Вы находитесь на странице: 1из 9

Compurers & Sfrucrures Vol. 26, No. 4. pi. 693701, 1987 004s7949187 s3.00 + 0.

00
Printed in Great Britain. Q 1987 Pergamon Journrlr Ltd

NONLINEAR FINITE ELEMENT ANALYSIS


OF ELASTIC FRAMES?
Kuo MO HSIAO and FANG Yu Hou
Department of Mechanical Engineering, National Chiao Tung University, Hsinchu, Taiwan,
Republic of China
(Received 3 November 1986)

Abstract-A very simple and effective formulation and numerical procedure to remove the restriction of
small rotations between two successive increments for the geometrically nonlinear tinite element analysis
of in-plane frames is presented. A co-rotational formulation combined with small deflection beam theory
with the inclusion of the effect of axial force is adopted. A body attached coordinate is used to distinguish
between rigid body and deformational rotations. The deformational nodal rotational angles are assumed
to be small, and the membrane strain along the deformed beam axis obtained from the elongation of the
arc length of the deformed beam element is assumed to be constant. The element internal nodal forces
are calculated using the total deformational nodal rotations in the body attached coordinate. The element
stiffness matrix is obtained by superimposing the bending and the geometric stiffness matrices of the
elementary beam element and the stiffness matrix of the linear bar element. An incremental iterative
method based on the Newton-Raphson method combined with a constant arc length control method is
employed for the solution of the nonlinear equilibrium equations. In order to improve convergence
properties of the equilibrium iteration, a two-cycle iteration scheme is introduced. Numerical examples
are presented to demonstrate the accuracy and efficiency of the proposed method.

INTRODUCIlON procedure which can remove the restriction of small


rotations. Here it is accomplished by using the body
Geometrically, nonlinear analysis of in-plane frames attached coordinate as shown in Fig. 1 to eliminate
is of considerable interest. Recently various beam the rigid body motion, and by using the total defor-
elements using many different formulation strategies mation to evaluate the element internal nodal forces.
and procedures to accommodate large rotation capa- The dominant factors in the geometrical non-
bility during the large deformation process have been linearities of thin structures are attributable to 6nite
proposed [l-7]. When evaluating these elements for rotations, the strains remaining small. For a structure
practical analysis, we should take the accuracy and discretized by finite elements, this implies that the
efficiency of these elements into consideration. motion of the individual elements to a large extent
Among these elements, it seems that the simplest is will consist of rigid body motion. If the rigid body
the one proposed by Yang [l]. In [l], an updated motion part is eliminated from the total displace-
Lagrangian formulation is used for large displace- ments, the deformational part of the motion is always
ment analysis. The element formulation is based on the small relative to the local element axes; thus, incorpo-
small deflection beam theory but with the inclusion of rated with the co-rotational formulation, the small
the effect of axial force; the element nodal forces are deflection beam theory with the inclusion of the effect
calculated by incrementation. This element was of axial force is adopted here to deal with the large
applied to a variety of examples of beams and frames rotations but small strains problems.
with large displacements and rotations in [l], and The numerical algorithm used here is an in-
reasonable results are reported. However, this cremental iterative method based on the Newton-
formulation, like most formulations, suffers from one Raphson method combined with an arc length
inherent drawback: it is restricted to small rotations control method[8,9]. In order to improve the
between two successive load increments during the convergence properties of the iteration, a two-cycle
deformation process. This limitation arises because iteration scheme is introduced. Nine numerical
the element displacement field is linearized with re- examples are presented to demonstrate the accuracy
spect to nodal rotation and the element nodal forces and efficiency of the proposed method.
are calculated by incrementation.
The purpose of this paper is to present a simple and FINITE ELEMENT FORMULATION
effective in-plane beam formulation and numerical
The following assumptions are made in the deri-
tThis is an extended version of the paper entitled ‘A vation of the nonlinear behaviour:
practical large displacements in-plane analysis of elastic
beams’ presented at ICCM conference, Tokyo, Japan, May, (1) X, Y (Fig. 1) is the plane of symmetry of the
1986. beam element and the external load.
693
694 Kuo MO Hmo and FANGYu Hou

polynomials in the body attached coordinates, the


transverse displacement of the beam axis may be
Current conflgurotion
given by

C(f) = N,d, + N26, + N,C2+ N,6,, (1)

where a bar over a quantity denotes it is defined in


the body attached coordinate. fi,, Bi, i = 1, 2 are the
nodal displacements and rotations. N,, i = 1, 4 are
shape functions and are given by

