MATRIX STRUCTURAL ANALYSIS Ronald L. SackFor information about this book, contact Waveland Press, Inc 4180 IL Route 83. ite 101 Long Grove, IL 60047-9580 (847) 634-0081 [email protected] www. waveland.com Copyright © 1989 by Ronald L. Sack Reissued 1994 by Waveland Press, Ine 10-digit ISBN 0-88133-824-9 13-digit ISBN 978-0-88133-824-9 Alll rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitied in any form or by any means without permission in writing from ihe publisher. Printed in the United States of America 13.12 W 10 9CONTENTS SYMBOLS AND NOTATION INTRODUCTION 11 1.2 1.3 Historical Context of Matrix Structural Analysis Matrix Structural Analysis and Classical Methods Discussion THE STIFFNESS METHOD USING THE 2.41 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 BASIC EQUATIONS The Basic Equations The Element Stiffness Matrix Nodal Equilibrium of the Structure Nodal Displacements Coordinate Transformations Element Forces Settlement, Initial, and Thermal Strains Static and Kinematic Indeterminacy Summary Problems THE STIFFNESS METHOD USING VIRTUAL WORK 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.9 The Principle of Virtual Work 3.1.1 Virtual Displacements for Particles and Rigid Bodies 3.1.2 Virtual Work for Deformable Bodies 3.1.3 Real, Actual, and Admissible Displacements 3.1.4 Discussion The Element Stiffness Matrix 3.2.1 Shape Functions 3.2.2 The Axial Force Element Coordinate Transformations The Structural Stiffness Matrix Nodal Displacements Element Forces Initial and Thermal Strains Comparison of Direct and Virtual Work Methods Problems 24 34 38 42 52 55 63 75 76 76 81 87 89 95 95 99 102 107 110 11 112 114 115 viiviii CONTENTS. THE STIFFNESS METHOD FOR BEAMS AND PLANAR FRAMES 4.1. The Prismatic Beam Element 4.1.1. The Basic Equations and 4.1.2 The Principle of Virtual Work and 4.1.3 Castigliano's Theorem (part I) and k 4.1.4 Element Forces 4.2 Nodal Equilibrium, Nodal Displacements, and Structural idealization 4.3 Equivalent Nodal Forces 4.3.1 Fixed-End Forces 4.3.2 Consistent Forces 4.4. Settlement, Initial, and Thermal Strains 4.5 The Prismatic Plane Frame Element 4.5.1 Flexure and Axial Deformations 4.5.2 Flexural and Axial Rigicity 4.5.3. Settloment, Initial, and Thermal Stains 4.6 Discussion 4.7. Problems THE STIFFNESS METHOD FOR THREE-DIMENSIONAL STRUCTURES 5.1 Space Trusses 5.2 The Three-Dimensional Bisymmetric Beam Element 5.2.1 Axial and Bending Effects 8.2.2 Torsion 5.2.3. The Element Stiffness Matix 5.3 Coordinate Transformations in Three-Dimensional Space 5.4 Grid Structures 5.5 Space Frames 5.6 Discussion 5.7 Problems SPECIAL TOPICS FOR THE STIFFNESS METHOD 6.1 Matrix Condensation 6.1.1 Release of Generalized Element End Forces 6.1.2. Substructural Analysis 6.2 Coordinate Transformations 6.2.1 Constraint Equations 6.2.2. Nodal Coordinates 6.2.3 Offset Nodes 121 122 123 127 129 131 132 142 143 148 150 157 157 162 166 168 169 178 179 186 187 188 191 193 196 201 203 204 212 213 215 219 223 224 227 233CONTENTS 6.3 6.4 Special-Purpose Elements 6.3.1 The Tapered Axial Force Element 6.3.2 The Tapered Beam Problems THE FLEXIBILITY METHOD 7a Td 7.3 74 rs 7.6 Te 7.8 7.9 The Flexibility Method Using the Basic Equations The Flexibility Method for Trusses The Flexibility Method for Beams The Principle of Complementary Virtual Work The Flexibility Method Using Complementary Virtual Work Relationships Between Statics and Kinematics The Element Flexibility-Stiffness Transformation Initial Strains, Thermal Strains, and Distributed Loads Automatic Selection of Redundants 7.10 Discussion 7.11 Problems APPENDIXES SOLUTION OF THE EQUILIBRIUM EQUATIONS. A.1 Direct Solutions Based on Gauss Elimination A.1.1 Solution by Elimination A.1.2. Solution by Decomposition A.1.3 Computer Implementation: A.1.4 Choleski Decomposition A.1.8 Frontal Solution A.2 Solution Accuracy A.2.1 Numerical Accuracy MATRICES. B.1 Definitions B.2 Equality B.3. Addition and Subtraction B.4 Multiplication B.5 Transpose B.6 Special Matrices B.7 Submatrices and Partitioning B.8 Determinants B.9 Simultaneous Linear Algebraic Equations B.10 Differentiation 240 240 245 251 251 254 257 260 266 268 272 273 276 281 282 285 285 286 289 293 296 296 298 299 302 302 302 302 303 305 305 305 306 308 310a CONTENTS B.11 Integration B.12. Bilinear and Quadratic Forms GENERAL REFERENCES. SYSTEMS OF MEASUREMENT ANSWERS TO SELECTED PROBLEMS INDEX 310 311 312 313 315 323SYMBOLS AND NOTATION Symbols are generally defined where they first appear. Some symbols have been used in different contexts to define several quantities. In general we have used lower-case symbols to indicate quantities associated with element coor- dinates and capital letters for global quantities. We use p (P) to indicate nodal applied forces (both loads and reactions) and u (U) to denote nodal displace- ments; k (K) contains the stiffness elements and f (F) denotes the flexibility matrix. Matrices are shown in bold print, and the elements of a column matrix are written within brackets, { }, to conserve space in the text. a Kinematics matrix (partitioned into a, and a,); matrix of coeffi- cients for polynomial A Cross-sectional area of a member b) Member width b Statics matrix (partitioned into by and b,) B_ Matrix relating nodal displacements to element strains d Column matrix of element deformations (d = aU) d’ Column matrix of initial element deformations e Element force transformation matrix for global displacements & Element force transformation matrix for local displacements E Modulus of elasticity (i.e., Young’s modulus) E Matrix of elastic constants f Element flexibility matrix F Global flexibility matrix G Modulus of elasticity in shear 1 Moment of inertia I Identity (unit) matrix J St. Venant’s torsion constant k Element stiffness matrix with elements k,, expressed in global co- ordinates k Element stiffness matrix with elements K,, expressed in local co- ordinates K Structural stiffness matrix with elements K, expressed in global coordinates L Length M_ Bending moment NDOF, NE, NN, Number of: degrees of freedom; elements; nodes; NR, NOK, NOS reactions; kinematic indeterminacies; static indeterminacies N_ Element axial forcexii SYMBOLS AND NOTATION N oO P D pe? P, P. pe q R s Fr T U.V.W, 4u,v,w Uy Vg We Wy VW; u u U Column matrix of shape functions Null matrix Column matrix of nodal element forces in global coordinates Column matrix of nodal element forces in local coordinates Column matrix of initial nodal element forces in global coordi- nates Column matrix of initial nodal element forces in local coordinates Column matrix of applied nodal forces in global coordinates (s = bP = b,P + b,X) Column matrix of known applied nodal forces in global coordi- nates Column matrix of unknown applied nodal forces in global coor- dinates Column matrix of initial forces in global coordinates Distributed load magnitude Column matrix of reaction forces Column matrix of element forces and reactions Temperature Transformation matrix Displacements at node i for a structure in the x, y, and z diree- tions, respectively Continuous functions expressing displacements in the x, y, and z directions, respectively Displacements at node i for an element in the x, y, and z diree- tions, respectively Displacements at node i for an element in the X, ¥, and Z direc- tions, respectively Column matrix of nodal displacements in global coordinates Column matrix of nodal displacements in local coordinates Column matrix of nodal displacements for the entire structure in global coordinates Column matrix of unknown displacements in global coordinates Column matrix of known displacements in global coordinates Shear force Work and complementary work done by external forces Strain energy and complementary strain energy Orthogonal cartesian global (structural) coordinates Orthogonal cartesian local coordinates Column matrix of redundant forcesSYMBOLS AND NOTATION xiii SUBSCRIPTS SUPERSCRIPTS i c +1 GREEK SYMBOLS @ (alpha) a8, y (gamma) I (gamma) 6 (delta) A (delta € (epsilon) 6 (theta) © (theta) « (kappa) x, (kappa) Danae v (nu) o (sigma) E (sigma) The node (point) associated with the quantity Degrees of freedom with known forces and unknown displace- ments Degrees of freedom with unknown forces and known displace- ments Quantity associated with the primary structure; used in the flexi- bility method Quantity associated with the structural redundants; used in the flexibility method Force (or moment) that is equivalent in an energy sense to a dis- tributed loading Force (or moment) required to give zero displacement at the point (i.e., a fixed-end force) The interval (element) associated with the quantity Quantity initially introduced by temperature, fabrication error, precambering, etc. Transpose of a matrix Inverse of a matrix Coefficient of linear thermal expansion Angles measure to a vector from the positive x, y, and z axes, respectively Shear strain Matrix of direction cosines; an orthogonal transformation Deflection; increment of a quantity; first variation of a quantity (a virtual quantity) Deflection; total change in a quantity ‘Translational strain Angle; rotation of node with respect to local coordinates Angle; rotation of node with respect to global coordinates Curvature of a beam Shear constant Direction cosines (i.¢., cos e, cos 8, cos y, respectively) Poisson’s ratio Normal stress Summation of quantitiesxiv SYMBOLS AND NOTATION 7 (tau) Shear stress ¢ (phi) Angle between axis of an element and the global coordinates GRAPHIC SYMBOLS Force 1 | S Moment Resultant force Resultant moment Reactive force Reactive moment YD PTED or Roller support Pinned support Fixed support Rigid connection Pinned connection SNTS 4MATRIX STRUCTURAL ANALYSIS1.1 INTRODUCTION Engineered structures must ensure the safety and welfare of the occupants and general public by performing in a prescribed manner. Strength requirements are accompanied by stiffness constraints to prevent excessive deflections, bouncy floors, outward-tilting walls, uncomfortable structural oscillations, and the like. Thus structural analysis and design are intertwined since behavior is affected by the arrangement of members and distribution of materials. New complex systems require more precise engineering; many major contemporary structures, such as the Boeing 747 aircraft, the Swiss Flesenau Bridge, and the Sears Tower, owe their existence to computer-oriented structural analysis and design. This chapter examines the origins and utility of matrix structural analysis, and its relationship to classical methods. HISTORICAL CONTEXT OF MATRIX STRUCTURAL ANALYSIS The airplane and digital computer are responsible for revolutionizing structur- al analysis. In the 1940s and 1950s structural engineers were confronted with two highly statically indeterminate systems: the swept-wing and delta-wing aircraft. The governing equations were cast ab initio (from the beginning) in matrix format, but this approach required solution of large sets of simulta- neous linear algebraic equations. At the time, relaxation methods were used extensively to solve the governing equations of structural behavior; therefore, the requirement to deal with great numbers of algebraic equations was an anathema to the engineer. Fortuitously, the University of Pennsylvania unveiled the 30-ton ENIAC digital computer in 1946. The invention of the transistor in 1947 and the silicon chip in 1959 were pivotal discoveries that accelerated the development of the digital computer and gave impetus to the structural analysis revolution. By embracing this new computing technology, the structural engineers of those two decades completely changed structural analysis. Trusses, beams, and frames were initially investigated, but in the