Rev.

A (Progress Report 15)

5/6/2003

Paper 1
Laterally supported aluminum flexural members with
symmetric cross sections
Yongwook Kim and Teoman Pekz

Abstract
The Specification for Aluminum Structures published by the Aluminum
Association does not fully account for the plastic-ultimate bending capacity: For
compressive component elements, the allowable stress is based on the yield stress.
In this study, possible modifications to the current specification are suggested so
that rather compact extruded flexural members can be evaluated more precisely. As a first
step, limit state stress equations are modified so that the plastic-ultimate capacity is fully
incorporated. Closed form solution of the ultimate shape factor is provided to set a new
cut-off corresponding to the ultimate stress. The ultimate shape factor is simplified for
practical design purposes. A parametric study of component elements using the finite
element analyses suggests a complete form of the modified limit state stress equations.
Using the modified limit state stress equations, more accurate approaches are proposed to
compute the plastic-ultimate moment capacity.
A parametric study of doubly symmetric I-shaped sections using the finite
element method shows that the approaches developed in this study are significantly better
than the current specification provisions. Flexural tests conducted in this research also
support the approaches.
To maintain a certain factor of safety against yielding in the allowable stress
design equations, an approach to use only a potion of the ultimate-plastic capacity is
suggested. Using this approach, a larger inelastic reserve capacity is recognized for the
material with a larger margin between the ultimate and yield stresses.

Paper 1 - 1

Rev. A (Progress Report 15)

5/6/2003

Introduction: The AA Specification for component elements

The Specification for Aluminum Structures published by the Aluminum
Association (2000a, abbreviated to the AA Specification) requires designers to check
whether the computed stress based on the applied loads is less than both of the allowable
stresses of a tension side component element as summarized in Table 1. The factors in
front of the tensile yield and ultimate stresses for the web incorporate the shape of a nonlinear stress distribution, which is often called the shape factor.
Table 1. Options to Compute Tension Limit State Stress of a Flexural Member by AA
(2000a)
Component element

Allowable stress based on

limit state of yield stress

Allowable stress based on

limit state of ultimate stressa

Tension flange of
structural shapes

Fty/ny

Ftu/nu

Tension web of
1.42Ftu/nu
1.30Fty/ny
structural shapes
a
Except for some of 2014-T6, 6066-T6 and 6070-T6 family materials, for which nu
is replaced with ktnu due to notch sensitivity. See AA (2000a) for details.
Fty = tensile yield stress
Ftu = tensile ultimate stress
ny = factor of safety on yield stress
nu = factor of safety on ultimate stress

However, the check against the limit state of the ultimate stress is not available when a
limit state stress is computed for a compressive component element. Compressive limit
state stress equations for a component element refer to yielding, inelastic buckling, and
post buckling ranges as shown in Figure 1.
For this reason, the ultimate-plastic capacity of aluminum members has not been
fully incorporated in the specifications, although aluminum alloys are hardening and
ductile materials.

Fig. 1. Limit state stress for a component element in the AA Specification

The limit state stress equations in this figure are expressed in terms of the equivalent
slenderness ratio (p) as defined in Equation (1), originating from Equation (2).

Paper 1 - 2

Rev. A (Progress Report 15)

p =

5/6/2003

3.266 b
b
= for = 1/3
t
kp t

(1)

2E
2E

2
p2
12(1 2 ) ( b / t )

(2)

Fcr = k p

For a flange or web with junctions between component elements idealized as simplysupported boundary conditions, equivalent slenderness ratios depending on plate buckling
coefficient kp are listed in the AA Specification as exampled in Table 2.
Table 2. Equivalent slenderness ratios in the AA Specification
(S.S denotes simply-supported boundary condition)

(a) AA 3.4.16

p = 1.6
t

(b) AA 3.4.15

p = 5.1
t

(c) AA 3.4.18

p = 0.67
t

The post-buckling equation seen in Figure 1 was proposed by Jombock and Clark
(1968) to take account of the non-linear stress distribution of a buckled plate. B and D
factors were determined by Clark and Rolf (1966) simplifying the inelastic plate buckling
equation proposed by Stowell (1948) based on experimental studies.
The cut-off in the yielding range was set against the yield stress without a
scientific basis. It can be assumed that the rather arbitrarily chosen cut-off for column
members to prevent a catastrophic failure after yielding was also employed for the
component elements according to Alcoa (1955). Due to the inelastic reserve capacity,
however, component elements of a flexural member are not as critical as the ones of a
column after yielding. For this reason, the cut-off equations for compressive component
elements are to be investigated.

