ARTICLE IN PRESS

Thin-Walled Structures 46 (2008) 1437– 1449

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Numerical investigation and design of aluminum alloy circular hollow

section columns
Ji-Hua Zhu a, Ben Young b,
a
School of Civil Engineering, Shenzhen University, Shenzhen 518060, PR China
b
Department of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong

a r t i c l e in fo abstract

Article history: This paper presents a numerical investigation of aluminum alloy circular hollow section non-welded
Received 19 October 2007 and welded columns using finite element analysis. A non-linear finite element model was developed
Accepted 10 March 2008 and verified against fixed-ended column tests. The column specimens were extruded from heat-treated
Available online 5 May 2008
aluminum alloys of 6063-T5 and 6061-T6, and the ends of the columns were transversely welded to
Keywords: aluminum end plates for the welded columns. The non-welded columns without welding of end plates
Aluminum alloys were also investigated. The welded columns were modeled by dividing the column into different
Buckling portions along the column length, so that the heat-affected zone softening at both ends of the welded
Circular hollow section
columns was included in the simulation. The initial local geometric imperfections of the columns were
Column
measured in this study. Geometric and material non-linearities were incorporated in the finite element
Design
Finite element analysis
model. The verified finite element model was used for a parametric study of fixed-ended aluminum
Heat-affected zone alloy circular hollow section columns. A comparison of the column strengths predicted by the finite
Parametric study element analysis and the design strengths calculated using the current American, Australian/New
Transverse welds Zealand and European specifications for aluminum structures was presented. The column strengths
were also compared with the design strengths predicted by the direct strength method, which was
developed for cold-formed carbon steel members. Design rules were proposed for aluminum alloy
circular hollow section columns with transverse welds at the ends of the columns. Reliability analysis
was performed to evaluate the reliability of the design rules.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction One disadvantage in using aluminum as a structural material is

Aluminum members are being used increasingly in structural
applications. The current American Aluminum Design Manual [1],
Australian/New Zealand Standard [2] and European Code [3] for
aluminum structures provide design rules for compression
members. Schafer and Peköz [4] developed a new design method
called the direct strength method (DSM) for cold-formed steel
structures. The test data used in the development of column
design for DSM were based on concentrically loaded pin-ended
cold-formed steel column for certain cross sections and geometric
limits [5,6]. The DSM has been adopted by the North American
Specification [7,8] for cold-formed steel structures. Zhu and Young
[9,10] showed that the DSM with some modification can be used
in the design of aluminum alloy square hollow section (SHS) and
rectangular hollow section (RHS) columns. For the purpose of
obtaining accordant design rules of different cross-sections, the
DSM was used in this study for the design of aluminum alloy
circular hollow section (CHS) columns.
 Corresponding author. Tel.: +852 2859 2674; fax: +852 2559 5337.
E-mail address: young@hku.hk (B. Young).
that heat-treated aluminum alloys could suffer loss of strength in
a localized region when welding is involved, and this is known as
heat-affected zone (HAZ) softening. Previous research [11,12]
indicated that welds have significant effect on column strength.
The test program presented by Zhu and Young [13] showed that
transverse welds at the ends of the CHS columns reduce the
column strength for nearly 46%. In addition, it was also shown
that the design rules in the current American, Australian/New
Zealand and European specifications are generally quite conser-
vative for aluminum alloy welded columns of circular hollow
sections [13]. Hence, it is necessary to obtain accurate design rules
for aluminum alloy columns containing transverse welds.
Finite element analysis (FEA) has been widely used in
structural design. Compared with physical experiments, FEA is
relatively inexpensive and time efficient, especially when a
parametric study of cross-section geometry is involved. In
addition, FEA is more convenient for investigation involving
geometric imperfections of structural members, whereas this
could be difficult to investigate through physical tests. Although
FEA is a useful and powerful tool for structural analysis and
design, it is important to obtain an accurate and reliable finite
element model (FEM) prior to a parametric study of FEA to be

0263-8231/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2008.03.006
 ARTICLE IN PRESS

1438 J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449

Nomenclature PFEA15 ultimate load predicted by FEA using 15 mm heat-

affected zone extension for welded column
A gross cross-section area PFEA20 ultimate load predicted by FEA using 20 mm heat-
COV coefficient of variation affected zone extension for welded column
D overall diameter of CHS Pm mean value of tested-to-predicted load ratio
DL dead load Pne nominal axial strength for flexural buckling
E Young’s modulus Pnl nominal axial strength for local buckling
e axial shortening Pu column strength
FEA finite element analysis Py yield strength of the section (fy A)
FEM finite element model r radius of gyration of gross cross-section about the
Fm mean value of fabrication factor minor y-axis of buckling
fy material yield strength t thickness of section
kc coefficient in the AS/NZS Standard VF coefficient of variation of fabrication factor
L length of specimen VM coefficient of variation of material factor
LL live load VP coefficient of variation of tested-to-predicted load
le column effective length ratio
Mm mean value of material factor z longitudinal coordinates
P axial load a1 factor in the proposed design equation due to welding
PAA unfactored design strength for American Aluminum (1.3(D/t)0.19)
Design Manual a2 factor in the proposed design equation due to welding
PAS/NZS unfactored design strength for Australian/New Zeal- (0.6(x/L)0.12)
and Standard b reliability index
Pcre critical elastic buckling load in flexural buckling, p2EA/ ef elongation (tensile strain) at fracture
(le/r)2 lc non-dimensional
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi slenderness for flexural buckling
Pcrl critical elastic local column buckling load ð P y =Pcre Þ
PDSM column strength calculated using the direct strength ll non-dimensional slenderness for
ffi interaction of local
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
method and flexural buckling ð P ne =Pcrl Þ
PDSM-W welded column strength calculated using the pro- rc local buckling coefficient specified in the Eurocode 9
posed design rules rhaz heat-affected zone (HAZ) softening factor specified in
PEC9 unfactored design strength for Eurocode 9 the Eurocode 9
PExp experimental ultimate load of column f resistance factor
PFEA ultimate load predicted by FEA s0.2 static 0.2% proof stress
su static ultimate tensile strength

carried out. A non-linear FEM for aluminum columns of SHS and
RHS with and without transverse welds has been developed by
Zhu and Young [9]. In this study, a non-linear FEM for aluminum
columns of CHS was developed and verified against experimental
results.
The purpose of this paper is firstly to investigate the behavior
and design of aluminum CHS columns using non-linear FEA. The
verified FEM is used for a parametric study of cross-section
geometries. Secondly, the current DSM is used for the design of
aluminum non-welded and welded columns of CHS. Thirdly,
design rules for aluminum welded columns of CHS are proposed
based on the current DSM. The column strengths predicted by the
FEA were compared with the design strengths calculated using the
American Aluminum Design Manual (AA), Australian/New Zeal-
and Standard (AS/NZS) and European Code (EC9) for aluminum
structures, as well as the DSM and proposed design rules. Lastly,
reliability analysis was performed to assess the reliability of these
design rules.
ended CHS columns without the welding of end plates. In this
paper, the term ‘‘welded column’’ refers to a specimen with
transverse welds at the ends of the column, whereas the term
‘‘non-welded column’’ refers to a specimen without transverse
welds. The testing conditions of the non-welded and welded
columns are identical, other than the absence of welding in the
non-welded columns. The experimental program included four
test series with different cross-section geometry and type of
aluminum alloy, as shown in Table 1 using the symbols illustrated
in Fig. 1. The measured cross-section dimensions of each specimen
are detailed in Zhu and Young [13]. The specimens were tested
between fixed ends at various column lengths ranged from
300–3000 mm. The test rig and operation are also detailed in
Zhu and Young [13]. The experimental ultimate loads (PExp) and
failure modes observed at ultimate loads obtained from the
non-welded and welded column tests are shown in Tables 2–6.
The test specimens were labeled such that the type of aluminum
alloy, test series, welding condition and specimen length could be
easily identified, as shown in Tables 2–6. For example, the label

2. Summary of test program

Table 1
2.1. Column tests and material properties Nominal specimen dimension of test series

Test series Type of material Dimension, D  t (mm)

