Proceedings of the

Annual Stability Conference

Structural Stability Research Council
Charlotte, North Carolina, April 11-14, 2023

Local buckling and design of aluminum I-shapes

Tristan Coderre1, Liya Li2, Sahar Dahboul3, Nicolas Boissonnade4

Abstract
Current standards use simplified approaches to predict the resistance of aluminum elements that
are not optimized to account for the effects of strain hardening, instabilities, and heat reduced
properties. This paper summarizes investigations towards the development of an alternative
design method for aluminum open cross-sections, based on the Overall Interaction Concept
(O.I.C.). This innovative design approach relies on the interaction between resistance and
stability, and also allows to consider geometrical and material imperfections. Moreover, it allows
to obtain direct, precise and consistent resistance predictions using continuous buckling curves.
A numerical finite element model was developed to accurately predict the cross-sectional
resistance of aluminum elements. Its efficiency was validated by comparing its numerical
resistance predictions to available experimental test data. Extensive parametric studies were then
conducted, allowing to study the impact of varying geometries, alloys, and load cases on the
resistance. Using the results from more than 2 300+ numerical simulations, O.I.C.-type design
proposals were formulated for the local resistance of extruded and welded aluminum sections.
The performance of the design proposals was evaluated by comparing their resistance estimates
to the reference numerical results and to resistance predictions from the Canadian, European, and
American aluminum design standards. The comparisons showed that the O.I.C. design proposal
leads to much more accurate and consistent results than these standards, while remaining simpler
and more efficient.

1. Introduction
The present paper addresses the behaviour, structural response and resistance of aluminium I-
shapes – either extruded or welded – under simple or combined load cases. Albeit marginally
used for heavy Structural Engineering applications, aluminium structural elements are gaining
interest owing to (i) an outstanding strength-to-weight ratio, (ii) its resistance to corrosion, in
particular in severe humidity and industrial conditions, considerably reducing maintenance costs,
(iii) great recycling abilities, (iv) an original manufacturing process (extrusion) complementary

1
MSc student, Laval University, <[Link].1@[Link]>
2
Post-Doc student, Laval University, <[Link].1@[Link]>
3
PhD student, Laval University, <[Link].1@[Link]>
4
Professor, Laval University, <[Link]@[Link]>
to more “traditional” ones (welding) and (v) stable resistance and ductility properties even at
very low temperatures, which shows particularly suitable in northern climates. All these qualities
leads to an increased use of aluminium members in construction projects, which causes an
improving need for suitable and economic modern design rules.
Nevertheless, most of the current design rules relative to aluminium structures (Aluminum
Association 2020; Canadian Standards Association 2017; European Committee for
Standardization 2007), usually based on steel structures’ codes, suffer a series of problems, from
a general lack of knowledge to inaccuracies. Among them, the underlying assumption of a plastic
plateau in the material response can be shown the source of many over-conservative resistance
predictions. Too, oversimplified, linear design equations appear poorly suited to aluminium
extruded shapes, which usually exhibit very complex geometries (Extrudex Aluminium Inc.
2020; Metra Aluminium Inc. 2019; Sapa Inc. 2008). Aluminium remaining a costly material, it is
therefore necessary to improve current design approaches in order to guarantee optimized
designs, to accurately consider the non-linear material response, to increase resistance
predictions accuracy, and to simplify the design process.
In this respect, this paper summarizes research efforts undertaken towards an improved, more
adequate design approach for aluminium structural sections, specifically for I-shaped sections.
The proposed approach is based on the Overall Interaction Concept (O.I.C., (Beyer 2017;
Boissonnade et al. 2017; Gagné et al. 2020; Gérard et al. 2021; Li et al. 2022)); the latter,
initially developed for steel members, is extended to aluminium shapes and accounts for (i) a
realistic stress-strain relationship, (ii) I-shapes form either extrusion or welding processes and
(iii) compact as well as slender geometries. The accuracy and performance of the proposal is
tested against reference Finite Element (F.E.) results. The non-linear shell models on which the
numerical studies rely are detailed in Section 2; in particular, the F.E. models are validated
against experimental tests on I-shaped specimens performed by Yuan (Yuan et al. 2015) in
Section 2.2. Then, a series of numerical computations varying the shapes, alloys and load
combinations are performed (Section 2.3). Next, Section 3 details the proposed O.I.C.-based
approach for both extruded and welded shapes. It further assesses the merits of this approach
against the reference numerical results and against existing design provisions from Eurocode 9
(European Committee for Standardization 2007), from the Aluminium Design Manual (A.D.M.,
Aluminum Association 2020), and from the Canadian Standard S157 (Canadian Standards
Association 2017).

2. F.E. models: development, validation and parametric studies

2.1 Basic features
Non-linear shell F.E. models were developed within software ABAQUS (Abaqus 2011). Typical
S4R quadrangular 4 nodes shell elements relying on Kichhoff’s bending assumptions combined
with reduced integration were used and placed at mid-thickness of each plate element within the
I-sections. Both Linear Buckling Analyses (L.B.A.), M.N.A. (Materially Non-linear Analyses)
and G.M.N.I.A. (Geometrically and Materially Non-linear with Imperfection Analyses)
computations were performed, providing accurate information on respectively buckling, yielding
and ultimate responses of the aluminium sections. State-of-the-art numerical techniques have
been employed, such as the subspace iteration method for L.B.A. (Bathe 2016), the Riks Method
for G.M.N.I.A. (Riks 1979), automatic loading strategies (Abaqus 2011), etc.

2
The section modelling of extruded sections was the subject of specific attention with respect to
web-flange areas, as extra material in the flange fillets can be shown to provide a non negligible
influence on resistance (Gérard et al. 2019; Gérard et al. 2021; Guo 1989; Li et al. 2022). In
these areas, two “problems” arise: (i) an overlap of material and (ii) the fact that the fillets’ areas
are not included. Accordingly, extra beam elements were introduced and positioned at the
centroid of the radius zone (see Figure 1). The area of these elements is set equal to the radius
area minus the overlapping area, and their section shape is chosen to be a square hollow section
so as to match all geometrical properties, and in particular closely represent torsional rigidities.
Besides, since an accurate representation of local buckling is important in the present study, a
truss system was added to the F.E. models to characterize the actual behavior of the web-flange
zone and its impact on the section’s response to buckling. The truss system, modelled through
relatively stiff spring elements, allows to maintain this area undeformed so that local buckling in
the adjacent web/flange plates initiates at the foot of the radius, similarly to the real section.
For the welded shapes considered herein, no such modelling refinements were adopted since
welds do not provide a meaningful effect on resistance. Yet, specific material properties in the
Heat-Affected Zone (H.A.Z.) were considered (Coderre 2022).
Flange nodes Overlapping area Spring elements

Centroid of Area of fillets radius

radius zone not included Additional
beam elements

Web nodes

a) b)
Figure 1: Modelling of web-flange area for extruded sections − a) Real geometry − b) Modelled geometry.