The ultimate cut-off and shape factor for the ultimate and plastic
flexural capacity
Rigorous analytic ultimate shape factor for rectangular web elements
The ultimate moment capacity of a member can be obtained if ultimate capacities
of component elements are known. For example, in the case of a doubly symmetric Ishaped section, the ultimate capacity of the flange element is the ultimate stress of the
used material, obviously. On the other hand, the ultimate capacity of the web is the
ultimate bending capacity. Since the stress distribution on the web is non-linear, the
moment capacity can be obtained through integration of the stress-distribution.
For the integration, the stress-strain relation is curve-fitted by the modified
Ramberg-Osgood equation:
f
f
= + 0.002
Fy
E

Paper 1 - 3

(3)

Rev. A (Progress Report 15)

5/6/2003

In order to use this equation, exponent n should be determined. Since the AA

Specification provides the yield stress, Youngs modulus and the ultimate stress, the
strain at the ultimate stress (referred to as the ultimate strain hereinafter, u) should be
obtained to determine exponent n as expressed in Equation (4).
log 500 u FEu

(4)
n=

F
u

log F

However, the statistical data of the ultimate strain is not available. Instead, the minimum
percent elongation is only available in the Aluminum Standards and Data (2000c),
abbreviated to ASD. Since the percent elongation is somewhat larger than the ultimate
strain, in general, the ultimate strain may be assumed from the percent elongation.
As Eberwien et al. (2001) showed, a moment capacity for a rectangular block
with non-linear stress distribution expressed by the modified Ramberg-Osgood equation
can be obtained using the Bernoulli-hypothesis. Based on the idea, the ultimate moment
capacity is computed and normalized by the yield moment capacity, resulting in a closed
form solution of the ultimate shape factor for a web element of a doubly symmetric
section:
2n
n
2
bh 2 Fu 1 Fu
1 n Fu
1 n + 1 Fu Fu
(5)
Mu =
+

2 u2 3 E 5002 2n + 1 Fy
500 n + 2 Fy E

2
bh Fy
My =
(6)
6
2n
n
2
Mu
3 Fu 1 Fu
1 n Fu
1 n + 1 Fu Fu
(7)
w =
= +

500 n + 2 Fy E
M y u2 Fy 3 E 5002 2n + 1 Fy

As seen in DOD (1994) and tensile coupon tests in this study, the typical value of
the ultimate strain (u) is approximately 6 to 8% for 6061-T6. For the same material, the
typical value of percent elongation is 12% as tabulated in the Aluminum Design Manual,
ADM in short, (2000b). Similarly, since the minimum value of the percent elongation is
8% for 6061-T6 as shown in ASD (2000c), the minimum value of the ultimate strain may
be approximately 4 to 6%. To understand the effect of the ultimate strain variation, shape
factors are computed for various ultimate strain values using Equation (7) in Figure 2.
From the figure, it is found that the shape factor variation is less than 1% if the ultimate
strain is set to either 50% larger or smaller than the minimum percent elongation.

Fig. 2. Variation of the shape factor for a rectangular web element with respect to the
ultimate strain (6061-T6)

Paper 1 - 4

Rev. A (Progress Report 15)

5/6/2003

Thus, the conjecture that the ultimate strain may be assumed from the percent
elongation is reasonable. In this study, the recommended value of the ultimate strain is
between the half and the same amount of the minimum percent elongation. The ultimate
shape factors are computed using Equation (7) for some 6000 series alloys in Table 3,
which are more frequently used for extrusions.
The obtained shape factors in Table 3 can be directly used as the proposed cut-off
for the limit state of the ultimate stress when those are multiplied by the yield stress.
Since the compressive stress distribution is approximately the same as the tension one for
doubly symmetric compact sections, the shape factors (w) in this study are also
compared to those for the tension web available in AA: Equation (8). It can be concluded
that the shape factor in AA (2000a) is quite close to the rigorous one proposed in this
study for 6000 series alloys.
Table 3. Material properties and ultimate shape factors for rectangular sections (webs) for
some 6000 series alloys (FIX DATA WITH 10000KSI E)

Alloy-temper

Fy a
(MPa)

Ea
(MPa)

Fu a
(MPa)

u b

nc

w d

6005-T5

241.15

69589

261.82

0.04

35.23

1.5976

1.0363

6061-T6, T6510, T6511

1.42 Ftu Fy

241.15

69589

261.82

0.04

35.23

1.5976

1.0363

6063-T5

110.24

69589

151.58

0.04

9.23

1.9468

0.9971

6063-T5

103.35

69589

144.69

0.04

8.74

1.9768

0.9944

6063-T6, T62

172.25

69589

206.70

0.04

16.01

1.7362

1.0189

6066-T6, T6510, T6511

310.05

69589

344.50

0.04

27.18

1.6227

1.0284

6070-T6, T62

310.05

69589

330.72

0.03

39.29

1.562

1.0313

6105-T5

241.15

69589

261.82

0.04

35.23

1.5976

1.0363

6351-T5

241.15

69589

261.82

0.04

35.23

1.5976

1.0363

6463-T6

172.25

69589

206.70

0.04

16.01

1.7362

1.0189

a.
b.
c.
d.
e.
f.
g.

