Você está na página 1de 9

ANALYSIS OF TRUSS-CABLE STRUCTURES

G. R. MONFORTON and N. M. EL-HWhft


Civil Engineering Department, University of Windsor, Windsor, Ontario, Canada N9B 3P4 (Received IO October 1978;received for publication 7 March 1979)

Abstract-The energy search method is applied to the analysis of general pin-ended truss and cable structures. Geometric and material nonlinearities are directly incorporated within the formulation, thereby accounting for large strains and displacements as well as con@uration changes due to the structural response. The principle of minimum total potential energy forms the basis of the analysis procedure and analytic strain energy expressions are presented for the individual elements of truss-cable structures. The total potential energy is obtainedby summing the energy contributionsfrom each member and solutions are generated using the conjugate gradient method. Examales illustratethe ease and versatilitv of the method which avoids the inherentproblems associated with solving large systems of nonlinearequatiok.

fNTRODUCTION

When a structure deforms considerably under loads, the resulting change in geometry complicates the analysis. The characteristics which are responsible for the complications include geometric nonlinearities (large displacements), material nonlinearities (large strains) and changes in the configuration of the structure (including buckling or fracture of components, loss of prestress in cables and guys resulting in the members becoming inactive in resisting the loads). The governing equations for such structures are nonlinear and the iterative solution techniques used are often subject to convergence difficulties; in addition, if some members of the structure become inactive (e.g. buckle) or return to service (e.g. unbuckle) during the loading history of the structure, the governing equations must be. redefined at every stage. An alternative solution technique, referred to as the energy search method, has been suggested and successfully applied in Ref. [ 11.This method circumvents the need to generate and solve governing equations; instead it involves a direct search for the minimum of the total potential- energy of the structure and makes use of powerful and readily available unconstrained minimization techniques. The energy search method, in conjunction with the finite element method and approximate potential energy expressions, has been used to predict the structural response of elastic truss, frame and shell structures which undergo finite displacements [ 1, 10,111. Bogner[2] refined the analysis presented by Mallett and Schmit[lO] for elastic tension structures and accounted for buckling of compression truss members and slackening of cables. Subsequently, Buchholdt, Das and Al-HilliD] applied the method to elastic cable nets with large deflections. In the following the formulations required for the elastic and inelastic energy search analysis of general truss-cable space structures are presented. Following the approach of[2], analytic expressions for the potential energy, based on the deformed geometry, are presented. The elements are assumed straight and prismatic between frictionless pinned ends and all loads are considered conservative in that their original directions are

preserved. Numerical examples are presented which demonstrate the potential and versatility of the formulations contained herein.
GENERAL FOltMtJLATION

Typical pin-connected tension and compression members in both the undeformed and deformed states are shown in Fig. 1. The deformation of such members is measured along the x-axis defined by the line joining the displaced position of the nodes p and 4. The undeformed length, L, of the elements is defined by the initial positions of joints p and q_measured in a global reference coordinate system (X, Y, 2): L = {(& - Xp, t (I$ - Y&Jt (2, - $,2}. (1)

Under loading, the joints undergo displacements (II, 6, I) measured with respect to the reference coordinate system and the distance between the joints in the deformed position is defined by S: s = {[& t l&) - (Xp t $)I2 t [( I$ t Q - ( Yp t Cp)12 t [(if, t G,J- (iiTp t 6#}2.

(2)

Strainenergy The following discussion is focused on presenting the formulation of the potential energy for the members under consideration. Expressions for the strain energy and its analytic gradient are obtained since both are required for implementation of the energy search technique used in the sequel to generate solutions. 1. Unbuckled elastic member. The strain due to deformations, cd, of an e&tic member which retMtiuS straight is expressed in terms of the deformation, II, measured along the deformed length of the member (i.e. in the x direction) and is given by e,, = duldx. (3)

Assuming an ideal linear elastic material, the stress in the member due to deformation and prestrain is (r = E(6 t 6,) (4)

tPresent address: Centre for Buildmg Studies, Concordii University, Montreal,Quebec, CanadaH3G lM8.

where Ed is defined in eqn (3). 4. is the prestrain and E


327

G.