Experimental results of aluminum alloy circular hollow
sections compressed between fixed ends have been reported by N-C1 6063-T5 50  1.6
Zhu and Young [13]. The test specimens were fabricated by N-C2 6063-T5 50  3.0
H-C1 6061-T6 50  1.6
extrusion using 6063-T5 and 6061-T6 heat-treated aluminum H-C2 6061-T6 50  3.0
alloys. The test program included 21 fixed-ended CHS columns
with both ends welded to aluminum end plates, and eight fixed- Note: 1 in. ¼ 25.4 mm.
 ARTICLE IN PRESS

J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449 1439

‘‘H-C2-W-L1000’’ defines the following specimen: the first letter
indicates the type of material of the specimen, where ‘‘H’’ refers to
the high strength aluminum alloy 6061-T6; and ‘‘N’’ refers to the
normal strength aluminum alloy 6063-T5. The second part of the
label indicates the cross-section shape of the specimen, where
‘‘C2’’ refers to a circular hollow section with nominal cross-section
dimension of 50  3.0 mm. The cross-section dimensions for other
sections are shown in Table 1. The third part of the label indicates
the welding condition. If a specimen has transverse welds to the
aluminum end plates, then ‘‘W’’ indicates a welded column
specimen; if a specimen was tested without welding to the end
plates, then the letter ‘‘NW’’ indicates a non-welded column
specimen. The last part of the label ‘‘L1000’’ indicates the length of
the specimen, where the letter ‘‘L’’ refers to the column length and
the following digits are the nominal length of the specimen in
millimeters (1000 mm). The non-welded and welded material
properties for each series of specimens were determined by
longitudinal tensile coupon tests as detailed in Zhu and Young
[13]. The measured material properties obtained from the coupon
tests are summarized in Table 7.

2.2. Measured local and overall geometric imperfections

t
In this study, the initial local geometric imperfections were
measured on four CHS specimens in length of 300 mm. The
specimens were cut from those specimens belonged to the same
batch of specimens as the column tests. Hence, the measured local
geometric imperfections are considered nearly the same order of
imperfections as the column specimens. A Mitutoyo co-ordinate
Measuring Machine (CMM) with an accuracy of 0.001 mm was
used to measure the initial local geometric imperfections. The
measurements were taken at the longitudinal quarter lines A–A,
D B–B, C–C and D–D of each specimen as shown in Fig. 2. Readings
were taken at regular intervals of 2 mm along the specimen
Fig. 1. Definition of symbols. length. The measured local imperfection profiles for Series N-C1 is
shown in Fig. 2. The vertical axis is plotted against the normalized
location along specimen length and the horizontal axis is plotted
against the measured local geometric imperfections. The max-
imum measured local geometric imperfections were 6.3%, 2.0%,
Table 2 10.4% and 16.0% of the section thickness for Series N-C1, N-C2,
Comparison of test and FEA results for non-welded columns
H-C1 and H-C2, respectively.
Specimen Experimental FEA Comparison Initial overall geometric imperfections were measured on all
specimens prior to testing, except for the short specimens of
PExp (kN) Failure mode PFEA (kN) Failure mode PExp/PFEA 300 mm in length, as detailed in Zhu and Young [13]. The
maximum measured overall geometric imperfections at mid-
N-C1-NW-L300 48.5 Y 46.3 Y 1.05
N-C1-NW-L1000 45.9 F 43.6 F 1.05 length were 1/1732, 1/1432, 1/562 and 1/854 of the specimen
N-C2-NW-L300 102.4 Y 87.1 Y 1.18 length for Series N-C1, N-C2, H-C1 and H-C2, respectively.
N-C2-NW-L1000 86.1 Y 76.0 F 1.13
H-C1-NW-L300 75.9 Y 69.6 Y 1.09
H-C1-NW-L1000 71.7 Y 65.6 F 1.09 3. Finite element modeling
H-C2-NW-L300 129.6 Y 124.3 Y 1.04
H-C2-NW-L1000 119.6 F 117.1 F 1.02
An accurate and reliable non-linear FEM for aluminum non-
Mean 1.08 welded and welded columns of square and rectangular hollow
COV 0.048
sections has been developed using the finite element program
Note: 1 kip ¼ 4.45 kN; F, flexural buckling; Y, yielding. ABAQUS [14], as presented by Zhu and Young [9]. In this study, the
FEM was used for the simulation of aluminum alloy circular
hollow section columns tested by Zhu and Young [13]. The
development of the FEM is detailed in Zhu and Young [9]. In the

Table 3
Comparison of test and FEA results for welded columns of Series N-C1

Specimen Experimental FEA Comparison

PExp (kN) Failure mode PFEA20 (kN) PFEA15 (kN) Failure mode PExp/PFEA20 PExp/PFEA15

N-C1-W-L300 35.9 HAZ 29.0 32.4 HAZ 1.24 1.11

N-C1-W-L1000 30.3 HAZ 28.0 31.3 HAZ 1.08 0.97
N-C1-W-L1650 27.9 F 25.1 27.3 F 1.11 1.02
N-C1-W-L2350 21.4 F 19.6 21.0 F 1.09 1.02
N-C1-W-L3000 17.3 F 14.8 15.4 F 1.17 1.13

Mean 1.14 1.05

COV 0.057 0.063

Note: 1 kip ¼ 4.45 kN; F, flexural buckling; HAZ, failure in the HAZ.
 ARTICLE IN PRESS

1440 J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449

Table 4
Comparison of test and FEA results for welded columns of Series N-C2

Specimen Experimental FEA Comparison

PExp (kN) Failure mode PFEA20 (kN) PFEA15 (kN) Failure mode PExp/PFEA20 PExp/PFEA15

N-C2-W-L300 69.4 HAZ 71.1 77.2 HAZ 0.98 0.90

N-C2-W-L1000 65.1 HAZ 61.9 64.2 HAZ 1.05 1.01
N-C2-W-L1650 48.6 F 45.5 46.9 F 1.07 1.04
N-C2-W-L2350 35.3 F 34.8 37.1 F 1.02 0.95
N-C2-W-L3000 28.3 F 27.5 28.4 F 1.03 1.00

Mean 1.03 0.98

COV 0.034 0.056

Note: 1 kip ¼ 4.45 kN; F, flexural buckling; HAZ, failure in the HAZ.

Table 5
Comparison of test and FEA results for welded columns of Series H-C1

Specimen Experimental FEA Comparison

PExp (kN) Failure mode PFEA20 (kN) PFEA15 (kN) Failure mode PExp/PFEA20 PExp/PFEA15

H-C1-W-L300 47.2 HAZ 44.1 45.2 HAZ 1.07 1.04

H-C1-W-L1000 39.0 HAZ 34.0 38.6 HAZ 1.15 1.01
H-C1-W-L1650 36.8 F 29.3 32.3 F 1.25 1.14
H-C1-W-L2350 25.3 F 21.8 23.3 F 1.16 1.08
H-C1-W-L3000 17.0 F 15.7 16.3 F 1.08 1.04

Mean 1.14 1.06

COV 0.064 0.047

Note: 1 kip ¼ 4.45 kN; F, flexural buckling; HAZ, failure in the HAZ.

Table 6
Comparison of test and FEA results for welded columns of Series H-C2

Specimen Experimental FEA Comparison

PExp (kN) Failure mode PFEA20 (kN) PFEA15 (kN) Failure mode PExp/PFEA20 PExp/PFEA15

H-C2-W-L300 88.0 HAZ 79.6 90.7 HAZ 1.11 0.97

H-C2-W-L1000 84.5 HAZ 80.1 89.0 HAZ 1.05 0.95
H-C2-W-L1650 66.9 F 59.4 61.6 F 1.13 1.09
H-C2-W-L2350 44.6 F 44.6 47.2 F 1.00 0.95
H-C2-W-L3000 33.0 F 32.7 33.2 F 1.01 0.99

Mean 1.06 0.99

COV 0.053 0.058

Note: 1 kip ¼ 4.45 kN; F, flexural buckling; HAZ, failure in the HAZ.