mid- 1950s a group at the Boeing Company demonstrated that the procedure could be extended to continua. Common usage now dictates that matrix structural analysis designates investigations of structures composed of articulated or dis- crete components, whereas the finite element method denotes analysis of con- tinua. Structural analysis of the carly swept-wing aircraft in 1947 depended upon work from the 1800s by James Clerk Maxwell and Otto Mohr. Their 1Chapter 1 INTRODUCTION method of consistent displacements is an example of a classical compatibility method yielding sets of simultaneous linear algebraic equations, with the struc- tural flexibilities as the coefficients and the forces as the unknowns. A new method, called the flexibility or force method, was formulated and distin- guished from traditional compatibility procedures by the fact that all quan- tities and equations were formulated initially as matrices and manipulated using the associated algebra; therefore, the operations are computer oriented. In 1953 the delta-wing aircraft was the impetus for a second computer- oriented approach to structural analysis. By broadening the scope of tradi- tional equilibrium methods and formulating the equations from inception using matrices, the structural engineers of the 1950s obtained a set of linear algebraic equations with the structural stiffnesses as the coefficients and the displace- meats as the unknowns. Thus the stiffness or displacement method was con- ceived, Interest in energy methods was also stimulated during this time, but structural mechanics has historically relied upon energy principles. Archi- medes (287-212 8.c), Leonardo da Vinci (1452-1519), and Galileo (1564-1642) each used some form of the work expression to substitute for the equations of equilibrium in lever and pulley systems. Johann Bernoulli (1717) was the first to suggest virtual displacement, and Maupertuis (1740) introduced the concept of measuring equilibrium of rigid bodies by minimizing the total system poten- tial. Leonhard Euler (1744) recognized that energy methods are an alternative approach for solving problems of structural mechanics and uscd minimization principles to investigate stable equilibrium for deformable bodies; he used expressions for strain energy suggested by Daniel Bernoulli. Lame (1852) derived the principle of conservation of energy and named it for his friend Clapeyron; he used actual forces, stresses, displacements, and strains. James Clerk Maxwell (1864) and Otto Mohr (1874) independently took the results of Lame, and, using a dummy load, investigated statically indeterminate trusses. Thus, the principle of virtual forces is also known as the Maxwell- Mohr method. Castigliano (1873) published the extremum version of Lame’s work. Since Lame used actual quantities, Castigliano’s theorem, part II, is valid only for linear elastic systems. Crotti (1878) and Engesser (1889) subse- quently extended this result, thereby making the minimization principle conform to the principle of complementary virtual work for nonlinear clastic systems. In 1954 J. Argyris and S. Kelsey formulated matrix structural analysis using energy principles. Matrix structural analysis emanated from physically directed thinking and was derived by satisfying the fundamental equations of structural mechanics; therefore, the application of energy principles was the next logical step in the evolution. Because of the initial popularity of the flex- ibility method during the 1950s, the corresponding principle of complementary virtual work was emphasized. In contrast, the stiffness method arises from the principle of virtual work, but early derivations represented this simply as an alternate choice of variables (ic., displacements as unknowns instead of forces). Subsequent work revealed that the stiffness method, based upon the principle of virtual work, is a numerically efficient procedure for implementing the clas-1.2 Section 1.2. MATRIX STRUCTURAL ANALYSIS AND CLASSICAL METHODS 3 sical Rayleigh-Ritz method, which was conceived in 1909. The finite element method owes its existence and wide appeal to this fact. MATRIX STRUCTURAL ANALYSIS AND CLASSICAL METHODS Thus matrix structural analysis has come to fruition since the 1940s, but its roots are in classical structural mechanics. Since the computer formulates and solves the equations, large structures can be investigated. We can use either compatibility or equilibrium methods and formulate the method using the fun- damental equations of structural mechanics or energy principles. Therefore, it is instructive to recall the basic principles and classical methods of structural analysis and observe their relationship to matrix structural analysis. ‘Structures must be in equilibrium, with their displacements in a compat- ible state and material laws satisfied. The structural engineer can investigate these primary behavioral tenets by either: (a) solving the fundamental equa- tions or (b) employing energy principles. Double integration, the method of elastic weights, and the moment-area method yield structural displacements from the fundamental equations. We use force-displacement relationships to assemble equations of structural response, thereby satisfying equilibrium, compatibility, and material laws. Compatibility methods mandate identifying statically indeterminate elements and imposing compatibility requirements, thus producing sets of equations with the structural flexibilities as coefficients and forces as unknowns, The method of consistent displacements and the three-moment equation are exam- ples of the compatibility method. Alternatively, by invoking equilibrium at points connecting structural ele- ments we can formulate sets of simultaneous linear algebraic equations with the structural stiffnesses as coefficients and displacements as unknowns. This approach begets equilibrium methods; the slope-deflection and the moment- distribution methods are two classical procedures in this category. Energy principles present an alternative approach for investigating struc- tural behavior. The principle of virtual displacements, the unit displacement theorem, and Castigliano’s theorem, part I, are examples of energy methods, wherein equilibrium is satisfied implicitly. In contrast, the principle of virtual forces, the unit-load theorem, and Castigliano’s theorem, part II, are comple- mentary virtual work theorems that satisfy compatibility implicitly. Virtual work theorems call for varying the displacement and correspond- ing strains, whereas complementary virtual work theorems require the forces and stresses to undergo variations. The former approach produces equilibrium methods, while the latter gives rise to compatibility procedures. For example, recall the method of least work for linearly elastic systems. By Castigliano’s theorem, part II, the partial derivative of the strain energy with respect to a force gives the corresponding displacement. If that displacement is zero (e.g., for a redundant reaction), we obtain what appears to be a minimum principle.Chapter 1. INTRODUCTION This, of course, is a compatibility method obtained from a complementary virtual work principle. For systems with many unknowns it is convenient to formulate the equa- tions from the beginning in matrix form; thus, subsequent manipulations are executed using matrix operations, which can be conveniently programmed in a computer language. If we use a compatibility method (either by solving the fundamental equations or invoking a complementary virtual work principle) we obtain sets of simultaneous linear algebraic equations involving structural flexibilities. In contrast, by formulating the solution using either the principle of virtual work or the fundamental equations of structural behavior, we mandate nodal equilibrium and obtain sets of simultaneous linear algebraic equations embodying the structural stiffnesses. The former approach is the flexibility or force method, whereas the latter is the stiffness or displacement method. We can program the force method to automatically identify statically indeterminate components or systems. We formulate the coefficient matrix of system flexibilities using a matrix triple product; one of the basic matrices required for this process expresses system equilibrium, while another simply contains element flexibilities. The computer executes a large number of oper- ations and consumes a great amount of time in formulating the global flex- ibility matrix. In the formative stages, the stifiness method suffered from engineers obsessed with the duality of the two methods. That is, since the basic equilibrium matrix of the flexibility method can be shown to define the system compatibility equations, we can formulate the global stiffness matrix using a matrix triple product in a fashion resembling that employed in constructing the global flexibility matrix. Equations can elegantly express this duality, but the approach is computationally inefficient. The global stiffness matrix is most efficiently formulated using list processing. By clinging to the duality of equa- tions, the carly pioneers nearly rang the death knell for the stiffness method. We now recognize duality for what it is: an interesting fact with few useful computational implications. Today, the stiffness method has almost totally supplanted the flexibility method. DISCUSSION Perhaps in a few years we will not be required to distinguish computer- oriented structural analysis by appending the adjective, matrix. This specula- tion is strengthened by the capabilities of the available computing hardware and software. In the early days of computers, the user was required to prepare punched cards, learn elaborate access protocol, and struggle with “turn- around time”; the microcomputer has eliminated all of this. In addition, many classical methods can be implemented on a standard spreadsheet, thus quali- fying them to be called computer methods. Larger problems simply require more computing power. The supercomputer offers solutions for mammoth systems, whereas intermediate-sized problems can be solved using some com-Section 1.3 DISCUSSION 5 ponent of the array of available equipment between the micro- and super- computer such as the mini- or mainframe computer. ‘The spectrum of approaches to structural analysis includes classical, approximate, and computer-oriented methods, and each has its function. By interpreting computer solutions using approximate analysis, the structural engineer can avoid the computer siren that lures acceptance of dubious machine-generated results with the implication that computer output is above question. Since nodes can be misplaced, members inadvertently omitted, and entire systems incorrectly modeled, it is wise to remember the old computer maxim: “garbage in, garbage out.”2 2.1 THE STIFFNESS METHOD USING THE BASIC EQUATIONS Most of today’s computer programs for analyzing structural response are based upon the principles of the stiffness method. In this chapter we will for- mulate this useful method using the basic equations: equilibrium, compat- ibility, and constitutive (material property relations). These equations must be satisfied within each individual element and throughout the entire structure. The solution of the equations given by the stiffness method yields the displace- ments of each prescribed point (node), and the internal element forces are sub- sequently obtained by backsubstitution. The objective of this chapter is to outline the stiffness method and apply it to trusses. We will initially illustrate the method with axial force elements in series and be able to analyze most types of two-dimensional trusses by the end of the