R. MONIUITON

and N. M. EL-HAKIM

//A*ksb
9.3
Fig. 1. Deformed: &f&tit ,~,+V,,i~+~~f,poi~+ii~,~~tY~.i,t~d

Tension and compression elements before and after deformation.

is the modulus of elasticity of the material. Correspondingly, it can be shown that the strain energy, U,, stored in a member having a cross-sectional area A is

is taken

as

u, =~EAsK,~ where
K,=e,,rt

(W

or
u = Cl t (c,2 t 2C26 - cy2 (W

(I-; >
e

where o and E are the stress and strain, respectively, and cf, c2, c3 are constants de&ted by c, =q,-cz/E c2 = zrqoy - u,p
cs = VP2.

w
@b) w

The axial force, F,, in the member is a constant given by F, = EA K,. (6) In the preceding portional limit, q 0.2% strain offset the linear portion is taken as

Note that K, is directly related to strain and is a nonlinear function of the nodal ~spi~ements by virtue of the definition of S (eqn 2). Depending upon the sign of K,, the force in the member may be tenshe (KI >O) or compressive (K, < 0). The gradient, VU,, of the strain energy can be obtained by partial differentiation of US with respect to the six displa~ment components of the nodes of the element (r&,, 9. 4, lip, ti,, 4) and is presented in Appendix A. 2. Unbuckled inelastic member, When the stress in a member exceeds the proportional limit of the material, the inelastic range of material behaviour must be considered in deriving the expression for the strain energy. In addition to the well known elastic-perfectly plastic stress-strain curve, several mathematical models have been adopted to represent the stress-strain relationship. For exampk, Greenberg161 used a compound curve which is initially linear up to the elastic limit followed by an expotential curve to the ultimate stress. Janatowski and Birnstiel[?I used a continuous smooth curve &ted to test results. Kennedy and Ku~r8~ also used a compound curve which is initially linear up to the proportional limit followed by a second degree parabola up to the ultimate stress. The method of analysis presented herein is capable of handling any of the above mentioned methods to represent the inelastic behaviour; as an example, the form used by Kennedy and Kumanan is used in the follows. The equation of the stress-strain curve between the proportional limit and the point of ultimate stress (Ea. 2)

equations, a,, is the stress at the prois the yieId stress determined usin8 the method and E is the elastic modulus of of the curve. The strain of deformation

Q = duldx + ; (dr&I.Q2

(9)

where the second order term is included to account for large strains in the inelastic range. Correspondingly, it

4. 2. Inelastic stress-shin

curve.

Analysis of truss-cablestructures can be shown that the force, & and the total strain energy, LG. stored in a member that remains straight and that is stressed beyond the proportional limit can be expressed as Fz = A[c, + (c,' t 2czKz - c#~] ;Ec,2tc,(K,-eP)t+2t2c2K2-c~)32
2

329

and the critical value of K3, denoted by K3=, which signities the bifurcation buckling point is K,c= - (;)f. Note that when K3, - K3 < 0, the member is unbuckled (w, = 0) and eqns (12) and (13) become identical to eqns (5) and (6). respectively. Also, when K3, - K3 20, the member has buckled (w,# 0) in which case the strain energy is given by eqn (12) and the force in the member is calculated by eqn (13). The analytic gradient, VU,, of fJ3 is expressed in terms of the nodal displacements and w, in Appendix A.

(W

- $

(c,3 t 2c2q

- c3)32 I

where K,=ct(,-$;(I-;). (*h)

In eqn (IO), eP and eP, are the strains at the proportional limit and the prestrain, respectively. Again, K2 is directly related to strain and is a non-linear function of the nodal displacements of the member. The gradient, CJ,, of the strain energy expression given in eqn (10) is presented in Appendix A. 3. Buckled elastic member. The strain energy formulation for the elastic buckling and post-buckling behaviour of compression members has been presented by Bogner[Z]. The detailed derivation of the strain energy is contained therein and is briefly summarized in the following. The strain energy of a compression member is expressed in terms of the displacements u and w measured in the local (x, z) coordinate system with origin at the end 0 of the deformed state (Fig. 1). The x axis is defined as previously and the z axis is normal to the x axis and lies in the plane of potential buckling. The transverse displacement, w, is a secondary deformation and is included in order to predict local member buckling. The strain due to deformation is taken as &, = du/dx t ; (dw/dx)2 - zd2w/dx2. (11)

Total potential energy The total potential energy, I&, of a structural system consisting of a total of M members and J joints is simply the sum of the strain energies of the individual elements minus the work done by loads applied at the joints of the assembly:
V
m-1.2....