Obviously, mesh density studies were performed prior to any use of the numerical models, the
goal being to find the best compromise between a sufficient accuracy and a reasonable
computation time (details can be found in Coderre 2022). Figure 2 shows various mesh densities
used beforehand when searching for the most suitable mesh density. Results were typically
G.M.N.I.A. computations on a regular computer, and Figure 3 proposes an example of results
obtained for ultimate loads of sections in compression or under major-axis bending moment My.
It is observed that Types I-III usually keep a G.M.N.I.A. computation below 5 min and that
Type V may reach up to 20 min calculation time. In terms of precision, Types III and IV meshes
have been shown to lead to results with a difference lower than 3% compared to Type V taken as
a reference. Accordingly, Type III mesh was selected for all F.E. calculations, namely within
numerical parametric studies (Section 2.3), for both extruded and welded sections.

3
 Type I Type II Type III Type IV Type V
Figure 2: Various mesh densities studied, from coarser (Type I) to finer (Type V).

25
Compression N
Computation time [min]
20 Major-axis bending My

15

10

0
Type I Type II Type III Type IV Type V

Mesh type [-]

Figure 3: Average computation time for G.M.N.I.A. calculations for the various meshes considered.
To adequately represent the material response of aluminium, stress-strain relationships followed
the so-called “Ramberg-Osgood equation” (Ramberg and Osgood 1943), which has further been
modified by Hill to take into account permanent deformation using the offset method (Hill,
1944), as presented in Eq. (1). This material model was used in all calculations, and alloys 6061-
T6, 6063-T6 and 6082-T6 were considered – the latter can indeed be shown to be the most
representative ones for Structural Engineering applications.

   n
 
   0.002   (1)
E  Fy 
 
Following the recommendations of most design standards, the Youngs’ modulus was fixed as
E = 70 000 N/mm2 and the remaining values needed in Eq. (1) were set according to Canadian
Standards’ S157 (namely exponent n), see Table 1 (Coderre 2022).
Table 1: Properties of the aluminium alloys used in Ramberg-Osgood equations.
Non H.A.Z. H.A.Z.
Alloy Fy [N/mm2] Fu [N/mm2] n [-] Fwy [N/mm2] Fwu [N/mm2] n [-]
6061-T6 240 260 55 105 165 8.1
6063-T6 170 205 19.5 55 115 -
6082-T6 260 310 20.7 110 190 6.7

4
 350

300

250

 [N/mm2]
200

150

100
6082-T6
50 6061-T6
6063-T6
0
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
 [-]
Figure 4: Stress-strain relationships for the various alloys considered.

Figure 4 presents the resulting - curves obtained for the different aluminium alloys that were
considered herein, namely along the parametric studies. They have been introduced in the F.E.
models by means of a total of 100 - pairs spanning the entire material response. This number
of points has proven to be sufficient to accurately represent the behavior of non-linear materials
in the numerical model (Gagné 2022; Coderre 2022).
Material imperfections have not been accounted for in models relative to extruded sections.
According to studies conducted by Mazzolani (Mazzolani 1994), residual stresses in extruded
aluminium products – irrespective of the shape or the heat-treatment – are usually pretty low and
have an insignificant effect of the load-bearing capacity (Mazzolani 1994; Coderre 2022).
Yet, residual stresses patterns as in Figure 5b were considered for welded profiles, following
Kristensen’s study on 6082-T6 aluminium sections (Kristensen and Moan 1999). This
distribution is also very similar to the patterns used by Zha and Moan (Zha and Moan 2001; Zha
and Moan 2003) and by Rigo et al. (Rigo et al. 2003). The compressive residual stresses, c were
carefully calculated so that every plate respects a self-equilibrium condition between tensile and
compressive stresses; also, note that reductions in mechanical properties in the H.A.Z. have also
been taken into account in the F.E. models, by carefully modifying node coordinates, allowing to
precisely align the end of the shell elements with the extent of each zone (Coderre 2022). Finally,
careful numerical checks have been conducted to confirm that the introduction of this residual
stresses pattern was indeed leading to patterns in self-equilibrium (Coderre 2022).

5
 2bhaz + tw

+ 0.75 F wy
- - -c

+ bhaz + t f

r= 0

+ bhaz + t f

Welded sections -c 0.75 F wy

a) b)
Figure 5: Residual stresses patterns – a) Measured by Mazzolani (Mazzolani 1994) – b) Considered in the F.E.
models for welded sections.
The support conditions of the F.E. models were defined as ideal fork conditions, meaning that
axial displacement, strong axis rotation, weak axis rotation, and warping were allowed. To
achieve these conditions, specific degrees of freedom of both ends’ nodes were constrained. The
nodes of the flanges and of the web were restrained in the direction perpendicular to their plane,
which also prevented local buckling of the plates that could occur due to stress concentrations at
load application areas. Rotation about the longitudinal axis was restrained at the four nodes
directly at the web-to-flange junctions, in order to prevent torsion and out of plane rotation.
Axial displacements were also fixed at the middle node of the (short) member to allow both ends
to move towards or away from this point. The boundary conditions applied to the model are
schematically represented in Figure 6a.
z
y x
"x-free" node

z
"x-constrained" node y x

"rotation around x restrained" node

a) b)
Figure 6: Boundary condition and load application − a) Modelling of end sections: transverse supports and
constraints conditions − b) Application of compression or major-axis bending through nodal forces.
Further, to guarantee that beam theory assumptions were effectively fulfilled (“plane sections
remain plane”), kinematic linear constraints were applied to the nodes of the end sections
(Figure 6a). With these constraints, the axial displacement of each node was controlled by the
four corner nodes, i.e., the nodes at the flanges’ tips. This also allowed the different loads used in
the study – compression (N), major axis bending (My), and minor axis bending (Mz) – to be
applied as punctual loads acting axially on the corner nodes, see Figure 6b.

6
Initial geometrical imperfections were also considered in the numerical models, through
adequately modifying nodes coordinates. Typical patterns adopted are schematically represented
in Figure 7 and follow sinusoidal distributions; 3 half-waves in each plate have been considered,
the half-wave length being defined as the average between the flat “local buckling length” of the
flange af and its web counterpart aw. Each plate was assigned an imperfection’s amplitude as
1/200th of either af or aw – several variations relative to the shape of the imperfections or their
amplitudes have been studied beforehand and led to the values proposed here (Coderre 2022).

aw
(aw + af ) / 2
web plate
af
aw
flange plate (aw + af ) / 2

local imperfection

L = 3 (aw + af ) / 2

Figure 7: Sine-shaped geometrical imperfections in web and flange plates.

2.2 Validation against experimental results
Analyses performed thanks to the numerical model were compared to experimental test results
on the local behavior of aluminium I-sections found in the literature. In this respect, the
experimental results from Yuan et al. on stub columns (Yuan et al. 2015) have been used for
validation purposes. All test specimens were fabricated from extrusion, and no radius area was
present. Obviously, all measured material and geometrical data have been implemented in the
numerical models, as reported in Table 2 and Table 3. Further, for each test specimen, Yuan et
al. measured the initial local geometrical imperfections at three regions of the cross-sections: one
at mid-length (section B), and the others close to both ends (sections A and B). Figure 8a and
Figure 8b illustrate the imperfections’ distribution pattern of specimens T6-300-200 and T5-240-
75. The schematic view of the imperfections and their amplitude is also shown.
On the basis of Figure 8, it was chosen in the validation studies to implement the initial local
geometrical imperfections as a single sinusoidal half-wave, coupled with the measured
amplitudes in each plate. The direction of the imperfections was based on the schematic view of
Figure 8c, and the measured amplitudes are reported in Table 2.