N, = l/4(1 - 5)2(2 + 5)

N2 = c/8(1 - r2)(1 - 5)

N, = l/4(1 + 5J2(2 - 5)

7 Ith equtllbrium configuration N4 = c/8(1 + r2)(l + t), (2)

where c = P, - f, is the current chord length of the


beam member; < = - 1 + 2$/c is a nondimensional
coordinate.
Note that due to the definition of the body at-
tached coordinate, the only non-zero nodal parame-
ters are Bi, i = 1, 2.
Initial conflguration
In the following, the symbol ( )’ is used for x-
x derivatives and { } for column matrix.
Fig. 1. The body attached coordinate and deformed beam If the arc length of the base axis is expressed by
member.
S=C2
(2) Euler-Bernoulli hypothesis is valid.
(3) The membrane strain along the deformed beam where c is the current chord length of the beam
axis is constant.
member and fi is given in eqn (1), then from assump-
(4) The axial strain (sum of bending and membrane tion 3, the membrane strain of the beam axis can be
strains) is small.
written as
Due to the assumption of small strain, the en-
gineering stress and strain are used for the measure CM = (S - So)
(4)
of the stress and strain. The element equations are so ’
formulated in the body attached coordinate as shown
in Fig. 1, and then transformed to the global coordi- where S,, is the initial arc length of the beam axis.
nate using standard procedure [l]. After this, the
system equations are obtained from the assemblage
of individual element equations. Element st@ness matrix
In this study, the element stiffness matrix is ob- The total element stiffness matrix is formulated by
tained by superimposing its bending, geometric and superimposing the bending stiffness matrix [R,] and
membrane stiffness matrices. The element internal geometric stiffness matrix [R,] of the basic beam
nodal forces are evaluated from the total defor- element, and the stiffness matrix [KM] of the linear
mational nodal rotations: the axial nodal force of the bar element. The derivation of these matrices is well
element is obtained from the elongation of the arc documented in the textbooks and thus will not be
length of the deformed beam axis, and the element repeated here. However, these matrices are given as

(5)
bending forces are calculated from the dot product of follows:
the bending matrix (including geometric stiffness

1
matrix) and the total deformational nodal rotations
of the element. 12 6L -12 6L

Kinematics of individual member _“1: _E2 -;; _;;*


If the undeformed and deformed axes of the beam
6L 2L2 -6L 4L2
member are assumed to be the cubic Hermitian
1
Nonlinear finite element analysis of elastic frames 695

36 3L -36 3L Since the virtual nodal displacements 6{&} are


arbitrary, the axial nodal forces are obtained from
-36 3L _$ -;; 1;; (6) eqn (11) as

3L - L= -3L 4L= {FM} =$4&{-l, I}. (12)


0

and
Because the assumption of the small strain, S/S, in
(7) eqn (12) is approximated to unity in this paper, and
eqn (12) is thus reduced to

where L = So is the initial arc length of the beam axis, {FM} = AEQ,{-1, I} (13)
F,, is the axial nodal force at node 1, A is the
cross-section area of the beam member, and E is for numerical computation.
Young’s modulus. The internal bending forces are calculated by
It should be mentioned here that the element
stiffness matrix obtained by the direct superposition
of [RB], [R,], and [R,] given in eqns (5x7) is the
{r?,1= K&(&J - (4:)) + W’,I{~, (14)
same as that used in [l], which Yang obtained by the
variation of strain energy.