(16)

where II is the strain energy of the mth member and is given by one of eqns (5), (10) or (l2), depending upon the type of behaviour that the member experiences in the structure; W is the external work done at node j equal to wo = $7, t P,:, t f,ti, (17)

The element strain energy can be obtained in terms of the six nodal displacements and the mid-span transverse displacement, w,:
U

3 2

&A

SK+!b4w2
2A-sJ =I

(l2a)

where $, f,, & are the externally applied loads at node j in the X, Y, Z reference directions and rl,, t?,, 4 are the corresponding displacements of the node in the reference directions. In general, the potential energy expression (eqn 16) can be written in terms of the independent degrees of freedom of the system; the degrees of freedom consist of all the nodal displacement components (measured with respect to the reference coordinate system) and the mid-span transverse deflections, w,, of members permitted to buckle (i.e. members which are capable of resisting compression forces). The determination of the total potential energy and its gradient follows directly from eqn (16) and consideration of the contributions of each element.
FSERGY

where K3=ep,t(l-$)t($.

Wb)

SEARCHsoLurioNs

The derivation of eqn (12) is based on linear elastic material behaviour where E is the modulus of elasticity, A is the cross-sectional area, I is the second moment of area for bending about the minor axis and l P. is the prestrain. It can be shown that the force in the member is constant along its length and is equal to F,= EAK3 (13)

where K < 0 for compression members. Also, the critical force, or buckling load, is
F,,=5 2EI

The principle of minimum total potential energy can be stated as follows: Of all displacement fields which satisfy geometric compatibility, those which locally minimize the total potential energy also satisfy the equilibrium conditions and are stable equilibrium positions. According to this principle, a structural analysis problem can be solved by an energy search procedure which involves finding the values of the displacement degrees of freedom which minimize the total potential energy. The use of energy &arch offers several advantages. First, the strain energy of individual members is relatively simple to construct (as seen in the previous section) and the strain energy of a structural system is equally simple to formulate since it is the scalar sum of the energy contribution of each element (eqn 16). Secondly, the method allows the direct use of unconstrained minimization

330

G. R. MONFCIRT~N and N. M. EL-HAKIM

software which is readily available in most computer systems. Furthermore, the energy search approach provides a natural means of detecting and accommodating geometric and material nonlinearities and changes in the configuration of the structure (e.g. tension members such as cables may have no stiffness to compression and may, under certain loading conditions, go slack; compression member may buckle). The conjugate gradient method and its extension by Fletcher and Reeves141 form the basis of the energy search algorithm used to generate solutions herein; in particular, the subroutine DFMCG from the IBM System/360 Scientific Subroutine Package was employed in conjunction with the variable scaling transformation proposed by Fox and Stanton(51. The algorithm requires a minimum amount of matrix operations and computer storage and was found to provide a computationally efficient solution procedure. The reader is referred to(4,5] for details of the theoretical developments. The most important characteristic of the energy search approach lies in the fact that the displacement degrees of freedom of the structural system are assumed known at every stage of the mimmization process. Starting from an arbitrary set of assumed displacements, the potential energy (eqn 16) and its gradient can be calculated by making use of eqns (2), (5), (10) and (12) and incotporating logical decisions which delne the state (i.e. tension, compression, elastic, inelastic, buckled, etc.) of each member. For example, if the displacements indicate that a compression member has buckled, then the strain energy of that member is calculated using eqn (12); if a tension member which has no stiflness to compression goes slack, then its strain energy is set to zero. A decision block, in the form of a 5ow chart, which outlines the selection of strain energy equations for the members considered herein is given in Fig. 3. The conjugate

gradient method then systematically generates a new set of displacements using only the values of the total potential energy and its gradient calculated from the previously known displacements. Using the new displacements, the process is repeated until the potential energy is minimized at which point the true solution has been obtained. It should be noted that the energy search procedure automatically accommodates a change in the state of members (e.g. slack members may come back into service, buckled members may unbuckle, etc.); in addition to detecting local buckling (i.e. buckling of an individual member), gross instability resulting from the accumulation of local effects can be predicted and is characterized by an unbounded minimum to the potential energy. Several numerical examples are provided in the following which demonstrate the versatility of the method. MAMPLE REWLTS 1. Cable roofs Structural suspension systems such as cable nets are being used with increasing frequency for the support of long-span roofs and provide efficient and practical configurations suitable for large exhibiting halls and stadiums. This type of structure is often more economical than conventional structures since the entire crosssection of the members can be used to its maximum capacity. However, deflections are often large, thereby requiring analysis techniques which are capable of accounting for geometric nonlinearities. Also, in order to compare the design and ultimate load capacities of cable roofs, the inelastic behaviour of the cables must be considered. The following examples illustrate the energy search procedure for predicting the structural response of cable roofs.