chapter. The approach developed in this chapter is extended in Chs. 4 and 5 to the analysis of two- and three-dimensional structures composed of axial force and beam elements. Throughout this book we will study only articulated structures composed of discrete individual elements. THE BASIC EQUATIONS A truss is composed of easily identified individual discrete members, termed elements, reaching from joint to joint. The basic equations of structural mechan- ics must be satisfied within each of the elements. In addition, the truss, com- posed of the assemblage of all the elements, must also respond in accordance with the basic equations. We can review the basic equations in their most simple form by examining the element in Fig. 2.1a. The element is prismatic with a cross-sectional area A, and it is subjected to forces that are colinear with its centroidal axis. When the forces are applied, the element remains straight, and plane sections normal to the member axis remain plane. We shall assume that the element undergoes sufficiently small displacements so that linear strain-displacement relations are valid, and the equilibrium equations relate to the undeformed state. Also, the element experiences small strains and is made from a material that is homogeneous, isotropic, and linearly elastic; thus, Hooke’s law is applicable. The element is in static equilibrium; hence, the axial forces at each end are equal in magnitude but opposite in direction (we will assume the memberSection 2.1 THE BASIC EQUATIONS 7 (a) p Pp p< oA (b) aw wt OH be Hoy, ©) Figure 2.1 (a) Axial force element; (b) element segment; (c) displacement of an element to be weightless throughout this discussion). The stress, o,, is constant over the cross section, and equilibrium must also be satisfied for any portion of the element. We observe from Fig 2.1b that o.A=p ay The displacement along the element, u, is a continuous function of x. This ensures that the displacements are geometrically compatible; that is, no rips or tears are introduced within the element by the displacement. The original small length, Ax, moves from points 0 and a to o’ and a’, respectively (sce Fig. 2.1c). Using the definition for strain gives the following strain- displacement equation: ig Hew SO fy LE Guten] — Aad 22) ee a Ax ox The constitutive equation (material behavior relationship) in this case is Hooke’s law; that is, o, = Ee, (23) where E is the modulus of elasticity. Combining Eqs. (2.1) through (2.3), and noting that u is a function of x only, yields the following relationship between the force and displacement: du = (24) Rearranging terms gives pdx = AE du (2.4a)8 Chapter 2. THE STIFFNESS METHOD USING THE BASIC EQUATIONS Integrating Eq. (2.4a) yields the force-displacement equation for this axial force element; that is, x 2 -4 +c, (25) where C, is a constant of integration. If the member is constrained at the left end, u(0) = 0, C, = 0; furthermore, at x = L, u = AL, which yields AE p= AL (26) where AL is the total deformation of the clement. If the element is incorpo- rated into a truss and both ends are displaced so that u(0) = u; and WL) = uj, then C, = u,, giving the following force-displacement relationship: AE P= (uj—u) (2.6a) Note that (uj — u) is the total deformation of the element, which is equal to AL in Eq. (2.6). Thus, by invoking the conditions of equilibrium, compatibility, and stress-strain we have obtained the familiar relationship between force and displacement for an axial force element. 2.2 THE ELEMENT STIFFNESS MATRIX The axial force element in Fig. 2.2 can deflect in either of the two coordinate directions and rotate about an axis normal to the plane of the paper. Throughout this book a deflection indicates the translation of a point on a deformable body, whereas displacement is used as a more general term to describe both translations and rotations. Since the element deformation (ie., the relative displacement of points on the element) occurs only in the axial direction, the rotation of the element can be calculated in terms of the deflec- ae (2.11) Second Element Displacement Case: 0», #0; End i of the axial force element is given a displacement in the positive y direc- tion while the other three displacements at the two reference points are not allowed to displace. This configuration is shown in Fig. 2.3b. Since the defor- mation along the element axis is equal to v, sin ¢, from Eq. (26) the axial compressive force at end iis pe aE v; sind (2.12) The x and y components of the force are 2 1? cos d = v, sin d cos 6 (2.13) AE and po? =p? sin g v, sin? @ (2.14) Equilibrium of the element mandates that AE £8) = —p2 = — “2 asin 6 cos 6 (2.15) AE and PS) = p= —T- vy sin? 2.16) Third Element Displacement Case: 440; u This deflected configuration of the element will result in an elongation equal to u, cos p. By invoking equilibrium and using Eq. (2.6), similar to what was done in the previous two displacement cases, the forces at the two ends are 3) p= 2.17) and PP = — (2.18)12 Chapter 2. THE STIFFNESS METHOD USING THE BASIC EQUATIONS Fourth Element Displacement Case: 440; u=H= For this last individual case, point j is given a deflection in the y direction while all other deflection components are constrained. Using the equations of equilibrium, along with Eq, (2.6), gives the constraining forces: (2.19) |E and ap =~ = 4 o, sin® g (220) General Element Displacement Case: 4 #0; #0; u, #0; #0 We are investigating a linear elastic system undergoing small displacements; therefore, the principle of superposition is valid. Thus if we simultaneously give all four of the displacement components nonzero values, the total con- straining forces are the sum of the forces from the four individual displacement cases; that is, Pai = PS + PD + PS + Po? = “ [lcos? 4)u; + (sin $ cos 4)v, — (cos? 4)u, — (sin $ cos xj] Pyi = Py + PS? + PW? + PS? = “= [(sin @ cos )u; + (sin? $v; — (sin d cos pu; — (sin? o)v,] ee es = BE) + + +P (220) 2 = [—(Cos? pu; — (sin $ cos g)v; + (cos? plu + (sin $ cos 9)e,] Byj = Py + yy + Py! + Py? = “ [—Gin ¢ cos du, — (sin? G)v; + (sin @ cos du, + (sin? #)v,] ‘These equations can be expressed in matrix form as Pri cos? ¢ sin@cos$ —cos'd —sing cos o|fu, P| AE] sin d cos sin? @ -singcos@ sin?’ d |] Pay —cos?@ sin # cos cos? $ sin $ cos @ || u, Py sin pcos — —sin? 6 sin cos g sin? g Lv, Pp k u or (2.22) p=ku (2.22a)Section 2.2 THE ELEMENT STIFFNESS MATRIX 13 Matrix equation (2.22) describes the relationship between the forces and the displacements at the ends of an individual axial force element that is orien- tated in a general position in two-dimensional space. The forces appear in the same position in the force matrix (p) as their associated displacements occur in the displacement matrix (u). The matrix k is the element stiffness matrix, and it has a number of unique characteristics that should be noted. A particular column of k contains the forces associated with a specific displaced configuration. For example, compare the second column of k and the forces computed in the second dis- placement case described previously. The rules of matrix multiplication mandate this; that is, if we set » # 0, with u,—u, =v, =0, the product ku gives the force matrix p that contains the forces calculated in the second dis- placement case. A specific column in k represents the set of equilibrating forces necessary to maintain a specific deformed shape; hence the sum of each column is zero. This is a characteristic of the axial force member, but it is not inherent in all element stiffness matrices. Note also that k is symmetri z kj = kj): a fact that can be demonstrated using the theorem of virtual work (see Sec. 3.2). The reader may wish to defer investigation of this property in detail until reading Ch. 3. Finally, we note that in general k has no inverse. That is, there is no unique set of displacements for an arbitrary p (unless these forces happen to be in equilibrium). This fact is easily explained by noting the physical problem. No constraints have been specified for the element. There- fore, it is possible for the member to display rigid-body displacements in addi- tion to deformations, and there are no unique displacements for a given set of applied forces. This problem of singularity with respect to solution is alleviat- ed when the element is adequately supported against rigid-body motion (two perpendicular translations and a rotation in two dimensions), or several members are combined into an assemblage that is constrained against rigid- body motion. This point is discussed in Sec. 2.4. Example 2.1 demonstrates the use of Eq. (2.22) for an individual axial force element that is adequately constrained and loaded at one node. EXAMPLE 2.1 Calculate the deflections and forces for the single strut in Fig. E2.1 using the stiffness matrix for the axial force member [Eq. (2.22)]; 4 = 1600 mm* and E = 200 GPa. The support symbol at point a can resist horizontal and vertical forces, while that at point b can resist horizontal forces (see “Graphic Symbols” in “Symbols and Notation’. Solution AE _ 1600 x 107°(200 x 10° AE, _ 600% 10 C0020) _. 6102 B 4 cos @ = 0,500 sin @ = 0.86614 Chapter 2. THE STIFFNESS METHOD USING THE BASIC EQUATIONS ae bo Figure E2.1 From Eq. (2.22), Pee 0.250 0.433 —0.250 —0.433 ][u, Pya | ,| 0433 0.750 —0433 —0.750]} », pa {80% | 0.250 0433 0.250 0.433 | uy Dy —0.433 —0.750 0.433 0.750] vs. But u, = v4 = uy = 0 and py, = —100 KN. From the fourth equation, —100 = 80 x 10%(0.750r,) by = — 1.67 x 107? m and the first three equations give Pra = 80 x 103(—0.4330,) = +57.7 kN Pya = 80 x 10°(—0.750%) = +100.0 kN Pay = 80 x 10°(0.433y,) = —57.7 KN 2.3. NODAL EQUILIBRIUM OF THE STRUCTURE The basic equations of structural mechanics must be satisfied throughout the structure. In the case of a truss, this is done within each individual element through Eq. (2.6a), but since a truss is an assemblage of several elements, inter- clement compatibility and equilibrium must be imposed. We will investigate how this is done by first examining two axial force elements connected in series. The principles involved with this problem will be extended to a two- dimensional truss to study the general form of the procedure.Section 2.3 NODAL EQUILIBRIUM OF THE STRUCTURE 15 Pw. U; Pras Ua Pur Uy oo a b A. Ly. Ey Figure 2.4 The assemblage with two axial force elements The assemblage in Fig. 2.4 is composed of two axial force elements as described in the previous section [note that (A, £,/L,) = k, for convenience]. This simple structure is constrained against displacements transverse to the element axes; these restraints are not depicted in Fig. 2.4 for clarity. The relationships between the forces P,4, Px», and P,, and the corresponding dis- placements U,,, U,,, and U,, will first be obtained by superposing the forces from three independent displacement cases, wherein each point is selectively given a nonzero displacement while the other points are constrained. This approach is similar to the procedure used for the single element in Sec. 2.2. Note that capital letters are used to designate forces and displacements at the points of the assemblage. This distinguishes these global values from the lower-case quantities for an individual element (ic. the notation used in Sec. 2.2), First Assemblage Displacement Case: U,#0; Uy Only element ab will be deformed; therefore, using Eq. (2.6a) we calculate the forces on the ends of the element. These are shown in Fig. 2.5a, along with the equilibrating forces exerted on the nodes according to Newton’s third law. The superscripts in parentheses on the forces denote the displacement case number. Figure 2.5 Two-element assemblage showing equilibrating forces for nodes and elements for three displacement cases: (a) U, #0; (b) U, £0; (c)U, 4016 Chapter 2. THE STIFFNESS METHOD USING THE BASIC EQUATIONS Second Assemblage Displacement Case: U,#0; U,=U,=0 Since both elements are attached to node b, continuity requires that deforma- tions and hence forces occur in both elements. By invoking Eq. (2.6a) we obtain the forces shown on the elements and nodes in Fig. 2.5b. Third Assemblage Displacement Case: U.