Table 7 Outside Inside

Measured non-welded and welded material properties of tensile coupons 1.0
Line A-A
Specimen E (GPa) s0.2 (MPa) su (MPa) ef (%) Line B-B
0.8 Line C-C
N-C1-W 73.0 71.3 120.9 9.9 Line D-D
Location, z/L

N-C1-NW 66.7 194.6 214.4 10.0

N-C2-W 71.6 75.3 109.7 6.9 0.6
N-C2-NW 67.1 185.9 207.7 10.4 D
H-C1-W 72.6 92.5 148.3 10.7 A C
H-C1-NW 67.1 286.7 310.1 10.7 0.4 B
H-C2-W 71.7 94.3 161.2 10.9
H-C2-NW 70.2 278.9 284.3 11.7
0.2
Note: 1 ksi ¼ 6.89 MPa; NW, non-welded tensile coupon; W, welded tensile A C
coupon. B
0.0
-0.2 -0.1 0.0 0.1 0.2
FEM, the measured cross-section dimensions, material properties Imperfection (mm)
and initial geometric imperfections of the test specimens were Fig. 2. Measured initial local geometric imperfection profiles for Series N-C1
modeled. The fixed-ended boundary condition was modeled by specimen of 300 mm in length.
 ARTICLE IN PRESS

J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449 1441

restraining all the degrees of freedom of the nodes at both ends of 140
the column, except for the translational degree of freedom in the Test
120
axial direction at one end of the column. The nodes other than the

Axial load, P (kN)

two ends were free to translate and rotate in any directions. 100
The material non-linearity was included in the FEM by specifying
80
the true values of stresses and strains. The plasticity of the FEA
material was simulated by a mathematical model, known as the 60
incremental plasticity model, in which the true stresses and true
40
plastic strains were calculated in accordance with ABAQUS [14].
The geometric imperfections were included in the FEM by using 20
the Eigenvalue analyses. The displacement control loading
0
method was used in the FEA that was identical to the loading
0 1 2 3 4 5 6
method used in the column tests. The S4R general-purpose shell
Axial shortening, e (mm)
elements were used in the FEM. The size of the finite element
mesh of 5  5 mm (length by width) was used in the modeling of Fig. 3. Comparison of experimental and FEA axial load-shortening curves for
the non-welded and welded columns of CHS. The welded columns Specimen H-C2-NW-L1000.
were modeled by dividing the columns into different portions
along the column length. Therefore, the HAZ softening at both
ends of the columns were simulated. The welded columns were 25
separated into three parts, the HAZ regions at both ends of the
columns, and the main body of the columns that are not affected 20
by welding. In this study, different length of HAZ extension at both

Axial load, P (kN)

ends of the welded columns were considered, that are equal to
15
20 and 15 mm. The material properties obtained from the non-
welded and welded tensile coupon tests were used for the main FEA20 FEA15
body and the HAZ regions of the welded columns, respectively. 10 Test

4. Test verification
It is necessary to verify the FEM. The developed FEM was
verified against the experimental results. For the non-welded
columns, the ultimate loads and failure modes predicted by the
FEA are compared with the experimental results as shown in Table
2. It is shown that the ultimate loads (PFEA) obtained from the FEA
are generally in good agreement with the experimental ultimate
loads (PExp). The ultimate loads predicted by the FEA are slightly
lower than the experimental ultimate loads. The mean value of
the experimental-to-FEA ultimate load ratio is 1.08 with the
corresponding coefficient of variation (COV) of 0.048 for the non-
welded columns, as shown in Table 2.
For the welded columns, both the ultimate loads predicted by
the FEA using the HAZ extension of 20 mm (PFEA20) and 15 mm
(PFEA15) are compared with the experimental results as shown in
Tables 3–6 for Series N-C1, N-C2, H-C1 and H-C2, respectively.
It is shown that the PFEA15 are in better agreement with the
experimental ultimate loads compared with the PFEA20. The mean
values of the experimental-to-FEA ultimate load ratio (PExp/PFEA20)
are 1.14, 1.03, 1.14, and 1.06 with the corresponding COV of 0.057,
0.034, 0.064 and 0.053 for Series N-C1, N-C2, H-C1 and H-C2,
respectively. The mean values of the load ratio PExp/PFEA15 are
1.05, 0.98, 1.06 and 0.99 with the corresponding COV of 0.063,
0.056, 0.047 and 0.058 for Series N-C1, N-C2, H-C1 and H-C2,
respectively.
The failure modes at ultimate load obtained from the tests and
FEA for each specimen are also shown in Tables 2–6. The observed
failure modes included yielding (Y), flexural buckling (F), and
failure in the HAZ. The failure modes predicted by the FEA
are identical to those observed in the tests, except for the
specimens N-C2-NW-L1000 and H-C1-NW-L1000. Fig. 3 shows
the comparison of the load-shortening curves obtained from
the test and predicted by the FEA for the non-welded specimen
H-C2-NW-L1000. It is shown that the FEA curve follows the
experimental curve closely, except that the loads predicted by the
FEA are slightly lower than the experimental loads in the post-
ultimate range. Fig. 4 also shows the load-shortening curves for
5
0
0 1 2 3 4 5 6
Axial shortening, e (mm)
Fig. 4. Comparison of experimental and FEA axial load-shortening curves for
Specimen N-C1-W-L2350.
the welded specimen N-C1-W-L2350. The load-shortening curves
predicted by the FEA using the HAZ extension of 20 and 15 mm are
shown in Fig. 4. In addition, Fig. 5(a) shows the photograph of
specimen N-C1-W-L2350 after the ultimate load has reached. The
specimen failed in flexural buckling. Fig. 5(b) shows the deformed
shape of the specimen predicted by the FEA after the ultimate
load. The resemblance of Fig. 5(a) and (b) demonstrates the
reliability of the FEA predictions.
5. Parametric study
The FEM closely predicted the experimental ultimate loads and
failure modes of the tested aluminum circular hollow section
columns, as shown in Tables 2–6. Hence, the model was used for
an extensive parametric study. The parametric study included 80
specimens that consisted of 16 series, as shown in Table 8. Each series
contained five specimens with column lengths of 500, 1200, 2000,
2700 and 3500 mm. The specimens were labeled such that the type of
aluminum alloy, section thickness, welding condition and column
length could be identified, as shown in Tables 9–12. For example, the
label ‘‘T5-t0.4-NW-L500’’ defines the following specimen:
 The first letter indicates the type of material of the specimen,
where ‘‘T5’’ refers to the aluminum alloy 6063-T5, and ‘‘T6’’
refers to the aluminum alloy 6061-T6.
 The second part of the label ‘‘t0.4’’ indicates the section
thickness of the specimen, where the letter ‘‘t’’ refers to
the section thickness and the following digits are the thick-
ness of the CHS in millimeters (0.4 mm). Table 8 shows the
 ARTICLE IN PRESS

1442 J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449

cross-section dimensions of each series using the nomencla-

ture defined in Fig. 1.
 The following part of the label ‘‘NW’’ indicates the welding
condition of the specimen, where the letter ‘‘NW’’ refers to the
non-welded column, and the letter ‘‘W’’ refers to the welded
column.
 The last part of the label ‘‘L500’’ indicates the length of the
column, where the letter ‘‘L’’ refers to the column length and
the following digits are the nominal length of the specimen in
millimeters (500 mm).

The material properties of the specimens of 6063-T5 alloy

investigated in the parametric study are identical to the material
properties of Series N-C1 in the experimental program for the
non-welded and welded material, whereas the material properties
of the specimens of 6061-T6 alloy are identical to the material
properties of Series H-C1 in the experimental program for the
non-welded and welded material, as detailed in Zhu and Young
[13]. The local imperfection magnitude was 10% of the section
thickness, and the overall imperfection magnitude was 1/2000 of
the column length used in the parametric study. The size of the
finite element mesh was kept at 5  5 mm (length by width) for
the non-welded and welded columns. The welded columns were
modeled with 15 mm HAZ extension at both ends of the columns.
The column strengths (PFEA) obtained from the parametric study
are shown in Tables 9–12.

