Journal of Constructional Steel Research 67 (2011) 14011410

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

journal homepage: www.elsevier.com/locate/jcsr

Seismically induced cyclic buckling of steel columns including residual-stress

and strain-rate effects
Charles-Philippe Lamarche a, , Robert Tremblay b
a

Department of Civil Engineering, Universit de Sherbrooke, Sherbrooke, QC, Canada

Group for Research in Structural Engineering, Department of Civil, Geological and Mining Engineering, cole Polytechnique de Montral, Montral, QC, Canada

article

info

Article history:
Received 30 April 2010
Accepted 25 October 2010
Keywords:
Buckling
Column
Earthquakes
Initial imperfections
Local buckling
Residual stresses
Strain rate
Yielding

abstract
Compression buckling tests were performed on four full-scale W-shaped column specimens to investigate
the buckling response of columns in multi-storey braced steel frame structures subjected to seismic
strong ground motions. The test protocols included monotonically and cyclically applied concentric and
eccentric axial loading. One test was conducted under dynamic cyclic loading. End moments were applied
on one cyclic test. The columns were W310 129 compact (class 1) sections made with ASTM A992 steel.
Weak axis buckling was studied and the column had an effective slenderness ratio of 48. The response
of the test columns was also examined using numerical simulations based on fibre discretization of
the member cross-section. Column residual stresses and strain rate effects on the material properties
were both characterized and accounted for in the numerical models. The study showed that steel
columns can sustain several cycles of inelastic buckling under seismic induced loading while maintaining
sufficient compressive resistance to support the applied gravity loads. Residual stresses affected the
column response only at the first buckling occurrence with a gradual reduction of the columns tangent
stiffness prior to buckling as well as a reduction of the columns compressive resistance. High strain
rates anticipated during strong earthquakes increased the column buckling and post-buckling strengths.
The cyclic buckling response of steel columns can be predicted adequately when using nonlinear
beamcolumn elements and cross-section fibre discretization provided that residual stresses and strain
rate effects are included in the modelling.
2011 Published by Elsevier Ltd

1. Introduction
Large axial compression loads are expected to develop in
columns of concentrically braced steel frames as a result of
bracing members developing their probable compressive and
tensile resistances when the structure is subjected to seismic
strong ground motions [1]. Similarly, the yielding of link beams
in eccentrically braced steel frames also induces high compression
axial forces in columns. Capacity design requirements have been
introduced in code seismic provisions to ascertain that the columns
are provided with sufficient strength to support their tributary
gravity loads together with the axial loads from the yielding
components of the system [24]. In multi-storey buildings, the
column design axial load at a given level is obtained by summing up
the contribution of all yielding braces or links above the level under
consideration, leading to very high axial loads in the columns that
may considerably affect the overall cost of structures.
In reality, such large seismic axial load peak demands are
expected to occur only a few times and to last for very short

Corresponding author. Tel.: +1 819 821 7704; fax: +1 819 821 7974.
E-mail address: charles-philippe.lamarche@usherbrooke.ca (C.-P. Lamarche).

0143-974X/$ see front matter 2011 Published by Elsevier Ltd

doi:10.1016/j.jcsr.2010.10.008

periods of time during a severe earthquake and the question can

be asked whether steel columns could, under certain conditions,
accommodate limited yielding excursions and/or even buckle
without adverse loss in load carrying capacity. For instance, Newell
and Uang [5] verified that columns can sustain a large cyclic plastic
flexural demand without losing their axial load capacity. Similar
experimental data for columns subjected to variable compression
axial loads exceeding the column compressive resistance do
not exist. Limited preliminary numerical simulations by the
authors [6,7] indicate that current capacity design provisions for
columns could be relaxed to some extent without a detrimental
impact on the structural integrity. Such a relaxation, if permitted,
could lead to substantial savings for new structures. The benefits
could be extended to existing structures that have not been
designed according to recently implemented capacity design
methods and for which column strengthening represents a costly
and challenging operation.
The seismic performance of structures designed for short duration buckling excursions must be carefully evaluated by means
of probabilistic structural collapse assessment studies before such
relaxation is implemented in practice. The methodology developed in the ATC-63 project [8] can be used to evaluate the margin of safety against structural collapse. The application of this

