. The book is also suited for those practicing engineers who have not been exposed to finite elements either in industry or in college. xiti3 xiv Preface After the general introduction in Chapter 1, an elementary review of matrix mathematics is given in Chapter 2. Chapter 3 reviews the basic structural theorems at a junior level, Chapter 4 introduces truss analysis and Chapter 5 introduces beam and frame analyses. Chapter 6 introduces tapered and curved beam elements, and Chapter 7 extends the formulations for bars, beams, and frames for free-vibration analysis. Chapter 8 extends the formulations for bars, beams, and frames for buckling and large deflection analyses. Chapters 1 through 7 are suited for a required course at the junior level in aeronautical, civil, and mechanical engineering disciplines. Chapter 8 can be taught in cither the first or second course. Chapter 9 introduces all types of plane stress and plane strain elements, together with solution procedures, a computer program, and practical sample solutions. Chapter 10 introduces axisymmetric solid elements and three- dimensional elements of various shapes. Chapter 11 discusses in detail the methods of numerical integration and introduces systematically a variety of Chapter 12 introduces various kinds of flat ter 13 gives Fortran programs, user's manuals, ta for static and vibration analyses of plane alysis using the constant strain triangle and the 16-degree-of-freedom rectangle in bending. Chapters 8 through 13 or 9 through 13 are suited for a second dual-level course (senior and graduate), required for those with a structural major and optional for those with a curved isoparametric elements. plate elements in bending. Chapt and sample input and output dat trusses and frames, and static an; structural minor. The author devotes this book to his wife, Dehleen, for her infinite patience and encouragement, and also to his daughters, Maria, 14, and Martha, 11, to aane Daddy’s guilt from spending too little play and homework time with them. The author is indebted to Richard H. Gallagher, his major professor during his Ph.D. studies at Cornell. It was Professor Gallagher's technical guidance, inspiration, professional example, and continuous encouragement set led the author, with confidence, to pursue a career in engincering edu- cation. The author is indebied to imany of bis undergraduate and grodtite students, including Rakesh K. Kapania and Sunil Saigal. Special thanks go to Kapania for his assistance in searching the literature, generating numerical data, and for providing valuable in-depth discussions and comments. ‘The technical typing and grammatical corrections by Nancy A. Stivers are gratefully acknowledged. itera tthor most graciously acknowledges the patience, care, and quality of editing and production of this book by Nancy G. Follender of Prentice-Hall. T. Y. YangCHAPTER-4 Introduction development of computers has completely revolutionalized research d engineering field. Conventional computers and graphics ‘The rapi and practice in every scientific an have rapidly branched out into supercomputers, minicomputers, computers. The dream that every engineering office and home would have a computer terminal and/or a microcomputer has become a reality. Personal computers in the 1980s are as popular as pocket calculators were in the 1970s and slide rules in the 1960s. Microcomputers are rapidly evolving to have unprecedented memory, speed and graphics capabilities. Following this trend, analysis and design methods that provide computerized solutions to scientific and engineering problems have rapidly been developing for increasingly routine daily use. In this book we focus on one such significantly developed method, the finite element method. ‘Although this method is applicable to many scientific and engineering fields, we deal only with the field of structural analysis and design. The finite element also been increasingly used as a method has long been a fertile research field. It has research tool for numerical experiment, Most importantly, the finite element method hasnow becomeapredominantstructural analysis and design tool which is used routinely by structural engineers. Because of the need for structural engineers to be familiar with the finite element method, it is necessary that structural engineering students take an introductory finite element course in their junior or senior year. tory course is gradually replacing the required traditional cours analysis in aeronautical, civil, mechanical, and other engineering Furthermore, it has become a tren d finite element co es on structural disciplines. urse be d that a secon 1Introduction Chap. 4 taught at the dual level (senior and graduate), required for those with a structural major and optional for those with a structural minor. This book provides more than enough material for these two courses. Furthermore, this easy-to-read self-contained book is suitable for those instructors who have been more used to conventional structural analysis methods. This book is ideal for practicing structural engineers who have limited or no background in finite through self-study. set or different patte appears to have been a element has a specific type of structural shape and is interconnected with the adjacent elements by nodal points. Acting at each nodal point a displacements (degrees of and disp cases of thermal, fluid, electrical, and other problems. Thus for each element a standard set of simultaneous equations can be formulated to relate these physical quantities. Physically assembling these elements to form the whole structure is equivalent to superimposing these element equations mathematically. The result is a large set of simultaneous equations which are suited for solution by computer. Upon implementing the loading and boundary conditions (for structural problems), the assembled set of equations can be solved and the unknown parameters found. Substituting these values back to each element formulation provides the distributions of stress and displacement everywhere within each element. 1.1 BRIEF REVIEW OF FINITE ELEMENT HISTORY Using the methods of discretization and numerical approximation to solve scientific and engineering problems is a fact of life. The concept of finite elements stems from the idea of discretization and numerical approximation. If we were to identify the evidence of the earliest finite element concept, we probably could loosely trace it back to the geometric approximations of pyramids by Egyptians some 5000 years ago. If we consider, for example, the numerical approximation of 7 as a starting point of finite elements [1.2], we could find some interesting historical records in China, Egypt, and Greece. The records show that the Chinese started approximating 7 in the first century a.p. A value of a equal to 3.1547 was evidenced in the design of a cylindrical volume measurer. In the second century a.p. astronomer Chang Heng of the Eastern Han Dynasty approximated a as 3.1466 (730/232) and V10. In approximately .p. 230, Wang Fan of the Country of Eastern Wu usedSec. 1.1 Brief Reviow of Finite Element History aw as 3.1556 (142/45). In the dynasty of Western Jihn (4.0. 265-317), Liu Hui in his comment on Mathematics—Nine Chapters used a regular polygon inscribed in a circle to approximate the circumference and he found ar to be 3.1416 (3927/1250) using a polygon of 3072 equal sides (finite elements). Thus some Chinese have since referred to 7 as Hui's ratio. In the dynasties of Former Sung and Southern Chi, mathematician Tzu Tzong Tze (A.p. 429- 500) determined that 3.1415926 < @ <3.1415927. To have obtained such an accuracy was equivalent to finding the side length of a polygon of 12,288 equal sides and the area of a polygon of 24,576 equal sides, both inscribed in a circle. The Ahmes paper shows that by 1500 n.c., the Egyptians were using V10 for 7. A still earlier papyrus, now in Moscow, indicates that the Egyptians used the correct formula for the volume of a pyramid and the area of a circle by about 1800 B.c. [1.2]. Archimedes (2872-212 n.c.), one of the greatest of the early mathematicians and inventors, used finite elements to determine the volume of solids [1.2]. More rigorously, if we consider some approximate solutions to elasticity and structural problems as the starting point of the finite element method, we can refer to a historical account of the developments by Timoshenko [1.3]. If we consider the emergence of the concept of frame analysis as the starting point of the finite element method, we can trace back to the works by Maxwell [1.4], Castigliano [1.5], and Mohr [1.6], among others, during the period 1850-1875. An account of the development starting from this period can be found in, for example, the text by Gallagher [1.7]. In 1915, Maney [1.8] in the United States presented the method of slope deflection, expressing the moments in terms of deflections and slopes at the rigid joints of frame structures. Such a formulation is in precisely the same form as the stifiness equations given in Chapter 5. A similar development was put forth by Ostenfeld [1.9] in Denmark. In 1929, Cross [1.10] made public his method of moment distribution for frame analysis. The method relaxes the joint moments very quickly and simply for an approximate solution or, with a little more labor, extends to any degree of exactness desired. The method dominated the practice in frame analysis and design for the following 35 years, Parallel to the earlier works on the analysis of frame structures, the concept of using lattice analogy to solve continuum mechanics problems also began to take form (see, for example, Refs. 1.11 to 1.15). In the early 1940s, Courant [1.16] proposed using piecewise polynomial interpolation functions to formulate triangular subregions as a special Rayleigh-Ritz variational method to achieve approximate solutions. Today’s practical finite element method is essentially a by-product of computers, The rapid development of the finite element method has kept close Pace with the swift development of computers ever since their inception in the early 1950s. Two early classic publications in the mid-1950s by Argyris and Kelsey [1.17] and Turner, Clough, Martin, and Topp [1.18] merged the* Introduction Chap. 4 initial concepts of discretized frame analysis and continuum analysis and kicked off,the explosive development in finite element method, Many reviews accounting for the growth of the finite element method are available (see, for example, Ref. 1.19). A thorough chronological review of the fundamental developments of the finite clement method is given in this text. Following the linear static formulation for each finite element, extensions for practical applications have continued to include the various physical effects and fields of vibration and dynamic response, buckling and postbuckling, geometrical and material nonlinearities, thermal effect, fluid-structural interaction, aeroelasticity, structure-acoustics interaction, fracture, laminated com- posites, wave propagation, structural dynamics and control interaction of aircraft and space structures, random dynamic response, and others. Following these developments, both research and commercially oriented computer programs have become available. The use of special-purpose and general-purpose finite element programs has become routine practice in structural engineering offices. A current major research thrust is directed to computer-integrated graphics, analysis, design, manufacturing, and automation. The finite element method occupies a predominant role and will have a significant impact on this research direction. Many fertile research areas within the domain of finite elements lie in this broad cross-disciplinary field. With the growing availability of microcomputers, it becomes apparent that a trend toward tailoring finite element methods and programs to fit microcomputers is in progress. 1.2 SOME SAMPLE APPLICATIONS In this book we introduce in great detail the various kinds of elements: truss bars; beam and frame members; tapered and curved beams; plane stress and plane strain triangles and rectangles; tetrahedra and hexahedra; and flat rectangles, triangles, and quadrilaterals in bending. The curved-thin-shell type of element is, however, beyond the scope of this text. The effects of vibration, buckling, and large deflection are included in appropriate sections for general illustration To illustrate the applicability of these elements in practical applications, a few example figures are shown. Figure |.1 shows a structure of a fuselage and a wing modeled using beam, plate, and shell elements. A detailed model showing a special component with a cutout illustrates the versatility of finite elements. This model can be used for the analysis of static stress, free vibration, landing impact response, panel and wing flutter, and optim weight and full strength. Figure 1.2 shows an example of a multi-story using beam, column, and plate elements, zation for minimum ti-story hotel building modeled This model can be used for static,Sec. 1.2 Some Sample Applications 6 vibration, earthquake response, static equivalent, and random wind response analyses [1.20] Figure 1.3 shows the outside configuration of a 1200-MW fossil fuel steam generator hung and tied to a supporting steel frame structure. The whole system is modeled using beam, column, and plate elements with 1860 degrees of freedom. The figure shows that the structure is vibrating in the third mode with a natural frequency of 1.1 Hz. The natural frequencies and mode shapes are used for earthquake response analysis [1.21]. Figure 1.4 shows the modeling of a column-supported cooling tower with a dimple type of local imperfection. The columns are modeled using column elements. The hyperboloidal shell is modeled using quadrilateral curved shell elements. To model the local imperfect area, smaller quadrilaterals are needed. Quadrilaterals of large and small sizes are connected by curved triangular “filler” elements. This model can be used for earthquake and random wind response analyses [1.22, 1.23]. If initial imperfection is taken into account, the behavior can become geometrically nonlinear. tft Typical elements Structural model g beam, plate, and shell elements for Fi jage and a wing modeled using ‘igure 1.1 Fuselage and a wing A nl eanaee static, dynamic, flutter, and optimization analyses. (Courtesy 0} Jr., NASA Langley Research Center.)‘ Introduction Chay 5 SAY UN, Say SNSS 4 WY Ny YY Figure 1.2 Multi-story hotel building modeled using column, beam, and plate elements for static, earthquake, and wind response analyses. Figure 1.5 shows a large flexible space structure and a single repetitive lattice cell. The structure can be modeled by a huge number of truss bars or a relatively small number of equivalent plate finite elements, each possessing the continuum properties of a few cells, This model can be used for dynamic analysis and control integration [1.24]. Figure 1.6 shows a radial belted tire modeled using axisymmetric laminated shell elements [1.25]. This model can be used for the analysis of static inflation with large displacements, free vibration about the inflated state, and dynamic response. For localized loads such as the road contact force, quadrilateral elements are needed. To illustrate the effectiveness of graphical display that can help us perceive and model structures of complex geometry, Fig. 1.7 shows someROSS iH SSSR, | s A M\AVVYY se RSS Yabo a / IN NiFf 8 Introduction Chap, 1 curved surfaces fitted by B-spline functions. The grid points and their coordin. ates are stored in a data base which becomes the input for finite clement modeling. Such a shell structure can normally be modeled using a large number of flat plate elements or relatively fewer number of curved shell elements. Figure 1.5 Space lattice structure modeled by truss bars or equivalent plate elements for dynamic analysis and control integration.Sec. 1.2 Some Sample Applications 9 Figure 1.6 Radial belted tire modeled using axisymmetric laminated shell elements for large defiection and vibration analyses, i ith gridpoints for shell Figure 1.7 Computer graphical display of curved surfaces with gridpoints finite element modeling, (Courtesy of Computer-Aided Design and Graphics Labora- tory, Purdue University.)10 1.3. 1.4, 15 1.6. 17 18. 1.9. 1.10. Lil. 1.16. Introduction —Chyy Day REFERENCES R. W., “The Finite Element in Plane Stress Analysis,” Proceedings, 29 e nic Computation, Pittsburgh, Pa., Sept. 1960, Introduction to Finite Element Analysis, McCirgy Clough, ASCE Conference on Electron Martin, H. C,, and Carey, G. F., Hill Book Company, New York, 1973, p. 2 ; Timoshenko, S. P., History of Strength of Materials, McGraw-Hill Book Company New York, 1953. Maxwell, J. C., “On the Calculations of the Equilibrium and Stifiness of Frames.» Philosophical Transactions, Vol. 27, No. 4, 1864, p. 294. ; Castigliano, A., Théorie de I'équilibre des systémes élastiques, Turin, 1879 (Engligg translation by Dover Publications). Mohr, O., “Beitrag zur Theorie der Holz- und Eisen Konstruktionen,” Zeitschrig des Architekten und Ingenieur Verienes zu Hannover, 1868. Gallagher, R. H., Finite Element Analysis—Fundamentals, Prentice-Hall, Ine, Englewood Cliffs, N.J., 1975. , Maney, G. B., Studies in Engineering, No. 1, University of Minnesota, neapolis, 1915. Ostenfeld, A., Die Deformationsmethode, Springer-Verlag, Berlin, 1926. Cross, H., “Continuity as a Factor in Reinforced Conerete Design,” Proceedings, ACI, 1929, and “Analysis of Continuous Frames by Distributing Fixed-End Moments,” Transactions, ASCE, Vol. 96, 1932, pp. 1-10. Wieghardt, K., “Uber einen Grenziibergang der Elastizitatslehre und seine Anwendung auf die Statik hochgradig statisch unbestimmter Fachwerke,” Ver. handlungen des Vereins z. Beforderung des Gewerbef leisses, Abhandlungen, Vol. 85, 1906, pp. 139-176. Riedel, W., “Beitriige zur Lésung des ebenen Problems eines elastischen Kérpers mittels der Airyschen Spannungsfunktion,” Zeitschrift fiir Angewandte Mathematik und Mechanik, Vol. 7, No. 3, 1927, pp. 169-188. . Hrenikoff, A., “Solution of Problems in Elasticity by the Framework Method,” Journal of Applied Mechanics, Vol. 8, 1941, pp. 169-175. McHenry, D., “A Lattice Analogy for the Solution of Plane Stress Problems,” Journal of the Institute of Civil Engineers, Vol. 21, 1943, pp. 59-82. - Newmark, N. M., “Numerical Methods of Analysis in Bars, Plates and Elastic Bodies,” in Numerical Methods in Analysis in Engineering, ed. L. E. Grintet, Macmillan Publishing Company, New York, 1949. 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. Argyris, J. H., and Kelsey, S., Energy Theorems and Structural Analysis, Buttet~ worths Scientific Publications, London, 1960 (collection of papers published in Aircraft Engineering in 1954 and 1955). memes M. J., Clough, R. W., Martin, H. C., and Topp, L. J., “Stifiness and Peflection Analysis of Complex Structures,” Journal of Aeronautical Sciences, Vol. 23, No. 9, 1956, pp. 805-823 SECHAPTER 2 Matrix Algebra and Linear Equations 2.1 MATRICES A rectangular matrix [A] of order m x n (m rows by n columns) is defined as Qn Ap + Ay Ba aay 2) a Al=[ay)=|% 2 an tAl= [a=] 4% 2 od Ami Am2 *** Amn where a, is the element at the ith row and jth column. If m If m If m = 1, it is a column matrix. n, it is a square matrix. 1, it is a row matrix. 2.2 ROW AND COLUMN MATRICES A matrix consisting of a single row is called a row matrix. It will be represented by semibrackets, as follows: WAL = Lan a2 ais an] (2.2) 12geo. 24 lar Multipliers 13 A matrix consisting of a single column is called a column matrix, Mt will be represented by bra follows: ay ay {A} =4 ay, (2.3) Am 2.3 ADDITION AND SUBTRACTION OF MATRICES Additions and subtractions can be performed only on matrices of the same order. Such operations are performed by adding or subtracting corresponding elements. For example, a a b, b, + + (% Se)e[oe te]-[Bik ete] es Matrix addition and subtraction are commutative. For example, [A] — [B] = -(B]+[A] (2.5) They are also associative. For example, ([A] + [B]) —[C] = [A] + ((B] -[C)) (2.6) 2.4 SCALAR MULTIPLIERS A matrix of order 1 x 1 is a scalar. A scalar is a quantity. Let k be a scalar; then kay, kay ++ kain ka, kar, kay, K[A]=| kas) - (2.7) Ket, Kya Ken, For example, if w-[ J] w- 2] oY 42 21 2a} + 18) a¢c1= [,% 2-2 A thenXN ew 4 Matrix Algebra and Linear Equations Chap,» 2.5 MATRIX MULTIPLICATION If matrix [A] is of order m x p, matrix [B] is of order p * n, and matrix (C) ig the product of matrices [A] and [B], matrix [C] is then of order m xn, The element at the ith row and jth column of matrix [C] is obtained by the following procedure: 2 ty =X auby er = ayy bij + irda; + Aiyby, +++ + + Aipbyy (2.8) For example, [ce] = [Al [By 2x2 2332 9 on can be written in detail as by by u 2] fan ai as]p 2? C21 C22, = 42, Azz 23. ag 63, B32 _ euee + ayzba1 + aizbs, 411012 + Qi2b22 + isbs2 42)b,, + A22b2, + A2sb31 21b12 + G2rb22 + Gayby9, | (2.9b) The matrix multiplication procedure can also be explained by using the following example in diagram form. If 10123 12-10 12 ‘ ‘ ' i [A]=]}0 0 -1 1 2 2} [B]= 392 2 1 41] (10a) 1 1 2 100 22 a) a Io. 2 and [c] = [A] [B] 3x5 3x66x5 (2.106)Matrix Multiplication * 4th column 2nd row > | 0 oo. The schematic layout shows how any element in the product is related to the row and column from which it is formed. For example, ‘ C24 = = Aaxdeg = 6 (2.10c) = Some notes for matrix multiplication 1. Matrices are in general not commutative in multiplication. [AJB] = [B][A] (2.11) wi-[t Et s)-[ 3] wra-[7 Sf} aeieo 2. Two matrices [A] and [B] can be multiplied together only if they are conformable ; that is, the number of columns in matrix [A] must be equal to the number of rows in matrix [B]. For example, if and only if P= L (2.12) if and only of Q= Rand P=L (2.13) where [C] is premultiplied by [B] and [B] is premultiplied by [A]. [A] is Postmultiplied by [B].16 Matrix Algebra and Linear Equations — Chay 3. The matrices are associative in multiplication. [A]LBI[C] = ((AJLB))[C] = [A}((BILC]) (2.14) For example, ae 1-1 0 10 wa=[ ] [B]=/O 1 -1 [c]=]o 1 2 ol Ito 14 7 2-1 27f! ° 41 carences=[3 -1 ‘| : ' -[} é] caesrey=[} 1] “10 -[f 3] 2. 2.6. FORTRAN STATEMENTS FOR MULTIPLYING AND PRINTING MATRICES The Fortran statements describing the multiplication of tc] = [A] [Bl MxN MxLLxN (2.15) are given as follows: HO 10 1 = 19m nO 10 J = dow CUIrs) = 060 BO 10K = deb Clr = CCE + ALLARD ® It is important to note that the matrices must be sufficiently dimensioned by DIMENSION statements. To print a matrix [A] of order M x N, the following Fortran statements ROK DD may be used: WRITEC6y 1) X97 TOLER > FURHATCSSX* LOHKMATRIX Ax) wert > BO 10 tet WELTE CS (ACE r SS) 9sed ND FORMAT CLIEL2. 4) The number “6” is the output tape number. The statements are based on a computer that allows us to print 132 characters on each line. The READ statements for inputting a matrix are similar to the WRITE statements.a geo. 2.7 Transpose of Matrix ” 2.7 TRANSPOSE OF MATRIX ‘The transpose of a matrix is obtained by interchanging the rows and columns of the matrix. If matrix [A] is of order m x n, its transposed matrix [A]’ is Srorder n Xm. For example, if [A] - (ze a, | 2x37 Lan an am then a1 an os fae =] an an (2.16) 4y3 Aas From the definition of the transpose of matrix, it is obvious that (Ary = Tal 17) (AT + 0B)" = (A) + EBD7 18) If matrix [A] is of order m x n and matrix [B] is of order n x |, the two matrices are conformable and we can form product [A][B]. For the transposed matrices [A]” and [B]’, the orders are n x m and | x n, respectively. We can no longer form the product for [A]"[B]’ but we can form the product for [B]’[A]”. So we know that ((AIB)? ¥ [AT"TB]" (2.19) Let us first consider the product [C] = [A][B] in diagram form, as shown in Fig. 2.1. We then consider another product, [D] = [B]"[A]’, in a similar diagram form, as shown in Fig. 2.2. It is seen in the two figures that the jth column. ith row baht 2.1 Diagrammatic explanation of rc © procedure for obtaining [C] = [AI].8 Matrix Algebra and Linear Equations ~— Chap, 9 ith column ith row Figure 2.2. Diagrammatic explanation of the procedure for obtaining [D] = 4, (By'TAY. elements ¢, and d, are both obtained as the sum of products of a pair of identical row and column (or column and row). Therefore, these two elements are equal. By definition, we found that [c]" (D] We then conclude that ((AIB])7 = [B]"TA]” (2.20) This formula can be generalized to transpose the product of more than two, say four, matrices by using the associative law, (CAIBILCIDY = ()" CALBIC)” = (D]"[C]"((ATB)" erererarny = 2h) Example 2.1 If | (c] = [AB] find [C)’ies ial Matric geo. 2.8 Special Matrices S Solu N 10 1 1 tar=tarei = [? i ‘If: 1 0 “| -B 42 1 33 0 -1 3 3 1 3 cy = te"=!2 0 1 =, . Solution 2 1 oi te yi: Cy’ =(B)"[Al’ = ib (cy" = (By'TAT i 0 <1 Le a J <2 0. 3 -% 2.8 SPECIAL MATRICES Square matrix. If m =n for matrix [A], then [A] is a square matrix. For a symmetrical matrix, es ay=a, and [A]=[A]™ (2.22) It will be shown in Chapter 3 through the use of Maxwell reciprocal theorem that all the stiffness and flexibility matrices for finite elements are symmetrical. For an antisymmetrical matrix, a,=-a, wheni#j (2.23) Zero or Null matrix. If all the elements in matrix [A] are zero, it is a zero matrix. Diagonal matrix (restricted to square matrix). If ay = 0 for i” j and a, # 0 for i = j, matrix [A] is a diagonal matrix. If a diagonal matrix [A] is of order, say, 5 x 5, we have a, 0 0 0 0 0 a 0 0 0 0 0 [A] =] 0 0 a3 0 (2.24) 5x«5 0 0 0 ass 0 0 0 O ass-e 20 Matrix Algebra and Linear Equations — Chap,» If [A] is a diagonal matrix and [A]Ox} = {e} the solution for the unknown vector becomes mate §= 12,3, The identity matrix is a special form of Identity matrix (unit matrix). lements are equal to unity. It is diagonal matrix where all the diagonal el denoted by [I]. If [I] is of order, say, 5 x 5, we have 10000 01000 tm =|0 0 100 (2.25) 5x5 10 001 0 00001 If matrices [A] and [I] are both of the same order, (2.26) (1[A] = [AJ] = [A] Scalar matrix. A scalar matrix is a special form of diagonal matrix If a scalar matrix [A] where all the diagonal elements are equal to a scalar. is of order, say, 5 X 5, and the scalar is, say, 4, we have 40000 04000 [A] =|0 0 4 0 0/= 410) (2.27) 5x5 10004 0 0000 4 Triangular matrix (restricted to square matrix). If in matrix [A] all the clements above the main diagonal are zero (i.e., ay = 0 for i J), the matrix is called an upper triangular matrix, For example, 21323 02710 0045 4 000 2 2 00001 If [A] is a lower triangular matrix and we have a set of matrix equations, [Al{x} = {e} (2.28) we can first find x, as ¢,/ai. We can then find xz, xs, - + - sequentially by the method of substitution. 2.9 MATRIX PARTITION (Submatrices) The array of elements in a matrix may be divided into smaller arrays by horizontal and vertical dash lines. Such a matrix is then referred to as a partitioned matrix, and the smaller arrays are called submatrices. For example, Ay, Ayr | Aya 1 [A] =] a ax | aa; | = [444 42 (2.29) where a, a a taa=[ " 3] tad={ a @2, 22 23. [An]=[a21 a2] [A22] = @33 2.10 ORTHOGONAL MATRIX (Restricted to Square Matrix) [ithe tanspose of a square matrix is equal to its inverse, the matrix isan ieee matrix. The obvious advantage of an orthogonal matrix is that its © can easily be obtained simply by transposing the matrix. Let us consider a square matrix of order, say, 3 X 3, a 2 ay TA] = | az, 22 aay (2.30) 4) Gy2 AySS 22 Matrix Algebra and Linear Equations Chap,» The property of orthogonality implies that [AJIA]" = [AJA]! = (00 (2.314) or a, Ar 43] Pan a2 aa 10 0 @, G2 a3 |] M2 Grr aye] =] 90 1 0 (2.31) 43, G2 Ass} Las 23 Ass oo} By evaluating the three diagonal elements in the identity matrix, we obtain ayy + ain + aizy= 1 ay +a + ay 1 (2.32) 2 + 35 By evaluating the three elements either above the diagonal or below the diagonal, we obtain 3443) + aay aa53 = 0 (2.32b) Thus we reach the following conditions of orthogonality for a matrix: 1. The sum of the squares of each element in each row (or column) must be equal to unity. 