Element internal nodal forces


The element internal nodal forces are calculated by
the total nodal deformational rotations. The axial
nodal force vector {F,,,} = {Fx,, Fx2}, see Fig. 1, can and {QOg)= (0, 8:, 0, 87)
be evaluated by introducing nodal virtual displace-
are shown in Fig. 1; 8:, i = 1, 2 are initial nodal
ments 8(&M) = a{&, &} at nodes 1 and 2 in the f
angles; [R,] and [R,] are bending. and geometric
direction, and equating the work done by the axial
stiffness matrices given in eqns (5) and (6).
nodal force {FM} going through the virtual displace-
ment 6{iiM} to the work done by the internal stress Equilibrium equations
resultant T going through the virtual strain S& (that
The nonlinear equilibrium equation may be
corresponds to the imposed virtual displacement)
expressed by
along the deformed beam axis as
{$f=W-1P]={O]? (15)
Tsr, dr, (8)
where {4) is the unbalanced force between the
internal nodal force {F} and the external nodal force
where c is the current chord length of the beam axis. 1(P); L is a loading parameter and {P} is a nor-
The stress resultant T may be given by malized loading vector, the direction of which needs
to be updated with respect to the current deformed
T = AE&, (9) shape of the structure when it is subjected to non-
conservative loads.
where A is the cross-section area; E is Young’s In this paper, a weighted Euclidean norm of the un-
modulus; &, is the membrane strain of the current balanced force [lo] is employed as the error measure
deformation. From eqn (4), the virtual strain SCM of the equilibrium state during the equilibrium iter-
corresponding to 8{iiM} can be expressed as ation, and the convergence criterion is given by

sr, = 6(S - S,)/S, II{4]II


P --
- Nil, s PIOIl (16)

1 where N is the number of the system equations and


pto, is a prescribed value of error tolerance.
(10)
NUMERICAL ALGORITHM

in which 6c is the variation of the chord length of the An incremental iterative method based on the
beam axis with respect to a{&}. Newton-Raphson method is adopted here. In order
Substituting eqns (3), (9) and (10) into eqn (8) gives to deal with the limit points and snap through, the arc
\ length of the incremental displacement vector is kept
s{t&,}‘{Fy) = 8{z&,}’ constant during the equilibrium iteration using
Crisfield’s method [8,9].
696 Kuo MO Hsuo and FANG Yu Hou

If the equilibrium configuration of the Zth in- From eqns (3), (4), (13), (14) and (22), the element
crement is assumed to be known, the system tangent internal nodal force can be evaluated in the current
stiffness matrix [Kr] can then be calculated at this body attached coordinate. the loading parameter 1
configuration and an initial displacement increment corresponding to the current configuration is given
A{q} for the next incremental step may be obtained by 1 = Iz’+ AL, where 1’ is the convergent loading
by using Euler predictor as parameter at the Zth increment and A1 the loading
parameter increment. Then the unbalanced force {4 }
(17) can be obtained from eqn (15). If the convergence
criterion eqn (16) is not satisfied, a displacement
where Arl is the initial incremental loading parameter correction {r} and loading parameter correction
and {qT} = [K,]-‘{P} is the tangential displacement 61[8,9] are added to the previous A{q } and Ad to
for unit loading {P}. For all increments other than obtain a new incremental displacement and incre-
the first, AI, is obtained much in the same way as that mental loading parameter for the next iteration. The
mentioned in [8], and is given by values of {r ) and 61 may be determined by

AA = ~A~l({q,}‘{q,})“‘, (18) (23)