1402

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

Table 1
Buckling test program.
Test

Type

Loading rate

Loading

1
2
3
4

Monotonic
Cyclic
Cyclic
Cyclic

Quasi-static
Quasi-static
Dynamic
Quasi-static

Central
Central
Central
Eccentric

methodology requires a good understanding of the buckling response of columns under constant gravity loads and repeated additional compression due to seismic effects, together with robust
numerical models that can reliably reproduce this behaviour. The
nonlinear beamcolumn element with cross-section fibre discretization available in the OpenSees framework [9,10] has been
used successfully to reproduce the cyclic buckling and tension
yielding response of steel bracing members [1113]. That model
does not include residual stress effects. While residual stresses
have limited effects on brace inelastic cyclic response and can be
neglected if properly dealt with at the macroscopic level, they can
lead to reductions of up to 30% of the compressive strength of structural steel columns made of shapes or built-up steel members [14],
sufficient to make the difference as to whether or not a column
will buckle under seismic loading. Lamarche and Tremblay [6,7]
implemented residual stress effects in the OpenSees model and
validated the implementation based on past compression tests on
steel columns subjected to monotonic loading. Validation was still
needed however for columns subjected to a cyclic buckling demand, including dynamic effects, as expected under seismic loading conditions.
This paper presents a test program that was carried out on
four full-scale W-shaped steel columns that were subjected to
four different displacement protocols producing inelastic buckling.
Ancillary tests were conducted to obtain material properties
under static and dynamic loading. The residual stresses were
also measured using the sectioning method. The response of
the column specimens was reproduced using the OpenSees
nonlinear beam column including residual stresses. The axial
loaddeformation response, member end rotations and strain
demand at critical locations are compared. The strain rate effects
were evaluated using a fibre cross-section analysis program. The
prediction of residual stress effects on column buckling strength
with the OpenSees model are also validated against data from past
tests on I-shaped built-up steel columns.
2. Test program
2.1. Objective and scope
In order to investigate the effects of residual stress and high
strain rates on the pre- and post-buckling compressive cyclic behaviour of steel columns, full-scale centrally and eccentrically
loaded column tests were performed. Four identical class 1 (compact) W310 129 columns made of ASTM A992 steel (Fy =
345 MPa) and 3725 mm tall were tested. The section and height
of the specimens corresponded to typical storey heights encountered in braced steel framed buildings. The four buckling tests
performed are summarized in Table 1. The experimental program
included one monotonic quasi-static centrally loaded buckling test,
one cyclic quasi-static centrally loaded buckling test, one cyclic dynamic centrally loaded buckling test, and, finally, one cyclic quasistatic eccentrically loaded buckling test.
Test 1 was performed according to Technical Memorandum #4
of the Guide to Stability Design Criteria for Metal Structures [15].
It aimed at obtaining the monotonic buckling curve of the W310
129 profile. Test 2 aimed at comparing the cyclic inelastic buckling
curve to the buckling envelope obtained in Test 1. In Test 3, the
same cyclic displacement protocol as in Test 2 was applied but at

Fig. 1. Test set-up: (a) testing apparatus, (b) strain gauges pattern at quarterheight and mid-height, (c) instrumentation. Two rows of potentiometers (four
in total).

a faster rate more representative of a seismic loading history. The

test was performed to investigate the effects of high strain rates
on the cyclic buckling behaviour. Test 4 was performed to investigate the effects of combined axial load and end moments on the
column cyclic buckling behaviour, as such conditions are more representative of the complex load combinations typically encountered in buildings during a seismic event. In all cyclic buckling
tests, i.e., Tests 2, 3 and 4, an initial static load corresponding to
approximately 60% of the nominal column compressive strength
was initially applied on the columns to reproduce gravity load effects. Cyclically applied axial displacements were then applied a
posteriori to simulate the seismic demand on a column in the postbuckling range up to a compressive axial deformation of 20 mm,
corresponding to 0.53% of the column height. In the case of Test 3,
the cyclic displacement protocol was dynamically applied assuming a building with a natural period of T = 1 s, typical of lowto mid-rise concentrically braced frames [12,16,17]. These types of
building represent a vast proportion of the building stock in North
America.
Ancillary tests included four tensile coupon tests to determine
the material properties of the web and flanges, residual stress
measurements, column initial out-of-straightness measurements
in the plane of buckling, and high velocity tensile coupon tests
to quantify the effects of strain rates on the column steel yield
strength.
2.2. Test set-up and instrumentation
The column tests were performed in a 12 MN (2700 kip)
capacity Tension/Compression MTS load frame in the HydroQubec Structural Engineering Laboratory at cole Polytechnique
de Montral. The specimens were mounted between cylindrical
bearings simulating pin-ended conditions for weak-axis buckling
and fix-ended conditions for strong-axis buckling. These two
12 MN capacity hardened steel cylindrical hinges with 250 mm
radii are illustrated in Fig. 1(a). These bearings were designed
so that the centre of the cylinders coincides with the centroid
of the column cross-sections at the column ends. Hence, the
effective length of the column specimens about the weak axis was

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

equal to h = 3725 mm, leading to an effective slenderness ratio

of 48. The instrumentation used comprised (Fig. 1(b) and (c)):
3 potentiometers to measure the out-of-plane displacements at
h/4, h/2, and 3h/4 (h is the specimen height = 3725 mm), 4
potentiometers, one at each corner of the bottom hinge to measure
the axial shortening of the columns, 2 inclinometers to measure
the end-rotations, and 8 strain gauges positioned at h/4 and h/2
(a total of 16 strain gauges), i.e., one strain gauge on either side
(interior and exterior faces) of each of the 4 half flanges positioned
at 25 mm from the flanges tips. The built-in 12 MN capacity load
cell of the test frame was used to measure the applied axial load.
2.3. Ancillary tests
The test program also comprised four tensile tests performed
on coupons taken out of the flanges and web of a virgin piece of
the W310 129 profile. The coupon tests were performed in accordance with Technical Memorandum #7 of the Guide to Stability Design Criteria for Metal Structures [15]. The average measured
modulus of elasticity is E = 203 000 MPa. A summary of the measured mechanical properties from the four tensile coupon tests is
presented in Table 2. In the table, h is the strain value at the onset
of strain-hardening, n is the strain value at the onset of necking,
i.e., when the cross-section area of the coupon starts to decrease
and the load begins to drop, and max is the maximum strain value
reached during the test. Idealized bi-linear models were calibrated
from the tensile coupon test results. These values were determined
using an equal plastic energy criteria between strain values h and
n . These two kinematic strain-hardening ratios b = Esh /E for the
flanges and the web are: b = 0.0055, and b = 0.0036, respectively.
Residual stresses were determined in accordance with Technical Memorandum #6 of the Guide to Stability Design Criteria for
Metal Structures [15]. The residual stress measurement pattern included 31 individual strips that were 305 mm long and 28 mm
wide: 11 in each flange and 9 in the web. The cutting pattern
is presented in Fig. 2(a). Gauge holes were drilled on both sides
of the strips to avoid faulty measurements caused by curvature
of the strips developing after cutting from the profile. Measurements were taken using 150 and 250 mm DEMEC mechanical strain
gauges. The measured values are presented in Fig. 2(a). In the figure, the two sets of residual stress values are in good agreement. As
expected, in the case of hot-rolled profiles, the tips of the flanges

