Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
ELSEWER
COmpUtlPlr
methods
in applDed
me&aNiGs and
engineering
Deparrment,
Received
The Universit)
19 August
1996;
of Queensland,
Abstract
Instead of the traditionat load-following process, this paper presents large deflection, post-buckling analysis of the threedimensional etastic frame from a dynamic point of view. A co-rotationat formulation combined with small deRection beam theory
with the inclusion of the effect of axial force is adopted. The pre-buckling anafysis employs the constant arc length method together
with the modified Newton-Raphson iteration method and the extrapolation technique to improve the convergence behaviour. As
the limit state being examined is passed, the previous converged solution is adopted to start the nonlinear dynamic analysis, based
on the average acceleration of the Newmark algorithm with a slow rate of load incrementation, to trace the load-deflection path
beyond the limit point. As a result, the snap-through problem is overcome without down loading. The ef6ciency of the damping
coefhcients on the control of dispiacement overshooting and vibration, which are the common phenomena using this procedure
under conventional structural damping, is also studied and appiied to all the numerical examples. This study is demonstrated by
application to typical three-dimensional frames which show snap-through behaviour in static analysis with any increment of load
after the Limit point. @ 1998 Elsevier Science S.A.
1. Introduction
A limit point corresponding to the primary path and a bi~~r~~~~~~
point corresp~~~ding to the secondary
path have been recognised as the critical state where the determinant of the tangent stiffness matrix
vanishes. However, the determinant criterion alone cannot distinguish a bifurcation point from a limit
point. The incremental work criterion [l] is found to be superior in order to trace both the primary and
secondary paths. In the presence of imperfection either in load or in geometry, a bifurcation point can be
degenerated into a limit point [2--G].The direction and location of the imperfection can be searched for
by the eigenvalue approach by Meek and Loganathan 151,or alternatively the energy method by Chan
(41.There is no static equilibrium state adjacent to the limit point if the external loading is continuously
increased. That is, any increase in the loading causes a sudden snap in the displacements. In order
to trace an adjacent load~~splacement
equ~I~~rium path, the convcnt~onal load following process has
been used for decades. However, this procedure is somewhat artificial and it does not tell us, whether
the solutions obtained correspond to the states at which the actual structure is coming to rest. It is of
interest to determine what happens to the structure if loading is continued beyond the limit point and if
an adjacent load-displacement
path can be traced without decreasing the load. A reliable way to solve
this particular problem is to simulate the dynamic behaviour of the structure as soon as it has reached
an unstable critical state. This was achieved previously by the authors for two-dimensional frames [BJ
Corresponding
author.
236
Methds
Appl.
2. Formulation
strategy
(x,-y+;i)
0,3+
(-q+g)
0223
(2)
where 0,i, Hzj are the member rotations
at ends 1 and 2 about
member length. The axial change in length ub due to the bowing
LLb =
du
J (->
l/2
SC >
2
dx, + l/2
hl
dx,
(3)
The axial load P is then given as
p=-
EAe
1
EA
+ 30
w,,
430223
+2@3,
m,',
42022
2fgJl
(4)
(b) Member
intermediate
237
and forces
displacements
and forces
FlO
AS10
wi
A
/
F6
AS3
-x1
AS6
x3
(c) Member
Fig. 1. Member
nodal displacements
deformations
and forces
and associated
forces.
A&, 0, and GJ are, respectively, the torsional moment, angle of twist and the torsional rigidity associated
with the member. The total potential energy !P is expressed as
~ Ml3013
=
y
-
(H,:,
Ml343
+ g
M23022.1 -
Ml242
[(2q25
M23023
-
6363
M22&2
+ y2
Ml2012
+ 2g3)
+ P4,
M22Q22
+ (q22
62022
(6)
+2&I
Mn=
(t;+rl:;il)
o13+ (T-3
1 and
2 about
the j axis,
023
and
(10)
Note that P, Mt, M13, M23, M12. M22 are the member
2.1. Incremental
stiffness
kll
22
k3.l .
AMu
ku
AM22
AM,
sym.