Minimum values from ADM (2000b) and ASD (2000c)

50% of the minimum percent elongation listed in ASD (2000c)
Equation (4)
Equation (7)
1.42Ftu/Fy is the ultimate shape factor for tension side web by AA: Table 1
Test coupon diameter or thickness up through 12.7mm
Test coupon diameter or thickness between 12.7 and 25.4mm

Simplified ultimate shape factor for rectangular web elements

The rigorous analytic expression for the ultimate shape factor of a rectangular
web element in Equation (7) is too complicated for practical design purposes. Thus, a
simplified expression such as the one used in the AA Specification (2000a) for tension
component elements, Equation (8), is necessary.
F
u = w = 1.42 tu
(8)
Fty
Although the performance of Equation (8) is found to be satisfactory for some
6000 series alloys in Table 3, it has never been investigated for a wide variety of alloystempers. In ASD (2000c), more than a thousand of materials are listed according to types
of alloys, tempers, dimensions of tested specimens, orientations of tested coupons taken

Paper 1 - 5

Rev. A (Progress Report 15)

5/6/2003

from specimens and product types of specimens (plates, pipes or extruded shapes).
Among these data, 986 alloy-temper combinations are chosen for the investigation. The
bases to choose the data are as follows: First, minimum material properties required for
Equation (7) should be available. Second, the ultimate strain should exceed 1.5%, when it
is assumed to be the half of the minimum percent elongation. Third, the orientation of the
test coupon should be longitudinal or not specified.
5

Equation (7)

4.5

Equation (8)

w = 1.42

4
3.5

Ftu
Fty

(current AA)

Equation (9)

2.5

w = 1.25

Ftu
+ 0.2
Fty

(proposed)
1.5
1
0.5

alloy-temper = no transverse direction,

u = 0.5 x percent-elongation
total number of data = 986
1

1.5

2.5

u > 1.5%
3.5

Ftu Fty
Fig. 3. Comparison of the ultimate shape factor approximations using Equations (8) and (9)
for a plate under bending

For the selected alloy-temper combinations, the shape factors are computed using
Equation (7) and plotted with respect to the Ftu/Fty ratio in Figure 3. From the figure, it is
clear that Equation (8) becomes unconservative as the Ftu/Fty ratio increases. Thus, a
more precise curve-fit is proposed for the ultimate shape factor:
F
u = w = 1.25 tu + 0.2
(9)
Fty
For practical design purposes, Equation (9) could be used instead of Equation (7)
for the ultimate shape factor of a rectangular web element under bending.

Parametric study for component elements and proposed AA

design equations
In order to show a possible modification to the design equation for the yielding
range, a parametric study is conducted for component elements of doubly symmetric Ishaped sections using the finite element method.
The finite element program, ABAQUS developed by Hibbitt, Karlsson and
Sorensen, Inc. is used for the analyses. Four-noded general-purpose shell elements with 4
integration points are used to take into account the large thickness variation in a series of
Paper 1 - 6

Rev. A (Progress Report 15)

5/6/2003

models. The alloy and temper is assumed to be extruded 6061-T6 with minimum material
properties in Table 3 except for the ultimate strain. To observe the variation of the
analysis results, two ultimate strain values are used; 4% and 8%. According to the
variation of the ultimate strain, the Ramberg-Osgood exponent (n) and the shape factor
for the web (w) vary.
For material model, isotropic hardening is used, since this study deals with
monotonic loading. In this case, once a stress reaches the ultimate stress, the stress
remains constant as the plastic strain exceeds the ultimate strain. Due to uncertainty after
this stage, it is assumed that the whole member reaches the failure when the von-Mises
stress at a single point of a member reaches the ultimate stress. This occurs when the
member is too compact to buckle. On the other hand, the failure of the member can also
be initiated by buckling when the member is less compact. In this case, the peak load of
the member is obtained before any point of the member reaches the ultimate stress. In this
study, these two possibilities of failure are considered simultaneously to find an ultimate
load factor.

Fig. 4. Boundary and loading conditions for finite element analyses

Fig. 5. Parametric study results for component elements (a) web (b) flange

Paper 1 - 7

Rev. A (Progress Report 15)