#0; U,=Us 0 This displaced state is similar to that in the first case in that only one element has been deformed. Applying the appropriate governing equation gives the forces shown in Fig. 2.5c. General Assemblage Displacement Case: U,#0; U,#0; U, #0 If nonzero values of U,, U,, and U, are imposed simultaneously, the relation- ships between these displacements and the corresponding forces are obtained by superposing the results from the three independent displacement cases. We must observe the positive coordinate directions shown in Fig. 24, and note that since the elements are connected at node b, the total force at that node is obtained by adding the forces from each individual element. This gives Pay = PO) + PQ + PO) = kU, — kU, + OU, Py, = PY) + PQ) + PQ) = kU, + (ki thaW,— kU. (2.23) P= PC) + PO) + PO) =0U, — kaU, + ba U. In matrix form these force-displacement equations become Pee: ky —k 0} U, Payl=|—ki ki tk, kal] Uy (2.24) Po 0 —k, k, LU, \ K U or P=KU (2.25) where P = column matrix of nodal forces U = column matrix of nodal displacements K = stiffness matrix of the assemblage (also referred to as the structural stiffness matrix) A perusal of Eq. (2.24) reveals that a given column of the stiffness matrix represents the equilibrating forces for a prescribed displaced shape of the assemblage (i.¢., a situation that is analogous to what we observed for the single element in Sec. 2.2). Thus, K, is the force P,, due to a unit displace-Section 2.3 NODAL EQUILIBRIUM OF THE STRUCTURE 17 ment of U, (with U, = U, = 0), K,2 is the force P,, due to a unit displacement of U, (with U, = U, =0), etc. Since all forces act in the x direction, a given column must sum to zero for assemblage equilibrium. This suggests that the stiffness matrix can be generated directly from the forces required to maintain equilibrium by imposing selective displacements at each node. Before proceeding, Eq. (2.22) must be modified for the special case in which the elements are aligned with the x axis. In this case ¢ = 0° (see Fig, 2.2). There can be no displacement normal to the element axis because there is no resistance to force (ie., stiffness) in that direction. Thus, by ignoring the degrees of freedom in the y direction (ic., py;, Pyj, U1, and v, are all zero in this case and all reference to these has been omitted), we have pa] EL 1 1 els oT] 5 P k u Thus, for the case of elements all aligned with the x axis, k is the 2 by 2 matrix shown in Eq. (2.26). Returning to a discussion of the assemblage in Fig. 2.4, we note that Eq. (2.23) shows the equations of nodal equilibrium for the assemblage, and these results indicate a convenient method for assembling the structural stiffness matrix K. Nodal equilibrium is assured by summing the stiffness coef- ficients from the individual elements that are connected to a given node (this observation will be verified subsequently for more general trusses). This pro- cedure is indicated symbolically in Fig. 2.6, and we note the process can be envisaged as a simple overlay of the individual element stiffness matrices [with the form shown in Eq. (2.26)]. Thus, we obtain K by starting with a null square matrix, with a size equal to the number of degrees of freedom for the problem, and placing the individual element stiffness matrices, k, in the rows =k Figure 2.6 Assembling the global stiffness matrix for two axial force elements in series (see Fig. 2.4)18 Chapter 2. THE STIFFNESS METHOD USING THE BASIC EQUATIONS and columns to which they relate. Wherever the k’s overlap, addition of the matrix elements is implied, and if no entries from any k are overlaid in a particular location of K, the entry remains zero. It is important to note that this is not matrix addition in the conventional sense, but it is sometimes referred to as submatrix addition (ie., K is the matrix into which the sub- matrices, k, are added). The combination of the individual element stiffness matrices into the structural stiffness matrix for the entire structure is referred to as the assembly or merge process. Thus, the assembly process is a convenient method for obtaining K which obviates having to calculate numerous equi- librium configurations for the assemblage. Numerical results for a two-clement assemblage are presented in Example 2.2. EXAMPLE 2.2 Obtain the structural stiffness matrix for the two-element assemblage in Fig E22 (displacements are possible only in the x direction); k, = (A,E,V/L, = 30,000KN/m and ky = (42 E3)/L2 = 15,000 kN/m. P, — Py —— —_{—a—a<<&_{_ooooooo > a i > ki ¢ Figure E2.2 Solution The following individual element stiffness matrices are obtained by substitut- ing into Eq. (2.22): where the superscripts on the k’s designate the element. By starting with a null 3 by 3 matrix and adding k*” and k® using the submatrix procedure implied in Fig. 2.6, we have z +2 0 K = 15,000] —2 3-1 0 -1 1Section 2.3 NODAL EQUILIBRIUM OF THE STRUCTURE 19 Node n Node / oo © Figure 2.7 (a) Truss showing key nodes; (b) element connectivity at node i; (c) free-body diagram of node i (for clarity only the x force components are shown) Equilibrium must be enforced for each independent force component that exists at every nodal point of a structure. The truss in Fig. 2.7a has two degrees of freedom per node, and this requires that we use k as shown in Eq, (2.22). Consider the elements connected to node i in Fig. 2.7b. The force- displacement equations for each of the three axial force elements attached to20 Chapter 2. THE STIFFNESS METHOD USING THE BASIC EQUATIONS this node can be expressed using Eq. (2.22) as follows: pe) [RE kia Mls ke |[ wr Dir) | Ri: Kis Wis Khe |] oy os (2.27: Bi | ah Res Rha |) Geis) pi} Lk, kis ks kta JLo, wok) [ARE eh en TL] Pe pps |_| Ast eh ea eh | on = (2.27b) pe | [eg eh deh ak |, G21) pot] Lesh ahha oe] Pet ki ki kta |fus pin] | mS eke | chal ge ae ee 2.216) pi) |S, ks, ksh || oy pin) Lat ka kis Ree where the stiffness entries are those from Eq. (2.22); the subscripts denote the position of the entry in the matrix, and the superscripts indicate the end nodes of the element. For example, the element designated by in has the local axes directed from node i to node n with the corresponding implied counter- clockwise angle ¢ (see Fig. 2.2). No superscripts are required for the displace- ments since for compatibility all elements must be connected to a common node. For example, u(! = uy" = ui" =u; =U; (since the nodal displacement refers to the global coordinates). ‘The following describes equilibrium in the x direction only; a similar argument could be made for the y direction. The free-body diagrams for the elements connected to node i are shown in Fig. 2.7c. Only the x forces are shown for clarity. The force applied externally to the node in the x direction is P,;. The equation of equilibrium in the x direction gives Pat = Das + Pui + PS Upon substituting from Eqs. (2.27a) through (2.27c), the equation becomes Pag = (ky + kT + RYU, + (Sg + ABS + RP) + AS Uy + kV + KS Uy + ASV + ATU + KTV (2.28) or Pup = Kyg i + Ky Vit Kyo Up + Kap Vit KygUin + Kar Ven + KygUy + Kar by (2.29) where Ky =RS; +k +88, and Ky = REG + M+ and so forth. That is, the displacement in the x direction at node i (U)) is considered as the gth degree of freedom for the truss, while the y displacement at node i (V,) is the hth degree of freedom, etc. The stiffnesses contributed to a given degree of freedom by the various structural elements must be added into the appropriate location in the structural stiffness matrix (K). This process is portrayed in Fig. 2.8 and illustrated for a three-member truss in Example 2.3Section 2.3 NODAL EQUILIBRIUM OF THE STRUCTURE 21 gh +— Column number U,V, *—— Displacement K Force P,, —= (row number ¢) Ky— Figure 2.8 Arrangement of the global stiffness matrix EXAMPLE 2.3 Obtain the structural stiffness matrix for the three-member truss in Fig. E2.3; Agy = 600 mm?, Ay, = 1000 mm?, A,. = 800 mm?, and E = 200 GPa. Stiffhesses contributed by element ab Stiffnesses contributed by element be Stifinesses contributed by element ac (b) Figure E2.322 Chapter 2. THE STIFFNESS METHOD USING THE BASIC EQUATIONS Solution For all elements AE/L = 40,000 kN/m. For element ab, ¢ = 0°, cos # = 1.00, sin ¢ = 0.00. Thus uw UW oh 1.00 0.00 —1.00 0.00 0.00 0.00 0.00 0.00 — 1.00 0.00 1.00 0.00 0.00, 0.00 0.00 0.00, 036 -—048 —0.36 0.48 —048 0.64 0.48 —0.64 —036 0.48 0.36 —0.48 048 —0.64 —0.48 0.64 Kk’ = 40,000] For element ac, $ = 90°, cos ¢ = 0.00, sin = 1.00, and U, Y, U, Vv. 0.00 0.00 0.00 0.00 0.00 1.00 0.00 ~1.00 0.00 0.00 0.00 0.00 0.00 100 0.00 1.00 ‘ks = 40,000] Note that the degree of freedom associated with each matrix column has been labeled to help with visualization. Combining the element stiffness matrices as shown in Fig. E2.3b gives the following force-displacement matrix equation for the truss: Pas 100 0.00 -1.00 0.00 0.00 o00]fu, P, 000 1.00 0.00 0.00 0.00 ~1.00]} ¥, P., |= 40,000]-100 0.00 1.36 ~048 -0.36 0481] U, Py 000 0,00 -0.48 0.64 0.48 —064]} K, Pee 000 0,00 -036 048 0.36 —048 HU, Pye 000 1.00 048 -064 048 164]] ¥, Discussion Each individual stiffness matrix is written using Eq. (2.22), and the structural stiffness matrix is constructed by assembling these three matrices using the logic developed in the foregoing discussion. Note that the assembly of the element stiffness matrices cannot be accomplished by simple matrix addition because the element k’s and K are not conformable with respect to addition. The graphic assembly representation in Fig. E2.3b assumes that K is initially null and the element stiffness matrices are added to80 Chapter 3 THE STIFFNESS METHOD USING VIRTUAL WORK y L., 16 kN (b) Figure E31 (a) Rigid body; (b) vertical virtual displacements; (c) horizontal virtual displacements Applying the theorem of virtual work, ie, dW, = 6W,, + 6W,, =0, and grouping terms gives 4/4 2 Ju + [Pa 5 G P) +8- fu a 4(4 [Fn +5G)- 4 3 2 3 ah, +($7,— 2 eu 6 42), (24 +3 Py — 10-2), =0Section 3.1 THE PRINCIPLE OF VIRTUAL WORK 87 Since du, , 6u,, and du, are arbitrary and independent, each of their coefficients must be zero for the above equation to be satisfied. These conditions give P, + ky = 0 9 —kimy — kyu, = 0 P. + kz, =0 The solution of these three equations gives u, = 0.60 in. (+); P, = —3 kips (+); and P, = —6 kips (+ ). Discussion By applying the principle of virtual work to this statically indeterminate structure both displacements and reaction forces are obtained. This occurs because the reaction forces are dependent upon the deformations of the structure, and the necessary number of equations for solving for all unknowns are formulated by giving each of the independent degrees of freedom of the structure a virtual change. Note that only reaction forces were obtained by applying the theorem of virtual work to the statically determinate structure in Example 3.1. In computing dW, the real element forces result from the real displacements. Since u, =u, = 0, only u, will produce forces within the elements. In contrast, for the virtual displacements selected and shown in Fig. E3.2c, du,, 6u,, and du, all participate in giving virtual deformations. See the next section for a detailed description of these terms. 