AFFDL-TR-67-194 AN ITERATIVE METHOD FOR THE ANALYSIS OF LARGE STRUCTURAL SYSTEMS V. B. VENKAYYA This document has been approved for publi release and sale; its distribution is unlimited.AFFDL-TR-67-194 FOREWORD This report is the result of an in-house effort under Project 1467, “Structural Analysis Methods,” Task 146705, “Computer Methods.” The work was carried out in the Advanced ‘Theory Group of the Theoretical Mechanics Branch, Structures Division, Air Force Flight Dynamics Laboratory, Wright-Patterson Air Force Base, Ohio, Mr. V, B, Venkayys, FDTR, was project engineer. ‘The repart was submitted by the author in October 1967 for publication, ‘The author expresses his appreciation to members of the Advanced Theory Group and the Analysis Group for their suggestions during the preparation of the manuscript, ‘This technical report has been reviewed and is approved, us Air Force Flight Dytamios Laboratory iAFFDL-TR-67-194 ABSTRACT An iterative method to analyze large structure! systems subjected to static and thermal loadings is developed, This method is essentially the displacement method of analysis in which the finite elements are grouped into convenient size substructures, The analysis proceeds with one substructure at a time, and a process of iteration establishes the con- tinuity of the system. The speed of convergence depends on the nature of the stiffness matrices of the substructures, To illustrate the method, two frame structures are analyzed. One is an 1600 degree of freedom system and the other a smaller system, The purpose of the larger one is to estimate the computational time involved, The results of the smaller system (by iterative analysis) are compared with those obtained from direct analysts, and the results are nearly the same, The main advantages of the iteration method are that the entire computational scheme can be carried out without exceeding the core capacity of the computer, and the method is more economical with respect to computational time than the direct analysis, ittCantracke Public ReleaseAFFDL-TR-67-194 TABLE OF CONTENTS SECTION 1 Introduction I Review of the Existing Methods 1, Analysis by Substructures 2, Tridiagonalization and Recursion 3, Iteration I Substructures and Iteration IV _Explanation of the Proposed Method V __ Griteria for Convergence VI ‘Results and Conclustons References PAGE 16 19 23AFFDL-TR-67-194 FIGURE 1 2 Ta Ib Ma 1m Ma 1b ILLUSTRATIONS ‘Typical Substructuring Schemes Plow Diagram for Iterstion on Substructures Direction of Positive Member Forces Nature of Convergence of Last Substructure Displacements Five-Bay, 100-Story Frame ‘Ten-Bay Gable Frame TABLES Five-Bay, 100 Story Frame: Joint Coordinates, Forces, and Displacements (Iteration) Five-Bay, 100 Story Frame: Member Forces (Iteration) ‘Ten-Bay Gable Frame: Joint Coordinates, Forces, and Displacements (Iteration) ‘Ten-Bay Gable Frame: Member Forces (Iteration) ‘Ten-Bay Gable Frame: Joint Coordinates, Forces, and Displacements (Direct Analysis) ‘Ten~Bay Gable Frame: Member Forces (Direct Analysis) vi PAGE 10 2B 38 20 a1 25 27 31 34 37 40AFFDL-TR-67-194 SECTION I INTRODUCTION In recent years there has been an upsurge in the development of methods to analyze large structures. Now it is possible with the aid of digital computers to anelyze quite large struc- tural systems. The displacement and force methods, in conjunction with the finite element approach, are firmly established as convenient methods of structural analysis, but as structural systems become larger and more complex, the discretization errors tend to be more prominent making the finite element analysis ineffective. This situation can be remedied, to some extent, by deriving an adequate element representation (Reference 1). At present, three types of elements are in use, each serving a different category of structures. A straight or curved element is suitable for frame structures, the plate elements are used for two- dimensional problems, and the tetrahedron element is proposed for the analysis of three~ dimensional problems. In any discrete representation of a continuous system the accuracy of the analysis improves as the representative mesh becomes finer. However, the finer mesh rapidly increases the number of degrees of freedom of the system. The large systems, handled as a whole, will strain the computational capabilities of even the largest computers. This is particularly so in the analysis of shell structures, It is possible to handle such large systems using auxiliary storage devices such as magnetic tape. Since the retrieval time when using such devices is excessive, the cost of computation may often be prohibitive unless storage devices are used with discretion, The proposed iterative method alleviates some of these problems by analyzing a part of the structure at a time, The method is simple, readily amenable for modification, and ideally suited for automated analysis. It can be used with equal facility for one-, two-, and three-dimensional problems if a suitable element stiffness matrix is derived, It is par ticularly convenient when the analysis has to be repeated a number of times with different loading conditions and minor configuration changes in some sections of the structure. The structure is divided into a number of substructures of convenient size, The analysis proceeds, sequentially, from the first substructure to the last. The compatibility of displace~ ments and equilibrium conditions between the substructures are established by a process of iteration, ‘There are a number of methods available in literature for the analysis of large structural systems (References 2 through 9). A brief review of some of these methods is presented in Section Il and a detailed discussion of the existing methods is in Reference 2, The theoretical basis and details of the proposed iterative method are discussed in Sections III through V, To illustrete the effectiveness of the iterative method, a 5-bay, 100-story building frame ‘and a 10-bay gable frame are analyzed, The results are presented in Section VI.AFFDL-TR-67-194 SECTION 1 REVIEW OF THE EXISTING METHODS 1, Analysis by Substructures In a finite element analysis, a structure is viewed as an assembly cf simple structural elements joined together at nodal points or joints. A simple element is one whose force- displacement properties can be determined, at least to a degree of approximation, by the known methods of analysis. The element stiffness and the stiffness of the structure with respect to its joint displacements are related by Equation 1 (Reference 10): Ke ya ha, a in which K and k; are the stiffness matrices of the structure and the it element, respectively; aj is the matrix that relates the i‘? element and structure displacements; and m is the total number of elements in the structure, The relation between the Joint forces and their displace ments is given by: R= Kr (2 Rand r are matrices of joint forces and their displacements, respectively. Since the size of the K matrix is directly proportional to the number of joints, the finer division of tho structure results in a larger K matrix, In a substructure analysis (Ref- erence 3), the simple elements are grouped into a fewer number of larger elements called substructures, The relation between the structure