Engineering Structures 95 (2015) 127137

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Buckling behaviors of
compression

section aluminum alloy columns under axial

Mei Liu, Lulu Zhang, Peijun Wang , Yicun Chang

School of Civil Engineering, Shandong University, Jinan, Shandong Province 250061, China

a r t i c l e

i n f o

Article history:
Received 30 January 2015
Revised 27 March 2015
Accepted 28 March 2015
Available online 10 April 2015
Keywords:
Aluminum alloy column
Irregular shaped cross section
Buckling behavior
Thin-walled structures
Direct strength method

a b s t r a c t
Thin-walled columns with
section are used widely as columns in aluminum alloy framed structures,
offering high strength-to-weight ratios and convenience in connection with maintaining walls. In this
section were studied experimentally and numeripaper, thin-walled aluminum alloy columns with
cally to investigate the buckling behavior and to assess the accuracy of current design methods. A nite
element model (FEM) was developed and used to perform parametric studies after being veried by tests.
Effects of plate thickness on elastic buckling stress was studied using nite strip method (FSM) and to
nd the potential buckling failure mode at a given length. Tested ultimate strengths were compared with
those predicted by the current American, European and Chinese specications on aluminum alloy structures and the Direct Strength Method (DSM) on thin-walled structures. Following reliability analysis, the
design strength predicted by current design specications were found to be generally conservative,
whereas DSM offered more accurate results.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
For its high strength-to-weight ratio, better corrosion resistance
and exural manufacture procedure through extrusion, aluminum
alloy members are being widely used in structural applications [1].
Summers et al. [2] performed a series of uniaxial tension tests on
AA5083H116 and AA6061T651 after simulated re exposure
and developed empirical laws for residual yield strength. Fogle
et al. [3] quantied the response and failure of 5083-H116 and
6082-T6 aluminum plates under compression load while being subjected to a re. Rasmussen and Rondal [4] proposed a column curve
to predict the strengths of the extruded aluminum alloy column
failed at exural buckling. Based on the FEM parametric studies on
buckling behaviors of re exposed aluminum alloy columns,
Maljaars et al. [5] found that EN 1999-1-2 [6] did not give an accurate
prediction for exural buckling strength of re exposed aluminum
columns. A new design method was proposed for the re resistance
design of aluminum alloy columns design considering the stress
strain relationship of aluminum alloys at elevated temperatures.
Manganiello et al. [7] evaluated the inelastic exural behavior of aluminum alloy structures through numerical method and proposed a
method for the ultimate strength of the rotational capacity of a
cross-section in bending. Maljaars et al. [8] studied local buckling
of compressed aluminum alloy at elevated temperatures through
Corresponding author. Fax: +86 531 88392843.
E-mail address: pjwang@sdu.edu.cn (P. Wang).
http://dx.doi.org/10.1016/j.engstruct.2015.03.064
0141-0296/ 2015 Elsevier Ltd. All rights reserved.

tests. Adeoti et al. [9] presented a column curve for extruded members made of 6082-T6 aluminum alloy. Yuan et al. [10] investigated
the local buckling and postbuckling strengths of aluminum alloy Isection stub columns under axial compression. Their research
results showed that current design codes were conservative to predict ultimate strength of aluminum alloy columns. Su et al. [11,12]
carried out a series of stub-column tests on box sections and two series of experiments on aluminum alloy hollow section beams. The
deformation based continuous method gave more accurate prediction for the ultimate strength. Wang et al. [13] carried out tests on
the columns of 6082-T6 circular tubes. Zhu and Young [14] presented tests results of aluminum alloy circular hollow section columns with and without transverse welds and assessed the
accuracy of the design rules in the current specications. These
researches greatly advanced the mechanism of buckling behaviors
of extruded aluminum alloy columns.
Many countries have already published design codes for aluminum alloy structural members, such as EN1999-1-2 (EC9) [6],
American Aluminum Design Manual (AA) [15], Australian/New
Zealand Standard [16], and Chinese Design Specications for
Aluminum Structures (GB50429) [17]. To make full use of structural material, the cross-section of an aluminum alloy member is
usually made up of thin-walled plates. These design codes follow
the element approach to calculate the buckling strength of each
thin-walled element considering effects of local bucking. In design
of thin-walled steel structures, the effective section is usually
determined through the effective width method [18]. While the