2, The sum of the products of corresponding elements between any two rows (or columns) must be equal to zero. 2.11 COORDINATE TRANSFORMATION MATRIX In the finite element formulation, it is often required to transform the formulation from a set of local coordinates to a more general set of global coordinates. The matrices for coordinate transformation have the special property of being orthogonal, and such a property simplifies the transformation procedure. Three examples of coordinate transformation matrices are given below. These three coordinate transformation matrices are important because they will be used in the following chapters. éxample 2.2 Two-dimensional coordinate transformation A set of old coordinates (x, y) and a set of new coordinates (x', y") are shown in Fig. 2.3, The new coordinates are resulted from a counterclockwise rotation of the old coordinates by an angle 0._ 211 Coordinate Transformation Matrix gee: Figure 23 Rotation of a set of two- gimensional Cartesian coordinates by an angle 8. For an arbitrary point A, the new coordinates (*’, y') can be related to the old ones as follows: x’ = CB + BA =x cos0+ysin@ y' = AE — DE = y cos @~ xsin@ {x} en{3} (2330) (2.33a) or in matrix form where cos@ sin ‘| (233¢) —sin @ cos 0 m-[ Matrix (T] is a coordinate transformation matrix. This matrix satisfies the conditions of orthogonality that cos? 6 + sin? @=1 1 (2.334) (-sin 6)? + cos? @ (-sin 6)(cos 0) + (cos 6)(sin 8) = 0 Thus we know that this coordinate transformation matrix (T] is an orthogonal matrix. Example 2.3. Two-dimensional coordinate transformation matrix for axial-flexural eam element An xial-flexural beam element, oriented along the local x'-coordinate axis, is shone joints 24 The x’ axis is oriented at an angle 6 from the global coordinate x axis. At att? of the clement, there exists an axial force Xs, a transverse shear force Y4, and con ti8 Moment M}. These forces and moment can be trans! Sr a "sponding to the global coordinates X;, Y2, and Mz, respectively,24 Matrix Algebra and Linear Equations cy, Chap 4 Y, ¥j 4 x, {1% 5 oA Te, My @ t i ' y | x 1 ' \ Figure 2.4 Transformation of joint force vectors from local coordinates (x’, y’) to global coordinates (x, »). Fig. 2.4. The relations are as follows: Xs cos@ sind 077 Xz ysl =|-sing cose of} ¥, Ms ° o 1 jlag, x =IT ¥2 (234) Mz. It is again seen that the transformation matrix [T] is an orthogonal matrix. Example 2.4 Threc-dimensional coordinate transformation matrix A set of old coordinates (x, y, z) and a set of new coordinates (x’, »’, 2’) are shown in Fig. 2.5. Let i, j, k and i’, j’, k’ be unit vectors in the directions of the respective axes, and let A be a general point in space having coordinates (x, y, z) and (x’, y’, z') in the respective systems. We can write for vector OA R= OA =xi+ pj+ zk (2.35a) The components x’, y’, z' are obtained from the dot products of vector R with the unit vectors i’, j,k’. Get je ke iy ii “| 2 Geko jek’ ke ] Uz hom mip x =[b om mliyb=oy (2.356) omy ns}lz zsec. 2.11 Coordinate Transformation Matrix 6 Figure 2.5 Unit vectors in two sets of three-dimensional Cartesian coordinates. where [,, my, 3 bh, m2, Ma: Is, ms, ns are the direction cosines of the x’; y"; 2’ axes relative to the x, y, z axes, respectively. Matrix [T] is the transformation matrix. It is necessary to show that the transformation matrix [T] is orthogonal. We first write the new unit vectors i’, j', k’ as components of the old unit vectors i, j, k; v= hit mjt+nk j= bit mjtnk (2.35¢) K = Lit mj+ nk Because the two coordinate systems are rectangular systems, we have -k ies =R+mitni =B+mi+n} =B+m3t+n3 =I (2.354) Also, jek = = hh + mm, + nny = bh + mymy+ nny = bh + msm, + nym =0 (2.35e) ‘The conditions of orthogonality are thus obtained for the transformation matrix [T]SS Q_—_—_ « 26 Matrix Algebra and Linear Equations Chap y 2.12 DETERMINANT (Restricted to Square Matrix) A determinant of a square matrix [A], is a scalar value function denoted by Gy Ain" Aim det [A] = |A| = (2.36) uy 2.12.1 Minor and Cofactor The first minor of a determinant det [A], corresponding to the element a, is defined as the determinant obtained by omission of the ith row and jth column of det[A]. Let us denote this minor by M,. For example, if 13579 24 6 8 2 la4j=|i 2.3 3 2 2 21 2 3943 43 i the first minor for a2 is defined as 1579 } ma-|2 © & 2 i S22 1) 8 ! 3.3 43 If the first minor M, is multiplied by ( 4, it becomes the cofactor of the term ay. Thus Ay = (-1)'9 My For example, Row weune a, ancienSec. 212 Determinant S 2.12.2 Expansion of Determinant (Laplace Expansion Formula) The determinant of a matrix [A] can be found by the repetitive use of the Laplace expansion formula: det [A] = ¥ audu (ican be any row and i is not to be summed) mxm ke (2.37a) or det [A] = D aydy (j can be any column and j is not to be summed) mxm ' (2.37) Det [A] is written in terms of the sum of the products of the elements in the ith row (or jth column) and their corresponding cofactors. For example, ay M2 As det[A]= |a2 22 23 3) 32 33 = an Ay + ay2Ai2+ aAis +1[@22 4 ve2[ aa @ asfan @. aap] 2) 4 a2] 2} +a] 432 433 a3, Ass a3, Gaz = ayyaze(—1)" 55 + ar@23(— 1)! Fase — @yaday(—1)"* ass 1 = y33,(—1)!"2ay) + @y3421(— 1)!" ag9 + y3@2a(— 1)!" = Gy, 2233 + Ay 242343) + 13421432 — A13@22A31 — G11 A23432 — A12G21 433 __ In other words, the foregoing operation is in the following “familiar” diagram form:28 Matrix Algebra and Linear Equations cng, | Example 2.5 Find oom - oe Solution. By the diagram method, 123 3 1 0| =14+0+9-0-0-6=4 od or by the Laplace expansion formula, 123 3 10 ort. = 1-1? 3 0 4/3 1 0 {| eacn 0 | 10 a i °| +2(-1) pee 11 -2 | +33 ‘| ol ol = (-1)?- 1+0—2-3(-1)?-1+3-3(-1)?-1+3+1(-1)°-0 =1-6+9 4 2.12.3 Properties of the Determinant The determinant of a matrix and the determinant of the transpose of that matrix are equal, |A| = |A|”. 2. \(CATBD| = [AlIBI- 3. Inter minant. 4. If all the elements in a row, or a column, in a matrix are zeros, the determinant is zero. 5. If two rows or two columns in a matrix are identical, the value of the determinant is zero. This is a sufficient condition, but not a necessary condition. For example, deter- Ne| gec. 2.13 Matrix Inversion = 6. If matrix [B] is obtained from matrix [A] by adding a multiple of one. row of [A] to another (or a multiple of one column to another), then |A| = |B). For example, 112 jaj=]2 1 1 121 If the second row is replaced by the sum of the second row and twice of the first row, we have 2 5| =4=|al 1 2.13. MATRIX INVERSION (Restricted to Square Matrix) ‘There are many methods available for matrix inversion. Three popular methods are illustrated here. Their properties are also discussed. 2.13.1 The Adjoint Method Definition of adjoint matrix: The adjoint of a matrix [A], written as adj [A], is of the form Ai An Am Ay Ax Ama adj[A] = =[Ay]" (2.38) which is defined as the transpose of the’matrix’of cofactors. Example 2.6 3.21 Ate faj=]2 4 2 ( [: 1 :| (\ Find adj[A] for Solution. First we form the cofactors: Ay =6 Ay = 2 A\3 = -10 3. An=3 Ax =30 Matrix Algebra and Linear Equations 6 -3 0 adj(AJ=| 2 3 —10 3 8 2.13.2 Derivation of the Equation for Matrix Inv Chap » Then sion The inverse of a matrix [A], expressed by [A] ', is defined such that (AyAy' = 10 First let a matrix [P] of order m x m be defined as [P] = [A] adj[A] The term at the ith row and jth column of the matrix [P] is Py= X andy = z Ay = det [A] which is precisely the Laplace expansion equation for finding the determinant. Thus Pir = Poa = Pas ** = Pmm = |Al ity, Py = auAy, = 0 & The proof of py = 0 will be given subsequently. From the explanation above, it is readily seen that oe lal Al all zeros lal. all zeros Al (l= |Alo0 adj[A] A)——— = [I ae We premultiply both sides by [A] ': ol payin lal2. 2 Matrix Inversion geo. 213 a Finally, We arrived at the equation for matrix inversion: adj [A] (Ay! == 9 det [A] (2.39) Example 2.7 Invert the matrix [A] as defined in Example 2.6 in Section 2.13.1, A 6-3 0 1 _ adi CA] _ J [A] -aa-[ 2 3 «| +12 -10 3 8 3-4 0 34 4] sg Proof that pj =0 when i # j. Let [A] be a3 x3 matrix. We can evaluate py for both i = j and i # j. Case 1: Fori=j=1, » Py = Pu= uta A, +a 23 @x 422) Say + ais 433 3, Qs 41 42 43 = det} a, 422 423 (2.40a) 43, G32 33, Case 2: For i=2%j=1, 3 * e Par = LY G2xArx = a2Aiy + @22A12 + a23Ai3 a Py = a2. a3 a2, aa @y, Ar = ay, — a2 + aos 32 433 3, G33 a3, 432 3, yp a3 =det}as; a2. a2;| =zero (2.406) 43; 32 33 Since the first row and the second row are identical, we conclude that py = 0 for i # j In general, to evaluate py with i # j is equival matrix with two identical rows of elements ay with k ent to finding the determinant ofa =1tomMatrix Algebra and Linear Equations 32 Cy Chap.» 2.13.3 Gauss—Jordan Elimination Method This method can be explained by performing an example. Let it be desired to find the inverse of the matrix 3.2 1 [A]l=]2 4 2 (241) 3°12 TABLE 2.1 Matrix Inversion by Gauss—Jordan Elimination Row [A] a Explanation uy | 3 2 1 1 oO (2) 2 }4 | 2 o| 1] 0 B] 1] 2 0 1 [4] 1 i a 4 oO 0 (ij+a, [5] o; ss] 3 }3] 0 0 | (2)-anl4 [6] o | -1 1 | -1 | o | a | B)~asl4) =[3]— 314) 071 0 1 4 ~- é ° [5] + as. =[5] +$ [8] 0 ° 3 a & 1 {6] — (-1)17] [9] 0 0 1 3 4 2 | (8) + ag2 = (8) +3 [10] o 1 | 0 2] $)/4 | m-49 tay 1 o ° 4 -al 0 [4] — $10] ~ 419] In Table 2.1 the identity matrix [I] is subjected to the same process as matrix [A]. Row [1] is first divided by a), to obtain row [4]. Row [4] is then multiplied by a, (i = 2,3) successively and subtracted from the remaining equations. Thus all elements but a,, in the first column are eliminated. Such a process (called pivot condensation) is then performed to eliminate all elements but ai and aa» in the second column. The process is continued until matrix [A] is reduced to an upper triangular matrix with I's along the diagonal (see rows [4], [7 and [9]). We then start from row [9] backward to eliminate the last element from row [7] and last two elements from row [4]. We finally obtain an identity matrix as given in rows [11], [10], [9], 1 0 0 010 t oot -3 The foregoing process is equivalent to premultiplying the left-hand-side matrix [A] by [A] | to obtain an identity matrix [I]. The same process applied to the right-hand-side identity matrix is equivalent to performing (Ay! =[ay'sec. 2.13 Matrix Inversion 33 Thus the inverse of the matrix is obtained as s -4 0 (4y'=| § 4 -} (2.42) 444 2.13.4 Cholesky’s Method The Cholesky method for inverting a matrix is explained step by step using the following example: 3°21 [A] = [: 42 (2.43) 3°12 Step 1: First, let us define [L] = lower triangular matrix (T] = upper triangular matrix with 1’s along the diagonal We then find matrices [L] and [T] such that [A] = [LIT] (2.44a) 322.7 hy, 0 0 Tote ts 24 2/=/hb & Oo 0 1 ty (2.44b) 3°12 ky 2 bs 0 0 1 The nine unknown elements in matrices [L] and [T] are obtained by. evaluating the nine known values in matrix [A]. We evaluate these values in acolumnwise direction starting from the first column. Thus for the first column, we obtain or 3=h Hh =I For the second column, we obtain 2=Mnte=3h2 t2=3 4 = bytia + lar = (2)8) + bee 1 = biti + be = B)G) + be For the third column, [! Hhits=3hs ts =4 2 = hits + betes = (2)G) +33 [; = bits + hates + bs = (3)(4) + (-DQ) + bsu Matrix Algebra and Linear Equations — chyy Chap,» We find 3.0 0 14 4 [L}=]}2 § 0 (m]=]o 1 3 (245) 3 -1 3 001 Step 2: Procedure for obtaining [A]"': [A] = [LIT] (AT)! = (LTT [AT TAIT] = [ATL] (ry! = (art) (2.46) In Eq. (2.46) we already have matrix [L]. If we know [T]', we can find [A] by following the similar procedure in step 1. Because of the triangular nature, the inverse of matrix [T] can easily be found by simple backward substitution. The inverse [T] ' is also an upper triangular matrix with I's along the diagonal. Step 3: Find [T]"' from the equation Of = (rir 1 0 0 Pf s)fen te hs ol =]0 1 3]] uo uaa as (2.47) oot oo U3; M3233, The nine elements in matrix [T]"* are obtained by evaluating the nine known values in the identity matrix [I]. We evaluate these values in row-wise direction and start from the last row. Thus for the last row, we have or ° Us, = a2 = 0 U3 = 1 For the second row, OS uy) +3 Ma = 1 =u t+it2 ua = 1 O= ua tu ay = 4 For the first row, P= uy + Gu + $v un = 1 E 2 1 O = ya + Stan + 32 r 2 1 O = ty + Stas + 33geo. 2.13 6 we have 1-3 07 (mr'=]o 1 -3 (2.48) o 0 1 Step 4: Find [A]"' from the equation (1)! = Carty (2.49a) or 1-3 0 bu be b3s]P3 0 0 Oo 1 -S|)=] 6b, be b3]]2 § 0 (2.49b) o 0 1 bs, baz ba JL3 -1 3. Matrix Inversion ‘The nine elements in matrix [A]"' are obtained by evaluating the nine elements in matrix [T]-'. We evaluate these values in a column-wise direction starting from the last column. Thus for the third column, For the second column, b3=0 For the first column, 1 = 3b), + 26,2 + 3,3 0 = 3by, + 2do2 + 3b25 0 = 36s; + 2bs2 + 3b33 Finally, we obtain ° [Al (2.50) 2.13.5 Properties of the Three Methods for Matrix Inverse |. The adjoint method is the easiest one to use for a matrix of order less than 5, Throughout this text, the matrices for all the noncomputer examples are smaller than 5 x 5. This method is indeed the one to use. This method is convenient when the matrix elements contain variables instead of pure numbers, as is the case for most examples.36 Matrix Algebra and Linear Equations ch, Chap 9 v The Gauss-Jordan method is a well-known method of elimination fo, the solution of simultaneous equations. It has been adopted by IBM f,. the matrix inverse subroutine MINV in the System/360 Scientific Sup. routine Package [2.1]. Such a subroutine is usually implemented in the IBM 360 series computers. 3. Cholesky’s method is a perfectly general systematic procedure for matrix inversion. The entire numerical process is simple substitution. It is used for example, by Argonne National Laboratory in the subroutine LINEQ} in EISPACK. The EISPACK package is well implemented in many CDC computers. 2.14 LINEAR EQUATIONS A set of linear simultaneous equations can usually be solved by one of the following two popular methods. 2.14.1 Matrix Inverse Consider, for example, the following three linear equations with three unknowns: 3x, + 2x2 +x = 10 | 2x, + 4x, + 2x, = 16 (2.51a) 3x, + x2 + 2x, = 11 In matrix form, we have | 3.2 1 xX 10° 1 2 4 2}) x9 =) 16 (2.51b) 3 1 2 Xs. IL or in symbolic form, [Aloe = te} (2.510) ‘The solution is obtained by the use of the matrix inverse: *1 xp = [AT {ec} } % 4-4 07 (10 1 “| ea =] 6 4 -3]f16b =f2 (2.52) =n - — a 3tlu 3Use of Fortran Subroutines for Matrix Inversion a7 geo. 218 2.14.2 Triangular Matrix and Back Substituting ‘The example above can also be solved by first obtaining an upper atrix with I's along the diagonal. This is done by the methods of triangular m | described in Table 2.1. From rows [4], [7], and [9], we have elimination 1 4ypmy pay 8 O 1 4iixebaseap aad (2.53) 0 0 1flxs. cs 3 From row [4], ¢: =’ From rows [5,7], ¢2 = [16 ~ (2)(9)1/3 = % From rows [6-9], ¢s = [11 — (3)4) —(-D@V3 We first find from the third equation, x=3 (2.54a) Substituting x; into the second equation gives Xp = C2 — 3x3 =2 (2.546) Substituting x; and x: into the first equation gives xX = ¢, — 3x2 —fx3 = 1 (2.54c) 2.15 USE OF FORTRAN SUBROUTINES FOR MATRIX INVERSION Nowadays, there is a library of subroutines available in almost every computer system, Matrix inversion subroutines are among the most popular. To perform matrix inversion, we can simply call such subroutines. ‘Alternatively, we may choose to use our own subroutines for matrix inversion together with our main program. For such purposes two example programs are given below. Both subroutines are based on the popular Gauss- Jordan method. Example 2.8 Subroutine MINV MINV is a subroutine available in the IBM Scientific Subroutine Package (SSP) for the 360 series systems [2.1]. The subroutine contains 90 noncommented statements. © use this subroutine, we simply have to use the following statement: COLL MINY Car— 38 Matrix Algebra and Linear Equations — Chap » where A = before the call statement, [A] is the matrix to be inverted; after the cay) statement, [A] is the inverse matrix N = order of the square matrix [A] D = determinant fed back from the subroutine L = column matrix of order N for use in the subroutine M = column matrix of order N for use in the subroutine It is important to note that because a storage-compression technique is used in this SSP, matrices [A], {L}, and {M} must be dimensioned exactly, that is, neither greater nor smaller than N, TIMENSION ACNEND® LOM» MOND Example 2.9 Subroutine MATINV Subroutine MATINV can be found in Ref. 2.2. It contains 72 noncommented statements. The usage is as follows: CALL MATIN Car Ne By My te rity where A = before the call statement, [A] is the matrix to be inverted; after the call statement, [A] is the inverse matrix N = order of matrix [A] B = before the call statement, {B} is the vector of constants of the simultaneous equations; after the call statement, {B} is the solution vector of unknowns M = number of sets of constant vector D=determinant fed back from the subroutine; when ID = 1, [A] is a nonsingular matrix, and when ID = 2, [A] is a singular matrix In the main program, the matrices must be dimensioned as DIMENSION ACLrL >, BKL1) tis important to note that L must be equal to or greater than that which appeared in the dimension statements for A and B in the subroutine. 2.16 EIGENVALUES AND EIGENVECTORS In the analysis of free-vibration and buckling problems, the finite element formulations are in the form of a set of simultaneous homogeneous equations: that is, the constant vector contains only zeros, It is of the form (Al (6 -A (B] i} = 0} mxmmxlomxmmxio mx1 Ges) where the m values of A are the eigenvalues, and corresponding to each eigenvalue A, we have a solution for {x} which is the eigenvector, In thesec. 218 Eigenvalues and Eigenvectors 9 free-vibration problem, eigenvalues are the squares of natural frequencies and corresponding to each eigenvalue, the eigenvector gives the nondimensional mode shape. In the buckling problem, the eigenvalues are the buckling loads and the eigenvectors give the corresponding mode shapes. Matrix [A] is the stiffness matrix and matrix [B] is the mass matrix for a vibration problem and the incremental stiffness matrix for a buckling problem. Equation (2.55) can also be written in such a form that matrix [B] is replaced by an identity matrix [I]. This is done simply by premultiplying by the matrix [B]"' on both sides: [C]{x} — ALL}{x} = {0} (2.56) It is important to note that matrix [C] is the product of two symmetrical matrices [A] and [B]"'. It is commonly a nonsymmetrical matrix. Two methods for the solution of eigenvalues and eigenvectors for the equations above are given next. 2.16.1 Direct Solution Let it be desired to solve for the following equations: 3.2 147m 10 077% 0 221 Xp —-AlO 1 0 Xt = {0 (2.57a) o 1 ifles 0 0 Xs 0 We first add the two matrices: a= d 2 1 x 0 2 2A 1 Xap = 40 (2.57b) 0 1 1-A Xs, 0 or in symbolic form, (Ditx} = (0) (2.576) The value of x; (or x2, or x3) may be solved by using Cramer’s rule, 0 2 1 0 2-A 1 o 1 1-al_o |D| |D| (2.58) x= In order to have a solution for x;, determinant |D| must be equal to zero. This is a necessary and sufficient condition. We thus set |D| = 0, which yields the_-.:.it*:, * = il 40 Matrix Algebra and Linear Equations ~— Chap, “nontrivial solution” (2.59) From this cubic equation we find three roots for the three eigenvalues: Ay = 4.7913 Ao = 1.0 (2.60) Ay = 0.2087 | Corresponding to each eigenvalue, there is an eigenvector (x1, x2, x3). } If we assign a value to any of the three unknowns, say x; = 1, we can find the relative nondimensional magnitudes of the other two variables x, and x2. Thus for A; = 0.2087, x3 = I, the original equations become (3 = 0.2087)x; + 2x. +1=0 2x, + (2 — 0,2087)x.+1=0 (2.61) 0+ x, + (1 — 0.2087)(1) = 0 Solving any two of the three equations gives the eigenvector x 0.2087 x2) = 4 -0.7913 (2.62a) Xs. 1.000 q We can also find for A, = 1.0 and x, = 1, a -0.5 {=} =1 00 (2.62b) Xs 1.0 For A, = 4.7913 and x; = 1, x 4.7913 1.000 xt = 413.7913 or 30.7913 (2.62c) X 1 0.2087 2.16.2 Iteration method (Power Method) For buckling problem, usuall: buckling load) is needed. For vibrai is usually of significant interest, ly only the lowest eigenvalue (the critical tion problem, the lowest natural frequency There is a simple iteration method that givesEigenvalues and Eigenvectors an goo. 2.16 the largest eigenvalue and the relative eigenvector. If we divide the eigenvalue equation by A, we obtain [BH x} — ¢[A}{x} = {0} (2.63) where # = 1/A. This means that the largest eigenvalue for pw is the lowest ue for A, that is, the lowest buckling load or frequency. Using this method vali We first assume the Jet us now solve for the same equations in Section 2.16. initial values for the eigenvector to be Ix} = [1 0 0] (2.64a) ‘Then we perform the simple multiplication 3.2 17/71 3 1.0 (Chx}=]2 2 1]Job=12) =3 0.6667 (2.64b) 0 1 1Jlo 0) 0 We perform multiplication again using the normalized eigenvector. 321 1.0 4.3334 1.0 2 2 1 [2 0.6667) = 43.3334) = 4.3334) 0.7692 o 1 1j}lo 0.6667 0.1539. A similar process yields 1.0 1.0 1.0 1.0 sonsfares «7a 730 arero{ a0 1 ersorm2} 0.1967 0.2062, 0.2082 0.2086 1.0 1.0 1.0 4.79102 0.7914 4.79153 0.7913 4.79134 0.7913 (2.64c) 0.2087 0.2087 0.2087 Finally, we obtain the converged largest eigenvalue 4.7913 and the corresponding eigenvector: 1.0; 0.7913; 0.2087. This method converges rapidly when the two largest eigenvalues are not too close. This method fails when the two largest eigenvalues have the same modulus. It also fails when the initial eigenvector is orthogonal to the expected eigenvector. _ To write a computer program for this straightforward iteration method is very simple. We have only to multiply the matrix by the eigenvector using the Fortran statements described in Section 2.6 and compare the eigenvectors between cycles of iteration for convergence. This iteration method, often called the power method, can be used to extract the remaining eigenvalues and eigenvectors as well, assuming that there are no repeated eigenvalues. The second highest eigenvalue Az and the cot sponding eigenvector can be obtained based on the matrix [D] = [C]~ 41 rre- (1).7 a2 Matrix Algebra and Linear Equations chy, Similarly, once A, and A known, the third highest eigenvalue A, andy, corresponding eigenvector can be obtained based on [D] = ((C] ~ a,c} Al), and so on, The choice of the starting eigenvector is important, , detailed description of the method and a Fortran program are given in Ref. 2 We can now find the remaining eigenvalues and eigenvectors for th example above. Based on [D] =[C] ~ Ai[1] with A, = 4.7913, we can find, by iteration, the following solution: -0.5 —3.79134 0.0 1.0. where A, — A, = —3.7913 or Az = 1.0. Based on[D] = ((C] — A,LI)((C] - a.m), we can find, by iteration, the following solution: 0.2087 3.62624 —0.7913 1.0 where (A3 — A,)(A3 — Az) = 3.6262 or A; = 0.2087. These solutions are the same as those obtained by the direct method. 2.17 USE OF FORTRAN SUBROUTINES FOR SOLVING EIGENVALUE EQUATIONS The subroutines for solving eigenvalue equations are usually implemented in the computer system libraries. The purpose and use of such subroutines are demonstrated by using subroutines NROOT and EIGEN (available in Ref. 2.1) as an example. NROOT has 69 and EIGEN has 105 noncommented statements. We have only to call NROOT, which in turn calls EIGEN. The two subroutines have to be used simultaneously. Purpose: Subroutine NROOT calculates the eigenvalues A, and the matrix of eigenvectors of a real square nonsymmetric matrix of the special form [B]''[A], where both [B] and [A] are real symmetrical matrices and [B] is positive definite. The eigenvalue problem is in the form [1A - arei] 3 = (0) (2.65) Subroutine EIGEN computes eigenvalues and eigenvectors of a real symmetrical matrix. Use: UIMENSION ACHrMd > BCAeND + VALUE CH)» YECTOR EMH) CALL NROOT (My Ar Br VALUE® VECTOR?. If the subroutine for solving for eigenvalues and eigenvectors of a real nonsymmetric matrix is not available in the system library, use NROOT and EIGEN as described in Section 2.17. 2.8. Determine the eigenvalues and the corresponding eigenvectors of the following systems of equations by hand calculation. . lo 2 1 10 0 x 0- 10 1 fo | {=| - {9} 1 10. oo. x 0. M6 -2 10 07] (™ 0 [2 ra fe o| {=|-{2} -12 24 13 ool xy 0 26, Determine the largest eigenvalue and the corresponding eigenvector of the two Systems of equations in Problem 2.5 using the iteration method.CHAPTER 3 Basic Structural Theorems 3.1 PRINCIPLE OF SUPERPOSITION The principle of superposition states that when a linear structure is subjected to a number of loads, the combined effect of these loads is equal to the sum of the effects of each load applied separately in any sequence. This principle can be described by the following example. A linear structure is shown in Fig. 3.1. The structure is subjected to, say, three concentrated loads P,, P2, and P;. Corresponding to the same locations and same directions as the three loads, there are three displacement components with magnitudes q,, q2, and qs, respectively. According to the principle of superposition, the displacements can be written as = firPi + fi2P2 > fisPs 42 = fy Pi + Sr2P2* fasPs ‘i ds = farPy + fa2Pa+ fasPs where fy is the flexibility coefficient or the influence coefficient that define the displacement at i due to a unit load at j. Equation (3.la) can be writt! in matrix form as a 3.10 Loy ¢ a) [ha fa faye Pr al ab =| fa for fos |) Pa : a) Lt fe follesWork Done by @ Load System sec 32 Figs 31 Linear clastic body under there concentrated bouche. or in symbolic form as fa) = (HP) (3.1¢) where [1] is called the flexibility matrix of the influence coefficient matrix. Equation (3.1¢) can also be wntten in inverse form, = (32a) fe=t7' (3.2b) where [hk] is called the unffnen: maine “Stiffness” and “flexibility” are words with opponite meanings As ts evudent from Eq. (3.2b), the stiffness matrix is the inverse of the Meubelity matre 22 WORK DONE BY A LOAD SYSTEM Inthe derivation of stiffness matnit equations of equilibrium for a finite element Cl eee a fe apg ser ne omit eo lm deriving the expression for the work done by a system of loads, it is “sumed that cach load is applied gradually from zero to its final value and that cach load-dispiacement reianonship is linear. Thus the work done by “as load is equal to the triangular area covered by the load-displacement line.sc Sh Basic Structural Theorems cy, , If we use the system in Fig. 3.1 as an example, the work done by ih, three loads is W = $Piqi + 3Paga + 3P sds 1 =41P, Po Psi} az (3.3a) % or symbolically, westria = | 3.3 MAXWELL-BETTI RECIPROCAL THEOREM The reciprocal theorem was first introduced by J. C. Maxwell in 1864 for a | particular case and a general proof of the theorem was given by E. Betti in 1872 (see Ref. 3.1). | Figure 3.2a shows a cantilever beam subjected to two concentrated loads P, and P; applied at two arbitrary points | and 2, respectively. Let it be defined that fi, = deflection at point 1 due to a unit load at point 1 fiz = deflection at point 1 due to a unit load at point 2 ‘for = deflection at point 2 due to a unit load at point 1 fez = deflection at point 2 due to a unit load at point 2 According to the reciprocal theorem, fi2=Su G4) | Figure 32 (a) Cantilever beam unde iis two concentrated loads; (b) Pr PP! before P,. 