Professional Documents
Culture Documents
An Investigation of Buckling Phenomenon in Steel Elements - P - 2018
An Investigation of Buckling Phenomenon in Steel Elements - P - 2018
net/publication/329267500
CITATION READS
1 9,592
2 authors, including:
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Abdullahi Maikanti Baru on 29 November 2018.
ABSTRACT
The stability of columns and structural stability of steel structures has turn out to be of principal interest with the debut
of steel material as a construction material and the advancement of higher strength steels. The most prominent failure
mode of steel column structures is Buckling. Buckling phenomenon can be defined as the structural instability that is
usually associated with high compressional loading which results in a failure mode of structural element that is long
and slender. Two types of buckling; bifurcation type and deformation-amplification type were studied considering
their failure modes. The numerical investigations of steel columns with different boundary conditions, and slenderness
ratio both with the effect of imperfections were carried out in finite element code ABAQUS. A linear perturbation
analysis was conducted for both the models and their eigenvalues were compared with the Euler buckling load. The
eigenvalues obtained were used as the buckling loads for the non-linear Riks analysis that includes the eigenmodes
and the imperfection scale factors with the Ngleom toggled on to account for the geometric and material imperfections.
It was learnt that the fixed-fixed boundary condition provides a stiffer response and high resistance to the imperfections
when compared with the other boundaries. The effect of the imperfection was more evident in the slenderness ratio
model, higher percentage difference with the critical buckling load was noticed and it was because the length of
columns is a function of the imperfection.
Keywords: Imperfections, Boundary conditions, Slenderness ratio, Deformation-amplification type buckling, Elastic buckling
Fig. 3.2a. Typical effective length factors K= Lcr/L used in 3.3 FE Model Verification
the model from Davison and Owens [11]
The Euler equation for elastic buckling was used
analytically to achieve theoretical buckling loads
Commercial FE code ABAQUS 6.16 was used to offered and it is stated as:
obtain the modelling of columns. In the 𝜋2 𝐸𝐼
ABAQUS library, ABAQUS [12], the 𝑁𝑐𝑟 = 𝐿2
(Equation 3.3)
(*BUCKLE) technique was used to perform the
Where 𝑁𝑐𝑟 is the first critical buckling load
FE tests. This undertakes the eigenvalue analysis
technique expressed in the earlier section. The conforming to the first eigenvalue and buckling
aim of the FE analysis was to predict the first mode. E, the modulus of elasticity (225 GPa), I,
critical buckling mode from the eigenvalue the moment of inertia of the cross-section about
analysis because it will be same as the Euler load. the weak axis, K is the theoretical effective length
The same cross-section as stated before was used factor, and L is the column length. FE buckling
for the boundary conditions and the slenderness analysis results was associated with analytical
ratio which had different column lengths, both solution acquired through the Euler equation for
situations had imperfections that is respect to
each member. This assessment was made to see
the accuracy of the finite element model. When a
shear-flexible element is used, the finite element
model tends to be less accurate.
Table 2 shows the result of the comparison
between the Euler and eigenvalues for the case of
the boundary conditions model but for the
slenderness ratio model since only one constant
boundary condition (Pinned-Pinned) was used it
was found that there would be similar
correlations with respect to the boundary. Strong
correlations are said to be achieved. Figure 3.3
shows the samples of the expected buckling
(a) (b)
modes for a slenderness ratio model (L6000)
where this represents the length of the column as
6000mm. The columns were rendered for the
sake of visualization although all of them are line
(2D).