3.1.3 REAL, ACTUAL, AND ADMISSIBLE DISPLACEMENTS — The principle of virtual work for deformable bodies involves the virtual work of the external forces (6W,) and the internal virtual work or virtual strain energy (6¥). The physical meaning of the former scalar quantity is illustrated in ig. 3.4, and the latter is shown in Fig. 3.5. The virtual strain energy density is portrayed in Fig. 3.6. In each case, the virtual quantity is the result of a virtual displacement. That is, the virtual displacement yields virtual deforma- tions of the element (dd) and corresponding virtual strains (de). The virtual displacements need not be those experienced by the structure, but they must be kinematically admissible. For example, the single axial force element in Fig. 3.8a is constrained at node a, but by envisaging the reaction force at a as another applied force we can impose an admissible displacement state as shown. Thus, even though u, must be zero, the virtual displacement at a need not duplicate the support conditions of the structure. The admissible displace- ment state in Fig. 3.8c is continuous between nodes a and b, and does not introduce any rips or tears in the element. In contrast, the displacements in Fig. 3.8d would not be admissible because there is a discontinuity (a tear) at point c, and this violates the kinematic conditions of the element. The element in Fig. 3.8a is uniformly tapered from end to end, with the cross-sectional area decreasing linearly. By making an imaginary cut through the element and envisaging a free-body diagram of everything to the right of the cut, we deduce that the element will have an axial force of P throughout itsSection 3.4 THE STRUCTURAL STIFFNESS MATRIX 109 Solution For element ab (,» = 36.87°; see Fig. E3.5b), wo e608 day vs in Las é ”), For element ac (bye = For element ad (pag = 315°), The internal virtual work is OW,=¥, | Geto" dV = 6e"0" A, Ly im Sym mn =Y 6e""EA, Ly Choosing the virtual strains with the same form as the real strains gives f IE - owe [ ou(2) COs? ay + te oo) in ig OSes L Jap L Jay AE . AE 3 +0, iu( 4) sin day 608 bay + Ps ou(42) in? a] it uf 42) [cul 42) cos tam Se FE) sin 608 Py ft L -nsiu{!2), sn duce out iE), su] L Joa L Jo cos? Poe ¥ In units of kips per inch: (AE/L)q, = 290; (AE/L),. = 302; (AE/L)qg = 769. Also, cos dap = 0.8; sin Gay = 0.6; COS dae = 1.0; COS hag = 0.707; sin bag = —0.707. Thus, BW, = (185.6, + 139.20, + 302.0u, + 384.5u, — 384.50,)5u, + (139.2u, + 104.40, — 384.5u, + 384.5u,)ov, ‘The external virtual work done by the applied loads is 5W, = 50 du, + 80 5v, From the principle of virtual work, 51, = 5W,. Since du, and dv, are arbi- trary, independent, and nonzero, we can equate the coefficients of these twoSection 4.2 NODAL EQUILIBRIUM, NODAL DISPLACEMENTS 141 EXAMPLE 4.7 Calculate the nodal displacements and element forces for the beam of Example 4.3 if it is supported as shown in Fig. E47. Okin 1740 kin 435 kein Ok-in 10.0k | Ties im | (b) Figure E4.7 Solution See Example 4.3 for dimensions, etc. From Example 4.3 (using appropriate entries from rows and columns 1, 2, 4, 5, 6, and 8 of K), —10 0.79271 68.966 68.966 0000.0 00000-0000 ||, 0 68.966 8000.0 4000.0 00000-00000 00000 |] 0, ~ 102] 8965 4000.0 160000 —68.966 40000 00000 |] @, -15 | 0000.0 0000.0 — 68.966 1.58542 0000.0 68.966 |} 0 0000.0 0000.0 4000.0 0000.0 -16000.0 40000 |} @, 0 0000.0 0000.0 0000.0 68.966 4000.0 80000 |} @, Thus Y, —0.94612 in. ‘e,. 0.00689 rad ©, |_| 000254 rad Key 0.00000 in. e. —0.00073 rad Oo 0.00036 rad Using Eq. (4.48) and the displacements, the element forces are {Vv M, M,}={-100k Ok-in, —1740k-in.} {V My, MJ" = {125k 1740 k-in, 435 k-in.) {V M, Mj ={-25k —435k-in. 0 k-in.} See Fig. E4.7b.154 Chapter 4 THE STIFFNESS METHOD FOR BEAMS AND PLANAR FRAMES The axial force required at each end of the element to maintain the element in an undeformed position is +n2 +2 AT Pe f o’b dy = J (Te Ae »p dy =AExT, (4.77) =n wa The moment that must be applied at the ends of the element to prevent rota- tion of the cross section is pena oma Me -/ ExT (y)by dy = Ea | —ny2 (t + a vy dy= fear (4.78) aya The beam would deform into the shape in Fig. 4.9 with a constant curva- ture; therefore, there will be neither rotations nor displacements if M? is applied to the left and right ends in a clockwise and counterclockwise direc- tion, respectively. The axial force P¥ will be neglected, since it is assumed that the neutral axis remains undeformed for conventional beam theory. Hence, the initial force matrix for the beam element is Bi 0 mg] ElaAT | —1 my (4.79) Py h 0 ng 1 and the corresponding initial force matrix for calculating final element forces is Bo ° me = Seat =1 (4.80) CH 1 EXAMPLE 4.10 The beam of Example 43 is fixed at nodes a and d. The structure in Fig. E4.10 is uniformly heated along its length with T, — T; = 120°F. Calculate the nodal displacements and the element forces. h = 12 in.; «= 6.5 x 10 ° in,/(in.- °F). [}-—— 34.5 fi = 43.5 h ——+ (a) 2260 kip: ft 2260 kip ft 2260 kip ft 2260 kip: ft 0 0 0 o 0 0 (b) Figure £4.10Section 4.4 SETTLEMENT, INITIAL, AND THERMAL STRAINS 155 Solution EIaAT _ (29 x 10° ksiX1200 in.*)[6.5 x 10° in,/fin. -°F)](120°F) a 12 in. = 2260 kip- in. Hence for each element Pri 0 cA -1 ™) = 2960] Poi 5 it 1 These initial fixed-end forces are merged to give Py, which in this case is a null matrix. Since P, is also null, solving the structural force-displacement equa- tions yields U, = 0. That is, all the nodal points of this structure remain in their undisplaced positions. For all elements ay’ 6 im, | = 80 +8 = 2260 k- in| —1 1 See Fig. B4.10b. Discussion Initial forces and the procedure embodied in Eqs. (4.62), (4.63), and (4.65) are used. This thermal distribution results in zero displace- ments, but since the structure is statically indeterminate, bending moments are imposed along its length. The initial forces for self-straining effects can also be obtained using the theorem of virtual work. In Sec. 3.7 [see Eq. (3.92)] we calculated that r= [ BT Ee? dV + [ Blo? dv Wot hot In the case of a prismatic beam element B is given in Eq. (4.27). For a thermal distribution that is uniform along the beam axis and varies linearly from the lower to the upper surface (see Fig. 4.9), ¢” is given by Eqs. (4.75) and (4.76a). Substituting B and T()) into Eq. (3.92) yields By a= | Biz Ary w--[ 9/8 Ba( Ta + 72) av 13 Baa,158 — Chapter4 THE STIFFNESS METHOD FOR BEAMS AND PLANAR FRAMES, Daiji) Figure 4.10 The plane frame element the beam element [Eq, (4.19)]. This yields Bui A 0 0 -A 0 07fii B 2 6g 1 SE, as eo a |i my ot 46° 0 af ar || 6; z E ic (4.81) by) L|-A 0 0 A 0 2 3 1 6 10 Bi “Er ° = 61 6 5 iy oF 4 0 Ls, Usually, each node in a plane rigid frame has three independent gener- alized displacements that must be considered (two translations and one rotation), and usually it is necessary to transform the generalized element forces and displacements into the common global coordinate system. For the beam element in Fig. 4.11 the generalized forces at node i are transformed as follows: Pui Bu= P,. C08 $ + py, sing Pai Sin + Py COS (4.82) and Similar transformations can be expressed for the generalized forces at node j; thus for both nodes of the element coop sing = 0 0 0 07 [Pas -sing cose 0 0 0 OT py 0 0 1 0 0 o}] m 48 - 0 0 0 se sng offp,| “8 0 0 0 -sing cosd Of] py o 0 0 0 0 1jLm182 — Chapter5 THE STIFFNESS METHOD FOR THREE-DIMENSIONAL STRUCTURES Solution The element stiffness matrices are obtained using Eq. (5.1). Invoking nodal equilibrium for the structure and enforcing the support conditions yields 0 413.443 139.200 0.000 | — 302.083 0.000 [Uy 0 139.200 818.444 0.000 0.000 0.000 || % o|= 0.000 0.000 197.973 0.000 0.000 |} ¥, 0 — 302.083 0.000 0.000 549.550 185.600 |] U, — 120 0.000 0.000 0.000 185.600 139.200 || ¥, Hence Us 1.825 y =0.310 W,|=| 0.000 | in, u, 2.354 v, 4.001 Using Eq. (5.9) the element axial forces are s“! = eu = +167 kips (t) s! = eu” = +167 kips (t) s* = eu = +160 kips (t) s = eu = —200 kips (c) s = eu = —200 kips (0) Discussion The results for this simple statically determinate structure can be checked using the methods of statics and strength of materials. Note that all rotational degrees of freedom have been excluded from K,,, since they have zero stiffnesses and if included would render the equations ill conditioned with respect to solution; W, has been excluded for the same reason, There is a stiffness term for W, because elements ad and bd extend out of the x-y plane. EXAMPLE 5.2 Analyze the space truss in Fig. E5.2. Ay = Agg= 10 x 10~* m?; all other element areas are 30 x 10 * m?; E = 200GPa.Section 5.1. SPACE TRUSSES 183 Figure E5.2 Solution The element stiffness matrices are obtained using Eq. (5.1). Invoking nodal equilibrium for the structure and enforcing the support conditions yields 0 82735 —57735 00000 +~—25000 00000 00000 Uy 0} | -s7735 107740 00000 0000 + 00000-00000 |] 0} | 00000 00000 57735 00000 00000-00000 |W, 0} | —25000 00000, 00000 =—:118510 75624 00000 U, —60 00000 00000 00000, 75624 66679 00000 ve 0. 00000 00000 00000, 00000 00000 57735. Ww Hence Uy 1.588 % 0851 W,| 4-3] 0.000 U, 3291 |™ ¥, 4.632 Ww, 0.000, Using Eq. (5.9) the element axial forces are +426 KN (t) = +42.6 KN (t) +39.0 KN () —369 KN (0) = —67.1 kN (0) —369 KN (c) = -67.1 KN (0)186 Chapter 5 THE STIFFNESS METHOD FOR THREE-DIMENSIONAL STRUCTURES The rank of the stiffness matrix is one if| Ky,| = 0, that is, if cos? a sin? 2 =0 or if (sin 2a)? = 0 This is true if « = 0, 7/2, x, etc. If «= 0, the rank of the augmented matrix is two; therefore the equations are inconsistent and we have a mechanism (see Fig. E5.3b). If « = 7/2, the rank of the augmented matrix is one; therefore the equations are consistent and the structure is stable since it consists of two coincidental axial force elements (see Fig. E5.3c). 5.2 THE THREE-DIMENSIONAL BISYMMETRIC BEAM ELEMENT Three-dimensional structures constructed from flexural members exhibit a more complex behavior than a space truss. In the most general situation there are six degrees of freedom per node, and each element is subjected to axial forces, torsion, shear, and bending about two orthogonal axes, Consider the prismatic beam element in Fig. 5.3; its cross section is symmetric with respect to each of its two principal axes, j and 2. When arbitrarily oriented in three- dimensional space, at each node the element must sustain six generalized forces (Pe, Bys Bes i, iy, i.) and will exhibit the corresponding six degrees of freedom (i, #, #, 0, ,, 6). All these quantities are shown in Fig. 5.3 directed in their respective positive local coordinate directions, with the moments and their corresponding rotations indicated as vectors with double arrowheads. This linear elastic element has four distinct response modes: (1) tension- compression in the % direction; (2) flexure in the X-) plane; (3) flexure in the Figure 5.3. The beam element with six generalized forces per nodeSection 6.1. MATRIX CONDENSATION 219 (©) With a hinge on both elements: = 2EL , 3B _ 27EI DB" 3p 8b 8 “=~ oie? Discussion This structure can be analyzed using three different com- binations of conventional and hinged beam elements, and all have been demonstrated for computing the generalized displacements. In using two hinged elements, only ¥, is obtained, but in each of the other approaches the rotation at point b that is associated with the conventional beam element is obtained from the calculations. Only the computation of element forces has been demonstrated for a hinged element between nodes a and b with a conventional beam element between nodes b and c (Fig. E6.2b). This required the e matrix corresponding to the stiffness matrix in each case. We suggest that the reader calculate the element forces for the other two idealizations as an exercise. 