AP
axes.
by partial
k16
A@13
kxi
A823
k 36
A42
kc,
A022
differentiation
(11)
A&
Ae
2.2. Lumped
matrix K, is obtained
deformations.
Hence
.
k
AMa
matrix
AMu
member
force and tangent
stiffness matrix
be referred
to in [17]. The element
can be obtained
by the transformation
from the member basic
global coordinates
to member
intermediate
force/displacement,
and from this to global nodal
(Fig. 1). As mentioned
above, attention
must be paid to the updated transformation
rotation of a joint in a space frame analysis cannot be treated as a vector. Here, use
orientation
matrix to account for the finite rotations.
Details can be found in [17].
matrix of the structure K is then obtained by standard assemblage.
ma.u matrix
239
about xi, x2 and x3 axes are given by plfl? + 13), 1/12pA13 and 1/12pAi3,
respectively, and half is apportioned to each node.
The total mass M of the structure is then obtained by standard assemblage after transformation from
local axes to global.
[20].
C=aM+flK
(12)
to two
3. Numerical algorithm
The numerical algorithm is similar to that used in [6]. However, Eq. (19) of [6] is not valid for the finite
rotations in the present three-dimensional
frame analysis because of their non-vectorial characteristic.
Again, use is made of the Orans joint orientation matrix in dynamic analysis already used in static
analysis. From Eq. (17) of 161, the small rotation increments with respect to the global X1,X, and
X3 axes, A&$ where i = 1,2,3, can be predicted. Then, the incremental rotation matrix for a joint is
expressed as
[hR]=
[~~~~
z3
_~~~j
(13)
(14)
Hence, the translational displacements and the joint orientation matrix must be updated properly so as
to evaluate the updated tangent stiffness matrix correctly when using NR iteration method and updating
the final displacement vector.
4. Numerical examples
The Williams toggle frame shown in Fig. 2 has been studied previously by the authors in [6], where
the displacements are overshot by the inertia force and decay with the damping provided. It is used
here as the example to demonstrate the efficiency of the damping coefficient on the control of the
overshooting. A mass density p = 0.282 lb/in3 is used with a time step of 0.05 s and a load-increment rate
of 0.002632 lb per second. Several damping coefficients were tested. Firstly, C = 0.7579M + 0.000306K
where the coefficient is evaluated according to the first two lowest natural frequencies of the original
structure. Secondly, C = 0.1346M + 0.000364K where the coefficient is evaluated based on the frequency
of the structure around the limit state. The third trial uses C = O.lK. Finally, C = 21.255M is adopted.
The displacement versus load curve of the final case is depicted in Fig. 3. The results around the snapthrough point A in all the cases above are given in the figures in Table 1. It is found that dynamic
vibration exists in the first three cases and is absent in the last one. This shows how the displacement
overshoot can be numerically controlled by using a relatively large damping given in terms of the mass
only. The damping coefficient for the additional
examples have been chosen based on this result.
240
EA = 1.855
X 1061b
El = 9.27 X 1 031b=in2
Fig. 2. Geometry,
material
properties
00
01
11,
04
DEFLECTION,
Fig. 3. Load-displacement
Table 1
Effection
of damping
C = 0.7579M
coefficient
+ 0.000306K
4.2. Clamped
shallow
on the control
C = 0.1346M
of displacements
+ 0.000364K
response
frame
,,I
03
02
of toggle
of the toggle
05
(I)
frame.
overshooting
C = O.tK
C = 21.255M
arch
Fig. 4 illustrates
a clamped shallow arch with a concentrated
load acting at its apex. A finite element
solution has been presented
by Hsiao [21] and Sandhu et al. [22]. The results of the present study are
obtained
by using two elements for half the arch. In the nonlinear
dynamic analysis, a mass density of
0.000078 N/mm3 is used and a Rayleigh damping of C = 15.0 M is applied to numerically
control the
overshoot.
The time step is 0.025 seconds. The load is increased linearly with the rate of 0.034897 N/s.