128

M. Liu et al. / Engineering Structures 95 (2015) 127137

aluminum alloy members usually have complex shape cross sections, the effective width method appears tedious because it needs
iterations for the effective width dependent on stress distribution
across the section. At this kind of circumstance, the effective thickness method [19] is more feasible.
Schafer and Pekz [20] developed DSM for predicting the ultimate strength of thin-walled steel structural members. The DSM
had been adopted by AISI [21,22] now. The design equations of
DSM was proposed by curve tting the test data and FEA results
on open section thin-walled structural members such as channel,
lipped channel with web stiffeners, Z-section, hat section and rack
upright section. Unlike the traditional design method uses the
effective section, DSM uses whole section to calculate the ultimate
strength, which provides rational analysis procedure for irregular
shaped section and allows section optimization. Zhu and Young
[23,24] found that the modied DSM could be used in the design
of square hollow section (SHS) and rectangular hollow section
(RHS) aluminum alloy columns. Aluminum alloy extruded members usually have complex sections to include as many functions
as possible. However, the applicability of DSM in aluminum alloy
member with irregular shaped section has not been investigated
yet.
This paper presented experimental and numerical investigation
on the buckling behaviors of aluminum alloy columns with
shape cross-section under axial compression. The structural component with the studied cross section is usually used as columns
in an aluminum alloy framed structure, as shown in Fig. 1(a). The
Fiber Reinforced Plastic (FRP) wall can be easily xed in the channel of the section, as shown in Fig. 1(b).The FSM software CUFSM

Fig. 1. Application of aluminum alloy column with

Fig. 2. Stressstrain relationships.

[25] was used to illustrate effects of plate thickness on the buckling

strength and the potential buckling mode at a given length. The
FEA software ABAQUS [26] was used to obtain the ultimate
strength of the member considering effects of initial geometric
imperfection and the elasticplastic properties of aluminum alloy.
Current design codes were assessed through the comparison of FEA
results with predictions by AA [15], EC9 [6], GB50429 [17] and
DSM [21,22], as well as AISI by substituting the material properties
of steel with those of aluminum alloy.

section. (a) Layout of the aluminum alloy columns. (b) Connection of the aluminum alloy column with FRP.

Table 1
Material properties.
Specimens

Area A
(mm2)

Length L
(mm)

Elastic modulus
E (GPa)

0.1% proof stress

f 0:1 (MPa)

0.2% proof stress

f 0:2 (MPa)

Ultimate strength
f u (MPa)

Ultimate
strain eu (%)

Parameter
n

T01a
T01b
Mean value

113.46
110.25
111.86

399.6
400.8
400.2

67.07
70.06
68.57

187.2
200.5
193.9

193.1
207.6
200.4

233.2
235.4
234.3

9.7
7.9
8.8

22.0
19.9
20.9

129

M. Liu et al. / Engineering Structures 95 (2015) 127137

ln 2
 

ln

f 0:2
f 0:1

The ultimate tensile strain of the tested aluminum alloy was

only about 8.8%, which indicated that the aluminum alloy members might encounter ruptures where the tensile strain was relative high. The stressstrain relationship obtained from coupon
tests was shown in Fig. 2.
2.2. Column test

Fig. 3. Column section dimension.

2. Experimental studies
2.1. Material properties
Material properties of the specimens were determined by tensile coupon tests. The mean value of the elastic modulus and yield
strength of the aluminum alloy were 68.57 GPa and 200.4 MPa,
respectively. The stressstrain relationship of aluminum alloy
was described by RambergOsgood expression [27],

r
r
e 0:002
E

n

f 0:2

where E and f0.2 were elastic modulus and nominal yield strength
(stress at 0.2% plastic strain), respectively. The index n was used
to describe the shape of the inelastic portion of the stressstrain
diagram, as listed in Table 1. The value of n was calculated by

The test specimens were fabricated by extrusion using 6063-T5

aluminum alloy. Two column lengths including 350 mm and
190 mm were studied. The section dimension was shown in Fig. 3.
It had one symmetric axis and was made up of four cantilever and
one stiffened plates.
7 specimens were tested to investigate the inuence of column
length, direction of global buckling and initial geometric imperfection on buckling behaviors. Since the short tested specimens all
would fail at local buckling but not interaction of local and global
buckling without initial imperfection, two levels of initial lack of
straightness, 0.70 mm (l/500) and 5.00 mm (l/70), were exerted
through rolling machine before tests. The initial geometric imperfection was listed in Table 2.
A servo-controlled hydraulic testing machine was used to apply
compressive force by displacement control at a constant rate of
0.1 mm/ min. The ends of the specimens were milled at and
2 kN was applied before recording in order to ensure full contact
with the hydraulic machine bearing plates. Two LVDTs were used
to measure the axial displacement and the horizontal deection
of the specimens, as shown in Fig. 4. The applied load and LVDTs
readings were recorded at 5 second intervals during tests.