and substructure sti'finesses is given by Equation 1, except that k is now the stiffness of the substructure instead of the element, and now the size of K is smaller because of the fewer joints, With tho aid of simple elements, the force-displacement properties of the substructure can’ be determined prior to the assémbly of K, By using substructures, the sizes of the matrices to be inverted can be reduced to be a8 small.as desired: however, this reduction in size is normally accompanied by an increase in arithmetic operations on smaller matrices. As the size of the structures becomes large, the computational errors (round-off errors) in these matrix operations terd to be cumulative. , Tridiagonalizatien and Recursion In a finite clement analysis it is necessary to number the system coordinates and the element coordinates. By a proper numbering of the system coordinatas, one can obtain a banded stiffness matrix. Each row of this banded matrix usually consists of three elements or submatrices placed symmetrically close to the main diagonal. The remaining olements are all zero. A banded matrix can be inverted by the method of recursioa or partitioning, in which the entire matrix need not be handled at one time, One limitation of the tridiagonaliza- tion method is that aot all structures lend themselves for convenient numbering to obtain a banded matrix, 3. Iteration ‘The moment distribution method developed by Hardy Cross (Referenve 11) was the most commonly used iterative method before the advent of computers, A similar method, but in a form more suitable for computer use, was originally developed by Kani (Reference 5) andAFFDL-TR-67-194 subsequently improved by others (References 2, 6, and 7). Kani’s method is presented in excellent detail in Reference 2. Reference 9 prosents another iterative method that is con- venient for the analysis of rectangular frames. The method presented in this report is also an iterative method, but it is more general in that it is not limited to frame structures. SECTION II SUBSTRUCTURES AND ITERATION As in any discrete element analysis, the total structural system is divided into a finite number of simple elements, joined together at the nodal points. These elements are further grouped into substructures, each substructure consisting of an arbitrary number of simple elements (see Figure 1). For convenience, the following restrictions are imposed on sub- structuring: 1. A given sub-structure has common joints with at most two other substructures. 2, The substructures are numbered in sequence so that an it) substructure shares joints with (i-1)" and (i+1)t substructures only (see Figure 1). 3. The total number of joints in the first substructure should be more than the sum of the Joints shared with the second substructure and the joints on the natural boundary. In this report, the unyielding supports are referred to as the natural boundary. 4, The last sub-structure should have an adequate number of joints on the natural boundary to prevent rigid body motion. ‘The first two restrictions are not essential to the method but they help in orderly presentation and simplification of programming. ‘The force-displacement relations for the itt substructure can be written in the form: ao. ll ia Rr andr are the column matrices of joint forces and displacements, respectively, of the it substructure, and (0 is its stiffness matrix.AFFDL-TR-67-194 ‘SUBSTRUCTURES 4, Multistory Frame SUBSTRUCTURES b. Multibay Gable Frame Figure 1. Typical Substructuring SchemesAFFDL-TR-67-194 & 2 8 2 & 2 3 zB We AN ae WN SAS ZIV 2 [= FZ SONAFFDL-TR-67-194 ‘The joints of a substructure are classified into four categories. For instance an it” sub- structure may have: 1. Joints sharing with (i-1)"" substructure 2. Joints belonging to it substructure alone 3. Joints sharing with (+1) substructure 4, Joints on the natural boundary Since the displacements of the joints on the natural boundary are zero, they are not included in Equation 3. Equation 3, partitioned to reflect this classification, takes the form: oe 0 a (R), (1), ! (hg | Why a), ees pana as 4-8 Py] = | Wy | Khe | Ky (r «) (Rs (gy | (K)gg | (Ks Cs ‘The superscript refers to the number of the substructure, The parentheses () indicate that the quantities are submatrices. The subscript refers to the three-fold partitioning indicated, ‘The submatrix (2), is the displacement matrix of the Joints common to ¢i-1)"™ vans ow structures, Similarly (r),") is the displacement matrix ofthe joints common to 1" and (1+1)"* substructures. (2), is tho displacement matrix of tho joints belonging to 1" substructure alone, Equation 4 is rewritten as three separate equations: oe odd cd + oc el (50) ee = los aed s «Bil (so re = eel + Bete + aol (Se The forces (R),", (ip, and (R)y” are given the following interpretation: the (R), forces are the sum of the external forces applied at the respective joints and the reactions 3") transmitted from (i-1)® substructure. The submatrix (R),") is simply the applied crAFFDL-TR-67-194 forces at the respective joints. The forces (R),") are the reactions from the (1+1)"™ sub- structure to the i" substructure. (R), does not include the applied forces at the respective jointe, With this interpretation, Equations Sa and Sb can be written in the form: wt G i i“ cht we cL fee} Jer oc, -- = [---4-|---L EE --4 tr) 16 wwf t¢ F, and F." are external forces applied at the respective joints. From Equation 6c, (R)3(")) cP cg cf? + 0? ccf? m ‘The substitution of Equation 7 in Equation 6 yields: AD AD _ lO fion ia ‘The notation in Equation 8 is as follows: can be written as: i od? tof] dt 0 ; xk) tad | | 0 In deriving Equation 8 from Equations 6 and 7, the displacement matrices (1)3"") ‘and (r)g) are replaced by (2),") and (r);""2), respectively. This is acceptable because they represent 7AFFDL-TR-67-194 the common joint displacements. Also some of the matrices are expanded by adding rows and columns of zeros without altering the meaning of the equations, For a given structure and loading, the matrices Cy"), cp", cg", and cg" are constant, Actually, the latter three matrices do not depend on the loading. Equation 8 forms the basis for writing the recurrence formula for iteration. According to this equation, to calculate the ith substructure displacements, the displacements of the sub- structure before, the substructure after, and its own displacements are required. Since it is assumed that in each cycle, calculations start from the first substructue and proceed to the ast, (-1)" substructure displacements are known at the time the i” substructure displace ments are calculated, The displacements r“) and r*), on the right sice, are obtained from the previous cycle. Then the recurrence formula for iteration takes the form: og] fo cn a_i sien me Bet 3 Cy ‘The subscripts on the displacement matrices refer to the cycle of iteration, i.e., to calculate +1)" cyoie values of x, the qe+2)" cycle values of r'"!) and the k*” cycle values of x") and r'*)) are used. Equation 9 forms the mathematical basis for iteration, For the first substructure, the second term on the right is elways zero. Similarly the fourth term does not exist for the last substructure, The actual method can be outlined by referring to Equation 9 or Equations 6 and 7 which are all derived from Equation 5. The method using Equation 9 has the following steps: 1, The iteration starts with the assumption of zero displacements for all joints, This starting cycle may be considered as the zero cycle, 2, The first cycle displacements, rj), 1%. ----------- 1, are calculated, successively, with the aid of Equation 9, Since the displacements of the previous cycle are all zeros, the last two terms of Equation 9 are zeros for all substructures In this cycle, The old values of the displacements are replaced by thedisplacements calculated in the present cycle. 