Journal of Constructional Steel Research 104 (2015) 155–166

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Concentric tubular steel braces subjected to seismic loading: Finite

element modeling
Madhar Haddad
Department of Architectural Engineering, United Arab Emirates University, P. O. Box 15551, Al Ain, United Arab Emirates

a r t i c l e i n f o a b s t r a c t

Article history: Steel buildings are susceptible to damage during earthquakes if an unreliable bracing system is used. A well-
Received 22 February 2014 designed and detailed concentric bracing system is needed for steel buildings in a seismically active area. Failure
Accepted 10 October 2014 of a concentric bracing member occurs at the mid-length plastic hinge. A refined finite element model has been
Available online 29 October 2014
developed to simulate the hysteresis behavior of bracing members under cyclic loading including fracture. The
model provides similar hysteresis behavior to previous (Shaback and Brown [1]) and two new experiments
Keywords:
Finite element model
(Tremblay et al. [2]). The specimens were subjected to different loading protocols. It was found that an initial im-
Concentric tabular steel braces perfection affects the pre-buckling and first buckling cycles but has no effect on the following cycles. The greater
Cyclic behavior the initial yield stress of the HSS, the earlier is the occurrence of local buckling. The cumulative plastic strain is
Ultra-low-cycle fatigue greater at the outer surface than at the inner surface of the compressive corners/web of the mid-length plastic
Cumulative plastic strain hinge where fracture initiates. Significant local rotation follows the same trend as the significant plastic strain
Fracture of the same element where fracture initiates.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Concentric braces can be used in steel and concrete buildings to
minimize lateral storey drift by absorbing the input energy of wind
and ground movements. Under these movements, lateral drift is greater
in unbraced buildings than in braced buildings. For symmetrical inelas-
tic lateral loading history, lateral drift for a single diagonal brace is less in
tension than in compression due to brace buckling (the behavior of
concentric braces is different in tension from compression). Therefore,
concentric braces are commonly used in opposing pairs in buildings in
one bay or in adjacent bays.
Hollow structural steel sections (HSS) have been frequently tested
under reversed axial displacements, as such sections are popularly
used as bracing elements. The objective is to understand the hysteresis
behavior and predict the fracture life of such members in an attempt to
improve the performance of these braces when buildings are exposed to
seismic excitations. These tests (Jain et al. [3]; Black et al. [4]; Lee and
Goel [5]; Liu and Goel [6]; Tang and Goel [7]; Walpole [8]; Shaback
and Brown [1]; Tremblay et al. [9]; Elchalakani et al. [10]; Goggins
et al. [11]; Uriz [12]; Yang and Mahin [13]; Fell et al. [14,15]; Han et al.
[16]; Tremblay et al. [2]; Haddad et al. [17]; Roeder et al. [18]; Takeuchi
and Matsui [19]; lai and Mahin [20]) have revealed that fracture occurs
after a few (1 to 4) cycles of local buckling due to the high strains and
rotations at the compressive corners/web of the mid-length plastic
hinge. After the brace has buckled, most of the tensile elongation and in-
elastic strains occur within the mid-length plastic hinge region.
E-mail address: madhar@uaeu.ac.ae.
Finite element models have been developed for the same purposes.
The model of Haddad et al. [21] suggested that fracture life is related
to the total axial displacements without repeating the unloading parts
of cycles. The crack void growth model (CVGM) with the triaxiality
effect of strains is also able to predict the fracture life of HSS bracing
members (Kanvinde and Deierlein [22]; Myers et al. [23]; Fell et al.
[15,24]). A modified CVGM model is tested here for the CAN/CSA-
40.21-98, class C, Grade 350 W [25] and for ASTM A500, Grade C [26],
steel tubular braces (HSS) under cyclic loadings. Another objective of
the current work is to provide a refined material model that is able to
predict the exact hysteresis behavior and to overcome a shortcoming
of previous models, specifically the overshoots in the axial compressive
resistance. The effect of an initial imperfection on the hysteresis behavior
of HSS braces is investigated.
2. Design of specimens
All specimens (1B, 2A, 2B, 3A, 3B, 3C, 4A, 4B) of the Shaback and
Brown tests [1] listed in Table 1 were designed according to the CAN/
CSA-S16.1-94 (CSA 1994) standard [27] except for the fillet weld,
which was designed according to Korol [28]. The tensile capacity of
the connection (gusset plate) is greater than the tensile capacity of the
specimen. Rectangular gusset plates were inserted into longitudinal
slots in the flanges of the HSS at both ends. The gusset plates were
welded to the HSS along the slot. The front end of the gusset plate was
welded to the HSS at the slot. Hence, no cover plates were used at the
net section of the brace.

http://dx.doi.org/10.1016/j.jcsr.2014.10.013
0143-974X/© 2014 Elsevier Ltd. All rights reserved.
156 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166

Table 1
Properties of the HSS specimens, gusset plates and fillet weld.