1403

Table 2
Tensile coupon test results.
Test

h (%)

n (%)

max (%)

Fy (MPa)

Fu (MPa)

f1
f2
Average

1.7
1.9
1.8

22
20
21

30
27
29

370
372
371

516
504
510

w1
w2
Average

2.7
2.3
2.5

21
21
21

27
27
27

349
351
350

454
452
453

and the web are in compression. The maximum compression stress

value at the centre of the web is about 40% of the nominal yield
strength value. The maximum compression residual stress value
at the tip of the flanges in compression is about 20% of the nominal
yield strength value.
The initial out-of-straightness profiles presented in Fig. 2(b)
were measured using a Leica TPS400 total station (theodolite).
The initial out-of-straightness was measured in the plane of
buckling at nine points equally spaced along the height of the
columns. When measuring out-of-straightness, the columns were
in their position in the test apparatus while no axial load was
applied. Measurements were taken on each flange, 25 mm from
the flange tips. Each measurement point was surveyed twice.
Therefore, each measurement point in Fig. 2(b) is the average
of four measurements. Positive values in the figure are in the
same direction as positive lateral deflections measured in the tests.
Detailed information about the tensile coupon tests, residual stress
measurements and initial out-of-straightness measurements can
be found in [18].
The effects of high strain rates on the magnitude of the yield
strength Fy were investigated through tension testing of coupons
fabricated from the same steel strips that were used to measure
the residual stresses. All strips tested came from the flanges of
the steel shape. The test results are presented in Fig. 2 along with
empirical curves to predict the ratio between the dynamic and the
(d)
static yield strength (Fy /Fy ). The first prediction curve presented
was proposed by Wakabayashi et al. [19]:
(d)

Fy

Fy

= 1 + m log

| |
0

(d)

(1)

where Fy is the dynamic yield strength for a strain rate , Fy , is the

quasi-static yield strength under a strain rate 0 = 50 106 s1 ,

a
b

Fig. 2. (a) Measured residual stress values; (b) initial out-of-straightness (weak axis).

1404

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

b
b

Fig. 3. Effect of the strain rate on the magnitude of Fy : (a) linear scale; (b) log10
scale.

and m = 0.0043 is a dimensionless empirical parameter. That

expression was used to capture the strain rate effects on the
yield tensile strength of steel bracing members [20,21]. For that
particular set of parameters, Eq. (1) was found to overestimate the
(d)
measured values of Fy . The empirical equation proposed by Pan
et al. [22] was also investigated:
(d)

Fy

Fy

= 1 + Ae(B/Fy )

(2)

where
A = a1 + b1 log( ) + c1 log2 ( )

(3)

B = a2 + b2 log( ) + c2 log ( ).

(4)

In Eqs. (3) and (4), parameters: a1 = 0.0226, b1 = 0.0094, c1 =

0.0011, a2 = 77.7183, b2 = 0.0693, and c2 = 0.5952. In
(d)
the case of Eq. (2), Fy is overestimated for high strain rates.
(d)

An attempt was made to better predict Fy by adjusting the

parameters of Eq. (1) based on a least square fit. The parameters
obtained from the least square fit are: m = 0.00225 and 0 =
200 106 s1 . The goodness of fit using those updated parameters
is clear in Fig. 3.
3. Finite element model
The finite element software OpenSees was used to simulate the
behaviour of columns. In these numerical simulations, the column
cross-section was discretized using 50 fibres: 20 for each flange
and 10 for the web. It is noted that buckling occurred about the
weak-axis and only one row of fibres over the flange thickness
was therefore sufficient. Buckling about the strong axis would
require more than one row of fibres over the flange thickness to
capture the gradual penetration of yielding in the column flanges
upon buckling. In Fig. 4(a), the residual stress pattern adopted in
the model is compared to the averages of the measured values.
Sixth and fifth order least square polynomial fits were used to
determine the stress value to assign to each fibre of the flanges
and web, respectively. A 1.8 MPa constant stress offset had to be
added to the residual stress pattern in the web so that the internal
axial forces and internal moments within the cross-section reached
equilibrium. The finite element model comprised 8 elements using
5 integration points per element. The end conditions were taken
as pinnedpinned. The initial out-of-straightness was taken as
presented in Fig. 2(b). The Steel02 hysteretic material law available

Fig. 4. Residual stress pattern used for the analyses; (b) coupon test result used to
calibrate hysteretic parameters [24].