3, The displacements of the subsequent cycles are caloulated by repeating Step 2. The iteration zay be continued until the absolute difference between the displacements calculated im two successive cycles is sufficiently small, Convergence cf the method depends on certain conditions discussed in Section V. Equations 6 and 7 can also be used as the iterative approach to cornpute programming. The iteration starts with the first substructure: 1, Assuming the displacements of the common joints between first and second substruc- tures to be zero, the remaining displacements of the first substructure ave calculated by Equation 6, 2. ‘The reactions on the second substructure are calculated by substituting the displace- ments determined in the previous step in Equation 7.AFFDL-TR~67-194 3, The reactions and external forces on the second substructure are combined. 4. Steps 1 through 3 are repeated for each substructure, always replacing the old dis- placements by the newly calculated displacements, One cycle of iteration is completed when the procedure is repeated on all the substructures. 5. For subsequent cycles, Steps 1 through 4 are repeated except the displacements calculated in the previous cycle are used for common joint displacements instead of zeros. Iteration is continued until the difference between thedisplacements calculated in two succes- sive cycles is sufficiently small. The author found the latter approach to be more convenient for computer programming than direct use of Equation 9, In Section IV, the method is presented in detail for computer use, The presentation includes determination of element forces from the joint displacements obtained by iteration, SECTION IV EXPLANATION OF THE PROPOSED METHOD The method is briefly outlined in Section Illand is further explained here in a suitable form for computer use, The flow diagram shown in Figure 2 summarizes the basic steps involved in iteration and each step is explained in the following paragraphs: a, Input Input information is similar to any discrete element analysis but includes information pertaining to substructuring. (2) Structure Information Give the total number of elements, joints, substructures, natural boundaries, etc, (2) Substructure Information List the number of elements and joints in cach substructure, the joints shared with the neighboring substructures, and the joints on the natural boundary, In addition, list the first and last elements and joints for each substructure,AFFDL=TR-67-194 INPUT 1 STRUCTURE INFORMATION 2 SUBSTRUCTURE INFORNATION ‘3 LOADING INFORMATION £ INITIALIZE JOINT DISPLACEMENTS INITIALIZE SUBSTRUCTURE STIFFHESS, KOYCLE = [acini nee ASSEMBLE STIFEWESS OF 1™ SUBSTRUCTURE —— INCORPORATE KATURAL BOUNDARIES | oo PARTITION STIFFNESS MATRIX | Le Ee FIND INVERSE OF [x] | Py Teen} BRING 1™ SUBSTRUCTURE INFORMATION DETERMINE DISPLACEMENTS! Titae t DETERMINE REACTIONS OW (+1) SUBSTRUCTURE [COMBINE REACTIONS WITH FORCES ON (1+ 1) SUBSTRUCTURE RESTORE DISPLACEMENT DETERMINE ELEMENT (QUTPUT J. ELEMENT FORCES: MATRIX TO FULL SIZE FORCES 2, JOINT DISPLACEMENTS: Figure 2. Flow Diagram for Iteration on Substructures 10APFDL-TR-67-194 {8) Joint Coordinates Give the coordinates of each joint (two in case of plane frames and three otherwise) with reference to a convenient set of coordinate axis. (4) Element Information For frame members give the relative values of EI and AE and for plate elements, give E, v , and thickness, E and y are elastic constants and 1 and A are the moment of inertia and cross-sectional area, respectively, (5) Loading Information Give joint forces if they can be determined directly or make provision to determine the Joint forces. A subroutine can be written to determine the joint forces. yb, Initialize Joint Displacements For the starting cycle, assume the displacements of all joints to be zaro, This cycle may be regarded as zero cycle. c, Initialize Substructure Stiffness Matrix Allot a block of memory space, large enough to accommodate the stiffness matrix of the largest (in terms of number of joints) substructure, Initialize this memory space to zeros before starting the assembly of the stiffness matrix. d. Assemble Substructure Stiffness Matrix Assemble the substructure stiffiess by adding successively the stiffness of each element till all the elements of that substructure are completed, This step constitutes the important part of finite element analysis and is explained in two parts, (1) Element Stiffness Matrix First, determines the clement stiffness matrix with respect to its own coordinate system, The procedure for determining stif mess of standard elements such as frame members and triangular plate elements can be found in many references on matrix structural analysis (References 2, 10 and 12), As an example, the stiffuess matrix of a plane frame member with constant section is given in Equation 10, The numbering of the element coordinate system is shown in Figure 3, AE/L oe oe symmetric ° TELS. Vaasa [i]: ° eee ae i -AE/L ° ° AEM. ° seve -6ei? ° evs ° eer? = 2E1/L 0 -6e1® ae 1AFFDL-TR~67~194 JOINT NUMBER MB MA < MB JOINT NUMBER MA SYSTEM COORDINATES Figure 3. Direction of Positive Member Forces (2) Transformation and Addition ‘Transform the element stiffness matrix from its own coordinate system to a general co- ordinate system before adding to the structure stiffness. Both these operatiors are accom= plished by addition of matrix, to the structure stiffness. Yj is given by: Y= a! ky oj a ky ts the it element stiffness matrix with respect to its own coordinate system, The dis~ placements of the {'P element and the structure are related by the a, matrix. The a, matrix is of the size pxq, where p and q are the dimensions of element and structure stiffness matrices, respectively. Since q is usually largo, a, 1s also large but very sparsely populated, To avoid storing the entire a; matrix, the transformation and addition are carried out ‘i separately in the following manner: First, the element stiffness matrix is transformed by the relation Yo wp By a) 12AFFDL-TR-67-194 Now, % is the transformed element stiffness matrix and is the same size as the original matrix, The matrix p; for a plane frame member Is given by: By 413) A, Is the direction cosine matrix of the element coordinate system with respect to structure coordinates. For a plane frame member, , is given by: cos8; 0 0s) 6, is the inclination of the member ineasured from the x-axis of the system coordinates, The transformed matrix yj (Equation 11) is partitioned into four quadrants (plane frame) as shown in Equation 15. 