241
P
f
Y
1
d
k-_
a
-X
-I
L = 863.6mm
b = 25.4mm
a = 27.69mm
d = 14.99mm
R=3381.lmm
E = 6.895 X 104MPa
Fig. 4. Geometry,
material
property
of the shallow
arch.
Fig. 5 indicates that by the current method, the adjacent displacement can be traced after the limit point
with the slow increment of load. The snap-through point A given by the usual load following process is
reached herein by the nonlinear dynamic analysis. The amplified figure around the snap-through point
is presented in Fig. 6 with a corresponding change of scale. Overshooting of the displacements from the
static equilibrium position is well controlled numerically by the large damping. The solution after the
snap-through point is almost identical to that obtained from the conventional nonlinear static analysis
using the down loading technique.
10
.
.
Present analysis
JSSandhu et al.
20
30
DEFLECTION, v(mm)
Fig. 5. Load-displacement
response
of the shallow
arch.
242
J. L. Meek, Q. Xue/Comput.
34 90
35 10
35 00
DEFLECTION,
Fig. 6. Enlarged
figure around
the snap-through
35 20
v(mm)
arch.
The two-hinged deep arch depicted in Fig. 7 is studied with a concentrated load applied at its apex.
This problem has been treated extensively by Huddleston [23], Da Deppo and Schmit [24], Wood and
Zienkiewicz [2.5], Tan [17] and Loganathan [26]. The results from some of the literature are plotted in
Fig. 8. In this research, the structure is modeled with eight elements. It is found that this structure
exhibits multiple instability phenomena and imaginary roots occur in Eq. (10) of [19] when using the
arc length method. Better control can be achieved by reducing the arc length and the arbitrary external
load, or artificially suppressing the equilibrium iteration and simply using pure incrementation. The
buckling modes of the structure are characterised by two asymmetric modes, one out-of-plane and one
in-plane, and by a symmetric in-plane limit point. In the present analysis, all these modes have been
traced, the lowest of which being the out-of-plane asymmetric mode (Fig. 8). The imperfect approach
is used in which a small perturbing load is introduced to degenerate the bifurcation point problem into
a limit point problem so as to find the secondary path. The out-of-plane buckling is traced with a force
0.1 P in the 2 direction at the crown and the hinge supports rotate freely about the Z axis only. The
in-plane asymmetric buckling mode is found with a moment 0.00133 PL about the Z axis at the crown.
In the nonlinear dynamic analysis, a mass density of p = 0.282 lb/ in3 and a time step 0.1 s are used. The
Rayleigh damping of C = 3.0 M, C = 3.0 M and C = 5.0 M for the out-of-plane, in-plane asymmetric
and in-plane symmetric buckling modes, respectively, are applied. The corresponding load is increased
linearly with the rates of 0.00638 lb/s, 0.010505 lb/s and 0.01291 lb/s, respectively.
b__~
~~~
L = 160
Lo
thickness
R=looln
a=lo,n
Fig. 7. Geometry,
material
---I
t =1I
areaA=,
In2
E=lOX1061b/1n2
property
J. L. Meek, Q. Xue/Comput.
10
20
30
40
50
60
70
243
80
DEFLECTION,v(in)
Fig. 8. Load versus
the horizontal
displacement
The curves for load versus the vertical deflection of the different modes are drawn in Fig. 8. The
enlarged figures around the corresponding snap-through points are shown in Figs. 9, 10 and 11, respectively. The overall deformation histories of the arch under the different buckling modes are presented
in Figs. 12, 13 and 14, respectively, and the projections of Fig. 14 on XY and ZY planes are plotted in
Table 2. Results indicate that the unsymmetric buckling mode curves finally join the symmetric branch.
Overshooting of the displacements due to the inertia force is perfectly controlled numerically by the
damping for the two asymmetric buckling modes. With the current damping used, the displacement
overshoots only 0.37% from the snap-through point for the in-plane symmetric buckling mode.
zz1294.70
9
3
1294.65 -
1294.60 -
1294.55 -
83.0
83.1
83.2
83.3
83.4
DEFLECTION,v(in)
Fig. 9. Enlarged
figure
around
the snap-through
mode.