6.1.2 SUBSTRUCTURAL ANALYSIS — Some structures are too large to be analyzed as a single system, and they can be investigated only after division into a number of smaller units or substructures. For example, in the case of an airframe as shown in Fig. 6.2, it is not unusual to require thousands of nodal points to completely describe the response of the entire structure. Each of the substructures defined is assigned to an individual engineering team, and the analysis of these separate units can proceed in parallel. With the smaller problem sizes, the engineers can make more convenient and reliable data checks. Also, the equations can be reasonably and economically manipulated on limited computing hardware. The response of the total airframe system is Figure 6.2 Typical substructures for an airframe222 — Chapter6 SPECIAL TOPICS FOR THE STIFFNESS METHOD EXAMPLE 6.3 Analyze the beam in Fig. E6.3 using substructural analysis; ! = 1200 in.‘ and E = 29 x 10° kips/in.?. 20 kips 10 kips 50 kip: ft 20 kips 10 kips 80 kip: ft e © a Substructure A Substructure B Figure E6.3 Solution The stiffness matrix for each beam element is 2.4167 145.00 —2.4167 145.00 145.00 11600. —145.00 5800.0 —2.4167 —145.00 2.4167 —145.00 145.00 5800.0 —145.00 11600. ki = 107 where w= {v, 6 r% 8} For substructure A, Eq, (6.13) is 0 2.4167 —145.00; 00000. —2.4167 —145.00 || Vv. 11600. ; 00000. 11600. 5800.0 || ©, -20 —2.4167 145.00 | — 145.00 00000. Y 0 — 145.00 5800.0; 5800.0 23200. || 9, Using Eqs. (6.14), (6.15), and (6.16) gives the following condensed force- displacement relationships for substructure A: =100]_f 7.5521 —18125 ]{ ¥, 600.]~L-18125 435000]. @,224 Chapter 6 SPECIAL TOPICS FOR THE STIFFNESS METHOD local transformations are subsequently carried out while calculating element forces. The transformation matrices are square if the matrices of variables in the two coordinates are of the same order, that is, T is m by n if @ and ware both n by 1, and u=Tu or u=T0 since T is orthogonal in this case. The same type of transformations apply to the force matrices, that is, p=Tp and p=T'p If the number of degrees of freedom of an element differ in local and global coordinates (¢.g., a and u are different orders), the theorem of virtual work can be applied to obtain the following contragredient transformations: u=T' and p=Tp In this case T is a rectangular matrix. Transformations can be used in other ways to deal with a number of special situations. For example, if several degrees of freedom in a structure are dependent, we can apply the theory of transformations to impose these con- straints on the structural force-displacement equations. Structural supports that are not aligned with the global coordinate system also impose unique demands on the force-displacement equations. Such skew supports can be treated using a set of nodal coordinates located at the subject node; the affected degrees of freedom are transformed into the nodal coordinates and the pre- scribed skew boundary conditions are imposed on the transformed equations. Coordinate transformations can also be applied to handle offset nodes; for example, if a structural member has a cross section that is not doubly sym- metric and the centroidal axis and shear center do not coincide. The flexure and torsion are coupled; the stiffnesses at nodes must be transferred to a common nodal point. 6.2.1 CONSTRAINT EQUATIONS A constraint indicates any type of relationship between generalized displacements. An unyielding support is a constraint that is described by putting the corresponding degree of freedom into the partition U,, and its effect does not appear in the partition K,,. Another type of constraint occurs if there is a dependency between several generalized displacements; this condition must be imposed on the structural force-displacement equations before solution. For example, the girders of the plane rigid frame in Fig. 6.4a are axially rigid and it is required to indicate this fact by noting that all the horizontal nodal displacements at a given floor level are equal. Another type of constraint occurs for the structure in Fig. 6.4b, which has experienced a support settlement; here the relationships for the vertical displacements of the foundation are prescribed. Constraint254 = Chapter 7 THE FLEXIBILITY METHOD and, from Eq, (7.11), U, = Lf, fol(fs + fa) Pan From Eq, (78), the forces in the ERLE Teta a tr fithl—fh The governing equations are written for the constrained assemblage; therefore, in the context of the stiffness method, Eq, (7.11) refers to the govern- ing equations with boundary conditions imposed. Comparing Eqs. (7.11) and (2.34) reveals that F=K;} (7.13) Since f; = 1/k;, F = 1/ky + k2), and K;, = (k, + k,), Eq. (7.13) is identical to the results obtained using the stiffness method for this two element assemblage with nodes a and c constrained [see Eq. (2.24)]. 7.2. THE FLEXIBILITY METHOD FOR TRUSSES We will apply the flexibility method as displayed in the previous section to the truss of Example 2.7 (see Fig. 7.3a). Since element ab transmits zero force the truss could be considered to have one statical redundancy, but two redundants will be included to illustrate the effects of internal (i¢., element) and external (ie. support) redundancies. For this truss P={P,. Pye Pxg Pyg} and X = {X, X,}. A column of the b, matrix contains the forces in the elements for a unit value of the associated external force; for example, the fourth column corresponds to element forces required to equilibrate a horizontal force at node d (see Fig. 7.3b). Similarly, a column of the matrix X is composed of the element forces that equilibrate a unit value of the associated redundant force (e.g, Fig. 7.3c). For this structure, Eq. (7.1) is a 0 0 0 0 % st -1 0 0 O|fP,. eo 1 8 Ueda ofp] ow st v2 0 ¥2)— OO || Pa é X 0 0 0 0. 0Section 7.2. THE FLEXIBILITY METHOD FOR TRUSSES 255 — 50 kips Figure 7.3 (a) Truss with two static redundants; (b) primary structure with applied unit force; (c) equilibrium solution for unit redundant X, From Example 2.7 we note that L/AE is 1/200 for horizontal and vertical elements and ,/2/200 for diagonal elements; therefore, the flexibility matrix of all elements is 1 0 0 0 0 oO o 1 0 0 0 0 r-L}/® © 1 0 0 0 ai 20}0 0 0 1 0 0 : 0 0 0 0 y2 0 0 0 0 0 0 y2256 Chapter 7 THE FLEXIBILITY METHOD Substituting the basic matrices by and b, from Eq, (7.14) and f from Eq. (7.15) into Eq. (7.12) yields the 4 by 4 flexibility matrix, F = K;/, identical to that shown in Example 2.7. Alternatively, we could use the bo matrix associated with P,4: the only applied force with a nonzero value. That is, using only the third column of by from Eq. (7.14) yields ={0 00 -1 y2 9 (7.16) Therefore, 1 Foo = 3g [3828] 1 For = Ffo = 395 [2707 0.000] (7.17) F -* 4828 —0.107 +1200 | -0.707 1.000. 2.135 d i= . an 0 (7.18) Since P,,= 50 kips, Eq. (7.11) gives U, = 0.534 in., which is identical to the result obtained from the stiffness method in Example 2.7, We obtain the clement forces from Eq. (7.8). Using bo from Eq. (7.16) along with b, from Eq. (7.14) and the matrices of Eq. (7.17) we have bo — b\FiFio = {0.000 0.442 0.442 —0.558 0.789 —0.625} stb ged got gh gal hey 0.0 221 221 —27.9 39.4 —312} kips (7.19) and Positive and negative signs indicate clements in tension and compression, respectively. The two redundants can be calculated using Eq. (7.6), along with the results shown in Eq, (7.17); thus {X, Xz} = {0.625 0.442}50 = {-31.2 -221}kips (7.20) The negative sign associated with both redundants indicates that we assumed the directions incorrectly. That is, element be, X,, is in compression, and the horizontal reaction at node b, X, acts in the negative coordinate direction. We will resolve the dilemma of the possible sizes of the bo matrix after deriving the equations using the principle of complementary virtual work. For now it is sufficient to observe the outcome using the single column by matrix. Using the 4 by 4 stiffness matrix from Example 2.7 and condensing (see Sec. 6.1) all degrees of freedom with the exception of Uy, we obtain K,. = 0.4687(200). Note that K,.' = 2.134/200. That is, in general we obtain the global flexibility matrix (F) corresponding to the inverted condensed stiffness matrix associated with the degree(s) of freedom included in bo, as indicated by Eq. (7.13).258 Chapter 7 THE FLEXIBILITY METHOD redundants as the moments at nodes a and c (see Fig. 7.6c). Using Eq. (7.21), the flexibility matrix for all elements (ie, ab, be, and cd) is 2-1 0 0 0 0 -1 2 0 0 0 0 tL] o 0 2-1 0 0 foi] 0 0 -1 2 0 0 2) 0 0 0 0 2-1 0 0 0 O-1 2 With this choice of flexibility matrix, Eq, (7.1) for the structure is 0 an) me L 11 me “2 2 2 be me = é Put -} -} [%] (7.24) med, 0 ot mit 0 0 -1 0 0 0. ‘ I on ptt [eee b (b) Figure 7.6 (a) Beam with two static redundants; (b) primary structure with applied unit force; (c) primary structure showing redundant moments268 7.6 Chapter 7 THE FLEXIBILITY METHOD Substituting Eqs. (7.65) and (7.66) into Eq. (7.64) yields SP'U + 8X7 d, = SPT d + 8X"b7 d (7.67) Thus 8P"(U — b3d] + 8X"[d, — by d] =0 (7.68) Since 8P and 8X are arbitrary, nonzero, and independent, Eq. (7.68) can only be satisfied if the expressions in the square brackets are each zero, that is, U=bid (7.69) and d, =bid (7.70) Substituting d = fs = fb, P + b,X) (7.71) into Eq. (7.70) and noting that d, = 0 we have X = —(b7fb,)~ 'bTfby P = —Fy}'F oP (7.72) Substituting Eqs. (7.71) and (7.72) into Eq. (7.69) yields U = (b} fo — bp fb, Fi1'F o)P = (Foo — FosF i Fio)P. =FP (7.73) where F = Foo — FoiFitFio (7.74) This result is identical to that obtained using the basic equations [see Eq. (7.12)]. RELATIONSHIPS BETWEEN STATICS AND KINEMATICS In Ch. 3 we noted that for conjugate vectors (e.g, P and U or s and d) there are corresponding contragredient transformations. That is, s=bP (7.75) implies the existence of the corresponding transformation U=b"d (7.76) Equilibrium for the statically determinate truss of Example 2.5, which is illus- trated in Fig. 7.12a, yields (see Fig. 7.12b) * -1 0 OlfPy. sx l=] $0 $]/P,, 7.7) x o -1 0f[PsSection 7.6 RELATIONSHIPS BETWEEN STATICS AND KINEMATICS 269 ° (b) Figure 7.12 (a) Statically determinate truss; (b) equilibrium solution for unit applied force P,, Therefore, from Eq. (7.76), U, 1 = olfa V.J=| 0 O —-1]] d& (7.78) U,, o ¢ olla Similarly, using the contragredient rule, the transformation d=aU (7.79) implies P=als (7.80) For the truss in Fig. 7.12a, a -1 0 1)fu, ad {=| 0 O #1% (7.81)270 Chapter 7 THE FLEXIBILITY METHOD Therefore, Pry -1 0. off s® P,,{=| 0 0 —1]| s¥ (7.82) Pas 1 2 oll In Ch. 2 we discussed the analog between static and kinematic indeterminacy. The truss in Fig. 7.12a has three kinematic indeterminacies corresponding to the degrees of freedom ¥,, U,, and ¥, ; therefore Eq. (7.81) could have been written as follows: d=[a, ale] (7.83) where U,={U, V, Us} (ie, the unconstrained degrees of freedom), U, = {V, U. V,} (/e, the constrained degrees of freedom), a, is the kinematic matrix associated with U, [ie, a shown in Eq, (7.81)}, and a, is the kinematic matrix associated with U, ; that is, 0 0 Oo 43 4 a=|-3 -; ¢ (7.84) o o 1 We recall from Eq. (7.55) that for an individual element d=fs= + s (7.85) Therefore AE s=kd=""d (7.86) Thus for the entire truss of Fig. 7.12a, - ke 00 Hae se]=]0 x 0 fa (787) st Oo 0 eller s = k a Substituting Eq. (7.79) into Eq. (7.85) yields s=kaU (7.88) From the principle of virtual work (i., 6W, = dW, and Eqs. (7.79) and (7.88), SU'P = d's = SUTaTkaU Thus aTkaU (7.89)A Al SOLUTION OF THE EQUILIBRIUM EQUATIONS The task that consumes the most computer time in executing the stiffness method of analysis is that of solving the equilibrium equations for the unknown generalized displacements. In the case of