5) If the plane frame member connects the joints MA and MB (Figure 3), the first quadrant elements of Yj are added to the structure stiffness elements, starting with address (SMA, 3 MA). The second, third, and fourth quadrants are added to the elements starting with the addresses (3MA, 3MB), (3MB, 3MA), and (3MB, 3MB), respectively, Equation 16 shows the location of ; in relation to the general stiffness matrix. The cross marks (+) indicate the locations to which yj is added. [K] is the structure or substructure stif mess matrix being assembled, e, Incorporate Natural Boundaries Introduce the natural boundaries by eliminating the rows and columns of the stiffness matrix corresponding to the displacements on the natural boundary. The resulting stiffness matrix is a condensed matrix in relation to the one assembled in the previous step. Another procedure to introduce natural boundaries is to substitute zeros in place of the off diagonal elements of the rows and columns corresponding to the boundaries, and ones for the diagonal elements. This way, the size of the matrix remains the same and time saved in condensing the matrix is expended in operating with larger matrices. 13AFFDL-TR-67-194 1 2 3 3MA 3MB SVOINTS) ama (voit f, Determine Displacements 1, 70) is the displacement matrix of the 1" substructure in the &! eycle of iteration, De~ termine It using Equation 6, 7," does not include the displacements of the common joints between 1 and (i+1) substructures. However, to determine 1,"), the displacements of the common joints are necessary (see Equation 6). The common joint displacements are cal culated, in the <1)" cycle, as part of (i+1)'" substructure displacements, and they are used to determine r,", it 1s important to replace r(), by 7,(") so that for the next cycle the latest values are available, g. Determine the Reactions on (i+1) substructure and Combine ‘Them With the Applied Forces Substitute the displacements 1," and the common joint displacements from the previous cycle in Equation 5c to obtain the reactions on i" substructure from (i+1)" substructure. ‘The same reactions, but in opposite direction, are transmitted from i to (i#1) substructure, These reactions are combined with the extemal forces applied on (i+1) suastructure, 4AFFDL-TR-67-194 Steps c through g are repeated on all the substructures, At the end of last substructure one cycle of iteration is completed. The procedure is repeated until the difference between the displacement matrices calculated in two successive cycles is relatively small. The criterla for convergence and some modifications to the iteration process are dis- cussed in Section V. At the end of iteration only the structure displacements are available. Often it is necessary to determine element forces in addition to the displacements, The next two steps indicate the procedure for calculating element forces. h, Restore the Displacement Matrix to Full Size The displacement matrix determined by iteration 1s a condensed matrix because it does not include the joints on the naturel boundary. These joints have zero displacements. Expand the condensed displacement matrix by introducing zeros in place of natural boundary dis- placements. The enlarged displacement matrix is used to determine clement forces in the next step, i, Element Forces Determine the i'* element forces by substitution of the displacement matrix in the equation: «7 ‘The matrices kj and a; are the same as defined in Section II, Sj is the matrix of element end forces and r s the full displacement matrix. Storage of a large a, matrix can be avoided by using Equation 16 in place of Equation 17: Sf kar (a) #; is previously defined in this Section, The matrix r' consists of only the displacements of the joints connected by the element, Equation 18 gives the element forces due to joint dis- placements only. If there are forces acting directly on the element, 8, is given by: S = keair- SP 180) s? is the matrix of fixed end forces directly acting on the element, J. Output ‘The type of output depends on the problem and, in general, includes the element end forces and/or stresses, the joint displacements, and the applied forces. 15AFFDL-~TR-67-194 SECTION V CRITERIA FOR CONVERGENCE In Sections Ill and IV it was tacitly assumed that iteration using Equation 9 (or 6 and 7) converges, However, convergence depends on the nature of matrices C,", c,"), c4"), ana cq"), The matrix Cy) is a function of the properties of the structure ard the loading, while Cyt, Call), and C,() depend only on the properties of the structure, Each of these matrices is a product of two matrices, Thefirstmatrix, In all four cases, is the inverse cf the stiffness matrix with respect to the displacements r“) (sce Equation 6). Since the elements of this inverse contain the term “EP” in the denominator, they are small compared to 1. For stable elastic structures, it is reasonable to assume that the elements of C;"), cy), 05", ana ,Y) are small in magnitude, compared to unity, These comments on the nature of the elements of “©” matrices help in arriving at certain conclustons rogarding the nature of convergence. In this discussion, the elements of a given structure are grouped into three substructures, ‘This restriction is necessary only to simplify the discussion. The conclusions reached with this assumption are applicable when there are more thin three substructures, The x" cycle displacement matrix of the if” substructure, according to Equation 9, is as followe: gf fen got cen Oe aD eD feO— cD lt o Applying this relation to the fixst substructure, the k" cycle displacement matrix is written as: Os D- Pay i The constants Cy and C, are zero for the first substructure, For the seccnd substructure the displacement matrix takes the form: e i _ 2, 2P_ Py " (0) mc? eH? led Bu where @ . o@)_ gle wes cs = cp gy In the expression for 1), only r,) andthe displacement matrices of the thir. substructure are retained, Ifthe elements of 'C” matrices are small compared to unity then the elements 16APFDL-TR-67-194 of their products would be even smaller. In such a case, it is reasonable to neglect the terms containing higher products, For convenience the expression for r,,) is rewritten in the form: 1, Fe, nil] & Pane fF & (3) +. 2. + APP 2) hg £2) a) ee, is) (Sc) 3 3 where h¢x)") = products of the order greater than x)”, with a similar notation the third Substructure (last) displacement matrix is written as: (Oe [rare oF yar Bde n nod ]@ + [oP x2 2h - 9 WE am BA] (94) where (3) 2 of) (3) 12) Pee ae altar E the products of the order x)” and higher are neglected, the displacement matrices 14) and 23°) would de the same, to the assumed order of approximation, Similarly, ifthe aus of the order x@” and higher are neglected, it can be shown, from Equation 9c, that rg) and r5@) would be the same, If the same reasoning is extended to the first sub structure, tho displacements 1, and xg) axe approximately equal, The conclusion from this discussion is that if three oycles are necessary to obtain the approximate displacement matrix for the last substructure then at loast n + oycies are necessary to obtain similar approximation for the first substructure displacements (a is the number of substructures), As an example to filustrate the mode of convergence, tle displacements of the last sub- structure are plotted against the number of cycles in Figure 4, The iteration approaches