6 ¢ = oe h,axwell-Betti Reciprocal Theorem, oss M ve 49 Equation (3.4) can be proven by applying the two loads in two different quences: . ee ; sean ading sequence 1: The load P, is first applied. As shown in Fig. 3.2b, tne work done is Wi = 3Pig = PSP (3.5) ye toad P; is then applied. As also shown in Fig, 3.2b, the additional work done is W. = 3Piga+ Prat = 3Po( forP2) + Pi(fi2P2) (3.6) Work W; and work W; are equal to the areas covered by the two load- Gisplacement lines as shown in Fig. 3.3. It is seen in this figure that after the final value of P, is reached, P, is maintained constant throughout the additional jefection qi. Thus the additional area covered by the load-displacement line is Piqi instead of 3Piqi- Load 1 4 ah a Displacement Displacement Figure 3.3. Load-displacement lines with P; applied after the final value of P, is reached. The total work is obtained by summing up W, and W2: = (3.7) Loading sequence 2: If P2 is applied first, the work done is W, = 3P.q2 = 3P2fr2P2 (3.8) The load P, is then applied; the additional work done is W, = Pig + Pras Th = BP fi Pit PofaPr (3.9) © total work done is (3.10)Basic Structural Theorems — Cha, a ior i he total work done is independ tural behavior is linear, t 5 aen f Te ree ot loading. By equating the two work expressions obtaineg e ‘e in Eqs. (3.7) and (3-10), it is shown that Sian te G4) i int 1 due to a unit | Physically, Eq. (3.4) states that the detection ont it load a int 2 is equal to the def ion at point 2 due to au) vad at point a een the example of the cantilever beam, we can generalize the reciproca, theorem as follows: 1. The concentrated loads can be in terms of forces (pounds) and/or moments (inch-pounds). 2. The corresponding displacements can be in term: (inches) and/or rotational displacements (radians). 3. As a result of the generalized loads and displacement coefficients can have the dimensions in./Ib, in./in.-Ib, ra and so on. 4, Equation (3.4) can be generalized as fy =f. Thus the reciprocal theorem Shows that the structural flexibility matrix is symmetric, and so is the stiffness matrix. s of direct displacements s, the flexibility d/lb, rad/in.-Ib, Example 3.1 ‘An arbitrary beam is shown in Fig. 3.4a. When a downward load of 100 Ib is applic at B, the deflections and slopes at A, B, C, and D are measured and are given in Table 3.1. Find the deflection at B due to a counterclockwise moment of 300 in.-Ib applied at B and a downward force of 60 Ib applied at C as shown in Fig. 3.4b. 300 inl Figure 3.4 (a) Arbitrary beam loaded # B; (b) the same beam loaded at Bw (bysec. 34 Energy Theorems ms ‘TABLE 3.1 Measured Deflections and Slopes Due to a Downward Load of 100 Pounds at B A B D Clockwise slope, @ (rad) 0.006 10,021 0.006 -0.032 Downward deflection, q (in.) 0.060 0.240 10.5601 0.0 Solution. From Table 3.1 we first find the needed flexil slope 62 =fssP_ or 0.021 rad = (fgy)(1001b) Sow = 0.00021 rad/Ib deflection gc = fesPs or 0.56 in. = (few)(100 Ib) Ses = 0.0056 in./Ib ity coefficients. From the reciprocal theorem Sac = Sew ‘Thus we can find the deflection at B due to the loads as shown in Fig. 3.4b. 4p = farMa + facPc = (0.00021 rad/Ib)(—300 in.-Ib) + (0.0056 in./1b)(60 Ib) = 0.273 in. ‘The negative sign in front of the bending moment 300 in.-Ib indicates that it is applied in a direction opposite to the positive direction (clockwise) defined for 6p Table 3.1. 3.4 ENERGY THEOREMS 3.4.1 Principle of Virtual Work (Due to Virtual Displacement) One particle. In variational mechanics, we imagine displacements to occur when in reality no such displacements exist. These fictitious displacements are called virtual displacements and the work done by these imaginary displacements is referred to as the virtual work. ___ The virtual displacements are assumed as small so that there will be no significant changes in geometry. Thus the forces may also be assumed to remain unchanged during the virtual displacements. ___ Let us consider a particle undergoing a virtual displacement Sq as shown in Fig. 3,5a. The particle is subjected to a system of, say, three forces Py, Pay and P,, respectively. The resulting virtual work can be expressed a 8W = P, 6g, + P2 5q2 + Ps 54s GADfray a, Opi chee Avup Sirs ane Fi Fei re the abers cacnpanenia wd Le fetione. romnserively. Teeceuse the Tine once d Ace Temes The Seer cane Terr A CIPLUML 21m paLUVELULL, Pees rats sertinrial dum nf a i eqantt fateh at ek sn aT au jeu thie Ey ene on. ber hs famine at tego Ue practice Ao every anid sulin cert ao nanaele ie athe viru! mark dame, By othe Fe yenishce ter any virraa! dien lace nor Auber ot partisan Che oP are bated inet SyseR Tapreesne rie sctines Of Uae publivles tin cash erie) cud dievefose wo On Paleo mya l uel Gu pee tence Abie plicaeaeerun 12 sheer a ur, Peru dinedepuriita myer Tt the rye Au cenmitieeriams . tiq 02 WAV cue rhyée yoda, bse emer: whens aw? wDealeens§ te = fie marticle J FE = errmiad sek at mreciiul Tom oPere ya. He enti 2 fttay pe eri Le numibaren pautlegue NOTOY Theorems 2 W is, for simplicity, defined as the summation of the work for the where f infinite number of particles. gystem ©! Continuum (elastic structure). A continuum is defined as a domain ch matter exists at every point. We may think of a continuum as consisting jn whit | in Wri finite number of particles. of an ‘Theorem. A necessary and sufficient condition for the equilibrium of a-system’ of particles-or a continuum system is that the virtual work done by For an elastic continuum system or an elastic structure, the work done by the internal forces W' is equal and opposite to the strain energy U stored in the system during the deformation: weer 15 substituting Eq. (3.15) in Eq. (3.14) and simply using W to represent the external work W* results in Let us consider a structure subjected to a set of m external forces P,,P.,..., Py...., P, and apply a virtual displacement 6q. This virtual displacement is allowed to take place in the structure in such a manner that 5q is continuous everywhere but vanishes at all points of loading except under P. Thus only the 6g, absorb work from P,; the other external forces do not do any work. __ Due to 6g, the virtual work done is Pq. According to the principle of Virtual work as described in Eq. (3-16), this virtual work equals the virtual strain energy: bU = P. dq, ay (3.17a) ©r, in the limit,5a Basic Structural Theorems — Chap, 4 If instead of virtual displacement 5q, we introduce a virtual rotation 84, a the point where a moment M, is acting, Eq. (3.17a) can be written as Mate (3.17) Equations (3.17a) and (3.17b) are known as Castigliano’s first theorem, after Alberto Castigliano, who published the theorem in his thesis for the engineer's degree in 1879 (Ref. 3.2). It is noted that this theorem is the foundation for the derivation of the finite element stiffness equations, especially for sophisticated elements for which the stress-strain equilibrium approach becomes difficult. It is also important to note that this theorem also applies, when the structural behavior is nonlinear. 3.4.3 Principle of Complementary Virtual Work (Due to Virtual Force) Figure 3.6 shows a load-displacement relation curve. We define W* and U* as the complementary work and complementary strain energy, respectively. Again 6 is variational operator denoting a virtual quantity. From this figure we see that | 6W* = q6P (3.18) Force P Figure 3.6 Load-displavement curve and Displacement q the associated works and energies. The principle of virtual work also holds for the case of complementary work and strain energy: 5W* = 6U* (3.19) In this case the virtual work is that due to a virtual force. 3.4.4 Casti iano’s Second Theorem Let us again consider a structure subjected to a set of n external force P,, P2,-+,P+++, Py Ifa virtual increment SP, is given to the external for*Theorems sec. 34 Energy os p, the complementary work will be increased by 6W*, does any complementary work is P, because the other for, Thus the increase in complementary energy can be obt: The only force that es are not changed. ‘ained, following Eq. ) (3.18), as ‘ 8W* = gap, (3.20) ; Based on the principle of complementary virtual work as given in Eq. (3.19), © y 5U* = 4 dP, (3.21) *h assuming the structural behavior to be linear, the load-displacement curve in Fig. 3.6 becomes a straight line. Then 6U = 8U* = gi 8P, (3.22a) In the limit, Eq. (3.22a) becomes ou (3.22) ind oP, “ aly ‘my Which is known as Castigliano’s second theorem. For linear structural behavior, the principle of superposition can be used. 4s, Equation (3.22b) can, alternatively, be proven by applying 64P, first and after- ward all the external forces. The load 6P, first produces an infinitesimal displacement so that the corresponding work is a small quantity of second order and can be neglected. When all the forces P;, P:, P3,... are then applied, the work (or strain energy) in addition to that produced by all the forces is aU = G dP, (3.22a) which arrives at the same conclusion as given in Eqs. (3.22a) and (3.22b). If we consider a moment M, and a rotation 6, instead of P, and q, respectively, Eq. (3.22b) can be generalized to au = 3.22 aM, cad a ant In applying Castigliano’s theorems, we must be careful that: 1. The displacement g, (or rotation 6,) must be referred to the same location at which the concentrated force P, (or moment M,) is acting. 2. qi (or 0,) must be referred to the same direction as P; (or M,)- 3. The strain energy U must be formulated for the entire structure based on all the external loads. In some literature, Castigliano’s first theorem is often referred to as the ‘heorem of virtual work and Castigliano’s second theorem is often simply Tefetred to as Castigliano’s theorem, and on occasion, as Castigliano’s firstat Basic Structural Theorems Chay, theorem. In that case, the principle of least work is referred to as Castigliang, second theorem, The principle of least work will be described in connectio, with Example 3.5. 3.4.5) Strain Energy Expressions Castigliano’s first theorem is the foundation for deriving finite elemen, stiffness matrices. Its application will be demonstrated in subsequent chapter, Castigliano’s second theorem is very useful in finding displacements ang analyzing statically indeterminate structures. Its application is demonstrate in this chapter by a variety of examples. Before demonstrating examples, we must introduce the strain energy expressions for several basic structural ele. ments. It is assumed that the reader is familiar with the derivations of these ea ‘Gad For a beam in bending: For a beam in shear: yd / u= i sae (3.236) For a bar in tension: : 'S? dx 2 ue f Tea (3.230 For a bar in torsion: 13.238) For a panel in shear: ¢ dx dy v= rary / (3.238 i 2Gt | Area where bending moment (in.-Ib) modulus of elasticity (psi) moment of inertia (in.*)* shearing force (Ib) E/2(1 + v), shear modulus (psi) cross-sectional area (in.?) axial force (Ib) wisting moment (in.-Ib) yusacnms3a) 3b) te) Energy Theorems “east 87 J = torsional moment of inertia dependent on the geometry of the cross section (in.*) q = shear flow (Ib/in.) hickness of the panel or skin (in.) length of the bar or beam (in.) Example 3.2 rind (a) the vertical displacement » and (b) the horizontal displacement u at joint B for the two-bar truss shown in Fig. 3.7a P @ «) © Figure 3.7. (a) Two-bar truss; (b) joint B as a free body; (¢) fictitious horizontal force Q applied at joint B. Solution, (a) For the vertical displacement v at B, let us consider joint B as a free body, as shown in Fig. 3.7b. L Fhorizontat = EF, = 0 S, cos 30° + S.cos30°=0 Sy © Feentiea = E Fe = 0 5; sin 30° — S, sin 30° = P Sp Hence Sil, Su PF “ZEA 2EA EA of (b) In order to apply Castigliano’s theorem to find a horizontal displacement ieint B, a horizontal force must be acting at B. Let us apply a fictitious force Q atInde! pif 22 - vat 58 Basic Structural Theorems Chap, B horizontally. Considering joint B as a free body as shown in Fig. 3.7c, we have LF, =0 — S,cos30°+ S, cos 30° = @ DF.=0 — S,sin30°— S, sin 30° = P which gives s-P+2 and) Se Pp eee vi v3 The strain energy for the truss is 2 2 2 2 Sik , Sik meh [(p+2) +(-e+2) ] 2EA 2EA\ V3. Va U= TEA After the execution of Castigliano’s theorem, we can set the fictitious load Q equal to zero. We thus find Example 3.3 j Find the deflection v at the midspan of a simply supported beam subjected to a } uniformly distributed load w as shown in Fig. 3.8. ] 2 ' ! ' | wdbjiny t | | 7 | Z 1 , | Figure 3.8 Simply supported beam under 1 Ry a uniformly distributed load w. Solution. We first apply a downward fictitious force Q at midspan. The bending moment at x from the left end is wx?tal toi dist geo. 34 Energy Theorems a following Castigliano’s theorem, Dropping out the fictitious load Q, we finally obtain Swit 384ET v If the beam is subjected to a concentrated load P at midspan but no distributed load, the deflection at midspan can be obtained by dropping out w and replacing Q by P: Pe ° 48ET Example 3.4 ‘As shown in Fig. 3.9, a shaft with one end free and the other end rigidly fixed is subjected to a variable twisting moment indicated by the ordinates to the curve. Find the angle of twist of the section at a distance 1/2 from the free end. Figure 3.9 Cantilever shaft subjected to a distributed torque. Solution. We first apply a fictitious twisting moment T; at a distance 1/2 from the free end. The strain energy is Wag? dx i (T+ TP dx a +| —4- . [, 2GI Jy, 26S 260 Basic Structural Theorems — Chay 5 The angle of twist is obtained as Example 3.5 Figure 3.10a shows a cantilever beam supported by a truss bar at the hinged joint 2. The bar can be considered as a spring with an elastic constant of k = EA/L (Ib/in,), Find the compressive axial force R in the truss bar. Solution. We can separate the truss bar and the cantilever beam as shown in Fig. -10b. Thus the beam is subjected to an upward force R at joint 2, whereas the truss bar is subjected to a downward force R at joint 2. For the cantilever beam, the moment equation is 2 M=Rx-** and w bin) 0) Figure 3.10 (a) Uniformly loaded beam-truss structure; (b) deflections of the beam and the truss bar at their common joint (2), respectively. ‘Thus the displacement at joint 2 is au 4 RE wt! 3EI 8EIgen 34 Energy Theorems 6 seo: § for the truss bar with elastic spring constant of k or EA/L and axial force of R, the sain eneray is, from Eq. (3.230), The displacement at joint 2 is Because the two forces R, one acting to the beam and the other to the bar, are opposite in direction, so do the two corresponding displacements v, and v;. Compatibil- ity between the beam and the bar at joint 2 requires that vt 2=0 (3.24) or RP wit | RL 3EI 8EI EA 3Awl* R= _ 8(PA+3L1) Ib If we let U represent the strain energy of the total structure (U, + U2), Eq. (3.24) can be thought of as arriving at by setting au aR (3.25) Equation (3,25) can be generalized as that the derivative of thé total strain energy U with respect to a force (R,) or a moment (M,) in any redundant member of a statically indeterminate structure is zero. In other words, R, or M, has such a value that U is 4 minimum, Equation (3.25) is known as the principle of least work, also referred to a5 Castigliano’s second theorem in some literature. Example 3.6 Figure 3.11a shows a panel-stiffener or skin-rib structure. For the three vertical stiffening bars; EA = 2.5 x 10° 1b, For the four horizontal stiffening bars, EA = 1.6 x 10° Ib. For a ‘wo panels, G = 4 x 10° psi and ¢ = 0.032in. Assuming that the panel does not luckle, find the vertical deflection v at point A due to a load of 10,000 Ib. Solation Based on Castligliano’s theorem, we can find the contributions of deflection iat the strain energy of each structural member and then sum these contributions memiin the total displacement at point A. To formulate the strain energy for each Taunbess we must discretize the structure and consider every member as a free body. ac hulls Of the equilibrium forces acting on each member are shown in Fig Lb. is the stening bar is acting as an axial force member. The value of the shear flow q ‘me everywhere in the two panels.Basic Structural Theorems cy, Nap. 4 2 = 10,000 Ib 1 = 0,032 in, aan bod hod re @ rod ' i 2 2 (b) Figure 3.11 (a) Panel-stiffener structure: (b) discretized free bodies (shear flow q is the same everywhere). Member I: Ata distance x from the top end of the stiffening bar (1), the axill force is ce aS Le 2h aP 2h - =a" Sdx 2 4 s(28) 4 '“ oP aP J, 2(BA), (EA), Jo *\ap) ee def fee) x Pl, ; (EA), i ( at) (a1) ae = TBA, ~ 900° S=noterencos chap 3 Ae Member # Ata distance x from the bottom end of the stiffening bar (4), the at force &S and ae oP 4 au_ot "(Px x & = 9p” (EADs I (- if \(-7) a Ph, = 3(EA)a 0.020 in. Ata distance x from the left end of the stiffening bar (2), the axial ‘Member force is Px S= “2h and * <3 ; i; ( )( 7 2 GP” EA Jo \2h)\ 2h) PR = o(BA),A 0.0185 in. Member 8: The rectangular panel is under constant shear flow everywhere. at aon eee if 8 aP AP Jo 2 (a * P( ) “5 ), i 22h aedy = PE — 0.0260 in ~ 4Gth and 4 f Because this structure is symmetrical, we do not have to repeat the calculations ‘or the members that have the same strain energies. The total deflection at point A is =v tt oyte ts al = 2v, + v4 + 402 + 205 = 0.156 in. REFERENCES 31. 7, Tinoshenk,S. Py Strength of Matra Pas Elementary Theory and Problems, a2, py Yan Nostrand Company, Ine New York, 1958, pp. 336 and 352. Timoshenko, S. P., History of Strength of Materials, McGraw-Hill Book Company, ew York, 1953, p. 289.ic Structural Theorems — Chap, 4 PROBLEMS 3.1. A frame structure is shown in Fig. P3.la. For an upward load of 100 1b, the deflections w and the two rotations aw/ax and aw/ay for points A, B, C, and p | are given as follows. The positive sign conventions are also shown in the figure, Find the vertical deflections at A due to a system of loadings as given in Fig. P3.1b, 2 (bw) @ 60ii b \ NS) 80 inet { (+ 2 direction) 120 Ib an @) Figure P3.1 Location A B c D Deflection, w (in.) 0.120 0.340 Rotation, aw/ax (rad) 0.008 =0.008) Rotation, aw/ay (rad) 0.010 0.010 3.2. A delta-wing panel has been loaded with 160 Ib at point 1, as shown in Fig. P32 The data for the vertical deflection w, the rotation about the x axis 6, and the rotation about the y axis 6, have been measured for every grid point. Some of the selected data are shown below. Positive sign conventions for w, 6, and are also shown in Fig. P3.2a. Find the deflection at point | due to the loads shown in Fig. P3.2b.chop. 3 Problems 65 2€ o Figure P3.2 Location 1 2 3 4 Defie 1.080, 0,380, 1.92 Rotation, 8, (rad) 0.012 0.024 Rotation, @, (rad) 0.080 0.042 0.096, hr BAA (3: \ 3. 3) Find the vertical displacement at joint B of the steel truss as shown in Fig. P3.3. The modulus of elasticity for steel is E = 30 x 10° psi. The cross-sectional areas for the bars are A, = 0.75 in.*, Az i As = 0.6 in., and Ay = Aq = 0.5 in’. In the analysis of an aircraft skin-stringer structure as shown in Fig. P3.4a, we can simplify the structure to be a one-dimensional bar and spring system with both ends fixed, as shown in Fig. P3.4b. The bending stiffness of the stringers is accounted for by the spring with elastic constant of k (Ib/in.) and the in-plane stiffness of the skin is accounted for by the bar with axial spring constant of FA/I(Ib/in.). Let a and AT be the thermal expansion coefficient and the rise in temperature, respectively, find the reaction force at the end of the bar or the compression force in the spring.Basic Structural Theorems — Chay, P= 20 kips \ : @ | _ EA, AT k (b/ing y | (oe © Figure P34 y) i salon 3.8.)For a simply supported beam with length | and under a uniformly distributed \ ~” load w (Ib/in.), find the deflection and slope at a section at [/4 from the left end |, 93.6) Find the vertical and horizontal displacements at point A for the structural system of an L-shape frame and an elastic spring as shown in Fig. P3.6. The frame his a constant bending rigidity EZ The strain energy due to axial force in member BC need not be considered. sy. 34, Find the horizontal deflection and the angle of rotation at the free end oft quarter ring beam with one end fixed as shown in Fig. P3.7. The beam hi | constant bending rigidity EI and EI < EA. Note: cos* @ = (1 + cos 20)/2- 3.8. Find the radial deflections at sections | and 2 for the diametrically loaded circult! beam as shown in Fig. P3.8. Hints: (1) Problem 3.7 is helpful; (2) becaus® ? symmetry, we have only to consider half of the ring beam and the rotation * section 2 vanishes (6; = 0).3 Problem nop: 8 w aab/in), EL c= FF abjiny Figure P3.6 EL, Figure P3.8 39. Figure P3.9 sho} end free. The be: respectively, and is under a uniforml: vertical deflection at the ws a horizontal circular beam with om am is of constant bending and torsional rigidi ly distributed load w (Ib/in.). Find the free end. One application of the curve anet @) 1 = 0,032 in, ~ /™~2 DN J 2000 Ib Figure P3.10 Figure P3.9 ¢ end fixed and the other ies EI and GJ, -d beams can be 18 in,3.10. Basic Structural Theorems Chas a found in the design of a bridge to connect two highways that are not ali Balconies and stairways are among other applications. Hint: At an a distance ¢ from the free end, the bending moment is M = [* and the twisting moment is T = J} R[1 — cos (¢ — 0)]wR do. An aluminum panel-stringer structure is loaded as shown in Fig. P3.10. The typ stringers ABC and FGH are fixed at A and F, respectively. Panel 3 is also fixeg at the edge AF. For all the stringers and the stiffener bars, EA = 2X 10° tb ang for all panels or skins, G = 4 x 10° psi and f = 0.032 in. Find the horizontal ang vertical deflection components at E. igneg, ; MBUlep Rsin ($ — O)wRayCHAPTER 4 Truss Bar Elements 1 TRUSSES A truss is a structure composed of straight bars connected at their points of intersection by means of momentiess joints called pins or hinges. All loadings are assumed to be applied only at these points of intersection. Thus each straight bar is subjected only to axial force, not to bending nor twisting moments. The stress corresponding to the axial force in the bar is called the primary stress in the bar. The truss bars are often connected by riveted or welded joints. The rigidity of the joints indeed induces bending or twisting moments to the joints and consequently to the bars. Such moments superimpose additional stresses to the bars, which are called secondary stresses. However, if the bars are carefully arranged so that their center lines meet in one point at each joint, we shall find that the presence of secondary stresses due to the rigidity of the joints usually does not greatly affect the magnitudes of the primary stresses. Thus it becomes a common practice to assume that the joints are pinned or hinged (i,e., momentless). The plane truss is one of the most important of all structural forms. Examples of application can be found in bridge trusses, roof trusses, truss beams and girders, and truss walls. The simplest plane truss is the one that Consists of three bars forming a triangle. Such a triangular truss is stable, Whereas a four-bar rectangular truss is unstable. If at each of any two joints °f a triangular truss an additional bar is connected lying in the plane of the Criginal truss and the two bars are joined at their outer ends, we have a five-bar, 69a Truss Bar Elements Chop, , four-joint truss with two triangles. If this process is continued, we will haye a stable truss with triangular net. Thus for a stable truss, it can be shown tha, n=2j-3 (4.1) where n is the number of bars and j is the number of joints. A plane truss with fewer bars will be unstable. One with more bars will be statically indeterminate. The bars in excess of the number required for a stable truss are called redundant bars. Three-dimensional or space trusses are also an important structural form, Examples of application can be found in roof trusses for arenas or coliseums, truss structures for supporting huge steam generators, truss towers for supporting power line cables, and truss structures for large, flexible space satellites and the like. A truss bar with hinged ends can be considered as the simplest possible form of structural finite element. The concepts of stiffness matrix equations, assemblage, and solution procedure can best be illustrated in the case of trusses, 4.2. STIFFNESS EQUATIONS FOR A TRUSS BAR ELEMENT IN LOCAL COORDINATES A truss bar element with length L, modulus of elasticity E, and cross-sectional area A lying along a local x axis is shown in Fig. 4.1. The two ends or joints are referred to as nodal points and are numbered | and 2, respectively. There are two forces X, and X, acting in the x direction at nodal points 1 and 2, respectively. Corresponding to the two forces, there are two respective dis. placements u, and u,. These displacements are often called degrees of freedom. There are a total of two degrees of freedom for this truss bar element. Figure 4.1. Truss bar element g along the local x axis. It is necessary to derive a set of two equations in matrix form to relate the two forces X; and X, s to the two displacements uw, and u;. The derivation ca be done either by an energy approach or a stress-strain equilibrium approach The energy approach is more general and more powerful, especially for sophisticated types of finite elements. The stress-strain equilibrium approaet is simple and physically clear. But it can be applied only to simple fini? elements. To use the energy approach, it is necessary first to define a displa* ment function for the element.ns, nal nts 12, lis- geo, 42 In Loval Coordinates n 4.2.1 Displacement Function and Shape Functions For a bar element with constant axial stress or constant axial strain, the axial displacement u(x) at a distance x from nodal point I may be assumed as varying linearly with x, or u(x) = aytarx = a+ [MM ) % (4.2) Byu, +E) where a, and @ are two constants to be determin conditions, z Uy 7 ae (%) Atx=0, fe fur u(x) = u(0) =u, = a, Atx = L, u(x) = u(L) =u: = a, + ab uy = My b Substituting the results for a, and a, into Eq. (4.2) and rearranging the equation gives the final form of displacement function: a, = u(x) = fi(x)en + i) ua (4.3a) with (AG)=1-7 and Best 7 (43) where f,(x) and f,(x) describe the distribution or shape of the displacement associated with the degrees of freedom wu, and up, respectively. They are referred to as shape functions. Physical insight into the displacement function (43a) and the shape functions (4.3b) can be explained by the following examples. ___ First, let us check Egs. (4.3a) and (4.3b) by setting x = 0 and L, respectively. For x = 0, we find f,(x) = | and fa(x) = 0, which yield u(0) = us. For X= L, we find f,(x) = 0 and f,(x) = 1, which yield u(L) = u2. Next, let it be required to find the displacement u at x = L/4and x = L/2 for a bar element with u; = 0.04 in. and uz = 0.08 in. For becomes this case, Eq. (4.3a) = z u(x) = (: = 2) (0.04) +7 (0.08) y the two nodal point*7a Truss Bar Elements — Chap. 4 } With the values of x = L/4 and L/2, we have | «(2) = 0.05 in. | Dp j u (5) = 0.06 in. 4.2.2 Stiffness Equations by Energy Method For the case of a uniaxial stress or strain, the strain is defined as e= lim u(x + Ax) = u(x) _ 9M 4 (4 sx) (44) aro Ax ax 4x where Ax denotes an infinitesimal length. For the present axial force bar element, the axial strain can be obtained by substituting Eq. (4.2) or (4.3a) into Eq. (4.4): _ Devils ~ (45a) or € = fila) + foe) ue (4.56) | Equation (4.5a) states that the strain is a constant. The primes in Eq. (4.5b) indicate derivative with respect to x. The axial force is S=cA=EeA= FAX ax = BALSi(x) us + f2(x) uo) (4.6) The strain energy expression is obtained by substituting Eq. (4.6) into Eq. (3.23¢), “Sax _ EA (* i = ies a 2 | i 3EA 7 2 |, Sim + Pian)? dx (an Applying Castigliano’s theorem gives rues = (ea [nn dx) us +(Ba fj Sfede) ws == (caf arax)u +(e J pra) ue xA) bar 3a) Sa) ~5b) .5b) 4.6) Eg 4.) goc.42__In Local Coordinates 73 or in a matrix form, {x} 7 ie ki] fu a Key a>} ue (4.8a) {X} = [kHu} aa) where [Kk] is called the stiffness matrix, in which the stiffness coefficients are defined as or symbolically, L = BA f SU )Sj0) dx (4.9) with i= 1 to 2 and j = 1 to 2. Mathematically, Eq. (4.9) shows that ky = ky, so that the stiffness matrix is symmetric. Physically, we have already shown that the stiffness matrix is symmetric due to the reciprocal theorem. As will be seen in subsequent chapters, Eq. (4.9) is ina form quite typical for obtaining stiffness coefficients for a finite element with given shape functions. Substituting the shape functions (4.3b) into Eq. (4.9), the stiffness equations are finally found: (4.10) The term EA/L is the axial stiffness of the bar. The bar behaves like a spring with spring constant equal to EA/L in the units of pounds per inch. 4.2.3 Stiffness Equations by Stress-Strain Equill Method Let us first assume that the bar element is fixed at nodal point | but free at nodal point 2, as shown in Fig. 4.2a. This means that u, = 0. gO FALL © Xp, 20 tne @ © Eat © a «> 1d at nodal point 1; (b) bar element fixed at i nodal Figure 4.2. (a) Bar element fixes Point 2.74 Truss Bar Elements Chap, g Because EA/L is the spring constant in pounds per inch and u, =0, x= Au, (4.11a Equilibrium in the x direction requires that -FA,,, (4.116) L ) Let us then assume that the bar element is fixed at nodal point 2 but free at nodal point 1, as shown in Fig. 4.2b. This means that u, = 0. Then EA xa (4.11¢) Equilibrium in the x direction requires that EA x= -x,=-FAu, (4.114) Superimpose the two cases: combining Eqs. (4.11a) to (4.114), stiffness equations identical to that given in Eq. (4.10) are obtained. 