6. Design approaches

6.1. Current design rules for aluminum structures

The American Aluminum Design Manual (AA) [1], Australian/

Fig. 5. Comparison of experimental and FEA deformed shapes for Specimen N-C1-
W-L2350: (a) experimental and (b) FEA.
New Zealand Standard (AS/NZS) [2] and European Code (EC9) [3]
for aluminum structures provide design rules for aluminum
columns with and without transverse welds. The design rules in
the AA Specification for calculating the design strengths of non-
welded aluminum columns are based on the Euler column
strength. The inelastic column curve, based on the tangent
modulus, is well approximated by a straight line using buckling
constants [12]. The buckling constants were obtained from
Tables 3.3-3 and 3.3-4 of Part I-B of the AA Specification.
Local buckling strength can be calculated using an empirical

Table 8
Cross-section dimensions of the series for parametric study

Series Type of material Diameter, D (mm) Thickness, t (mm) Area, A (mm2)

T5-t0.4-NW 6063-T5 75.4 0.4 94.2

T5-t0.6-NW 6063-T5 75.6 0.6 141.4
T5-t1.0-NW 6063-T5 76.0 1.0 235.6
T5-t2.0-NW 6063-T5 77.0 2.0 471.2
T6-t0.4-NW 6061-T6 75.4 0.4 94.2
T6-t0.6-NW 6061-T6 75.6 0.6 141.4
T6-t1.0-NW 6061-T6 76.0 1.0 235.6
T6-t2.0-NW 6061-T6 77.0 2.0 471.2
T5-t0.4-W 6063-T5 75.4 0.4 94.2
T5-t0.6-W 6063-T5 75.6 0.6 141.4
T5-t1.0-W 6063-T5 76.0 1.0 235.6
T5-t2.0-W 6063-T5 77.0 2.0 471.2
T6-t0.4-W 6061-T6 75.4 0.4 94.2
T6-t0.6-W 6061-T6 75.6 0.6 141.4
T6-t1.0-W 6061-T6 76.0 1.0 235.6
T6-t2.0-W 6061-T6 77.0 2.0 471.2

Note: 1 in. ¼ 25.4 mm; NW, non-welded column series; W, welded column series.
 ARTICLE IN PRESS

J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449 1443

Table 9 Table 10
Comparison of FEA and design strengths for non-welded columns of aluminum Comparison of FEA and design strengths for non-welded columns of aluminum
alloy 6063-T5 alloy 6061-T6

Specimen FEA Comparison Specimen FEA Comparison

PFEA (kN) PFEA/PAA PFEA/PAS/NZS PFEA/PEC9 PFEA/PDSM PFEA (kN) PFEA/PAA PFEA/PAS/NZS PFEA/PEC9 PFEA/PDSM

T5-t0.4-NW-L500 16.5 1.10 1.10 1.27 0.98 T6-t0.4-NW-L500 23.3 1.11 1.11 1.42 1.06
T5-t0.4-NW-L1200 15.9 1.05 1.05 1.27 0.97 T6-t0.4-NW-L1200 22.5 1.07 1.07 1.43 1.08
T5-t0.4-NW-L2000 15.3 1.02 1.02 1.31 1.01 T6-t0.4-NW-L2000 21.6 1.03 1.03 1.50 1.16
T5-t0.4-NW-L2700 13.9 0.96 0.96 1.31 1.05 T6-t0.4-NW-L2700 18.6 0.94 0.94 1.47 1.15
T5-t0.4-NW-L3500 11.5 0.91 0.91 1.29 1.08 T6-t0.4-NW-L3500 13.0 0.91 0.91 1.30 1.06

Mean, Pm – 1.01 1.01 1.29 1.02 Mean, Pm – 1.01 1.01 1.42 1.10
COV, VP – 0.073 0.073 0.014 0.045 COV, VP – 0.084 0.084 0.052 0.044
Reliability index, b – 2.72 2.52 3.67 2.96 Reliability index, b – 2.66 2.46 3.88 3.31

T5-t0.6-NW-L500 26.3 1.07 1.07 1.16 0.97 T6-t0.6-NW-L500 37.9 1.09 1.09 1.31 1.01
T5-t0.6-NW-L1200 25.4 1.03 1.03 1.18 0.99 T6-t0.6-NW-L1200 36.7 1.05 1.05 1.34 1.03
T5-t0.6-NW-L2000 24.2 1.01 1.01 1.21 1.05 T6-t0.6-NW-L2000 34.8 1.03 1.03 1.40 1.11
T5-t0.6-NW-L2700 21.5 0.99 0.99 1.20 1.08 T6-t0.6-NW-L2700 29.0 0.98 0.98 1.38 1.15
T5-t0.6-NW-L3500 17.4 0.91 0.91 1.20 1.08 T6-t0.6-NW-L3500 19.8 0.92 0.92 1.24 1.07

Mean, Pm – 1.00 1.00 1.19 1.03 Mean, Pm – 1.01 1.01 1.33 1.07
COV, VP – 0.059 0.059 0.016 0.052 COV, VP – 0.064 0.064 0.049 0.053
Reliability index, b – 2.81 2.60 3.31 2.98 Reliability index, b – 2.82 2.61 3.63 3.14

T5-t1.0-NW-L500 45.4 1.02 1.11 1.02 1.00 T6-t1.0-NW-L500 66.2 1.03 1.10 1.15 1.00
T5-t1.0-NW-L1200 44.1 0.99 1.08 1.05 1.02 T6-t1.0-NW-L1200 64.3 1.00 1.07 1.19 1.04
T5-t1.0-NW-L2000 41.3 1.03 1.03 1.08 1.07 T6-t1.0-NW-L2000 59.9 1.06 1.06 1.25 1.15
T5-t1.0-NW-L2700 36.3 1.00 1.00 1.09 1.09 T6-t1.0-NW-L2700 49.2 0.99 0.99 1.26 1.17
T5-t1.0-NW-L3500 29.0 0.91 0.91 1.13 1.09 T6-t1.0-NW-L3500 34.8 0.97 0.97 1.24 1.13

Mean, Pm – 0.99 1.03 1.07 1.06 Mean, Pm – 1.01 1.04 1.22 1.10
COV, VP – 0.047 0.074 0.036 0.038 COV, VP – 0.034 0.051 0.039 0.067
Reliability index, b – 2.84 2.60 2.78 3.16 Reliability index, b – 3.00 2.80 3.31 3.12

T5-t2.0-NW-L500 92.1 1.00 1.13 1.02 1.02 T6-CHS2.0-NW-L500 134.7 1.00 1.12 1.02 1.01
T5-t2.0-NW-L1200 89.1 1.00 1.09 1.04 1.03 T6-CHS2.0-NW-L1200 130.5 1.01 1.08 1.06 1.06
T5-t2.0-NW-L2000 83.2 1.04 1.04 1.06 1.08 T6-CHS2.0-NW-L2000 121.2 1.07 1.07 1.14 1.16
T5-t2.0-NW-L2700 73.1 1.01 1.01 1.08 1.10 T6-CHS2.0-NW-L2700 99.1 1.00 1.00 1.19 1.17
T5-t2.0-NW-L3500 58.1 0.91 0.91 1.12 1.09 T6-CHS2.0-NW-L3500 66.7 0.93 0.93 1.15 1.09

Mean, Pm – 0.99 1.03 1.06 1.06 Mean, Pm – 1.00 1.04 1.11 1.10
COV, VP – 0.048 0.079 0.037 0.034 COV, VP – 0.050 0.072 0.064 0.062
Reliability index, b – 2.85 2.59 2.73 3.21 Reliability index, b – 2.87 2.66 2.75 3.17