the true value quite rapidly in the beginning (within 10% of the true values in less than four cycles), then the rate of approsch slows down considerably. It is evident from Figure 4, that the iterated values do not even remain on one side of the true value, Instead, they oscillate about the true values with decreasing amplitude, Figure 4 is a qualitative representation of the iterative process; the details vary with the structure. It is evident, therefore, that convergence starts from the last substructure and proceeds to the first sequentially, To take advantage of this property, a modification is proposed to the iteration scheme. In the original scheme each cycle of iteration starts from the first sub- structure and proceeds to the last successively. In the modified scheme, however, the cycles of iteration are classified as full cycles and truncated cycles, i.c., (a) a full cycle of iteration starts from the first substructure and proceeds to the last as in the original scheme, and () the truncated cycle consists of ‘‘n’? subcycles. The first suboycle is a full oycle, The second subcycle ends at {n-1)"" substructure, the third subeycle stops at (n-2)"™ substructure, urAFFDL-TR-67-194 ITERATED DISPLACEMENTS: ACTUAL DISPLACEMENTS i 2 3 4 #5 6 7 8B CYCLES OF ITERATION Figure 4, Nature of Convergence of Last Substructure Displacements and so on until the n'® subcycle ends at first substructure. After three or four full cycles, the displacements of the last substructure have attained 2 measure of stability. Now a truncated cycle is introduced. At the end of the truncated cycle, the displacements of all the substructures are within 10 to 15% of the actual values, ‘Then, the iteration is continued with full oycles. The latter step smooths out the iteration process after: the truncated cycle, In the modified scheme, the displacement matrix obtained in every cycle after truncated cycle is a reasonable approximation, To obtain « similar approximation, at least 2n cycles are necessary in the original scheme. The author adopted the modified scheme in which the first three are full cycles, the fourth is truncated, and the subsequent ones are full cycles, 18AFFDL-TR-67-194 SECTION VI RESULTS AND CONCLUSIONS ‘Two frame structures were selected to illustrate the iterative method, The first structure 4s a S-bay, 100-story frame shown in Figure § and analyzed for the following three different loading conditions: +. 595, 1, Lateral loads (in the direction of positive x-axis) on the joints 1, 7, 13, 20, 2, Moments, on all the joints, in the clockwise direction, 3. Combination of the first two loads, Table la gives the displacements of the joints on floors 100, 50, 10, and 1. For each joint there are three lines of displacements resulting from the three loading conditions, Table Ib gives the internal forces of the members on and under floors 100, 50, 10, and 1, Once again each member has three lines of internal forces resulting from three loading conditions. The total computational time for the three loading conditions is approximately 22 minutes on IBM 7094, During this time, in addition to the displacements of the 600 joints, internal forces of the 1100 members are caloulated, The estimated time for each additional loading condition is less than a minute, The second structure is a 10-bay gable frame shown in Figure 6 and also analyzed for the three different loading conditions defined in Figure 6, Tables Ua and Ib give the joint displacements and the member forces, respectively, for this frame. The computational time, excluding time of compilation, is about 30 seconds, The increase in time for an additional loading condition is almost too insignificant to report, For the purpose of comparison, the results of direct analysis for this frame are presented in Tables Ila and IIb, I is evident from these tables (iI and II) that results of iteration are not too different from those obtained from direct analysis. The computational time for direct analysis is about 22 seconds, which is less time than for iteration. However, for larger struotures this difference in computa~ tional time becomes increasingly in favor of iteration. The following two examples attest to this fact, The computational times for a 20-bay gable frame (analyzed for 3 loading conditions) are 1 minute and 68 seconds and 2 minutes and 20 seconds, respectively, for iteration and direct analysis. A more significant difference is shown in the analysis of a S-bay, 10-story frame (180 degrees of freedom). The direct analysis of this frame for three loading condi- tions requires 5 minutes and 30 seconds. The same frame is analyzed by iteration in 1 minute 10 seconds, In developing utilitarian programs, the computational times can be reduced further with increased effort in programming, Stiffness matrices are, in general, sparsely populated and the computational time can be reduced by avoiding operations with zero elements. two or more substructures are similar, the time can be reduced by avoiding repetition in assembly of matrices wherever possible. The author has not taken advantage of these properties in analyzing the two frames. A more significant change in computational time can be effected by balancing the size and number of substructures. The time for inversion of matrices increases rapidly with the size of substructures while the number of cycles of iteration are proportional to the number of substructures. By relaxing the first two restrio- tions on substructuring (see Section II), the substructures can be reduced to the con- venient size, Now r““)) mnay be interpreted as the displacement matrix of the substructures 19APEDL-TR-67-194 Losin 1: LATERAL ronees oMtr (>) aE LOADING 2: MOMENTS ONLY (10P) fee toaoino 8: COMBINATION OF OTH USE, THIS FIGURE MITH TABLES Te BNO T> tle a ag MEMBER NUMBERS is ae ,X/e 1g 13. @, & carom ‘SUBSTRUCTURE 7 ¥ ¥ iB 6 f 2 hu w Bs WP Se pe -—)—_ jt —} 22 + ie it a i 2 let 2 + + + + 2 G 2— 2 2 > ge +} —? Feat t O i + > » as Peace >—_-—_}} # aoa Cael > as 8 +t Peep Dae 8 G > iz z U > ss 3 +— ee eee ees ee 2— > + b -—_ >—_ +} * 2 > > Z cs > 2 ir tea Pee > er le is e ie . > eusstwueneiao —"]yg Perec pas 5@ 40°0"s 200-0' | Figure 5. Five-Bay, 100-Story Frame 20AFFDL-TR-67-194 ourery a1qup Seg-ues, “9 amg “amt ONY om ‘amr ‘ort S379¥L HLIM uNOLd SIL 3501 SOVOT OM 3HL 40 NOILYNIGWOD :€ ONIGVO? (401) AINO SLN3WON 2 ONI¥OT (NMoHS NOWLOBUI HL NI}d “SovOT TVOLLABA :1 ONIGWOT a 107,08 © o SUAENNN LNIOr- SH3QWAN 43ENW-AFFDL-TR-67-194 around and before i** substructure. Similarly r“*) is the displacement matrix of the substructures around and after i substructure. The word “‘around’’ refors to the substruc~ tures sharing joints with i substructure, “Before” and “after refor to the sequence of analysis. In conclusion, the advantages of the iteration method are: 1, From the flow chart (Figure 2), it is evident that the method is simple and demands Little sophistication in programming. All computations can be carried out without exceeding the core capacity. 2, Since the recurrence formula (Equation 9) was derived for a general structure, the method is applicable to plate and shell structures as well as frames, 3. Since an approximate result is available at any stage of iteration after the truncated cycle, the method can be used to great advantage in preliminary design. 