where the sign is chosen following an approach due and


to Bergan and Soreide [1 1] in which the sign follows
that of the previous increment unless the determinant Aa2 = (A{q) + {r))‘(A{q} + {r)), (24)
of the tangent stiffness matrix has changed the sign,
in which case a sign reversal is applied. Au is the where [K,] may be the tangent stiffness matrix at some
incremental arc length used for the next increment, known configuration. This procedure is repeated
and is determined by until the convergence criterion is satisfied.
It should be mentioned that during the 6rst few
Au = C,(.ZD/J,)1’2Aa, (19) equilibrium iterations, the values of the element
nodal forces obtained from eqn (13) may be several
and
orders larger than their convergent values for certain
C,IAa/Aa,IC,, (20) problems. This may cause difficulty of convergence or
even divergence for a large increment. In order to
where Aa, is the arc length at the Zth increment; JI is overcome this difficulty, a two cycle iteration scheme
the number of iterations required to achieve equi- is employed for each increment. In the iterations of
librium at the Zth increment respectively; JD is the the first cycle, the F,, used to obtain [R,] in eqn (6)
desired number of iterations; the ‘safety factor’, C, , and {FB> in eqn (14) is the convergent values of the
lies between 0.7 and 1.0, and the ‘cutout parameters’ previous increment, and the element axial nodal
C, and C, are chosen to be 0.2 and 1.5 respectively forces {FM} are obtained from the current defor-
to avoid yielding a too large or too small incremental mation by using eqns (3), (4), (9) and (13). The
displacement. iterations are repeated until a convergent solution is
Let 6:, i = 1,2, Xi, Y:, i = 1,2 and f’, y’shown achieved. In the second cycle, the procedure is the
in Fig. 1 be the total deformed nodal angles (includ- same as that of the first cycle, except that the
ing the initial nodal angles), global nodal coordinates convergent Fxl just obtained from the first cycle is
and the axes of the body attached coordinate re- used to replace the convergent F,, of the previous
spectively for a single element at the equilibrium increment used in the first cycle. The convergent
configuration of the Zth increment, and A{q,} be the solution of the second cycle is used as the solution for
incremental nodal displacement vector of a single each increment.
element extracted from the system displacement
increment A{q }, and given by
NUMERICAL STUDIES
A{qe}={AU1,AV,,Ae,,Av,,Av,,Ae,}, (21)
Cantilever beam with two lateral loadr
where AU,, AVi and ABi, i = 1, 2 are the incremental The large deflection behaviour of the cantilever
nodal displacements and rotations in the global co- beam with two lateral loads shown in Fig. 2 is
ordinate. Then the current nodal coordinates Xi, Yi studied. This example is chosen because earlier com-
of the element are obtained by adding AU, and AVi parisons of numerical results of different methods are
to Xfand Y,r respectively. Let a (Fig. 1) denote the available [5,12,13]. The error tolerance ptol = lo-’ is
angle of rigid body rotation measured from 2 ’ axis to used for equilibrium iterations, and the stiffness
the current 2 axis, then the current deformed nodal matrix is updated only for the first four iterations of
angles are given by each cycle per increment. The end displacements
corresponding to P = 0.85 obtained by the various
8,=8:+Ae,-a, i=l,2. (22) methods are taken from [5] and listed in Table 1. In
Nonlinear finite element analysis of elastic frames 697

Table 1. Comparison of solutions for cantilever beams


(Deflection at C) Length
Number of Number of
Methods elements increments Horizontal Vertical
Elastica 0.3081 0.655 1
Ream-Column 2 20 0.308 1 0.6491
Jennings-NR 2 20 0.2980 0.6497
Powell-NR 10 20 0.0765 0.3436
Argyris 2 20 0.3521 0.7314
Martin 2 20 0.3505 0.7219
Wen-Rakimzadeh 2 20 0.3496 0.7110
Present 2 2 0.3047 0.6543