Table 2
Comparisons of buckling strengths obtained from tests and FEA.
Specimen

Length L (mm)

Direction of initial deection

Magnitude of initial deection (mm)

Test
PEXP

FEA
PFEA

Comparison
PEXP/PFEA

C01
C02
C03
C04
C05
C06
C07

350
350
350
350
350
350
190

Y
X
Y
X
Y+
X

0.7
0.7
5.0
5.0
5.0
0.7

60.48
68.78
42.56
43.50
39.54
65.28
66.02

60.24
64.32
46.24
47.46
43.53
64.32
61.91

1.00
1.07
0.92
0.92
0.90
1.01
1.07

Mean
COV

0.99
0.072

Note: X stands for axis of symmetry; Y stands for axis of non-symmetric axis; + and  stand for the positive and negative direction of the axis, respectively.

Fig. 4. Test arrangement.

130

M. Liu et al. / Engineering Structures 95 (2015) 127137

2.3. Test results

2.3.1. Failure modes
Deformations of the specimens after tests were shown in Fig. 5.
Since the section had only one symmetric axis, the specimens
failed at exuraltorsional buckling when it buckled around the
symmetric axis; and failed at exural buckling when it buckled
around the non-symmetric axis. For the 350 mm specimens, two
failure modes were observed during tests: (1) exural buckling,

as shown in Fig. 5(a), (c) and (e); and (2) exuraltorsional buckling, as shown in Fig. 5(b) and (d). The 190 mm specimens failed
at local buckling, as shown in Fig. 5(f) and (g).
The initial geometric imperfection would affect the failure
mode of the column. The specimens with initial deection toward
x axis (non-symmetric axis) failed at exuraltorsional buckling
around the symmetric axis, as shown in Fig. 5(b) and (d). While
the specimens with initial deection toward y axis (symmetric
axis) failed at exural buckling around the non-symmetric axis,

Fig. 5. Deformation of test specimens. (a) Deformation of specimen C01. (b) Deformation of specimen C02. (c) Deformation of specimen C03. (d) Deformation of specimen
C04. (e) Deformation of specimen C05. (f) Deformation of specimen C06. (g) Deformation of specimen C07.

M. Liu et al. / Engineering Structures 95 (2015) 127137

as shown in Fig. 5(a), (c) and (e). Because of the smaller ultimate
tensile strain of aluminum alloy, cracks were observed at the tension side of specimen C04 and C05, as shown in Fig. 5(d) and (e).
Specimen C07 failed at local buckling nally because of its short
length, as shown in Fig. 5(g).
2.3.2. Ultimate strengths
The magnitude of the initial imperfection had signicant effects
on the ultimate strength of the column, as listed in Table 2. The
0.7 mm initial lack of straightness was found appropriate to trigger
global buckling. While the initial lack of straightness with 5.00 mm
was too large, thus the buckling strength reduced a lot due to the
P-4 effect.
Length of specimen C03, C04 and C05 was 350 mm and the
magnitudes of initial imperfection was 5.0 mm. Measured ultimate
strengths were 42.56 kN, 43.50 kN and 39.54 kN, respectively. The
difference from the average value were very small, which were
1.8%, 4.0% and 5.8%, respectively. However, specimen C03 and
C05 failed at exural buckling and specimen C04 failed at exuraltorsional buckling, which showed that the failure mode had
no effects on the exural buckling strength.
The average ultimate strength of specimen C01, C02 and C06
with 0.7 mm initial deection was 64.58 kN, which was much
greater than the average value of specimen C03, C04 and C05 as
41.80 kN.

131

and column end. Hard contact was dened to prevent the penetration of the column into the rigid plates. The tangential behavior
was dened by friction formulation. The friction coefcient was
set as 0.47 [28]. In order to obtain a stable numerical solution, a
pseudo-dynamic was adopted with a default dissipated energy
fraction of 2  104 that are suitable for most applications. FEM
is shown in Fig. 8.