5/6/2003

The boundary conditions for component elements of doubly symmetric I-shaped

sections are idealized as shown in Figure 4. To avoid singularity of the stiffness matrix,
one longitudinal degree of freedom is restrained at the span center. Equal and opposite
loadings are applied at the loaded edges, at which rigid beam elements are attached.
Prior to the non-linear analyses, elastic eigen-value analyses are conducted to
generate initial geometric imperfections. The maximum amplitude of imperfections is
determined based on the industrial production limitation provided by ASD (2000c).
The results of the parametric study for component elements are shown in Figure
5. For most ranges, the finite element analyses are close to the current AA Specification.
However, for yielding range (width/thickness S1), significant difference is detected. For
very low width-to-thickness range, the ultimate shape factors are employed for the
proposed cut-off: 1.086 (= Fu/Fy) for the flange element, 1.598 for the web element (if u
= 4% is used). For the remaining part of the yielding range between the proposed and
current cut-off values, it is found that a linear extension of the inelastic buckling range
(S1 width/thickness S2) equation can be used. These modifications work well
compared to both series of finite element analyses based on 4% and 8% of the ultimate
strain.
The AA Specification equations modified in this study will be used as the limit
state of the ultimate stress, while the current ones as the limit state of the yield stress. The
symbolic design equations of component elements are summarized for the two limit
states in Table 4.
Table 4. Design equations and borders of equations
limit state stress
limit state stress
limit state stress
limits
limits
component
b
b
b
limit state of
S1 S2
S2
S1
S1
S2
element
t
t
t
B Fy
Flange
Ff = Fy
yield stress
D
(current AA)
B 1.3Fy
Web
Fw = 1.3 Fy
k BE
b
k1 B
F= 2
D
F = B D
b
t

D
B Fu
Flange
Ff =Fu
t
D
ultimate stress
(proposed AA)
B w Fy
Web
Fw = w Fy
D
Note. 1. F = Ff or Fw
2. B,D, k1 and k2 for a flange differ from those for a web. See AA (2000a) for details.
3. Equation (1) for (Table 2)
4. Equation (7) for w

The moment capacity evaluation approaches

In the AA specification, to compute the moment capacity of a member the limit
state stress of each component element is multiplied by the elastic section modulus of the
entire cross section. Among obtained moment capacities, the minimum one is chosen as
the allowable moment capacity of the member. This is denoted MMCA (the Minimum
Moment Capacity Approach) hereinafter. Since possible interactions and stress
redistributions between component elements are disregarded, this approach is expected to
be rather conservative.

Paper 1 - 8

Rev. A (Progress Report 15)

5/6/2003

As an alternative in AA, limit state stresses obtained from all component elements
can be averaged according to contributory area. The averaged stress is multiplied by the
section modulus to compute the moment capacity. This is called weighted average stress
approach (denoted WASA).
The weighted average stress equation was first introduced by Jombock and Clark
(1968) to compute the crippling strength of aluminum trapezoidal formed sheet members:
Equation (10),
F A +F 1 A
Fwt = f f 1 w 6 w
(10)
Af + 6 Aw
where Ff = limit state stress for the flange, Fw = limit state stress for the web, Af = entire
compression side flange area, and Aw = entire web area.
Although the WASA was verified through experiments by the researchers, a
theoretical basis of the weighted average method has never been investigated. Thus, the
accuracy of the method is questionable for various geometric shapes other than the
formed sheet members.
The theoretical basis of the WASA is investigated for a doubly symmetric Isection shown in Figure 6. Multiplication of both the denominator and numerator of
Equation (10) by (hc/2)2 and simplification assuming that the flange thickness is
relatively small result in Equation (11):
M + M w hc M u hc
Fwt = f
(11)
=

I f + Iw 2
I 2
where Mf = moment capacity of the flanges, Mw = moment capacity of the web, If =
moment of inertia of the flange elements, Iw = moment of inertia of the web element, Mu
= total moment capacity and I = total moment of inertia.

Fig. 6. Contributions made by component elements to the moment capacity of an -section

Equation (11) impies that the weighted average stress is an approximatly

linearized bending stress mesured at the mid-thickness of the flange. Therefore, to obtain
an appropriate total moment capacity, the corrections shown in Table 5 have to be made:
The correction in Table 5 is insignificant when the flange thickness is relatively
small, such as the ones of the thin-walled formed sheet members used for experimental
evidences by Jombock and Clark (1968). However, it is known that extrusion is more
economical than cold-forming, in general. In addition, sections consisting of component
elements with large width-to-thickness ratio are not suitable for extrusion due to
production difficulty: For details, see Kissell et al. (1995). For these reasons, standard
sections listed in the ADM (2000b) are mostly made of relatively thick component

Paper 1 - 9

Rev. A (Progress Report 15)

5/6/2003

elements, falling into the yielding or, at least, inelastic buckling ranges. Thus, the
modifications shown in Table 5 as well as Table 4 are significant for extruded sections.
Table 5. Correction in the current WASA
current WASA
proposed WASA
(WASA)
(WASA2)
h
Mu = (Fwt)(S)
hc
Note. S = section modulus = I /(h/2)

Mu = (Fwt)(S)

As an alternative to the proposed WASA, the Total Moment Capacity Approach

(TMCA) is also suggested in this study. In the TMCA, the limit state stress from each
component element is multiplied by the contributory section modulus to compute the
contributory moment capacity. All the moment capacities from component elements are
added to obtain a member moment capacity. For example, the moment capacity of an Ishaped section shown in Figure 6 can be expressed as Equation (12) based on the TMCA.
M u = Ff Af hc + 16 Fw Aw ho
(12)
Since contributions by all component elements are made to a member capacity in both the
WASA2 and TMCA, these are expected to be more accurate than MMCA.