7
 Table 2: Measured geometrical properties of specimens and test vs F.E. results.
Geom. Geom.
Die h b tf tw L imp. imp. Nu,Exp. Nu,F.E. / Nu,Exp.
Specimen Alloy
toolset [mm] [mm] [mm] [mm] [mm] web flange [kN] [–]
[mm] [mm]
T6-300-150 1 300.1 149.4 10.89 7.79 897.9 0.09 -0.07 1353.0 1.05
T6-300-200 1 300.1 199.2 10.87 7.67 899.4 -0.52 -0.13 1569.5 1.03
T6-280-160-R 2 279.5 159.4 9.9 6.94 839.8 0.27 0.17 1139.6 1.07
T6-350-110-R 3 349.3 111 10.06 7.72 1052.9 0.39 0.09 1057.8 0.90
6061-T6
T6-350-150-R 3 350.7 149.9 10.01 7.95 1049.4 -0.40 0.18 1271.0 0.96
T6-350-200-R 3 350.6 199.8 10 7.93 1049.8 -0.42 0.13 1442.2 0.98
T6-450-150-R 4 449.5 147.4 10.21 8.4 1351.5 1.45 0.08 1135.7 0.95
T6-450-200-R 4 449.9 199.7 10.5 8.64 1348.1 1.03 -0.33 1365.1 1.03
T5-240-75 5 239.8 74.7 6.96 4.44 719.6 -0.51 -0.19 254.1 0.93
T5-240-118 5 239.8 117.3 6.9 4.35 720.3 0.39 -0.22 340.2 0.98
T5-240-168 5 239.3 168.7 7 4.25 720.3 0.38 -0.24 446.8 0.91
T5-240-208 6063-T5 5 239 206.9 6.92 4.11 719.6 0.27 -0.49 520.6 0.83
T5-270-100 6 269.1 100.3 5.96 3.59 809.1 0.30 -0.21 295.1 0.91
T5-270-145 6 269.3 144.5 5.96 3.64 808.6 -0.23 -0.18 315.6 1.07
T5-270-178 6 268.9 178.0 5.83 3.63 809.0 0.30 -0.33 332.0 1.07
Mean 0.98
C.O.V. 0.08

Table 3: Measured material properties of specimens.

E Fy Fu n
Alloy Die toolset Location
[N/mm2] [N/mm2] [N/mm2] [–]
Flange 70 100 279.9 296.2 28.0
1
Web 68 900 258.1 289.7 25.8
Flange 72 600 257.8 291.5 25.8
2
Web 70 000 270.5 298.5 27.1
6061-T6
Flange 70 800 248.0 285.6 24.8
3
Web 72 500 264.8 297.9 26.5
Flange 68 300 231.7 260.6 23.2
4
Web 69 100 249.7 291.4 25.0
Flange 63 300 142.5 185.3 14.2
5
Web 67 200 154.6 201.6 15.5
6063-T5
Flange 63 700 170.2 215.6 17.0
6
Web 64 600 168.4 218.0 16.8

8
 a) b) c)
Figure 8: Measured local imperfections from (Yuan et al. 2015) − a) Distribution pattern for specimen T6-300-200 –
b) Distribution pattern for specimen T5-240-75 – c) Schematic view.
The experimental setup, in particular the support conditions, were recreated numerically as
accurately as possible (fixed ends with warping restrained, end rotations prevented). Table 2
reports the results obtained for each specimen and allows comparing each test result with its
numerically-predicted counterpart. Overall, the numerical results are seen to be in quite good
agreement with the experimental sources and are usually safe-sided. In average, some 2%
difference between sources is reported, with a Coefficient of Variation (C.O.V.) of 0.08, which is
excellent. As a complement, Figure 9a provides examples of typical stress-strain cures (obtained
from load-displacement data) where a good correspondence between the numerical sources and
the reference test data is observed, in terms of peak load, initial stiffness, post peak behaviour,
etc. Figure 9b further illustrates the similarities between experimental and numerical failure
modes. Accordingly, the F.E. models are deemed fit for being substituted to physical testing and
have been thoroughly used within numerical parametric studies, as detailed in the next
paragraphs.
250 T5-240-118 F.E.
T5-240-118 Exp.
T6-350-150-R F.E.
200 T6-350-150-R Exp.
Stress  [N/mm2]

150

100

50

0
0.000 0.005 0.010 0.015 0.020

Strain  [-]

a) b)
Figure 9: Examples of test vs F.E. − a) Load-displacement curves − b) Failures modes.

9
2.3 Parametric studies
The numerical models being validated, large amounts of both L.B.A. and G.M.N.I.A. numerical
simulations were performed to collect information on (i) the critical buckling response of such
aluminium I-shaped sections and on (ii) their ultimate (failure) loads, respectively. Overall, some
2 368 non-linear shell F.E. computations have been performed, with the following parameters
being varied:
 Cross-section types and shapes (Table 5): 30 I-shaped extruded sections selected from
Canadian and American extruders’ sections catalogs as well as 29 welded aluminium
cross-sections inspired by bridge girders (designation starts with “BG”), column shapes
(“HEA”) or invented sections (“IS”) have been considered;
 3 different heat-treated aluminium alloys: alloys 6061-T6 (Fy = 240 N/mm2), 6082-T6
(Fy = 260 N/mm2) and 6063-T6 (Fy = 170 N/mm2). Only the first 2 alloys wee considered
for welded sections, owing to limitations in fabrication and in welded area performance;
 Multiple load cases: axial compression N, major-axis bending My, minor-axis Mz and a
series of combined loading situations as characterized by Figure 13 and Table 4 (further
explanations on Figure 13 are given later in the paper).

n  N Npl  N  A  Fy  (2)

my  M y M pl,y  M y Wpl,y  Fy  (3)

mz  M z M pl,z  M z Wpl,z  Fy  (4)

tan   my  n  cos    mz  n  sin   (5)

tan   mz my (6)

Table 4: Combined load cases considered in this study.

φ [°] n [–] Load case
0.4 N + My
0
0.8 N + My
0 My + Mz
30 0.4 N + My + Mz
0.8 N + My + Mz
0 My + Mz
50 0.4 N + My + Mz
0.8 N + My + Mz
0 My + Mz
70 0.4 N + My + Mz
0.8 N + My + Mz
0.4 N + Mz
90
0.8 N + Mz