2.3.3. Loadaxial displacement curves

Loadaxial displacement curves obtained from tests were
shown in Fig. 6. The axial compression load increased linearly with
the increase in axial displacement at rst. The axial load kept
increasing till it reached the vertex of the loadaxial displacement
curve, where global buckling occurred, as shown in Fig. 6(ac).
3. Finite element model and verication
3.1. Finite element model
The FEA software ABAQUS [26] was used to predict the buckling
behavior and ultimate strength of the aluminum alloy column. S4R,
a 4-node reduced integration shell element, with 5 mm  5 mm
mesh size was used after mesh sensitivity analysis considering
accuracy, CPU time and memory. The loadaxial displacement
curves of the FEM with different mesh size were shown in Fig. 7.
Three curves almost coincided with each other indicated that mesh
size with 5 mm  5 mm was accurate enough to predict the behavior of the studied column. The general purpose shell element S4R
gave robust and accurate solutions and allowed transverse shear
deformations. The classical metal plasticity model was applied. It
used standard Mises yield surfaces with associated plastic ow.
Perfect plasticity and isotropic hardening denitions were both
available. It was simple and adequate for common applications
including crash analyses, metal forming, and general collapse studies. The normal stressstrain data obtained through tensile tests
was converted to true stress (Cauchy stress) and true strain (logarithmic strain). Poissons ratio of aluminum alloy was assumed to
be 0.3.
Meshed by C3D8R, 8-node linear reduced integrated structural
brick element, the rigid plates were used to apply axial load to
column through contact interaction between test machine and
column end. The mesh
size was selected to be
10 mm  10 mm  10 mm. The rigid plates were made of steel
with Youngs modulus of 2.05  105 MPa and behaved elastically.
One rigid plate was xed and the other could only move axially
toward the column to apply the axial compressive load. Contact
algorithm was dened between the surface of the rigid plates

Fig. 6. Loadaxial displacement curves obtained from tests and FEA. (a) Loadaxial
displacement of specimen C01, C02 and C06 with 0.7 mm initial imperfection. (b)
Loadaxial displacement of specimen C03, C04 and C05 with 5 mm initial
imperfection. (c) Loadaxial displacement of specimen C07 without initial
imperfection.

132

M. Liu et al. / Engineering Structures 95 (2015) 127137

Fig. 7. Mesh sensitivity analysis.

Fig. 8. FEM of the aluminum alloy column.

The initial imperfection of the column followed a half-sine

curve with the maximum deection at mid-span with magnitude
of 0.7 mm. The model was adjusted by modifying the coordinates
of nodes.

3.2. Effect of residual stress

The residual stress of extruded aluminum alloy members is
usually lower than 20 MPa [29]. The distribution of residual stress
across the studied section was shown in Fig. 9(a). The maximum
magnitude was 15.7 MPa. Stress distribution across the section
and deformation shape of the member with and without residual
stress were nearly the same, as shown in Fig. 9(b) and 9(c). The
axial loadaxial displacement curves coincided with each other
before reaching the ultimate strength, as shown in Fig. 9(d). The
ultimate strength was 59.06 kN with residual stress and 60.48 kN
without residual stress. The difference was only 2.40%, which
showed very little inuence from residual stress. Residual stress
was ignored in the following FEM parametric studies.

3.3. Model verication

The measured material properties in Table 1 were adopted in
FEM. The comparisons of ultimate loads obtained from tests and

Fig. 9. Effect of residual stress. (a) The distribution of initial residual stress. (b)
Stress distribution of the column without residual stress. (c) Stress distribution of
the column with residual stress. (d) Axial loadaxial displacement curve.

133

M. Liu et al. / Engineering Structures 95 (2015) 127137

FEA were listed in Table 2. The mean value of Test/FEA was 0.99.
The associated coefcient of variation was only 0.07.
There existed differences between the loadaxial displacement
curves obtained from FEA and tests, as shown in Fig. 6. However,
the ultimate loads and the corresponding axial displacements
when the columns buckled globally were almost the same.
Furthermore, the loadaxial displacement curves obtained from
FEA clearly showed the turning points on the ascending branch
which reected the reduction of the axial stiffness due to the local
buckling.
Deformation of specimen C01 and C07 at global buckling failure
obtained from tests and FEA were shown in Fig. 10. The local buckling and global buckling deformation agreed very well with each
other.

