Professional Documents
Culture Documents
steel structure.
Contents
1. Introduction ........................................................................................................................1
Reference .................................................................................................................................39
1
1. Introduction:
Development of an efficient FE model to represent local buckling as well as combined
local/distortional buckling under different boundary conditions of cold-formed steel column is the
main objective in this paper. This paper use different softwares to simulate one model to see the
difference between different softwares and analyze reason and compare to each other which is
better. Elastic buckling analysis is the most commonly used technique to define the shape and
distribution of initial imperfections. The lowest Eigenmode is often selected as the shape of the
initial geometry of the structure [1]. However, this is not always the case and selecting an
appropriate Eigenmode is sometimes very challenging in some boundary conditions since it should
represent the buckling modes which are triggered in the member. Kwon and Hancock [2] found
that the mode of imperfection had a more significant effect on the behaviour of their sections than
the amplitude of initial imperfections. Yu and Schafer [3] used the finite strip software CUFSM to
obtain the shape of initial imperfections. In the present study, it was found to be difficult to obtain
an appropriate Eigenmode representative of pure local buckling along the constant axial force. It
was, therefore, decided to generate the initial geometric imperfections manually with the aid of
finite strip software CUFSM [4] and sinusoidal functions rather than using the conventional method
of feeding the Eigenmodes from the ABAQUS elastic buckling analysis into the ABAQUS nonlinear
analysis. The sensitivity of FE models to the magnitude of initial imperfections has also been
addressed in this paper using different CDF values proposed by Schafer and Pekz [5] for
imperfection amplitudes.
2
and even impossible to use sometimes leads to a much shorter computing time to
because of the demanding computing find a solution with nearly the same accuracy.
facilities.
Large quantity of input data which can lead to Very small amount of input data due to the
mistakes. smaller number of meshing.
Large quantity of output. Normally Easier to specify only those nodes which
displacements of all the nodes are listed. displacements and stresses are required.
Difficult to program and a very big Due to the reduction in the number of
computational requirement. degrees of freedom, the computational
requirements are smaller.
Geometric Imperfections:
I considered the geometric imperfection in Abaqus. To define an imperfection based on the
deformed geometry of a previous static analysis (Unstable collapse and postbuckling analysis),
specify the result file and step (and, optionally, the increment number) from a previous static
analysis. The response of some structures depends strongly on the imperfections in the original
geometry, particularly if the buckling modes interact after buckling occurs. Hence, imperfections
based on single buckling mode tend to yield nonconservative results. By adjusting the magnitude
of scaling factors of various buckling modes, the imperfection sensitivity of the structure can be
assessed. Normally, a number of analyses should be conducted to investigate the sensitivity of a
structure to imperfections. Structures with many closely spaced eigenmodes tend to be
imperfection sensitivity, and imperfections with shapes corresponding to the eigenmode for the
lowest eigenvalue may not give the worst case.
The imperfect structure will be easier to analyze if the imperfection is large. If the imperfection is
small, the deformation will be quite small (relative to the imperfection) below the critical load. The
response will grow quickly near the critical load, introducing a rapid change in behavior. On the
other hand, if the imperfection is large, the postbuckling response will grow steadily before the
critical load is reached. In this case the transition into postbuckled behavior will be smooth and
relatively easy to analyze.
Material Modeling:
Material nonlinearity in the cold-formed steel beams was modeled with von Mises yield criteria
and isotropic hardening. Measured stress-strain relations taken from tensile coupons from the
beams were employed. All other components were modeled as elastic, with E = 29500 ksi (203 GPa)
and = 0.3, except for the hot-rolled steel tubes and the loading beam which used an artificially
elevated modulus (10E) so that they would effectively act as rigid bodies. Residual stresses were
ignored. [7]
3
3. Comparison between each method under different simulation software:
In this paper, I use the members cross-section 600S162-54, boundary condition are: one end is
fixed, the other end constrain the displacement in 3 different ways, C-C, C-F, C-S. Uniform pressure
applied in the cross-section other than constraint one. On the other hand, set the yield stress is
50ksi as well.
4
Figure 3: Strip stress distribution
The shape functions for the transverse direction are assumed to be the same polynomial function
for every boundary condition. On the other hand, in the longitudinal direction, trigonometrical
functions are taken. These functions have to satisfy the pre-set boundary conditions. The out of
plane displacement will use a shape cubic polynomial function for all boundary conditions.
Therefore, the expressions for general displacements are as follows:
Where = and is the half-wave number. is the function for the longitudinal direction,
which varies depending on the boundary conditions.