Specimen HSS section, (mm) HSS length (mm) Total specimen Gusset plate cross sectional Fillet weld length, HSS E (GPa) HSS Fy (MPa)
length, L, (mm) area, wgtg, (mm2) Lw, (mm)

1B1 127 × 127 × 8 3350 3452 5715 300 191 421

2A 152 × 152 × 8 3950 4040 6350 350 202 442
2B 152 × 152 × 9.5 3950 4028 7620 350 196 442
3A 127 × 127 × 6.4 4350 4456 5080 300 196 461
3B 127 × 127 × 8 4350 4446 5715 300 191 421
3C 127 × 127 × 9.5 4350 4414 6350 300 202 461
4A 152 × 152 × 8 4850 4944 6350 350 202 442
4B 152 × 152 × 9.5 4850 4914 7620 300 196 442
RHS-192 152 × 152 × 9.5 3829 3905 11,201 290 200 345a
CHS-1 273 × 9.5 4109 4198 13,068 450 235 317a

CAN/CSA-40.21-98, Class C, Grade 350 W. Rectangular gusset plates, Fy (gusset plate) = 300 MPa (hot rolled). Asymmetrical displacement history. E480XX Electrode.
ASTM A500, Grade C. Tapered gusset plate, ASTM A572 Grade 345. Symmetrical displacement history. E490 XX Electrode. a Specified minimum yield stress.
1
Shaback and Brown [1].
2
Tremblay et al. [2].

Similarly, a slotted connection was adopted for specimens RHS-19 and CHS-1 are (16730, 13580), (23218, 20246), and (17818, 15362),
and CHS-1 of the tests by Tremblay et al. [2], listed in Table 1. For respectively.
these connections, the net section at the end of the slots was reinforced The gusset plates were connected to the HSS by the nodes on their in-
with cover plates to avoid fracture at this critical location. The connec- terfaces. The cover plates, if present, were connected to the HSS by tie-
tions were designed and sized according to the AISC 2005b [29] and type multi-point constrains (MPC) at nodes on the longitudinal sides of
the AISC seismic provisions [30], respectively. The gusset plate dimen- each cover plate. Fixed brace end conditions were used at both ends of
sions a and b are obtained following the modified Uniform Force Meth- the specimens. Neither the weld nor the slotted hole was modeled. The
od using the minimum possible gusset plate thickness tg with the
working point being located at the intersection of the beam bottom
flange and the adjacent column inner flange to minimize the gusset
plate dimensions (Sabelli [31]). All failure modes are verified including
net section fracture including shear lag effects, tear-out failure in the
brace and in the gusset plate, tension yielding on the Whitmore section
of the gusset plate, and failure of the welds.
The optimal brace inclination angle for maximum shear rigidity
with respect to the horizontal axis of Tremblay [2] tests is θ = 35°
(Moon et al. [32]). The results of design for all specimens are listed in
Table 1 and in the figures in the Appendix. For RHS-19 and CHS-1,
a = 513 mm and b = 359 mm.

3. Loading protocols

All specimens analyzed herein were subject to quasi-static cycles of

reversed axial displacements. Two displacement protocols were used:
the first is asymmetric in the experiments of Shaback and Brown [1]
and the second is symmetric in the experiments of Tremblay et al. [2]
as shown in Fig. 1(a) and (b). For the Tremblay et al. specimens, the
loading sequence was developed in terms of interstorey drift. The max-
imum normalized axial deformation of all braces is shown in Fig. 1(c).
The axial deformation of the brace is normalized with respect to the
brace length between the end hinges, LH.

4. Finite element model description

The use of a four-node quadrilateral shell element (S4R) in modeling

steel braces accounts for large strains and rotations and allows for
changes in thickness with deformation. S4R is a reduced integration el-
ement with three translational and three rotational degrees of freedom
per node. This element is suitable for analyzing thin and thick members.
Seven integration points were adopted through the thickness of the S4R
element so that the nonlinear stress distribution was estimated. Trian-
gular shell elements were avoided since they can behave in too stiff a
manner. A high mesh density with element size less than the thickness
of the HSS was used at the mid-length plastic hinge. The element aspect
ratio in this region was 1. The number of (nodes, elements) used in Fig. 1. Displacement protocols for specimen 4A (a), RHS-19 and CHS-1 (b), and maximum
modeling each specimen of the Shaback and Brown tests [1], RHS-19 amplitudes for tests of Shaback and Brown [1] and Tremblay et al. [2], (c).
 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166
details of the slotted hole at the net-section should not affect the brace
fracture response since fracture developed at the brace mid-length. The
reference point was selected so that the brace model lies in the first
quadrant. The normal of each shell element in the HSS was directed to
the outside of the HSS. The models of the brace connections are shown
in Fig. 2 for the Shaback and Brown tests [1], RHS-19 and CHS-1.
The finite element analysis was performed in two stages using the
Abaqus 2011 code [33]. The first mode elastic buckling analysis was cre-
ated in the specimens in the first stage. The initial imperfection to match
the measured camber and the cyclic end displacements was then
Fig. 2. Brace connection models for specimens: 4A (a), RHS-19 (b), and CHS-1 (c).
157
applied in the second stage where nonlinear inelastic analysis was
adopted to simulate the hysteresis loops. The Newton–Raphson method
was used to solve the nonlinear equilibrium equations in the stability
based approach with a maximum of 1000 increments in each step.
The isotropic hardening material model is frequently used for most
analyses with metals. However, this material model is not able to simu-
late the actual shift in the yield surface, as it changes the yield surface
size in all directions in stress space so that the yield stress decreases
or increases with plastic straining. The combined isotropic kinematic
hardening material model was used in the finite element analysis
presented here with true plastic stress–strain data equal to a half
cycle. This model is capable of simulating the expansion, contraction
and the shift of the yield surface in stress space to model the brace be-
havior under cyclic loading. The yield surface is defined by the function