3.4 Non-linear Finite Element Analysis
(Nonlinear Buckling Analysis)
Effective Equivalence
Normal Boundary (Top Assumed Boundary
Length factor between assumed
and Bottom) (Top and Bottom)
(K=Lcr/L) and normal
Pinned-
1 Pinned-Roller U1=U2=0, U1=U2=0 U1=U2=0, U1=0
Pinned
Fixed- U1=U2=0,
0.85 Fixed-Roller, U2 U1=U2=UR3=0, U1=0, U2= − 3.5 × 10−5 mm
Pinned U1=U2=UR3=0
Fixed-Free 2 - U1=U2=UR3=0 -
Encastre-XSYMM, U1=U2=UR3=0, U1=U2=U3=UR1=UR2=UR3=0,
Fixed-Fixed 0.7
U2 U1=U2=UR3=0 U1=UR2=UR3=0, U2=−5.12× 10−5 mm
Fixed-
1 Pinned-Roller U1=U2=UR3=0, U1=0 U1=U2=0, U1=0
Roller
4. Results and Discussion from the eigenvalue analysis. As observed, the
modes of different lengths were equal with other
Restraining connection of beam-column stiffness lengths. For example, in L1000, mode 2 kept
plays a main function on the stability behavior of changing and becomes a higher mode in the next
columns. The columns were loaded length. The mode 3 in L1000 becomes the mode
concentrically therefore the deflections happen at 6 in L2000, it can be argued that the mode 9 of
low loads in the applied load direction and means L3000 will be equal to the mode 6 of L2000. The
that transverse deflection doesn’t happen until mode 1 in L1000 becomes mode 2, mode 3, mode
bifurcation or buckling load is attained. 4, mode 5, and mode 6 in the respective lengths
L2000, L3000, L4000, L5000, and L6000.
The column remains straight till the critical load It is believed that if L7000 existed then its mode
and at the critical load, the bifurcation of 7 will be equal to the mode 6 of L6000. The mode
equilibrium happens that leads the column to 1 of L1500 becomes mode 2, mode 3, and mode
reach instability and buckle. 4 in the respective lengths of L3000, L4500, and
It can be said that in all the plots, the structure has L6000. This goes back to the sinusoidal
undergone a loading from its base state of imperfection pattern. It is normal in eigenvalue
equilibrium to a new form of equilibrium state. analysis to have some modes equal, some factors
Hereafter, it can continue with the path which may have played a role. In this case, the only
consume the minimum of energy, Farshad [22]. variable was the length, therefore it is the reason
for this discrepancy. This is witnessed in the
paper of Hald and Nielsen [23] which studied the
4.1 Eigenvalues buckling of thin-walled cylinder.
Table 2 shows the critical buckling load of the 4.2 Riks Analysis
columns (2600 mm) extracted from the
eigenvalue analysis. The critical buckling load
from Euler equation is with respect to the weak Figure 4.2a shows the load against out of plane
axis (x-x) (horizontal). deformation for fixed-fixed, fixed-pinned, and
It shows the lowest eigenvalues corresponds fixed-roller classified as stiffer boundary
closely with the Euler equation and hence are the conditions, and Figure 4.2b shows the load
buckling load of the column. against out of plane deformation for fixed-free
There are huge variations for the fixed-fixed and and pinned-pinned classified as weaker boundary
fixed-pinned when their actual boundary conditions. From the elastic range, the average
conditions were used. A decrease of about critical buckling load was 63.28 KN for the stiffer
32.34% and 48.9% exist when their eigenvalues boundary and was 21.179 KN for the weaker
are compared. boundary. Fixed-fixed was greater than the Euler
The problem that caused this difference was buckling load and its lowest eigenvalue, but for
because a vertical translation needs to be allowed, the rest of the boundary conditions, they were
hence the vertical restraint was neglected. To get lower than their Euler buckling load and their
more accurate result, a vertical counter lowest eigenvalues as expected. There was an
displacement (U2) was applied and it makes the average decrease of about 12.024% and 21.675%
eigenvalue to become negative. Hence, the when their Euler buckling load and their lowest
absolute value was considered. The counter eigenvalues are compared for the stiffer boundary
displacement values assumed for the fixed-fixed and weaker boundary respectively. The reason
and fixed-pinned to be −5.12× 10-5 mm and − 3.5 for the discrepancy of fixed-fixed may be argued
× 10-5 mm respectively. to be due to the boundary condition which is most
stiff and made the effect of the imperfection to be
Table 3 shows the critical buckling load of irrelevant. Overall, the discrepancy follows a
columns at different heights which was extracted pattern that shows the less stiff the boundary
condition, the more the effect of imperfection discrepancies could be due to the different
becomes visible (i.e. decreased buckling load). lengths which is with respect to the imperfections
but for the long columns the variation seems to be
The average bifurcation point for the stiffer constant which can be argued that it is due to the
boundary was reached at 90.148 KN and the cross-section and column slenderness. The
average final maximum load at 102.895 KN average bifurcation point was reached at 88.785
while for the weaker boundary the average KN and the average final maximum load at
bifurcation point was reached at 23.0127 KN and 102.283 KN for the short and intermediate
the average final maximum load at 27.876 KN. columns, while for the long columns the average
The greater the connection restraint, the more the bifurcation point was reached at 10.7802 KN and
load and in turn less deformation, and vice versa. the average final maximum load at 12.778 KN.