Fig. 10. Comparison of buckling modes obtained from tests and FEA. (a) Specimen
C01 failed at the exural buckling. (b) Specimen C07 failed at local buckling.

Based on the comparisons in ultimate strengths, loadaxial displacement curves and deformations obtained from FEA and tests, it
was concluded that the presented FEM could accurately predict the
behavior of aluminum alloy columns with
section subject to
axial compression. The validated FEM was used thereafter in the
parametric studies to assess current design codes for aluminum
alloy columns.

4. Section optimization
The
section was made up of 6 plates, as shown in Fig. 11.
FSM software CUFSM was used to obtain the elastic buckling stress
of the column with different plate thickness and column length.
Effects of plate thickness on elastic buckling stress were shown
in Fig. 12. It was seen that the thickness of plate t2, t3 and t6 had no
inuences on global
buckling stress, as
shown in
Fig. 12(b), (c) and (f). With the increase in t1 thickness, the column
global buckling stress increased. However, the increments were
very small, as shown in Fig. 12(a). Except for small thickness, the
increase in the thickness of t4 and t5, did not lead to the
corresponding increase in global buckling stress, as shown in
Fig. 12(d) and (e). That was, for the column section studied, it
was not an effective way to improve the global buckling stress
by increasing the plate thickness.
While for short columns less than 100 mm, t1 thickness greatly
affected the buckling stress and bucking modes, as shown in
Fig. 12(a). When t1 was 0.2 mm or 0.3 mm, it only occurred local
bucking. When t1 was 0.4 mm or 0.5 mm, it occurred both local
and distortional buckling; however, the local buckling dominated
the column failure. When t1 was 0.6 mm or 0.7 mm, it occurred
both local and distortional buckling and the distortional buckling
dominated the column failure. Under this circumstance, the buckling capacity of the column could not be improved by increasing t1
thickness. When t1 was 0.7 mm, the local buckling stress exceeded
aluminum alloy yield stress. That was, no local buckling occurred
and the whole plate was effective.
Plate t2 was a cantilever plate and was vulnerable to local buckling. When t2 was less than 0.9 mm, short columns with length
being less than 100 mm failed at local buckling, as shown in
Fig. 12(b). When t2 was greater than 1.0 mm, the column failed
at distortional buckling. The increase in t2 thickness did not
increase the local or distortional buckling stress of the column.
Plate t3 was a stiffened plate in the section. When t3 was
greater than 0.5 mm, the increase in t3 thickness did not lead to
the increase of the critical buckling stress accordingly, as shown
in Fig. 12(c).
Plate t4 was a cantilever plate and was vulnerable to distortional buckling. The critical elastic distortional buckling stress
was signicantly inuenced by t4 thickness. With the increase in
t4 thickness, the critical elastic distortional buckling stress
increased. And the column behavior was controlled by local

Fig. 11. Plates division.

134

M. Liu et al. / Engineering Structures 95 (2015) 127137

buckling when t4 was 1.5 mm. Furthermore, it was not efcient to

increase t4 thickness when it was greater than 1.5 mm since the
critical elastic distortional buckling stress had exceeded the yield
stress of aluminum alloy, as shown in Fig. 12(d).
Pate t5 was a stiffened plate and its local buckling stress kept
constant when t5 increased from 0.5 mm to 2.0 mm. However,
the distortional buckling stress increased along with the increase
in t5 thickness, as shown in Fig. 12(e). The elastic local buckling
stress was less than the elastic distortional buckling stress when
t5 was greater than 1.5 mm. That was, increase in t5 thickness
had no signicance under this circumstance.

The existence of plate t6 did not increase the buckling stress of

the column, as shown in Fig. 12(f). Plate t6 was not a structural part
in the section. It was to satisfy the leak proof requirement.

5. Current design codes

5.1. Current design codes for aluminum alloy column
EC9 [6], AA [15] and GB50429 [17] provide design rules for aluminum columns with traditional section shapes. AISI [21,22] for

Fig. 12. Effects of plate thickness on elastic buckling stress of the column. (a) Effects of t1 on elastic buckling stress (thickness of other plates are 1.5 mm). (b) Effects of t2 on
elastic buckling stress (t1 = 0.6 mm, thickness of other plates are 1.5 mm). (c) Effects of t3 on elastic buckling stress (t1 = 0.6 mm, thickness of other plates are 1.5 mm). (d)
Effects of t4 on elastic buckling stress (t1 = 0.6 mm, thickness of other plates are 1.5 mm). (e) Effects of t5 on elastic buckling stress (t1 = 0.6 mm, thickness of other plates are
1.5 mm). (f) Effects of t6 on elastic buckling stress (t1 = 0.6 mm, thickness of other plates are 1.5 mm).