the solutions obtained by various methods using


2c IP 159P
20 increments. A comparison of Martin’s method,

E+=i
E.30x106
Y
c Wen-Rahimzadeh’s method and the present method
is given in Fig. 2 in which a beam-column solution
is also plotted as the exact solution. Apart from the
A’02 results of the present study, the results in Fig. 2 are
I . 0.000167
reproduced from [5]. The numbers (i,j) in Fig. 2
represent the number of iterations required for the
0 : Present salution first and second cycles of the two-cycle iteration
A: Beam -column
scheme respectively.
B: well (5 steps)
C: Martin (20 steps1
0: Won (2 steps) Portal frame
The structure-load system is shown in Fig. 3. Each
V/L
member of the frame is represented by three equal
elements. Four loading cases are considered. At each
Fig. 2. Comparison of results for cantilever beam subjected
to two lateral loads. increment, the stiffness matrix updating is only per-
formed at the first two iterations of each cycle. The
average number of iterations per increment is about
this example, the elastic’a solution [14] may be re- nine with ptO,= 10e4. In Fig. 3, the present results are
garded as the ‘exact’ solution. The beam-column plotted together with the results given by Oran and
solution [q is obtained using the tangent stiffness Kassimali [13]. It is seen that the present study uses
matrix given by Oran [ 151and the Newton-Raphson only two increments for the loading case II = 0.001,
(NR) method. It is seen that the present solution, three increments for n = 0.01 and 0.1, and five in-
although it uses only two elements and two in- crements for n = 0.5.
crements, is as accurate as, if not even more so than, Although the agreement among most of the results

I n ~0.001

120
Pb
ap
-

Ii
nP
B C
c
6
G
A D
I

-e- Present study I a310.1


Ill4
-h [I31 A * I I.77 in2

I I I I I I I I
D 20 40 60 60 100 I.20 140
Horizontal deflection of point B

Fig. 3. Load-deflection curves for portal frame.


698 Kuo MO H~IAOand FANGYu Hou

shown in Fig. 3 is quite good, there is a significant


deviation for the case n = OS. One limit point is
obtained in the present study but none is reported in
[13]. The reason for this discrepancy may be that the
number of elements used in [13] is insufficient (the
number of elements is not mentioned in [13]). US0
The structure has also been analysed by the au- Lx IOm

thors using the discretization of one element per b=lm


h *O.lm
member and four elements per member. It may be of
interest to mention that the results obtained using
three and four elements per member are nearly the I
same, and the results obtained by one element per
member are nearly the same as that obtained in [13].

Cantilever beam with an end moment


This example considered is a cantilever beam sub-
jected to a concentrated moment at the free end as
shown in Fig. 4. The beam was discretized by ten
elements. The results shown in Fig. 4 are obtained by
using only four increments with pto,= 10-4. The
number of iterations is about six per increment. As
-- Load P”F tl61
can be seen, the agreement with the analytical sol-
ution is quite good. This example is extensively
studied in the literature to demonstrate the efficiency Displacement (ml
of numerical methods and the large rotation capabil-
Fig. 5. Cantilever beam subjected to conservative and
ity of the beam, plate and shell elements. To the nonconservative loading.
authors’ knowledge, it is not reported in the literature
that the cantilever beam is bent into a full circle by
using only four increments, or less. force P,, whose direction is always perpendicular to
the current neutral axis of the beam. This beam is
Cantilever beam with end force discretized by ten equal elements, and the error
The cantilever beam portrayed in Fig. 5 is selected tolerance is chosen to be 10a4. The results shown in
to study the effect of nonconservative loading. To this Fig. 5 are obtained by using four increments for each
end, two loading conditions are considered, the first case. The average number of iterations per increment
of these being that of a standard conservative load P, used is five for the first case and eight for the second
with 6xed direction normal to the initial straight case. Also shown in Fig. 5 are the solutions in [16]
beam. Nonconservative loading is simulated by a obtained by using ten increments. Very good agree-
ment between the solutions of the present study and
,h [16] is observed.

Simply supported beam


Here the elastic post-buckling behavior of a simply
E = l.2x105kN/m2 supported beam shown in Fig. 6 is studied. A small
Y = 0 transverse load was added to act as an initial imper-
L= IOm fection. The number of elements used is four for half
b: Im the beam and the error tolerance used is 7 x 10m4.
h = O.lm
Results are shown in Fig. 6, where they are compared
with the ‘Elastical’ solution [17]. The number of
iterations for each cycle is also given in parentheses
- Analytical solution beside each point on the graph. It can be seen that
five increments are used and the average number of
iterations for each step is eight.
This example was also analysed by Epstein and
Murray [2], who used 23 increments. To the authors’
knowledge, it seems that of all the researchers who
have studied this example, only the present authors
use less than ten increments.
Displacement (ml
Clamped-hinged deep arch
Fig. 4. End displacements of cantilever beam with end
moment. The clamped-hinged deep arch depicted in Fig. 7
Nonlinear finite element analysis of elastic frames 699