10
Table 5: Chosen cross-sections for parametric studies.
h b tf tw r b / tf h / tw
Manufacturing Cross-section
[mm] [mm] [mm] [mm] [mm] [–] [–]
I3x2 76.20 63.50 6.35 4.78 9.53 5.0 14.6
I3x2- 76.20 63.50 5.00 3.76 6.27 6.4 18.9
S687 76.20 50.80 6.35 4.78 9.53 4.0 14.6
MS186 95.25 30.15 3.18 1.83 1.65 4.7 50.3
I4x2 101.60 76.20 6.35 4.78 9.53 6.0 19.9
I4x2- 101.60 76.20 12.00 9.02 9.53 3.2 9.9
S181 101.60 53.98 3.81 4.06 3.81 7.1 24.1
S208 111.13 50.80 6.35 6.35 6.35 4.0 16.5
I6x4 152.40 101.60 7.37 4.83 7.62 6.9 30.0
I6x5- 152.40 101.60 16.00 10.48 7.62 3.2 13.0
I6x4- 152.40 101.60 6.00 3.93 6.21 8.5 37.3
WF6x9 152.40 152.40 12.70 9.53 15.88 6.0 14.7
I7x5 177.80 101.60 9.53 7.14 11.13 5.3 23.6
I7x5- 177.80 101.60 6.50 4.87 7.59 7.8 35.2
Extruded I8x6 203.20 101.60 9.53 7.14 11.10 5.3 27.1
shapes I8x6- 203.20 101.60 6.00 4.50 6.99 8.5 43.8
S263 203.20 133.35 7.82 5.84 8.13 8.5 33.5
S263- 203.20 133.35 6.50 4.85 6.75 10.3 40.6
I6x9- 203.20 203.20 12.70 9.53 15.88 8.0 20.0
I10x9- 254.00 127.00 8.75 6.00 9.83 7.3 40.9
I10x11 254.00 152.40 12.70 9.53 14.27 6.0 25.3
I10x11- 254.00 152.40 11.33 8.50 14.27 6.7 28.5
I8x7- 254.00 127.00 9.53 7.14 11.10 6.7 34.2
I10x11-- 254.00 152.40 7.00 4.27 6.83 10.9 57.8
I12x15 304.80 165.10 15.88 11.10 15.88 5.2 26.0
I12x15- 304.80 165.10 14.30 10.00 14.30 5.8 29.1
I12x14 304.80 177.80 15.75 7.87 10.16 5.6 36.7
I12x14- 304.80 177.80 12.00 6.00 7.74 7.4 48.8
I14x16 355.60 203.20 15.24 7.62 10.16 6.7 44.7
I14x16- 355.60 203.20 11.00 5.50 7.33 9.2 62.7
BG390 390.00 300.00 35.60 20.60 0.00 4.2 17.2
BG500 500.00 250.00 42.90 10.70 0.00 2.9 42.7
BG500- 500.00 250.00 57.10 14.30 0.00 2.2 31.0
BG600 600.00 250.00 28.60 19.10 0.00 4.4 29.9
BG600- 600.00 250.00 22.90 15.30 0.00 5.5 37.7
BG800 800.00 300.00 34.30 17.10 0.00 4.4 44.8
BG900 900.00 400.00 57.10 28.60 0.00 3.5 29.5
BG900- 900.00 300.00 42.90 22.90 0.00 3.5 37.4
BG1000 1000.00 400.00 57.10 22.90 0.00 3.5 41.2
BG1100 1100.00 400.00 57.10 28.60 0.00 3.5 36.5
BG1100- 1100.00 400.00 45.70 22.90 0.00 4.4 46.0
BG1200 1200.00 500.00 71.40 28.60 0.00 3.5 39.5
BG1200- 1200.00 500.00 57.10 22.90 0.00 4.4 49.9
HEA100- 192.00 200.00 16.00 10.00 0.00 6.3 17.6
Welded
HEA120- 228.00 240.00 16.00 10.00 0.00 7.5 21.2
shapes
HEA140- 199.50 210.00 19.10 12.40 0.00 5.5 14.5
HEA180- 213.80 225.00 11.90 7.50 0.00 9.5 26.9
HEA220 210.00 220.00 11.00 7.00 0.00 10.0 28.4
HEA260- 250.00 260.00 16.30 9.80 0.00 8.0 23.8
HEA320 310.00 300.00 15.50 9.00 0.00 9.7 32.7
HEA320- 310.00 300.00 20.20 11.70 0.00 7.4 24.8
HEA450- 440.00 300.00 27.30 15.00 0.00 5.5 27.5
HEA550 540.00 300.00 24.00 12.50 0.00 6.3 41.3
HEA550- 540.00 300.00 36.00 18.80 0.00 4.2 26.8
IS7x5 177.80 101.60 6.50 4.90 0.00 7.8 35.0
IS263 203.20 133.35 6.50 4.90 0.00 10.3 40.1
IS10x9 254.00 152.40 7.00 4.30 0.00 10.9 57.4
IS12x14 304.80 177.80 12.00 6.00 0.00 7.4 48.8
IS14x16 355.60 203.20 11.00 5.50 0.00 9.2 62.7

11
3. Proposed O.I.C.-based design approach
3.1 Bases and background
The F.E. results have served as reference to assess the merits of a new design procedure based on
the O.I.C. approach (results in Section 3.2 and in Section 3.3). The O.I.C. method relies on the
definition of a relative slenderness – denoted here as ̅λL for local (cross-sectional) resistance –
that takes the balance between plastic resistance and the section’s local instability. After ̅λL has
been determined, the method uses original cross-section buckling curves to determine a (local)
penalty factor χL that decreases the plastic capacity taken as reference owing to buckling effects
and imperfections. These various steps are illustrated on Figure 10.
Cross-sectional local behaviour (L)

Rcr,L Rpl
2 key factors: calculated by advanced tools (or by formulae)

Cross-section buckling curve

R pl
L L 
R pl Rcr , L
Rcr,L
Cross-section local Design check:
buckling curve  L  R pl
L Rb , L   1.0 (Eurocode 3 format)
M
Rb , L     L  R pl  1.0 (AISC format)
L

Figure 10: O.I.C. design flow chart for cross section resistance.
It is important to note that some coefficients may or shall be calculated by tools – the O.I.C.
approach is among the first ones to offer the possibility to rely on numerical tools for the
calculation of key coefficients. In Figure 10, Rcr,L stand as the load ratio by which the initial
loading shall be multiplied to reach the (local) critical state; Rcr,L therefore corresponds to an
L.B.A. calculation. Rb,L is the factor by which the initial loading shall be multiplied to reach the
ultimate section capacity, and is therefore obtained through a G.M.N.I.A. computation. Use of
such load factors or load ratios is made necessary to address combined loading situations through
the same simple concept. In situations where a simple load case is considered, R-ratios can
usefully be substituted by their internal forces counterparts (e.g., under compression N, Ncr,L and
Nb,L can be used in lieu of Rcr,L and Rb,L, respectively). Within the present study and in an effort
to ensure accuracy and consistency with the sources of the results presented, ABAQUS was used
to provide numerically-computed values of Rcr,L and Rb,L. As for Rpl factor, which represents the
load factor to reach the plastic capacity (all fibers yielded), a separate fiber-based cross-section
model was programmed specifically, with the ability to account for (i) the section’s accurate
geometry and (ii) a lower yield stress in the H.A.Z. for welded sections. Accordingly, plastic
capacities through factor Rpl were quite accurately taken into account. Further details and
background information on the O.I.C. approach may be found in (Boissonnade et al. 2017;
Gérard et al. 2021; Li et al. 2022; Coderre 2022).
As a particular point of the proposed approach, the selection of an appropriate buckling curve is
essential. Observations showed that there cannot be a unique, safe-sided curve unless it stands as
a (very) conservative one. Exceedingly simple, this was deemed inappropriate, inaccurate and
uneconomical and specific analyses were led to understand how a series of cross-section

12
buckling curves should be defined, i.e., identify parameters responsible for different resistance
responses beyond the ones already involved in the O.I.C. approach.
This led to identifying a geometrical parameter, denoted as γ (see Eqs. (7) and (8)), to be
responsible for the scatter in observed results. Therefore, a series of buckling curves with a
dependency on γ have been established and are detailed in Section 3.2 for extruded sections and
along Section 3.3 for welded shapes.
2
 h   b  t 
 E       w  for extruded shapes (7)
 t w   tf   tf 
0.8
h  b  t  3
W      w  10 for welded shapes (8)
 tw   tf   tf 
3.2 Design proposals for extruded sections
In a first step, O.I.C.-based design equations have been developed for extruded I-shapes. For
compact sections with ̅λL < λ0 where λ0 is a reference (limit) slenderness, a strain-based approach
is suggested so as to better account for strain hardening resistance reserves. A relationship
between ̅λL and the ratio peak / y is first established, where peak is the strain level at peak load
and y is the strain at 0.2. This ratio stands as a sound indicator of the potential benefits from the
pronounced non-linear material response of aluminium alloys, and allows, through a 2 nd step that
links peak / y to χL, to provide resistance predictions higher than the usual, conventional plastic
capacity. Figure 11a and Figure 11b illustrate this strain-based approach for compact aluminium
I-shapes under major axis bending moment My.

