You are on page 1of 13

Engineering Structures 37 (2012) 1–13

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Experimental and modeling study of high-strength structural steel under


cyclic loading
Gang Shi a,b,⇑, Meng Wang a, Yu Bai c, Fei Wang d, Yongjiu Shi a, Yuanqing Wang a
a
Department of Civil Engineering, Tsinghua University, Beijing 100084, China
b
State Key Laboratory for Disaster Prevention in Civil Engineering, Shanghai 200092, China
c
Department of Civil Engineering, Monash University, Melbourne, VIC 3800, Australia
d
College of Civil Engineering, Chongqing University, Chongqing 400045, China

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

Article history: In order to study cyclic performance of high-strength structural steel and establish an appropriate con-
Received 17 March 2011 stitutive relationship, experiments were carried out on seventeen Q460D steel specimens subjected to
Revised 5 December 2011 different loading patterns. The mechanical responses of high-strength structural steel were evaluated
Accepted 8 December 2011
and discussed including stress–strain relationship, failure modes, ductility and hysteretic performance.
Available online 2 February 2012
A constitutive model was further established for uniaxial cyclic loading and implemented in ABAQUS
through its user subroutine interface – UMAT. After justification of the modeling results by the experi-
Keywords:
mental measurements from various loading conditions, the proposed model was applied in nonlinear
High-strength steel
Hysteresis performance
time history analysis for steel frames using fiber beam element method. Both experimental and modeling
Cyclic loading results showed that the responses of high-strength structural steel under cyclic loading and monotonic
Material constitutive model loading were different and the necking and fracture behavior would occur in advance for the former
Finite element analysis because the accumulated damages reduced the ductility of steel.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction constitutive model of high-strength structural steel under cyclic


loading forms the basis for further structural seismic response
High-strength structural steel (standard value of strength modeling (see the path of research process in Fig. 1).
P 460 MPa) has been increasingly used in high-rise buildings, Steel stress–strain relationship is commonly defined as bilinear
bridges and long-span and space structures, such as Landmark or tri-linear model in literatures [8,9]. Those models, however, can-
Tower in Japan, Dusseldorf–Ilverich bridges in Germany and not satisfy cyclic loading conditions [10]. More accurate models
National Stadium (Bird’s Nest) in China, providing great economic have been developed to characterize the material mechanical
and social benefits [1,2]. The change from normal to high-strength behavior for such loading conditions. A three-parameter model
structural steel would inevitably lead to differences in structural was proposed by Ramberg and Osgood [11] and widely used for
seismic performance. Seismic scenarios correspond to a reversed metal materials to describe the envelop curve of stress–strain rela-
cyclic loading in material level, which is also called extremely low tionship in cyclic loading scenarios. Nathaniel and Krawinkler
cycle fatigue [3–6]. Experimental studies of mechanical responses [12,13] developed a mathematical model for uniaxial cyclic
of high-strength structural steel under reversed cyclic loading stress–strain behavior of steel, with an assumption of stress
condition provide basic knowledge for evaluating its seismic bounds resulted from stress hardening, softening and mean stress
performance. relaxation. Chaboche [14,15] proposed a cyclic constitutive model
Those experimental results also need to be characterized for fur- to describe the cyclic behavior based on elastic-plasticity theory.
ther implementation of numerical modeling in structural level. In Atkan et al. [16] and Wang et al. [17] gave a mathematical model
most cases, analysis of full scale three-dimensional finite element for steel rebar under reverse loading according to test results,
model is time-consuming and inconvenient to extract internal which may also be applicable for structural steel. Dong and Zhang
forces. With the help of fiber beam method [7], ‘‘line’’ finite element [18] summarized cyclic stress–strain relationship of structural
model is often used in seismic analysis, as illustrated by the proce- steel in both uniaxial and three-axial loading conditions. Shen
dure shown in Fig. 1. In this procedure, uniaxial stress–strain et al. [18,19] developed a model to include both monotonic and
cyclic behavior with a consideration of steel yield plateau. Maria
⇑ Corresponding author at: Department of Civil Engineering, Tsinghua University, et al. [21] developed a methodology for knee braced steel frames
Beijing 100084, China. Tel.: +86 10 62797420; fax: +86 10 62788623. by means of the kinematic theorem of cyclic plastic collapse.
E-mail address: shigang@tsinghua.edu.cn (G. Shi). Nogueiro et al. [22] verified a hysteretic model considering the