a dynamic analysis the percentage of solution time is even larger than that for a static solution. The structural engineer is typically concerned with: idealization of a structure into its individual structural elements; efficient assembly of the element stiffness matrices into the global stiffness matrix of the system; and interpreting the calculated nodal displacements and element forces. The feasibility of a struc- tural analysis depends upon an efficient algorithm for solving the algebraic equations of equilibrium. In the late 1960s an airframe manufacturer pushed the state of the art of equation solvers with an investigation of the wing-body intersection of the 747 airplane; this involved approximately 2500 equations. Today, structures with 10,000 equilibrium equations are routinely solved. Computer solution algorithms are usually formulated by numerical analysts, but the structural engineer must be aware of the cost, accuracy, and sources of errors associated with the solution process. Equipped with a knowl- edge of a solution algorithm we can: minimize analysis time by appropriately numbering the nodes or elements; interpret errors; and understand the limi- tations of the results. Sets of simultaneous linear algebraic equations can be solved using either direct elimination methods or iterative solutions. During the emergence of the stiffness method of analysis the latter solution algorithms were widely used; today these have been largely supplanted by Gauss elimination. The structural equilibrium equations have special qualities that make them amenable to efficient solution algorithms. That is, the coefficient matrix: is symmetrical; includes many zeros; contains nonzero entries clustered about the main diago- nal; and has positive terms on the main diagonal. These respective character- istics can be succinctly described by stating that the global stiffness matrix is symmetric, sparse, banded, and positive definite. Solution errors are intro- duced by the computer, the algorithm, and the structural idealization. An added benefit of a well-executed solution package is that it can frequently detect an unstable structure. DIRECT SOLUTIONS BASED ON GAUSS ELIMINATION We are interested in obtaining the generalized displacements associated with the static nodal equilibrium equations of a structure. In this section we will 285286 — Appendix A SOLUTION OF THE EQUILIBRIUM EQUATIONS r pi Figure A.1 A three-element beam structure discuss methods of direct solutions based upon Gauss elimination. For the sake of simplicity we will write the force-displacement equations for a beam element [see (Eq. (4.19)] as follows: By 2 6 -12 6]f a mjL|_EI| 6 4 -6 2/16L — (ALL) by | B|-12 -6 12 6]! 5, (Any aL. 6 2 -6 461 Thus we can write the force-displacement equations for the supported beam of Fig A.l with El/L = 1 and L = | as follows: My 8 -6 2 oo, 0 P,|_|-6 24 0 6lly}_}-1 M.| | 2. 0 8 2/0,] | 0 (A2) M 0 6 2 4jlo, 0 A.1.1_ SOLUTION BY ELIMINATION We will solve the equilibrium equa- tions in Eq. (A.2) using Gauss elimination as follows: STEP 1 Subtract a multiple of the first equation from the second and third equations so that the first column of K contains all zeros; that is, we multiply the first equation by —3/4 and subtract it from the second, and the third equation is replaced by 1/4 times the first subtracted from the third. This gives Py i a Nv ° 39 3 eo, 0 Ory 2 Sly }_f-a / 3 15 e | | 0 (Aa) "3 F *IeJ} | o 0'6 2 4Section A.2 SOLUTION ACCURACY 301 Table A.1 Solutions for three-spring assemblage of Fig. A.6 Number of significant digits used Five Six Exact Four Five +R/O Six +R/O Seven U, 0.666686 Nosolution 0.50002 0.66666 0.645162 0.665869 0.666667 U, 0.666657 Nosolution 0.49999 0.66663 0.645132 0.666637 0.6666370 Therefore 1.0000 =L"'p= ¥obors [ise] Aa) (0,000089190 -tye and D'Y= ee 1 (A41) Solving L'U = D~'Y yields Us] _ [0.50002 al a [pisees | (a42) If our hypothetical computer performs arithmetic operations by rounding off the fifth significant digit and then truncating, we observe the following: L remains unchanged; D,, becomes 3.0000, while other entries in D are unal- tered; wipes aa a= (pees Aa) and L'U = D™'Y yields Us] _ [0.66666 [t'| a [Rees | (A.44) The displacements for the assemblage of Fig. A.6 using various calculation strategies are summarized in Table A.1B MATRICES B.1 DEFINITIONS A matrix is a rectangular array of elements arranged in rows and columns as follows: My Fz A=[a,)=|% 72 (Bl) Gm m2 where a,, is the element in the ith row and jth column. The above matrix is order (ie., size) m x n (ie., m by n). The row index is always given first. If m=n, A is a square matrix of order n; the elements a,1, 422, .-., Gyn lie on and form the main (or principal) diagonal. If n = 1, A is a column matrix of order m x 1, and if m = 1, A is a row matrix of order 1 x n. B.2 EQUALITY es A = [a,;] and B =[b,;] are equal if they have the same order ys that is, x+2y]_[2 3x+4y] L8 implies that x + 2y = 2and 3x +4y =8 B.3 ADDITION AND SUBTRACTION The sum of two matrices A = [a,;] and B = [b,] is A+B=[a, +b] =[e]=C (B2) where it is implied that A and B are the same order. Thus, 18 69 4 4 3°4/+] 8 SJ=)}tl 9 72 12 3 19 5. 302Section B.4 MULTIPLICATION 303 In general, matrix addition has the following properties: A+B=B+A (commutative law) (B.3a) A+(B+©)=(A+B)+C (associative law) (B.3b) The mull (or zero) matrix 0, consisting of only zero elements, implies that A+0=A and A+(-A)=0 (B4) where —A =[—a,]. Note that —A = [—a,] = (—1)[ay], where we multiply each element of the matrix by the scalar —1. In general multiplication of a matrix by a scalar implies that kA = Kay] = (kay) (BS) Thus matcix subtraction is similar to matrix addition since A-B=A+(-1)B=A+-B (B6) For example, 18 69] [-s -1 3 4/-| 8 sl=}-s -1 7 2) [iz 3) L-s -1 B.4. MULTIPLICATION If A is a matrix of order m x n and B a matrix of order n x p, then the matrix product of A and Bis a matrix of order m x p, that is, A B=C =tea=[ S06, (B7) (mxm) (np) mx) where i= 1, 2,...,m;j=1, 2,..., p. Note that for matrices A and B to conform for multiplication in AB, the number of columns in A must equal the number of rows of B. When the product is written AB, A premultiplies B, and B postmultiplies A. For example, if 7 10 135 ~ 0 tasl7 "| ap-[| | 8-11 24 6H] 5 9 [eect 1(10) + 3(—11) + 50) ~ |=) + 48) + (-9) (10) + 4-11) + 60) —B 23 —36 —24 then310 Appendix 8 MATRICES the upper left 2 x 2 submatrix gives le i ; | (12\(7) — (8X11) = 84 — 88 = —4 B.10 DIFFERENTIATION The operation dA/dx indicates that we must differentiate the matrix A element by element with respect to x. For example, if 2x3 4x? 5x sinx cos*x 2sinx 8x 3x? ne therefore, 6x? 8x 5 aA T=] esx —2e0sxsin x 2.c0s.x * 8 6x sx* Partial differentiation is also performed by operating upon each element. B.11 INTEGRATION We integrate a matrix by operating upon each element. Thus if 2x3 4x? Sx A=|sinx cos? x 2sin x 8x 3x? x therefore, ae ax? sx? ]* 2 a 2 = 1 [aa cos x (2x +sin 2x) —2cos x x6 4x2 x = 4. LB 4b si 4 3 2 =| t—cost GAL + sin 2L) 21 00s 1 4 vBSYSTEMS MEASURE D Although the structural engineer in the U OF MENT ited States must work with pro- ducts, documents, institutes, and industries that use U.S. Customary System (USCS) units almost exclusively, the civil engineering profession is working toward the official adoption and use of the System International (SI) units. It therefore behooves the student to develop a facility for working with both systems. Tables D.1 through D.3 provide a ready reference for the SI units used in this book. More complete information on the entire SI system is avail- able in many references Table D.1 Selected SI units SI symbol Unit Quantity Base units: Length meter m Mass kilogram kg Time second s Derived units: Area meter? Density, mass kilogram/meter? Force newton Moment of force newton-meter Moment of inertia, area meter* Pressure, stress pascal Temperature Celsius Work, energy joule Supplementary units: Plane angle radian rad Table D.2. Sl unit prefixes Multiplication Multiplication Prefix Symbol __factor Prefix Symbol _ factor exa E 10** deci* da 10°" peta Pp 10's | centi# oc 10°? tera oT. 10 milli om 10-3 gia = G 10° | micro 10-° mega M 10° | nano n 10°? kilo k 103 | pico P 10-1? hectot oh 10? femto 107" deka’ da lot | ato a 10718 * Use should generally be avoided 313314 Appendix D SYSTEMS OF MEASUREMENTS Table D.3_ Some conversion factors Length Lin, = 25.40 mm 1 mm = 0.03937 in. = 00254 m 1 m = 3937 in. 1 ft. = 304.8 mm 0.003281 fi = 0.3048 m 3.281 ft Area 1 in? = 6452 x 10? mm? 1 mm? = 1.550 x 107? in? = 6452 x 10-* m? 1 m? = 1.550 x 10? in? 1 ft? = 9.290 x 10¢ mm? 1 mm? = 1.076 x 1075 ft? = 9.290 x 107? m? 1m? = 1.076 x 10° f? Moment of inertia Lint = 4162 x 10° mm* 1 mm* = 2.402 x 106 in = 4162 x 10-7 m* 1 m* = 2.402 x 10° in* Force and force per unit length 1 Ibf = 4.448 N 1 N = 0.2248 Ibf 1 kip = 4448 x 109 N LKN = 2.248 x 10? Ibf 1 Ibf/ft = 14.59 Nim 1 N/m = 6.853 x 10°? Ibfjft 1 kip/ft = 14.59 x 10° N/m 1 KN/m = 6.853 x 10? Ibf/ft Bending moment 1 Ibf-in. = 0.1130 N-m 1. N-m = 8.851 Ibf- in. 1 Ibf- ft = 1.3558 N-m 1 N-m = 0.7376 Ibf- ft I kip: ft = 13558 x 10° Nem 1 KN: m = 0.7376 x 10? Ibf- ft Stress 1 Ibffin.? (psi) = 6.895 x 10? Pa 1 Pa = 0.1450 x 107? psi 1 kip/in.? (ksi 6.895 x 10° Pa 1 kPa = 0.1450 psi 1 Ibf/it? (psi) = 47.88 Pa -? psf Temperature te = (9/S)tc + 32E ANSWERS TO SELECTED PROBLEMS Chapter 2 2.2 (a) Y= —10.0 mm;(b) P,, = 216 KN 2.3 (a) V,=0.40in.;(b) P,, = 165 kips 2.4 (a) and) —16 4 -8 0 2 | K=3x10) 9 gg ay | NI 0 0-1 2 © uy] fi2 u, | =| 28 | mm Uy 31 Py, = —576 KN @ [c.]-[e)== [fe]= [care es (e) U, =8.8 mm Pz) [ —1152 P,,|=| 1517] kN Poa -554 27 036 048 = 0.00 0.00 048 1.64 0.00 —1.00 0.00 0.00 1.00 000 —1.00 0.00 K=2x 10) 499 100 0.00 100 000 a0 |*N™ =036 -048 1.00 000 136 0.48 048 —064 0.00 000 048 0.64. 2.8 (a) Equilibrium (b) The forces Pxa, Phas Pars Pro» Pac s Pye due toa U,= 1, while all other nodal displacements are zero. xand y directions is satisfied. (©) The diagonal terms are the forces required to impose a unit displacement at the corresponding degree of freedom. (@) The truss is not restrained against rigid-body motion. 315316 ‘Appendix E ANSWERS TO SELECTED PROBLEMS 2.9 IfV, = 1,while U, = 0, P,, = (/3AE/2L) and P,, = 2AE/L: U,] _ PL [0.0824 v,|~ AE [0.9643 =0.55P (tension) 2.42 1,000 0.000 —1.000 0.000 0.000 1.000 0.000 0.000 Kry= 200) 1000 0.000 1354 0.354 0.000 0.000 0354 1.354. 2.13 (a) 1.72 0.00 -036 -100 0.00 0.00 1.28 048 0.00 0.00 K,,-—2x 10*| —0.36 048 1.72 —036 —048 =1.00 0.00 -036 136 0.48 0.00 0.00 -048 048 0.64. o 1.72 0.00 -0.36 —100 0.00 0.00 1.28 0.48 0.00 0.00 K,,=2% 104 -036 048 1.72 -036 048 =100 0.00 -0.36 220 0.85 0.00 0.00 -048 085 0.80. 2.15 s* = 22.5 kips (c); s* = 75 kips(c) 2.16 (@) 15-05 -10 00 | -05 15 00 00 Kry= 8x10) i 00 1s os 00 00 05 15. (®) K,, same asin part (a). (© s4 = 386 KN (0); s™ = 218 KN (¢) 2.17 (a) 20 00 05 —05 xu4E] 09 20 -05 05 L| 0s -05 15 00 -05 05 00 15. (b) 94 = 11.79 kips (0) 2.19 Ww] _[ 13.72 v.]=L s10}™™ s* = 90.0 KN (0) 2.21 | a [ 0.00 YJ [-0.32 s = 6.0 kips (c) 2.23 U.]_PL[ 0.7155 v, |” AE | —0.5822 | N/m N/mAppendix E ANSWERS TO SELECTED PROBLEMS 317 2.25{U, VU, WU. ve Us % Uy Vv, U, YU) 10°*{114 -1009 56 —347 -197 —346 —293 —49 —98 157 —S1 168 —113} (in) s* = 1.73 kips (c); s* = 2.00 kips (1); x = 2.08 kips (t); s* = 1.66 kips (c); s = 1.89 kips (c); s/ = 2.13 kips (1); s' = 1.52 kips (6); s/*= 2.11 kips (0); = 1.83 kips (0) 2.26 v, 1.281 Us 6.734 % 0.497 U, |= 10-5) 1.538 | in % -2.616 U, 2.893 by. ois s = 0.260 kips (c); s* = 0.165 kips (c); s* = 0.372 kips (t); s™ = 0.127 kips (0); s = 1.005 kips (t); s** = 0.928 kips (c); s** = 0.442 kips (t); s** = 0.315 kips (t) 2.28 (a) U.)