 3

1=2
d d
f σ i j; k ¼ Si j −α i j Si j −α i j −σ y0 −k ¼ 0 ð1Þ
2
where σij is the stress tensor, αdij is the deviatoric part of the backstress
tensor Si j, σy0 is the initial yield stress, and k is the hardening parameter
that defines the size of the yield surface.
It should be noted that when k = 0, Eq. (1) reduces to purely kine-
matic hardening and when αdij = 0, purely isotropic hardening is obtain-
ed. The shift (translation) of the yield surface can be subtracted from the
corresponding stresses according to different kinematic hardening
rules. A review of the plastic hardening models (Prager [34]; Ziegler
[35]; Mroz [36]; Dafalias and Popov [37,38]; Armstrong and Frederick
[39]; Chaboche [40]; Bower [41]; lemaitre and Chaboche [42]), revealed
that Ziegler's [35] kinematic hardening rule is sufficient to model the
shift in the yield surface in stress space for the Shaback and Brown [1]
HSS brace tests.
In Eq. (1), the translation of the yield surface αdij is subtracted from
the corresponding stresses Sij. The increment of the deviatoric part of
the backstress tensor according to Ziegler's [35] kinematic hardening
rule is defined as
d H
d

p
dα i j ¼ S −α i j dεeq ð2Þ
σy ij
where
dσ
H¼ ð3Þ
dεp
is the plastic hardening modulus, and
σ y ¼ σ y0 þ k ð4Þ
is a stress quantity associated with the expansion or contraction of the
yield surface.
Fig. 3. True stress–strain curves for specimen 4A (walls, corners, gusset plate).
158 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166

Fig. 4. Strain gauge deployment, local buckling and fracture at the mid-length of specimen 2 (Haddad et al. [17]).

The shift in the yield surface is a function of the plastic hardening (Fadden [47]). Strain hardening is found upon reverse bending of sheet
modulus, the initial yield stress and the plastic strains. The increment metals (Zhao and Lee [48]). Hence, greater plastic strength values with a
in plastic strain according to the associated flow is defined as low plastic hardening modulus are present in the test results for
coupons at the corners compared to the walls of the HSS. Specimen
p ∂f p RHS-19 was modeled using the stress–strain curve from the walls only
dεi j ¼ dε ð5Þ
∂σ i j eq as corner coupon tests were not available. Modeling large scale braces
using the stress–strain curves from the walls was sufficient to produce
where the increment of the equivalent plastic strain is defined as the hysteresis behavior of the brace. The area of the corners is small
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi compared to the cross-sectional area of the HSS.
p 2 p p An extension to the Rice and Tracey model [49] was the Kanvinde
dε eq ¼ dε dεi j : ð6Þ
3 ij and Deierlein CVGM model [22], which predicts fracture for ultra-low-
cycle fatigue (ULCF) braces under cyclic loadings. The model is based
on the following equation (fracture occurs in braces when the degrada-
5. Refined material and fracture models tion in monotonic capacity is less than the cyclic growth in demand):

The engineering stress–strain curves obtained from coupon tests


(Bubela [43]; Aguilera et al. [44]; Moreau, et al. [45]) were converted exp −λCVGM ε p ηmonotonic ≤
to true stress–strain curves that were used here in the finite element X Zε2 X Zε2 ð7Þ
models of the Shaback and Brown tests [1] and specimens RHS-19 and
tensile−cycles
exp ðj1:5T jÞdε t − compressive−cycles
exp ðj1:5T jÞdεc :
CHS-1. Representative true stress–strain curves for the walls and cor- ε1 ε1
ners of specimen 4A in addition to the gusset plates are shown in
Fig. 3. Thus, the material effects of the corners are considered in the
modeling of the square HSS braces. The true stress–strain curves in On the demand side, the equivalent compressive strains are
the post-necking range were obtained by changing the plastic harden- subtracted from the equivalent tensile strains and added incrementally
ing modulus through a trial and error procedure. Beyond necking, the for every cycle of loading taking triaxiality, T, into account. Such triaxiality
stress state is neither uniaxial nor uniform and the equivalent stress de- is absent from the Haddad et al. model [21]. Triaxiality is the mean stress
parts from the experimentally-derived true stress (Mirone and Corallo over the effective or the Von Mises stress. Cycles are considered compres-
[46]). For the Bubela [43] and Aguilera et al. [44] tests, coupons were ex- sive when triaxiality is negative and vice versa.
tracted from HSS tube wall segments, corner segments and gusset The capacity side represents the degradation of the monotonic tensile
plates. By using these curves, different material models were utilized capacity, ηmonotonic, due to cyclic loading according to the equation,
for the walls, the corners and the gusset plates, rather than modeling
the brace as a whole. Cold working seriously affects the ductility in the

ηcyclic ¼ exp −λCVGM εp ηmonotonic ð8Þ
corner regions of the HSS suggesting that low cycle fatigue is a concern