Parametric Study of I-Shaped Sections

To validate the improvements discussed above, a parametric study is conducted
for doubly symmetric I-shaped sections using the finite element analyses. The width of
the entire flange ( w in Figure 6) and the depth between centerlines of flanges ( hc in
Figure 6) are maintained as 254 mm, while uniform component element thicknesses vary
so that wide ranges of width-to-thickness variations of component elements can cover
most of the standard sections listed in the ADM (2000b). The length of the members is
fixed as 2540 mm. The boundary conditions are determined as shown in Figure 7. Other
details regarding the finite element models are the same as those for the parametric study
for component elements.

Fig. 7. Model geometry of an I-shaped section for a parametric study

As a first step, the influence of the ultimate-plastic capacity consideration in the

AA Specification is investigated as shown in Figure 8. The approach abbreviations used
in Figure 8 are denoted as follows: For example, MAA-U-WASA is the moment capacity
evaluated by AA using the conventional WASA, considering the limit state of the
Paper 1 - 10

Rev. A (Progress Report 15)

5/6/2003

Ultimate stress, while MAA-Y-WASA is the moment capacity considering the limit state of
the Yield stress. The horizontal axis is the slenderness factor. The vertical axis is the
moment capacity obtained from the finite element analysis normalized by the moment
capacities using the approaches specified in the graph.
As seen in this figure, an approximately 7% improvement in the member capacity
is observed by the consideration of the ultimate-plastic capacity. In addition, the variation
of the data significantly decreases when the ultimate-plastic capacity is considered.

/M

approaches

1.3
1.2
1.1

FEM

M FEM / M approaches
M FEM / M A A -Y -WA SA

mean

c.o.v

0.9

1.209

0.056

0.8

M FEM / M A A -U-WA SA

1.141

0.029

0.3

0.4

0.5

0.6

0.7

0.8

= Fy Fcr
Fig. 8. Influence of the ultimate-plastic capacity consideration in AA (u = 4%)

FEM

/M

approaches

1.3
1.2
1.1
1

mean

c.o.v

0.9

M FEM / M approaches
M FEM / M A A -U-WA SA

1.141

0.029

0.8

M FEM / M A A -U-WA SA 2

1.04

0.024

0.3

0.4

0.5

0.6

0.7

0.8

= Fy Fcr
Fig. 9. Influence of the modification in WASA (u = 4%)

Comparison of models obtained by WASA and WASA2 defined in Table 5 is

shown in Figure 9. The AA Specification approach is improved approximately by 10% of
the member capacity due to the modification of WASA. In addition, the difference
between the AA specification and the finite element analysis is significantly decreased.
In Figure 10, the currently available two approaches in AA (2000a, MMCA and
WASA) are compared to the ones developed in this study (WASA2 and TMCA). As seen
in the figure, a significant improvement is made for a wide range of slenderness. Since
the difference between the WASA2 and TMCA is insignificant, either approach could be
Paper 1 - 11

Rev. A (Progress Report 15)

5/6/2003

employed. The moment capacities obtained from all four approaches as well as the finite
element analyses are listed in Table 6.
Table 6. Parametric study results
Fy
MAA-Y-MMCA MAA-Y-WASA MAA-U-WASA2 MAA-U-TMCA MFEM-4% MFEM-8%
=
My
My
My
My
My
My
Fcr

tw (mm)

tf (mm)