4, ‘The increase in computational time for additional loading conditions is a small fraction of the time required for one loading, 5. The effects of minor configuration changes can be localized without drastically altering the tot#l computational scheme. Because of these advantages, the method is particularly attractive for use with structural synthesis programs, The current limitation is that only static and therraal loadings can be handled, Another method with similar capabilities to handle dynamic loading ‘s therefore necessary for the analysis of large structural systems, 22AFEDL-TR-67-194 10, i. 2 REFERENCES T, E, Lang, Summary of the Functions and Capabilities of the Struotural and Matrix i iter 7 ‘am, Jet Propulsion Laboratory, JPL Technical Report 32-1075, 1 April 1967, ‘M, F, Rubinstein, Matrix Computer Analysis of Structures, Prentice Hall Inc., 1966, J, 8. Przemieniecki, ‘Matrix Structural Analysis of Substructures,’” AIAA Journal 1, No, 1, January 1963, C, J. Meissner and R, 8, Levy, ‘Flexibility Method of Coupling Redundant Complex Structures,”” Journal of Structural Division, ASCE 82, No. St 6, December 1963, pp. 325-364, G, Kani, “Die Berechnung Mehrstockiger, Rahmen,’’ 5th edition, Konrad Wittwer Verlag, 11956, (English translation, N, Y., F. Ungar, 1957), Sa eeeaaeeaanetreaaneaaat B, N. Thadani, ‘Solution of Complex Multistoried Structures,”” Structural Engineer, Institution of Structural Engineers, London, P, Lustgarten, “Iterative Method in Frame Analysis,’ Journal of Structural Division, ASCE 89, No, St 2, April 1963, Pt 1, pp. 75-94, rarer eeerer anata eens B, E, Gatowood and N. Ohanian, ‘‘Tridiagonal Matrix Method for Complex Structures,"” Journal of Structural Division, ASCE 91, No. St 2, April 1965, pp. 27-41. R, W, Clough, E, L, Wilson, and 1, P, King, ‘Large Capacity Multistory Frame Analysis Programs,” Journal of Structural Division, ASCE 89, No, St 4, August 1963, pp, 179-204, J. H. Argyris and S, Kelsey, “Energy Theorems and Structural Analysis,” Butterworth ‘Scientific Publications, London, 1960. H, Cross, “Analysis of Continuous Frames by Distributing Fixed-End Moments,"’ Transactions ASCE 96, 1932, pp. 1-10. R, K, Livesley, Matrix Methods of Structural Analysis, Pergamon Press, Macmillan, Nir) Meares 10042 rere eeeAFFDL~TR-67-194 LEGEND FOR ACCOMPANYING TABLES All frames are analyzed for I/A = 20/144 FT? TABLES Ia, Ha, Ifa 1 4 The first ooluma refers ta joint numbers shown in the applicable figure (Figure 6 or Figure 6), ‘The second and third columns refer to joint coordinates in feet, Reference axes are shown in the applicable figure (Figure 5 or Figure 6). Columns 4, 5, and 6 refer to applied joint forces and moments in Kips (or lbs) and ft- Kips (ft-lbs), respectively, in terms of quantity P, Columns 7 8, and 9 are the joint displacements in terms of quantity P/EI, The units are kkips (or Ibs) and feet, EI |s the lowest flexural rigidity. TABLES bb, Ib, tb 1, 2 The first column refers to the member number shown in the applicazle figure (Figure 5 or Figure 6), ‘The last two columns are the joint numbers to which the member is zonnected, (Figures 5 or 6) Column 8 gives the relative values of the flexural rigidity EI, Columns 2 through 4 are the member forces at the end MA, These “orces are in terms of P, For interpretation of signs, see Figure 3. Similarly, columns 5 through 7 are the member forces at the end MB, (Note: MA is always less than MB.) 24APFDLATR-87-198 19 30290" 19 321 13 ieooaee to 369 1D seseEe to asezere 00 3300" 20 syessre 0 dzoeees 20 ae0ore 2 gazes fo dyetere 20 co sz0e6r2 20 3sests aeusia 2 oH a at : a 2 3 ° 2 200%905~ 7 D cooraes~ 3 ocorocs- 2 *° a 7 3 oy 2 3 0 (worzvus SINENADVTASIG ANY ‘S49HOE ‘SELVNIGHOOD INIOC 'FICVUL AUOLS OOF “AYE-AALS ooatace coaven 2000» co2roce poatee ceoteyoy soe 31 APEDL-TR-S1-104 % ° 2 5 to % "0 3 “0 “e evoroute +e 109 19 guruzre co goreort 2 “0 00 3skse9= 10 Souusey 5 6 To seeee> 26 ewaee o ‘0 aorses- —ooatace 208 to gaorri- co azecort autor "2 $0 Sseesct= te Stzosry —Goerat= 3 00 35696t1- £0 s4t90%1 ——_o0001- 2 To-daveerz= To 3oaeue> ——Booeot= 2 Mn starsi= 20 deeeeee "0 2 dovtaee- —ogatart ws To-auave? Boasot- 3 Op asersnr Bo aucias 2 ° poore8= —ooatoe wes 0 a0zvor1 —ggorai- “ > ast 2 Basra "0 3 poatase- acarar oes 20 aeasoe1- fo szacort ——co0z01~ 2 poost $0 duveers= 0D auveses te aoeuyey © Saosol= 3 oy 2 iesesrt= TS BSoess Ue Stecete *2 2 00% vo0ta66- * ses 20 a0sezs2- zo ayotus- yo 2ee0cz ——aooror~ “0 7 To jocey-t- 10 auutss- eo 3eueeet_ ——o0s01- 2 ° 20 Btostez= 0 SHeeors- So settee *0 ° ° pooro0s- —nontooe ons 20 yooutsz- 0 atzctt- v0 seeoerz ——_oo0tot~ 2 2 To Stgeert- To 300aee1- Co azecey ceva 2 ° 20 doieore> 0 Sotvere> $0 aaeLE Te "0 °° ° oo*s06- —o9arost os yo aycocrs ——ooosor~ 2 a fo grersst —— Soosal= 2 2° 50 dete “0 2 0 pootace= 009702 ws {0 dooeest= Sootar= to 20 dessore= *3 2 eooraos- 009%08 as yo pecoesz ——coorot~ P “0 fe deuce booso1- 2 7 So gzost-e 3 2 ° 2004205 909°» as 20 aevecte— yo aevoerz ——ogorar~ 2 ons 1 Seursst= fe Seuerst—— Saorot= 2 ° 20 destzee— 80 S90ut-2 *0 * ooo" 00*008= + us watuviow aewesia xeusio anon aosow0s cS » ator (eneo) ot ateveAFEDL-TR-61-104 a cere 0001 oo0ror oooror ooorte ooorte oon 2 aez0tee soo3e128 89 360% Toeasa ponasteset Op douters — Gquaetaeet= fondeuerez teogsouz" to: co aso 23 desaeee fe ite 29 aston bo aesurte~ 03 He Ge Sezeore= 89 3600" oe seo0sre= oo aac 207306207 90 a2ceztz— fosteese Te-3ucyea- bo auetees Jocaevenrin OC aetaete m0 mons 33004 Wwe avon GoM sa2eod WIGAN aNVEE AMOUS on HVE. a mave rocaervs roodeee ot 501g “n90t AAPFDL-TR-67-194 06 ate ue oe woe oe we oe oe 206 oe oe oocese ooaree averse coors o00*52 aootse noose ” 19 auseesy 13 Se00es= 63 asenr 13 Behera 10 agezers 03 derek 13 dosters to 26120" 89 auoe 19 agteers BEER 1a sioeees 03 suesoet 13 ieetes to asnerse 13 Sveeeeae Toate op aMeee Noedetaces 00 Sen0n to acute Te-geotor= to agsterte To-dvotors- leet to atzeors Te-avecores To 3etyort= 00. aenucry To-astesey 00 saucers 00 asvey sz Toodeecrse 06 aeeet ez So-saeoese 00 364 oo ain oo ate 00 aevent2 To-deaewee €0 Suet, foodeen te goszor- Te Basest= 1 azole to az9s9et- 60 stocete= fo 392cet tergeeaeet to gousert ob S609K0e $6 aeooe"t 3g youz0re te-s0006| zo-atoro"« forgets zonseeserz deoateotee CEES sonascct loostsey zeracest fordeoee 00 se908 20-36990 Go seor3 sows wD (ou) a grave ie aveceee o atoorrs Xe as00t +s We ieesees Teor to aseeers te asszeee 16 tessa to 3aueuty- © gone Xe Soeoeeee 6 ce Anak Te-aueryee— Go ausezre= io-detsecy— to ayztcet qoateort To-d90t0-% ioautsy o-3¥010%F aerate to atzvort Ie-deeeore to gthost Yom3st9¢e4~ op aeeteey oo asenyez— 1o-d2eeyee Oe seyetee= spa nytet- toraeecora to i to a ro-gesaes 123see9e1 rockove Toezeveess— Aocdeeee st 19 zoxsoet~ 60 Ss09e ee 19 30619¢5~ torzeo2622— eocstotsse= tosher BSIREE rooaezzrce ssolsat ronatasere ebnsezetee terdeoeeed oo aes0r ose 268 28AFFDL-TR-67-194 os 26 ss ws as ts oooree oourss 200% oo0ess voor oocroe v00r0¢ 3 1D doeeees 20 acstorr 23 a1ztott 03 Sot 13 30uasee 1p asvtote 1p acecoty 13 Hesse snauon 00 aosre9 09 Seems To 3ozeei- 10 azgvore~ 10 s1tcore- 18 e100"e= to attyert= to se9uett- Fepeere: 00 38956 0 sze9es9 Toedezses? 