244
J.L. Meek.
Q. Xue/Comput.
Methods
,,,*k ,
Appl.
,
82.5
82.0
83.0
DEFLECTION,v(in)
figure around
the snap-through
point
B of the in-plane
asymmetric
mode.
640.6
640.4
640.2
____I
6400
6398
639.6
82.4
/I,
,/
82.5
82.6
82.7
OEFLECTION,v(in)
figure around
-50
deformation
the snap-through
50
asymmetric
X(h)
the symmetric
buckling
mode
mode
deformation
245
the in-plane
asymmetric
buckling
mode
, A(in)
k
Fig. 14. The overall
4.4. Star-shaped
shallow
deformation
Z(in)
the out-of-plane
asymmetric
buckling
mode.
dome
246
Table 2
The out-of-alane
asvmmetric
Projection
buckling
historv
in XY and ZY okme
of Fig. 14 on XY plane
Projection
all dimensions
jY*
in mm
/
82.16
v20
A
7
_
I_
250
25%. I
~_I_..
_, .
433
._
A_& x u _
4,
433
ix=0918X104mm4
ly=2377X104mm4
lr=0295X104mm4
E = 3.03
G=1.096X103N/mm~
material
of Fig. 14 cm ZY plane
X 103 N/mm2
property
of the star-shaped
dome.
load-increment rates are 0.056 N, 0.00607 N and 0.005864 N per second for load conditions (a). (b) and
(c), respectively. Results are shown in Figs. 16, 18 and 20 and their details around the corresponding
snap-through points A, I3 and C are plotted in the enlarged Figs. 17, 19 and 21 for load conditions (a),
(b) and (c), respectively. Again, adjacent load-deflection responses are traced in all cases after the limit
point up to the snap-through point without down loading. The overshooting effect of the inertia force on
the displacements is also well controlled. The snap-through points achieved here are in good agreement
with previous research, especially for case (b) which gives better results than those in [17], where the
same element model is used, and is much closer to the results of [15], where a 72-elements modef was
used.
247
.
A
A
A
___
Present analysis
H.S.Tan
Kuo-MO Hsiao
15
10
I
30
25
20
DEFLECTION, v(mm)
Fig. 16. Load-displacement
response
of the star-shaped
frame
under
concentrated
central
load.
2-T
C563.0
3
-I
562.5 -
562.0 -
561.5 -
561.0
t
32.4
32.5
32.7
32.6
DEFLECTION, v(mm)
Fig. 17. Enlarged
figure
around
the snap-through
point
A of the star-shaped
frame
under
concentrated
central
load.
J. L. Meek, Q. Xue/Comput.
248
600
f
iT
,,,
I!
8 500
400
300
.,
.;
Present analysis
H.S.Tan
Kuo MO Hsiao
/
a
200
i
i
Pi2 P
.
P/Z
;
loo
0
0
30
20
10
40
DEFLECTION, v(mm)
Fig. 18. Load&displacement
response
of the star-shaped
frame
under
symmetrical
load
615
610
605
figure around
the snap-through
point
B of the star-shaped
frame
under
symmetrical
load.
249
$j 500
400
300
Present analysis
HSTan
Kuo MO Hsiao
A
.
200
100
L,
I<
10
>
11
,I
I,
30
20
DEFLECTION, vfmm)
Fig. 20. Load-displacement
response
of the star-shaped
frame
under
unsymmetrical
load.
589
586
585
k
I
I
37.6
37.7
37.8
37.9
38.0
38.1
DEFLECTION, v(mm)
Fig. 21. Enlarged figure around
the snap-through
point
C of the star-shaped
frame
under
unsymmctrica1
load.
250
v = 0.3
G = 2.652 X 1 ON/m
E = 6.895 X 1ON/m*
Xl
A=6.5X10-4m2
a = 0.75 m
f = 0.6 m
L = 6.0 m
Equation of surface:
XT+ X2,+ (Xs + 7.2)2= 60.84
Xl
material
property
of the geodesic
dome
2140
K\
i?