25 1.4
6061-T6
6063-T6
6082-T6 1.3
20
peak / y = 1
1.2
peak / y [-]

Proposal 6061-T6
15
L [-]

Proposal 6063-T6
1.1 6061-T6
Proposal 6082-T6
10 6063-T6
1.0 6082-T6
Proposal 6061-T6
5 0.9 Proposal 6063-T6
Proposal 6082-T6
0 0.8
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 5 10 15 20 25 30

L [-] peak / y [-]

a) b)
Figure 11: Principle of two-step procedure for L ≤ 0 (major-axis bending cases).
For more slender shapes, i.e., ̅λL > λ0, the proposed approach more classically relies on a direct
̅λL – χL approach, through a so-called “modified Ayrton-Perry” approach (Ayrton and Perry
1886; Boissonnade et al. 2017; Li et al. 2022) – cf. Table 6 for definitions and coefficients. In the
proposed equations, parameter L stands as a general (cross-sectional) imperfection factor while
 aims at specifically accounting for plate post-buckling effects; note that both L and  are
functions of geometrical parameter .

13
  in [0.006;0.1[
 in [0.1;0.15[  in [0.006; 0.17[
 in [0.15;0.26[  in [0.17; 0.3[
 in [0.26; 0.47[  in [0.3; 0.47[
Proposal  = 0.1 Proposal  = 0.17
Proposal  = 0.15
1.4 Proposal  = 0.3
1.4 Proposal  = 0.26
Proposal  = 0.47
Proposal  = 0.47 6082-T6
6082-T6 1.3
6063-T6
1.2 6063-T6
1.2 6061-T6
6061-T6
Proposal 6082-T6
Proposal 6082-T6

L [-]
L [-]

1.1 Proposal 6063-T6

1.0 Proposal 6063-T6
Proposal 6061-T6
Proposal 6061-T6
1.0
0.8
0.9

0.6 0.8
0.0 0.5 1.0 1.5 2.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2
L [-] L [-]
a) b)
Figure 12: Performance of proposed approach for extruded sections – a) Compression – b) Major-axis bending.

Figure 12a and Figure 12b propose a series of results over the entire ̅λL domain, i.e., for compact
and slender shapes, for the cases of simple compression N and major-axis bending My. It can be
observed that the proposed approach provides quite accurate yet safe resistance estimates when
compared to the reference F.E. data – note however that the figures only plot a few lower bound
curves of the O.I.C. proposal (since the buckling curves are γ-dependent) while each individual
data point possesses it own unique γ value, so that the curves plotted are usually safe-sided for
the interval of γ they correspond to.

Figure 13: 3D resistance surface.

14
Table 6: Design proposal for aluminium cross-sections under simple load cases.
Manufacturing
Load cases Compression N Major-axis bending My Major-axis bending Mz
Types
For  L,N  0  0.55 : For  L,My  0  0.60 : For  L,Mz  0  0.60 :
Alloy 6063-T6 L,N  -0.067 - 0.24  ln( L,N )  L,My  1.27 - 0.46   L,My  L,Mz  1.22 - 0.35   L,Mz
Alloy 6061-T6 L,N  1.19 - 0.35   L,N  L,My  1.12 - 0.21  L,My  L,Mz  1.08 - 0.14   L,Mz
Alloy 6082-T6  L,N  1.44 - 0.79   L,N  L,My  1.40 - 0.66   L,My  L,Mz  1.22 - 0.36   L,Mz
For  L,N  0  0.55 : For  L,My  0  0.60 : For  L,Mz  0  0.60 :
Extruded
cross-sections
  
L  0.5  1   L   L,N  0   L,N

   

L  0.5  1   L   L,My  0   L,My    
L  0.5  1   L   L,Mz  0   L,Mz


  
 L,N   L,My   L,Mz 
Ayrton-Perry   
L     L,N
2
L     L,My
2
L  L2   L,Mz
format L L

L  L  L 
0.01  0.25 E 0.4  0.3 E 0.01  0.18 E 0.3  0.1 E 0.2 E 0.4  0.2 E

For  L,N  0  0.50 : For  L,My  0  0.45 : For  L,Mz  0  0.60 :

Alloy 6061-T6  L,N  1.12 - 0.24   L,N L,My  1.12 - 0.27   L,My  L,Mz  1.10 - 0.16   L,Mz
Alloy 6082-T6  L,N  1.37 - 0.73   L,N  L,My  1.18 - 0.40   L,My  L,Mz  1.24 - 0.40   L,Mz
For  L,N  0  0.50 : For  L,My  0  0.45 : For  L,Mz  0  0.60 :
Welded
Cross-sections   
L  0.5  1   L   L,N  0   L,N

   

L  0.5  1   L   L,My  0   L,My    
L  0.5  1   L   L,Mz  0   L,Mz


  
Ayrton-Perry  L,N   L,My   L,Mz 
  
format L  L2   L,N L  L2   L,My L  L2   L,Mz
L  L  L 
0.04  0.65 W 0.5  1.50 W  0 0.06  0.55 W 1.4 1.5 W 0.13  0.2 W 0.15 1.4 E
As for combined load cases, the proposed O.I.C. approach relies on the definition of a 3D
resitance surface (see Figure 13), where each axis refers to a relative amount of an internal
force – either relative axial compression n or relative major or minor-axis bending moments my
and mz, cf. Eqs. (2) – (4). This surface is such that any load combination leading to a point below
this surface indicates that the design under the considered forces is safe-sided, while a point lying
on this surface means that maximum resistance is just reached – obviously, loading points
outside/above this surface denotes that capacity is exceeded and that loading shall be reduced. A
spherical system of coordinates shows appropriate in this context, and use of angles  and  as
defined in Eqs. (5) and (6) is useful and efficient. In this respect, Eq. (9) may suitably be used for
combined load cases. In Eq. (9), q factors allow for adjustments in the proposed resistance
surface at both (i) the overall level (q1 affects the general shape of the 3D surface), or (ii) a more
localized level, through factors q2 to q6 which enable local tuning of the surface where a given
internal force is dominant. For example, q2 locally modifies the shape of the resistance surface
when axial compression is dominant. The q factors have been optimised for providing the best
compromise between accuracy, simplicity and safety, and their proposed values are summarized
in Table 7.
1
  q1

    L,N  cos q2     L,M y  sin q3   cos q4    
q1 q1 q1
 L,combined  sin   sin 
q5 q6
 (9)

L,M z 

Table 7: O.I.C. design factors for combined load cases.