4.3 STIFFNESS EQUATIONS FOR A TRUSS BAR ELEMENT ORIENTED ARBITRARILY IN A TWO-DIMENSIONAL PLANE Having derived the stiffness equations for a horizontal truss bar element lying along a local coordinate axis, we now consider a general case of a bar element oriented arbitrarily in a two-dimensional plane. We can derive the stifiness equations by using an energy approach combined with a coordinate transfor mation approach, a pure coordinate transformation approach, or a stress-strain equilibrium approach. 4.3.1 Energy Method Combined with Coordinate Transformation A truss bar element lying along an % axis is shown in Fig. 4.3. The * axis is a local coordinate axis and the axes (x, y) are the global or referent coordinate axes. The x axis is oriented at an arbitrary angle measured counterclockwise from the x axis. To distinguish between the two coordins systems, bars are added: to all the symbols referring to the local ¥ axis. in Of global coordinate system, cach nodal point has a horizontal force X, a vert force Y, a horizontal displacement u, and a vertical displacement v. Thus eat! element has a total of four degrees of freedom: uy, 0), v2, and v2.goo. 43, Two-Dimensional Plana PA Figure 4.3 Local (% 7) and reference (x, y) coordinate systems for a truss bar element. From Fig. 4.3 we see at nodal point 1, %, =u, cos +0, sind (4.12a) and at nodal point 2, i, = uz cos b + ws sind (4.12b) If we use the symbols A=cosé and w=sing Equations (4.12a) and (4.12b) can be written as a@) _[A 2 0 O]|u - {@}-[2 OA °| u, (413) ‘ As explained in Chapter 3 in connection with Eqs. (3.3a) and (3.3b), the strain energy or the work done is equal to one-half of the nodal forces multiplied by the corresponding displacements, U =3(X ia, + Xzi2) (y) aw Equati Pe A - ‘uation (4.10) can be rewritten for an element in a local ¥ axis, {=} -#4[ 1 Ate) (4.15)°76 Truss Bar Elements Chap, Substituting Eqs. (4.13) and (4.15) into Eq. (4.14) gives u,|7PA 0 uy y =FA v, we 0 [ 1 “YE we 0 | vy 2L ju{ |o all-1 jlo 0 a wl)u vw) LO vy uy) PAP Aw AR Aw] (tt EA)», Ag 2 Ap —p? |) 0, 20 Vas “2 en ee as ‘s (19) 7 v2) [Am yw? Aw pw? |e The stiffness equations can be obtained by using Castigliano’s theorem: ou, 2 2 ou, » Ab \ A —Ap| | uy au A a Ae a? : _ |e: | _ zal 28> # we el Oy ait A au Ey na ae > us. ; ou, Be ( Me 2 au \ het = 7 oe. Me me \ Aw » v. This stiffness matrix has an easy-to-memorize pattern. We can simply remember the stifiness matrix as Lk (4.18a) where al (4.18b) he 4.3.2 Coordinate Transformation Method As shown in Fig. 4.3, the two-dimensional bar element has a total of four degrees of freedom. If we create an additional nonexistent force Y and a displacement 0 in the ¥ direction at each nodal point, we can relate the fout degrees of freedom in local coordinate directions to those in global coordinaté directions as a A pw 0 O}{u al _|-n a 0 oljy, (4.198) cy 0 0 A plu, D2 0 0 -p Alle_ 1) aply 18a) 8b) in. a Two-Dimensional PI geo td ne jane or symbolically, {= [Tia (4.19) where [T] is the coordinate transformation matrix. This matrix was also derived in Example 2.3. We can relate the four nodal forces in local coordinate directions to those in global coordinate directions by the same coordinate transformation matrix as that derived for displacements: {P= (7) (4.19¢) Because the transformation matrix [T] is of order 4 x 4, it is necessary to augment the stiffness matrix for local coordinates as given in Eq. (4.10) from order 2X2 to 4x4. This can be done simply by putting zeros in the additional rows and columns corresponding to the nonexistent forces Y,, Y» and displacements 0, 52. xX, 10 -1 O}fa ¥,[_FA|0 0 0 ofa, X%(~ 2 |-1 0 1 olla G.20a) Ye 0 0 0 offs or symbolically, {F} = [kl{a} (4.20b) Substituting Eqs. (4.19b) and (4.19c) into Eq. (4.20b) gives (T1F} = Cea} or {F} = (1) '(kKITHa} Because the transformation matrix [T] is an orthogonal matrix (i.e., its inverse is the same as its transpose), we have {F} = [kl{a} (4.21a) a mrern | coun This stiffness matrix is in a congruent transformation form. It is in precisely the same explicit form as that given in Eq. (4.17). 4.3.3 Stress-Strain Equilibrium Method with Let us first assume that the bar element is fixed by a hinge at nodal point ‘and supported by a roller at nodal point 2, as shown in Fig. 4.4. The rollerited (atin Yerba! meee ene by: it mes thar eo Learns ne mene rectal js wcclnoten Tis Figs td. charged: He heel wire Alonnatinn oo mpeng a an The stad Paice duuivuleas i such ehougativas de BAR Ls Po tapeing eeinarae my in | alah ES ee“ panes Bide: BS bene Satu, 14. aA Ta Taig = au: wo awe Inet Bory and ceming tan the Guulhbe un . EA ae a8 £,_ Jn a Two-Dimensional Plane god r7) gquations (4.22a) and (4.22b) can be written in the following form: x Av Yi( _ BA )-ap X> L ) a2 ( {ust (4.23) Yo Au iris readily seen that Eq. (4.23) yields precisely the third column of the stiffness given in Eq. (4.17). matrix . By the same procedure, we can derive Ist column Uy = wh, v, = uy = v2 = 0 2nd column} of [k] by setting }v, = v,, u, = uy = v, = 0 4th column y= 03,0; =v, Sy = 0: This derivation and the subsequent procedure for assemblage and solution are due to Turner et al. [4.1]. This pioneering work was done during the emerging stage of electronic digital computers. The reader is referred also to Ref. 4.2. 4.3.4 Properties of the Two-Dimensional Bar Element Stiffness Equations Equilibrium. Let us consider the two-dimensional truss bar element shown in Fig. 4.3 as a free body. The stiffness equations for this element are given in Eq. (4.17). We see in the stiffness matrix that the coefficients in the first row are the same but opposite in sign as the coefficients in the third row. The same relation holds between the second and fourth rows. If we multiply cut the matrix equations, we find that X,=-X, and Y=-Y2 So the equilibrium in both the x and y directions for the free body is satisfied. We then take the moment about nodal point 1, CEM = XaLbye — ¥2LA ca Y= BAL(HA2e + AP) + HAN? + dw?) OL + (02 — AZ )ua + (Aw? = Aw?) v2] =0 Whi ati. a ch satisfies the moment equilibrium conditions. Singularity. In the stiffness matrix [k], the first and third rows are the Same by Tl k] t opposite in si second and fourth rows. Thus [k] is Singyl,, OPPOsite in sign, as are the seco s say ants Possible : It cannot be inverted and a solution to Eq. (4) ., ———5 Truss Bar Elements — Chyp 4 Physically, singular stiffness matrix means that the element, without an support, is an unstable free body. The element can become stable and thy stillness matrix can become nonsingular if the element is properly supportey One of the ways to stabilize the element is to support it as shown in Fig, 44 In that case, u; = v, = v2 = 0. If we consider only the third equation in the stiffness equations (4.17), we find that WEA iE {Xa} = {uz} The reduced stiffness matrix is of order | x 1 and is no longer singular. 4.3.5 Equation for Axial Force If we know the nodal displacements for a bar element, either given or solved, the axial force can be obtained directly from Eq. (4.17). Let us multiply out the third and fourth equations in Eq. (4.17), X= FA a%(uy— a) + eles — 21) Yo ruin — uy) + 27(v2 — 0,)] If we use S to designate tensile axial force, then S=XA+ Vou = FOr We = 4) + Ces 0) EA Ate = m1) + (02 — 01) = (spring constant)(total axial elongation) where (u;—u,) and (v,—v,) are, respectively, the horizontal and vertical components of the axial elongation. In matrix form, the tensile axial force is given by If the value obtained for S is negative, the axial force is in compression:In Three-Dimensignal Space a seo. 44 44 STIFFNESS EQUATIONS FOR A TRUSS BAR ELEMENT ORIENTED ARBITRARILY IN THREE-DIMENSIONAL SPACE Let us now consider a bar element oriented arbitrarily in three-dimensional space. There are two sets of Cartesian coordinates: the local coordinates (3,2) and the global or reference coordinates (x,y,z). The local % axis coincides with the axis of the bar. Let ,, m,,m; 2, m2, m2; Is, ms, ny be the direction cosines of the ¥; ¥; 2 axes relative to the x;y; z axes, respectively. It has been shown in Example 2.4 that x hom, n(x ¥ bom miyy (4.25) z booms ns|lz The bar element in space has three forces and three displacement degrees of freedom (X, ¥,Z and u,v, w) at each nodal point in the global x, y,z directions, respectively. The element in local coordinates has only an axial force X and an axial degree of freedom @ at each nodal point. Let us first augment the 2X2 stifiness matrix equations in local coordinates to 6 x6 stiffness matrix equations by creating two more nonexistent forces ¥,Z and displacement degrees of freedom #, at each nodal point in the local 5, Z directions, respectively. Thus Eq. (4.10) becomes x 1 0 0 -1 0 07fa/ ¥; 000 0 0 O|lxy, Z,(_EA}| 0 0 0 0 0 Oj;Jw, Xf L|-1 0 0 1 0 Oj)a C262) Ya 000 0 0 Offa Zz 000 =O 0 OJlws or symbolically, {F} = [kia} (4.26b) Based on Eq. (4.25), the six nodal degrees of freedom in local coordinates can be related to that in global coordinates as — Zz A) Dead a hom on 0 0 O7(uy a Lom m 0 0 O|})on |* w _ EA) hm; ny 0 0 Oflw |” (4.27) it L}O 0 Oh m ml)u o, 0 0 0 bh m mi}e Wy 0 0 0 bh m mj lw| | 2 Truss Bar Elements Chap, or symbolically, {q} = [TKa} (4.275) Where [T] is the transformation matrix. It is an orthogonal matrix (i.e, jy, transpose is its inverse). The six nodal forces in local coordinates can be related to those in globa) coordinates by the same matrix, {F} = (THF) (4.28) By using the same procedure as that used in deriving Eq. (4.21a) for the two-dimensional case [i.e., by substituting Eqs. (4.27b) and (4.28) into Eq (4.26b)], we obtain for the three-dimensional case {F} = [k]{q} (4.29a) with (k] = (7) (KT) (4.296) In an explicit form, Eq. (4.29a) is obtained as x My yn on “| BA [ fe a1 wt (4.302) Xx L ko ua . Ys vb 2 We with Room mh [ke] =|im, mi mn (4.30) nl omny nyt ‘The subscript is 1 throughout matrix [Ko]. For simplicity we can drop out the subscript Po olm al [ko] =] im om? mn (4.300) nl mn n? where /, m,n are the direction cosines between the ¥ axis and the x, y, 7 axes. respectively. It is observed that this 6 x 6 stiffness matrix is in a pattern similar to the 4 <4 stiffness matrix in Eq. (4.17). This matrix is also a singular matri If we know the nodal displacements for this element, either given 0 solved, the three force components X>, Y>, Z> at nodal point 2 can readily b& obtained from Eq. (4.30a). The tensile axial force S is the resultant of thes? athods of Assemblage and Solution ga 48 MO 83 ree forces: Letting J, m,n be the direction cosines of the element axis relative thre“. x, yz axes, respectively, and making use of the fact that ? + m?+ n? = | yields f S = Xj1+ Y¥,m+ Zn EA wu — uy =p Ulin} yo v (4.31) W.-W, which is in a form similar to Eq. (4.24). 45 METHODS OF ASSEMBLAGE AND SOLUTION DEMONSTRATED BY A THREE-BAR TRUSS. 4.5.1 Assemblage To demonstrate the method of assemblage, an example of a three-bar truss as shown in Fig. 4.5a is chosen. All bars have the same length / and axial rigidity EA. The truss has three nodal points and six degrees of freedom. To formulate the total system, we must formulate the individual element first. Table 4.1 lists the direction cosines and sines for the three elements. Poy V3 Py fi © Py vy Pq, 02 @ (oy Figure 4.5 (a) Three-node, six-degree-of-freedom, three-bar truss; (b) boundary and loading conditions for the truss. F ‘r element 1-2, we have, from Eq. (4.17) and Table 4.1, x 10 -1 0]fu Yi[_FA| 0 0 0 Ojju (4.32a) x.f 7 |-1 0 1 0})m Y2 0 0 0 ofl»| Truss Bor Elements TABLE 4. Direction Cosines and Sines for Rach ment Element a ” Me 12 o 1 0 1 0 . WD js 1 43 23 120 +} 7 i 7 1 3 ! 4 1-3, 60" 2 2 4 a For element 2-3, Xe -¥3 -1— V3 | (us 1 v3 3 V3 -3| Ju X3{ at -1 v3 1 -v3 | | 4a -3 WB 34 ky For element 1-3, X, 1 V3 -1 -V3]{u, ¥,|_ BA | V3 3 -v3 -3\Jo, X(~ 4F | -1 -v3 1 V3] )us we v3 -3 V3 3} ley sivaiwe | ® Chap 4 (4.32b) (4.32c) We must bear in mind that the X’s and Y’s are the internal nodal forces, When the three sets of element stiffness equations are assembled, the sum of such internal forces in either the x or the y direction at each nodal point is equal to the We must also beat he P, = X, of element 1-2 plus X, of element 1-3 P,= Y, of element 1-2 plus Y, of element 1-3 P; = X, of element 1-2 plus X, of element 2-3 Ps Ps ¥, of element 1-2 plus ¥; of element 2-3 X; of element 1-3 plus X; of element 2-3 Ps = Y; of element 1-3 plus ¥; of element 2-3 mind that at each nodal point the ction. Thus 1 D (4.33) ) to (4.32c) to obtain the internal forces X and Y for ae element, and then to sum them up as indicated itbo) 2 Methods of Assemblage and Solution 7 ‘gee: 45 9 (433) , | Ga & 4 ah Tu Pp v3 @,_ 09 0} -v3 =3 |} pl fA) 74 OED valli v3 [Ju p{ 4+ 9 0 -v3 3\) v3 3 o (4.34) P -1 -¥30 -1) V3 141 V3-v3} }u, na -v3 -3) V3 -3 V3-V3_ 343 Jl The foregoing assemblage procedure is equivalent to move three 4 x4 element stiffness matrices into a 6 x6 total stiffness matrix as illustrated in Fig. 4.6. For elements 1-2 and 2-3, each 4 x 4 stiffness matrix remains as a block during: assemblage. For element 1-3, the 4 x 4 stiffness matrix is separated into four blocks (2 x 2 submatrices) during assemblage. The positions where each block moves into depend on the related degree of freedom numbers. In portions where two blocks overlap, the element stiffness coefficients are superimposed. YA Ato EA +-H EFSSSS) EF 4 ka) Lkp5, tky.5] Total [K] Figure 4.6 Assemblage of the three element stifiness matrices. 4.5.2 Solution for the Truss Problem __ Lett be assumed that the truss is supported and loaded as that shown in Fig. 4.5b, We are to find the values of the six displacementdegrees of freedom, the four reaction forces, and the three bar axial forces. The supporting conditions are u; = v2 = us = %3=0. Only o and ws temain to be solved. Let us rearrange Eq. (4.34) so that v1 and u, and the corresponding P; and P, appear first in their respective columns. First, let us rearrange the positions of rows as V3 =! P, Ps Pil_ EA Py Ps Pp. | |86 Truss Bar Elements Chop , ‘Then let us rearrange the positions of the columns as “3 (4.35) Multiplying out the first two rows gives 0) _EA[3 0O]f{v {o al [3 3] (nh ac which gives vy 4l 5 0 0 4Pl {0 {i} ~ 15EBA [3 sf} ~ SEA {} (2a It is noted that Eq. (4.36a) can be obtained from Eq. (4.34) without going through the rearranging and partitioning processes described in Eqs. (4.35a) and (4.35b). This is done simply by crossing out those columns and correspond. ing rows in the total stiffness matrix which are related to the zero degrees of freedom; that is, Eq. (4.36a) is obtained by crossing out the first, fourth, fifth, and sixth columns and rows in the total 6 x6 stifiness matrix. Having found v, and w2, we can find the reaction forces by multiplying out the third to sixth rows in Eq. (4.35b): P, v3 -4 4 P,| __ BA o -V3| fo) Pe \AN3 4 = = (437) Ps ar |-V3 -1) les 5) -1 Ps 3 V3, v3 Now we have all six P’s. They satisfy equilibrium if we consider the whol truss as a free body. The axial forces S;.2, S2.3, and S3., for the three bar elements are found by using Eq. (4.24): EA u— uy Si2= ie mal f ‘ il v2 — Oy 4Pl EA — P =—-l o fsa =4F (tension) tof § En-some Basic Treatments of Stiffness Equations Flas mal {Out sec. 46 Dy — v2 ~,(~4Pr EA|_1 V¥3|)=—| 2p -*4|-4 3 ae Ss (tension) EA ene aA pm Saaz Wis mn {22} = FALE 4 |{o} =o Fl2 2 os” Having obtained the axial forces, the reaction forces may be found. by considering equilibrium of each joint instead of using Eq. (4.37). Thus the formulation of the 6 X6 total stiffness matri i net (436a).is necessary. 4 SOME BASIC TREATMENTS OF STIFFNESS EQUATIONS After we have been familiar with the solution procedure of Section 4.5, it is easier to understand some basic procedures of treatment of stiffness equations as explained in the following symbolic form. 4.6.1 Treatment of Boundary Conditions Let it be assumed that the stiffness matrix equations assembled for a finite element structural system are obtained as {P} = [Klta} (4.38) Based on the prescribed zero displacement boundary conditions, Eq. (4.38) may be rearranged and partitioned as demonstrated in obtaining Eqs. (4.35b): {ry = [e pel {x = i} (439) PJ ~ LK: Kool la: = 0. re } contains the unknown unconstrained nodal degrees of freedom, contains the constrained nodal degrees of freedom (zeros), {Pi} contains t! * external loads related to the unconstrained nodal degrees of freedom, and el Contains the unknown reaction forces related to the constrained nodal Sees of freedom. Where {q, {a:¥ a Truss Bar Elements — Chay , Multiplying out Eqs. (4.39) gives {Pi} = (Kia (44a) and the unknown reaction forces are given by {P2} = (Ko Ha} = [Kai Kid {Pit (4.40¢) It is a common and efficient practice to formulate only the smaller matrix [Ky] instead of the whole matrix [K] based on the zero-displacement cond; tions. Matrix [K),] in Eq. (4.40a) is sufficient to yield all the unknown degrees of freedom so that the reaction forces can be obtained by using the necessary element stiffness equations instead of matrix [K>,]. In computer programming, to formulate only [K,:] instead of [K] is essential in saving DIMENSION and computer storage. so that 4.6.2 Reduction Procedure In an actual problem, not every unconstrained nodal degree of freedom | is subjected to external load. The load vector {P,} in Eq. (4.40a) may containa | certain amount of zeros. Based on these zeros, a reduction procedure may be developed. For simplicity, let us rewrite Eq. (4.40a) by dropping out the subscripts {P} = Ka} (441) where the vector {q} may no longer contain zero degrees of freedom With the knowledge of the zero loads, we may rearrange and partition Eq. (4.41) as P, | [xe Kel fait 5 = 42) ica Ky Kod la: « where {P,} contains the actually applied loads and possibly some selected zeros, {P2} is a zero load vector whose size can be chosen to be less than the actual number of zero loads, {q,} contains the degrees of freedom correspont ing to the loads {P,}, and {q,} contains the degrees of freedom correspondint to the zero loads {P2}. Multiplying out Eq. (4.42) gives | {P,} = [Ki Haut + [Kiqo} (4.439) {Ps} = {0} = [KayKau) + [K (4.430) From Eq, (4.43b), we can write ( Aae) = “0 KooT (Karan)oe some Basic Treatments of Stiffness Equations Pat geo 48 oted that matrix [K22] is a square-symmetric nonsingular submatrix. tess (P,} and {P2} are of the same size, matrix [K2] is a nonsquare matrix. Unk is square, itis nonsymmetric and often singular, [Ree ipstituting Eq. (4.43) into Eq. (4.43a) yields {P.} = [Kia} (4.44a) 18) = 0K.) 06.01) (4.44b) a}= (RNP) (4.44c) After the displacement vector {q,} is obtained, a back-substitution of {q,} into Eq. (4.43c) will give {qo}. By the use of this reduction procedure, we do not have to invert the whole matrix [K]. Instead, we invert two smaller matrices, [K22] and [K]. The sum of the sizes of [K32] and [K] is equal to that of [K]. For a large set of stiffness equations, this reduction procedure can result in a considerable saving of computing time. It is 9 with Hence 4.6.3 Joints with Prescribed Displacements Instead of Loads In certain problems, a nodal point may be subjected to prescribed displacements instead of loads. Forced fit during structural construction is an example. Sometimes, certain nodal displacements may be known from measurement. Let us start with Eqs. (4.41), for which the boundary conditions have already been accounted for. The equations are rearranged and partitioned according to the prescribed displacement conditions, (r= [er chat ad (44s) where (q,} contains the unknown unconstrained nodal degrees of freedom, ‘&) = {@} contains the prescribed nodal degrees of freedom, {P,} contains the teow external loads corresponding to {q,}, and {P2} contains the unknown res required to produce the prescribed degrees of freedom {q2}- Multiplying out Eqs. (4.45) gives {P,} = (Ki Has} + (Kika) (4.468) 7 {P2} = [Kz Kay} + [Kz2]@) (4.466) “unknown degrees of freedom may be found from Eq. (4.46a): fa.} = [Ki J '({Pi} — [Kio t@)90 Truss Bar Elements Chap, The unknown forces corresponding to the prescribed degrees of freedom ma, be obtained by back-substitution of {q)} into Eq. (4.46b). 4.6.4 Initial and Thermal Forces in Truss Bars During construction of a truss structure, a certain bar could be fabricateg with improper length. Thus the bar is forced into its position between tg joints by applying some initial extension or compression. Under such , ‘condition, some axial forces are induced in the bars even in the absence o external loads. Such a condition may also arise when the structure is subjected to a temperature change. The stress analysis for such a condition is divided into two steps and the results of the two steps are then superimposed. Step 1: Let all the joints be assumed as restrained from displacements and the axial forces developed in all bars due to initial extensions (or compres. sions) or temperature changes are determined. The axial force for a typical bar element is given by S= pant (4.474) where the overlength AL is positive when the bar is longer than the distance. L between its two joints. The axial force S is positive when in tension, The negative sign is introduced to indicate that the bar is in compression if it overlength is suppressed by restraining the joints in their original positions In the case of a temperature rise AT, Ss —EAa AT (4.476) where a is the coefficient of thermal expansion (strain per degree) and ATi positive when the temperature rises. Again, the negative sign is introduced t? dicate that the bar is in compression if its thermal expansion due to posititt AT is suppressed by restraining the joints in their original positions. For a typical element as shown in Fig. 4.3, the nodal forces required © restrain the nodal points or to maintain the initial axial forces S as obtain’ in Eq. (4.47a) or (4.47b) are determined as x —A A x Yi Ke AL | Ss = Sey. a 448) xX, ar ea(42) Jt of BAw AT “| ( ¥ ” KB [eos 30° sin ws mi} = 5 (tension) EA From symmetry we know that P Sus=5 Example 4.2 Indeterminate truss solved by reduction method A symmetrical indeterminate truss subjected to a pair of parting forces is shown in Fig. 4.8. All bars have the same axial rigidity EA. There is no connection between the two diagonal bars. Find the displacements at all joints and the axial force for each bar. Solution. The truss is composed of six bar elements, four nodal points, and eight degrees of freedom, as shown in Fig. 4.8. The element stiffness matrices are obtained based on Eq. (4.17) and Table 4.2. Pas Ys Figure 4.8 Truss and the numbering of its nodal points and degrees of freedom— seo.47 Examples ” 4 TABLE 4.2. Direction Cosines and Sines for Rach Element Element a # ” Au Length l 1 1 1 1 . is = ; t 5 1 i 2 R 2 2 2 1 1 1 1 1 3,14 45% -= t t -+ 1 a 2 2 2 2 a 1 1 1 1 1 3400 -BS 5 ; - 1 v2 v2 2 2 2 13 o 1 0 1 0 0 vu 2-4 —90" o -1 o 1 ° v2 For element 1-2, ‘ ‘| Ki Ta) -1 -1) (u, HaSV A t ) y,[_ga|\i1 1{-1 -1])e, X,{ 2 ]-1 -1 1 1|)w Y; ter -1 1) ly a For element ~j with (i,/) = (2,3) and (1,4), eat fe ryt pe ah EA\=1 1) 1 -1|)a 4 aij-1 1 1 -1|)y tis r-r-1 tly For element 3-4, Xs 1 oot -t -¥]\(us f y,|_ BA; 1 ot -1 -1|)o \ X,(- 2b |-b -2 1 1} ) uy Ya. -1 -1 1 ly For element 1-3, X, 10 -1 O}(u, Lie A Yi [BA] 0 OF 0 Oi) 0% ofisv > lor x,( V2i}-1 0 1 Of}uy Ys 0 0 0 Olly For element 2-4, X, 0 00 uy . ? y,|_ BA]O 1 0 -1])o +A» \ ' x,( V21]0 0 0 Of)us 2 v4 o -1 0 tthyTruss Bar Elements Chay , Kia kaa Fo haa @ ib) Figure 4.9 Positions the six element stiffness matrices move into during assemblage. Figure 4.9 shows the positions where the six 4x4 element stiffness matrices move into during assemblage. Both 8 x 8 matrices shown in Fig. 4,9 are to be superim, Posed. Because the associated degrees of freedom in the total stiffness matrix are Scauenced from 1 to 8, each angle ¢ in Table 4.2 is measured at i for element i) with i [1 oj {" a =o S35 (tension) ‘The reader should check all these axial forces to see whether they satisfy equilibrium at all joints. ‘uss with a joint sliding along # slope uss with four joints fixed and one joint ‘All bars have the same axial he displacements at joint 1, d sines for each Example 4.5 Statically indeterminate tr pees, five-joint statically indeterminate tr rai bie slide along a 45° slope is shown in Fig. 4.11. eos Find the reaction force normal to the slope, # cater xial forces in all bars. Table 4.4 lists the direction cosines ant ints are fixed, the structure has only (WO degrees Solut ion. Because four out of five joi e assembled as of freedom. 1 om. ‘The total stiffness equations can b fy -[se ee]y | 700 Truss Bar Elements cp, 4 Figure 4.11 Truss with a joint confined to slide along a slope. where 8 1 Because joint | is confined to slide along a 45° slope as shown in Fig. 4.11, we havt | v= —uy ii a TABLE 4.4 Direction Cosines and es for Each Element y Element * A a ae rd Aw Length 1 o ! 3 1 3 8 21 t 2 2 4 4 4 3 1 3 1 ey 5 ‘ 1-3 30° = = => 23 i 2 2 4 4 4 i i 14 o 1 ° 1 0 0 f v3 = oo 1s we lB 3 1 ETE) 2 2% 4 4 4& hat Examples 10 ere is a reaction force R normal to the slope, the external loads are R v2 P, = Rsin 45° ~ P pecwuse th P, = Ros 45° = ifiness equations become rR v2 ™ DAaked Lo Pie ~maprres paall se wAclat or ‘The sti 3-3 v2 |_EA 3 a = FAL T 8V nn P| Reaction force R can be eliminated, jy EA P= (12+ 6V3) G7 the displacements of joint 1 are 4 PL_4(2-V3) PL uy =v, — oe © 6+3V3 EA 3. EA and the reaction force normal to the slope is obtained from either of the two stiffness equations: : pa 2829) p The axial forces in the bars are me 3 \feze} Gv3=5)P ension) 21 3 1.2 = 2Jle-o 5, <23EA [2 ‘| {3 - ait _ (9 = S3)P (compression) re 2h 2 2) le. ey (compression) [ 3 a /S)F (compression) , 73 3102 Truss Bar Elements, hap 4 The reader should check these axial forces to see whether or not they satisfy equi bri at joint 1. vm Example 4.6 Truss bar with a temperature rise A three-bar truss is shown in Fig, 4.12. All bars have the same axial stiffness 4, | the temperature of bar 1-2 is increased by an amount of AT above a certain speci uniform temperature of the truss, find the displacements of joint 1 and the axial fo, in the three bars. fied re, Figure 4.12 Three-bar truss with bar 1-2 subjected to a temperature rise AT. Solution. This problem can be solved according to the two steps described in Sectin, 4.6.4 Step 1: Assuming that joint 1 is restrained, the axial forces developed in the three bars due to the temperature rise AT in bar 1-2 are, from Eq. 4.47b, 13 = Sia =O Siq=-EAa AT where the primes are used to indicate that the forces are for step 1 and the negative sign indicates that bar 1-2 is in compression when joint 1 is restrained. The nodal forces required to restrain nodal point 1 are, from Eq. (4.48), EAa AT v2 EAa AT X= 2(—cos 135° 450 ¥, = Sj.2(-sin 135°) If there is more than one bar with temperature rise, the restraining forces must summed up at each common joint. Step 2: Based on Table 4.5 and Eq. (4.17), the total stiffness equations assembled as EAa AT earl" Fale realsconoluding Remarks = 0 ” 5 Direction Cosines and Sines for Each Element TABLE “ES a * “ MH Length = £ -1 1 ; ) “d 135° 2 2 2 aye 1 a 90° t i ! 