Fig. 3. Decision flow diagram.

Analysisof truss-cablestructures An orthogonal cable network having the shape of a hyperbolic paraboloid shown in Fig. 4 was analyzed by Thornton and Bimstiel[l2]. The cross-sectional area of each cable is 1.0 in* (6.45 cm*), the modulus of elasticity is E = 24 x 10psi. (165.6 x lo kN/m*) and the horizontal component of prestress in all cables is 50 kips (222.5 kN). Each joint is subject to a vertical load of 1 kip (4.45 kN), except joint 7 which has a vertical load of I5 kips (66.8 kN) and a horizontal load of 10 kips (44.5 kN) in the Y direction. Thornton and Bimstiel used a method of continuity for the solution and a linear elastic material was assumed. Using the energy search approach presented herein, the total potential energy is a function of 75 nodal displacements. Sample results for the vertical deflection of the nodes and the horizontal component of the force in the cables are given in Table 1 where they are shown to agree with the results presented by Thomton and Bimstiel. Comparison of the geometrically linear and nonlinear solutions indicate that the cable tensions may vary by more than 100% for the example presented. i
12.0

331

12.0 D
L_________-_________-J

AB-24Oft; AD=l20ft
A

AC is axis of antisymmetry

8 Spaces

x 40. 320

0 1

v
Fig. 5. Nonorthogonal hyperbolicparaboloidnet (I ft = 0.305m).

Doubly symme+r~c

Coordinate

of joint in

ini?ial pcmtwn indicated

Fig. 4. Orthogonal hyperbolicparaboloidroof (1 ft = 0.305m).

Although results for non-orthogonal cable roofs are limited, no additional provisions need be introduced into the energy search formulation to account for non-orthogonality. The non-orthogonal hyperbolic paraboloid roof shown in Fig. 5 was analyzed by Kumanan[9] using an incremental load method to account for large deflections and material nonlinearity. The stress-strain curve used is the same as that used by Kennedy and Kumanan[Il] and presented in eqn (7) herein with the following material properties: E = 24.0 x lo6 psi (165.6 x 10kN/m*); a, = 124.9X lo3 psi. (861.8 x lokN/m*) a, = 155.0x lo3 psi. (1070 x ld kN/m*); a, = 250.0 X lo psi. (1725 x ld kN/m*) where a. is the ultimate strength of the cables. Failure of the roof is defined as the point at which the stress in any one cable reaches the ultimate stress. The initial horizontal component of prestress in each cable was 50 kips (222.5 kN) and the cross-sectional area of all cables is 1.25in* (8.06 cm?. Using the energy approach and considering anti-symmetry about the diagonal AC, the structure involved 108 degrees of freedom when equal vertical loads are applied to each joint. Results for vertical loads of 1 kip (4.45 kN) at each

Table 1. Results for orthogonalcable net (Fig. 4) Joint number 34 1 3 7 13 19 23 25 26 Verticaldisplacement(ft) (1 ft = 0.305m) Linear Nonlinear
0.0 0.0

Horizontaltension component(kips) (1 kip = 4.45kN) Member Linear Nonlinear 34-l l-3 3-7 7-13 13-19 19-73 23-25 25-26 33.651 42.531 92.994 71.213 29.996 24.818 23.7% 23.242 45.916 46.005 46.240 36.609 36.622 36.629 36.710 37.708
(45.947)t

(O.O)t
(0.445) (1.368) (3.750) (1.664) (0.%3) (0.558) (0:22& 10.01

0.475 1.662 4.514 I.%3 0.978 0.468 0.142 0.0

0.441 1.371 3.752 1.669 0.%2 0.561 0.224 0.0

(46.056) (46.264) (36.657) (36.679) (36.697) (36.710) (37.706)

Ylhomtonand Bimstiel[121.