Fig. 5. Axial displacement brace–gauge ductility (a) and axial load brace–gauge ductility (b) for specimen 2 of Haddad, et al. [17] tests.
 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166
Fig. 6. Critical significant plastic strain/rotation for specimens 4A (a,b) and RHS-19 (c,d).
εp
where, ηmonotonic ¼ ∫ exp ð1:5T Þdεp and εp is the cumulative plastic
0
strain. The monotonic capacity degrades according to the degradation
coefficient λCVGM and can be obtained from standard coupon or notched
bar tests.
The calibration of ηmonotonic and λCVGM based on notched bar tests
(Fell et al. [15]) is representative of the localized behavior that predis-
poses fracture initiation at the mid-length of the brace. This explains
the high significant plastic strains on the capacity side in the finite ele-
ment model by Fell et al. [15]. The stress–strain curves for the corners
(Bubela [43]) and for the notched bars (Fell et al. [15]) are similar
where early ultimate tensile capacity is attained and degraded thereaf-
ter. This degradation leads to a lower plastic hardening modulus for the
corners than for the walls. Consequently, the cumulative plastic strains
will be greater for the corners than the walls. However, there is a lack of
experimental tests to estimate the values of ηmonotonic for the corners of
the HSS. As a result the coupon tests from the flat parts of the HSS are
used to estimate the values of ηmonotonic.
Two cases are considered here for fracture life prediction on the de-
mand side. In the first case, the significant plastic strains are taken to be
compressive minus tensile strains. In the second case, the significant
plastic strains are taken to be tensile minus compressive strains. In
Fig. 7. Cumulative/significant plastic strain of element outer/inner integration points at
location of fracture initiation for specimen 4A.
159
either case, strains are considered compressive when triaxiality is less
than −1/3. The −1/3 is a revised simplification for positive maximum
principal stress (Huang and Mahin [50]). The maximum principal stress
is positive for damage evolution in cyclic loading (Lemaitre [51]). For
cyclic loading, a slightly varied modified form of damage evolution by
Lemaitre [51] was used by Huang and Mahin [50]:
(
m 1
dD ¼ ðcyÞ dεp for TN− ; and
3
0 otherwise
2  
σ eq 2 2
y¼ ð1 þ ν Þ þ 3 ð1−2ν ÞT ð9Þ
2Eð1−DÞ2 3
Zεp
Dc ¼ dD
0
where c and m are material constants, E is the elastic modulus, ν is the
Poisson's ratio, σeq is the effective or the Von Mises stress, dD is damage
increment and Dc is the critical damage limit.
In the Shaback and Brown [1] tests, axial strains in compression are
greater than axial strains in tension as discussed below (referring to
strain gauges deployed at the mid-length plastic hinge for specimen 2
of the Haddad et al. tests [17]). In addition, the applied compressive dis-
placements are greater than the applied tensile displacements of the
brace as shown in Fig. 1(a). Therefore, in the first case, the significant
plastic strains are taken to be compressive minus tensile strains. Plastic
strains are accumulated during compressive cycles in the model of
Myers et al. [23]. For the Shaback and Brown tests the significant plastic
strain is always positive (for each hysteresis cycle, compressive strains
are always greater than tensile strains). In the second case, tensile
strains minus compressive strains are considered. This is based on the
basic principle of fracture mechanics that crack propagation is driven
by a tensile stress.
Specimen 2 of Haddad et al. [17] was subjected to the same loading
history as the specimens in the Shaback and Brown tests [1]. High elon-
gation strain gauges were deployed at the neutral axes and corners of
the plastic hinge location in the middle of the specimen, as shown in
Fig. 4. The axial brace–gauge displacement ductility and the correspond-
ing axial brace–gauge load ductility are shown in Fig. 5(a), and (b). As
160 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166
may be seen, gauge ductility is greater than brace ductility. For speci-
men 2, cracking started at the compressive side corners/web and
progressed into the flanges as shown in Fig. 4. Similarly, in the Shaback
and Brown experiments, two different forms of small cracks, which ini-
tiated the main crack, were observed to occur either at the compressive
corners of the hollow structural section (HSS) or in the compressive
web at the mid-length of the specimens, normal to their longitudinal
axis.
For all of Shaback and Brown experiments [1], λCVGM is equal to 0.11
and ηmonotonic is between 2.65 and 2.97. Calibrating the capacity side of
the finite element models based on coupons from walls of HSS results
in an average ηmonotonic value of 2.8 which is compatible with ASTM
A572 Grade 50 steel, AW50 (Kanvinde and Deierlein [22]). The yield
and ultimate strengths of the Shaback and Brown braces [1] are similar
to AW50 steel with Fy = 423 MPa and Fu = 494 MPa. The fracture life of
the Shaback and Brown specimens [1] are similar to the predictions of
the finite element model when considering tensile minus compressive
strains and vice versa. A representative prediction of fracture life for
specimen 4A is shown in Fig. 6(a), and (b). For all braces, the fracture
cycle was not altered when using values of λCVGM up to 0.7. However,
fracture will shift to the compressive side of the same cycle when
using larger values for λCVGM in the range (0.11 to 0.7). There was no dis-
tinctive value for λCVGM to separate between compressive and tensile
fracture in the model. This distinctive value is not needed since fracture
in all braces occurred during the tensile side of the cycle when λCVGM is
equal to 0.11.
For RHS-19 and CHS-1 of the Tremblay et al. experiments [2],
λCVGM is equal to 0.32 and η monotonic is 2.2 and 2.4, respectively
which is compatible with ASTM A572 Grade 50 steel, AP50
(Kanvinde and Deierlein [22]) with F y = 388 MPa and Fu =
588 MPa. The fracture life of these two specimens was the same
Fig. 8. Deflected shape and local buckling of specimen 4A: a) half-bow shape, b) half-V shape, c) local buckling and small size bent.
when triaxiality is less than − 1/3 or zero. Thus triaxiality could be
related to the asymmetry of the displacement history for the Shaback
and Brown experiments [1]. The prediction of fracture life for RHS-19
is shown in Fig. 6(c), and (d).
Cumulative plastic strain is used to specify the location of crack
initiation for all specimens. The cumulative plastic strain and corre-
sponding significant plastic strain are greater at the external integra-
tion point (outer face) than at the internal integration point (inner
face) of the element where the fracture initiated, as shown in Fig. 7
for specimen 4A. The axial and rotational capacity of the element in
consideration is exceeded at the outer faces due to cyclic loading
leading to fracture initiation at the external face of the element
(Fig. 8).
In all cyclic brace tests cracks initiate at the outer or external free
surface of the compressive corners/web of the mid-length plastic
hinge as shown in Fig. 11(a) for specimen 4A.
An interesting result is the significant local rotational demand
that is found to have the same trend and value as plastic strain
for the same element where the fracture initiates (Fig. 6). This
would suggest replacement of the significant plastic strains by the
significant local rotations on the demand side. A small size bent
at the compressive corners and web at the mid-length plastic
hinge is usually present upon compressive buckling. This bent is
usually severe at the compressive corners (Figs. 4 and 8) due to
the geometric nature of the local buckling and bending. The bent
opens during tensile loading and concaves during compressive
buckling, leading to crack formation due to the cyclic rotational de-
mand and reduction in the cross-sectional area at those locations.
Therefore, cyclic bending tests on small scale notched specimens
are needed to estimate the local-rotational capacity that is absent
from literature.
 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166 161