in OpenSees was used to model the steel. The Steel02 material is

a modified version of the well known GiuffreMenegottoPinto
model [23]. The Steel02 properties used in the models were based
on the tensile test results presented in Table 2, i.e., Fy = 350 MPa
and b = 0.0036 in the web, Fy = 371 MPa and b = 0.0055 in the
flanges and E = 203 000 MPa. The isotropic hardening parameters,
i.e., a1 , a2 , a3 , and a4 , amongst other parameters, were calibrated
based on the cyclic tensile test results presented in Fig. 4(b) [24].
Those parameters were taken as: R0 = 25, cR1 = 0.925, cR2 =
0.15, a1 = a3 = 0.005 and a2 = a4 = 1.0.
4. Buckling test results
4.1. Test 1: monotonic quasi-static test
The monotonic quasi-static buckling test was performed
according to the Technical Memorandum #4 of the Guide to
Stability Design Criteria for Metal Structures [15]. In the elastic
range, the displacement was imposed at a rate such that the
loading rate was equal to 6.89 MPa/min (1 ksi = 6.89 MPa) based
on the elastic properties of the W310 129. Buckling occurred
about the weak axis, as expected. After the buckling had occurred, a
downward displacement was monotonically applied at a constant
rate in the displacement-control mode until the load reduced
to approximately 40% of the ultimate load (60% of the nominal
column compressive strength). At the end of the test local buckling
was observed at the column mid-height. The measured axial force
as a function of the applied axial displacement during the test,
presented in Fig. 5(a), is compared to the OpenSees prediction
with (0 = 0) and without (0 = 0) the inclusion of the initial
stresses. The ultimate load reached during that test was Pu =
5607 kN. The maximum loads reached in the OpenSees models
( =0)
with and without residual stresses are Pu 0
= 5330 kN, and
(0 =0)
Pu
= 5782 kN, corresponding to relative differences of 4.9%
and +3.1% respectively. A similar comparison of the out-of-plane
displacement at column mid-height is presented in Fig. 5(b). In
Fig. 5, the global behaviour of the column modelled with OpenSees
is in excellent agreement with the test data. Residual stresses
resulted in a gradual reduction of the columns tangent stiffness, as

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

Fig. 5. Test 1: (a) Axial load vs. axial displacement; (b) Axial load vs. out-of-plane
displacement at column mid-height.

well as a decrease of the column compressive resistance. In each

plot, the response of both numerical models gradually converges
to the same values after bucking has occurred. This is because the
residual stresses do not greatly affect the behaviour at a given
section once all the material has experienced plastic deformations,
either in tension or compression.

1405

Prior to buckling, the column remained elastic and all

strain measurements were negative (compression) and equal in
amplitude. When buckling initiated, weak-axis bending of the
column started, which influenced the strain gauge readings. At the
load Pu = 5607 kN, the strains on the concave side (intrados)
exceeded the strain at yield in compression whilst the strains
on the convex side (extrados) remained in the elastic range.
As the out-of-plane displacement of the column increased and
the load slowly decreased, the bending moment at mid-height
increased and the strains switched from compression to tension
and eventually exceeded the yield strain in tension at a load P =
4300 kN. It is visible in Fig. 6 that the plastic excursions measured
by the strain gauges in both the compressed or the stretched flange
pairs do not occur simultaneously as would have been expected
theoretically. This is mainly attributed to small variations in the
material inelastic properties between the column flanges and to a
smaller extent because very consistent strain measurement results
were obtained in the elastic range, small geometrical imperfections
in the specimen and/or end conditions. At a load P = 4300 kN,
when yielding was reached in the strain gauge on the convex
(tension) side, local buckling initiated shortly after on the concave
(compression) side of the cross-section at a load P = 3750 kN. This
explains the differences observed between the strains measured on
the interior and exterior sides of both compressed flange tips.
In Fig. 7, the average strain values measured on the interior
and exterior sides of the flange tips are compared to those
obtained from OpenSees using the kinematic strain-hardening
ratios determined from the ancillary tests, i.e., b = 0.0055
and 0.0036 for the flanges and the web, respectively. Fig. 7(a)

Fig. 6. Strains at the flange tips at the column mid-height (Test 1).

Fig. 7. Axial load vs. average strain at h/2 on: (a) the compression side and (b) the tension side. Influence of the strain-hardening parameter b on the axial load vs. average
strain response at h/2 on: (c) the compression side and (d) the tension side.

1406

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

and (b) show the strains at mid-height. The results from the
numerical model begin to diverge from the measured response
at a load of 4300 kN in the post-peak region. This coincides
with initiation of yielding on the convex side and, thereby, the
beginning of plastic rotation in the plastic hinge at the column
mid-height. The differences can be partly explained by the fact
that no imperfections other than initial out-of-straightness in the
plane of buckling were included in the finite element model. More
importantly, strain-hardening plays a key role in the amplitude
of strains in plastic hinge regions, as well as on the length of the
plastic hinge. Therefore, the properties assumed in the numerical
model may also have contributed to the divergence of the strain
results at the column mid height. To illustrate the sensitivity
of the strain prediction to assumptions made for the strainhardening parameter, additional simulations were performed
using parameter b = 0 (no kinematic strain-hardening) and b =
0.01 (higher kinematic strain-hardening). The simulation results
are presented in Fig. 7(b) and (c), where it is clear that strainhardening plays a key role on the amplitude and distribution of the
inelastic demand in the plastic hinge region: a higher kinematic
hardening ratio reducing the peak demand at the column midheight to the level observed by spreading the inelastic demand over
a longer column segment (longer plastic hinge region).

Fig. 8. Test 2 at the end of test: (a) Overall buckled shape, (b) Flange local buckling
at column mid-height.