25.400

42.342

1.000

1.023

1.309

1.305

1.375

1.378

0.233

25.400

31.750

1.000

1.031

1.277

1.279

1.347

1.340

0.289

25.400

21.158

1.000

1.047

1.223

1.229

1.290

1.280

0.384

25.400

15.875

0.971

1.037

1.157

1.163

1.240

1.230

0.473

25.400

12.700

0.892

0.990

1.105

1.111

1.200

1.190

0.566

12.700

42.342

1.000

1.012

1.289

1.281

1.314

1.305

0.259

12.700

31.750

1.000

1.017

1.251

1.249

1.280

1.270

0.335

12.700

21.158

1.000

1.025

1.171

1.172

1.210

1.200

0.469

12.700

15.875

0.954

0.992

1.085

1.087

1.150

1.140

0.581

12.700

12.700

0.871

0.929

1.013

1.014

1.090

1.090

0.679

8.458

42.342

1.000

1.008

1.282

1.272

1.290

1.275

0.268

8.458

31.750

1.000

1.011

1.241

1.238

1.256

1.240

0.351

8.458

21.158

1.000

1.017

1.151

1.151

1.180

1.170

0.506

8.458

15.875

0.948

0.975

1.056

1.057

1.100

1.100

0.644

8.458

12.700

0.863

0.905

0.975

0.975

1.020

1.030

0.767

6.350

42.342

1.000

1.006

1.276

1.267

1.276

1.260

0.349

6.350

31.750

1.000

1.009

1.233

1.230

1.240

1.220

0.358

6.350

21.158

1.000

1.013

1.136

1.136

1.170

1.150

0.524

6.350

15.875

0.945

0.966

1.034

1.035

1.080

1.080

0.677

6.350

12.700

0.860

0.892

0.946

0.947

0.998

1.000

0.817

1.5
M TEST / M approaches

1.4

FEM

/M

approaches

1.3
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5

0.3

M FEM / M approaches

mean

c.o.v

M FEM / M A A -Y -MMCA
M FEM / M A A -Y -WA SA

1.248

0.055

1.209

0.056

M FEM / M A A -U-WA SA 2
M FEM / M A A -U-TMCA

1.04

0.024

1.04

0.021

0.4

0.5

0.6

0.7

0.8

= Fy Fcr
Fig. 10. Comparison between the current and proposed approaches (u = 4%)

All the finite element computations in Figures 8 to 10 are made when the ultimate
strain is 4%. The percent symbol in the subscript of FEM analyses in Table 6, 4% or 8%,
Paper 1 - 12

Rev. A (Progress Report 15)

5/6/2003

represents the ultimate strain value used. When the ultimate strain is 8%, the results do
not change significantly as compared in Table 6.

Experiments and FEM Simulation

To support the approaches developed in this study, physical tests are conducted
for three doubly symmetric AA standard I-shaped sections; I-3x1.64. The alloy and
temper of the specimens is 6063-T6, of which the minimum material properties are listed
in Table 3. Since the study is based on the strength of component elements, continuous
lateral supports are required. However, such supports are virtually impossible for
practical tests. For this reason, a parametric study is conducted using the finite element
method to find an appropriate lateral support spacing so that the ultimate load factor and
the corresponding displacement are similar to those of continuous one: The study results
in 304.8 mm (12 in.) for the lateral support spacing. The determined test setup is shown
in Figure 11 together with the dimensions for the tested specimens.

Fig. 11. (a) Dimensions of section -3x1.64 (b) schematic test setup side-view
All dimensions are in mm and not to scale

The residual deformation is shown in Figure 12 when the specimens were

removed from the test frame. A single ripple is formed near the span center in each
specimen.
The bending test setup is simulated numerically using the finite element method.
Bi-linear spring elements are attached between spreader plates and the specimen so that
only compression can be transferred. This is to simulate the contact behavior of the actual
test setup, in which the spreaders were simply placed on the specimen without any
moment connections such as welding and bolting.
Two types of elements are used: Four noded linear shell elements with reduced
integration are used for the SHELL model, while twenty noded quadratic hexahedral
solid elements with reduced integration for the SOLID model. The SOLID model is
composed of two layers of solid elements through the thickness.
Initial geometric imperfections are generated using elastic eigen-value analyses
with a maximum amplitude based on either the actual measurement in this study (0.048
mm, model FEM 1) or the industrial production limitations (0.127 mm, model FEM 2 to
3) by ASD (2000c). In order to fairly compare the finite element simulations with the
tests, the median of five tensile coupon test results obtained from one of the specimens is
introduced into the finite element analyses.

Paper 1 - 13

Rev. A (Progress Report 15)

5/6/2003

Fig. 12. Residual deformation of the tested specimens

Fig. 13. Deformed shape near failure using finite element method simulation (SOLID)
1.4

1.2

LF = Mu /My

0.8

0.6

DESCRIPTION IMP MAX-LF DISP-XL MODEL

TEST 1
1.2
109.2
TEST 2
1.275
120.7
TEST 3
1.301
127.7
FEM 1 0.048 1.238
121.2
SHELL
FEM 2 0.127 1.218
101.9
SHELL
FEM 3 0.127 1.204
131.1
SOLID
DISP-XL = Displacement at the maximum LF

0.4

0.2

50
100
SCVD (Span Center Vertical Displacement, mm)

150

Fig. 14. Load factor-displacement result comparison for -3x1.64

A deformed shape near the failure from one of the finite element simulations
exampled in Figure 13 is similar to the ones from the physical tests shown in Figure 12.
The load factor-displacement curves obtained from both physical tests and the finite
element simulations show close agreements each other as plotted in Figure 14. The
variation in the test results would be mainly due to the variations of the material
properties as well as initial geometric imperfections.

Paper 1 - 14

Rev. A (Progress Report 15)

5/6/2003

The average of the maximum load factors obtained from the physical tests is
compared to the current AA approaches (MMCA and WASA) and those developed in
this study (WASA2 and TMCA) in a dashed oval of Figure 10. The test results follow the
trend of the parametric study, which supports the validation of the proposed approaches
in this study.