6. Validation with experiments
The refined finite element model was validated against the exper-
iments of Shaback and Brown [1] (1B, 2A, 2B, 3A, 3B, 3C, 4A, 4B) and
Tremblay et al. [2] (RHS-19 and CHS-1). Sample axial hysteresis
loops for specimens (1B, 3A, 3C, 4A, RHS-19, CHS-1) are presented
in Fig. 9(a). The corresponding lateral hysteresis loops are presented
in Fig. 9(b). The overshoots on the compression side of previous
models presented in literature indicate that the Bauschinger effect
is not modeled properly. Neither the Drain 2DX nor the OpenSees

Fig. 9. (a): Numerical versus experimental axial hysteresis loops. (b): Numerical versus experimental lateral hysteresis loops.
162 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166
Fig. 9 (continued).
physical theory models of HSS braces account for local buckling. In
addition, the Drain 2DX brace model does not consider the progres-
sive degradation of the tangent stiffness during cycles and the grad-
ual spread of plasticity along the brace length (Chen and Tirca [52]).
For a stretched tabular brace, the Bauschinger effect refers to the fact
that the axial compressive yield stress is smaller than the axial ten-
sile yield stress. The greater the initial yield stress, the smaller the
tangent modulus, consequently the earlier the occurrence of local
buckling, as shown in Fig. 9(a) for specimen CHS-1 for instance. A
small increment in the initial yield stress can have an effect on the
occurrence of local buckling. The overshoots are absent from the
compressive hysteresis behavior in the current refined model, but
the degradation in the tangent stiffness is modeled properly. More-
over, the refined finite element model presented here was able to
predict the behavior of the brace in terms of compressive and tensile
strength, stiffness for both loading and unloading parts, out-of-plane
deformation, energy dissipation and fracture life. The experimental
versus numerical buckling load, tensile resistance, global and local
buckling cycles in addition to fracture cycle are listed in Tables 2
and 3. Excellent agreement is seen between the experimental and
the finite element results. The percent error is below 6% for the
buckling load, 4% for the tensile resistance and 6% for the energy
dissipation capacity among all specimens considered in the current
study. Part of the previous difference in the energy dissipation capac-
ity between the finite element models and the experimental results
of Shaback and Brown [1] is due to frame slippage. The experimental
and finite element fracture life results at instance of fracture
were essentially the same. The behavior of specimens (1B, 2A,
2B, 3A, 3B, 3C, 4A, 4B, RHS-19) was similar. Therefore, specimen
4A was selected here as a representative of the previous nine
specimens.
Specimen 4A (HSS152 × 152 × 8.0 mm brace 44944 mm long
with 250 × 25.4 mm gusset plates and free length of 47 mm) exhib-
ited elastic behavior up to and including the fourth cycle. During the
fifth compressive cycle, the specimen buckled laterally in a bow
shape as shown in Fig. 8(a), at axial loads 1459 kN and 1381 kN in
the finite element model and the experiment, respectively with a
corresponding error of 5.7%. Two plastic hinges occurred in the
 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166 163

Table 2
Experimental versus numerical buckling loads, tensile resistance and fracture cycle.

Specimen Buckling load (kN) Tensile resistance (kN)

Experimental Numerical % Experimental Numerical %

error error

1B 1159 1132 −2.1 1647 1706 3.6

2A 1507 1504 −0.2 2165 2203 1.8
2B 1721 1683 −2.2 2624 2659 1.3
3A 864 824 −4.7 1462 1499 2.5
3B 952 997 4.7 1632 1664 2.0
3C 1011 955 −5.6 2284 2276 −0.4
4A 1381 1459 5.7 2132 2208 3.6
4B 1435 1371 −4.5 2585 2627 1.6
RHS-19 N.A. 1415 N.A. N.A. 2324 N.A.
CHS-1 N.A. 1975 N.A. N.A. 2991 N.A.