_[ 0.109 # v, |= | 2126 |™™ @ U,] _[-0.517 v,|>L-7942}™™" (©) All displacements are divided by two; element forces are the same as in part (a). 0.0690 in. ; s* = 33.33 kips (c) —0.020 in; s® = 19.66 kips (c) 0.000 in.; s** = 0.0 kips 0.084 in.; ** = 29.8 kips (c) ‘2 =—0045in.; s* = 15.9 kips (t) (© U, = 0.00 in; s* = 35.4 kips (t) 2.32 (a) {% VY Us & U, Ve Uy, Vp= {104 -412 241 645 512 —441 000 —441} mm 39 KN (0); © = 41.0 KN (0); s = 36.2 KN (0); s#=450 KN (0); s = 27.1 KN (t); s* = 73.0 KN (©); s* = 68.4 KN (t); s*! = O.0kN Minh Ke Us Ve Ue Ye Ur Vi= {0.38 0.00 0.90 —1.65 0.00 —088 0.00 —0.88} mm 1 KN (0); sf = 13.6 KN (t): s¥ = 17.0 KN (: = 85 KN (t); 7 =0.0kN Chapter 3 3.2 P,q = 2.75 kips; Px, = 1.00 kips; Py, = 0.25 kips 3.4. P,, =7.5kips; Py, = —22.5kips; P,. = 105.0 kips: P, = 40.0 kips 3.6 sh —27.9kN(c) 3.8 = 75 kips(c)320 Appendix E ANSWERS TO SELECTED PROBLEMS 4.25 ©, = IM,/24E1; mg? Ma/6; ig? = My/2; ti? = My/4 4.27 @, = —1.441 x 10-4 rad (ia iy iy} = B12KN- 332KN-m —10.24kN-m} (3 my mm.) = QL84KN 10.24 KN-m — 18.88 kN: m) 4.29 (a) —99k -113k-f —135k-f} 13.5k 135k-ft 135k-ft} {a May i) {Dy Thy fi yee ye V, W,} ={0.076 0.093 0.103} in 93 kips (t) 5.4 s = 28.28 KN (t); s™ = sf = 28.28 KN (c); sf = 20 KN (1); all other element forces zero 5.6 sit? ~ 1.94 kN (0); 51"? — 0.65 KN (0); s*14? = 2.83 kN (0); S83 = 3.64 KN (c); $7" = 0.94 KN (1); 999? = 3.64 KN (0); 0992 = 96393 = 0); st = 3.75 KN (0) 5.8 (W, We We Wy Mis) = {-0.134 -0.105 —0150 -0.168 —0.168} in. 5.13 (W, We} ={-0.549 —0:825} in, 5.15 {0,, ©, @,,}1073 = {5000 —0.865 —6.667} rad {Pre Psa Peo sa Mig Mag) = {0 54 50 30 —67 283} (kips and kip- in.) 5.17 {V, Ox Ge) {0508 in. 0.00365 rad —0,00480 rad} 5.19 (a) (%, O4 On Ke Ob= 10-°{—2.257 1.1285 —0.871 —4.128 1.128} (mand rad) i]! [s0 7 f s0 7 [so 74P 50 76 fig |_| 171 co | fo 20 | versa my | ~Jiso |? | uma}? | aa 15.0 an m) Js; 0 13.29. 13.29. 0 (local 5 axis coincides with y axis) 0) (% On On KO} = 10°*{-2.708 1.354 —0.697 4.092 1.354} (m and rad) BY [60 ]" 60 ]* 60 ]* 60 Je ms 1.37 0 0 —137 ia | “130 ]* | 137} * | asr | * | aso | &N and kN fits ° 7.63. 7.63 0 (local j axis coincides with y axis)‘Appendix E ANSWERS TO SELECTED PROBLEMS 321 Chapter 6 6.1 x, -4£[ 3 2 Us] _ 2PL[2 4L[-2 6! [wy ali 6.2 U,=—Kjg'K,,U,, where U,-{U, %} and U,-(U, ¥%} U]__AE[ 07929 -02071] L [-10 00] PL [4.828 v.J~ ~ TL |-0201 07929] 4E| 0.0 0.0} 4E | 2.000 PLE 3.828 “AE | -1.100, 6.3. Place the hinge on element ab. Therefore, Ky, = 4 x 10* KN/m; Py = 16 kN-m; @,, = —4 x 10-* rad; (@,_ = 1.7578 x 10-* rad if hinge is on be). Element forces: {j, 7} ={619kN 8.44 kN-m)} and {18KN OkN-m —24kN-m} 6.5 Place the hinge on element ab. {Ke @.} = {0.053 in, 0,000 rad}; p,, m, = —400 kip in. 6.7 U,=U, = PB/I2EI; m, = PL2 6.9 See Prob. 4.27. 6.10 See Prob. 4.29(a). 1333 kips; 61 U, 1.832 —2992 0.168] -, vey io“ -1.160 6962 0.840 a UU 0.168 3008 1.832 ||,” Vy —0840 7038 1.160] 6.12 See Prob. 2.12. 6.13 See Prob. 2.13. 6.15 See Prob, 2.26. 6.16 See Prob. 2.32. 6.17 Geometrically unstable, but truss can support loading because equations are con- sistent 6.20 {U, V, Vj) ={0 0 4.53 mm);s" 0; s* = 725 KN (t) 6.21 Ny = 2Ax/L)? — Hx/L)? + 1; By = by(Ax/L) — 1/2: ky, = 40SEbA?/D Chapter 7 7.1 See Prob. 2.12. 7.4 See Prob. 4.5 7.5 See Prob. 4.9. 7.7 See Prob. 5.2.INDEX Accuracy, 298, 299 Active column, 294, 298 Active column solution. See Skyline reduction ‘method Active degrees of freedom, 288, 297, 298 Actual behavior, 137 ‘Actual displacements, 87, 89, 264 Admissibility, 87 Alterrating symbol, 306 Approximate solution, 93, 95, 114, 193, 241, 242 ‘Archimedes, 2, 260 Argyris, J. H., 2 ‘Articulated structures, 6 Assemblage of elements, 6 13, 15, 24, 107, 113, 114, 134, 135, 137, 144, 251, 253, 254, 263-265, 267 Assembly of element matrices, 18, 22, 23, 32, 56, 108, 132, 135, 137 ‘Augmented matrix, 184, 185 kinematic, 281 ‘Axes, See Coordinate systems Axial deformation, 8, 9, 83, 84, 86, 122, 157, 166 rigidity, 122, 162,'168 Axial force element, 81, 82, 99-101, 157, 159 assemblage of, 14, 19 flexibility matrices, 251, 266 force transformation matrix, 39, 48, 111, 114 forces, 38, 48, 49, 111, 114, 167 initial forces, 47, 112, 13, 166 stiffness matrix, 8, 9, 12, 13, 37, 95, 106, 179 tapered, 87-89, 240-242 three-node, 97, 102, 241 Backsubstitution, 287, 291 Bandwidth. See Half-banéwidth Beam element, 122, 126, 127, 128, 132 distributed loading, 142, 148 element force matrix, 131, 145 flexibility matrices, 257 idealization, 136 initial force matrix, 154 sign convention, 132 stiffness matrix, 125 strain energy, 130, 131 tapered, 243-245 thermal distribution, linear, 153, 154, 155 Beam element in three-dimensional space, 186 bending about the J axis (in X-£ plane), 187 stiffness matrix, 191-192 Bending moment, 131 Bending strain energy, 130 Bernoulli, Daniel, 2 Bernoulli, Johann, 2, 75 Bilinear form, 311 Boundary conditions, 32, 33 kinematic, 92, 93 Boundary element, 233 Castigliano’s theorem (part 1), 3 stiffness matrix from, 129-131 Castigliano’s theorem (part 11), 2, 3 Cholesky decomposition, 296 Clapeyron, B.P.E., 2 Compatibility methods, 2, 3, 32, 53, 251, 281 ‘Complementary external work, 261 Complementary strain energy. 261 Complementary virtual strain energy, 262, 265 ‘Complementary virtual strain energy density, 263 Complementary virtual work, 2-4, 260-265 external, 262, 264, 265 internal, See Complementary virtual strain energy principle of (theorem of), 253, 256, 260, 262- 263, 264, 266, 267, 274 Complete polynomials, 95, 242 constant strain, 242, 245 rigid-body modes, 242, 244 Condensation, matrix, 213, 214, 217, 221, 288, 297 of stiffness matrix, 213-216, 221, 226 Conjugate vector, 104, 268 Connectivity of elements, 33, 38 Consistent displacements, 2 Consistent equations, 184, 186, 279 Consistent forces, 143, 148 Constrained degree of freedom, 24, 33, 270 element, 272 nodes, 24, 25, 110 structure, 27, 57 Constraint, 224-227 equations of, 165, 224-227 force transformation, 226 Constraint forces, 34 Contragredient transformation, 104-106, 180, 198, 224, 226, 236, 237. 268, 269, 311 Convergence, 242-243 compatibility of elements, 243 completeness, 242 Coordinate systems clement, See Local slobal, 9, 34, 37, 103-107, 113 local, 9, 34, 103-107 nodal, 34, 103, 224, 227-231 nondimensional, 97, 99 structural. See Global Coordinate transformations, 223-224, See also Transformation for constraint equations, 225 323INDEX Frames plane, 157-168 space, 201-203 Frame element element force matrix, 160 flexibility matrix, 260 force-transformation matrix, 160 sign convention, 159 stiffness matrix, 158 Frontal solution, 296-298 Galileo, 2, 75 Gauss elimination, 213, 276, 277, 279, 285-290, 297, 298 Generalized coordinates, 93, 94, 103 Generalized displacements, 122, 127, 158, 159 Generalized forces, 122, 131, 132, 138 release of, 215 Global coordinates, 9, 34, 37, 103-107, 113, 158, 159, 179, 193 Global flexibility coefficients, 253 Global flexibility matrix, 256 Global statics matrix, 269, 270 Global stilTness coefficients, 38 Global stiffiness matrix, 32 Grid structures, 196-200 Grid clement foree-transformation matrix, 198 stiffness matrix, 197 Half-bandwidth, 293, 298 Heated structures. See Thermal effects Hinge, internal, 215-217 Hooke's law, 6, 7, 100, 128, 252, 266 Idealization of structures, 121, 136, 137, 169, 299 M-conditioning, 163, 182, 184, 233 Indeterminacy kinematic, 52, 53, $4, 270 Static, 46, 49, $2, 53, 54, 251, 270 Initial deformations, 273-275 Initial forces, 113, 114, 145, 150, 153, 200 using virtual work, 153 Initial force matrix element, 47, 112, 113, 134 global, 48, 113 Initial strain, 42, 46, 49, 112, 150, 153, 204 Interface nodes, 220, 221 Internal strain energy, 89, 93 Internal virtual work, 83, 87, 100, 134-135. See also Virtual strain energy torsional element, 191 Kelsey, S., 2 Kinematic admissit y, 87 325 Kinematic boundary condition, 92, 93 Kinematic equations, 280 Kinematic indeterminacy, 52, 53, 54 Kinematic matrix, 24, 270-272, 276, 279 Lame, G., 2 Lamella roofs, 201 Linear algebraic equations, 308-310 Loads between nodes, 144, 148 equivalent nodal, 143, 148 uniformly distributed, 142-144, 148 Local coordinates, 9, 34, 103-107, 157, 159, 179, 193 Matrices addition of, 302-303 ¢ law of addition, 303 associative law of multiplication, 304 commutative law of addition, 308 commutative law of multiplication, 304 conformability for multiplication, 304 determinant of, 306-308 differentiation of, 310 distributive law of multiplication, 304 equality of, 302 integration of, 310 inversion of, 308, 309 minor of, 307 multiplication of, 303, 304 order of, 302 partitioning of, 306 rank of, 309 scalar multiplication of, 303 submatrix of, 305, 307 subtraction of, 303 Matrix adjoint, 308 augmented, 184 banded, 293, 294 coefficient, 308, 309 cofactor of, 307, 308 column, 302, 311 diagonal of, 302 nonsingular, 308 notation, 302 null, 303 orthogonal, 36 row, 302 singular, 308 skew symmetric, 305 square, 302 symmetric, 305 transpose, 305 triangular, 287, 290, 291, 300, 308 unit, 304 Matrix structural analysis, 1, 3, 165 Maxwell, J.C. 1,2326 INDEX Maxwell-Mohr method, 2 Maupertuis, 2 Member. See Element Member forces. See Element forces Merge. See Assembly Mohr, O., 1, 2 Moment-curvature, 123, 128, 244 Moment distribution, 157, 163 Moment-resisting connection, 236 Nodal coordinates, 34, 103, 224, 227-231 Nodal displacements, 24-25, 110-111, 132 Nodal equilibrium, 14, 107, 110, 113, 114, 132, ns solving equations of, 25 Node, 6 free (unconstrained), 25, 110 interface, 220, 221 interior, 220 master, 234 offset, 224, 233, 234, 239 slave, 234 supported (constrained), 25 Notation, xi-xiv Numerical accuracy, 299 precision, 298, 299 round off, 299, 301 Offset connections, 236-239 Orthogonality, 36, 37 Orthogonal matrix, 36, 195, 196, 224, 229 Partitioning of matrices, 25, 26, 213, 221, 225, 230, 271, 276 Pivot elements, 290, 292 Plane frame element, 157 transformation matrix, 160, 167 i force matrix, 166 stiffness matrix, 158 Plane rigid frames, 157-168, 224 Precambering, 45, 52, 153, 166 Prestrains, 273, Primary structure, 257, 263 Quadratic form, 311 Rank of matrix, 184, 186 Rayleigh-Ritz method, 3 Reaction forces, 25, $3, 87, 111 Real displacements, 87, 89, 90, 92 Redundancy external, 254 internal, 254 Kinematic, $4, 271 static, $4, 270, 271 Redundant displacements of, 252, 267 forces, 251, 284, 256, 257, 259, 260, 267 solution for, 253 Redundant force, automatic selection of, 276-281 References, 312 Release of generalized force, 215, 217 Rigid body, 78, 79 Rigid-body displacement, 97 Rigid-body motion, 78, 242, 244 Rigid link, 234-238 Round off, 299, 301 St. Venant torsion constant, 189 Self-straining, 150, 155, 165 Shape functions, 95-97, 101, 127, 241, 244 complete polynomials, 95 Shear, 131 Shear center, 234 Sign convention for axial deformation, 252 for axial force, 29 for bending moment, 12 for redundants, 256 for shear, 124 Significant digits, 299-301 Simultaneous linear equations, 308-310 Singular matrices. See Ill-conditioning Skyline, 293, 295 Skyline reduction method, 293-296 Slope-deflection equations, 157, 168 Solution accuracy, 298-301 by decomposition, 298-301 by elimination, 289-293 process, 25 Space frame clement stiffness matrix, 192 Space iruss element force-transformation matrix, 181 Stability, $3, 184, [86, 279 Static condensation, See Condensation, matrix Static indeterminacy, 46, 49, 32, 53, 34, 251, 270, 274 Statics matrix, 272 Stiffness coefficients, 13 Stiffness equations. See Force- ‘equations Stiffness matrix axial force element, 8, 13, 38, 106 beam element, 125 element using virtual work, 95, 101, 127, 129 frame element, 158 global using virtual work, 134-135 structural, 16, 107-108 iapered axial element, 240-241 tapered beam clement, 243-245 three-dimensional beam element, 191-192 isplacement