4.2. Test 2: Quasi-static cyclic test

Test 2 aimed at studying the cyclic inelastic behaviour of
the steel columns. During the test, an initial static axial load of
3000 kN was applied to the column to reproduce the effect of the
gravity loads present during the earthquake. This load corresponds
to 53% of the column nominal axial yield strength, 64% of the
unfactored column axial strength, as determined using CSA S16
Standard [25], and 62% of the nominal compressive strength of
the column determined with the AISC Specification [26]. This high
axial load ratio was selected to represent the critical case of a
column supporting large gravity dead loads. Cyclically applied
displacements were then applied in the displacement-control
mode to simulate the seismic induced demand that is expected
when a braced frame oscillates and the axial load exceeds the
column buckling strength. Fig. 8 presents the specimen after the
test. Buckling occurred about the weak axis and local buckling of
the flanges was observed at the end of the test at the column midheight, as illustrated in Fig. 8(b).
In Fig. 9(a), the axial load as a function of the applied axial displacement during the test is compared to the predictions with the
OpenSees models with (0 = 0) and without (0 = 0) the inclusion of initial stresses. For an unknown reason, the four potentiometers used to measure the axial displacements in Test 2 were
noisy, but otherwise, operated well. This low level noise is visible in Fig. 9(a). The ultimate load reached during that test was
Pu = 5399 kN. The maximum loads reached in the OpenSees mod( =0)
els with and without residual stresses are Pu 0
= 5179 kN and
(0 =0)
Pu
= 5624 kN, respectively corresponding to relative differences of 4.2% and +3.8%. As expected, the maximum load predicted with the model is lower when including initial stresses. In
Fig. 9, despite small discrepancies, the global behaviour of the column as modelled with OpenSees is in excellent agreement with
the test data. At the end of the test, the column could still carry the
gravity load (3000 kN) in spite of the 140 mm out-of-plane permanent deformation (=0.038h).
Fig. 9(c) shows the strain gauge measured on the interior and
exterior sides of the flanges at the four corners of the cross-section.
Strain gauges 5 to 8 were located on the intrados of the buckled
column. The initiation of flange local buckling on this side of the
cross-section is revealed by the deviation between the readings
from strain gauges located on opposite faces of the flanges.

Fig. 9. Test 2: (a) Axial load vs. axial displacement, (b) Axial load vs. out-of-plane
displacement at h/2, and (c) Strains measured on the interior and exterior sides of
the flange tips at h/2.

This occurred after the 2nd (gauges #7 and 8) and 3rd (gauges #5
and 6) loading cycles, at strains of 0.019 and 0.024, respectively,
which approximately corresponds to 1013 times the actual flange
steel yield strain (= 371 MPa/200 000 MPa = 0.00186). Such a
good local buckling response
was expected as the column flange

b/t = 7.72 = 0.32 Fy /E (based on nominal Fy = 345 MPa),

which corresponds to the CSA S16 limit for I-shaped for class 1
sections and to 84% of the limit prescribed in the AISC Specification
for compact I-shaped sections. The imposed axial displacement at
the onset of flange local buckling was equal to 1012 mm. This
corresponds to a column shortening of approximately 0.3% of the
column height, or 1.7 times the column yield axial deformation.
The local buckling response is not included in the OpenSees

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

1407

c
Fig. 10. Comparison between Test 1 and Test 2 (a): Axial-load vs. applied
displacement; (b) Axial load vs. average strain on compression and tension sides
at h/2.

fibre cross-section model. Comparison between the numerical and

experimental results indicates that local buckling in these centrally
loaded column tests did not impact significantly on the column
post-buckling axial strength.
Fig. 10 presents a comparison between Test 1 and Test 2. In
Fig. 10(a), except for a small difference being noted between the
measured peak buckling loads, likely due to differences in the
initial out-of-straightness, and the difference in the unloading
sequence between the two tests, the envelope responses of both
columns are nearly identical. The envelope of the strain responses
from the two tests are also in excellent agreement (Fig. 10(b) and
(c)). The results in Fig. 10 suggest that applying cyclic compressive
loads or imposing cyclic inelastic compressive displacements does
not affect the buckling and post-buckling resistances of steel
columns with compact (class 1) cross-sections.

Fig. 11. (a) Axial load vs. applied displacement in Tests 2 and 3; Strain rates
measured at column mid-height in: (b) Test 3 and (c) Test 2.

4.3. Test 3: dynamic cyclic test

Fig. 12. Axial load vs. average strain on the compression and tension sides at
column mid-height in Tests 2 and 3.

Test 3 was performed to investigate the effect of high

seismically induced strain rates on the cyclic buckling behaviour
of the column. In that test, the same displacement protocol as in
Test 2 was applied but at a faster rate more representative of a
seismic loading history. The measured axial load as a function of
the applied axial displacement during the test was compared to
the results from Test 2 in Fig. 11(a). The ultimate load reached
during Test 3 was Pu = 6004 kN, an increase of 11.2% compared
to Pu = 5399 kN measured in Test 2. This higher buckling strength
was expected due to the increase in steel yield strength resulting
from the higher applied strain rate. The strain rates measured at the
column mid-height in Test 3 are presented in Fig. 11(b) and (c) as
a function of the applied axial displacements. A comparison of the
average strains at the flanges tips in tension and in compression for
Tests 2 and 3 is also presented in Fig. 12. The episodes of high strain
rates that occurred during Test 3 coincide with loading sequences,
when downward axial displacements were applied, as depicted
in Fig. 11 whilst unloading sequences generate much lower stain
rates. The maximum strain rates were recorded during and just
after the first buckling occurrence. The peak strain rates then
gradually decreased in subsequent cycles. This is attributed to the
fact that as column bending becomes more pronounced, the strains
in the plastic hinge become less concentrated as plasticity spreads

away from the centre of the column. In Fig. 12, the compression
strain demand at the column mid-height is also significantly
reduced when higher strain rates are applied. This trend does not
seem to exist on the tension side but one strain gauge on this side
started malfunctioning after the third cycle, at 7500 (11 mm
axial displacement), and formal conclusions cannot be drawn on
this aspect.
As an attempt to model the effect of strain rates at the material
level on the column strength, the Steel02 base code was modified
to account for the shift in the yield envelope at each time step
of the analysis as a function of the measured stain-rate history.
Eq. (1) with the updated parameters: m = 0.00225 and 0 = 200
106 s1 was used in the numerical simulations. The modified
Steel02 material with residual stress and strain rate capabilities
was introduced in the RD-SAP (Rate Dependent Sectional Analysis
Program) sectional analysis tool that was specifically developed
for this study. RD-SAP is a standalone program coded in the
Matlab language that is independent of the OpenSees software. The
program is based on the EulerBernoulli plane section hypothesis
and uses fibre sections. The axial strain and curvature responses
measured in Test 3 were used as input in RD-SAP to compute the
resisting axial load and bending moment histories. The parameters
used to model the steel material were the same as in the

1408

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

b
b

Fig. 14. Test 4. (a): Axial load vs. applied displacement; (b) Axial load vs. out-ofplane displacement at h/2.