332

G. R. MONFORTON and N. M. EL-HAKIM

node, corresponding to 5 psf (239.5 N/m) of plan roof area, are presented in Table 2 which shows close agreement with the results obtained by Kumanan. The ultimate load capacity of the cable system was determined by incrementing the load at each joint. Using the energy search technique the ultimate load was determined to be 49.2 kips (219 kN) per joint with a maximum deflection of 22.8 ft (6.95 m). These results compare to 49.1 kips (218.5 kN) per joint and 22.5 ft (6.86m) as determined by Kumanan[9]. It should be mentioned that no provisions were made in the formulations to account for the occurrence of unloading of cable segments, but the load-stress history indicated no case of cable unloading in the inelastic range. 2. Truss-cable system The suspended dome-truss structure shown in Figs. 6(a) and (b) is studied as a final example. The structure consists of a 120ft (36.6 m) diameter shallow truss dome 6 ft (1.83 m) in height (Fig. 6a). The dome has 42 tabular aluminum members with outside diameter of 4.5 in. Z=4.114in4 A = 1.718in* (11.1 cm*), (11.4 cm), (171.2cm4) and a weight of 2.02Ibs/ft (265 N/m). The shallow dome is suspended from the nodes on the outer circumference by 12 slender hangers, each 12ft (3.66 m) long, A = 0.1 in* (0.65 cm*) to a horizontal network of orthogonal prestressed cables (Fig. 6b). The prestressed cable network (high tensile steel wires, A = 0.5 in* (3.23 cm*)) provides the strength necessary to support the loads involved in the analysis. The prestress force in each cable of the network is taken equal to 25 kips (111.3kN). The behaviour of this structural system demonstrates the effectiveness of the analysis and permits the investigation of a snap-through buckling phenomenon as well as the post-buckling behaviour coupled with the response of a typical cable network. The total potential energy of the structure is a function of 162 degrees of freedom (120 nodal displacement components of the structures 40 nodes, plus 42 possible midspan buckling amplitudes of the dome bars); symmetry was not taken into consideration, however, complete symmetry in displacements and stresses resulted in all loading conditions. In the following discussion, loaddisplacement histories are generated and the search procedure uses the solution for the previous loading condition as the starting point for the current loading condition. For gradual load increment&ion this feature is

t---------_l20 v t

Z i-3&l i-_

60-

Fig. 6(a). Suspended shallow truss dome (I ft = 0.305m).

T
/

30
240 p!

Fig. 6(b). Horizontal prestress cable network (1 ft = 0.305m).

Table 2. Vertical joint displacement for nonorthogonal cable net (Fii. 5) Joint number 1 2 3 4 5 6 7 ! 10 11 12 Displacement (f0 0.1969 0.5001 0.3705 0.8077 0.7144 0.4415 1.0685 0.7902 0.9979 0.4532 1.2482 1.1908
(0.2ofm

Joint number 13 14 15 16 17 18 19 21 20 22 23 24

Displacement (ft) 1.0231 0.7539 0.4018 1.3175 1.2644 1.1104 0.8662 0.5558 0.2303 1.2482 1.1901 1.0220
(1.0312)t

Joint number 25 26 27 28 29 30 31 33 32 34 35 36

Displacement (ft) 0.7529 0.4012 1.0685 0.9966 0.7884 0.4518 0.8077 0.4398 0.7130 0.5001 0.3694 0.1960
(0.7558)t

(0.5182) (0.3826) (0.8293) (0.7322) (0.4509) (1.0882) (0.8016) (1.0150) (0.4582) (1.2629) (1.2034)

(0.7570) (0.4013) (1.3286) (1.2735) (1.1153) (0.8662) (0.2246) (0.5517) (1.2629) (1.2030) (1.0300)

(0.4006) (1.0882) (1.0137) (0.7997) (0.4567) (0.8293) (0.4491) (0.7308) (0.5182) (0.3815) (0.2060)

tKumanan [9] [ 1ft = 0.305ml.

Analysisof trusstabie structures somewhat helpful as a computer time-saving device; however, solutions are obtained essentially independently for each loading case, since the method of solution itself is not a load incrementation scheme. Two cases of behaviour response are studied: Case (i). The dome is suspended only from the circumferential joints. Initially the dead weight of the structure is considered as concentrated at the joints. The load-displacement history for node I is shown in Fig. 7, while the load-force histories of members A, C, D and E are shown in Fig. 8. Inspection of the figures indicate that the relationships are linear up to a superimposed load at node 1 of 6lOlbs (2715 N). At this load members A and B buckle and is accompanied by a snap-through buckling phenomenon of the complete dome structure resulting in a configuration which is inverted with respect to the original configuration. The large displacements that result is shown in Fig. 7 where it is seen that node 1 snaps from a small deflection of 0.925 ft (0.282 m) to a large deflection of 12.89ft (3.93 m). Inspection of Fig. 8 reveals that all the dome members which were in tension prior to the snap-through become compression members and all dome members