q factors q1 q2 q3 q4 q5 q6
Extruded cross-sections 9 0.19 1.2 0.13 3.5 8
Welded cross-sections 5 0.80 1.3 0.6 2.2 3

Figure 14a and Figure 14b summarize the 1 170 F.E. results for various load combinations
against which the accuracy and safety of the proposed approach can be tested. Figure 14a
examines the overall accuracy of the O.I.C. proposal through plotting the frequencies of
the L,O.I.C / L,F.E. ratio along various accuracy intervals. Ratios < 1.0 denote safe O.I.C.
resistance predictions, while ratios above unity point towards unsafe estimates. As can be
observed, the proposed design approach here again provides excellent results, precise and safe-
sided, with a very good consistency level as indicated by a narrow standard deviation. Figure 14b
proposes to split the results by load cases, and allows to identify the situations where predictions
are either unsafe or over-conservative. A vast majority of the results fall in between the targeted
L,O.I.C / L,F.E. ratios, which are highlighted by the green and the red dotted lines. The prediction
for results from combined load cases with N + My + Mz generally tend to be a bit unconservative,
with some results below the 1.0 line. Most of the results are quite close to the 1.0 line and,
therefore, are not so unconservative – this shall easily be compensated by usual safety factor
values, see final design step in Figure 10.
 Safe Unsafe 1.2
600

L,O.I.C. / L,F.E. [-]

1.0
Frequency [-]

400

0.8
200 N + My L,F.E. = L,O.I.C.
0.6 N + Mz 10% safety
My + Mz 3% unsafety
0
N + My + Mz
]

.8 5]

.1 5]

]
]

] 0.4
5

.2
;1

]1 .05
.8

]0 0.9

]1 1.1
8

.9

1
;0

;1
.

.
;0

;0

;1

;1
5
;

;
.9

5
75

5
0 200 400 600 800 1000
]1
8

.9

.0

.1
]0
.

.
]0

]0

]0

]1 L,O.I.C. / L,F.E. [-] Case #

a) b)
Figure 14: a) Accuracy of proposal for combined load cases (extruded sections, all cases included) – b) Performance
of proposal based on load cases.
For a more detailed analysis of the results, Table 8 compiles statistical results obtained with the
O.I.C. design proposal for extruded sections, sorted per load combination. It shows that for
simple load cases, the design proposal presents a great accuracy. All three simple load cases have
an average close to 1.0, and a very low standard deviation, the highest being as low as 3.0% for
pure compression. All standard deviation values reported are excellent, and virtually no
problematical unsafe result is reported – only very few values are indeed higher than 1.05. The
most conservative value for simple load cases is reported as 0.86 for pure compression, which is
still considered as fairly accurate.
For combined load cases, Table 8 also shows that the proposal meets a very good accuracy. The
average L,O.I.C / L,F.E. ratios range from 0.93 to 1.0 and the highest standard deviation is only
6.0%, meaning that the results are very consistent and continuous. Results for N + Mz cases
present the best performance, and, as expected, the most complex N + My + Mz combinations are
associated with the “worst” performances: even if it has a really low standard deviation of 4.5%
and an average of exactly 1.0, almost half of the results (48%) have a value higher than 1.0.
However, this number decreases drastically to 14% for values over 1.05, meaning that almost all
the results are located between 1.0 and 1.05, which is considered accurate. As mentioned
previously, a safety factor will eventually complement the proposal, shifting the unconservative
results towards safer results.
The overall performance of the design proposal for extruded sections is therefore very
satisfactory. Including all results, the average L,O.I.C / L,F.E. ratio is 0.97, with a standard
deviation of 5.6%, meaning that the vast majority of the results are grouped just under a value of
1.0. Indeed, only 10% of all the results are below 0.9, and only 7% are higher than 1.05. When
looking at unsafe values over 1.1, this value drops to less than 2%.

17
 Table 8: Statistical results of proposal for extruded sections based on load combination.
Load case N My Mz N + My N + Mz My + Mz N + My + Mz All
Average 0.97 0.97 0.93 0.93 0.94 0.94 1.00 0.97
Maximum 1.01 1.06 1.02 1.10 1.05 1.13 1.18 1.18
Minimum 0.86 0.90 0.89 0.82 0.79 0.82 0.86 0.79
C.O.V. 0.03 0.02 0.03 0.05 0.06 0.06 0.05 0.06
Values < 0.8 0 0 0 0 3 (2%) 0 0 3 (0%)
Values < 0.9 3 (3%) 0 4 (4%) 34 (19%) 39 (22%) 64 (24%) 7 (1%) 151 (10%)
Values > 1.0 7 (8%) 4 (4%) 1 (1%) 9 (5%) 20 (11%) 25 (9%) 257 (48%) 323 (22%)
Values > 1.05 0 1 (1%) 0 4 (2%) 0 15 (6%) 75 (14%) 95 (7%)
Values > 1.1 0 0 0 1 (1%) 0 8 (3%) 15 (3%) 24 (2%)
Values > 1.2 0 0 0 0 0 0 0 0
Total cases 90 90 90 180 180 270 540 1440

800
Ref. = A.D.M. Safe Unsafe
Ref. = Eurocode 9
Ref. = S157
600 Ref. = O.I.C.
Frequency [-]

400

200

0
]0.25 ; 0.3]

]0.3 ; 0.35]

]0.35 ; 0.4]

]0.4 ; 0.45]

]0.45 ; 0.5]

]0.5 ; 0.55]

]0.55 ; 0.6]

]0.6 ; 0.65]

]0.65 ; 0.7]

]0.7 ; 0.75]

]0.75 ; 0.8]

]0.8 ; 0.85]

]0.85 ; 0.9]

]0.9 ; 0.95]

]0.95 ; 1]

]1 ; 1.05]

]1.05 ; 1.1]

]1.1 ; 1.15]

]1.15 ; 1.2]

]1.2 ; 1.25]

L,Ref. / L,F.E. [-]

Figure 15: Performance of proposed approach vs existing ones (Eurocode 9, A.D.M. and S157) for combined load
cases – extruded sections.
The pertinence of the proposed approach was also tested against existing design approaches,
namely the the European Standards (Eurocode 9 Part 1-1 (European Committee for
Standardization 2007)), the American Standards (cf. Aluminum Design Manual (Aluminum
Association 2020)) and the Canadian Standards (CSA S157-17 (Canadian Standards Association
2017)). In this respect, reduction factors χL – which stand as a direct meaure of resistance – have
been calculted as recommended in these standards and compared to the proposed approach and
to the reference F.E. results. Figure 15 first proposes summary histograms of frequencies of the
χL,Ref. / χL,F.E. ratio, where χL,Ref. sucessively refers to one of the standards or to the O.I.C.
proposal. Overall, this figure shows that (i) the O.I.C. approach proposes, by far, the best
resistance predictions both regarding accuracy, consistency and safety and that (ii) all other

18
design approaches yield quite conservative resistance estimates, where the predicted resistances
can even be doubled sometimes.
N + My L,F.E. = L,Eurocode 9 N + My L,F.E. = L,A.D.M.
N + Mz 10% safety N + Mz 10% safety
My + Mz 3% unsafety My + Mz 3% unsafety
1.2 1.2
N + My + Mz N + My + Mz
L,Eurocode 9 / L,F.E. [-]

L,A.D.M. / L,F.E. [-]

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4
0 200 400 600 800 1000 0 200 400 600 800 1000
Case # Case #
a) b)
N + My L,F.E. = L,S157
N + Mz 10% safety
1.2 My + Mz 3% unsafety
N + My + Mz
L,S157 / L,F.E. [-]

1.0

0.8

0.6

0.4
0 200 400 600 800 1000
Case #
c)
Figure 16: Performance of proposed approach vs existing ones (Eurocode 9, A.D.M. and S157) for combined load
cases (extruded sections) – a) Results for Eurocode 9 – b) Results for A.D.M. – c) Results for S157.