E = 10.3 x 106ib/in2
V’O X-section
0.753’
m 0.243”
H (lb)
1002003C#400500600i

- The Elostico tl71


0 Present

0
I 0.1
I
02
I
0.3
I i
0.4
1

A/L
Fig. 6. Load-deflection curve for simply supported beam.

is subjected to a point load at the crown. Thirty 0.2 0.3 0.4 0.5 0.6
elements are used for discretization and the error V (in.)
tolerance ptol = 10e3 is used for equilibrium iteration.
Fig. 8. Williams toggle: geometry and load-deflection
Eleven incremental steps are used in the analysis, and curves.
the average number of iterations per step is seven.

c@%r
The crown load deflection curves presented in Fig. 7
“‘P show good agreement with the analytical results 1181.
V Williams toggle frame

/pr t
Here the Williams toggle frame [19] shown in
R
Fig. 8 is studied. Due to symmetry of geometry and
.c- a ; A.. deformation, only half of the frame is analysed. Five
elements are used for discretization and the error
R = 100, jd =215’ tolerance 10m4is used for equilibrium iterations. Five
t=1, EI - IO6 increments are used for this example and the average
number of iterations per increment is about seven.

Lee’s frame
This example deals with the snap through of the
frame shown in Fig. 9. This frame is discretized by 20
equal elements. The results shown in Fig. 10 are
obtained by using 20 increments with the error
tolerance 7 x IO-*. The average number of iterations
per increment is about seven. It should be mentions
that at the difficult point A (Fig. 9) the error tolerance
used is 3 x 10e3. Also shown in Fig. 9 are the results
reported in [7] and (201. The number of increments
used is not mentioned in [;1.

Circular ring
- - Analytical [I81
Consider a circular ring subjected to uniform pres-
Present solution
sure as shown in Fig. IO. The direction of the pressure
is always perpendicular to the current surface of the
0.2 0.4 0.6 0.8 I 1.2
ring. The ring has an initial imperfection of the form
U/R, V/R
wi= w,cos2/?.
Fig. 7. Chunped-hinged arch: geometry and load-
deflection. Two cases of initial imperfection, W,/R = 0.008
700 Kuo MO HSIAOand FANG Yu Hou

24 cm CONCLUSIONS
1 96cm
I
This paper describes a practical formulation and
procedure for the geometrically nonlinear finite ele-
ment analysis of in-plane framed structures. The
element stiffness matrix is obtained by superimposing
the bending and geometric stiffness matrices of the
elementary beam element and the stiffness matrix of
the linear bar element in a body attached coordinate.
The nonlinear formulation has been based on the co-
rotational formulation by which the major geometric
nonlinearities were shown to be embodied in the
coordinate transformation when element assemblage
is formed.
In the formulation presented here, a body attached
coordinate is adopted to distinguish between a rigid
body and deformational rotations and accumulate
the latter incrementally along the deformational path.
The membrane strain of the beam element is assumed
to be constant along the beam axis and is evaluated
from the elongation of the arc length of the beam
axis. The element axial nodal forces are obtained
from the membrane strain, and the element bending
forces are evaluated from the dot product of the
element bending stiffness matrix (including geometric
stiffness matrix) and the total deformational nodal
Fig. 9. Load-deflection curves for Lee’s frame. displacements. Due to the definition of the body
attached coordinate, the nodal rotations are the only
non-zero deformational degrees of freedom. This
and 0.024, are studied. Due to double symmetry, only formulation removes the restriction of small nodal
one-quarter of the ring is discretized in the present rotations between the successive increments.
study. The analysis is carried out with 26 equal It has been demonstrated by nine examples that the
elements and error tolerance 10m4. For both cases present formulation and procedure are valid for very
four increments are used and the average number of large displacements and rotation increments. Despite
iterations per increment is about five. In Fig. 10 the the fact that the formulation is very simple, highly
results of the present study are compared with those accurate solutions are obtained. It is believed that the
reported in [21]. use of the simple element and procedure in this paper
may represent a valuable engineering tool for the
solution of nonlinear in-plane framed problems.