Fig. 13. Responses obtained from the RD-SAP models: (a) resisting axial load vs.
applied axial displacement history, (b) resisting moment vs. curvature history.

OpenSees analyses, including the residual stress pattern presented

in Fig. 4(a). In Fig. 13, the response computed with RD-SAP is
compared to the experimental member force results at the column
mid-height for Test 3. The responses presented in Fig. 13 are only
for the three first cycles due to the malfunction of strain gauges on
the tension side under larger imposed deformations. In Fig. 13(a),
the axial load response computed with RD-SAP without taking into
account strain rate effects is lower for the first buckling occurrence.
The computed axial load response taking into account strain rate
effects is very accurately modelled in the first two cycles, i.e., the
pre-buckling cycle and the first buckling cycle. The axial resistance
predicted by RD-SAP exceeds the measured values in the third
cycle due to local flange buckling that initiated at the end of the
second cycle in the test, invalidating the plane section assumption
of the RD-SAP model. In the momentcurvature plots (Fig. 13(b)),
the increase in moment resistance due to strain rate is also well
captured for the first two cycles of Test 3.
The simplified strain rate dependent sectional analysis presented herein shows that the increase in buckling strength that was
observed in Test 3 is mostly the result of the higher strain rates imposed in the plastic hinge region of the column. This effect is clearly
more pronounced at the first occurrence of buckling and reduces in
the post-buckling range. In braced steel frames, this effect has positive consequences as the columns will offer greater compressive
strength. However, the buckling strength of the diagonal bracing
members will also experience similar amplification, increasing the
seismically induced force demand on the columns. Nonlinear time
history analyses of braced frame structures, with models that account for strain rate effects on member strength, will need to be
performed to evaluate the net effect on the inelastic response and
stability of braced steel frames under dynamically applied seismic
loading.
4.4. Test 4: Eccentric quasi-static cyclic test
Test 4 aimed at studying the cyclic inelastic behaviour of steel
columns including end bending moments. The end moments were
induced by applying the axial load with an eccentricity e = 20 mm

in the plane of buckling. As in the previous cyclic tests, an initial

static load of 3000 kN was applied prior to applying the cyclic loading. The eccentricity in the column was established such that the
end moment would be equal to 25% of the plastic resisting moment of the section (M = 0.25Mp ), when the axial load reaches
an ultimate value Pu = 3935 kN as determined from the member strength and stability interaction equation in the CSA S16 standard [25]. In this calculation, a resistance factor, = 1.0, and an
average yield strength Fy = 370 MPa were assumed. The axial
load as a function of the applied axial displacement during the test,
presented in Fig. 14(a), is compared to the OpenSees models with
(0 = 0) and without (0 = 0) the inclusion of initial stresses.
A similar comparison is made between the out-of-plane displacements (Fig. 14(b)).
The ultimate load reached in Test 4 is Pu = 3985 kN, which
compares very well with the CSA S16 prediction of 3935 kN. The
relative difference between the test and the predicted values is
+1.2%. At the ultimate load, the bending moment at the column
mid-height obtained experimentally from the axial load multiplied
by the combined initial out-of-straightness and the measured outof-plane displacement at the mid-height is 197 kNm, corresponding to 24.7% of Mp , assuming Fy = 370 MPa, which is also very close
to the assumed value of 0.25Mp . In Fig. 14, the global behaviour of
the column as modelled with OpenSees is in good agreement with
the test data. The maximum load reached in the OpenSees mod( =0)
els with, and without residual stresses are Pu 0
= 4026 kN, and
(0 =0)
Pu
= 4213 kN, corresponding to relative differences of 1.0%
and +5.7% respectively. As expected, the maximum load including initial stresses in the model is lower than the one where the
residual stresses are not accounted for. As in the previous tests,
the prediction with 0 = 0 is unconservative. After the third cycle,
the response of the model including residual stresses converges to
the result from the numerical model without residual stresses and
both predictions overestimate the post-buckling strength by 11% at
an axial displacement of 25 mm. Fig. 15 presents the axial strains
in the column flanges at h/2. Because of the eccentric loading, the
instants where strains on the compression and tension sides reach
the inelastic range are closer together than in the case of concentric loading. In Fig. 15, the strain gauge readings on the intrados
indicate that local flange buckling occurred before global buckling
as strains on the exterior face of the column flange (gauges #5 and
#7) started to deviate from the respective readings on the opposite flange side (gauges #8 and #6) before the ultimate load was

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

1409

Fig. 15. Measured strains at flange extremities at mid-height (Test 4).