Further, Figure 16a to Figure 16c propose a load case-dependent view of the same results, for
each standard. It is observed that Eurocode 9 generally underestimates the carrying capacity for
every combined loading situation, in particular when My and/or Mz are dominant. In average,
Eurocode 9 provides some 30% over-conservatism, and leads to quite scattered predictions.
Similarly, A.D.M. specifications lead to quite over-safe resistance estimates, yet a little less
conservative than Eurocode 9 and still very scattered. S157 is seen to lead to the worst results in
terms of accuracy; the results are also very scattered and never get any closer than 10% on the
safe side, which is rather uneconomical.

19
All these results ought to be compared to the ones reached for the O.I.C. proposal presented on
Figure 14b, to emphasize how significantly better the proposed approach stands in all relevant
aspects of accuracy, consistency and economy.
3.3 Design proposals for welded sections
Design equations for welded sections have been derived in the same way as for extruded shapes,
namely with (i) a proposal based on the peak / y ratio for the most compact shapes, (ii) a
generalized Ayrton-Perry format for the slender shapes and (iii) the definition of a 3D resistance
surface for combined load cases. Also, for each load case, a scatter in the results was found
associated with the various dimensions of the sections, so that a similar γW factor was defined as
in Eq. (8), in order to propose a series of (local) buckling curves.
Figure 17a illustrates the results obtained for this proposal for welded sections in compression,
while Figure 17b proposes results for sections bent about the major-axis My. As can be observed,
the O.I.C. proposal provides very good resistance predictions in comparison to the reference F.E.
data points. Further, Figure 18a plots a statistical summary of the results obtained for combined
load cases; here again, the O.I.C.-based approach is seen to lead to accurate, consistent and
generally safe-sided predictions. In Figure 18b, the individual load case results are detailed, and
it is again observed that the proposed approach gives quite satisfactory results, whatever the load
case. Especially for the most complex N +My + Mz cases, the performance of the proposed
approach is seen very good.
 in [0.018; 0.06[
 in [0.06; 0.14[  in [0.018; 0.12[
 in [0.14; 0.22[  in [0.12; 0.2[
 in [0.22; 0.35[  in [0.2; 0.35[
1.2 1.3 Proposal  = 0.12
Proposal  = 0.06
Proposal  = 0.14 Proposal  = 0.2
1.1 1.2
Proposal  = 0.22 Proposal  = 0.35
Proposal  = 0.35 6082-T6
1.0 1.1 6061-T6
6082-T6
Proposal 6082-T6
L [-]

L [-]

0.9 6061-T6
1.0 Proposal 6061-T6
Proposal 6082-T6
Proposal 6061-T6
0.8 0.9

0.7 0.8

0.6 0.7
0.0 0.5 1.0 1.5 2.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2
L [-] L [-]
a) b)
Figure 17: Performance of proposed approach for welded sections – a) Compression – b) Major-axis bending.

Table 9 gives further details on how the results are distributed per load case with relevant
statistical data. Overall, the O.I.C. proposal is 5% conservative on the 928 F.E. reference cases
considered, with a C.O.V. lower than 9%, which is excellent. All average and C.O.V. values can
be seen to be outstanding, given the complexity of the problem and the many variables
considered (very different geometries, alloys, load conbinations, etc.). The proposal is also seen
to provide very little over-safe predictions (less than 4% of the 928 results exceed 20% over-
conservatism) as well as limited unsafe resistance estimates (6% of the results are over the 10%
limit beyond which usual safety factors compensate for unsafe predictions).

20
 Table 9: Statistical results of proposal for welded sections based on load combination.
Load case N My Mz N + My N + Mz My + Mz N + My + Mz All
Average 0.95 0.96 0.95 0.94 0.94 0.92 0.97 0.95
Maximum 1.01 1.02 0.99 1.15 1.19 1.18 1.21 1.21
Minimum 0.87 0.83 0.9 0.78 0.74 0.71 0.78 0.71
C.O.V. 0.033 0.042 0.023 0.081 0.099 0.098 0.092 0.087
Values < 0.8 0 0 0 2 (2%) 14 (12%) 22 (13%) 1 (0%) 39 (4%)
Values < 0.9 5 (9%) 6 (10%) 0 29 (25%) 32 (28%) 74 (43%) 90 (26%) 236 (25%)
Values > 1.0 3 (5%) 6 (10%) 0 24 (21%) 23 (20%) 33 (19%) 112 (32%) 201 (22%)
Values > 1.05 0 0 0 15 (13%) 13 (11%) 17 (10%) 58 (17%) 103 (11%)
Values > 1.1 0 0 0 7 (6%) 9 (8%) 9 (5%) 35 (10%) 60 (6%)
Values > 1.2 0 0 0 0 0 0 5 (1%) 5 (1%)
Total cases 58 58 58 116 116 174 348 928

N + My L,F.E. = L,O.I.C.
N + Mz 10% safety
300 My + Mz 3% unsafety
Safe Unsafe
N + My + Mz
250
1.2
200
L,O.I.C. / L,F.E. [-]

1.0
Frequency [-]

150
0.8
100
0.6
50
0.4
0

0.2
]0.95 ; 1]

]1 ; 1.05]
]0.7 ; 0.75]

]0.75 ; 0.8]

]0.8 ; 0.85]

]0.85 ; 0.9]

]0.9 ; 0.95]

]1.05 ; 1.1]

]1.1 ; 1.15]

]1.15 ; 1.2]

]1.2 ; 1.25]

0 200 400 600

L,O.I.C. / L,F.E. [-] Case #
a) b)
Figure 18: a) Accuracy of proposal for combined load cases (welded sections, all cases included) – b) Performance
of proposal based on load cases.

21
Similarly to extruded sections, the histograms in Figure 19 further illustrate how much of an
improvement the proposed O.I.C. approach can represent compared to existing design
provisions. It is in particular interesting to note that (i) other design approaches provide very
scattered, inconsistent predictions and that (ii) some results show extremely conservative: as an
example, one can cite the non-negligible amount of results from Eurocode 9 or S157 having
χL,Ref. / χL,F.E. ratios in the vicinity of 0.33 or lower, which indicates that the failure loads
predicted by these codes may actually be multiplied by at least a factor 3 to reach the reference
ultimate capacities of the F.E. models.
200

Ref. = A.D.M. Safe Unsafe

Ref. = Eurocode 9
Ref. = S157
150 Ref. = O.I.C.
Frequency [-]

100

50

0
]0.25 ; 0.3]

]0.3 ; 0.35]

]0.35 ; 0.4]

]0.4 ; 0.45]

]0.45 ; 0.5]

]0.5 ; 0.55]

]0.55 ; 0.6]

]0.6 ; 0.65]

]0.65 ; 0.7]

]0.7 ; 0.75]

]0.75 ; 0.8]

]0.8 ; 0.85]

]0.85 ; 0.9]

]0.9 ; 0.95]

]0.95 ; 1]

]1 ; 1.05]

]1.05 ; 1.1]

]1.1 ; 1.15]

]1.15 ; 1.2]

]1.2 ; 1.25]
L,Ref. / L,F.E. [-]

Figure 19: Performance of proposed approach vs existing ones (Eurocode 9, A.D.M. and S157) for combined load
cases – welded sections.
As a complement to Figure 19, Figure 20 breaks down these results per individual load cases
(only combined load cases are reported on this figure), and allows to further illustrate that (i) all
existing code proposals provide extremely conservative, over-safe resistance predictions and that
(ii) they all lead to very scattered results, although the A.D.M. approach is seen to be the best
among them. When compared to Figure 18b, it further illustrates that the O.I.C.-based proposal
brings significant improvements in accuracy/economy, consistency and safety.