REFERENCES

1. T. Y. Yang, Matrix displacement solution to elastica


moblems of beams and frames. ht. J. Soliak Struct. 9,
‘829-842 (1973).
2. M. Epstein and E. W. Murray, Large deformation
in-plane analysis of elastic beams. Compur. Struct. 6,
l-9 (1976).
3. R. D. Wood and 0. C. Zienkiewicz, Geometrically
nonlinear finite element analysis of beams, frames,
arches and axisymmetric shells. Comput. Srruct. 7,
725-735 (1977).
P,= E/4(+3 4. M. H. El-Zanaty and D. W. Murray, Nonlinear finite
element analysis of steel frames. J. Struct. Engng AXE
109(2), 353-j68 (1983).
5. R. K. Wen and J. Rahimzadeh, Nonlinear elastic frame
analysis by finite element. J. S&t. Engg ASCE 109,
1952-1971 (1983).
6. C. Chebl and K. W. Neale, A finite element method for
elastic-plastic beams and columns at large deflections.
Comput. Struct. 18, 255261 (1984).
17. C. Cichon, Large displacements in plane analysis of
elastic-plastic frames. Comput. Struct. 19(5/6), 737-745
Fig. 10. Load-deflection curves for circular ring. (1984).
Nonlinear finite element analysis of elastic frames 701

8. M. A. Crisfield, A fast incremental/iterative solution 15. C. Oran, Tangent stiffness in space frames. J. Stat.
procedure that handles snap-through. Comput. Struct. Diu. AXE 99(ST6), 973-1001 (1973).
13(1-3), 55-62 (1981). 16. G. Horrigmoe and P. G. Bergan, Nonlinear analavsis of
9. T. Y. Chang and K. Sawamiphakdi, Large deflection free-form shells by flat finite elements. Comput. Meth.
and post-buckling analysis of shell structure. Cornput. Ap,ul. Mech. Enana 16. 11-35 (1978).
Me&. Appl. Me&. E&ng 32(-3), 31 l-326 (1982): 17. S: P. Timoshenko and J. M. Gere,’ Theory of Elastic
10. A. K. Noor and J. M. Peters. Tracina uost-limitina Stability. McGraw-Hill, New York (1961).
point with reduced basis technique. co&w. Meth: 18. D. A. Da Deppo and R. Schmidt, Instability of
Appl. Mech. Engng. 28, 217-240 (1981). clamped-hinged circular arches subjected to a point
11. P. G. Bergan and T. Soreide, Solution of large displace- load. Trans. ASME., 894-896, Dec. (1975).
ment and instability problems using the current stiffness 19. F. W. Williams, An approach to the nonlinear behav-
parameter. In Finite Elements in Nonlinear Mechanics, iour of the members of a rigid jointed plane framework
Vol. 3, pp. 647669. Tapir Press, Norway (1978). with finite deflection. Q. J. Mech. Appl. Maths. 17(4),
12. A. M. Ebner and J. J. Ucciferro, A theoretical and 451-469 (1964).
numerical comparison of elastic nonlinear finite element 20. S. Cescotto, Etude par elements finis des grands de-
methods. Comput. Struct. 2, 1043-1061 (1972). placements et grands deformations. Doctoral thesis,
13. C. Oran and A. Kassimali, Large deformations of Liege (1977/1978).
frames structures under static and dynamic loads. 21. S. Kyriakides and C. D. Babcock, Large deflection
Comput. Struct. 6, 539-547 (1976). collapse analysis of an inelastic inextensional ring under
14. R. Frisch-Fay, A new approach to the analysis of the external pressure. Int. J. Solidr Struct. 17, 981-993
deflection of thin cantilevers. J. Appl. Mech., 87-90, (1981).
March (1961).

C.&S. 26,4-K

Вам также может понравиться