Boundary conditon Shape function
Simply-Simply p y
Yp sin
a
Clamped-Clamped p y y
Yp sin sin
a a
Simply-Clamped
( p 1) y p 1 p y
Yp sin ( )sin
a p a
Clamped-Free 1
( p ) y
Yp 1 cos 2
a
Clamped-Guided 1
( p ) y
2 p y
Yp sin sin
a 2a
5
A typical signature curve after analysis are shown in the following:
6
Figure 6: Calculated section properties
After set the material and dimension of cross section, then use the general boundary condition
solution type for the half-wave length, because traditional BC only fit for S-S.
. Clamped-Simply supported B.C.
Set 20 eigenvalues and make m=2 means only two term of triangular series will be used for the
specified half- wave length in the finite strip method for the C-S boundary condition.
7
Figure 8: Load condition figures
8
Figure 10: CUFSM results for C-S boundary condition, mode 1
. Clamped-Clamped B.C.
This time I use m=4 in the C-C boundary condition. Because after analysis, the result with m=4 can
be closer to the Abaqus analysis result without consider geometric imperfection
9
After analysis, we get the results as shown in the following, means the local buckling controls and
when applied load is 0.28682*50ksi=14.341ksi, the elastic critical local buckling will occurs.
. Clamped-Free B.C.
This time I use m=1 in the C-F boundary condition. Because under this boundary condition, it seems
like the simply supported B.C., relative weak than other two B.C., so use minimum m value.
10
After analysis, we get the results as shown in the following, means this time global buckling
controls and when applied load is 0.063387*50ksi=3.17ksi, the elastic critical local buckling will
occurs.
11
Figure 15: Section details and material property
12
. Clamped-Simply supported B.C.
Set 20 eigenvalues and make m=2 means only two term of triangular series will be used for the
specified half- wave length in the finite strip method for the C-S boundary condition.
13
Open all the base vectors to analyze:
14
. Clamped-Clamped B.C.
This time I use m=4 in the C-C boundary condition. Because after analyze and comparison, it is
more close to Abaqus analysis result without consider geometric imperfection.
15
. Clamped-Free B.C.
This time I use m=1 in the C-F boundary condition. Because under this boundary condition, it seems
like the simply supported B.C., relative weak than other two B.C., so use minimum m value.
16
Then I summary the result in following table, with make all the other conditions are same and only
change the cross-section.
Load factor C-S B.C. C-C B.C. C-F B.C.
C-section without 0.28682 0.28946 0.063387
round corner
C-section with round 0.30151 0.30151 0.059504
corner
Difference (%) 5.12% 4.16% -6.13%
From the table, I can make a conclusion that section with round corner can improve load factor a
little under C-S and C-C boundary conditions and reduce a little under C-F B.C.
17
Load and boundary condition:
Apply load equal to the yield stress, means apply 50ksi to the cross-section. For the thin-wall
structures, I use an equivalent load application method, treat the uniform load as a concentrate
load which apply on the center of gravity of cross section which is not encased, as shown in the
following. So P=-50ksi*0.554=-27.7k, negative value because the force is pressure. For the fixed
end, I use ENCASTRE (U1=U2=U3=UR1=UR2=UR3=0) in ABAQUS.
18
. Clamped-Simply supported B.C.
For the simply-supported end, apply the pressure and constrain displacement in x and y direction
19
. Clamped-Clamped B.C.
All the conditions are same as C-S situation except for the B.C. of where load applies. This time,
constrain x,y direction of displacement and 3 direction of rotations. But only accept displacement
in z-direction.
20
. Clamped-Free B.C.
This time delete the boundary condition at where load applies.
After that, we can get the outcomes as shown in the following
21
Figure 34: Geometric imperfection consideration
In C-S boundary condition, I apply the horizontal load very near the simply-supported side. Because
it is simply-supported, so I cant apply load in the exactly cross-section because the x-directions
displacement has been restrained. I apply load as shown in the following:
22
Figure 36: Abaqus result with consideration of Geometric imperfection for C-S
23
Figure 37: Geometric imperfection application method for C-C.
After apply the geometric imperfection, we get the results as shown in the following, we can see
it converges when increment=63, so the load factor is 0.5863 at this condition, means buckling
stress is 50ksi*0.5863=29.315ksi
24
Figure 38: Abaqus result with consideration of Geometric imperfection for C-C
25
After apply the geometric imperfection, we get the results as shown in the following, we can see
it converges when increment=12, so the load factor is 0.05962 at this condition, means buckling
stress is 50ksi*0.05962=2.981ksi
Figure 40: Abaqus result with consideration of Geometric imperfection for C-F
26
I summary results in following table.
C-C C-S C-F
Abaqus without consider 0.29038 0.28998 0.0607
geometric imperfection
Abaqus with consider 0.5863 0.3598 0.0596
geometric imperfection
Difference (%) +102% +24% -1.8%
From the table above, I can make a conclusion that the C-C B.C. has the most significant effects
with consider geometric imperfection.