1 14 se 2 2 3 vat ar sternal loads are equal in magnitude but opposite in sign to the restraining pete the exEETT Eq. (4.57). The final solution for displacements at joint 1 is thus fovoes as BH otic’ BAa AT ‘The axial forces due to these displacements are a EA 1 1 Wa u | if Al-s +f = "| = EAa AT he Sa Bil Va Vallu-»J~ 2 sts-=4 10 fers} (/2 - 1)EAw AT 7 3-2 st, = FAIL a\{ecet iv “ ale V2d los- 1 where the double primes are used to indicate that the forces are for step 2. ‘The final axial forces in the three bars are obtained by superimposing the forces obtained in both steps, —v2 7 5) Sia=Sigt+St.=—2 ate EAa AT (compression) Sis = Si.g+ Si =(V2—1)EAa@ AT (tension) a Sig = Stat Sis = = v2 EAa AT (compression) Ps i . + a st een axial forces satisfy equilibrium at joint 1 while joint 1 is subjected to no ‘al loads. 48 CONCLUDING REMARKS Inthi 7 tion, "chapter we have demonstrated the basic procedures for element formula- emblage, and solution of the finite element stiffness method by the Dr104 Tru Bar Elements Chap, ‘ use of truss bar elements and various kinds of examples. The sign convention, 8, numbering s stem, assemblage steps, and methods for solving simultaneo,, equations are introduced in a consistent and systematic fashion which is idea)), suited for computer programming. ¥ of solution introduced in this chapter are summai ‘The advantages of the truss bar element formulations and the method, ed as follows: 1. The orientation and length of each bar element can be arbitrary (i-e., the geometry of the truss structure can be arbitrary). 2. The boundary conditions can be arbitrary. For example, a joint can be allowed to displace along a locus such as that demonstrated in Example 4.5. 3. The loading conditions can be arbitrary. The given conditions can also be a mixture of loadings and displacements or merely displacements a any joint 4. Indeterminacy or redundancy due to extra bars presents no difficulty in the assemblage. Thus the indeterminate structures can be solved just as straightforwardly as the determinate ones with little or no additional effort. 5. More fixed boundary conditions mean fewer degrees of freedom and consequently, less computational effort. This is contrary to the common traditional method, such as the method of least work, where more fixed boundary conditions mean higher degrees of indeterminacy and more complication in the solution of the problem. 6. The special class of problem of truss with misfit or temperature gradient can be solved as straightforwardly as the ordinary cases. For the fundamental development of the methods described in this chapter, the reader is referred to Ref. 4.1. For the method of superposition that treats the thermal stress problems, the reader is referred to any good text on elasticity (for example, Ref. 4.3). A Fortran program for the static analysis of plane trusses, a user’s manual, and the input and output data for an example of a five-bar truss are given it Section 13.1. 4.1, Turner, M. J., Clough, R. W., Martin, H. C., and Topp, L. J., 4.2. 4.3. REFERENCES stiffness at Deflection Analysis of Complex Structures,” Journal of Aeronautical Science, V 23, No. 9, Sept. 1956, pp. 805-823. Martin, H. C., Introduction to Matrix Methods of Structural Analysis, McGraw-Hil Book Company, New York, 1966, Chap. 3. 4 Timoshenko, §. P., and Goodier, J. N., Theory of Elasticity, 2nd ed., McGraw" Book Company, New York, 1951, Chap. 14. vyproblems. 105 chap. 4 PROBLEMS ‘asymmetrical truss is shown in Fig, P4.1, Each member has the same FA. Find 4a. o> Gisplacements of joint 1 and all the member axial forces. Also check the equilibrium of joint 1. Figure P4.1 42. A symmetrical steel truss is shown in Fig. P4.2. The modulus of elasticity for steel is E = 30 10° psi. The cross-sectional areas are Aj.2 = Azs= 2.0in.?, A, 3 = Ass = 1.2in.?, and A,., = 0.8in.?. Find the displacements at joints 2 and 3, all the member axial forces, and the reaction forces at joint 1 or 4. Also check the equilibrium of joints 2 and 3. P= 100 kips 20 in. 20 in. 60 ii Figure P4.2 43. A three-bar aluminum truss is shown in Fig. P4.3. The modulus of elasticity for aluminum is E = 10 x 10° psi. The cross-sectional areas are Aj.> = 0.36in.”, A.» = 0.4 in2, and A,., = 0.6 in2. Find (a) the total assembled stiffness equations, () all the nodal displacements, (c) the axial stress (in psi) in each bar, (d) the Feaction forces at joints 1 and 2, and (e) the equilibrium at joint 3. Nore: A computer or at least a calculator is needed for inverting the 3 x 3 stiffness matrix.y 108 Truss Bar Elements — Chop , 10 Kips al Figure P4.3 4.4. A five-bar statically indeterminate truss is shown in Fig. P4.4. Each bar has the same EA. Find nodal displacements, member forces, and reaction forces. Also check equilibrium at every joint. The reduction procedure is suggested. The symmetrical truss as shown in Fig. P4.5 is loaded in an antisymmetrical way Each bar has the same EA. The stifiness matrix can be reduced from 4 x4 10 2x2. Find the nodal displacements, member forces, and the reaction force: Also check the equilibrium of every joint. Figure P4.4 Figure P4.5 4.6. A symmetrical truss is shown in Fig. P&S. Each bar has the same EA. 1 stiffness-matrix-can-be reduced-to-3-X3-by making-use_of symmetry. Find nodal displacements, member forces, and the reaction forces. Also check equilibrium at every joint. 4.7. A symmetrical truss is shown in Fig. P4.7. Each bar has the same EA ™ problem is similar to Problem 4.6 with the addition of four bars but m0 degrees of freedom. Find the nodal displacements, member forces, and ret" forces, Also check the equilibrium of every joint.,oplern® 107 Figure P4.7 48. The truss as shown in Fig. P4.8 has eight degrees of freedom which can be reduced to four by using symmetrical conditions. The four equations can be solved by using reduction procedure. Each bar has the same EA. Find the nodal displace ments, member forces, and reaction forces. Also check the equilibrium of every [ © z - | bis ‘a Z 7,2) io | Figure P4.8 (30 Substituting the solution for a’s in Eq. (5.5) into Eq. (5.2) gives 2 3x? 2x? Ba x v(x) = v, + x0, Fra Batten 7% ax x x type +7 pet a After rearrane: : : ¢ ‘anging, we obtain the final form: ? (5.6) ql B(x) = fix) vy + fal) Os + Aa) 02 + ACO2 Beam and Plane Frame Elemente cy, 05 where (32) % >» wen=a(#) =2(3)) The functions obtained in Eq. (5.7) are called shape functions. If we vary the value of x from 0 to L, we obtain four curves as shown in Fig. 5.2 for the four shape functions, respectively. Physically, each of the four shape functions represents the deflection curve for the beam element produced by setting the corresponding degre: of freedom to be one and the rest degrees of freedom to be zero. In other ae AG) fo A- a =ariin ai | os Be et Figure 5.2 Plots of the four shape ee tions associated with v;, 4, v2, and & aL a beam clement,orm Straight Beam Element uni "3 sett ord OY v=! and 0,=v,=0,=0 o,=1 and v=, = 6,=0 v,=1 and v,=0,=0,=0 6,=1 and v,=6,=0v,=0 (x), Al), A>), and fa(x), respectively. iain fy t the deflection 1 js seen that : y ‘urve represented by the displa ic & %6) is obtained as the linear superposition of the cue funion degrees of freedom. wo ‘the stiffness equations for this beam element can be obtained by usin cpigliano’s theorem, introduced in Chapter 3: . au aq where F; is the nodal force or moment, q, the nodal displacement or rotation degree of freedom, subscript i the degree of freedom number, and U the strain (5.8) energy. The strain energy for a beam element with uniform cross section has been given in Chapter 3 as EI (* (a°v\* u==) (S$ 2 J, (2) dx (5.9) Equation (5.9) indicates that we need the expression 2 Fea SU 0, + F588, + L802 + FU) Os where x . 4 F(x) = —7 + 675 (5.10) 5 6 x AQ) =~ 2FS : 2 ee I(x) = —T + 67v 14 Beam and Plane Frame Elements cy, ubstituting the displacement function (5.6) into the strain ener, expression (5.9) and then performing partial differentiation of the enc”! expression with respect to the degree of freedom v, as indicated in Castigliang, theorem (5.8) gives | aU_ EL y= Bt" a, 2 Jo Javi\a iL El f LAT Ce) 01 + F502) 8, + L306) v2 + SAX) OST (x) dx ° Kyo) + ky28) + ky302 + ky482 (5.11) where L L Ky, = EI i FICOFTR) dx ky = erf FiC)GE(X) ax ° ° a L (5.12) ka = er| SFICOLS) dx Ka = er f Fie FC) ax ° ° Equations (5.12) can be generalized to the standard form L —T = er f FUCYUG (x) dx Gu} which gives the stiffness matrix for the beam element. As an example, le i= j= 1 and we find that 2 k= er | (-£ preg 3) dx 72x? | 48x*\* = F(s6x ate *) EI =125 By using the same procedure, we obtain the stifiness equations, 12 6 12 6 ‘9 PL oP © \\* 6 6 M|ielrt * ~E 2 |\ si y. BN S12 812 si 2 EL Bor i|| 6 6 M, 7 -~ 4 | [+» unitorm Straight Beam Element 8! 00 ymboticallYs : = fea) (5.14a) h s with high numbers of degrees of freedom such isin the case of plates and shells, the solution for shape functions may become fambersome. It is customary to use existing mathematical polynomial func- ca (ez. Hermitian polynomials) as the shape functions, However, for those displacement functions whose shape functions are hard to find, we can simply sperate on the constants (called the generalized coordinates) instead of the shape functions. Such a procedure is demonstrated by using the present beam element as an example. We first rewrite the expression 4*v/ax" based on the original form of the displacement function (5.2), 3 ax? = 2a; + 6ayx (5.15) Substituting Eq. (5.15) into the strain energy expression (5.9) gives _Eift 5 =F J, Gas + 6aux)? ae = 2EI(a3L + 3a3a,L? + 3aiL*) 00 0 0 F(a 1 00 0 0 [ja = la, a, 2 21a: a2 4 al) 9 gem 6EIL?|\ a, 0 0 GEIL? 12EIL*) la, = sla] [ka} (5.16) The elements j je ation produ [k] are obtained by using the following partial differentt- ot 2 yas p= 1.2,34 (5.17) Substituti da, da, ing Eq, (5.5a) for the constants into Eq. (5.16), we obtain (5.18) U =S{q\(T)' kT Kat 116 |+8 Beam and Plane Frame Elements cy, can write the strain energy expression as one-haip 4 Alternatively, we and the corresponding nog the sum of the products of the nodal forces "i degrees of freedom: YY, U=Sly 0 v2 9) ay Slo 0, 2 |% M2 = HlaliF (5.19) Substituting the stiffness equations (5.14a) for the nodal forces, we have U = 31a) UkKa} (5.20) Although in separate forms, the two strain energy expressions in Eqs (5.18) and (5.20) should be the same. Thus we conclude that Ok] = (1) (KIT) Go -3L 2};/0 0 0 0 Lv 0 o 0 _fIjo Le -2r* L||0 0 0 0 o FB ° 0 T\o 0 3£ -2}]0 0 4L 6£?|\-3L -2L7 3L -L? o 0 -2? Lilo o 6? 12L*}L 2 L =2. 2 20 Satae 2 RS) PoL PL. 6 6 _EI| Lb ia aan = is 6 (5.21) L L S 2 a 4 which is the same as that obtained in Eq. (5.14) by using the shape functions. 5.1.3 Properties of the Beam Element Stiffness Equations Equilibrium. If we consider the beam element as shown in Fig. 5.1 ® a free body, it must be in equilibrium under the four nodal forces which at related to the four nodal degrees of freedom by the stiffness matrix in Eq. (5.14) It is seen in the stiffness matrix [k] that the first row and the third are exactly identical but with opposite signs. This means that forces Y: 2 Y, are equal in magnitude but opposite in direction. So the equilibrium " the transverse direction of the beam is satisfied.52 Beam Element Oriented Arbitrarily soc Ww We then lake moment about nodal point 2 ~M,= Y,\L—-M,-—M, = (first row)(L) — (second row) — (fourth row) E[/12 6 6 ToT gat -4~290, 12,66 ty +(-Z+p+f)a+6-2-00] ay =0 {\ which again satisfies moment equilibrium condition. Singular matrix. Because the first and third rows in the stiffness matrix [k] are only different by a sign, [k] is a singular matrix. There is no inverse for matrix [k] and consequently there is no solution. The fact that the element stiffness matrix is singular can be interpreted i physically. The element, without support, is an unstable free body. It will be stable and the stiffmess matrix will no longer be singular if it is properly +t supported. For example, if the element is clamped at nodal point (1) (i.e., | », = 6, = 0), there are only two degrees of freedom, v, and @,, left. The stiffness equations then reduce to plage ko a MJ LI} 6 4 {le L where the stiffness matrix is no longer singular for the supported stable element. From the reasons explained here, we know that the stifiness matrix for a finite element is always singular. nd 5.2 BEAM ELEMENT ORIENTED ARBITRARILY IN TWO-DIMENSIONAL PLANE (PLANE FRAME ELEMENT) the beam elements are no longer In the case of pl ame structures, ee os o-dimensional plane. Such horizontal. They ¢ itrarily oriented in a tw hm y can be arbitrarily orien a 0 lements are subjected to axial force, shear force, and bending moment To “ope with the plane frame problem, ‘an element must possess three degrees : Xn se a i h an element 1 " “ tively, and an angle of rotation 0. Su , 5 Corresponding to the degrees of freedom % % % there is shown in Figns Beam and Pi Frame Elements cy, Ds Yas Up ales, ty Yo Figure 5.3. Arbitrarily oriented plane frame clement, are two forces X, Y, and a bending moment M, respectively, at each nodal point. This element is oriented arbitrarily at a counterclockwise angle ¢ with the global x axis, It has modulus of elasticity E, cross-sectional area A, moment of inertia I, and length L. To derive the 6 x6 stiffness matrix for this element, we first combine the 2 x 2 stiffness matrix for the truss bar element and the 4 x 4 stiffness matrix for the beam element to a 6 x 6 matrix. We then transform this 6 x 6 matrix from the local coordinates to the global coordinates (general reference coordinates). 5.2.1 Axial-Flexural Beam Element long the Local X When a truss bar element and a beam element are combined, we obtain the axial-flexural beam element shown in Fig. 5.4. The stifiness Figure 5.4 Axial-flexural beam clement oriented along the x axis.ne (5.22) This stiffness matrix is composed of four submatrices among which two are zero submatrices. The two nonzero submatrices are along the main diagonal. One is associated with the axial behavior and the other is associated with the nog flexural behavior. Such an arrangement of the submatrices indicates that the bi axial stiffness submatrix and the flexural stiffness submatrix are uncoupled. one, In other words, the solution for the axial displacements and the solution for the transverse deflections and rotations can be carried out separately and independently. When this element is oriented in the two-dimensional plane with an angle ¢ with the x axis, the stiffness matrix in Eq. (5.22) must undergo a coordinate transformation procedure. In that case, the axial and flexural submatrices are no longer uncoupled. For the convenience of assemblage, it is desirable to number all the degrees of freedom at each nodal point in a certain sequence. For this purpose, we rearrange Eq. (5.22) as pew ef, ts] my th y, % M, im = (5.23) XX uw I y, bv M, 6 )120 Beam and Plane Frame Eloments chy. 5 or symbolically, 4By=(eMay any |? where the bars indicate that the formulation is with reference to a set of loca coordinates. 5.2.2 Coordinate Transformation An axial-flexural beam element with its longitudinal axis lying along the # axis is shown in Fig. 5.5. The axes (X, 9) are the local coordinate axes ang (x, y) are the global or reference coordinate axes. The X axis is oriented at an angle ¢ measured counterclockwise from the x axis. To distinguish betweey, the two coordinate systems, bars are added to all the symbols for the noda| forces and degrees of freedom that are with reference to the local coordinates Xi, ‘Le My. 6, OMN- , on | Yu Figure 5.5 Transformation of an axial-flexural beam element from local (% 3) © global (x, y) coordinates.ye Oh Beam Element Oriented Arbitrarily am MY goc. 62 the six nodal forces and bending moments in a local coordinate system Sy ae transformed into those in a global coordinate system: x, ty xy lad 7, ML (5.24) Xx ¥2 ngy M x We, here A and 42 are direction cosines and sines defined as ety k A=cosd ni ing . (5.25) "i wasing Equation (5.24) can be written in symbolic form as {F} = [THF} (5.26) This transformation matrix was derived in Chapter 2. It has been proven to be an orthogonal matrix, that is, cr! =r" (5.27) Equation (5.26) can be written in inverse form, {F} = (1) '{F} = (117 {F} (5.28) ___ Inthe meantime, we can also transform the six nodal degrees of freedom in the local coordinate system into those in the global coordinate system: {a} =[Tl{a} (5.29) where the transformation matrix [T] is the same as that for the nodal forces and bending moments as obtained in Eq. (5.24). fay ig SUPstituting Eq, (5.26) for the vector {F) and Eq, (5.29) for the vector a} into the stiffness Eq, (5.23a) in a local coordinate system, we obtain (THF} = [kITKa} (5.30) Making use of the orthogonal property of matrix [T], we obtain The prog 7 {F) =(T)"(RUT Ha} (5.31) (52,)/0duct [EYTRIET] is called a congruent transformation. Substituting Eq. or matrix [k] and Eq. (5.24) for matrix [T] into Eq. (5.31), we finally obtain fain the gi 5 - i a Systeme stiffness matrix for a plane frame element in global coordinate Er122 Boam and Plane Frame Elements cy, ap 5 5.2.3 Stiffne: Equations The stiffness equations for a plane frame element oriented ay , ie pees : angle & measured counterclockwise from the x axis are of the following fo," RAP + a x pe yu symmetric (5.32) where R is defined as the ratio between the cross-sectional area and the moment of inertia (A/J). 5.3 APPLICATION OF BEAM ELEMENTS The method of assemblage and procedure of solution for using beam elements are introduced with the aid of illustrative examples. Beam Example 1 General procedure. Using two beam elements to mode! the beam structure as shown in Fig. 5.6a, find the deflection shape, the reaction forcss and moments, and the shear force and bending moment diagrams. We first idealize the beam by using two beam elements as shown in Fig. 5b. The total system has three nodal points and six degrees of freedom. To formulate fo! the total system, we must formulate the individual elements first. For element 1-2, we have symmetric zh 4 a (sa 6 12 1B & 6 2 woe 4] [ae53 123 ‘ D nh Qn er @ 7 a «i, me i P29, Psa P85 Figure 5.6 Beam and two-element Pid Puy Tt modeling. (b) For element 2-3, we have dor Ya symmetric b M2 EL 4 6, ¥ Tl o3 3 (5.34) ° “Ee vc % emt 3 3 Ms i 2 -7 4] [& It must be borne in mind that the ¥’s and M’svare the internal forces and 1m moments. When the two sets of element stiffness equations are assembled, the sum of ale) ‘udhiatgmnabésteononmomentsakeantnodalin loads P's at the same nodal point as shown in Fig. 5.6b. Thus slat R=, P,=M, P; = Y; of element 1-2 + Y; of element 2-3 an P, = M; of element 1-2 + M; of element 2-3 | Ps= ¥3 P= My t each nodal Point Ttmust also be borne in mind that the deflections » and rotations @ al are still the same as those for the assembled system. on the understanding of these two ground rules, the method of lage is simply to sum up the Y's and M's as indicated in Eq. (5.35).frow, — Omdried y Ce 4 Coburer Vws 124 Beam and Plane Frame Elements Shap 5 ‘The equations for Y's and M's are given in Eqs. (5.33) and (5.34), PD vy | 2 P, 4 » & z s (525, Po % Pe a] | o ‘The dashed lines bracket the two element stiffness matrices. The overlapped poric, contains the stiffness terms that resulted from the superposition of the terms in the ty individual element stiffness matrices. We have three boundary support conditions: v= 0, = 0 =0 and the following external loading conditions: P, =unknown reaction force at point 1 P;=unknown reaction moment at point I P,=—P (opposite to the positive y-direction) (538 Pa= PL S P,=unknown reaction force at point 3 Corded Ps Following the boundary conditions (5.37), we rearrange Eq. (5.36) such thats, 61, vs (or Py, P2, Ps) are together. We first rearrange the sequence of the rows: Py 2 _§ 6 -+3-_% 2/(% ve Lv L veo Ps a+? 2 3 8 -; 2 |\a Ps a 0 0 ; 2 -3 4 || op ie af t|2 £ 2 £ 5 o|le P $ 4 -= 2 0 0 fies oe} Lp oo Bz a allemat 53 Application of Beam Elements “T $06: 126 sen rearrange the sequence of the columns | we v (5.40) mS Romie ig ety) Now we have six equations that can solve for six unknowns, three on each side of the equations. Multiplying out Eq. (5.40) gives 5 -P PLY == é (5.41) F 0 (sa) 4 t oefe PI le 3a nv Fs = and atsl PB, Nn RI® 0 0, (5.42) Itis noted that (5.41) can be obtained by simply crossing out the first, second, and fifth rows as well as columns in the system stiffness matrix in Eq. (5.36). Thus we do not have to go through the tedious process of rearranging rows and columns to obtain such reduced equations. In conventional structural theory, boundary constraints increase the number gi dcarees of redundancy and consequently complicate the problem by increasing the number of equations. In the finite element method, boundary constraints reduce the numb er Of degrees of freedom and consequently simplify the problem by decreasing © number of equations.126 Beam and Plane Frame Elements cy, aD 5 Equation (5.41) may be solved by the method of matrix inverse The Aadjoj method given in Sec. 2.13.1 is recommended. i 18 30 [ Dd, 28 aoe P a 18 51 39 a,b = = 25-35] dee 276EI | L v L 6 0 39 7 L Po PL 276EI (345 Equations (5.37) and (5.43) provide the solution for all six degrees of freedom. 7), corresponding deflection curve for the beam is shown in Fig. 5.7. At this point, if y, want to know the deflection at a certain point, we can simply substitute the coord value of the point and the nodal degree of freedom values into the displaceme: function (5.6) of the element. Inflectio: point Shear Figure 5.7 Deflection curve, shear diagram, and moment diagram for the beam in Fig. 5.6.ation of Bear Elements Applic! wp godt ws al forces and bendin, C1 internal shear & Moments are obtained by iin ne wmities of nodal point degrees of freedom into ach ince tating the sno mat Xs sols ae iment 1-2, we have, from Eq. (5.33), ror ele _2 6 Yi my, M, =Fie2 '|_B ==!) not needed ct L 6 (5.44) 2 ay (Sq 6 M2 “L orelement 2-3, we have, from Eq. (5.34), ¥ 3 3 3 37) (-10PL 1P = eS LL || 276Er 46 3 3 33 PL? TPL M; 5S 4 -5 2]{ ae 2 _EI| Lb L 276EL 23 (5.45) y, Fra| pees ES ie ee. o | |-2 ; P EO L 46 3 3 —9PL? M; > 2 -> 4 L L 276ET H ‘The sign conventions of the Y"s and M’s are defined in Fig. 5.1. With the results obtained in Eqs. (5.44) and (5.45), we can plot the shear force and bending moment diagrams in Fig, 5.7. It is seen that the amount of drop in shear force at nodal point 2is caused by the external load P and it is equal to P. The amount of drop in bending moment at point 2 is equal to the external moment PL, The point of zero moment is the point of inflection which is customarily termed as hinge by structural engineers. The external reaction forces and bending moments at the supports can be obtained from Eq, (5.42): 12 6 53P f P ars =10 46 EI] 6 3 | PL? _ )21PL P t\-n 2? 2 L [{ 276EI 46 R 3 3 _3}|_9 ere | , “poe rite as respectively, in Eqs. Which, of course, agree with those obtained for Yi, Mi, and Ys. Sere eaten oF eas) and (5.45). The use of Eq. (5.42) is actually not necessary. 1 tifiness equations for a system, itBeam and Plane Frame Elements gy, hap. Beam Example 2 Method of matrix partitioning and reduction. ,, three elements to model the beam structure as shown in Fig. 5.8, find the dena? shape, the shear force and bending moment diagrams, and the reaction force, ,." moment. Dead Fro Y* fe i oo ae gos} C 8, P92 PaO eeeiik | i Figure 58 Beam and three-element modeling. For this problem, the three-element stiffness matrices are the same. The total stiffness equations are assembled as 12 1 P zB 1 symmetric y—9 2 \ 6 ' Ps 3 i ab = Py mee A ' va = 1 v2 | 6 | wen | Paar ' 6 lai (546) Ps o ot Ps a 6; es Fs ee 5 4| e wewei Elements ‘Application of Beam ni goss 129 loading conditions are dary and pout ( %1= 6, = (5.47) total stiffness matrix may be reduced from 8 x8 to 44 by crossing out the rows hy golumas corresponding to the zero degrees of freedom. a 0) {4% 2 |]6, ~6l) | (5.48) -F || 4 || oe It is recommended to assemble the element stifiness equations and take into account the zero degrees of freedom simultaneously [i.e., to form Eq. (5.48) directly without forming Eq. (5.46)]. Such a step is crucial in computer programming because the size of DIMENSION can be reduced and consequently core storage can be saved. Equation (5.48) can be solved by inverting the 4X4 stiffness matrix. Instead, we can use a partition scheme by taking advantage of the zero-loading conditions so that we only have to invert two matrices of smaller sizes. Such a reduction procedure may result in easier hand computation or less computer time. To illustrate this procedure, wwe first write the partitioned matrix equations (5.48) in symbolic form: (5.49) Multiplying out Eq. (5.49) yields (0} = 1K, HQ} +(K KO.) (5.50) {P2} = [2 KQ)}+1K221Q2} (5.51) Equation (5.50) can be rewritten as a (Q,}= [Kn '062H2} (5.52) oe at_we should invert [K,,] instead of [K,2]- Submatrix [K,2] is Jalltgonal and it may well be a singular or non-square matrix. Substituting Ea. (5.52) 0 Eq. (5.51) gives [P2] 2.) =[K{Q2} a teduced matrix. (5.53)130 Beam ond Plane Frame Elements It is seen that in the reduction method, we have only {0 invert the guy, UK.) and reduced matrix [K] instead of the full matrix [K]. Substituting the syn defined in Eq. (5.48) into Eq. (5.53), we obtain "se oa 2) 6 EN TL 6 cle: = a4 (54 From Eqs. (5.53) and (5.54), 130 41 onal) epeeupy 2] 21 een oa={*} <1 Pa=F i 9 [lo 1a Ee pe {haze 21 J J 7 cer a1 (S55) 14 ‘The rest of the degrees of freedom are obtained by substituting Eq. (5.55) into Eq. (550) ee {e} = SL 15 {et Os, « es Oa, SE eis) pln PL) i4 EIT) 2 o 7 ‘Thus we can construct the deflection curve shown in The elements ~9 ¥, P —P)-2L M;| _\PL 7) 9 Y,{ )-P‘AI 131 inne r a ‘Moment Figure 5.9 Deflection curve, shear diagram, and moment diagram for the beam in Fig. 5.8. ‘The resulting shear and moment diagrams are plotted in Fig. 5.9. These diagrams Provided the reaction forces and moments: eo upward 7 areal P= M,== counterclockwise iP ee daiayard 1 Ee ea prard ple 3. Treatment of symmetrical and antisymmetrical loading }conditions. Using three elements to model the beam structure as 10, find the deflection shape, the shear force and bending moment n forces and moments. m, the total stiffness matrix equations are precisely the same as 6). Because of the antisymmetrical type of loads, the boundary132 Beam and Plane Frame Elements Chap. ie a i. bot Py, 02 Po, 9 F504 Pivy Pav i vs Pry Figure 5.10 Antisymmetrically loaded beam and three-element modeling. and loading conditions can be specified as Y= 0, = 0, = 0,=0 R2= 03 Le.