22
 N + My L,F.E. = L,Eurocode 9 N + My L,F.E. = L,A.D.M.
N + Mz 10% safety N + Mz 10% safety
My + Mz 3% unsafety My + Mz 3% unsafety
1.2 N + My + Mz 1.2 N + My + Mz
L,Eurocode 9 / L,F.E. [-]

L,A.D.M. / L,F.E. [-]

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2
0 200 400 600 0 200 400 600
Case # Case #
a) b)

N + My L,F.E. = L,S157
N + Mz 10% safety
My + Mz 3% unsafety
1.2 N + My + Mz
L,S157 / L,F.E. [-]

1.0

0.8

0.6

0.4

0.2
0 200 400 600
Case #
c)
Figure 20: Performance of proposed approach vs existing ones (Eurocode 9, A.D.M. and S157) for combined load
cases (welded sections) – a) Results for Eurocode 9 – b) Results for A.D.M. – c) Results for S157.

4. Conclusions
This paper presented an original O.I.C.-based approach for the design of aluminium I-shaped
sections under simple or combined load cases. For simple load cases, this design proposal relies
on a specific, strain-based 2-step approach for predicting the resistance of compact aluminium
shapes, and on a so-called “extended Ayrton-Perry” format for the most slender shapes, both
providing direct capacity estimates. For combined loading situations, the suggested approach
relies on a 3D resistance surface and on the use of a system of spherical coordinates. Advanced
non-linear shell F.E. models were also developed then validated against more than 15 test results.
Then, some 2 300 numerical simulations were performed to gather a large dataset of reference
results that were used to investigate the adequacy of the proposed design approach. The latter is
proved very satisfactory in terms of accuracy, economy, safety and consistency. Compared to
existing design provisions for aluminium profiles in Eurocode 9, in the American A.D.M.
recommendations or in the Canadian Standards S157, the suggested approach is seen to

23
significantly improve design predictions, making this proposal an interesting candidate for an
alternative design proposal intended at practitioners.

5. References
Abaqus (2011). “Abaqus 6.11:” Dassault Systemes Simulia Corporation, Providence, RI, USA.

Aluminum Association (2020). Aluminum Design Manual: 2020, Aluminum Association,

Washington, D.C.

Ayrton, W.E. and Perry, J. (1886). “On struts:” The Engineer, Vol. 62.

Bathe, K. J. (2016). “Finite Element Procedures”, 2nd Edition, Prentice Hall, Pearson Education.

Beyer, A. (2017). On the Design of Members with Open Cross Sections Subject to Combined
Axial Compression, Bending and Torsion, PhD thesis, Lorraine University.

Boissonnade, N., Hayeck, M., Saloumi, E. and Nseir, J. (2017). “An Overall Interaction Concept
for an alternative approach to steel members design:” Journal of Constructional Steel
Research, Vol. 135, pp. 199–212.

Canadian Standards Association (2017). CSA S157-17: Strength design in aluminum.

Coderre, T. (2022). Development of an alternative design method for aluminium open cross-
sections using the Overall Interaction Concept, MSc Thesis, Civil and Water Engineering
Department, Laval University.

European Committee for Standardization (2007). Eurocode 9: Design of aluminium structures -

Part 1-1: General structural rules, Commission of the European Community, Brussels,
Belgium.

Extrudex Aluminium Inc. (2020). Standard Shapes Catalogue.

Gagné, A.-S. (2022). Design of stainless steel welded I-sections by means of the Overall
Interaction Concept, MSc Thesis, Université Laval, Québec, Canada.

Gagné, A.-S., Gérard, L. and Boissonnade, N. (2020). “Design of stainless steel cross-sections
for simple load cases with the O.I.C.:” Journal of Constructional Steel Research, Vol.
168, p. 105936.

Gérard, L., Li, L., Kettler, M. and Boissonnade, N. (2019). “Recommendations on the
geometrical imperfections definition for the resistance of I-sections:” Journal of
Constructional Steel Research, Vol. 162, p. 105716.

Gérard, L., Li, L., Kettler, M. and Boissonnade, N. (2021). “Steel I-sections resistance under
compression or bending by the Overall Interaction Concept:” Journal of Constructional
Steel Research, Vol. 182, p. 106644.

24
Guo, Z.T. (1989). “Bending and Torsion of Thin-Walled Members:” Chinese Architectural and
Industrial publishing House Beijing.

Hill, H.N. (1944). Determination of stress-strain relations from “offset” yield strength values.

Kristensen, O.H.H. and Moan, T. (1999). “Ultimate strength of aluminium plates under biaxial
loading:” Proc., Fifth Int. Conf. on Fast Sea Transportation, Society of Naval Architects
and Marine Engineers.

Li, L., Gérard, L., Kettler, M. and Boissonnade, N. (2022). “The Overall Interaction Concept for
the design of hot-rolled and welded I-sections under combined loading:” Thin-Walled
Structures, Vol. 172, p. 108623.

Mazzolani, F. (1994). Aluminium Alloy Structures, CRC Press.

Metra Aluminium Inc. (2019). Catalogue FR - Profilé en “I” avec rayons.

Ramberg, W. and Osgood, W.R. (1943). Description of stress-strain curves by three parameters.

Rigo, P., Sarghiuta, R., Estefen, S., Lehmann, E., Otelea, S.C., Pasqualino, I., Simonsen, B.C.,
Wan, Z. and Yao, T. (2003). “Sensitivity analysis on ultimate strength of aluminium
stiffened panels:” Marine Structures, Vol. 16, No. 6, pp. 437–468.

Riks, E. (1979) "An Incremental Approach to the Solution of Snapping and Bucking Problems,"
International Journal of Solids and Structures, Vol. 15, pp. 529-551.
Sapa Inc. (2008). Hydro Catalog Standard Shapes.

Yuan, H.X., Wang, Y.Q., Chang, T., Du, X.X., Bu, Y.D. and Shi, Y.J. (2015). “Local buckling
and postbuckling strength of extruded aluminium alloy stub columns with slender I-
sections:” Thin-Walled Structures, Vol. 90, pp. 140–149.

Zha, Y. and Moan, T. (2001). “Ultimate strength of stiffened aluminium panels with
predominantly torsional failure modes:” Thin-Walled Structures, Vol. 39, No. 8, pp. 631–
648.

Zha, Y. and Moan, T. (2003). “Experimental and Numerical Prediction of Collapse of Flatbar
Stiffeners in Aluminum Panels:” Journal of Structural Engineering, Vol. 129, No. 2, pp.
160–168.

25