formula which was first developed by Clark and Rolf [15]. 6.2. DSM for aluminum alloy non-welded columns
The design rules in the AS/NZS Standard for calculating the
design strengths of non-welded aluminum columns are generally The DSM has been proposed by Schafer and Peköz [4] for
identical to those in the AA Specification, except that the AS/NZS laterally braced flexural members undergoing local or distortional
Standard reduces the yield load of the column using a parameter buckling. Subsequently, the method has been developed for
kc which is not included in the AA Specification. The EC9 Code concentrically loaded pin-ended cold-formed steel columns
adopts the Perry curve for column design, and values of the undergoing local, distortional, or overall buckling [5,6], which
imperfection factors are listed in Table 5.6 of the Code. The allows for interaction of local and overall buckling as well as
effects of local buckling on column strength are considered interaction of distortional and overall buckling. Zhu and Young
by replacing the true section with an effective section. The [10] reported that the modified DSM can be used for the design of
effective cross-section is obtained by employing a local buck- aluminum alloy columns of SHS and RHS. The CHS is investigated
ling coefficient rc to reduce the thickness of the element in the in this study. As summarized in the North American Specification
section. (NAS) [7,8] for cold-formed steel structures, the column design
The strength of aluminum column with transverse welds rules of the DSM that considered the local and overall flexural
(welded column) depends on the location and number of welds buckling are shown in Eqs (1)–(3). The values of 0.15 and 0.4 are
[1]. For CHS columns with transverse welds at the ends only, the the coefficient and exponent of the direct strength equation,
design equations given by the AA and AS/NZS specifications are respectively, that were calibrated against test data of concentri-
identical to the design equations of non-welded columns. cally loaded pin-ended cold-formed steel columns for certain
However, the design strength of CHS welded columns is calculated cross sections and geometric limits:
using the welded mechanical properties and the buckling
PDSM ¼ minðP ne ; P nl Þ (1)
constants obtained from Table 3.3-3 of Part I-B of the AA
Specification regardless of temper before welding. The EC9 Code 8 
> l2c
uses a factor rhaz to consider the weakening effects of welding on > 0:658 P y
> for lc p1:5
< !
column strength, and rhaz is equal to 0.60 and 0.50 for the 6000 Pne ¼ 0:877 (2)
>
> Py for lc 41:5
Series alloys of T5 and T6 conditions, respectively, as shown in >
: l2c
Table 5.2 of the Code.
 ARTICLE IN PRESS

1444 J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449

Table 11 Table 12
Comparison of FEA and design strengths for welded columns of aluminum alloy Comparison of FEA and design strengths for welded columns of aluminum alloy
6063-T5 6061-T6

Specimen FEA Comparison Specimen FEA Comparison

PFEA PFEA/ PFEA/ PFEA/ PFEA/ PFEA/ PFEA PFEA/ PFEA/ PFEA/ PFEA/ PFEA/
(kN) PAA PAS/NZS PEC9 PDSM PDSM-W (kN) PAA PAS/NZS PEC9 PDSM PDSM-W

T5-t0.4-W-L500 7.9 1.27 1.32 1.11 0.47 1.06 T6-t0.4-W-L500 10.3 1.32 1.32 1.37 0.47 1.07
T5-t0.4-W-L1200 7.8 1.26 1.31 1.15 0.48 1.07 T6-t0.4-W-L1200 10.3 1.32 1.32 1.45 0.49 1.10
T5-t0.4-W-L2000 8.0 1.28 1.33 1.24 0.53 1.10 T6-t0.4-W-L2000 10.3 1.32 1.32 1.56 0.55 1.15
T5-t0.4-W-L2700 7.6 1.30 1.30 1.29 0.57 1.16 T6-t0.4-W-L2700 9.7 1.31 1.31 1.65 0.60 1.22
T5-t0.4-W-L3500 7.3 1.36 1.36 1.44 0.68 1.34 T6-t0.4-W-L3500 9.2 1.35 1.35 1.91 0.75 1.46

Mean, Pm – 1.29 1.32 1.24 0.54 1.15 Mean, Pm – 1.32 1.32 1.59 0.57 1.20
COV, VP – 0.028 0.018 0.104 0.161 0.101 COV, VP – 0.010 0.010 0.132 0.194 0.131
Reliability index, b – 4.10 4.02 2.86 0.21 2.98 Reliability index, b – 4.26 4.04 3.41 0.32 2.84

T5-t0.6-W-L500 13.2 1.34 1.47 1.07 0.49 1.03 T6-t0.6-W-L500 17.7 1.32 1.43 1.27 0.47 0.99
T5-t0.6-W-L1200 12.5 1.27 1.40 1.06 0.48 1.00 T6-t0.6-W-L1200 17.6 1.32 1.42 1.33 0.49 1.02
T5-t0.6-W-L2000 12.7 1.36 1.42 1.16 0.55 1.07 T6-t0.6-W-L2000 16.5 1.37 1.41 1.44 0.53 1.02
T5-t0.6-W-L2700 12.4 1.41 1.41 1.24 0.62 1.17 T6-t0.6-W-L2700 15.6 1.39 1.39 1.57 0.62 1.15
T5-t0.6-W-L3500 11.7 1.42 1.42 1.41 0.73 1.32 T6-t0.6-W-L3500 13.4 1.41 1.41 1.85 0.73 1.32

Mean, Pm – 1.36 1.42 1.19 0.57 1.12 Mean, Pm – 1.36 1.41 1.50 0.57 1.10
COV, VP – 0.046 0.020 0.120 0.180 0.115 COV, VP – 0.030 0.010 0.155 0.186 0.124
Reliability index, b – 4.19 4.34 2.55 0.34 2.75 Reliability index, b – 4.32 4.33 2.97 0.30 2.62

T5-t1.0-W-L500 25.2 1.50 1.68 1.04 0.56 1.06 T6-t1.0-W-L500 32.8 1.51 1.68 1.25 0.49 0.95
T5-t1.0-W-L1200 25.7 1.53 1.72 1.11 0.60 1.12 T6-t1.0-W-L1200 33.0 1.51 1.69 1.33 0.54 1.01
T5-t1.0-W-L2000 25.4 1.62 1.70 1.19 0.66 1.17 T6-t1.0-W-L2000 32.5 1.61 1.66 1.46 0.62 1.10
T5-t1.0-W-L2700 23.8 1.62 1.62 1.27 0.72 1.22 T6-t1.0-W-L2700 29.3 1.57 1.57 1.58 0.69 1.18
T5-t1.0-W-L3500 21.3 1.57 1.57 1.43 0.80 1.31 T6-t1.0-W-L3500 25.1 1.50 1.50 1.85 0.83 1.35

Mean, Pm – 1.57 1.66 1.21 0.66 1.18 Mean, Pm – 1.54 1.62 1.50 0.63 1.12
COV, VP – 0.034 0.037 0.124 0.143 0.081 COV, VP – 0.032 0.051 0.157 0.209 0.141
Reliability index, b – 4.90 4.89 2.58 0.86 3.27 Reliability index, b – 4.84 4.68 2.94 0.57 2.51

T5-t2.0-W-L500 63.2 1.88 2.11 1.16 0.70 1.17 T6-t2.0-W-L500 82.9 1.90 2.13 1.30 0.62 1.05
T5-t2.0-W-L1200 63.1 1.88 2.10 1.22 0.73 1.21 T6-t2.0-W-L1200 82.7 1.90 2.13 1.39 0.67 1.11
T5-t2.0-W-L2000 62.9 2.00 2.10 1.34 0.82 1.27 T6-t2.0-W-L2000 82.5 2.05 2.12 1.59 0.79 1.23
T5-t2.0-W-L2700 54.4 1.85 1.85 1.34 0.82 1.22 T6-t2.0-W-L2700 66.8 1.80 1.80 1.63 0.79 1.18
T5-t2.0-W-L3500 46.5 1.71 1.71 1.49 0.87 1.26 T6-t2.0-W-L3500 54.1 1.60 1.60 1.88 0.88 1.28

Mean, Pm – 1.86 1.97 1.31 0.79 1.23 Mean, Pm – 1.85 1.95 1.56 0.75 1.17
COV, VP – 0.056 0.093 0.096 0.089 0.031 COV, VP – 0.090 0.125 0.146 0.137 0.078
Reliability index, b – 5.41 4.91 3.13 1.67 3.85 Reliability index, b – 4.89 4.38 3.18 1.27 3.28