= 0, We first remove the four rows and columns that are related to the four zero degrees of freedom given in Eq. (5.57): 6 L e 2 0, 0 | Ju=—v. 8 | | =6 Replacing v; by —v2, 0; by 62, and multiplying out Eq. (5.58) yields = 36 ee BP £ 6 0 a _etl| 10 | (vy wea |eass 6 i ere is 6 = 10on of Beam Elements 133 Applic reat the antisymmetrical conditions (v= ~v,) and (0, 0,) is simply ‘ne pe (hurd column from the first and add the fourth column to the second ae in Eq. (5-58). In the resulting four equations (5.59), we see that the first rpemats Igns are parallel (0 the second set of two equations, We can choose aol ee For the first sets & o oI 5 — es SS 10 Ce} » fel _PI? [-s) 0; 0.5 162EI| 3 G20) rom this example we can generalize the method of reduction of stiffness matrix m igan relate two degrees of freedom. For example, if we know blo re solution is whenever a1 =(—6) qo the method of reduction is simply to The internal shear forces and bending moments are found by substituting the foregoing deflections and rotations into the clement stiffness equations. We find for clements 1-2, 2-3, and 3-4, respectively, % 13 Y: -14 ys 13 M,|_P) 6L M,{_ P}-7L M,|_P) 72 ¥2(— 27-13 ¥;( 27) 14 ¥,{— 27 )-13 M, 1L Ms 9 Mz 6L ‘The deflection curve, shear force, and bending moment diagrams are plotted in Fig 5.11. The reactions, forces, and moments are given by Yi, Mi, Ya and Ms. _Beam Example 4 Alternative solution for Beam Example 3. For symmetrical or antisymmetrical problems, it is a common practice to analyze only one Portion of the structure cut out from the lines of symmetry oF antisymmetry. For Beam Re on 3, we can model only half of the beam by using two elements as shown in The bound: fi 5 ; conditions for th t antisymmetrical case are lary conditions for the present sy’ Sn Eetastiaell Ure Ae pera By i 0, = 0,=05=0 , ( 742 from esis such boundary conditions, we can reduce the total stiffness equations KGa 3o<3, Thus we have only to assemble a 3 X3 stiffness matrix by colleche iiee terrae ia the two element etifiness matrices instead of forming the largerCreberres 5 Fee oyrety or Rc Chas Inflection point oy Shear Figure 5.11 Deflection curve, shear diagram, and moment diagram for the beam in Fig. 5.10. Ps0y toh De a modeling for f of the beam shown in Fig. 5.10Bere spiarrsnioniotiesmisiemente a 6 x6 stiffness matrix. ze a put unnecessary Sa Py 3 * oO eeree2s |) iG CLD QD 2 6 4 4.8 4 Pym et|-S4 4,8 4 7 v LL ical 24 4 8 5 we ae salle! or = 08 esas Fe ie ||| see ea OF amas Nl a 2 4/46 24 0 5 4 844s P 80 ca ees Sains nt ~ 3592 ET needed 216 05 eats) L yi 13, Y2 —14 M,(_P) 6 oe M,|_ P)-7L ¥2(— 27 )-13 Y,(~ 27) 14 Ma 2D My 0 _ Treatment of elastic supports. Let us consider the same mple 1 with the addition of two elastic springs: an extensional @ (Ib/in.) and a rotational spring with constant 6 (in-Ib/rad).136 me Elements — ¢y, ie i 5 fates AN O(a) > E 7 P28) PaO Pon Pury Paiva P. Figure 5.13 Beam with elastic supports and two-element modeling. z The effect. of the two elastic springs is to produce a resistant force «j | av, and a resistant moment —f0, in addition to the existing external load: Such elastic force and moment always have a negative sign which means thy they are acting in the directions opposite (resistant) to the positive direction of v and 8, respectively Without considering the springs, the stiffness matrix for the beam system is iv 7 in Eq. (5.41). Taking the spring effect into account, we add the spring force and mome:| 7 to the existing loads in Eq. (5.41): is 33 —— GL L\\" e1| 3 PE et | 23 sat c\-n ® 2 |\e ( a 3 ; 0-60 2 2 alle 1 a ia 6 Replacing the spring force and moment to the right-hand side of the equations (§* results in we v2 3EI ; PLy=| — ee LC 6 3EI 0 L on e des It is seen that in order to include the effect of a spring associated with th ic al of freedom i, we simply have to add the spring constant to the stiffness term ki& ‘of Beam Elements pplication i SE ‘ 1 r the stiffness matrix. We see in Eq. (5.62) tha | of the st 4. (5.62) that the stiffness t viol ont gonal have the same physical meaning as the elastic spring conan ge ye main positive. eat av Eq. (5.61) into Eq. (5.63), we can solve for the unknowns hed 2 pti 2 si! v 12 _£ L _ & Wl 92° = 240ET| LE ~p |\Pe 6 35 os 7 Silo -6L PL? { =} 240ET | se internal shear forces and bending moments are obtained by substituting the jal isplacements and rotations into the stiflness equations for elements 1-2 and 10 4.3, respectively. yi 123 Yo 27 iad m(_p) aml 4 JMal__p jaar ine %[ 120-123 ¥3(~ 120)-27 3 the ti M; T6L, M, 19z: ‘give | the resulting deflection curve and shear and moment diagrams are shown in Fig. 5.14. omit.) Tae force at point 2 and the bending moment at point 3 due to the springs are _ a A ward ees \aonr) 5 UP sEY Pl pe, =-==| PL : (ssl z ) == clockwise ~48EI)~ 12 ‘which agree with the amount of jump at point 2 in the shear diagram and the amount otjump at point 3 in the moment diagram. an Example 6 Beam with prescribed displacements. _In engineering Ocmameagtete are cases where beams are given initial deflections and rotations due sre peanges such as thermal expansion or misfit. For such type of problems, we quantities: i the total stiffness matrix into four submatrices based on two groups sal) ana ie ¢ known degrees of freedom (or the corresponding unknown seastige "| oeedamy te Known forces and moments (or the corresponding unknown degrees Al é . though ge imple is shown in Fig. 5.15. Let it be assumed that a smalll hole is punched “deegign ™4depth point of cross-section 2. This point is then pushed upward by edt] (Medal aa A and held by a momentless pin. It is desired to find (a) the slope 1% Wthe syesi™t 2: (b) the force required to produce the deflection A at point 2, an “a and moment diagrams. bh138 plus the special condition that By imposing the boundary conditions, we can reduce the total stiffness matrix is 6 x6 to 2x2. Thus we only have to assemble a 2X2 stiffness matrix by collectin? pertinent terms in the two-element stiffness matrices instead of assembling the te u Spans | Aivranw, Wy dnd ¢ Cau, toad ere raed) Fara Bearn and Plane Frome Elornants, a 0, fu Inflection point Moment Fig. 5.14 Deficction curve, shear diagram, and moment diagram for the beam in Fig. 5.13. Two elements are used to idealize the beam. The boundary conditions are 2) = 6, =v3= O3= =A Figure 5.15 Beam with initial deflection at midspan.cation of Boar Elomonts tn Apoll 83 see easary 6X6 stiffness matrix eo wen i ie {Pt EN Le Le { Py L| 6 12 tf + h uit 48 ot : a 6 6 fect # PoL fap P=0J L)6 |, te (5.64) L tiplying out Eq. (5.64) yields aul 36ET. 6EI Peery Bore (5.65) EI, | 1261 patty & (5.66) Equation (5.66) gives A 2L which is the slope (clockwise) at point 2. Substituting the solution for 9 into Eq. (5.65) gives 33EIM t a LD te | Which is the force required to produce an upward deflection A at point 2 The internal shear forces and bending moments are obtained by substituting the nodal displacements and rotations into the two element stiffness matrices, respectively YN =15 Y2 18 M,|_EId)-7L M,| _ Er) 8b y,(~ DB) 1s and y,(— Bb al M, —8L, Ms, 10L. The deflection curve and the shear and moment diagrams are shown in Fig. 5.16 The negative slope at point 2 shows that element 1-2 is bent with more curvature than foment 2.3. This agrees with the physical reality that the former element has less nding rigidity than the latter. The jump at point 2 in the shear diagram has a value * 33E14/L, which is the force required to hold point 2 at a distance 4 trom its rinal position | connge?2™ Example 7 Beam connected with truss bars. An example of a beam win ’°€4 to truss bars is shown in Fig, 5.17, The beam is assumed to be inextensible boty | Pending rigidity £1, ‘The truss bars are assumed to have hinge connections at ends. The axial rigidity EA of both truss bars is assume — fee d to be equal to 24E1/L?.Boam and Plane Frame Elements — ¢ 140 Shag lutlection point TT Figure 5.16 Deflection shape, shear diagram, and moment diagram for the beam in Fig. 5.15. P Figure 5.17 Beam connected to two truss bars,Gordes F pan the daniy ? oer B 4 ppiioation of Beam Elements hat strate the us first formulate the total stiffness equations for th " et woes © beam modeled by elements: di di yo elemposing the boundary conditions that we have {rh EL Pi) EL We then formulate the total stiffness equations for the truss bar elements, Imposing the boundary conditions and the inextensible beam condition that (5.67) us= =us= =m=0 we obtain, from Eq. (4.17), IEEE GeO COTE) on Superimposing Eqs. (5.67) and (5.68), we obtain the stiffness matrix for the total system, 108EI EA 18ET DP 2 LE \{m 18EI12ET lial i oe (5.69) where the sum of the unbalanced vertical force P5 at beam joint 2 and the unbalanced vertical force PS at truss joint 2 must be equal to the net external load Py, or ~P in this case. Equation (5.69) can be solved by setting EA=24EI/L*, roms ies a a (A - PL? {4 o,J 1116Er) 18 120\l oJ 186EIl 3 Lv The axial forces in the two truss bars are obtained by using Eq. (4.24), EA oe nif Me wal AP (S4=- Leos (150°) sin cisoni{—"} 31 on sunt (both in tension)Boar and Plane Frame Elements Nae Chay rhe internat shear forces and bending moments for the two beam clements arg Y, 7 Y, 20 ae) lat y,("31)-7 nA aul 24 hal | » 6L. curve, shear, and moment diagrams are shown j 5.18, The unbalanced force at point 2 duc to the load P and the vertical componrt Of the axial forces in two truss bars is equal to 27P/3\. This force is equal 1," mount of jump in the shear force diagram for the beam the | point NS ‘The resulting deflection Shear 20P 31 Figure 5.18 Deflection curv ‘ i jection curve, sh a iis. 0 eat diagram, and moment diagram for the beam 5.4 BEAMS UNDER DISTRIBUTED LOADS In the examples i i Eee a een Te the load: ints. In practical engineering applicati”” Peer ioch aes concentened: Fonsicnieassewa ara transform! i loads so that we the no points. Two common methods are suggested as ree ood Loads dor Distribute yns UN 143 00 i a E " work-Equivatent Loads at work-equivalent load method, we set the work produced by the inthe Wal concentrated loads to be equal to the work produced by the hs" Taputed load. This method is convenient when the distributed load ual ise ribed by a mathematical function. ape des an Oe sidering @ beam element, the work done by the equivalent but oa nodal loads is in the form 101 vk - a Oo ihe other hand, the work done by the distributed loads due to the deflection trite beam element can be obtained as W=321¥i MY, Mj (5.70) . ws] P(x) v(x) dx (5.71) eee here the deflection function is defined in Eq. (5.6) as vy @ v(x) = Lie) A(x) file) fax) . (5.6) 02 The work given in Eq. (5.70) should be equal to the work given in Eq. (571). Thus we have L BG j POX)fi(x) dx ° L M j P(x) f2(x) ax =) pt (5.72) Y2 J P(x)fs(x) dx ° L M, i) P(x) f(x) dx ° "in general form, ie ‘ F P(x) fi(x) dx (5.73) Where ; Work. denotes the degree-of-freedom number. We see in “Ativalent load associated with a certain degree of free! Eq. (5.73) that the .dom i is obtained106 Beam and Plane Frame Element, ff by integrating the product of the distributed load function and the ; 4, function over the element length. Such a definition can be generalized y, types of finite elements, such as plate and shell finite elements th Because such loads are found on the basis of functions that are ¢, with the assumed shape functions, they are often called consistent i, ads Beam Example 8 Beam under uniformly distributed load. fig, shows a beam with both ends simply supported and under uniformly distribycy P. Let it be desired to find the slope at the end (1) and deflection at the mids." Up ny Arias Aart Eek (erat Vows o\1 ee ; \ 4 > te ee) ® 4 o.+7\ a 1 Z io i—+| Figure 5.19 Beam under uniformly distributed load. Because ofsymmetrysswetcangmadeljonly half of the beam and use only o» beam element with length J/2. The boundary conditions for this beam element ate The stiffness equation for this case is 8 24 na {xt aie iE {e ca Ya 24 96 | lv EF oP The work-equivalent loads for the downward uniform distributed load —p can bi obtained by using Eq. (5.72): Equation (5.74) can be solved as GD Fed db vat ny Ctuk~ {a} = c ~~ 2467 |S! 16ted Loads 145 tion obtained using Casti lano’s theorem, demonstrated in 69 Cantilever beam under linearly vai spler beam under a linearly varyi gload. Fi 1g distributed load a 12 6 fy of Joy MJ” LI6 a 4 L (5.77) fal, ah notions used to find the work equivalent loads med as the exact solution to the differentmee’ — Boam and Plane Frame Elements 146 Chop This is why the answers we obtained by using only one elem, equation of beam (5.1). are exact. De 5.4.2. Statically Equivalent Loads In the statically equivalent load method, we have only to find 4p, concentrated loads that are in statical equilibrium with the distributed joaq, This is done by simply assigning the total distributed loads over one half the element to one nodal point and those over the other half of the clemen, to the other nodal point. Although this approximate method appears to 5, crude, it is eonvenient and versatile. It can cope with complex and arbitrari distributed loading conditions in an approximate sense. This method provide results with good accuracy when sufficient number of elements is used. Beam Example 10 Beam under uniformly distributed load. Let us consid; the same simply supported, uniformly loaded beam shown in Fig. 5.19. Because symmetry, we have to model only one half of the beam. First, let us model the left half of the beam by one element 1-2 with len; L= 1/2. The total load on the beam element is pL. We apply pL/2 to each of the ty nodal points. The stiffness equations are h 6 aa . | _ er i ie rT] 6 12|ley Ure, LL. UrED Hence uz & {at ee ey » a »,J T2EI} 6, 4 L£ Substituting L by 1/2 in Eq, (5.78) and multiplying out the two equations gives a {*} _ ph )32 6 % EI) 1 96 Compared with the exact solutions given in Eq. (5.75), slope @, and defle nv, are Obviously smaller in abs 25% for @, and 20% for v,, respectively. the one-element results jolute magnitude. The errorsSide sti eg ete ie 147 be j Let us then model the left half of the beam by using two beam elements, 1-2 and cash with length L= 1/4, For each element we apply a concentrated load pL/2 ae end. The stiffness equations are ae 0 4 £ | 2 0 é pe | ‘ 6 24 12 P| gr Ee | * | Bl) We o 2 tole [-§\lo 2 ié with ° 2p 2 | =p le vs ‘The equations can be rearranged and partitioned as \ 6 0 4 2/— 0 e (Pe 1 6 0 2 o -= I> 8 \ ic 6. 4 \ ee a 6 ' 24 12 1 ee ce a zalnes ze ee ee A rae rae) | By the method of substitution described in Eqs. (5.49) through (5.53), we obtain (h-alt aN} (5.80) and —pL Pl _ se 16 ia un = BJ, i The solution is wl BPs 7 {Pe pL* [27 -— php = ~oRy (3.82) oJ ~oeLi1 16) )-2e( ~~ i2erlas148 Boar and Plane Frame Elements At this stage, we can substitute 1/4 for Land obtain the final solution 9 1024 191 pl’ } 1536 wai 5 128 ii 256 on Compared to the exact solutions, we find that by using two elements the errors aq reduced to 64% for 0, and 5.00% for vs, respectively. It is expected that the errors will be reduced even further as the number elements used is increased. With the aid of a computer, results are obtained for cas with various numbers of elements and are compared with exact solution in Table 5, ‘The negative signs in the table indicate that all the results are less than exact solutio: in absolute magnitude. It is seen that for a three-element model (for half of the bea the deflection reaches a satisfactory level of accuracy. TABLE 5.1 Percentage Errors of Maximum Deflection and Slope for a Uniformly Loaded Simply Supported Beam Number of Elements for Error in Midspan Error in Slope Half the Beam Deflection (%) at End (%) 1 —20,000 =25.000 2 =5.000 6.250 3 -2.220 -2.776 4 1.250 =1.563 5 0.800 =1,000 6 =0.560 —0.688 7 =0.408 =0.510 8 =0.313 0.391 9 =0.247 =0.309 10 0.200 —0.250 5.5 APPLICATION OF PLANE FRAME ELEMENTS The method of assemblage and procedure of solutions for using plane {ra elements are introduced with the aid of illustrative examples. Frame Example 1 Inextensible square frame. A square frame subje to a pair of parting forces P is shown in Fig, 5.21. It is assumed that the frame mem are inextensiblesand that thesrightanglesiatitherjoints"arexpreserved: Let it be desijeation of Plane Frame, 56 Applicat Elements, ae 149 a I SSG i Figure 5.21 Inextensible square frame. to find the deflection shape, the shearing force diagram, and t diugram. Because the square frame is a structure of double symmetry, only a quadrant need be modeled and two elements, 1-2 and 2-3, are used. The reason that we assure the members to be inextensible is to eliminate the axial degrees of freedom so that we can obtain a simpler example for demonstrating the necessary procedure. The present formulation for the plane frame elements is by no means to be limited by such an assumption. . The boundary conditions for the two-element model shown in Fig. 5.21 are the bending moment Di OL tp = ky = Oy due to symmetry and due to inextensibility As a result, we have only to formulate a set of three stiffness equations which are assembled from the two element stifiness equations defined in Eq. (5.32). The direction sines and cosines are for element 1-2 A=0 and w A=1 and »=0 forelement2-3 Thus we have pe 4 2 0 = & 0 2180 Beam and Plane Fra The solution is Pp Fj ? ‘ PL? | 2 f o,f * ener] 7 not needed |4 0 48EI | é The internal shearing forces and bending moments can be obtained by su the resulting nodal displacements and slopes into the element stifiness equation For element 1-2, we have 12 6 Pp — -= L x zB iB 5 2 xi 0 ° 0 0 3PL AN le 55 a 5 ° 8 - eco eT A Sie | 2 Z a 2 Y> 0 oO oO 0 6 PL M; = 4 2 L : 8 J For element 2-3, we have a 0 0 0 6 12 Y> 5 on 0 6 4 — 5 2 soilless eC peyl et | 0 0 4sEi/ | 0 6 12 He “Tr Lc 3L 6 M; 2 “Er ® Ei j not needed The deflection curve, the shearing force, and the bending moment diagrams are sho in Fig. 5.22. It is important to note that the forces X> and X3 in element 2-3 are obtained zero, which obviously should be equal to ~P/2 and P/2, respectively. This is ca¥s by the imposition of inextensible assumption that u, = 0. Let us now disrega®d condition of inextensibility. Frame Example 2 Extensible square frame. _ If the members in the sat! frame shown in Fig. 5.21 are not assumed to be inextensible, we have one extra 4&8! of freedom (u; # 0) in addition to what we had in Frame Example |. Thus we hav‘pplication of Plane Frame Elements Inflection point Shear Moment 5.22 Deflection shape, shear m, and moment diagram for the in Fig. 5.21. four stiffness equations. P f 2 e }° lle | 0 S 6 0 vs By using the method of partitioning and substitution as described in Eqs. (5.49) E53), we obtaii { } ale i ‘} Ds. uyBeam and Plane Frame Elem, Mens 182 and The solution is SAL? | P __PL _ SPL? {“} L 1 oar 2 2BA 48Er uJ EBA i 1} o mele J 2EA and | | Ets} vj 4881 (3L. (54 and v; are the same as those obtained in the inextensible cues, (5.83)]. The solutions for u; and w; are, however, different. It is seen that Wen eee te axial rigidity EA is included in the stiffness matrix of element 2.3 by sa, u, # 0, the magnitudes of the displacements 4, and uw, are both increased by an ee of PL/2EA. ‘When the new values for the displacements and rotations are substituted the two element stiffness equations, respectively, we find The solutions for 03 P - x, — ans 2 % F y, 0 Ys 0 3PL wm pel fae |e =j\ Pp and =) p Xs a Xs > Yo 0 ¥3 oO PL PL Mi US my oS caer oe found in the inextensible case, X, and X3 now possess a —P/2 an , Fespectively, instead of ‘axi is problen mene nea y, of zero. The axial forces for this pro! Frame Example3 Antis: i i ymmetrical Z frame. Figure 5.23 shows an me zi frame under two concentrated loads P at points 2 ‘and 3. All membes! Se Aten eer naa rigidity EA, and bending rigidity EJ. Let it be requ sha; fad the deflection shape and the shear foree and bending moment diagrams ® > |sation of Plane Framo Elements applic ‘sa a gue $23 Antisymmetrical Z frame. P 2 cas Whey The fixed conditions at points 1 and 4 provided that by wy an any itutede We have a set of six stiffness equations: 0 Re 0 £ ° $ is y _£l ze -$ 8 = 0 2 8s of * -2 0 =e Ree 0 -$ uy be o =k 0 0 Rig $ ns oe |" $ 0 2, -< $ 8 || atle This i , tear is @ problem of antisymmetrical frame under a pair of loads acting along antisymmetry. The displacements and rotations have the following relations: u, = Uy 2 = V3 0,=—05yp Cha Boam and Plane Frame Elemen, 1 ts nditions can be imposed by subtracting the Such antisymmetrical co foung from the first, adding the fifth column to the second, and subtracting the ,j,1, Sy from the third in the stiffness matrix above: Oh 24 ° R+G 0 0 } AZ ma Fas L 12 6 bd 0 = —> 6 EI L L . ac 2 L 24 0 R-Z 0 0 12 6 | §& v L 12 6 a 2 6 ° EB L The first or fourth equation gives u,=0 The second and third (or fifth and sixth) equations give the same results: {3} Pi? 7 0,5“ 6ET )—7 It is seen that the axial rigidity EA (or R) does not appear in the sole Clearly, they can be disregarded at the beginning of the formulation. This is dit? the fact that all three members carry only bending moments and shear forces, 10% force. The deflection shape is shown in Fig. : 5.24. The forces and moments for eles® 1-2 are obtained by multiplying its stiffness matrix by the six nodal degrees of freeé* x 0 nN 1 M, = x,(7F 0 ¥2 -1 M, £ 3 \ # Because of antisymmetr a 'Y, the shear forces men sof sufficient for us to plot the she, Tioment diego eae tame a : ame a in Fig, 5.28 ‘arand moment diagrams for the whole frame oe Figure 525 shaumble 4 Inextensible square frame connected to tS iy! igure 5 San inextensible square frame braced by four truss b* hof Plane Frame Elements oy "i pplication on? 156 Inflection — \ point ra v3 shear P PL 2PL Si lute duet ome Moment lems dat PL 3 3 Figure 5.24 De i : a 2 ewe 824 Deflection shape, shear diagram, and moment diagram for the frame tds ate conne “ay FL Foren ee frame by hinges. The square frame has constant bending abtisidity with Bate of solution, all truss bars are assumed to have the same | db required to find 2V2EI/L?. The frame is loaded by a pair of parting forces P. Py nd the deflection shape and the shear force and bending moment iagtams fi a for the frame, and the axial forces in the truss bars. ‘4, The quadrant contains nO" ty Because of g: “ane dene aiminietry, only a quadrant need be analyze oundan , 1-2 and 2-3, and a truss bar element, 1-3, as shown in Fig. 5.25. 'Y condi f , . m ~ itions for the quadrant are ve 1 = Us = 0, = we 9,=0 — due to symmetry due to inextensibility of the frame memberpe Ene cie ee Uy Beam and Plane Frame Elemen, s 156 Shy ge dl tt 7 ) p———>F P L tt Fa Figure 5.25 Inextensible square frame braced by truss bars, As a result, we have only to assemble a set of three stiffness equations for the frame elements and a set of two stiffness equations for the truss bar element For the two-frame-clement system, , 12 6 XS tO lfm EI| 6 6 0 p=—|-— so -- 6, LIL L Ga 6 12 Ys -2 Giles | ‘ . Boe | For the truss bar element with length V2L and axial rigidity 12V2EI/L?, we have, int | Eq. (4.17), | Xt) _ EAT § -3]f[m]_oeff 1-1] fu sth wSTJorl-1 1 = i | ysJ~ J2zL-$ sJlos. Bl-1 1} los Due to symmetry, the loading conditions for the quadrant are , P X\+Xf = external load = —= Yst+ ¥5=0 ‘The total stiffness equations are obtained by the superposition of Eqs. (5.85) and 6" a 6 6 \ 2 “rT "Bll" 0 6 r g 8 1\|)% 4 ° Ao } gor)ion of Plane Frame Elements cat 4, se epee i ML 50 fh , olution 'S Tre p 108 = i uy 2 } | 144 pot needed I neede ) 8 L ” | 84 0 bs sion curve for the structure is shown in Fig. 5.26. The internal shear forces es moments are obtained by substituting the results for and DEM “they, Inflection point (3 ave, fat Shear a ae Moment pur, ® $26 i teen, ung Deflection shape, shear me ‘| Fig sgment diagram ‘for the a>y o Beam and Plano Frame Elemeny Chay rotations into the two element stiffness equations, respectively, Xi 6 xX, 0 y, 0 Y2 2 M|_PJSLU ang J MLLP J-“ ) xy 16) 6 Xy 16 )o Y ° Ys -2 ' M, L Ms 3L The shear force and bending moment diagrams are plotted in Fig. 5.26, The axial force in the truss bar is found by using Eq. (4.24), EBA PL’ { * ° sin 45° Sis at Leos 45° sin 45°] 96E1|-7 -22P “8 In the shear diagram we can see that there is a jump of 3P/4 at nodal point 1. 1, force is equal to the unbalanced force between the external load P and the horizon, components P/4 of the tensile forces in the two truss bars. Such tensile forces in in, bars also cause a jump of shear force of P/4 at nodal point 3. (tension) Frame Example 5 Portal frame with inclined columns. Figure 5.27 shon! a portal frame with columns inclined at 45°. All three members have the same lens L, axial rigidity EA, and bending rigidity EZ. A horizontal load P is applied at jos 2. Let it be desired to find the deflection shape and the shear force and bending mone: diagrams. QO « @® Lb Li vz Figure 5.27 Portal frame with inclined columns. 4 Boom and Plane Frame Elements 160 Chg |g int | Moment pone AE 211 + 6/2) Figure 5.28 Deflection shape, shear diagram, and moment diagram for the frame in Fig. 5.27 equal; that is, P\=Py ) To satisfy this requirement, we split the load P and make P BP, ot 2 of . “ s not,” The split of the load makes the loading condition antisymmetrical, which does course, change the problem. aot cation of Plane Frame Elements application 161 tne first set of three equations in Eqs. (5.87 = ) to obtain © eras adore the 33 stillness matrix, excluding BI/ 1 nour solution, L, is an eter ue fhe oe 144. K| = (132+ 722) S4+—8 [KI =( ate gjoint method of inverse, we obtain the solution ihe a sing # 138-472, iA 5R pea Bi 2 2 L/El 42+36V2 ofa “SRt not needed | ) 0 (6—3V2)R_72(1+V2) a EE 0 and c 5 i pray Hr SRL? +6(23+ 12: 2) vy =} -sR1°+6(7+ 6V2)L (5.88) 288+ 24(11 + 6V2) RL? = a 0. 3(2 V2) RL? -72(1 + V2), Because this is an antisymmetrical problem, the results obtained in Eq. (5.88) ae sufficient for us to plot the deflection curve as shown in Fig. 5.28. Substituting the results obtained for nodal displacements and rotations into the siiiness equations for elements 1-2 and 2-3, respectively, we obtain the nodal forces and bending moments. For element 1-2, we have x, -R, 6 _w2 2 i L y R632 2 L iB 2 , x R_6 32 2 2 L 02 % R,6 WN? 2 re. L M, Ea 4 = 24-211 +6V2) RU? 24-2(4+3V2)R 21 +2V2)(6+ RL*)L 48+4(11+6V2)RL? ) 2442111 +6V2)RL* —244+2(4+3V2)RL [-124+ (443 2)RL\L162 or Beam and Plane Frame Elements For element 2-3, we have Xx ¥2 For element 3-4, where the forces Y, element 1-2. It is desirable at this as a free body: P 48+ 4(11+6V2) RL? i} LF.=p-2_P 22 P 2 (4+3V2) RL7JP 24+2(11+6V2)RL? (1+202)(6+ RL?) PL 24+2(11+6V2)RL? xX By, [=12+ (44302) RL7}PL 48 +.4(11+6V2)RL? 0 24—2(44+3V2)RL? [12-(44+3V2) RIL 0 —24+2(4+3V2) RL? [12-(443V2)RL7IL N+¥s=¥,-¥, <0 7 L , = Mo (4) ~ ¥s(1+V2)L—~ M,— My = P(E) Yil+V2)L—2M, (5.0) (sal) and bending moments M, and M, are given in Eq. (5.8) Point to check equilibrium by considering the entire fra™ m0 8 Foyjieation of Plane Frame Elements Sh, my gs APP 163 expression for Y; and M, given in Eq. (5.89), we can easily show that eg on th - E My =0 pasic bending problem such as this, the effect of axial displac } Fora Minay contribute very little to the deflection, ncaa emt le, ° AL? (Si RLS = 2 tensible assumption, we can obtain a modified set of results for displace- ns, forces, and bending moments as those obtained in Eqs. (5.88) to done by simply dividing both the numerators and denominators of those ssions by RL? and then setting RL? equal to infinity. lacements and rotations we have 2 PL?/EI { a », b=—— 4] sz (5.92) 6, 24(11 +62) acta (sq iS seen that uy and v, are equal in magnitude but opposite in sign. It means that nodal point 2 is deflected in a direction at 45° with the horizontal line or perpendicular tothe axis of element 1-2. Within the limit of small deflection theory, such a conclusion agrees with the assumption that element 1-2 is inextensible. For element 1-2, we have posed on inex! ments, rotatio (591). This is cesulting expre For disp! P x 72 y, ‘1 211 +62) (3a is (1+203)PL (5.93) 2(11 + 6V2) is =X, 30) . 