:
1
120 I
,iAYM
L
100
60 -
60
40 -
20 -
O~,,,,,,,,~
0.0
0.1
0.2
0.3
0.4
0.5
06
0.7
0.8
0.9
DISPLACEMENT,
response
of the shallow
geodesic
dome
10
w(m)
at node
0
0.0
~,11111,,1111111111
0.1
0.2
251
0.3
0.4
0.5
0.6
0.7
0.6
0.9
DISPLACMENT, w(m)
Fig. 24. Load-displacement
2
Y
response
of the shallow
geodesic
dome
at node 2.
140
z
9
2
120
60
0.0
0.1
II/
I,
0.2
1 ,
0.3
I,
0.4
0.5
0.6
DISPLACEMENT, w(m)
Fig. 25. Load-displacement
response
of the shallow
geodesic
dome
at node 8.
252
/
/
40.
:
20.
'I
,
OV1",""""'
000
",','I
002
0.04
006
'1,"
0.08
',I",
0.10
012
DISPLACEMENT,
response
of the shallow
geodesic
w(m)
dome at node 20
the structure
shape corresponding
to the marked points A to M in Fig. 23 and for further loading is
illustrated
in Table 3. The results of the first three snap-through
points are in excellent agreement
with
that obtained in [26]. Unfortunately,
the authors arc unable to compare later results, because in [26] an
imperfection
of unrecorded
magnitude
was used and some data appears to be incorrect from the authors
point of view.
5. Conclusions
A more realistic post-buckling
behaviour
of elastic three-dimensional
frames has been described in
this paper by using a nonlinear
dynamic analysis with small load incrementation
after the limit point.
The results show that there is no sudden snap-through
in the load-deflection
path without down loading.
Overshooting
of displacements
due to the inertia force can be well controlled
by using a relatively large
damping factor which is related to the mass only. The snap-through
point achieved in this way without
decreasing
the load is very close to that obtained from the conventional
nonlinear
static analysis, using
load-following
procedures
after the limit point, carried out by previous researchers.
The results following
this point are almost the same as those obtained by the nonlinear
static analysis, since the load increment
in the dynamic analysis is very small. Snap-back
behaviour
of the vertical displacement
observed from
the static analysis in the geodesic dome example vanishes in the present dynamic analysis where the
load is not decreased.
The purpose of this paper has been to demonstrate
an alternative
approach
to the current loadfollowing techniques
for determining
the load-deflection
behaviour
past the limit point, where loads
are slightly increased.
It is shown that the nonlinear
dynamic analysis is capable of handling the large
rotations
experienced
in the three-dimensional
frame analysis by using small linear increments
in the
co-rotational
approach
(see Sections 4.3, 4.4 and 4.5). The example of the deep arch is interesting
because it shows that the final configuration
reached after the out-of-plane
buckling behaviour,
using
the co-rotational
matrix to accommodate
the large rotations,
is the same as that reached during the
in-plane buckling modes. Through such snap-through
phenomena,
the present dynamic approach must
be considered
a more realistic representation
of the structure behaviour.
From the examples included,
it would appear that it is relatively simple to implement
the technique.
Animation
of the process gives
a most helpful insight into the snap-through
phenomena.
253
References
nonlinear
analysis of space frames by an incremental
iterative technique,
Comput.
111J.L. Meek and H.S. Tan, Geometrically
Methods Appl. Mech. Engrg. 47 (1984) 261-282.
PI W.T. Koiter, Post-buckling analysis of a simple two-bar frame, in: Recent Progress in Applied Mechanics (Almquist and
Wicksel, Stockholm,
1966) 337-354.
(1974) 1-16.
[31 W.T. Koiter, Current trends in the theory of buckling, Buckling of Structures
numerical
procedure
for nonlinear
analysis of frames exhibiting
a limit or a bifurcation
point,
[41 Siu Lai Chan, A generalized
Int. J. Space Struct. 6(2) (1991) 99-114.
Large displacement
analysis of space frame structures,
Comput. Methods Appl. Mech. Engrg.
[51 J.L. Meek and S. Loganathan,
72(l) (1989) 57-75.
254