Fig. 10. Variation of energy dissipation with cycle number.

free length in the gusset plates. Compressive resistance was notice- both strength and stiffness of the brace, due to local buckling and
ably reduced in compressive cycles following the fifth cycle as the Bauschinger effect.
shown in Fig. 9(a). The rate of reduction became smaller when the The location of the plastic hinges for specimens (1B, 2B, 3B, 3C,
cycle number increased as noted with all other specimens. In the fi- 4A) was at the exact mid-length of those specimens in both the ex-
nite element model and the experiment, the deflected shape of the periments and the finite element models, as shown in Fig. 11(a),
specimen was sinusoidal until the end of the seventh compressive and (b) for specimen 4A for instance. This location was shifted at a
cycle. Local buckling occurred in the ninth compressive cycle. distance of less than the size of the HSS for specimens 2A, 3A and
After the specimen buckled locally, the curvature was concentrated 4B as shown in Fig. 11(c), and (d) for specimen 3A for example. In
in the end connections and the central hinge region. The shape of the experiments and the finite element modeling, the stress–strain
the specimen transformed from a bow shape to a kinked shape (V curve for the corners is different from that of the walls of the HSS.
shape) as shown in Fig. 8(b). A small size bent is usually present The difference is higher at the early yielding stages for the corners
at the compressive corners of the local buckling region as shown than the walls of the HSS (Fig. 3). After brace buckling, the plastic de-
in Fig. 8(c) that causes fracture initiation due to high plastic strain mand will be concentrated at the compressive corners where high
demand. The specimen failed on the tenth cycle of tension loading yield capacity is present at the brace mid-length. This will force the
at 1034 kN and 1059 kN, in the finite element model and the exper- demand to be shifted at a small distance where yielding is easy to
iment, respectively. Specimen 4A reached maximum displacement grow in the compressive corners/web leading to formation of the
levels of 28.9 mm in tension and 75 mm in compression. The exper- mid-length plastic hinge at a small distance from the mid-length of
imental and numerical results of the axial stiffness at zero axial dis- the brace. The initial imperfections with size and length of the HSS
placement were the same. in addition to end conditions would play a role in this shift since
The area under the axial hysteresis loops represents the impor- not all specimens exhibited this behavior.
tant attribute of energy dissipation capacity of the brace during the Specimens 3A, 3B and 3C were the same in terms of HSS length
load reversals. As an example, the experimental and corresponding (4350 mm) and gusset plate length (350 mm) and thickness
numerical energy absorbed by specimens 4A, plotted against cycle (25.4 mm). These 127 mm HSS specimens had thicknesses of 6.4, 8,
number is shown in Fig. 10. A slight increase in the energy dissipa- and 9.5 mm, respectively. Specimens 3A, 3B and 3C fractured on
tion, not exceeding 5 kN·m, occurs during each of the partial yield- the tensile side of the ninth, thirteenth and seventeenth cycles, re-
ing cycles (cycles 3 and 4). A significant increase in the rate of spectively. Fracture life was inversely proportional to the thickness
energy dissipation occurs thereafter, due to buckling and yielding of the HSS. The axial and rotational capacity of the HSS increased
of the bracing member. However, the rate of the energy dissipation when reducing the compactness ratio. However, ηmonotonic is of an
begins to decline, then degrades rapidly due to the reduction in average value of 2.8 since it represents an integration point in a
tiny element. Seven integration points were implemented through
the thickness of the element. The number of cycles to fracture in-
creased as the width-to-thickness ratio became smaller in the
Tremblay et al. [2] and Haddad et al. [17] tests on HSS braces fabricat-
Table 3
Experimental versus numerical energy dissipation capacity and peak points. ed from ASTM A500, Grade C [26] and CAN/CSA-40.21-98, class C,
Grade 350 W [25], respectively.
Specimen Energy dissipation (kN·m) Cycle number
The use of shell elements in modeling tubular braces is more ap-
Experimental/numerical propriate than using solid elements. A brace shell model can be eas-
Experimental Numerical % error Global Local Fracture ily created by sketching the geometric boundaries and converting
buckling buckling them into surfaces. It is more convenient to use multiple integration
1B 357 352 −1.4 5 9 11 points through the thickness of one shell element than stacking
2A 300 293 −2.3 5 8 9 solid elements with one integration point for each solid element
2B 443 447 0.9 5 9 10 in addition to the computational expense of using solid elements
3A 178 168 −5.6 4 8 9
which are stiffer than shell elements. Solid elements are not able
3B 406 404 −0.5 5 11 13
3C 739 742 0.4 4 14 17 to simulate the kink at the mid-length of the brace, while this be-
4A 366 351 −4.1 5 9 10 havior is evident when using shell elements (Santagati et al. [53]).
4B 636 661 3.9 5 9 12 The use of full or reduced integration shell elements did not alter
RHS-19 N.A. 863 N.A. 8 24 24 the hysteresis response of the brace. Thus, S4R was used in the anal-
CHS-1 N.A. 441 N.A. 8 21 23
ysis of all models presented here due to its computational
164 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166

Fig. 11. Initial failure at mid-length of specimen 4A in experiment (a), in finite element model (b) and specimen 3A in experiment (c), in finite element model (d).

advantage and to avoid the possibility of shear and membrane
locking. However, refining the mesh at the mid-length plastic
hinge way below the thickness of the HSS led to non-convergence
during the first severe local buckling cycle due to the softening be-
havior of these elements.
Initial imperfection was found to affect the pre-buckling cycle
and the buckling cycle but had no effect on the following cycles.
The effect of initial imperfection for specimen 4A is shown in
Fig. 12(a), and (b). The full hysteresis behavior seen in the first and
the second buckling cycles; initial imperfection = 0.00135 m, was
not seen in any cyclic test of HSS braces. Once buckling occurs, the
initial imperfection had no effect on the cycles that followed the
first buckling cycles. In modeling, the acceptable hysteresis behavior
of braces (without the full hysteresis loops) complies with the
smallest initial imperfection of 0.00145 m here for specimen 4A, for
example.
3. The cumulative plastic strain is greater at the outer surface than the
inner surface of compressive corners/web of the mid-length plastic
hinge where fracture occurs.
4. The significant local rotation follows the same trend as the
significant plastic strain of the same element where fracture
occurs. This suggests replacing the significant plastic strain by
the significant local rotation in predicting fracture life of braces.
5. The greater the initial yield stress of the HSS the lower is the
tangent modulus; consequently the earlier the occurrence of
local buckling.
6. Initial imperfections affect the pre-buckling cycle and the buckling
cycle but have no effect on the following cycles.
7. Cyclic bending tests on small scale notched specimens (e.g.: sec-
tions of the webs, corners and welds) are recommended to esti-
mate the local-rotational capacity at which cracking occurs in
the HSS.