Fig. 16. (a) Residual stress pattern of the 14H202 profile. Model vs. test results [6]: (b) stub column test; (c) centrally loaded column tests: KL/r = 60 and 90.

reached. In Fig. 14, the numerical models overpredicted the column resistance in the post-buckling range. Local buckling effects
are not accounted for in the OpenSees fibre cross-section model.
Further study is needed to investigate if the observed difference
can be attributed to this modelling simplification.
5. Columns made from welded flame-cut steel plates
To further examine the capability to predict the pre- and
post-buckling behaviour of steel columns including initial stress
effects, predictions with the OpenSees model were also validated
against data from tests performed on 14H202 columns at Lehigh
University [27]. The test columns were made of ASTM A572
(grade 50) welded flame-cut steel plates. The experimental study
included residual stress measurements, tensile coupon tests to
determine the base yield stress, stub column tests on short
columns, and pin-ended centrally loaded tests on columns having
slenderness ratios KL/r of 60 and 90. The measured residual
stresses and cross-section dimensions of the 14H202 columns
along with the residual stress pattern used in the finite element
model are presented in Fig. 16(a). The column cross-section was
modelled using 100 fibres, 40 for the flanges and 20 for the
web, and 8 elements were used along the column height, with
4 integration points per element. The end conditions were taken
as pinnedpinned and an initial half-sine wave profile was used
to represent the column out-of-straightness [6]. The simulation
results for the stub column test are presented in Fig. 16(b) for
two cases: with and without residual stresses. The softening effect
upon yielding due to the residual stresses is well predicted by the
model. In Fig. 16(c), the numerical results are compared to the
measurements taken during the two centrally loaded column tests.
Buckling developed about the column weak axis in both tests. The
same failure mode was observed in the simulations and a very good
correlation is achieved between test and numerical results when
including residual stresses in the analyses.

6. Summary and conclusions

A test program was conducted to examine the buckling and
post-buckling responses of a W310 129 column and when
subjected to monotonic and cyclic axial loading, with special
attention to the effects of residual stresses and seismically induced
high strain rates. The column specimens were made of ASTM A992
steel and had an effective slenderness ratio of 48. In the cyclic tests,
a static load equal to approximately 60% of the column nominal
compressive strength was initially applied prior to imposing cyclic
axial displacements in the post-buckling range up to a compressive
axial deformation of 0.53% of the column height. Quasi-static
and dynamic cyclic tests were carried out. End moments were
applied in one of the quasi-static cyclic tests. The test results
were compared to numerical predictions obtained from OpenSees
models built with nonlinear beam column elements and fibre
discretization of the cross-section. Residual stresses were included
in the numerical models. Opensees numerical predictions were
also compared to results from stub column and centrally loaded
column tests performed on built-up I-shaped columns made
from flame-cut steel plates. Predictions of the strain rate effects
on column buckling and post-buckling axial resistances were
validated against test data using a fibre cross-section numerical
model accounting for both strain rate effects on the yield strength
and residual stresses. The main conclusions of the study can be
summarized as follows:

In all column tests described in this article, column inelastic

buckling occurred about the weak axis with the formation of
a plastic hinge at the column mid-height.
The columns subjected to monotonic and quasi-static cyclic axial displacements in the post-buckling range were able to carry
the applied gravity loads up to axial deformations of 0.53% of
the column height and out-of-plane displacements at the column mid-height of 3.7% of the column height. Comparison of
the results from the monotonic and cyclic tests showed that the

1410

C.-P. Lamarche, R. Tremblay / Journal of Constructional Steel Research 67 (2011) 14011410

post-buckling behaviour of the columns studied was not influenced by the imposed inelastic cyclic displacements. These results suggest that steel columns with class 1 (compact) I-shaped
sections may be able to accommodate limited yielding excursions and inelastic buckling occurrences during an earthquake
without adverse loss in the load carrying capacity.
The increase of steel yield strength due to high strain rates
could be characterized through dynamic ancillary coupon tests.
Comparison of the quasi-static and dynamic cyclic column tests
showed that high strain rates resulted in an increased column
buckling and post-buckling compressive strength. Numerical
simulations showed that this increase was due to the increased
axial and flexural resistance of the column in the plastic hinge
region that resulted from the strain rate increase in yield
strength in the plastic hinge region.
In the centrally loaded cyclic tests, local flange buckling
developed in the plastic hinge region at large inelastic column
axial displacements corresponding to an axial ductility of 1.7.
When end moments were applied in the test, local flange
buckling occurred before global buckling of the column.
The OpenSees model with residual effects was capable of
reproducing the buckling and post-buckling response of
W-shaped and built-up I-shaped steel columns in monotonic
tests and quasi-static cyclic tests. The numerical model showed
that residual stresses reduce the compressive resistance of
columns, especially at the first buckling. Residual stress effects
tend to gradually diminish in the post-buckling range, which
suggests that they may not be critical for prediction of the
collapse of structures once the column capacity is exceeded. The
comparison between numerical and experimental results also
showed that the kinematic strain-hardening ratio specified in
the numerical model plays a key role on the strain response in
the plastic hinge region of columns.
The fibre cross-section model used in the OpenSees model did
not account for strain rate effects on the yield strength and
local buckling of the column cross-section elements. In the test
program, strain rate effects on the column compressive strength
were found to be significant and the model should be modified
to account for these effects so that more realistic predictions
of the seismic response of braced steel frames could be
obtained when an inelastic response is expected in the bracing
members and columns. Comparison between the numerical
and experimental results indicated that local buckling had
negligible effects on the post-buckling strength of the columns
studied when centrally loaded, but that these effects could be
more pronounced under combined axial and bending moments.
Incorporating local buckling effects in the OpenSees model
would therefore represent a useful implementation.
In future studies, the numerical models should be validated
against experimental data where axially loaded columns are
simultaneously submitted to transverse displacements, as is
the case in actual braced frame structures subjected to strong
seismic events.