Low effective length factor increases column The response of the short columns seems to be a
buckling load claimed by Hailemichael and Liivo stiffer one and resulted in low deflections as
[24] and can be achieved by bracing of the expected since the material law controls them.
column. The response of the intermediate columns seems
to be less stiff and hence has more deflections
than the former since it is in the transitioning
The carrying capacity of an imperfect column is stage. The response of the long columns is the
smaller than the Euler load, regardless of how most less stiff and hence has higher deflections.
small the initial imperfection is, Fernandez [25]. This shows that the Euler equation governs
The boundary effect affects the imperfection, buckling of the long columns as low loads caused
material nonlinearity, intermediate, and long the buckling and with high deformations.
columns.
Table 2: Critical Buckling loads of the columns
In this work, L1000, L1500, and L2000 are with respect to their boundary conditions
classified as short columns, L2500, L3000, and
L3500 are classified as intermediate columns, From
and L4000, L4500, L5000, L5500, and L6000 are Eigenvalues Mode
Mode 2 Mode 3 Mode 4 Mode 5 Mode 6 Euler
(KN) 1
classified as long columns. The classification was Equation
based on their lowest eigenvalues. Pinned-
The short and intermediate length are sensitive to Pinned end
44.963 179.854 404.671 719.415 1124.09 1618.68 44.9636
imperfect shape that comprise overall buckling ~ Pinned-
while long lengths are the most sensitive modes. Roller
Usually, the failure shape of the short and Fixed -
intermediate beams is affected by the initially Pinned end
assumed geometric imperfection mode ~ Fixed- 68.718 203.116 404.669 673.398 1009.31 1412.39 62.2333
(magnitude and scale factors). With increase of roller and
slenderness ratio, load capacity decreases. U2
Fixed- Free
11.241 101.168 281.022 550.802 910.51 1360.14 11.2409
Figure 4.2c showed the load against out of plane end
deformation for the short and intermediate Fixed -
columns, and figure 4.2d showed the load against Fixed ~
out of plane deformation for the long columns. Encastre, 91.849 187.901 367.398 555.395 826.645 1106.52 91.7624
From the elastic range, the average critical U2 and
XSYMM
buckling load was 72.987 KN for the short and
Fixed -
intermediate column, and was 8.214 KN. There
Roller ~
was an average decrease of about 29.667% and 44.963 179.854 404.671 719.415 1124.09 1618.68 44.9636
Pinned-
36.48% when both their Euler buckling loads and Roller
their lowest eigenvalues were compared for the
short and intermediate columns, and long
columns respectively. The reason for these
Table 3: Critical Buckling loads of the columns
with respect to their lengths
Eigenvalues
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6
(KN)
L1000 303.923 1215.81 2735.58 4863.25 7598.83 10942.3
L1500 135.09 540.361 1215.81 2161.44 3377.26 4863.25
L2000 75.988 303.923 683.894 1215.81 1899.71 2735.58
L2500 48.632 194.53 437.692 778.12 1215.81 1750.77
L3000 33.773 135.09 303.923 540.361 844.314 1215.81
L3500 24.812 99.25 223.312 397 620.312 893.25
L4000 18.997 75.988 170.974 303.923 474.927 683.894
L4500 15.01 60.04 135.09 240.16 375.251 540.361
L5000 12.158 48.632 109.423 194.53 303.923 437.692
L5500 10.048 40.192 90.432 160.769 251.201 361.729
L6000 8.4431 33.773 75.988 135.09 211.078 303.923