M. Liu et al. / Engineering Structures 95 (2015) 127137

Table 3
Comparisons of ultimate strengths obtained from design codes and FEA.
Comparison

PFEA
PAA

P FEA
P AISI

PFEA
PEC9

PFEA
PGB

PFEA
PDSM

Mean
COV

1.21
0.103

1.18
0.083

1.47
0.101

1.62
0.145

1.13
0.107

cold-formed steel structures was also used to obtain the ultimate

strength of thin-walled aluminum alloy column by substituting
the material properties of steel with those of aluminum alloy.

135

The ultimate strength of the column under axial compressive

load is the minimum strength of global buckling, local buckling
and interaction of local and global buckling in AA [15]. Global
buckling strength is calculated based on Euler formula. Local buckling is the summation of the local buckling strength of each plate.
The interaction of local and global buckling is calculated when the
strength of local buckling is lower than that of global buckling. The
design strength is reduced when considering the effect of local
buckling in AA Specication.
EC9 and GB50429 adopt Perry curve for the design of axial compressive column. The stability factor of global buckling is

Fig. 13. Comparisons of results by FEA and design codes. (a) Comparison of ultimate strength obtained by AA and FEA. (b) Comparison of ultimate strength obtained by AISI
and FEA. (c) Comparison of ultimate strength obtained by EC9 and FEA. (d) Comparison of ultimate strength obtained by GB50429 and FEA. (e) Comparison of ultimate
strength obtained by DSM and FEA.

136

M. Liu et al. / Engineering Structures 95 (2015) 127137

calculated using PerryRoberson formula. Effective thickness

method is applied to consider the effect of local buckling. AISI considers the effect of local buckling using effective width method,
which is also used to predict the ultimate strength of the aluminum alloy column with
section in this paper.
DSM [21,22] assumes that the full section is effective when
calculating the ultimate strength of thin-walled structural members. Global buckling, local buckling, distortional buckling and
their interactions are all taken into consideration. The column
design rule of DSM for global buckling strength is

P ne

8

2
>
< 0:658kc P y


>
: 0:877
Py
k2
c

for kc 6 1:5
3
for kc > 1:5

where Pne is the global buckling strength of the column; Py is the

section yielding strength; kc is dimensionless slenderness ratio.

Fig. 14. Comparisons of ultimate strengths by FEA and design codes (t1 = 0.9 mm;
t2 = 2.0 mm; t3 = 1.0 mm; t4 = 2.0 mm; t5 = 1.0 mm; t6 = 1.0 mm).

5.2. Parameter studies

The veried FEM was used to obtain the ultimate strength of the
columns with various parameters. The obtained ultimate strengths
were compared with those calculated by design codes.
The 0.2% proof stress of the aluminum alloy obtained from tests
was adopted in FEM, which was 200.4 MPa. In order to eliminate
the effects of local buckling at column ends, the Multi-Points
Constraint (MPC) was used to simulate the xed boundary condition in FEM. The Beam type MPC was used ensuring a rigid beam
connection in order to constrain the displacement and rotation of
each slave nodes to those of the control point. The displacement
load was applied at the control point.
The parametric studies consisted of 615 specimens including 5
lengths and 123 sections with different plate thickness. Studied
column lengths were 600, 1200, 1800, 2400, 3000 mm, respectively. The section dimension remained unchanged while plate
thickness varied. The column lengths and plates thicknesses were
carefully selected to ensure global buckling occur.
Four groups of sections were studied and plate thicknesses in
each group were:
(1) group I: (33 specimens)
t1: 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, 1.2, 1.3, 1.4, 1.5 mm;
t2 and t3: 1.0, 1.5, 2.0 mm;
t4: 2.0 mm;
t5: 2.0 mm
t6: 1.0 mm.
(2) group II: (30 specimens)
t1: 0.8, 1.2, 1.5 mm;
t2 and t3:0.6, 0.8, 1.0, 1.2, 1.4, 1.5, 1.6, 1.8, 2.0, 2.2,
2.4 mm;
t4:2.0 mm;
t5:2.0 mm;
t6:1.0 mm.
(3) group III: (30 specimens)
t1:0.8 mm;
t2 and t3: 1.5 mm;
t4: 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.2, 2.4, 2.6, 2.8, 3.0 mm;
t5: 1.0, 2.0, 3.0 mm;
t6: 1.0 mm.
(4) group IV: (30 specimens)
t1: 0.8 mm;
t2 and t3: 1.0, 1.5, 2.0 mm;
t4: 2.0 mm;
t5: 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.2, 2.4, 2.6, 2.8, 3.0 mm;
t6: 1.0 mm.