Application to the AA Specification

It is shown in this study that the limit state stress based on the ultimate-plastic
capacity works well with both the finite element simulations and physical tests. However,
it is desirable to maintain a certain factor of safety against yielding in actual structural
designs.
As summarized in Table 7, the AA Specification (2000a) allows choosing the
minimum of the allowable stresses based on the limit states of the yield and ultimate
stresses for tension component elements. However, there is only one allowable stress
based on the limit state of the yield stress for compression component elements.
Procedure I, which is one of the proposed approaches shown in Table 7, is almost
the same as the current AA Specification except that allowable stresses for both limit
states are available not only for the tension but also for compression component elements
due to the development in this study. In this approach, the same factor of safety against
yielding is maintained as the AA Specification, which sets a fixed number, 1.65 for
building type structures. However, it does not seem to be reasonable to employ the
uniform safety factor regardless of the margin between the yield and ultimate stresses for
a wide variety of alloy-temper combinations.
Thus, an alternative approach to compute the allowable stress is proposed for an
additional inelastic reserve capacity: Procedure II in Table 7. In this approach, 25% of the
margin between the allowable stresses based on the yield and ultimate limit states is
added to the allowable stress based on the yield limit state. For this reason, the safety
factor against yielding varies depending on the ultimate to yield stress ratio. The safety
factor against yielding can be defined as
F
n y = y y
(13)
Fa
Table 7. Allowable stress in the AA Specification and the proposed approaches

approaches

allowable stress for tension

component element

allowable stress for compression

component element

current AA
Specification

Fa = min ( Fay , Fau )

Fa = Fay
Procedure I

proposed
approaches

Fa = min ( Fay , Fau )

(14)

Procedure II

Fa = Fay + 0.25 ( Fau Fay ) min (1.25Fay , Fau )

Note: Fa = the (final) allowable stress

Fay = the allowable stress based on the yield limit state
Fau = the allowable stress based on the ultimate limit state
See Table 8 for details.

Paper 1 - 15

(15)

Rev. A (Progress Report 15)

5/6/2003

Table 8. Details of allowable stress equations for (a) tension element (b) compression
element and (c) shape factors
(a)
AA Section

allowable stresses

Fay = y Fty n y

3.4.2
3.4.4

allowable stresses
AA
b t S1
Section

Fay =

3.4.15
3.4.16
3.4.18

Fau =

Fau = u Fty

(b)
allowable stresses
S1 b t S2

limit
S1

y Fcy

B y Fcy

ny

D
n
B n u Fcy

u Fcy

nu

( kt nu )

Fay =

1
ny

B D

Fau =

1
ny

B D
t

limit allowable stresses

S2 b t
S2

k1 B
k BE
Fay = Fau = 2
b
D
n y
t

(c)
AA Section

yield shape factor

ultimate shape factor

3.4.2

y = 1.0

u = Ftu Fty

3.4.4

y = 1.3

u = 1.25 Ftu Fty + 0.2

3.4.15, 3.4.16

y = 1.0

u = Ftu Fcy

3.4.18

y = 1.3

u = 1.25 Ftu Fcy + 0.2

b
b

Note a. In the AA Specification, u = 1.42 Ftu Fty .

b. Not available in the AA Specification
c. For other coefficients, see the AA Specification
2.2

2.2

k t = 1.00
k t = 1.10
k t = 1.25

2
1.8

2
1.8

~
ny = 1.65

1.6

1.6

n y

~
ny = 1.32

1.4

1.4

1.2

1.2

~
ny = 1.00

1
0.8
0.6

~
n
y
1

1.5

mean stdev

min

max

1.54

0.129

1.32

2.01

2.5

Ftu Fty

3.5

(a)

1
0.8

0.6

200

400

numbers of alloy-temper
(b)

Fig. 15. Safety factor of the tensile allowable stress (AA 3.4.4) for a plate under bending

Paper 1 - 16

Rev. A (Progress Report 15)

5/6/2003

In Figure 15a, the safety factors against yielding of a tension side rectangular web
element under bending are computed and plotted using solid circles for the previously
selected 986 alloy-temper combinations when Procedure II is used (AA Section 3.4.4). It
is found that the average safety factor (1.54) is only 6.7% smaller than the current fixed
safety factor (1.65). Solid and dashed curves represent the analytic expressions of the
safety factors for the materials with different notch sensitivities (kt). It is also noted that
the majority of the alloy-temper combinations are concentrated near the average as shown
in Figure 15b. In addition, the minimum safety factor is set to 1.32.