3 7 Me (44302) PL o ee k “Srelement 2.3, we have x, Swi xs _2(44+3/D) M; P —(443V2)L sa x,{ 4a1+6v2) © = 2(4+3V2) Ms —(4+ 3V2)Ey 164 Boar and Plano Frame Elements ni hap Based on Eqs. (5.93), (5.94), and (5.91), we can plot the shear force ang ,, Moment diagrams as shown in Fig, 5.28. "i, It is noted that in this problem the inclined columns are not perpeng;, the loads. They are subjected to transverse as well as axial loads, In order ntl, . 0 tran axial load, the axial rigidity ZA (or R) must be included in the stiffness may i . If EA is not included, the solution will be in. For example, the supposedly downward displacement v2 in Eq. (5.88) will be pon ' fs Ot, if R is set to zero. " as we have done in this proble: Frame Example 6 Two-story frame under uniformly distribute Figure 5.29 shows a two-story frame structure under uniformly distribute, faa S Nog | P on the second floor. The lengths and bending rigidities are marked in the fj igure, it be desired to find the deflection shape and the shear force and bending mon.” diagrams by using six frame elements. 7 es Tength Figure 5.29 Two-story frame underu 5 form load on the second floor. It has been shown in Sec. 5.4.1 that work-equivalent loads or consistent lox concentrated at nodal points can be obtained to replace the uniformly distributed los! p. Using Eq, (5.7) for shape functions and Eq. (5.72) for work equivalent loads, ** find for element 3-4, 6 1 Y, -= — 2 = ; it Ms = 1 1 ape =pee 3 oo} Y, 2 | 6 oa 1 L t My I EL 3 % where the length for the element is 21.sof Plane Frame Elements io ooo ie in Vertical eo wn it es feolumns 15 Contribute | litle ot alter the He ity | G eimie jest of four silfness equations: Hay | ot! pave le 3 i ‘ qeat24 ho i } 2 442412 : . (5.96) 0 122444 a i d 2 0 2 444424 | Lo sroted that beeause of the assumption of inextensibility, the forces Y, and Y, ulin Eq, (5.95) are not needed in the formulation (5.96). To impose the symmetrical conditions that 85 84 ‘simply subtract the fourth column from the first and the third column from the onl inthe stifiness matrix in Eq. (5.96): 2 8 20. 2 * pe 3 EI} 2 16 | | @ pie la 4 ia -16 | | . -2 2 et SU (or th list) two equations give the solution: ~ (ef fa { al (597) s EI\=10. yy substituting the ind bending moments are obtained by sues quatios {and the zero displacements into the individual stiline aT Heiner bsg ema shea "for ropnt Shear forces a ations » > anesy 168 Boar and Plane Frame Elements —¢, Chap 5 for elements 1-2, 2-3, 2-5, and 3-4, respectively. x, -3 Xs 27 Y, 0 a 0 M pl i M,\ __ pl | —8! ) xX, { 237) 3 X, (237) -27 } Y2 oO Ys 0 } M; 2 Ms) -191 (ss X 0 xX; o * Y2 0 Ys] 0 J M,\ _2pP J} 1 Ms; (20pP -T i x,{- 79 ) 0 Xs 79) 0 Ye 0 Ys 0 M; 1 Ma 1 : 7 At this point it should be remembered that element 3-4 is subjected to a uniform), distributed load. The bending moments M; and M, obtained in Eq. (5.98) for elemex 3-4 should be superimposed by the fixed-end moments (work equivalent momen caused by the uniformly distributed load. The real bending moment at point 3 actiy on element 3-4 is } dp? 38 pP = 33; pP’ (599) Such a superposition procedure is explained in Fig. 5.30. By the same token, we knov the real shear force at point 3 acting on element 3-4 is L ug ¥,=0422= pl (6.100 Because the computations are carried out by computers in practical applications it is popular to use statically equivalent concentrated loads. In that case, such! superposition procedure is no longer needed. The simplicity is, however, gained at! expense of the need for more elements. . Because of symmetry, the shear forces and bending moments obtained in E* (5.98) through (5.100) are sufficient for us to plot the shear and moment diagrams” the whole frame as shown in Fig. 5.31. Due to the assumption of inextensibility: ® axial forces are not obtainable from Eq. (5.98). It is obvious that the columns 1-3" 4-6 are under axial compressive forces of pl. The beam 2-5 is under an axial i force of 10pI/79 and the beam 3-4 is under an axial compressive force of 9pl/7: two values are obtained from the jumps in shear force at points 2 and 3, respec" shot Frame Example7 Arch with variable moment ofinertia. Figure Far 18 a parabolic arch with both ends fixed and under uniformly distributed 1004: a19pr2 237 | FawreS30_ Superposition of bending moments due to end rotations and uniformly ditrbuted load p. | Stoidl axis of the arch is represented by the equation | rl-(7)] (Ne his the rise and 21 is the width between two supports. The moment of inertia "his defined as I(x)= I, seca f(T" axth?\"? | ‘ =1(1+ ) en, g the moment of ais y is the ; Ii an iva qi Slope of the tangent to the parabolic curve an the arch bridge and “W ypgg eg TOMA Of the arch. Such design is widely Useé Of constructions, beUy Boam and Plane Frame Elements , Chap Inflection i point Shear | et 79 Moment 5.2 wr b >> LLL Q Figure 5.31 Deflection shape, 8 a diagram, and bending moment ¢ ad for the frame in Fig. 5.29. es for? In this example, the parameters are assumed to have the following val steel box arch: 1=83 ft—4 in. = 1000 in. P = 6000 Ib/ft = 500 Ib/in, E = 30 10° psisi diag™ es ft! of Plane Frame Elements application 169 Figure 5.32 Parabolic arch with variable cross section. For this symmetrical problem, we have to model only half of the arch. We use six different modelings with 2, 4, 6, 8, 12, and 16 elements, respectively. In each of thesix models, the nodal points are equally spaced along the x direction. The constant moment of inertia for each element is approximated by the average of their values at both ends. Statically equivalent concentrated loads are applied. For the case of inextensible arch, an exact solution is available in Ref. 5.1. The ‘olution gives the reaction forces and moments at the ends and the crown which are listed in Table 5.2. By setting A=1000J, the inextensible frame element solution is cbtined and given in Table 5.2. It is seen that both the deflections and bending ‘foments for this problem are so small that they can be considered as zero everywhere. - 's actually a pure axial force problem without bending. It is seen in Table 5.2 mi erate for reaction forces converge monotonically. In fact, these results are in 3% of accuracy at the eight-element level. A more realistic value of I(x)/1000 ibe" used for the cross-sectional area A. The results are presented in Table 5.3. It 'sseen that th oe ¢ deflections and bending moments are considerably larger in the extensible —TABLES2 Results for the Arch with Variable Moment of Inertia (Inextensible Case) N eae Te ad = ——S . jection lenge of Horizontal Vertical Bending Bending Detection Halt arac’ Reaction at Reaction at Moment at Moment # Crome rh End (kips) End (kips) End (kip-in.) Crown (kip-in.) a ; 442 442 0,002 oots sa ‘ a0 ao 0.012 0.06 os 8 480 480 0.020 oe 83 2 485 485 0.027 eo 8st 16 490 490 0.041 0.585 9.29 492 492 0.055 0.040 500 * °170 . for the Arch with Vartable Moment of Inertia (Extensibig sults for the Arc TABLE 5.3 Re — Bending Elements for Re eicips) End (kips) End (kip-in.) Crown (kip-in,) Halt Arch End —_—— + 1,595 11,489 = al asi éon 41,707 2 or aor 3634 10,466 é 416 dei 11913 10,066 . 481 Pat 15,264 2780: 1 as hee 17,412 9,697 u 5.6 CONCLUDING REMARKS The formulation and solution procedures illustrated in this chapter are on the stiffness method. The method finds displacements and rotations firg, which are then used to find the internal forces and bending moments at eag nodal point. Because the zero displacements and rotations result in reduction a simplification of the stiffness matrix, this method is particularly advantageos, for structures with high number of constraints or high degrees of redundang, The formulation and solution procedures are systematically presenu: so that they can be straightforwardly adopted for computer programming 4 Fortran program for the static analysis of plane frames, a user's manual, a the input and output data for an example of a stairway frame are given Sec. 13.1. By proper rearrangement, the stiffness formulations for beam element([E3 (S.14)] and frame element (Eq. (5.32)] can be reduced to the same form ast well-known conventional slope-deflections equations (see, for example, ke! 5.2). In this chapter, the formulations and the notations are the same as thst presented by Martin [5.3]. REFERENCES 5.1. Parcel, R.B.,and Moorman, R. B. B. John Wiley & Sons, Inc. Maney, G. A., Engineering Studies, No. 1 Martin, H.C., Introduction to Matrix Me: Book Company, New York, -. Analysis of Statically Indeterminate Sine") » New York, 1955, p. 430. BR University of Minnesota, 1915: thods of Structural Analysis, McG®™" 1966, Chaps. 4 and 8.8s fig, each and eau; znted 1A and an in (Es. sthe Rel hose tus oH | pat ee sf nm a PROBLEMS s in the stiffness matrix for a beam clement using the equation term: all 16 L ky = EI i) ST OOS G(x) dx stiffness formulation fia the Dea a eietlonsratell eae points, iagrams, and . sing ments 2 oment diagram for the proble: + Use work: ip ihe bending j far diswibuted! loads. Uchecs otherwise specified, all bean men, Soiatent pe Faalinextendible? with bending Ngidity EL ; pers are assu! frame element stiffness formulation given in Eq. (5.32), find (a) Using the Plane oe and rotations at all nodal points, (b) the shear force diagram, * the Se aia moment diagram for the problems given in Fig. P5.3. Use and (6) tne eng vor distributed loads. Unless otherwise defined, all irae wor NEE ae as inextensible with bending rigidity EL All right ancl ne nea ia be preserved alter bending. ass given in Eq. (5.14), (b) the shear force di ‘ms given in Fig. 5.2 find (a) the ae a eal Pe Qe a, Y S EG ny B — Figure P5.2vy Beam and Plane Frame Elemen, Sg 172 EA= (hy zi a Figure P5.2 (continued)\z Figure P53~~ _eeoo—_"_~—S—‘(_—‘i—‘—_éf v Beam and Plane Frame Elemane, 174 Figure PS.3 (continued)\ problems 9.5 rT Figure P5.3 (continued)’ CHAPTER ¢ Nonuniform and Curves Beam Finite Elemen,; In structural design, we often encounter beams with nonuniform cross section When the weight or the cost of material is of major concern, the beams ar designed with nonuniform cross sections. In civil engineering structural desis towers, arches, chimneys, and bridge girders are examples of tapered beams In aerospace structural design, nearly all the beams are tapered. For examp: wings are tapered for both weight and aerodynamic reasons. In automoiit structural design, examples can be found in the suspension system whet beams are designed in leaf-spring type. The shape of taper conforms toth moment diagram of a simply supported beam under a point load at midsp* Such a design can save material and produce the desirable structural damp Beams with curvatures are another form of structure that we oft eESeeAe ee design. In this chapter we limit our attention iminvelvede and bent only in the plane of the curvature so that no ton ¢c- Examples for application of curved beams can be found in ust rings, reinforcing rings for cylindric, i girders, hooks, and so on. It is com cross sections, eset” a common practice to mode ‘| nt eat It is a * La tay d repr tion usin, é meee Pered beam by a step to mater eesteimenta sath 4 constant moment of inertia. It is also © gam in a broken representation using straight“, elements. Such practices es May no i for ae problem unless & sufhcige Y Not, however, give accurate results for tat F F er it is a better practice to use strai mn of clgments is used. On the oth seoi® or curved elements, or even curved c{ment® With nonuniform F088 “5d . "Curved elements with nonuniform cross § 176sniform Straight Beam Element nun No! es: wien chapter the stiffness matrix for a beam element with in the rertit is derived. The displacement function is based jo injal as that used for the uniform beam element po"; demonstrated by an example. matrix is also derived for a circularly curved beam el 6 he tangential and the radial displacement functions are based Siement: sue ye The element is used to analyze a circular arch problem ands, th the straight beam element. Developments in the study of curved nts are discussed. nonuniform on the same ne 4 Superiority of jolynoial- Faced Wil seam finite eleme’ eq {NONUNIFORM STRAIGHT BEAM ELEMENT tts . , inthis section a beam element with nonuniform cross section as shown in fe 61 is considered. The element has the same degrees of freedom at each mad as those assumed for the uniform beam element ( a deflection v and ssope av/ax or 0). The element has modulus of elasticity E, length L, and avariable moment of inertia defined as ions, @ I(x) = sft + (2) ] (6.1) oa w= afi+(3)] Pore Yeu Yay 02 Figure 6.1 Nonuniform beam clement. here oe Ty i and @ are fo Fetam0'8 the val Sv erdia jal point | and r° ih Met ue of moment of inertia at nodal p Nement geometty- ers Lae! we) The dignccted to define the approximation of the clement OT. same 2) “IN useq pagement function for this element is assum a uniform beam element, (62) 1 1 2 2th Magy PC) = fix) oy + fale2) 01 +A P2 + ADM? the Sha . cf *Pe functions f(x) are given in Bq. (5-7):JY 178 Nonuniform and Curved Beam Finite Elements The strain energy for the element is uy =f rool where I(x) is now inside the integral. Substituting Eq. (6.2) into Eq. (6.3) and applying Castigliano’s the, eA) ix ax? the stiffness coefficient is obtained in the form L ky = Bf TXT) (x) dx 0 Chay (6, Oren, | (64 The only difference between Eqs. (6.4) and (5.13) is that I(x) is inside y, integral in Eq. (6.4). ‘The explicit form of the stiffness matrix is obtained by substituting Egy (6.1) and (5.7) into Eq. (6.4). 12 yy PO M 4Cx _ El Tr 6 Ys fs -7en M, IC where Cy = 143 ( u = 7(—— " atl Cy =1 +2r( 4 Ca = 1+ 7( 1+20( s 2 Ce=1+2: ce (oa 1 atl Cua tte atl at2 atl a+2~ “aeat 6 2.) + at2°at3, symmetric 4 4 eh a+3 6 7 12 9 - + atl at+2° a+3 5 6 9 9 at3 at+3, at+3. ) ) ) ) (6.54) (6st!Side, ‘ing p ication to Tapered Beam Example Applic 179 the constants Cy Ba. (6.58 nol ats ace’ i w ee sa) and (6.5b) are available in Ref. 6.1, ns (6-5 qual? a more general case in which the beam el os ‘na two-dimensional plane, the stiffness matri pita » ordinate transformation procedure de: a the lements are oriented “ Xx can be obtained by ribed in Section 5.2 or E we! In that case, however, the stiffness coefficients related to the two local 6 {1 displacement degrees of freedom must take into account the effect of i form cross section. Let it be assumed that the cross-sectional areas are on defined as s A(x) = a¥[ +s(2) ] (6) shete Ap s the cross-sectional area at nodal point | and s and B are geometric jiameters. Recalling and employing Eq. (4.9) gives My ta Sof k, =E f A(X)SIFi(%) dx We thus obtain the stiffness matrix related to the two local axial-displacement degrees of freedom: Xi) LE s 1 -1]fu tel e£a(irgt yl te} (6:7) This stifiness matrix is the same as that for a truss bar element with uniform Sos section with A = Ag[1 + s/(B + 1)]. For B = 1, A(x) varies linearly with 4A= A(1 + s/2), which is the value of A at midlength of the element. The cr of beams can, in general, vary differently from those defined in Eqs. howit@ (6.6). The definition of a completely general representation may, malts Prohibit the explicit formulation of the stiffness coefficients. The fon ike case can still be accommodated by the use of numerical integra- © computation of the element stiffness coefficients. 82 apy P| MCATION TO TAPERED BEAM EXAMPLE Ata 7 ia % ‘unc’ ©antilever beam is shown in Fig. 6.2a. The moment of inertia is ‘nd the gi’, defined by Eq. (6.1) with r=8 and a = 1. Let it be deste Te “form jeutction and slope at the free end by using both nonunto a! ion. “am elements and compare the results with exact solutioOY 180 Nonuniform and Curved Beam Finite Elements ts fou : 4 @ 30 {r © Figure 6.2 (a) Tapered cantilever beam; (b) modeling by two tapered elements; (c) modeling by two uniform elements. 6.2.1 Exact Solution Chap 5 | An exact solution for this problem can be obtained by direct integratio" of the differential equation for a beam, a EI(x) 3 = —Px (68) u Integrating Eq. (6.8) once and twice yields @ and », respectively. The et integration constants can be obtained based on the boundary conditions ! v= @=Oatx=L The exact deflection and rotation at the free end ate Pr 0.051166. Ely Pr > fs PET) 27 7 eB + | [r ~log (1+ r)] = 0.090668 2 Elo 6 PEI A derivation of Eq. (6.9) can be found in Ref. 6.2. (69) j aaSF Ces) aa Oy, 1 Application t© Tapered Beam Example 101 ge 3.2 Nonuniform Beam Elements 6.2: Let us first model the tapered cantilever beam by using equal-length red beam elements. . Pope jne element is used, we have only to form a 2 x2 stiffness matrix due | ine boundary conditions v, = 6 = 0. Based on r = 8 and a = 1, Eq. (6.54) } oth ives 22 1 te 12 |lo, 6 P {e peed 1 = FE{-oasoasi, oJ 118Bh| 11 30 o | Bll 0.09322 L a . Compared with Eq. (6.9), the errors for v, and 6, are —0.62% and 2.82%, respectively. A model of the beam using two tapered elements is shown in Fig. 6.2b. For the two elements, the moments of inertia are defined as noe rf) grit x hha= suf + os(2)] @ , “here L = 1/2 is the half-length of the beam. e my For element 1 — 2, we obtain, for Eq. (6.5), a iots ase TT: Cu=3 Cu=3 Cam Ca=% Ca=3 Cum of,Vy 182 Nonuniform and Curved Beam Finite Elements Chop, The boundary conditions are v; = @; = 0. We have a set of four Stifty equations: ty 360014 36 22 BaoP Bo hae iB a 14 a a 6 6, } =o | & eS “OL 36 14 36+84 _ 22-38 7 (6.19, Pes PLT Be L 2 22 22-38 = -=—= 16-24 | | 6, L § L 2 Multiplying out the last two rows of Eq. (6.10) and rearranging gives B 120 41L {et = sti [ 20 _ 55 |{3t 6 < oJ ~ 284|-- -s9|le, ct Multiplying out the first two rows of Eq. (6.10) and employing the relation! obtained in Eq. (6.11) gives rel a 6.2324 4.2324 iat P| _Eh| LF L lf om of L | 4.2324 4.7324 a, | L 4.2324 ; f2} ok 4.7324 7 [- 4 J 11-S581E%| 4.2324 6.2324 || 0 ~ L L , Because L = 1/2, ; {2} = BEY —o.aaasaly _ PP { -0.0510807 aS” El 0.36546 J~ Efol 0.091365 Compared with Eq. (6.9), the errors for v, and 6, are —0.17% and om respectively. We thus find that at the two-element level, the errors for %" 0, are already within 1%. 4 6.2.3 Uniform Beam Elements Let us now model the tapered cantilever beam by stepped repress using equal-length uniform beam elements. Each element has @ ay apoticatio' not 6,2¢. n Fig. 6.20. ne element is used, we have (oh (ey 12 EI) | P 7 6 1 P 30EI, om 1 nto Tapered Beam Example uy { PrP “En a 183 of inertia equal to its value at midlength. A two-element model is nt oF ) 15 a 10 Compared with Eq. (6.9), the errors in v, and @, are 30.3% and 10.3%, spectively. Iftwo elements are used, we can find that the errors in v, and 6, become 858% and 5.04%, respectively. The errors in v, and 6, for different numbers ofelements are shown in Table 6.1. It is seen that the errors are monotonically reduced as the number of elements is successively increased. The errors in v; ind #, become approximately 1% at the six-element level. The results obtained ‘section 6.2.2 using tapered beam elements are also shown in Table 6.1 for comparison. 1 lair TABLE 6.1 Percentage Errors in Deflection and Slope at the Free End ae of a Tapered Cantilever Beam Ul rm Elements Tapered Elements: ij i rin Tip Error in Nomber Error in Tip Error in Error in Femen Defiection Tip Slope Deflection Tip Slope — (%) (%) (%) (%) ‘ 2.82 2 30.30 10.29 0.62 at 3 8.58 5.04 =0.17 4 3.92 299 6 2.22 197 8 0.98 1,02 R 0.58 0.62 16 0.24 030 a O14 oT = tay le blem, the tapered of 82m eek’ 8:1 shows that for this tapered cantilever beam Prom ment a a comet is considerably more efficient than the wali m eptem requires a “lusion is important when a tapered beam oF ~Oy hy 5 184 Nonuniform and Curved Beam Finite Elements ation. For example, computation of g = ro ta yn; a large amount of computatio ; be BOD OF dyna. sponse of a structure consisting of tapered beams due to arbitrary fore function, such as that caused by an earthquake, may involve form) computational expenses. Such expenses can be reduced by minimiz,, ) number of degrees of freedom using nonuniform beam elements, 6.3 CIRCULARLY CURVED BEAM FINITE ELEMENTS Curved beams are a special form of curved shells. A study of curved be, is an important first step toward gaining insight into the more complex Although there have been considerable developments in the curved shel elements, they are beyond the scope of this text. Only developments in curs beam elements are introduced in this chapter. The basic difference between a curved and a straight beam is that, int small deflection theory, axial and flexural behaviors are coupled in the curs beam but not in the straight beam. Furthermore, in the finite element for tion, the displacement functions for curved beam elements must be ca of representing three rigid-body displacements: two orthogonal displaceme: and a rotation, all in the plane of curvature of the element. There are many circularly curved beam finite elements available. lis however, not within the scope of this text to introduce all these element detail. For simplicity and clarity, we simply introduce seven types of circ curved beam finite elements by describing the assumptions of displa’™ functions for tangential (uw) and radial displacements (v) and the asus degrees of freedom at each nodal point. To explain in depth, one ¢! " (clement 3) will be formulated and evaluated in detail. Finally, an arche®™ will be used to compare all these elements. Element 1: linear u and cubic v [6.3, 6.4] u a, + ass = 6 D= ay + aas + ass? + ags® d.o.f.’s at each node = u, v, av/as where s is the curvilinear distance measured from nodal poitt | Blemen( 2: linear wand cubic v with rigid-body displacement "oy 4 a1 Cos & — arsin & + ayR(cos B cos — 1) + 438 oh v= ai sin b + ar cos & + a,R cos B sing +ass7t aes | th d.o.fs at each node = u,v, (av/as — u/R) af curvature of the element, B nal & an angular variable measured from t where R is the radius of subtending angle, ¢ is »a wo ci soam Finite Elements snl cured ovo 165 _ tending angle, and (00/08 ~ u/R) is the it Folation of the wh aie # and cubic v [6.6] in yea tas + as" + ays" past aos + as? + ays 7 do.f’s at each node = u, du/ds, v, dv/as za sunt qubic u and quintic v [6.7] usatast as’ + ays? pH ds + O68 + 78? + dys? + ays* + ayos® (6.15) dio.f’s at each node = u, 4u/95, v, 90/08, 670/35” slment 5 quintic wand cubic 0 [6.8] ya, tas + ass? + ays? + ass* + ass> p= ay + ays + ays? + Ayes” (6.16) do.f’s at each nade = 4, du/as, au/as’, v, av/as Sement 6: quintic w and quintic v [6.8-6.10] u=a,tans tags? + ays? + ass* + ags* (6.17) 0 =a, + ags + ays? + ais? + ays? tans dost’s at each node = u,0u/as, au /as?, v, 30/05, 70/05" Bl 7 lment 7: constant strain and linear curvature [6.11] u=-a,sin@ + a, cos @ + as + asd + asd™ v= a, cos $ + asin p + ae 46% ae do. here p = Senet element, and (av/as — ¥/ ca ashen genes beam finite elements are a SP se inte elements and because significant cy pgts OF shell elements, the displaceme Halen si Clements are usually given in a reduc "ing ya lements. In the references listed abo Kl Sten, ee alent original unreduced disp! bei nd in age omPrehensive surveys of curved be" els. 6.8 and 6.12. *s at each node = u, v, (90/8 — u/R) near distanc R) is the rota! ve measured from tion of tangent ecial and simpler fOr as beet made s for theOO vy Nonuniform and Curved Beam Finite Eig, 186 6.13) for clement 2 with Bq. (6,18) 5, 5 ( st us compare Ed: © curvature « of tet oon tangential strain € and the Of the mj dd. For element 2 i are For element 7, Ay(as + ac) ‘The terms ay, ax, and a; are absent in € and « in both equations. They repr. the three rigid-body displacements (two orthogonal displacements and ai tion in the uv plane) which do not contribute to the strain energy. Sug representation appears simpler in Eq. (6.18) than in Eq. (6.13). The remai terms (a4, as, a.) include, in Eq. (6.18), a coupling term ag which occus both wu and v. We also observe that € is constant in Eq. (6.20) but cubicir in Eq. (6.19), while « is linear in s in both equations. Although the rigid-body displacement terms are not explicitly repress in the displacement functions for elements 3 to 6, it was shown in Rel that when the polynomial displacement functions are of sufficiently high (cubic in this case), the necessary rigid-body displacements are imp included. _ The stiffness matrices for elements 1, 2, and 3 were given expliti! Ref. 6.13. In the free-vibration analysis of circular arches performed it® si 3, far titer ageuracy in frequency was obtained by using clemet Shel by wei eh # convergence analysis of a centrally Laan than ae the ne mente 1 2, 3, and 7 performed in Ref. 6.12, it wi" four different tones of eres OF freedom increases, the solutions 1 is superior to 3, element 3 is superior sg 2, dud element 2 is 8 However, clement 1 is reasonably seed heey eee onto! EI) and not deep periot Pre nably good when the arch is inextensi6! a eat curved beet to demonstrate in depth the formulation and apple’, can be fortulated ora We choose an clement for which the stlla 4, lated explicitly and ae od for & applic: y aceura: ca btaine! application. Element 3 is chosen ite results can be ol y4 ja ouie Cubic Circularly Curbed Beam Finite Element 0 ey ~~ cusic-CUBIC CIRCULARLY CURVED BEAM FINITE Sy 6 4, 64 eMENT 6.4.1 Element Description and Displacement Functions A circularly curved beam finite element is shown in Fig, 6.3. The element 4 ,gconstant bending rigidity EU, axial rigidity £A, constant radius of curvature fe subtending angle B, and length L or RB. The angular variable # and gutance variable s or Ré are measured from nodal point 1. The element ssesses four degrees of freedom at each nodal point: a tangential displace- mnent t, a derivative of tangential displacement au/as or u,, a radial displacement v, and a derivative of radial displacement dv/as or v, or slope 6. t No eo” uw aa v * mL as), Maer Figure 6.3. Eight-degree-of-freedom circular beam element, in Re high’ i . . og inp The displacement functions for both the tangential and radial displacements for this element are of cubic polynomials in s as given in Eq. (6.14). The eight constants a are obtained by using the conditions of eight nodal degrees of freedom at both ends. After substituting the a's back into the displacement functions and factoring out each degree of freedom, we obtain ‘he displacement functions in the form similar to that given in Eq. (5.6), u = fi, + fas, + fala + fattes (6.21a) v= fv + fas, + fd2 + fads, here the shape functions are f=l-384+28 Sr= LE - 28 + &) (6.21b) Sy = 367-28 fr= L(-@ + &)| and Curved Beam Finite Element, Nonunifor 188 . and Ro - ‘ RB OB (63 ) 6.4.2 Strain Energy Expression i ral thin shells are well k cree etrain energy expressions for Bene ’ doen, for wee e, Ret 6 ta) The strain energy expression for a curved begs, e , Ref. 1 a special reduced form of that for a thin shell, BAL gee EE 2 u-FAf ease «? ds _ where € and « are the axial strain and curvature of the middle suris: respectively, with ou v v =i sats €-3s R R a (a tau Fe _ ly yy K" Ras as Substituting Eq. (6.23) into Eq. (6.22) gives U = Ut Uw + Vee a where 2. Ea EI(* i Uw = u'vds —— ‘o" we z Je sia J, we" ds EA (* Er [© Un = i? Y mS i ail a (was L EI(* vo? ds += “)* ° * 2 |, eas The energy expression i s L i i , flexural coupling, Yiu Uw, and Uy, are associated with ax! and flexural behaviors, respectively. 