Fig. 15. Comparisons of results by FEA and DSM.

5.3. Comparison of ultimate strengths

Table 3 gave the mean value and Coefcient of Variation (COV)
of the ratios of FEA results and design codes predictions.
Comparisons of ultimate strengths by FEA and design codes were
shown in Figs. 1316. Results of EC9 were very close to GB50429
since they both considered initial imperfections based on Perry
curve. Ratios of PFEA/PEC9 and PFEA/PGB were 1.47 and 1.62 with
the corresponding COV of 0.101 and 0.145, respectively. GB50429
considered a bigger imperfection, which leaded to the ultimate
strength being the most conservative among these design methods. AA predictions were also conservative with PFEA/PAA of 1.21
and COV of 0.103. Results by AISI were close to DSM and were relatively economic with PFEA/PAISI and PFEA/PDSM of 1.18 and 1.13 and
the corresponding COV of 0.083 and 0.107, respectively.
Fig. 13 and 14 compared FEA and design codes calculations for
section. From Fig. 13 and 14, it
aluminum alloy column with
was seen that current design codes were all conservative to predict
the ultimate strength. The column design curves were divided into
3 parts in AA. The curves of EC9 and GB50429 agreed well with
each other. And GB50429 were the most conservative prediction.
DSM exural buckling curve agreed well with that obtained
from FEA, as shown in Fig. 15. Fig. 16 showed the strength ratio distribution, and the majority of PFEA/PDSM was between 1.05 and 1.15.
Therefore, it was concluded that DSM exural buckling curve predicted the ultimate strength more accurately for aluminum alloy
column with
section under axial compressive load.

M. Liu et al. / Engineering Structures 95 (2015) 127137

137

References

Fig. 16. Strength ratio distribution.

6. Conclusions and suggestions

This paper presented experimental and numerical investigasection aluminum alloy columns
tions on the behaviors of
under axial compression. Seven axial compression tests were carried out. The ultimate strengths and failure modes were reported.
Columns with length of 190 mm failed at local buckling; and those
with length of 350 mm failed at global buckling. A FEM were developed and veried by tests in three aspects including ultimate loads,
loadaxial displacement curves and deformations.
Effects of plate thickness and column length on the elastic buckling stress were investigated by FSM. The thickness of plate t2, t3
and t6 had no signicant inuences on global buckling stress.
With the increase in t1 thickness, the column global buckling stress
increased slightly. Except for small thickness, the increase in t4 and
t5 thickness did not increase the global buckling stress.
Parametric studies on 615 specimens were presented using
FEM. Ultimate strengths obtained by FEA were compared with
those calculated by AA, AISI, EC9, GB50429, as well as DSM.
Results of EC9 were close to GB50429 because they both considered initial imperfections based on Perry curve. GB50429 considered imperfection greater than EC9, which leaded to the most
conservative predictions. AA was also a little conservative.
Results of AISI were close to DSM and they both agreed well with
FEA results. DSM exural buckling equation which considered the
gross area could predict the ultimate strength for
section column under axial compression more accurately and conveniently.
Acknowledgements
The supports given to this work by the Independent Innovation
Foundation of Shandong University (No. 2014HW013) are greatly
acknowledged.