7. Conclusions and recommendations

Acknowledgments

1. The refined finite element model was able to simulate the hysteresis The study was made possible by financial support of the Individual
behavior of tubular steel braces. Research Grant (Seed-AE-2012/G00001044) of the UAE University.
2. The refined CVGM model was able to accurately predict the fracture Many thanks to Professor Tom Brown and Brad Shaback, University of
at the mid-length of the brace. Calgary, for providing the experimental data.

Fig. 12. Sensitivity analysis to initial imperfection for specimens 4A.

 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166 165

Appendix A
166 M. Haddad / Journal of Constructional Steel Research 104 (2015) 155–166
References
[1] Shaback B, Brown T. Behaviour of square hollow structural steel braces with end
connections under reversed cyclic axial loading. Can J Civil Eng 2003;30(4):745–53.
[2] Tremblay R, Haddad M, Martinez G, Richard J, Moffatt K. Inelastic cyclic testing of
large size steel bracing members. Proc 14th World Conf on Earthq Eng, Beijing,
China; 2008 [Paper No. 05-05-0071].
[3] Jain AK, Goel SC, Hanson RD. Hysteresis behavior of bracing members and seismic
response of braced frames with different proportions. Report no. UMEE 78R3. Ann
Arbor, Michigan: Dept of Civil Eng, Univ of Michigan; 1978.
[4] Black RG, Wenger WA, Popov EP. Inelastic buckling of steel struts under cyclic load
reversals. Report no. UCB/EERC-80/40. Berkeley, CA: Earthq Eng Res Center, Univ
of California; 1980.
[5] Lee S, Goel SC. Seismic behaviour of hollow and concrete-filled square tubular
bracing members. Report no. UMCE87-11. Ann Arbor, Michigan: Dept of Civil Eng,
Univ of Michigan; 1987.
[6] Liu Z, Goel SC. Cyclic load behavior of concrete-filled tubular braces. J Struct Eng
ASCE 1988;114(7):1488–506.
[7] Tang X, Goel SC. Brace fractures and analysis of phase I structure. J Struct Eng ASCE
1989;115(3):1960–76.
[8] Walpole WR. Behaviour of cold-formed steel RHS members under cyclic loading.
Research report no. 96-4. Christchurch, New Zealand: Univ of Canterbury; 1996.
[9] Tremblay R, Archambault MH, Filiatrault A. Seismic performance of concentrically
braced steel frames made with rectangular hollow bracing members. J Struct Eng
ASCE 2003;129(12):1626–36.
[10] Elchalakani M, Zhao XL, Grzebieta RH. Tests of cold-formed circular tubular braces
under cyclic axial loading. J Struct Eng ASCE 2003;129(4):507–14.
[11] Goggins JM, Broderick BM, Elghazouli AY, Lucas AS. Experimental cyclic response of
cold-formed hollow steel bracing members. Eng Struct 2005;27(7):977–89.
[12] Uriz P. Towards earthquake resistant design of concentrically braced steel structures.
[Ph.D. Thesis] Berkeley, CA: Dept of Civil Eng, Univ of California; 2005.
[13] Yang F, Mahin SA. Limiting net section fracture in slotted tube braces. Steel tips
series, Struct Steel Education Council, Moraga, CA; 2005.
[14] Fell BV, Myers AT, Deierlein GG, Kanvinde AM. Testing and simulation of ultra-low
cycle fatigue and fracture in steel braces. Proc 8th US National Conf on Earthq Eng,
San Francisco, CA; 2006 [Paper No. 587].
[15] Fell BV, Kanwinde AM, Deierlein GG, Myers AT, Fu X. Buckling and fracture of
concentric braces under inelastic cyclic loading. Steel tips. Struct Steel Education
Council, Moraga, CA; 2006.
[16] Han SW, Kim WT, Foutch DA. Seismic behavior of HSS bracing members according to
width–thickness ratio under symmetric cyclic loading. J Struct Eng ASCE 2007;
133(2):264–73.
[17] Haddad MA, Brown TG, Shrive NG. Experimental cyclic loading of concentric HSS
braces. Can J Civil Eng 2011;38(1):110–23.
[18] Roeder Ch, Lumpkin E, Lehman D. A balanced design procedure for special concen-
trically braced frame connections. J Const Steel Res 2011;67:1760–72.
[19] Takeuchi T, Matsui R. Cumulative cyclic deformation capacity of circular tubular
braces under local buckling. J Struct Eng ASCE 2011;137(11):1311–8.
[20] Lai JW, Mahin SA. Experimental and analytical studies on the seismic behavior of
conventional and hybrid braced frames. Report no. 2013/20. Berkeley, CA: Pacific
Earthq Eng Res Center, Univ of California; 2013.
[21] Haddad MA, Brown TG, Shrive NG. Finite element modeling of concentric HSS braces
under cyclic loading. Can J Civil Eng 2011;38(5):493–505.
[22] Kanvinde AM, Deierlein GG. Micromechanical simulation of earthquake-induced
fracture in steel structures. Report no. 145. San Francisco, California: Dept of Civil
and Env Eng, Stanford Univ; 2004.
[23] Myers AT, Deierlein GG, Kanvinde AM. Testing and probabilistic simulation of ductile
fracture initiation in structural steel components and weldments. Report no. 170.