0141-0296/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.12.018
2 G. Shi et al. / Engineering Structures 37 (2012) 1–13

Nomenclature

Re l lower yield strength n0 cyclic hardening index


Rm tensile strength r~ r=ry stress of regularization
A elongation after fracture ~e e=ey strain of regularization
Jk impact energy ~
K cyclic hardening factor of regularization
Es young’s modulus ~
n cyclic hardening index of regularization
ry =fy yield stress k1 ; k2 ; k3 control the shape of monotonic loading curve
ey yield strain a; b0 ; b1 ; b2 control the shape of hysteresis skeleton curve
fu ultimate stress Ek tangent stiffness
fu2 nominal fracture stress (when the strain is 0.48) e0 strain of start point on e-axis
e1 strain as ultimate stress fu r0 stress of start point on e-axis
eu fracture strain as fracture stress fu2 ep strain of end point on hysteresis skeleton curve
fu1 15% of fu rp stress of end point on hysteresis skeleton curve
eu1 strain when stress is 15% of fu g proportional coefficient
En hysteretic energy until the strain is eu3 (gauge is 20 mm) ½K e  element stiffness matrix
Np plastic hysteresis loops ½K global stiffness matrix
De total strain amplitude ½D Jacobian matrix
Dee elastic strain amplitude DP load increment
Dep plastic strain amplitude fc concrete strength under compression
Dr stable stress amplitude
K0 cyclic hardening factor

Analysis process
Seismic Input
Research process

Actual project Substructure Member Section Material


2
Beam
Δ 1

2 2 Steel
Wall Connection 1 1

Column
Concrete

Fig. 1. Analysis and research process for multi-floor structure systems.

Table 1
Mechanical properties of Q460D steel.

Rel (MPa) Rm (MPa) A (%) Cold-formed behavior (180°) Jk (J)


466 568 24 d = 2a 281 281 332

Table 2
Chemical composition in percentage of Q460D steel.

C Si Mn P S Alt Alsol V Nb Ti CEQ pcm N


0.054 0.20 1.51 0.008 0.0013 0.031 0.028 0.036 0.072 0.011 0.35 0.15 39.4

pinching phenomenon, which could be used in steel connection model of high-strength structural steel under cyclic loading and
simulation. Nip et al. [23] have carried out a series of hollow sec- further demonstrate its applicability in the analysis of structural
tion specimens to study the cyclic response of tubular bracing system.
members of three structural materials and numerical modeling In this paper, a series of Q460D steel specimens are examined
has been performed. Cyclic constitutive models of concrete and under various loading patterns to evaluate its stress–strain rela-
composite sections have also been developed by other researches tionship, hysteretic behavior and energy dissipation capacity.
[7,18,24–28] and been considered for structural steel. These math- Based on the experimental results, a constitutive model is estab-
ematical characterizations mentioned above, however, were only lished for high-strength structural steel under uniaxial cyclic load-
justified for structural steel with normal strength not for high- ing to describe the monotonic loading curve, hysteresis skeleton
strength structural steel. Furthermore, those models have seldom curve and hysteresis criterion. This constitute model is further
been implemented in structural level, probably due to their implemented into finite element software ABAQUS by its user sub-
complicated expressions and considerable computation cost. It be- routine interface – UMAT. It is demonstrated that the model pro-
comes necessary to develop an efficient but accurate constitutive posed in this paper compares well with the experimental data
G. Shi et al. / Engineering Structures 37 (2012) 1–13 3