8
>
> Pne for ll p0:776
<"   # 
P nl ¼ Pcrl 0:4 P crl 0:4
>
> 1  0:15 P ne for ll 40:776
: Pne Pne
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where Py ¼ fy A; lc ¼ P y =P cre ; ll ¼ P ne =Pcrl ; A, gross cross-
section area; fy, material yield strength which is the static 0.2%
proof stress (s0.2) using the non-welded material properties in this
paper; Pcre, p2EA/(le/r)2, critical elastic buckling load in flexural
buckling for CHS columns; Pcrl, critical elastic local column
buckling load; E, Young’s modulus; le, column effective length;
and r, radius of gyration of gross cross-section.
The nominal axial strengths (PDSM) are calculated for the two
cases, as shown in Eqs. (2) and (3), respectively, where Pne refers to
the nominal axial strength for flexural buckling, and Pnl refers to
the nominal axial strength for local buckling as well as interaction
of local and overall buckling. The nominal axial strength, PDSM, is
the minimum of Pne and Pnl, as shown in Eq. (1). In calculating the
axial strengths, the critical elastic local buckling load (Pcrl) of the
cross section was obtained from a rational elastic finite strip
buckling analysis [16]. Fig. 6 shows the local buckling (cross-section
distortion) of CHS generated from the finite strip analysis. Fig. 7(a)
and (b) show the comparison of FEA and experimental results
against the direct strength curves plotted from Eqs. (2) and (3),
respectively, for the non-welded columns. The unfactored design
strengths (PDSM) calculated using the DSM are compared with the
numerical and test results of aluminum non-welded columns of
(3) CHS, as shown in Tables 9, 10 and 13.
6.3. Proposed design equation based on DSM for aluminum alloy
welded columns
Design equation was proposed based on the current DSM for
aluminum alloy CHS columns with transverse welds at both ends of
the columns (welded columns). The research reported by Zhu and
Young [17] indicated that the effects of transverse welds on aluminum
stub column strengths are varied with the overall diameter-to-
thickness ratio of the CHS. The HAZ softening factor proposed by Zhu
and Young [17] is used in this study. The unfactored design strengths
(PDSM-W) is obtained based on the column strengths (PDSM) calculated
using Eqs. (1)–(3) and multiplied by the factors a1 and a2 due to
welding, as shown in Eq. (4). The proposed design equation was
calibrated with the welded column strengths obtained from the
parametric study presented in this study, as well as the test results
reported by Zhu and Young [13]. The unfactored design strengths
(PDSM-W) are compared with the numerical and test results of
aluminum welded columns of CHS, as shown in Tables 11, 12 and 14:
PDSM-W ¼ a1 a2 P DSM (4)
0.19 0.12
where a1 ¼ 1.3(D/t) and a2 ¼ 0.6(30/L) ; PDSM-W, welded
column strength; PDSM, column strength calculated using
 ARTICLE IN PRESS

J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449 1445

Table 13
Comparison of test strengths with design strengths for non-welded columns

Specimen Experimental Comparison

PExp (kN) PExp/PAA PExp/PAS/NZS PExp/PEC9 PExp/PDSM

N-C1-NW-L300 48.5 1.04 1.16 1.07 1.05

N-C1-NW-L1000 45.9 1.08 1.12 1.11 1.11
N-C2-NW-L300 102.4 1.20 1.37 1.24 1.24
N-C2-NW-L1000 86.1 1.10 1.16 1.14 1.14
H-C1-NW-L300 75.9 1.11 1.25 1.13 1.13
H-C1-NW-L1000 71.7 1.16 1.17 1.20 1.22
H-C2-NW-L300 129.6 1.04 1.16 1.06 1.05
H-C2-NW-L1000 119.6 1.06 1.07 1.10 1.12

Mean, Pm – 1.10 1.18 1.13 1.13

COV, VP – 0.052 0.077 0.055 0.062
Reliability index, b – 3.31 3.26 3.25 3.38
Fig. 6. Local buckling (cross-section distortion) of CHS generated from finite strip
analysis.

Table 14
Comparison of test results and design rules for welded columns
1.5
Specimen Test Comparison
Eqn. (2) – DSM
Non-Welded FEA data
PExp PExp/ PExp/ PExp/ PExp/ PExp/
Non-Welded experimental data (kN) PAA PAS/NZS PEC9 PDSM PDSM-W
1
N-C1-W-L300 35.9 2.15 2.41 1.33 0.79 1.29
Pu / Py

N-C1-W-L1000 30.3 1.86 2.02 1.22 0.73 1.19

N-C1-W-L1650 27.9 1.87 1.86 1.32 0.80 1.23
N-C1-W-L2350 21.4 1.61 1.61 1.46 0.81 1.19
0.5 N-C1-W-L3000 17.3 1.43 1.43 1.69 0.93 1.33

Mean, Pm – 1.78 1.87 1.40 0.81 1.24

COV, VP – 0.156 0.203 0.129 0.089 0.049
Reliability index, b – 3.84 3.24 3.03 1.79 3.80
0
N-C2-W-L300 69.4 2.06 2.30 1.40 0.84 1.20
0 0.5 1 1.5 2 2.5 3 3.5 4
N-C2-W-L1000 65.1 1.99 2.16 1.44 0.87 1.24
λc N-C2-W-L1650 48.6 1.65 1.65 1.28 0.77 1.04
N-C2-W-L2350 35.3 1.35 1.35 1.33 0.75 0.97
N-C2-W-L3000 28.3 1.22 1.22 1.56 0.87 1.09
1.5
Mean, Pm – 1.65 1.74 1.40 0.82 1.11
Eqn. (3) – DSM
COV, VP – 0.226 0.276 0.077 0.065 0.101
Non-Welded FEA data Reliability index, b – 2.84 2.44 3.59 1.95 2.86
Non-Welded experimental data
H-C1-W-L300 47.2 2.18 2.45 1.43 0.71 1.15
1
H-C1-W-L1000 39.0 1.88 2.03 1.35 0.67 1.09
Pu / Pne

H-C1-W-L1650 36.8 1.97 1.97 1.68 0.83 1.27

H-C1-W-L2350 25.3 1.52 1.52 1.85 0.87 1.28
H-C1-W-L3000 17.0 1.17 1.17 1.91 0.94 1.35
0.5
Mean, Pm – 1.74 1.83 1.64 0.80 1.23
COV, VP – 0.228 0.270 0.151 0.141 0.084
Reliability index, b – 2.95 2.60 3.30 1.47 3.41

H-C2-W-L300 88.0 2.09 2.34 1.44 0.72 1.02

0 H-C2-W-L1000 84.5 2.10 2.24 1.56 0.79 1.13
0 0.5 1 1.5 2 H-C2-W-L1650 66.9 1.85 1.85 1.63 0.82 1.10
λl H-C2-W-L2350 44.6 1.42 1.42 1.77 0.84 1.08
H-C2-W-L3000 33.0 1.21 1.21 1.97 1.00 1.25
Fig. 7. Comparison of FEA and experimental data with direct strength method
Mean, Pm – 1.73 1.81 1.68 0.83 1.12
(PDSM) for non-welded columns: (a) flexural buckling and (b) interaction of local
COV, VP – 0.232 0.274 0.123 0.125 0.076
and flexural buckling.
Reliability index, b – 2.90 2.55 3.69 1.67 3.12

Eqs. (1)–(3); D, overall diameter of CHS; L, column length in mm,

but LX1200; and t ¼ thickness.
7. Comparison of numerical and experimental results with
design predictions
The nominal axial strengths (unfactored design strengths)
predicted by the AA Specification (PAA), AS/NZS Standard (PAS/NZS),
and European Code (PEC9) for aluminum structures, as well as the
current DSM (PDSM) and proposed design equation (PDSM-W) are
compared with the column strengths obtained from the para-
metric study (PFEA) and experimental program (PExp) [13], as
shown in Tables 9–14. The statistical parameters Pm and VP which
are the mean value and coefficient of variation (COV) of FEA and
experimental-to-predicted load ratios of each series of specimens
are shown in Tables 9–14. Reliability indices (b) of the design
rules for each series of specimens are also shown in Tables 9–14.
 ARTICLE IN PRESS

1446 J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449

The FEA results are also compared with the column design curves 120
obtained from the design rules, as shown in Figs. 8–23. The design PAA Non-welded FEA
100

Column strength, Pu (kN)

strengths are calculated using the material properties for each Flexural buckling
series of specimens, as shown in Table 7, and the 0.2% proof stress
80
(s0.2) was used as the corresponding yield stress. The design
strengths of non-welded column specimens are calculated using 60
the non-welded material properties. In calculating the design PAS/NZS
strengths of welded columns, PAA and PAS/NZS are calculated using 40 PDSM
the welded material properties, as specified in the AA and AS/NZS
specifications. Whereas PEC9 are calculated using the non-welded 20 PEC9
material properties as required by the EC9 Code. The design
strengths (PDSM-W) of the proposed design Eq. (4) are calculated 0
0 500 1000 1500 2000 2500 3000
using the non-welded material properties. The fixed-ended
Effective length, le (mm)