Acknowledgements
Funding for this research was provided by the Natural Sciences
and Engineering Research Council (NSERC) of Canada, the Fonds
Qubcois de la Recherche sur la Nature et les Technologies
(FQRNT) and Canada Foundation for Innovation. The authors wish

to express their appreciation to the technical staff at the Hydro

Qubec Structures Laboratory of cole Polytechnique de Montral.
References
[1] Richards PW. Seismic column demands in ductile braced frames. Journal of
Structural Engineering, ASCE 2009;135(1):3341.
[2] Comit Europen de Normalisation. CEN. Eurocode 8: design of structures
for earthquake resistancepart 1: general rules, seismic actions and rules for
buildings. Brussels (Belgium): EN 1998-1:2004:E; 2004.
[3] National Research Council of Canada. NRCC. National building code of Canada.
NBCC. 12th ed. Ottawa (ON); 2005.
[4] American Institute of Steel Construction. AISC. Seismic provisions for
structural steel buildings. Including supplement no. 1. Chicago (IL): ANSI/AISC
341-05; 2005.
[5] Newell JD, Uang C-M. Cyclic behavior of steel wide-flange columns subjected
to large drift. Journal of Structural Engineering, ASCE 2008;134(8):133442.
[6] Lamarche CP, Tremblay R. Accounting for residual stresses in the seismic
stability of nonlinear beamcolumns elements with cross-section fibre
discretization. In: Structural stability research council. SSRC. Proceedings of
the 2008 annual stability conference. 2008. p. 5978.
[7] Lamarche CP, Tremblay R. Experimental and numerical study of the potential
for column size reduction in capacity designed multi-storey braced steel
frames. In: Proc. 9th US national and 10th Canadian conf. on earthquake eng.
Paper no. 1590. 2010.
[8] FEMA. Quantification of building system performance factors. FEMA report
P695. Redwood City (CA); 2009.
[9] Neuenhofer A, Filippou FC. Geometrically nonlinear flexibility-based frame
finite element. Journal of Structural Engineering, ASCE 1998;124(6):70411.
[10] Mazzoni S, McKenna F, Scott MH, Fenves GL. Open system for earthquake engineering simulation (OpenSees), user command-language manual. Berkeley
(CA): Pacific Earthquake Engineering Research. PEER. University of California;
2006.
[11] Agero A, Izvernari C, Tremblay R. Modelling of the seismic response of
concentrically braced steel frames using the opensees analysis environment.
The International Journal of Advanced Steel Construction 2006;2(3):24274.
[12] Tremblay R. Fundamental periods of vibration of braced steel frames for
seismic design. Earthquake Spectra 2005;21(3):83360.
[13] Uriz P, Filippou FC, Mahin SA. Model for cyclic inelastic buckling of steel braces.
Journal of Structural Engineering, ASCE 2008;134(4):61928.
[14] Beedle LS, Tall L. Basic column strength. Journal of the Structural Division, ASCE
1960;86(ST7):13973.
[15] Galambos TV, editor. Guide to stability design criteria for metal structures. 5th
ed. New York: Wiley-Interscience; 1998.
[16] Lacerte M, Tremblay R. Making use of brace overstrength to improve the
seismic response of multistorey split-X concentrically braced steel frames.
Canadian Journal of Civil Engineering 2006;33(8):100521.
[17] Tremblay R, Poncet L. Improving the seismic stability of concentrically braced
steel frames. Engineering Journal, AISC 2007;44(2):10316.
[18] Lamarche CP. Development of real time dynamic substructuring procedures
for the seismic testing of steel structures. Ph.D. dissertation. Universit degli
Studi di Trento, Trento, TN, Italy & cole Polytechnique de Montral, Montral,
QC, Canada; 2009.
[19] Wakabayashi M, Nakamura T, Iwai S, Hayashi Y. Effects of strain rate on the
behavior of structural members. In: Proceedings of the 8th world conference
on earthquake engineering. 8WCEE. vol. IV. 1984. p. 4918.
[20] Tremblay R, Filiatrault A. Seismic impact loading in inelastic tension-only
concentrically braced steel frames: myth or reality? Earthquake Engineering
and Structural Dynamics 1996;25(12):137389.
[21] Filiatrault A, Tremblay R. Design of tension-only concentrically braced steel
frames for seismic induced impact loading. Engineering Structures 1998;
20(12):108796.
[22] Pan CL, Wu S, Yu WW. Strain rate and aging effect on the mechanical properties
of sheet steels. Thin-Walled Structures 2001;39(5):42944.
[23] Menegotto M, Pinto PE. Method of analysis for cyclically loaded reinforced
concrete plane frames including changes in geometry and non-elastic behavior
of elements under combined normal force and bending. In: Proc. of LABSE,
symposium on resistance and ultimate deformability of structures acted on
by well defined repeated loads. 1973.
[24] Black RG, Wenger WAB, Popov EP. Inelastic buckling of steel struts under
cyclic load reversals. Report no. UCB/EERC-80/40. Berkeley (CA): Earthquake
Engineering Research Center, University of California; 1980.
[25] Canadian Standards Association. CSA. CAN/CSA S16-09: limits states design of
steel structures. Mississauga (ON, Canada); 2009.
[26] American Institute of Steel Construction. AISC. Specification for structural steel
buildings. Chicago (IL): ANSI/AISC 360-05; 2005.
[27] Kishima Y, Alpsten GA, Tall L. Residual stresses in welded shapes of flame-cut
plates in ASTM A572 (50) steel. Report no. 321.2. Fritz Engineering Laboratory,
Lehigh University; 1969.