06 aetEee > 00 a9ezery oo asceee> 60 ascuery oo aer98"> Yoaeneare ob seis" 00 aceue Toeducse-2 was oo a zo ayecorz- io site 20 Bleoeet> 2b auuts 00 seven ts~ bt deste 00 atesuee 00 deste" to az0cery Ob dots y To 31enoes 2 aaecore 20 detent to-aeuuast= socatstter eooserce 0 seues tte Soe3za0¢ I= Go deers oo suesst— 2beaoenot= awa ey (6109) at stays Beseee9 aosz071 Bees deoweet susie abet 36086 beet averse Atos 0 s06u@ty- To-aeevere, 05 3e88s-y— yo aeveeet wedeene— to seuseet 10 azovorz to atucorz To-dbovece To ave? lo atty6; 1o-dove: to pa9uu 1o-d0ere oo are og abot 29 au99ety= pects 00 a2acery— oo asses 00 astuaty— Tongzesore= ou _aotavtye We-aevcace= So asus To-aoese-e= 00 39 2330 83 demos 29 auctor ig pues 20 Seger 19 acousy 00 aztec 89 aeetest 65 susie to azoceey- 09 dostyca- 18 seas 23 datoert= root 001 see re soot “ROE 20APEDL-TR-61-104 209 £09 oe 009 os uss 08 098 recs 965 566 oss 166 ooorss oo0rss conse 00%0¢ 00*0¢ oo0r0e 0006 000 29 3 3 z @ Fa 3 guenser dooney dressed Jeactss dross! aevpurt aise 30 i Rseert Becsore- Beste avstes9- og drones sarees Balvere BeLeeese ascsore~ Bt Bees9-3- Dee 2 sie anno Toodtees io atsoee 10 360 eonsusee 00 azeysre To-doezore op aetsee 0 aecrere Yo-dveteet 00 aLeyere 00 atecere 00 actenre To-abe bo azssere Tonaoszore 00 avesere To 38% To-380 20 soveyeze to Stree 20 savage $0 este 0 z0008", Br 10 aeevety 0G 3t1L0+9 to reese z0 20080" io anos 2 3eseeee 0 seuserz- To-dtsees— oo aes0re eo-3tese: bo aoeesrt= locateztsa= fednte tosaeasore 00 swore To-ateest Op aeisoet fo desss 20 asieree= saves WWW (omoo) a steve to aeritee= 10 daeoset= Go aviv Wo 3900°t co as009"s 16 at900et 00 sezuzes fo Seacers~ Sb Beers co asieety 6 ssatee to ayerzre~ ge 31079 Yo 2e2ee9- to seotees- awauow 16 tsb to 350: e-dseest to 3e2: zo-deew: foasect 230302 00 s65yyte~ To-dnerger= op atesyre To-deacost= oo aetoere= 00 aer9yte— To-3ve, oe sossere= 00 5604 to-398: oo ats to seretst 20 sovevee 13 stsse to aezve: 00 gzevo. to doses: 9 scoaes1 69 aczst-t 00 aueeoey 00 asuter? Toaaterset 00 sear 90 3s020 81 eonaties 0 scetort 00 acvaest~ To-aegeeete 08 deteert= Zo sezyyse 13 Sespece 20 asters aovos wre vott sot neor reo 01 s0or seo cot zeor teot cor soot 30AFFDLATR-61-194 To-a0zcoee to ersos2 Bisosse To gezuers 00 assure 10 atenest~ 10 doxet-z> 09 aveeerd, soriviow 10 azuv9- 20 aiszorz To-de90 20 aeeer-e 10 30089" fe auce-t- © 10 fo decree fo Suei-s— 20 aszeett 20 aeo0ee2 10 ateesee- 900 a60se-z- 2o-deeeese= 0 aeee"e= 20 sozees2- 3 3eveo-2 20 Sonia anwast0 aosi6 aoarer- ‘0 boorat= * * 3 o90%01- anawow—Avazw03 Gonvean wang 00°02 o00ro+ 00°07 o00°or “0 oovroz 0070 000-02 o00+08 ooo SINGMADYTAsA ORY ‘SOHO ‘SHUVNIGNOOD Alor ARV STEVD AVENE mm stava, 00002 oooreet ooovost vooroet 00001 000-008 000-05, o00r0z * om or anor aAFFDL-TR-67-104 To-aeetteg= ‘0 * wotavi0¥ °8 00 a0cts2- 20-2901 amines 00 agzees2~ eo-ateee-t 00 aueze-2- 20 sever 20 ates fe dette" 20 21020" breos (9) su atave, oooree 0002 00-02 o00r07 00°00 o00eor oo0-0se 000-056 oa0a2e 200006 00082 00022 00002 se se e 2 w o st n antor aAFFDL-TR-6T-104 co ae 3 2 70 * °° °° *0 0000s ze con °° o00rcz 00°00 “ 20 s0v06:2- 00:01 00021- 2 Te deetsce Soorer— “0 * Po daeeest= 8 00-1 53 onto» eoarary oe % 20 se te 2 °° °° *0 0 ° “0 o00*0ey eo 90 acyastz- 20 acrezrs sora ss : Toogoswest —@0 g1eog-e 000-01 ce * oo seene-z 20 aytel"t °e *e °° 00" o0ros we 20 anoayey o90"1~ “0 20 aeeotee se 8 to decetst> 20 atten 00"1~ * 00's —onnroze a sortvi0% Avrasia 0 sawn A304 x30 ” ” axtor (emoo) a stave 38AFFDL-TR-61-194 " or ooorot oo0or 00°01 x00 0-01 c00°0t 00°0 (200°01 zo-avyeee1 Eonaleesse- eocaeeor't docguitee- zo-aequure fonatee zonaneess to-aneee2 00 aceas7e- Oo seet-e- To-aseeers zo-asorstt- 0 aesey-t 80 3 00 ayezery= 60 Scere fo-Rote's= 20-30L89+9 80 deevoce— 80 aetzee= snawow wens gorgevyert- fo-3laess fo-Sze0r Toaussee yocdzeae-s- to-geresre~ rovgezeese- Toratseo-e= Lovzezsore- 29 azsoors- to-aenese2- focuisee"t 00 agzeesg- Lo-autee-s— Hocgeeteu- 09 Severe detent aos Wi aeaV0" ans (olMEaM) SOWA WIEWEN ‘aNVUA FTEVO AvE-NEL on save ro-aveeere soazeeess Tocaeevere To-ser ro-ave yoraee 0-290 To-39990-9~ tooseaeess aa08 WwW " ze u “APFDL-TR-61-104 a oz a ” a 2 a ” nu a “ 00 or o000t ooo 00°01 000% 00°01 oaarat cooror o00"or onto vooos n 2s0-s- To-seset"s 00 y00s1"2 00 aoeee To-deecess 00 seusyet 09 aweeest Toodersere 900 aetzere- 80 a¥a60%4— Toodgese-2= 0 sesorsy- 80 ag000¢y— to-dtss0°t— 20 aevoyrs~ 00 Seceges= tedster aNawon z0~3088 So-3028 ro-aye Tocaoeeese Eerass¥0rS to-ateesce Torponniss Toga Tocaseee ee ws tongoutere~ SE tecaeerser Kecgbetsrs. Tooayeese4 wens agvoa waxy (em09) on wtavt 1m To-auseore ent Bo-deattee to-gsut tocaoeds zo-s0ze9rs Zonaeouese doresters zonseces-2 Pocgeozest Zocdesiort To-acezyez— to-av004 lo-ao0es cocasseOre= Tovaciesse- yodseer se eonsuteess, Toodseeere 1O-86R TOL 00 azwsory To-dzooere 00 g9zsort To-asaere To-asueteL to-goztere Tocaeatget~ rosie soos Whew 2 a 2 ” oz o on a * weAFFDL-TR-67-194 ze ra of er 82 oy B 00°01 oooror 00°01 peosee ewes 31660"1— focgonte" Yordesee-s— 10-30090°6 yooaeas9"€ yocsuenest yor3seee Focgessoet= ro-azece: (emoo) an eave roasts footers Toca 10. 0 -3Ez 16" dteste ro-aooery fosaonte-2- 1-3 0908-5, 10-30090"5— to-susorz Yorases0et to-z2s6e erat EoaeL9e Tarateyss ro-at9see roads zo-accor-6 2ocgrteses— Toasisi-e toraeszice roogsenes se 38APEDL-TR-O7-104 Yo szetos2- 10 doxotee— 10 e0%5¢¢- to deree-e> oo sent" Bo setae ee 09 a5230-e— "9 oe te d2sa6" 00 aneeesz- eo-30snere 90 aeesesz- 2o-doceere- 00 aeeene to deeeese (asxavey conn ovor0z 00%0% (000°02 000% 000-02 ‘o00*02 SINIKIOVIASIC ANY ‘sxOWOR "RALVNINIOOD HOP TANYA S1AVO AYECHEL wm tay 00-002 00: 000*051 000°008 ooo 00-02 a n amorAPPDL-TR-61-104 to sores to-30e 10 3221 “0 % °8 to aex: 2o-6e to azsterz- 19 distesa- Toate 0 aseoes *0 * 00 9624 foneieest oo arene to aise zo asses fo aeeee fo deat 0 aotees 20 a2ece" %0 xoesia awawow— A-a3¥0 (e000) my steve x-a9¥04 ovorae o00"0" o00%02 000-0" 00°02 o00'0» o00+0z 00°09 00000" c00-00% oo0voue o00o6e ooorere 900008 voavoaz ooorase vo0rose ao0aee o e 2 ro ” amor 6AFFDL~TR-87-194 *0 co %8 te °° * * ov0-00, ve Seige Boee-t oa0'e— oooroz 000005 a ayove-1- zo 399062 00"1- 00"0- deere fe aot sooo 00°0- sores io deel bon t= 300°2- ——aooor ——coo-on» o* ‘0 so :0 co 20 2 0 * * °° 0 °° “0 000s se azesere- 00 agsaste- 20. angers o000- Geevare- —To-agwnest 00. acne o0-0- Gotzo-1= oo aoeou"e- 0 BTL T 00-0 oar 000s ca peuerz- 0. ab02 £9 a00gr'y 0071 00e0- Reese To arte 20 ayoviee Sore 500-0- 3eee"t to duet 20 deetent 800°". 805°8- —onaroy ——on0-oz aw antavio® sstesia xevast0 INanOk A304 H-BDWD - ~ amor (ooo) a stays ssAFFDL-TR-O7-194 ” 1 " a or or 00°01 00°01 00°01 00°01 0000 ooo 00°01 oo0ror 00 a9cby's— 80 aoseere= zo-zeszse1— 00 ae 00 aes 00 308! 