20 Fig. 11. Comparison of FEA and design column strengths for Series T5-t2.0-NW.
Non-welded FEA
Column strength, Pu (kN)

15 Flexural buckling 30
Non-welded FEA

Column strength, Pu (kN)

10 PAA PAS/NZS
20 Flexural buckling
PDSM
5
PEC9 PAA PAS/NZS
10
0 PDSM
0 500 1000 1500 2000 2500 3000
PEC9
Effective length, le (mm)
0
Fig. 8. Comparison of FEA and design column strengths for Series T5-t0.4-NW.
0 500 1000 1500 2000 2500 3000
Effective length, le (mm)

Fig. 12. Comparison of FEA and design column strengths for Series T6-t0.4-NW.
30
Non-welded FEA
50
Column strength, Pu (kN)

Flexural buckling Non-welded FEA

Column strength, Pu (kN)

20 40 Flexural buckling
PAA PAS/NZS
30
10 PDSM PAA
20 PAS/NZS
PEC9
PDSM
10
0 PEC9
0 500 1000 1500 2000 2500 3000
0
Effective length, le (mm)
0 500 1000 1500 2000 2500 3000
Fig. 9. Comparison of FEA and design column strengths for Series T5-t0.6-NW. Effective length, le (mm)

Fig. 13. Comparison of FEA and design column strengths for Series T6-t0.6-NW.

60 80
PAA Non-welded FEA Non-welded FEA
70
Column strength, Pu (kN)

50
Column strength, Pu (kN)

Flexural buckling 60 Flexural buckling

40
50
PAA
30 40
PAS/NZS
30 PAS/NZS PDSM
20
20 PEC9
10 PDSM
PEC9 10
0 0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000
Effective length, le (mm) Effective length, le (mm)

Fig. 10. Comparison of FEA and design column strengths for Series T5-t1.0-NW. Fig. 14. Comparison of FEA and design column strengths for Series T6-t1.0-NW.
 ARTICLE IN PRESS

J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449 1447

160 120
Non-welded FEA Welded FEA
140 PDSM
100

Column strength, Pu (kN)

Column strength, Pu (kN)

120 Flexural buckling

Flexural buckling
80
100 PAA
80 PAS/NZS 60
60 PDSM
40
40
PEC9
20 20
PAA PAS/NZS PEC9 PDSM-W
0 0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000
Effective length, le (mm) Effective length, le (mm)

Fig. 15. Comparison of FEA and design column strengths for Series T6-t2.0-NW. Fig. 19. Comparison of FEA and design column strengths for Series T5-t2.0-W.

20
Welded FEA 30
Column strength, Pu (kN)

PDSM Welded FEA

15 Flexural buckling

Column strength, Pu (kN)

Flexural buckling
20
PDSM
10

5 10

PAA P
AS/NZS PEC9 PDSM-W
0 PAA PAS/NZS PEC9 PDSM-W
0 500 1000 1500 2000 2500 3000 0
Effective length, le (mm) 0 500 1000 1500 2000 2500 3000
Effective length, le (mm)
Fig. 16. Comparison of FEA and design column strengths for Series T5-t0.4-W.
Fig. 20. Comparison of FEA and design column strengths for Series T6-t0.4-W.

30
Welded FEA
Column strength, Pu (kN)

Flexural buckling 50
PDSM Welded FEA
20
Column strength, Pu (kN)

40
PDSM Flexural buckling

30
10
20
PAA PAS/NZS PEC9 PDSM-W
0 10
0 500 1000 1500 2000 2500 3000
PAA PAS/NZS PEC9 PDSM-W
Effective length, le (mm) 0
0 500 1000 1500 2000 2500 3000
Fig. 17. Comparison of FEA and design column strengths for Series T5-t0.6-W. Effective length, le (mm)

60 Fig. 21. Comparison of FEA and design column strengths for Series T6-t0.6-W.
Welded FEA
PDSM
Column strength, Pu (kN)

50

40 Flexural buckling column specimens were designed as concentrically loaded

compression members, and the effective length (le) was taken as
30 one-half of the column length (L), as recommended by Young and
Rasmussen [18].
20 The reliability of the design rules for aluminum columns is
evaluated using reliability analysis. Reliability analysis is detailed
10
in the AA Specification [1], and the ratio of dead (DL) to live (LL)
PAA PAS/NZS PEC9 PDSM-W
0 load of 0.2 was used in the analysis. In general, a target reliability
0 500 1000 1500 2000 2500 3000 index of 2.5 for aluminum alloy columns as a lower limit is
Effective length, le (mm) recommended by the AA Specification [1]. If the reliability index is
greater than or equal to 2.5 (bX2.5), then the design is considered
Fig. 18. Comparison of FEA and design column strengths for Series T5-t1.0-W. to be reliable. The AA and AS/NZS specifications provide different
 ARTICLE IN PRESS

1448 J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449

80 long columns that are slightly unconservative, as shown in Tables

70 PDSM Welded FEA 9 and 10. The mean values of load ratios PFEA/PAA and PExp/PAA
Column strength, Pu (kN)

ranged from 0.99 to 1.10, with the corresponding COV ranged from
60 0.034 to 0.084, and the reliability index ranged from 2.66 to 3.31
50 Flexural buckling for all the non-welded column series. The mean values of load
ratios PFEA/PAS/NZS and PExp/PAS/NZS ranged from 1.00 to 1.18, with
40
the corresponding COV ranged from 0.051 to 0.084, and the
30 reliability index ranged from 2.46 to 3.26 for all the non-welded
20 column series. The design strengths calculated using the EC9 Code
are generally more conservative compared with the predictions
10
PAA PAS/NZS P given by the AA and AS/NZS specifications. The mean values of
EC9 PDSM-W
0 load ratios PFEA/PEC9 and PExp/PEC9 ranged from 1.06 to 1.42, with
0 500 1000 1500 2000 2500 3000 the corresponding COV ranged from 0.014 to 0.064, and the
Effective length, le (mm) reliability index ranged from 2.73 to 3.88 for all the non-welded
column series. The design strengths PDSM predicted by the current
Fig. 22. Comparison of FEA and design column strengths for Series T6-t1.0-W.
DSM are generally conservative and reliable for all the non-
welded column series, as shown in Tables 9, 10 and 13. The mean
values of the load ratio PFEA/PDSM for each series of FEA specimens
160 ranged from 1.02 to 1.10, with the corresponding COV ranged from
PDSM Welded FEA 0.034 to 0.067, and the reliability index ranged from 2.96 to 3.31.
140
In terms of the test results, the mean value of the load ratio PExp/
Column strength, Pu (kN)