[1] Mazzolani FM. Competing issues for aluminium alloys in structural

engineering. Progr Struct Eng Mater 2004;6(2):18596.
[2] Summers PT, Case SW, Lattimer BY. Residual mechanical properties of
aluminum alloys AA5083-H116 and AA6061-T651 after re. Eng Struct
2014;76(1):4961.
[3] Fogle EJ, Lattimer BY, Feih S, Kandare E, Mouritz AP, Case Scott W. Compression
load failure of aluminum plates due to re. Eng Struct 2012;34(1):15562.
[4] Rasmussen KJR, Rondal J. Strength curves for aluminum alloy columns. Eng
Struct 2000;22(11):150517.
[5] Maljaars J, Twilta L, Soetens F. Flexural buckling of re exposed aluminum
columns. Fire Saf J 2009;44(5):7117.
[6] EC9. Eurocode 9: design of aluminum structuresPart 11: General rules
general rules and rules for buildings. DD ENV 1999-1-1:2000. Final draft
October 2000. European Committee for Standardization; 2000.
[7] Manganiello M, De Matteis G, Landolfo R. Inelastic exural strength of
aluminium alloys structures. Eng Struct 2006;28(4):593608.
[8] Maljaars J, Soetens F, Snijder HH. Local buckling of aluminium structures
exposed to re, Part 1: tests. Thin-Walled Struct 2009;47(11):140417.
[9] Adeoti GO, Fan F, Wang YJ, Zhai XM. Stability of 6082-T6 aluminium alloy
columns with H-section and rectangular hollow sections. Thin-Walled Struct
2015;89(1):116.
[10] Yuan HX, Wang YQ, Chang T, Du XX, Bu YD, Shi YJ. Local buckling and
postbuckling strength of extruded aluminium alloy stub columns with slender
I-sections. Thin-Walled Struct 2015;90(1):1409.
[11] Su MN, Young B, Gardner L. Testing and design of aluminium alloy cross
sections in compression. J Struct Eng 2014;14(9):111.
[12] Su MN, Young B, Gardner L. Deformation-based design of aluminium alloy
beams. Eng Struct 2014;80(1):33949.
[13] Wang YJ, Fan F, Lin SB. Experimental investigation on the stability of
aluminum alloy 6082 circular tubes in axial compression. Thin-Walled
Struct 2015;89(1):5466.
[14] Zhu JH, Young B. Experimental investigation of aluminum alloy circular hollow
section columns. Eng Struct 2006;28(2):20715.
[15] AA. American aluminum design manual. Washington, DC: The Aluminum
Association; 2005.
[16] AS/NZS. Aluminum structures. Part 1: Limit state design. Australian/New
Zealand Standard AS/NZS 1664.1:1997. Sydney, Australia: Standards Australia;
1997.
[17] GB 50429-2007. Code for design of aluminum structures. Beijing: China
Architecture & Building Press; 2007 [In Chinese].
[18] Schafer BW. Review: the direct strength method of cold-formed steel member
design. J Constr Steel Res 2008;64(78):76678.
[19] Bulson PS. The treatment of thin-walled aluminum sections in Eurocode 9.
Thin-walled Struct 1997;29(4):312.
[20] Schafer BW, Pekz T. Direct strength prediction of cold-formed steel members
using numerical elastic buckling solutions. In: Proceeding of 14th international
specialty conference on cold-formed steel structures. Rolla, MO: University of
Missouri-Rolla; 1998. p. 6976.
[21] AISI. North American specication for the design of cold-formed steel
structural members. Washington, DC: American Iron and Steel Institute; 2001.
[22] Supplement to the North American Specication for the design of cold formed
steel structural member. Washington, DC: American Iron and Steel Institute;
2004.
[23] Zhu JH, Young B. Aluminum alloy tubular columnspart I: nite element
modeling and test verication. Thin-Walled Struct 2006;44(9):9618.
[24] Zhu JH, Young B. Aluminum alloy tubular columnspart II: parametric study
and design using direct strength method. Thin-Walled Struct
2006;44(9):96985.
[25] CUFSM: Elastic Buckling Analysis of Thin-Walled Members by Finite Strip
Analysis. CUFSM v4.05. <http://www.ce.jhu.edu/bschafer/cufsm/index.htm>;
2012.
[26] ABAQUS Analysis Users Manual, Version 6.10.1. ABAQUS Inc; 2009.
[27] Ramberg W, Osgood WR. Description of stressstrain curves by three
parameters. National Advisory Committee for Aeronautics; 1943. p. 902.
[28] Minshall H. CRC handbook of chemistry and physics. In: Lide DR, editor. 73rd
ed. Boca Raton, FL: CRC Press; 1992.
[29] Mazzolani FM. Aluminum alloy structures. 2nd ed. London: E & FN Spon; 1995.