San Francisco, California: Dept of Civil and Env Eng, Stanford Univ; 2009.
[24] Fell BV, Kanvinde AM, Deierlein GG. Large-scale testing and simulation of earth-
quake induced ultra low cycle fatigue in bracing members subjected to cyclic inelas-
tic buckling. Report no. 172. San Francisco, California: Dept of Civil and Env Eng,
Stanford Univ; 2010.
[25] CSA. General requirements for rolled or welded structural quality steel. CAN/CSA-
40.21-98 Rexdale, Ontario: Canadian Standards Association; 1998.
[26] ASTM. Standard Specification for Cold-Formed Welded and Seamless Carbon Steel
Structural Tubing in Rounds and Shapes. ASTM A500/A500M-07. ASTM Internation-
al, West Conshohocken, Pennsylvania, USA; 2007.
[27] CSA. Limit state design of steel structures. CAN/CSA-S16.1-94 Rexdale Ontario:
Canadian Standards Association; 1994.
[28] Korol RM. Shear lag in slotted HSS tension members. Can J Civil Eng 1996;23(6):
1350–4.
[29] AISC. ANSI/AISC 360-05, specification for structural steel buildings. Chicago, IL:
American Institute of Steel Construction; 2005.
[30] AISC. Seismic provisions for structural steel buildings. Chicago, IL: American Institute
of Steel Construction Inc.; 2005.
[31] Sabelli R. Seismic braced frames — design concepts and connections. An AISC short
course. North American Steel Const Conf. Montreal, QC; 2005.
[32] Moon KS, Connor JJ, Fernandez JE. Diagrid structural systems for tall buildings: char-
acteristics and methodology for preliminary design. Struct Des Tall Spec Build 2007;
16:205–30.
[33] Abaqus. User's manual, version, 11.6. Providence, RI: Hibbitt, Karlsson, and Sorensen,
Inc.; 2011.
[34] Prager W, Providence RI. A new method of analyzing stresses and strains in work-
hardening plastic solids. ASME J Appl Mech 1956;23:493–6.
[35] Ziegler H. A modification of Prager's hardening rule. Q Appl Math 1959;17:55–65.
[36] Mroz Z. On the description of anisotropic work hardening. J Mech Phys Solids 1967;
15:163–75.
[37] Dafalias YF, Popov EP. A model of nonlinearly hardening materials for complex loading.
Acta Mech 1975;21:173–92.
[38] Dafalias YF, Popov EP. Plastic internal variables formalism of cyclic plasticity. J Appl
Mech 1976;43:645–51.
[39] Armstrong PJ, Frederick CO. A mathematical representation of the multiaxial
Bauschinger effect. CEGB report, RD/B/N/731, Berkeley Nuc Lab, R&D Dept, CA; 1966.
[40] Chaboche JL. Time-independent constitutive theories for cyclic plasticity. Int J Plast
1986;2(2):149–88.
[41] Bower AF. Cyclic hardening properties of hard-drawn copper and rail steel. J Mech
Phys Solids 1989;37(4):455–70.
[42] Lemaitre J, Chaboche JL. Mechanics of materials. Cambridge: Cambridge University
Press; 1990.
[43] Bubela RK. An experimental and analytical study of chevron braced frames with
vertical slotted connections. [MSc Thesis] Vancouver, B.C., Canada: Dept of Civil
Eng, Univ of British Colombia; 2003.
[44] Aguilera J, Shaat A, Fam A. Strengthening T-joints of rectangular hollow steel sections
against web buckling under brace axial compression using through-wall bolts. Thin-
Walled Struct 2012;56:71–8.
[45] Moreau R, Roger CA, Tremblay R, Packer JA. Finite element evaluation of the
“modified-hidden-gap” HSS slotted tube-to-plate connection. Connections VII, 7th
Int workshop on connections in steel struct, Timisoara, Romania. 30 May–2 June;
2012 [10 pp.].
[46] Mirone G, Corallo D. Stress–strain and ductile fracture characterization of an X100
anisotropic steel: experiments and modelling. J Eng Fract Mech 2013;102:118–45.
[47] Fadden MF. Cyclic bending behavior of hollow structural sections and their applica-
tion in seismic moment frame systems. [PhD Dissertation] Ann Arbor, MI: Dept of
Civil Eng, Univ of Michigan; 2013.
[48] Zhao KM, Lee JK. Generation of cyclic stress–strain curves for sheet metals. J Eng
Mater Technol 2001;123:391–7.
[49] Rice JR, Tracey DM. On the ductility enlargement of voids in triaxial stress fields. J
Mech Phys Solids 1969;17:201–17.
[50] Huang Y, Mahin SA. A cyclic damage plasticity model: implementation and applica-
tions. Proc 10th Int LS-DYNA Users Conf. Dearborn, Michigan USA. June 8–10; 2008
[14 pp.].
[51] Lemaitre J. A course on damage mechanics. Springer-Verlag; 1992.
[52] Chen L, Tirca L. Simulating the seismic response of concentrically braced frames
using physical theory brace models. Open J Civ Eng 2013;3:69–81.
[53] Santagati S, Bolognini D, Nascimbene R. Strain life analysis at low-cycle fatigue on
concentrically braced steel structures with RHS shape braces. J Earthq Eng 2012;
16(S1):107–37. http://dx.doi.org/10.1080/13632469.2012.675840.