120 Flexural buckling
100
80
60
40
20
PAA PAS/NZS PEC9 PDSM-W
0 calculated using the AA specification (PAA) are quite conservative,
0 500 1000 1500 2000 2500
Effective length, le (mm)
Fig. 23. Comparison of FEA and design column strengths for Series T6-t2.0-W.
resistance factors (f) for compression members with different
failure modes. The resistance factor varies with the slenderness
parameter for flexural buckling failure mode. The resistance factor
is a constant and equal to 0.85 for local buckling or interaction of
local and overall buckling failure mode. The observed failure
modes of the columns in this study included local buckling,
flexural buckling, interaction of local and overall buckling, and
failure in the HAZ. Hence, the resistance factor of the columns
given by the AA and AS/NZS specifications ranged from 0.76 to
0.95. In calculating the reliability indices of the AA and AS/NZS
design rules, the resistance factor for the columns is chosen to be
equal to 0.85 for all failure modes in this study. The EC9 Code
provides a constant resistance factor of 1/1.1 ¼ 0.91 for compres-
sion members, which is used in the reliability analysis. The
reliability of the DSM and proposed design rule for aluminum
columns is also evaluated, and the resistance factor is equal to
0.85. The load combination of 1.2DL+1.6LL is used in the analysis
for the AA Specification, the DSM and the proposed design rule.
The load combinations of 1.25DL+1.5LL and 1.35DL+1.5LL are used
in the analysis for AS/NZS and EC9 specifications, respectively. The
statistical parameters Mm, Fm, VM, and VF are the mean values and
coefficients of variation (COV) of material and fabrication factors.
These values are obtained from Section 9 of Part I-B of the AA
Specification [1], where Mm ¼ 1.10, Fm ¼ 1.00, VM ¼ 0.06, and
VF ¼ 0.05. The statistical parameters Pm and VP are the mean value
and the coefficient of variation of test-to-predicted load ratios,
respectively.
For the non-welded columns, it is shown that the column
strengths predicted by the AA and AS/NZS specifications are quite
close with the numerical and experimental results, except for the
PDSM is 1.13, with the corresponding COV of 0.062, and the
reliability index of 3.38, as shown in Table 13. It is shown that the
current DSM could be successfully used in the design of aluminum
non-welded columns of circular hollow sections. The non-welded
column design curves predicted by the AA, AS/NZS and EC9
specifications, as well as the DSM are shown in Figs. 8–15 for each
non-welded column series.
For the welded columns, it is shown that the design strengths
3000 as shown in Tables 11, 12 and 14. For the column series of FEA
specimens, the mean values of the load ratio PFEA/PAA ranged from
1.29 to 1.86, with the corresponding COV ranged from 0.010 to
0.090, and the reliability index ranged from 4.10 to 5.41. For the
column series of test specimens, the mean values of the load ratio
PExp/PAA ranged from 1.65 to 1.78, with the corresponding COV
ranged from 0.156 to 0.232, and the reliability index ranged from
2.84 to 3.84. The design strengths calculated using the AS/NZS
Standard are more conservative than the predictions given by the
AA Specification, as shown in Tables 11, 12 and 14. The design
strength calculated using the EC9 Code for welded columns are
also quite conservative, but generally less conservative than the
predictions by the AA and AS/NZS specifications. The mean values
of the load ratio PFEA/PEC9 ranged from 1.19 to 1.59 for each column
series of FEA specimens, and the mean values of the load ratio
PExp/PEC9 ranged from 1.40 to 1.68 for each column series of test
specimens. The corresponding COV for the load ratios PFEA/PEC9
and PExp/PEC9 ranged from 0.077 to 0.157, and the reliability index
ranged from 2.55 to 3.69. It is also shown that the current DSM is
not suitable for the design of aluminum CHS welded columns. The
mean values of the load ratio PFEA/PDSM ranged from 0.54 to 0.79,
with the corresponding COV ranged from 0.089 to 0.209, and the
reliability index ranged from 0.21 to 1.67 for each column series of
FEA specimens, as shown in Tables 11 and 12. The mean values of
the load ratio PExp/PDSM ranged from 0.80 to 0.83, with the
corresponding COV ranged from 0.065 to 0.141, and the reliability
index ranged from 1.47 to 1.95 for each column series of test
specimens, as shown in Table 14.
The design equation based on the DSM for aluminum alloy CHS
columns with transverse welds at both ends of the columns (welded
columns) is shown in Eq. (4) of the paper. The unfactored design
strengths (PDSM-W) calculated using the proposed design rules
are generally conservative for the welded columns, as shown in
Tables 11, 12 and 14. The mean values of the load ratios PFEA/PDSM-W
and PExp/PDSM-W ranged from 1.10 to 1.24, with the corresponding
COV ranged from 0.031 to 0.141. The reliability indices for the design
 ARTICLE IN PRESS
J.-H. Zhu, B. Young / Thin-Walled Structures 46 (2008) 1437–1449
rules (PDSM-W) are greater than the target value of 2.5 that ranged
from 2.51 to 3.85 for each series of specimens. The welded column
design curves predicted by the AA, AS/NZS and EC9 specifications, as
well as the current DSM and the proposed design rules are shown in
Figs. 16–23 for each welded column series. It is shown that the
proposed design equation based on the DSM can be used for the
design of aluminum columns of circular hollow sections with
transverse welds at the ends of the columns.
8. Conclusions
Numerical investigation and design of aluminum alloy circular
hollow section columns have been presented in this paper. A non-
linear FEM incorporating geometric imperfections and material
non-linearities was developed. The FEM was verified against
experimental results. The column specimens were fabricated by
extrusion using heat-treated aluminum alloy of 6061-T6 and
6063-T5. The ultimate loads and failure modes predicted by the
FEM are in good agreement with the experimental results. A
parametric study was performed using the verified FEM that
included 80 specimens with the column lengths ranged from 500
to 3500 mm. The column strengths obtained from the experi-
mental and numerical investigations were compared with the
design strengths calculated using the current American, Austra-
lian/New Zealand and European specifications for aluminum
structures, as well as the current DSM that was developed for
cold-formed steel members. It is shown that the design strengths
predicted by the current DSM are generally conservative for the
aluminum non-welded columns of circular hollow sections.
Design equation was proposed based on the current DSM for
aluminum alloy circular hollow section columns with transverse
welds at the ends of the columns. It is also shown that the design
strengths calculated using the proposed design rules are generally
conservative for the aluminum welded columns. The reliability of
the current and proposed design rules was evaluated using
reliability analysis. It is shown that the proposed design rules
are reliable. It is recommended of using the proposed design rules
in the design of aluminum columns of circular hollow section
containing transverse welds at the ends of the columns.
1449
Acknowledgement
The research work described in this paper was supported by a
grant from The University of Hong Kong under the Seed Funding
Programme for Basic Research.
References
[1] AA. Aluminum design manual. Washington, DC: The Aluminum Association;
2005.
[2] AS/NZS. Aluminum structures. Part 1: Limit state design. Australian/New
Zealand Standard AS/NZS 1664.1:1997. Sydney, Australia: Standards Australia;
1997.
[3] EC9. Eurocode 9: Design of aluminum structures—Part 1-1: General
rules—general rules and rules for buildings. DD ENV 1999-1-1:2000. Final
draft October 2000. European Committee for Standardization, 2000.
[4] Schafer BW, Peköz T. Direct strength prediction of cold-formed steel members
using numerical elastic buckling solutions. In: Proceeding of 14th interna-
tional specialty conference on cold-formed steel structures. University of
Missouri-Rolla, Rolla, MO, 1998. p. 69–76.
[5] Schafer BW. Distortional buckling of cold-formed steel columns. August final
report to the American Iron and Steel Institute, Washington, DC, 2000.
[6] Schafer BW. Local, distortional, and Euler buckling of thin-walled columns.
J Struct Eng 2002;128(3):289–99.
[7] North American Specification for the design of cold-formed steel structural
members. American Iron and Steel Institute, Washington, DC, 2001.
[8] Supplement to the North American Specification for the design of cold-
formed steel structural member. American Iron and Steel Institute, Washing-
ton, DC, 2004.
[9] Zhu JH, Young B. Aluminum alloy tubular columns—part I: finite element
modeling and test verification. Thin-Walled Struct 2006;44(9):961–8.
[10] Zhu JH, Young B. Aluminum alloy tubular columns—part II: parametric study
and design using direct strength method. Thin-Walled Struct 2006;44(9):
969–85.
[11] Mazzolani FM. Aluminum alloy structures. 2nd ed. London: E&FN Spon;
1995.
[12] Sharp ML. Behaviour and design of aluminum structures. New York: McGraw-
Hill; 1993.
[13] Zhu JH, Young B. Experimental investigation of aluminum alloy circular
hollow section columns. Eng Struct 2006;28(2):207–15.
[14] ABAQUS analysis user’s manual, version 6.5. ABAQUS Inc., 2004.
[15] Clark JW, Rolf RL. Design of aluminum tubular columns. J Struct Div ASCE
1964;90:259.
[16] Papangelis JP, Hancock GJ. Computer analysis of thin-walled structural
members. Comput Struct 1995;56(1):157–76.
[17] Zhu JH, Young B. Effects of transverse welds on aluminum alloy columns.
Thin-Walled Struct 2007;45(3):321–9.
[18] Young B, Rasmussen KJR. Tests of fixed-ended plain channel columns. J Struct
Eng 1998;124(2):131–9.