333

which were compression members become tension members (including members A and B). As the load is increased from 610 Ibs (2715 N), the behaviour is again linear as indicated in Figs. 7 and 8. At a load of 124Olbs (5518N) the outer circumferential members D buckle. The buckling of members D again cause a change in the shape of the loaddeflection curves as the load is increased beyond 1240lbs (5518 N) (Fig. 7). Once that members D have buckled the forces in members D remain constant (Fig. 8) and the tension force in member E also remains constant in agreement with equilibrium considerations. The force in members A follows a linear relationship from the snapthrough buckling load of 610 Ibs (2715 N) until a load of 1870Ibs (8321N) is reached at which point members C also buckle. Although the structural behavior could have been monitored further, it was felt that the effectiveness of the analysis procedure has been demonstrated and that a practical alternative involved preventing the snapthrough buckling phenomenon as described in the following discussion. Case (ii). In addition to the 12 hangers of case (i), three more hangers (A = 0. I in2 = 6.45 cm) are intro-

05

075

IO

13 0

1325

I3 5

I375

I40

Displacement

(FT)

Fig. 7. Load-displacement history of node I; case i (I ft = 0.305m; I lb = 4.45N).

Compesvon

1600

Member E

Force in Members

(IBS)

Fig. 8. Load-force history of members A, C, 0, E; Case i (1 lb = 4.45 N).

334

G. R. MONFORT~N and N. hi. EL-HAKIM

duced between npdes 1,2 and 12 of the dome-and nodes 20,21 and 25 of the cable network, respectively. Starting from the dead weight loads of the dome and superimposing loads at node 1 the behaviour is predicted. The load-displacement histories for nodes 1,2,4 and 5 are shown in Fig. 9, while the load-force histories for members A, B, C and for the hangers at nodes I, 2 and 3 are shown in Fiis. 10 and 11, respectively. In all cases, the relationships are basically linear until the superimposed load at node 1 reaches 11601bs (5162 N). At this load, members A and B buckle and the hangers at nodes 2 and 12 go slack. Therefore, any increase in load at node 1 causes a redistribution of forces in the non-buckled members, since the buckled members cannot carry any additional load and also since the hangers at nodes 2 and 12 cannot take any compressions forces. It is therefore seen in Fig. 9 that the deflection of node I increases more rapidly as the load is increased beyond the load which causes buckling in members A and B. This is accompanied by a more rapid increase of the force in the hanger at node I and a decrease of the forces in the circumferential hangers as shown in Fig. Il. Since the hanger at node 1 resists a higher percentage of the superimposed load after buckling, less load is carried by the truss members as exemplified by the load-force

20301

Force III Hangers (IBS)

Fig. II. Load-force history of hangers at nodes I, 2, 3: Case ii (I lb = 4.45 N). history of member C (Fig. 10). The load-deflection

behaviour of nodes 2 and 4 after buckling (Fig. 9) is a result of the combined effect of buckling of the six truss members, the change of force levels in the hangers, the slackening of the hangers of nodes 2 and 12 and the change of geometry of the suspended structure. Note that the additional hangers preclude the possibility of a snap-through buckling phenomenon.
cofKLtJsloNs

05

075

I.0 D~s~loccment

I 25

15

175

(FT)

Fig. 9. Load-displacementhistory of nodes 1, 2, 4, 5; Case ii (I ft = 0.305 m; 1lb = 4.45 N).

3200

2800

24002OC016001200

800

400

400

800

lxx)

KC0

Force I Members

(IBS) (I lb =

Fig. 10. Load-force history of membersA, B, C; Case ii 4.45 N).