Conclusions
The cut-off based on the yield stress in the AA Specification is replaced with the
ultimate shape factor developed in this study, which is eventually simplified for practical
design purposes. With the shape factor, the compressive limit state stress of a component
element can be evaluated more precisely. In addition, the modification is more consistent
with the specification approach for tension elements.
Secondly, the empirically developed weighted average stress approach (WASA)
is investigated. From the study, it is found that the weighted average stress equation is an
approximation of a linearized ultimate bending stress at the centroid of the flange, while
the current specification equation is considered to be the bending stress at the extreme
fiber; a simple modification is proposed to the WASA.
As a result of the study, moment capacities of laterally supported flexural
members are evaluated more accurately, when a series of parametric study is conducted
for a wide range of slenderness using the finite element method. The study is also further
validated by the flexural tests of standard sections.
In addition, two procedures are suggested for the final form of the future
specification. In these approaches, a certain factor of safety against yielding is
maintained. Procedure II is more reasonable, since the amount of additional inelastic
reserve capacity used in the approach depends on the margin between the ultimate and
yield stresses. In the proposed approaches, the current frame of the specification is
maintained.

Acknowledgement
This study is sponsored by the U.S. Department of Energy and the Aluminum
Association. Their sponsorship is gratefully acknowledged.

References
The Aluminum Association (2000a). The Specification for Aluminum Structures.
The Aluminum Association (2000b). The Aluminum Design Manual.
The Aluminum Association (2000c). Aluminum Standards and Data.
Aluminum Company of America (Alcoa) (1955). Alcoa Structural Handbook, A Design
Manual for Aluminum.
Clark, J.W., Rolf, R.L. (1966). Buckling of Aluminum Columns, Plates, and Beams.
Journal of the Structural Division, ASCE, Vol. 92, Proc. Paper 4838.
Eberwien, U., Valtinat, G. (2001). The fullness method: A direct procedure for
calculation of the bending moment of a symmetrical aluminum cross section. The

Paper 1 - 17

Rev. A (Progress Report 15)

5/6/2003

8th International Conference in Aluminum (INALCO) in Munich, Germany, March

28-30
Jombock, J.R., Clark, J.W. (1968). Bending Strength of Aluminum Formed Sheet
Members. J. of the Structural Div., ASCE, Vol.94, No. ST2, Proc. Paper 5816.
Kissell, J.R., Ferry, R.L. (1995), Aluminum structures, John Wiley & Sons, Inc.
Stowell E.Z. (1948). A Unified Theory of Plastic Buckling of Columns and Plates.
NACA Technical Note 1556, National Advisory Committee for Aeronautics.
U.S. Dept. of Defense (DOD) (1994), Metallic Materials and Elements for Aerospace
Vehicle Structures. Military Handbook Vol. 1.

Nomenclature
f = the ultimate shape factor for flange
u = the ultimate shape factor
w = the ultimate shape factor for web
y = the yield shape factor
w8% = the ultimate shape factor for web when the ultimate strain is 8%
= variable strain
u = strain at the ultimate stress = the ultimate strain
= constant for equivalent slenderness ratio depending on the plate buckling coefficient
= slenderness factor = (Fy/Fcr)0.5
p = equivalent slenderness ratio
= Poissons ratio
Af = entire compression side flange area
Aw = entire web area
B = buckling formula constant intersecting vertical axis for zero width-to-thickness ratio
b = the width of a plate element or a flange element
D = buckling formula constant for slope of the inelastic buckling range
E = Youngs modulus
f = variable stress
F = limit state stress for the flange or web
Fa = the (final) allowable stress
Fau = the allowable stress based on the ultimate limit state
Fay = the allowable stress based on the yield limit state
Fcr = minimum buckling stress
Ff = limit state stress for the flange
Fp = limit state stress of a component plate element
Ftu = tensile ultimate stress
Fty = tensile yield stress
Fu = the ultimate stress
Fw = limit state stress for the web
Fy = the yield stress
h = depth of the web element or the entire depth of an I-shaped section
hc = depth of an I-shaped section between centerlines of flanges
ho = depth of an I-shaped section between inner surfaces of flanges
I = total moment of inertia

Paper 1 - 18

Rev. A (Progress Report 15)

5/6/2003

If = moment of inertia of the flange elements

Iw = moment of inertia of the web element
k1 = a coefficient to determine slenderness limit S2
k2 = a coefficient to determine limit state stress when width-to-thickness is larger than S2
kt = a notch sensitivity factor
kp = plate buckling coefficient
LF = load factor = Mu/My or Pu/Py
Mf = moment capacity of the flanges
Mn = moment capacity obtained from each approach
Mu = total moment capacity or moment capacity from FEM or test
Mw = moment capacity of the web
n = an exponent used for Ramberg-Osgood equation
nu = factor of safety on ultimate stress
ny = factor of safety on yield
n y = variable factor of safety on yield
S = section modulus
S0 = width-to-thickness distinguishing proposed yielding and inelastic buckling range
S1 = width-to-thickness distinguishing yielding and inelastic buckling range
S2 = width-to-thickness distinguishing inelastic and elastic buckling range
t = thickness of a plate element
tf = flange thickness
tw = web thickness
w = width of an entire flange of an I-shaped section

Paper 1 - 19