6.4.3 Stiffness Equations Substit g i S (6.24a) and (6.24b), and then performing partialvy ¢-Cuble Ciroularly Curved Bear Finite Element acu io ve strain energy With respect to cach of the ei 7 ight degrees of freedom, sot stifiness matrix equations for the clement is obtained ne 8 * ’ x au J au, ' ty au x ou, , ; aa Ukuul | Cue] x. aur I 2 jun us 6 Xs = (6.25a) tay M, Uke) | Eko] ¥ 5 ! v> ' M; au t 6, 30> \ su Where Xj and X4 are the counterpart forces in inch-pounds associated with the degrees of freedom u,, and u,,, respectively. The coefficients in the four 4*4 submatrices are obtained as se . @ : " Keay, = | EA( 1 + 3) SiSi as ° L - Kevoy = Kou =| EA pf, — of fi) as (6250) 3 " 1 =) oR wo na( ff belts ) as o AR 0 Where BA uo TRE primes indicate derivatives with respect to sand «= FIrises urn iiata leet om ie teed evinaenele att = Vinge “ Lau Ms Hae wre ayuda an Fb ei wr Tae ea per Peron it ba need tee wt arestiraqead © 13 dag a ANT Vines yes eo the om Uiul ae Leds) sau, ba (us Ro Uo mritTnea <-™MAy LNEN Hie tn Lused Teescircularly Curved Beam Finite Element pireu! pic cubic 189 ergy With respect to each of the eight de in ene} r Brees of freedom, trix equations for the element is obtained stra he a pat mines me 6, | . au ! u } xX au, I | au I i x! i 1 fs, a i ou, Dk) | kw] ey, a 1 Nis > 2 ‘ Xz uz ' , ' ' (6x te maa -ba- (6.25a) 1 vy Ma Me I M * [ku ) [ko] I (6.2 1 2 al Y; oo. au ! ah | 2 M, 305 L ' 2a tn i i i ON hee X} and X} are the counterpart forces in inch-pounds suede with “. 5 Re EA Tk == © symmetric L ’ 1 2k 2 3 6 1 6 “Lb 2 24 1b 1 2k 276 2 3 1 Lia 1 Li at Sf _-_9 =4% gee 10 LL 2 10 4 Lye 948 bie Lia Tot Lk 2 10 L 60° 2 ‘ Loo 1 . (62% A Fp atO. —ae 2 102 2 10 Lb -E_s Pia L,e# 4g 2 10 L 60 2 107 2] 138? | 120 ; aa a symmetric 11LB? 6a Lg? 4+ Se 20 "Lr 10s 4 13LB* 6a 138? . 12a 420° 7 35° 7 292 2 257 EB" 5g be? 6a LB’ 4a 140 210 Lo Ws f Equations (6.25c) are available in Ref, 6.15, in which a vibration of prestressed circular arches was performed. They were given earlier” 6.13 in a different form. It is seen in Eq. (6.25c) that for thin and arches, the thickness/radius ratio is small, so that a/R? « ((/R) <' | a/R* may be neglected from submatrix (k, For curved elements with nonuniform cross sectio curvature, an explicit form of stiffness matrix may then be obtained integration shall unde at = or variable iffness matrix may be difficult t Ol through Eq. (6.25b) by usingcurved Beam Example puication 10 curved B : . applic 10N TO CURVED BEAM EXAMPLE A ppLic 6s uate the performance of the cight-degree-of-fr valu ier ery example of a semicircular arch as shown in Fig, 3),anexa t3)s dom element 3 4a is analyzed are defined as A=1X1in?, R=17in., p— 2000 Ib, and jot jeme 0" psi different approaches are used to analyze this problem, Three 2000 Ib 1X Lin? 17 in, (6% " (a) (b) Figure6.4 (a) Circular arch under a central load; (b) half of the arch under analysis. 6.5.1 Exact Solution by Castigliano’s Theorem Due to symmetry, only half of the arch as shown in Fig. 6.4b need be wuulyed. At an arbitrary point B, the bending moment and axial force are, spectively, | M = PR sin @ — QR(1 — cos 8) — Mo | S = Psin 0+ Qcos@ a} Thestrain energy expressions are : ua ["° M’Rao (sae 2ET lo «DEA rein: S°°4US€ Of symmetry, the tangential displacement 1, and rotation 0, at 8re both zeros. Hence F+ on(3Z—2) + m(2-1)] ~ PR + oRr(Z-1)] =o 2Ww onuniform and Curved Bearn Finite Elermeny, 192 N Moy i equa imultancously give: sregoing two equations sim! y gives BRAC 2/4) — 201, Q=—LBtw RAC? — 8) + 7 mal (E-)e| ' ction at C is obtained as aU Ri( PRE QR_ Mo) +£(2+9) Solving for the fe The radial defle 4 2 EA\ 4 2 For R=17in, A=1x Lin, I= 1/12in.*, and P = 1000 1b, Q = 0.91591372P Ib Mg = 0.30379484PR = 5.1645123 P in.-lb bp, = 0.14152379 in. 6.5.2 Using Eight-D.0.F. Curved Beam Elements (Element 3) If one element is used to model half of the arch, the boundary condition are mann (Z)-0=(3).-° From Eqs. (6.25), the stiffness equations can be obtained as xt 356.150 —89.0375 —15.7263 ] (4%, {=} = 1of “san 356.150 15.7263 {| Ys -15.7263 15.7263 3.43724} | vs Inverting the matrix gives Us, 35.3144 2.12302 151.860 o {isl = | 2.12302 35.3144 —151.860 | 2 151.860 —151.860 4298.91 1000 which gives ¥ = 0,004299 in. with -97% error : The solution for the central radial deflection using 2, 3, 4 ee length elements is given in Table 6.2, The solution converges ™P" monotonically as the number of elements increases. The error reduces" than 1% at the four-element level (15 degrees of freedom). It is "°° better accuracy may be obtained if elements of unequal length at ¥ using smaller elements near the central load). ysedy | to Curved Beam Example ation noo wa ysing six-D-0-F. Straight Beam Elements 553 i 6 a Gj » nivnd. | interest CO test the simple straight beam elements for thi h yisotieine central radial deflection by using up to. t6 * Problem cats OFS given in Table 6.2 for comparison. itis seen thay anne? J ne Sooate ee Op degrees of freedom (not differing by more ang, ot” (eye MUO e better than the straight ones as n ore than 1), the j wey alememts ar 7 umber of d.o.f's - 19 gel cpLe 62 Convergence Study of the Two Types of Finite Klemen TABL for the Arch Problem ents _— a Curved Element 6-D.O.F. Straight Element | Number | Center Number | Number| Center _ a of, Deflection Error of of Deftection | Error : ans) DOF. (in.) (%) | Elements| D.0.F. (in) ra - i 3 0.004299 97.0 1 1 0.004800 | 96.6 2 7 0.114635 19.0 2 4 0.127804 9.69 3 I 0.137156 3.09 3 7 0.134743 4.79 4 1S 0.140381 0.81 4 10 0.137483 2.86 | 5 19 0.141123 0.28 6 16 0.139643 1.33 ‘ 23 0.141353 0.12 y 19 0.140127 0.99 1 27 0.141440 0.059 8 22 0.140447 0.76 & 31 0.141478 0.032 10 28 0.140829 0.49 F 35 0.141497 0.019 VW 31 0.140948 0.41 0 39 0.141507, 0.012, 12 34 0.141039 0.34 B 0.141513 0.0076 14 40 0.141167 | 0.25 a a7 0.141516 0.0055 15 43 0.141213 0.22 wwe 16 46 0.141250 0.19 _ Exact center deflection = 0.1415238 in. This example has been analyzed in a review paper by Dawe [6.8] using d ; ‘i isto 5, 6, and 7. Dawe's figure for the percentage error ' cen ‘sin gy Qa, number of degrees of freedom is shown in Fig: 8 ne stight cine, Straight beam elements are also plotted in Fae. oof degrees reedom Ments starts with quite good accuracy at al eases ppt COnverges very slowly as the number 0! ite superior. | The Jlement 6 is by far the most superior; element 7 18 ANY Tico a | Moone (Ch iS considered as thick and deep. In Ref. 6.8. TEINS Ti and i ‘hilgy ¢ £OF the interesting cases of thin and deep (f= 0-1 im Coe, “aiigiy seteNding angle = 30°), and thin and shallow. When the section ine pati? OF EA/ EL is relatively large, the arch is relatively enon in ie ttain eal deformation is more dominant than the axis! a reformation ‘No mone tBY: When the arch is relatively shallow, the Hsu! ore lominant j inant in the strain energy. ow num! “ f degrees of freedom >»194 Nonuniform and Curved Beam Finite Elements, Ch roentage extor in center deflection ° 10 20 30 40 i Degrees of freedom . Figure 6.5 Convergence study for the arch problem using six types of finite elements. In general, the axial-flexural behaviors are intr structures. ately coupled in the art It is recommended that we not be prejudiced against the «i displacement function uw; that is, the degree of accuracy or order of polynomik assumed for the axial displacement function u should be comparable to thit for the flexural displacement function v, 6.6 CONCLUDING REMARKS a A tapered beam clement and a tapered axial force bar element based : assumed variations of moment of inertia and cross-sectional area, respect’) have been introduced. The two elements can be combined and oriet! : siott! arbitrarily through coordinate transformation in two- and three-dimensi”! space for frame analysis. ef A tapered cantilever beam example shows that the tapered beam ¢le™ . ~ et is far superior to the uniform beam element: that is. be! the former gives accuracy than the latter for the same number of degrees of freedom a recommended that tapered beam elements be used for the problems of aart anid mia v th ces 195, refer with nonuniform Cros Sections. In the general cas; james "ross section Is such that explicit formulation in of Te ipited, numerical integration is recommended is Pus of circularly curved beam finite elements, all based o; oo Gever, VP ment functions, have been introduced. A review and ean goed disp ved beam finite elements are available in Refs. 6.8 and 6.12 son OF cGvantages of using displacement functions to formulate an element neo! on ‘on the energy principles, the effect of inertia force and initial ita ean be straightforwardly included for the analysis of dynamic, (ef finite deflection problems (see, for example, Refs. 6.13, 6.15, that the of stiffness 9b 6:10) | been learned, from the development of shell finite elements, that kite axial and flexural behaviors are coupled in the curved beam finite pecause ye displacement function for the tangential displacement has to be em ticated (of the same order) as that for the radial displacement. as Sere should note that numerical integration is a convenient and accurate | yith which to formulate the stiffness matrix when its explicit form is ‘ficult to obtain. The difficulty arises when the curved element assumes an aisly curved shape, itregular variation of cross section, or complicated displacement functions. , Because curved beam finite elements are a special and simpler form of ull finite elements, learning how to deal with curved beam elements is a logical first step toward an understanding of the shell elements. REFERENCES 6. Gallagher, R. H., and Lee, C. H., “Matrix Dynamic and Instability Analysis with Nonuniform Elements,” International Journal for Numerical Methods in Engineer- ing, Vol. 2, No. 2, 1970, pp. 265-276. Martin, H. C., Introduction to Matrix Methods of Structural Analysis, McGraw-Hill Book Company, New York, 1966, Chap. 5. Connor, J., and Brebbia, C., “A Stifiness Matrix for a Shallow Rectangular Shell Element,” Journal of the Engineering Mechanics Division, ASCE, Vol. 93, No. EMS, 1967, pp. 43-65. Gallagher, R. H., “The Development and Evaluation of Matrix Methods for Thin hel Structural Analysis,” Ph.D. thesis, State University of New York at Buffalo, 6. 65. Cay 62, 63, 64, , G., and Clough, R. W., “A Curved, Cylindrical Shell, Finite Element,” 55, AIA Journal, Vol. 6, No. 6, 1968, pp. 1057-1062. is * Bogner, F. K., Fox, R. L., and Schmit, L. A., “A Cylindrical Shell Discrete a ciment,” AIAA Journal, Vol. 5, No. 4, 1967, pp. 745-750. siecil|aeinite “Enter, G. R,, Lindberg, G. M., and Olsen, M. D.. “A Shallow stelle “Ment of Triangular Shape,” International Journal of Solids and Strw 2 NS: 8, 1970, pp. 1133-1156. _ &na jon 1) Ini Disa ery problems oe Kx) (4) Figure P6.4 go: Byuation (6.25a) for the eight-degree-of-freedom clement can be arranged in the following sequence: (elle tty F key kell LF = Lx) X54) la} = Lu, 4s, F}= LG ¥1 Mi X2 ¥2 MI where la} = luv % wa v2 2] Assuming that Xj = X4 = 0, the two degrees of freedom u,, and u,, can be written in terms of {q}. We can then obtain {F} = (kl {a} 6x1 6x66xI1 Find [k] in terms of the submatrices [ky :], [ki2]}, (Ko:], and [k22]. What difference do you see between the original eight-d.o.f. and the new six-d.o.f. elements in terms of interelement compatibility at the common nodal point? P , | Figure P6.7 angle = 30°, R = 17in., A= d the central deflection by lar element, and (c) one 0.223073 in. 63 i Pre P6.7 shows a circular arch with subtending *1in2, B= 30 x 10° psi, and P = 2 x 10°Ib. Fin “sing (a) Castigliano’s theorem, (b) one eight-4.o-f. circul Straight beam element. Hint: The exact central deflectionOe Ty CHAPTER - Free Vibration of ‘Truss Bar, Bex», and Plane Frame Finite Elemey, 7.1 ONE-D.O.F. SPRING-MASS SYSTEM The simplest example of a free vibration system is a one-degree-of-freeie spring-mass system, as shown in Fig. 7.1. When the mass is vibrating « distance x from the original static position, the equation of motion is | | kx + m¥ =0 & = spring constant Static position Figure 7.1 Free vibration of a 09% degree-of-treedom spring-muass system 1 where k is the stiffness of the spring and the dot represents derivat® respect to time, The term ky is theo ' he inertia force of the wer he tic force in the spring, whils © The inertia force always acts it NS. AL the static position, the as Lo aatre ing xia! Vibration of TWo-D.O.F. Truss Bar Element a i 199 he absolute value of velocity is ximum. At the extreme DOSILIONS Jitude), the absolute value of the Butt pero butt acceleration is 1 ome , ssealaler ximum but the city is zero. The mass accelerates as the spring force increases, in ne verte sense. a if we define oes vi (7.2) gq. (7-1) becomes ¥+ 07x =0 (7.3) ‘The solution to Eq. (7.3) is x = Acos wt + Bsin wt (7.4) Equation (7.4) states that the motion is harmonic with natural frequency w. ‘The constants A and B (amplitudes) in the equation are determined by the initial conditions. If this spring-mass system is subjected to a forcing function f(1), the equation of motion becomes kx + mi = f(t) (7.5) In this chapter, the equations of motion for the truss bar, beam, and plane frame elements are to be formulated in a form similar to Eq. (7.5) but with more degrees of freedom: Lk]ix} + [m]{x} = {£(t)} (7.6) where [k] is the stiffness matrix, [m] the mass matrix, {x} the vector of nodal sisplacements and rotations, {x} the vector of accelerations, and {f(t)} the vector of forcing functions. 72 AXIAL VIBRATION OF TWO-D.O.F. TRUSS BAR ELEMENTS The stiffness equations for a truss bar element with two axial displacement “egrees of freedom have been formulated in Chapter 4 for static analysis. A ass matrix is needed to extend the stifiness equations to equations of motion br dynamic analysis. There are two common ways to formulate such mass trix, 72.1 Lumped Mass Formulation Figure 7.2a shows a truss bar element with length L, modulus of elasticity * “ss-sectional area A, and mass density p. The simplest way to accour > nlvy Teus) | Truss Bar, Beam, and Plane Frame Finite Elon, » Vibration o| 200 Free = ' @ - ot B ———_@». m= PAL o Figure 7.2 Two-degree-of-freedom axial force bar element with mass lumped « igure 7: both ends. for the effect of inertia force is to split the total mass PAL and lump half : the total mass at each end like a dumb bell, as shown in Fig. 7.2b Re Including the effect of inertia forces at both ends, the equilibrium sits, | a equations as given in Chapter 4 become _pAL way eal 1 ee} os pAL.. tl-1 idle x. — Pai, | where the negative signs in front of the inertia forces (pAL/2)ii, and (pAL/? indicate that the inertia forces are acting in the opposite direction to (re the elastic forces X, and X3. Replacing the inertia forces to the right-hand side of Eq. (7.7), we sbi! | the equations of motion: {eI at fe ef 0 x Et-1 ilu, 2 lo 1 or symbolically, {X} = [kitu} + fmyfii} . It is seen that to formulate t all nodal masses along the main For simple harmonic m sinusoidally with respect he mass matrix [m], we simply put the ! diagonal, . . cements " tion, we assume that the displaceme™ to time with a natural frequency , tu} = (U}sin wt where {U} is the vector of amplitudes,pration of TWo-D.0.F. Truss Bar Elamen (y_ Axial Vibra nents 0) ations the accele {i} = —@{U} sin wt ) = —e{uh (7.10) ¢ Eq. (7-10) into Bq. (7.8), we obtain b-FeL es Tie} ean the case of free vibration, the forces X, and X; are zeros and Eqs. (7.11) Fetribe an eigenvalue problem. Je abstr b dy 7.2.2 Consistent Mass Formulation Alternatively, we can formulate the mass matrix by the use of Lagrange’s formula with the strain energy and kinetic energy expressions written in terms \jfthe displacement functions. If the displacement functions used in deriving sing the mass matrix are the same as those used in deriving the stiffness matrix, the mass matrix is called a “consistent” mass matrix [7.1]. It has been derived in Chapter 4 that the axial displacement function for this element is > hay tm u(x) = fi(x)uy + fol) up (7.12) where AlN AG) =1-F 5 ns (7.12a) x a AOD =F The strain energy expression is in the form L ua %Al (wOoP ax (7.13) 0 where the prime indicates derivative with respect to x. The kinetic energy “pression is in the form L y 7-24 { {wi(x)F dx (7.14) 2 Jo Laggan cauations of motion can be obtained by the use of the well-known “nge’s equation [7.2] as follow: 4(22) 4oU (7.15) dt\ag,) ~ a4—_ 202 Free Vibration of Truss Bar, Beam, and Plane Frame Finite Etemenrs where q, represents the ith degree of freedom. It is seen in f, the kinetic energy T is absent, it i the stiffness matrix. For the simplest ease of a single mass m, with velocity 4, ine, energy is known as iD 24. (7.15) s precisely Castigliano’s equation j,. des T= 1 Substituting this T into Eq. (7.15) and assuming the absence of strain te U gives t which describes precisely Newton's first law. Substituting the displacement function (7.12) into the str: } | kinetic energy expressions (7.13) and (7.14) and then performin, ations as indicated in Eq. (7.15), we obtain d(aT au ae) “au, d(pA e & -4(8) J, AHO ST AMC} de +24 [ wear sh twa ‘ain energy zy 8 the differen, az L = GPA) f Afi) tis + Ae} Fi (x)} dx | + 8A fiom + BOOM SCOr ae [oa [nonce ae pa [iconca ax |{@} +| ba i SiC) dx ea |" Fix) FC) ax |{" ‘} U2 = bm ml EL Ue, wl} - By the same token, X= Lomas mal {Bh Ly keai{ th where, in general form, . L my = pA J FODG(x) dx ‘f Two-D.O.F, Truss Bar Elements axial Vibration o! . 1 ky = BA I Fi) (x) dx (7.19) 7.19) was derived in Chapter 4, gui ing out the integrations as defined in Eqs. (7.18) and (7 19) for an = 1,2and then assuming harmonic motion with natural frequency 12 ip results 17 Ste, xX, EAL 1 -1 2PAL[2 1 uy =|—4 — wo (7.20) X LL-1 1 6 LI 2 up, frample 7.1 8Vaq [et it be desired to find the natural frequencies for a cai ten, eaxial direction by using one and two elements, res; he exact solution is, from Ref. 7.3, intilever bar vibrating freely in Pectively, as shown in Fig. 7.3 na /E SE where n= 1, 3, 5,... is the mode number, +-—_4____ x, on ete he alt hea _ 2 P2103 @ oO Patsy | i cantilever bar. Bure 7.3 One- and two-element modelings of # cantil ih On "Chem, J, we have re i ‘i s used, we “solution. If tumped mass formulation (7.11) i : tere = [Lt t]-etfs Ue204 en \d PI rame Finite Elemen: Truss Bar, Beam, an Vibration of equation: only to multiply out the second equation FA pAb (oh [ Loe Jun It is necessary The first mode frequency is j Ltt JE vs. 7 [E. 10% error | w= NV, 2LVp If consistent mass formulation (7.20) is used, we have for the second egy, The first mode frequency is ° — 12 fe vs E> 103% error TVG 2LVp in ‘Two-element solution, If lumped mass formulation (7.11) is used, we have i » P\=2 1-1 0 1 0 074% =0 ¥ AL > P,=0b= 34) -1 2-1-0) 0 2 of [fu=2 "4 P3=0 o-1 1 oot w=? Neglecting the first equation and following the eigenvalue solution procedure givens | Chapter 2, we have 2EAL 2-1 pAL[2 0 det |-—* — @PAL _ “ ale ‘] oa [ “| ° w(pAL/4) _ pL? 2EA/L OSE Let We have 41 = 0.293 and A= 1.707 The natural frequencies for the first two modes are wo, 1531 fe a f[E 1 L ? vs. 2h 7 > —2.53% error w 3095 fe 3a /E EN op p77 216% erroryr 3 Axial Vibration of Four-D.0.F. Truss Bar Elements geo? 206 ir consistent mass Formulation (7.20) is used, we have oat [P54 7 whl gtPALl4 1 j cl- 1J>° li a} >° Let My Ay = 0.1082 and Az = 1.320 The natural frequencies for the two modes are aE ye fF ns L Vo s 2LVp 56% error o VS. ae, Z 19.5% Ibis se senn den zal for this problem the lumped mass solution for natural frequencies fae ep com lower values toward the upper-bound exact solution while the consistent converges from higher values toward the lower-bound exact solution. Both > ‘elutions give about the same level of accuracy. r tig nee fetus for the macural frequencies for the first six modes by the use of up ents with consistent mass formulation are presented in Fig. 7.4. 73 Ax IAL VIBRATION OF FOUR-D.O.F. T! BAR ELeMENres 0.F. TRUSS .f, truss bar elements in free-v ibration cement derivative or axial st ar element has four de; The acy curac analysis ea and efficiency of the two-d. degree or tra be improved by adding a displa OT freedom edOm (4u/ax) at each end. Thus the b: he as shown in Fig. 7.5. © axial di " xial displacement function may be assumed as a cubic polynomical u(x) = ay + ax + ayx? + aux" (7.201)206 Bar, Beam, and Plane Frame Finite Elemgn,, of Truss Bar, ; Free Vibration o4 ost ey erro OLE m = mode number 0.05 F 6 4s 6 7 8 9 10 a2 14! Number of elements Figure 7.4 Percenta vibration, Figure 7, Four-degree-of freedom axial force bar finite elementOO | vibration of Four-D.O.F, Truss Bar Elements ’ axial 207 -onstants are obtained by using the following boundary condition pur CO! ONditions fo u(O) = uy WL) = wy ) = (2) i ax Xx. au(L) _ () ax ~ \ax), on substitution of the solution of the four constants back into Eq. (7.21) a rearrangement of the terms, we can obtain the displacement function in and re the form u(x) = filed + 100(2) + lod + foo( 24) 2 where the four shape functions are in forthe beam element in Eq. (5.7). The stiffness matrix and consistent mass matrix for this element can siraightforwardly be derived by substituting the shape functions in Eq. (7.22) isto Eqs. (7.18) and (7.19), respectively, and performing the integrations. The ‘wilting equations of motion are in ine following form: (7.22) Precisely the same form as those obtained x = symmetric] { w, | 1 2L () Xi — | cal 10 15 ax), x, -& 1 6 uw > SL 10 SL ° x! 1 Lol 2 3) ° 10 30-1015 | (Max), 156 symmetric ty : (22) AL | 22h aL x71 (7.23) 420 | 54 13L 156 th ati | -13L -34? -22L 4? (%). Orsy “™boticany, Sus a (7,238 he {X} = [ku} + [rm }{ii) or 4 with the Qo strain fi 2 forces associated with “sof i. X% are the counterpart forces associated te edom, respectively 7208 Free Vibration of Truss Bar, Beam, and Plane Frame Finite Elomen, © of free vibration, the displacement vector may be with circular frequency w. E ay For the as sinusoidal function of time becomes a set of eigenvalue equations, {0} = [[k] ~ @ tm] Hu} . used to find the natural frequencies of a ¢, } bar in axial vibration. The results are given in Fig. 7.6 [7.4] on a semi Jon Heenan ee solution by Przemieniecki [7.5] using the two-degree-of j, bar elements but with quadratic equations of motion is also given in Przemieniccki’s solution is significantly more accurate and efficient the i an Equation (7.24) is _ Quadratic equations by Preemieniecki [7.5] 4-d.o.f, finite element X, Firth, mode xy 100 | om a mode a mode Number of equationsjoxural Vib" F 206 ven in Figs 7.4 by using the two-depree-of-fy edo a fhe ons of motions Howeve, ii seen in iy ree earamber of degrees oF Freedom, the solution hee spel ts is order-of- magnitude improved in ie ee two-degree-of-freedom elements with quadra nt fine en ‘ eleme? yom ele asc an as Compared tic equations vt nution PY HT tion jexURAL VIBRATION OF BEAM ELEMENTS. 4 ‘e717 shows a four-degree-of-freedom beam element for which the d fat funtion and stifiness matrix have becn formulated in Cha fer 5, sary to derive a consistent mass matrix for this element, wise nsistant ‘The displacement function is in the form v(x) = file) oy + fox) 8; + A(x) vy + xl) 6 (7.25) ster the shape functions are defined in Eq. (5.7). 10 Samiti Es Yat Figure 7.7. Four-degree-of-freedom beam finite element. The kinetic energy for the beam element in flexural vibration is A(t 72 T af u(x) de (20) an : ‘he strain energy expression for the beam element in bending is (7.27) U al v'(x) de 2 Jo "here ti ae Prime indicates differentiation with respect 0% i. enengy Sapna tituting the displacement function (7.29) YN” tbe Ba eae the gl” (7.26) and the strain energy expression (7.27) 4m Ue (7.15) with tespes * itferent; eagh netiations as indicated in the Lagrans equation °F the four degrees-of-freedom, we obtain (7.28) {f(t} [ma + [kall —— — 210 Free Vibration of Truss Bar, Boar, and Plane Frame Finite Eign,,, 100 Vibration 6 ts mass and stiffness terms are obtained in the sim, where the mt(7.18) and (7.19), respectively: given in i oe val Sco lx) dx 1 ky = EI J SICMSTO) ax ! o : The resulting equations of motion (7.28) in explicit form are as follows. 12 6 ee vi ra Lt of tL v% 6 6 G 4 -2 2 a M\ rl T L 1 Tz 12 6 12 6 Yo “a -p BF ~E||™ 6 6 M; a Boop 4 | be ' 156 22k 06 54 13] (& pAL| 22L 41? 13L —3L?}} 6, 420| 54 13L 156 ~22L/) 8, -13L -327 -22L 41? (6, Equation (7.31) is due to Archer [7.1]. For the case of free vibration with natural frequency @, Eq. (7)! (7.31) takes the following form: {0} = [[k] — o’LmII{a} « which is suited for eigenvalue solutions. ov ‘The equations of motion given in Eq. (7.32) were used by Archerit” 7.1 to perform free-vibration analysis for beam with free-free and 5% supported boundary conditions. Results of percentage errors iat frequencies obtained by using up to six elements are given in Table © solutions by Fowler [7.6] based on three different kinds of mass 18", explained in the footnote of Table 7.1 are also given for compatiso™ is case studied, the improvement in accuracy by using consistent mass fo" is dramatic. fo wt - |tion of Beam Elements groxural Vibra ™m ' percentage Errors in Natural Frequencies fy ABLE 7A Percentag tral Frequencies for Heams Consistent Mass Matrix UMped Mass Matriy : Free-Freet ; Simply - want : Supported A ) a, | mode | Free-Free | Supponed | eee c 11.00 ‘ 0390 ~38.07 : 11.00 hy 0.081 20.68 2367 563 1.182 31.55 11.00 [a 0.026 10.79 -1555 3.18 3 0395 11.60 32.17 ~11.04 \ 3 1.83 28.27 | 4 172 11.00 feel 0.048 oon 6.64 =10.0 -214 2 0.320 0.167 11.01 ~16.19 121 3 0.969 0.792 987 ~20.36 ~1558 4 0.995 1.021 26.58 3 11.28 1.00 af. 1 0.024 0.005 451 1.16 -1.82 v 2 0.170 0.082 7.63 12.16 3.13 3 0.575 0.394 9.29 = 15.64 ~10.45 4 1.262 1.183 455 19.17 16.97 5 1.150 2.66 ~2587 6 10.96 11.00 “A, mass lumped at center of gravity of the element; B, half the mass lumped at ends of ment C, same ox A with the inclusion of mass moment of Inea, Source: Ref. 7.4 a ample 9 : ee en i ‘lement to idealize half of the beam as shown in Fig. 7.8, find the © Natu a olution ae ral frequencies, Compare the results with the exact solut ay y ~ PN pA + 2,3 *++-is the mode number.wid Frou Vibration of Truss Bar, Boam, and Plane Frame Finity ¢, r First mode Second mode Third mode Fourth mode Figure 7.8 Simply supported beam and its first four mode shapes. Solution. For the first and third modes as shown in Fig. 7.8, we have the same bout! a 1 conditions as for the one-clement model, 1 = 62 =0 Thus we have a set of two equations of motion: 4 © {c}- EI E »pAL[ 4? 31) {oy \ 0 L}|_6 12] © azoLise 156] | lor t Let _ @(pAL*y 420BTwv m Elements, vipration of Bee . al oxul? rie + guay =oEEE det r et) OH IBA 12 156A ) L v 45547 — 828A + 12=0 A =0.01461 and 1.8052 for the first mode, 2.477 EL 9.9085, EI (0.39% o- Vga P Via .39% error) _géfor the third mode, _ 27.5 EL 110.1 ‘EL (24% o= Via 2 Via fo error) ‘The second and fourth modes have the same boundary conditions for the oneelement model, ose ave v, =2=0 Thus we have a set of two equations of motion: {ol Ela 2 ely 4? eT 0 TL2 4) ° 420L-3z? 4r7iile 4-4, 2+ 3a] 2+3A 4-4A «| 7a? — 44 +12=0 und A = 0.2857 and 6.0 HK “nee we have for the second mode, 109s [EL was [Et (11% error) w ay 'Wfor the fourth mode, 50.2 2008) wn 202 (HF 208) 11% for the (27.2% error) first- and second-mode frequencies, errors of 0.39% and e N =2. hy, Nebr are the same as those shown in Table 7.1 at amale 7.3 ad. A lumped mass with m = 4pAL/35 a ath ends fixed. A ump pAL/35 aida 22 Shows a beam with Boh ONT 12 F1/L! are attached at the midspan. “Massless elastic spring with Com"ar, Beam, and Plane Frame Fin), y Froo Vibration of Truss B 2 ‘0 tpAL Figure 79 Fixed-end beam with @ mass and a spring attached at nid, it is desired to find the approximate values of the first- and second:mede i, frequencies by using two beam elements to model the whole beam, The boundary conditions are i == 6-0 y= Thus we have a set of two equations of motion 24 {oh - Eilps | _ j2eAL[312_ 0 {>} Oo iL 420L 0 8L7J| le, 0 8 To include the effect of attached spring and lumped mass, we simply adi»# m to the stiffness term k,, and mass term m,, respectively. 36 (\-2 D °|_ gspAtf360 0 {2} oy |e 420L 0 sL7J| le 0 8 Let wal’) 420ET 36 LL = 10a) 0 0 BLII-A 288(1~ 10A)(1— A) = 0 a Land |

FEA CH 1-6 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

FEA CH 1-6 PDF

Uploaded by

Copyright:

Available Formats

You might also like