27
Figure 41: C-section with round corner
For the simply-supported end, apply the pressure and constrain displacement in x and y direction
Then we get the mode 1 outcome in the following figure, because for buckling, we only consider
the minimum buckling mode. The color bar means resultant displacement of this member. And I
change the deformation scale factor to 3 just in order to see the deformation better. Eigenvalue
means the critical buckling force are eigenvalue*applied force in Abaqus. For mode shape 1:
28
eigenvalue=0.30919, buckling stress=50*0.30919=15.46ksi.
. Clamped-Clamped B.C.
All the conditions are same as C-S situation except for the B.C. of where load applies. This time,
constrain x,y direction of displacement and 3 direction of rotations. But only accept displacement
in z-direction.
29
Figure 45: Abaqus result for C-C boundary conditions (mode 1)
For mode shape 1: eigenvalue=0.30905, buckling stress=50*0.30905=14.453ksi.
. Clamped-Free B.C.
This time delete the boundary condition at where load applies.
After that, we can get the outcomes as shown in the following
30
The outcomes comparison:
Between CUFSM4 result and Abaqus result:
C-C C-S C-F
CUFSM4 0.30556 0.30151 0.059504
Abaqus 0.30905 0.30919 0.05077
Difference (%) +1.13% +2.48% -17.2%
Abaqus result between C-section with round corner and without round corner:
C-C C-S C-F
Without round corner 0.29038 0.28998 0.0607
With round corner 0.30905 0.30919 0.05077
Difference (%) +6.43% +6.624% -16.4%
From the table, I can make a conclusion that section with round corner can improve load factor a
little under C-S and C-C boundary conditions and reduce a little under C-F B.C. It is the same
outcome as analysis by the CUFSM4.
31
Figure 48: material property of member
32
Figure 49: mesh in Solidworks
. Clamped-clamped B.C.
Mode shape 1: Load factor is 0.51532, so the critical buckling stress is: 50ksi*0.51532=25.766ksi
33
Figure 51: Solidworks result for C-C B.C. (mode 1)
. Clamped-free B.C.
Mode shape 1: Load factor is 0.034889, so the critical buckling stress is: 50ksi*0.034889=1.7445ksi
I summary results between Solidworks results and Abaqus results with consideration of geometric
imperfection.
C-S C-C C-F
Solidworks result 0.28106 0.51532 0.0349
Abaqus with consider 0.3598 0.5863 0.0596
geometric imperfection
Difference (%) +21.9% +13.77% -70.8%
34
3.3.2 Simulate C-section without round corner
I design the members cross-section as shown in the following, and the length of member is 96
inches, all the parameters are the same as before only change the shape of cross-section:
35
Figure 54: mesh in Solidworks
36
. Clamped-clamped B.C.
Mode shape 1: Load factor is 0.7393, so the critical buckling stress is: 50ksi*0.7393=36.97ksi
. Clamped-free B.C.
Mode shape 1: Load factor is 0.059586, so the critical buckling stress is: 50ksi*0.059586=2.98ksi
37
I summary results between Solidworks results between with round corner and without round
corner:
C-S C-C C-F
Solidworks without round 0.28106 0.51532 0.0349
corner
Solidworks result with 0.46571 0.7393 0.0596
round corner
Difference (%) +65.7% +43.5% 70.8%
Base on the analysis result from solidworks, I can make a conclusion that the round corner will
increase the load capacity of member in each 3 B.C.s.
38
Reference
[1]. Moreyra ME, Pekz T, Finite element studies on lipped channel flexural members. In:
Proceedings of the twelfth international specialty conference on cold-formed steel structures.
University of Missouri-Rolla, St Louis, MO; 1994.p. 57-74
[2]. Y.B. Kwon, G.J. Hancock, Post-buckling analysis of thin-walled channel sections undergoing local
and distortional buckling
[3]. C. Yu, B.W. Schafer, Simulation of cold-formed steel beams in local and distortional buckling
with applications to the direct strength method
[4]. CUFSM4 Version 4.03 by Ben Schafer. Department of Civil Engineering, Johns Hopkins
University; 2006. http://www.ce.jhu.edu/bschafer/cufsm/
[5]. B.W. Schafer, T. Pekz, Computational modelling of cold-formed steel: characterizing geometric
imperfections and residual stresses
[6]. JOVICEVIC, J and OATE, E: Analysis of Beams and Shells Using a Rotation
[7]. C. Yu, B.W. Schafer, Finite Element Modeling of Cold-formed Steel Beams Validation and
Application.
[8] Alejandro Lifante Mira, Non-linear geometric elastic analysis of thin-walled beam by the finite
strip element method.
39