80 3524 00 Seieene- focaecor'e, to-acr9e+2 To-a0tre to-aeun9%e zossten to-sz026"¢ Yo-s0200"« rosaeete - Socseee! to-z69¢276- oH Toodoazess- Toostees 3 savor why Gasxiysy to00) Soczeazore focass91 10-361 FOcaeaoOr to-ayyeete socaepee-e Bocdeaore1- To-aseve-s To-gonse-2- to-s20287¢~ 00 ae4or-t- zo-gestLce-10-ass19"6 103 fouivscee Co-3e0Le"t Tocazsetse= eonakesers— LOcgesse-6 00 aus 00 as9ee-s- Yo-39266°s— ro-apicere 2oravese'e To-aeotere ro-goezess 0 seizes Tooaesee's— sxowos WIEKAN “ANWa IEVD AVENE a or 0AFFDL-TR-81-10¢ a ” @ 2 oz ow a 00°01 200¢0r 00°01 conor 00°ot 00-01 00°01 000%01 00001 00 aasee-y- To-azistee 00 3asrert to-aneeess 00 3eeze-4- 80 3su607— BO-ansee-z— 00 aeceorr 99 dogo" o-dsovee2 snewon 20c30t09"e— 0-369" Zomzousere— e0cgzetens toeneutarz Tooseneece to-aeany"2 Tonseart-e eocieneaee Torasocore Toate to-ague9"¢ —t0-209L"9- No-seerice 19-389" focaesers— Yo-aveeave= g0-319622- to-3t00675- fesatsese- socauueess poraeesent Tonaenbere- S672 90 3z050"1~ Tonawore-z e0c38621 wens awd wi woo) oan arava 0 azzogry~ 60 acots-s— 10-36916=2— 00 3¢ 1629 Bo-a2see"r ro-394 2o-3s100 srawow (ona to-azeotss cocgoeti ce gaat: cane IF x S0-3¢950" lo-asns9se— 1o-39sene focaoeee"s corgsver" 19-300 0-3 tora EOodetees Tosaseeee wows aos wiey % or et oo aAPFDL-TR-81-104 o-atoe1ss- To-asetess — to-a1oety — to-gtysee8 focgeeta-2 00 aozus-2— To-aetory ye oooror bo aesnere — Tocgesa"s- LOnaeose-s- 0D STOTT Tosser 1 Toogeno-e= oe Yocaieeert=To-atue-s ot To ateso*t- Y0-300907¢ | to-stsse“e- 90 aereerL— bo steered Ww-atuig-€ — Tondvotece- 09 _ oe —oooror 90 ateet-t= To-aeve"T — TO-SLULG"E- 0D 00 p1evcrz-to-aysst-z t0rs296L"E- 00 BLS4LTE- YO-avESTTZ- JO-azNsETE So Ssestes= Tosanceose—Lomserere 00 Sagar yooaete ez cdoror ao e918T 20-aeeate {ote Tosanee-s e 00 368867 10-2ELG0°= TOrSsEOETE- _LO-BLLSZTE_To-geLSOTZ —_Yo-D6E0C76 88 REGRETS Zorzestece> Zo-aeeusee — Lo-atnog-s- gonaeetere —20-30619"8— ez conver 90 3eee] —Tona1eoorl= TO-se1Gi-e- TO-a¥zs5"S —10-Bus9OTT —TO-36TSI“E ee 00 a99Les9- 10-3z6EerE —YO-a9zeITE- 00 3zEDeTG —TOMsROEETE- TO-aEELTS 80 seen Tocdesess2 — 1o-3090 rericteed tz cooter 00 sere eonaozbe°s — To-av9z1"9— yocaeeeno 2 8 sNawow waHs aoa Ww awawon waus aves wie eK (0) am staveUNCLASSIFIED, Secucity Classification ‘DOCUMENT CONTROL DATA - Ral ecu clousiNenton of 16 nay faba and Indexing motion mnt be werd nan th orev! ep ca TGeTamA TING AETIVITY Comers ma ipa nenOnT BECUMITY € Cham ‘Theoretical Mechanics Branch Structures Division Air Force Flight Dynamics Laboratory AN ITERATIVE METHOD FOR THE ANALYSIS OF LARGE STRUCTURAL SYSTEMS peaerem WOTEE ips of pe on aca weed 7 RORY ast name Wat rae aD Venkayya, V. B. fe REPORT ORE Yar Fava we. OF Paaet [75 Wo. OF REFS April 1968 49 (aRORT NOUBERD) a pnoster wo. 1467 AFFDL-TR-67-194 = ‘Task No, 146705 Fe BI ERAPPON? PCY Chay ear mambare Bal may De alge Te AVAICABILTFYUIMITATION NOTICES ‘This document has been approved for public release and sale; its distribution is unlimited. Air Fores Flight Dynainies Laboratory ‘Wright-Patterson Alr Force Base, Oblo 45133 19: ABETRACT An iterative method to analyze large structural systems subjected to static and therma? toadings is developed. This method ts essentially the displacement method of analysis in which the finite elements are grouped into convenient size substructures. The analysis proceeds with one substructure at a time, and a process of iteration establishes the con- timuity of the system. The speed of convergence depends on the nature of the stiffness matrices of the substructures. To illustrate the method, two frame structures are analyze One is an 1800 degree of freedom system and the other a smaller system. The purpose off the larger one is to estimate the computational time involved, The results of the smaller system (by iterative analysis) are compared with those obtained from direct analysis, and the results are nearly the same. The main advantages of the iteration method are that the entire computational scheme can be carried out without excoeding the core capacity of the ‘computer, and the method is more economical with respect to computational time than the direct analysis, DD .t2m. 1473 UNCLASSIFIED. Security ClassificationUNCLASSIFIED Security Clessification Weeration Substructures Finite Element Analysis Frames, plates, and shells Automated Analysis Structural Synthesis L, ORIGINATING ACTIVITY: Eater the name and addres of the contractor, subcontractor, grancee, Department of Be- fease activity or other organisation (corporate author) Tensing 2a, REPORT SECURTY CLASSIFICATION: Enter the over: security clansficetion of the epor. Indicate whether ‘Restricted Datu" Ie included" Marking le to be in accord lance with appropriate security regulations, 28, GROUP: Automat dounarading is epcifed in DoD Die ‘fective 5200,10 and Armed Forces Indunttval Manual Evter the group aunber" Also, wien spplic teelinee fave eeu wed Yor Groep Sand Gravp + 3. REPORT TITLE: Ester the complete report ttle in all Capital tetterm Titles in all caseo should be unclassified. mingfai ttle caanot "tle cleasifietion inal capita 9 following the tl 4. DESCRIPTIVE NOTES eer the trpe of ‘port eg, interim, proerens, summery, aia or final Gite tne Inciuaive davow wher a specilic reporting period te 5. AUTHOR(S) eter the name(s) of authors) a8 shown Fin the report, Enter (eat name, fra outer mldate nl if altarys abew rank and beanch of sevice’ The name of the principal sthor tan absolate minimsme requirement {6 REPORT DATZ: Enter the date of tho report as day, onth, weer ot moh, yesh It more han one date appedre Sn the separ tre cate ot publicetion. 7a. TOTAL NUMBER OF PAGES: The total page count Should follow normal papioatton procedares, ers enter the ‘Hinber Sf pages conteiing information. 7b, NUMER OF REFERENCES: Enter the tt Teferonces cited! inthe faports Sa, CONTRACT OR GRANT NUMBER: 1f sppropriate, ‘he appitcable umber ofthe contract or grant under whi the Fepon was wetten, 4b, te Sd. PROJECT NUMBER: Enter the eppropriate listary department identification, auch a8 project ramen, Subproject mumber, syotem mumberey tack number, ele 9a, ORIGINATOR'S REPORT NUMBER(S: Ener se ff ESP Sgicalng ety. "PS mer mat 90, OTHER REPORT NDMBERIS) 1 he copor hes been ‘seigned any ether report numbers (either by the ordinator Fb) the sponsor, algo enter this hamber Oy 10, AVAILABILITY /LIMITATION NOTICES: Enter any tine lation on farther dissemination of the report, other than those amber of TRETRICTIONS by security clasatication, ning standard mete w ‘Qualified sequesters may obtain sopies of this report tor: DDC: “Foreign amouncement acd diesgrination of this feport Uy DBC in not authorized. “U, 8. Government agence may cbtaln copies of thle report directly fom DISC. Other qualified DDC ‘bere shall request through @ o SU. 8 military agencies way obtein coptee of thie tepore directly from DDG. Other qualified seers hail request through o ‘Ail distrbation of thi report in 2ontolled Qual= fled DDC were shall reqeent Gee © i the japon has been farlahed tothe Office of Technical Services, Department of Commerce, for aslo to the publley Indl {te ths fact and enter the rice, it town, {LL SUPPLEMENTARY NOTES. Uae fr uation eplanee 32, SPONSD: 180 MILITARY ACTIVITY: Enter the name of the departmental project office Laborato’ aponsoring (pay Ing fon the research and development, Include address. 13. ABSTRACT! Enter an abatract giving « brief and factual ly desirable that the brtract of cleasifiedceporta te uncinnifed. “Such acegregh of the sstzact shall on Sith fe incication ofthe military security classification of the in- Fraattonin'the paregrePh, presented eu) (2). (2) 0° (0) “There is no limitation on the length ofthe ebstreck. ever, the suggested length se from 130 to 248 words 14, KEY WORDS: Key ment medel dew gnetion, trade name, military roraphie Toce!ton, may be ened t8 ‘be followed by an iiicaion of technicnl coe at of ints res, and weigh fe opto UNCLASSIFIED Security Classification