The energy search method has been applied to the nonlinear analysis of general truss-cable structures. Geometric nonlinearity was incorporated in the analysis by using nonlinear straindisplacement relations. Equilibrium is therefore based on the deformed geometry of the structure which permits the prediction of large nodal displacements and post-buckled contigurations; the formulation also allows the detection of general instabilities which result from the occurrence of unstable deformed nodal configuration due to the accumulation of local instabilities. The analysis presented also incorporates material nonlinearities. The method presented does not require the use of a load incrementation procedure followed by many researches to deal with nonlinear structural problems; load incrementation is avoided herein by direct incorporation of the nonlinearities into the formulation. The method of analysis developed represents an advance toward more realistic prediction of the behaviour of general cable and truss type structures. The direct search for the position of the minimum total potential energy function of the structure using the Fletcher-Reeves unconstrained minimization algorithm was found to be efficient and gives accurate solution for nonlinear problems. No convergence problems were encountered; solutions were obtained for every problem with a reasonable number of iterations. The energy search approach provides a natural means to accommodate changes in structural co&uration due to slackening of tension members and buckling of compression members. The method also was capable of dealing with yielding of tension members (material non-

Analysis of truss-cable structures

335 suspen247-270

linearity), and the determination of the ultimate load capacity of cable roofs. Finally, the formulation presented can serve as a basis to derive similar formulations for structural members with other than pin-connected ends; this would permit the analysis of stayed and guyed towers which are a combination of tension members and beam column members.
Aclmowkedgemenrs-The authors would like to thank the National Sciences and Engineering Research Council of Canada which financed this study under grant number A4126.

12. C. H. Thornton and C. Birnstiel, Threedimensional sion structures. 1. Stmctuml I%. ASCE 93(X?), (190.

M?ENDIxA Strain energy gradiefUs The analytical gradients of the element strain energies U,, U*, U, given by eqns (5). (IO) and (l2), respectively, are defined by vectors V Ui (i = I, 2.3) which can be represented by

au au au, au- au, au au- T f,f,~,t,7.+,au, au, aw, au, au, aw, aw,
Defining A,=(&+ri&(&+ri~)

(Al)

(A2a) (A2b) (A2c)

I. F. K. Bogner, R. H. Mallett, hf. D. Minich and L. A. Schmit, Development and evaluation of energy search methods in nonlinear structural analysis. Rep. AFFDL-TR65-I 13. Wright-Patterson A.F.B., Ohio (1965). 2. F. K. Bogner, Analysis of tension structures. Pmt. Second Conf. Marti Methods Sfmcru~l Me& AFFDL-TR68-150, pp. 1253-1270. Wright-Patterson A.F.B., Ohio (MB). 3. H. A. Buchholdt. N. K. Das and A. J. Al-Hilli. A aradient method for the analysis of cable structures with-flexible boundaries. Proc. Inf. Conf. Tension Roof Structures, London (1974). 4. R. Fletcher and C. hf. Reeves, Function minimization by conjugate gradients. Comput. 1. 7, 163-168 (PM). 5. R. L. Fox and E. Stanton, Developments in structural analysis by direct energy minimization. AIAA 1. 6(6), 1036-1042 (1968). 6. D. P. Greenberg, Inelastic analysis of suspension roof structures. 1. Sfrucrurol ZXu. ASCE, %(sTS), 905-931 (1970). 7. J. J. Jonatowski and C. Bimstiel, Inelastic stiffened suspension space structures. /. Sfrucrural Div. ASCE 96(ST6). 1143-l 166 (1970). 8. J. B. Kennedy and T. Kumanan, Elastic and inelastic analyses of pretensioned cable networks. Proc. Ninth Gong. ht. Association Btidge Structuml Engng pp. 433-N Amsterdam (1972). 9. T. Kumanan, Elastic and inelastic analysis of pretensioned cable-roof structures. Ph.D. Thesis, Department of Civil Engineering, University of Windsor, Ontario, Canada (1971). IO. R. H. MaJlett and L. A. Schmit, Nonlinear structural analysis by energy search. 1. St~crwal Llcu. ASCE 936T3). 221-234 ww. Il. L. A. Schmit, F. K. Bogner and R. L. Fox, Finite deflection structural analysis using plate and shell discrete elements. AIAA 1. 6(5), 781-791 (1968).

A,=(~G+iG,-(~~++it,, A,=(i,,+ri$-(&+G,,)

results in the following expression for the components of VUi:

(Ah)

(A34
where (I) Elastic Member (i = I): f,=U,+F,L; (2) Inelastic Member (i = 2):

g,=o

(A4)

h=U,tF&

2-k

>

g2=o

(A5)

(3) Elastic Buckled Member (i = 3): ,,=U,+P~L-~[~AEK,+(5)1N] (A6a)

g,=;($AE[K,+($]w,

(AW

Você também pode gostar