Extensometer prescribed in Chinese national standard [29]. Rel is the yield strength
and given as 460MPa for steel Q460 in the standard [29]. Q460 steel
R50 has three quality grades, as represented by ‘‘C’’, ‘‘D’’ and ‘‘E’’. Q460D

14

10 14 10
has the medium quality, which means medium impact toughness

34
behavior.
Effective length 14 Totally, 17 dog-bone specimens of Q460D structural steel were
60 30 20 30 60 prepared according to ASTM E466 [30] and the resulting dimension
200 was shown in Fig. 2. It should be noted that in order to minimize
the local buckling effects and achieve a larger axial compressive
Fig. 2. Geometric details of specimen (unit in mm). strain; a relatively smaller length-to-width ratio should be adopted
for specimen preparation. With the consideration of extensometer
installation and loading facilities, the length-to-width ratio was
Actuator taken as 1.43 (being the ratio of 20 mm to 14 mm, see Fig. 2) in this
Specimen study, which is much smaller in comparison to a value of 2.0 sug-
gested in ASTM E466 and E606 [30]. The tests were stopped pre-
maturely once any local buckling occurred therefore no local
buckling behavior was involved in the following data interpreta-
+ tion. The specimens were loaded in an INSTRON machine (Model
- 1343, with universal tension, compression and torsion fatigue
capacity) as shown in Fig. 3. Strain was measured by tension and
compression extensometer with a gauge length of 20 mm (see
Q460D
Extremely Low C ycle Fatig ue Tests o n hig h -performan ce
Fig. 3). The range of the extensometer is 50% in tension and 25%
structural steel Q 46 0D

Load cell Extensometer in compression.


A series of loading patterns were programmed in a displace-
ment control mode as summarized in Fig. 4. The displacement con-
trol was achieved through the extensometer measurements, i.e.
Fig. 3. Experimental setup and instrumentation.
strain values were obtained from the displacements measured by
from both material level with various loading conditions and struc- the extensometer and then divided by the gauge length 20 mm,
tural level. In addition, structural models are established for steel and those values were used to control the experimental program.
frames subjected to severe earthquake in order to conduct nonlin- In addition to those in Fig. 4, H1-1–3 (three specimens) and H2-
ear time–history analysis using the proposed constitute model, and 1–2 (two specimens) (see Table 3) were respectively defined for
then the resulting structural seismic responses are compared with monotonic tension and compression loading. By studying material
those based on the traditional bilinear constitutive model. mechanical responses under a variety of cyclic loading conditions,
cyclic mechanical properties of steel were obtained and discussed
below. They were further used to validate a constitutive modeling
2. Experimental study for high-strength structural steel under cyclic loading condition in
Section 3.
2.1. Experimental description
2.2. Monotonic properties
The mechanical properties and the chemical composition of
Q460D steel provided by the manufacturer (the mill certificate data) Table 3 and Fig. 5a demonstrated that the elongations (more
were shown in Tables 1 and 2, those values satisfy the requirements than 25%) of high-strength structural steel were satisfactory in

3.4
Step 0.5 Step 0.5
2.0
ε(%)
ε(%)

ε(%)
ε(%)

t t t t
1.0 1.0 1.0
H3-1 (twice) H3-3 (once)
H3-2 (once) H3-4 (twice) H4-1 (once) H5-1(once)
2.0 5.0 Step 0.5 Step 0.5
Step 0.5 1.5 2.0
ε(%)

ε(%)
ε(%)

ε(%)

t t t t
-2.0

H5-2 (once) H6-1 (once) H7-1 (once) H9-1 (once)


Step 0.5
4.0 “Once” means that each strain amplitude (peak
Step 1.0
strain) was achieved only one time
1.5 “Twice” means that each strain amplitude was
1.0 achieved twice, i.e. each loading cycle was
repeated before the new loading cycle .
H7-2 (once) t H8-1 (once) t

Fig. 4. Cyclic loading patterns.


4 G. Shi et al. / Engineering Structures 37 (2012) 1–13

Table 3 half circle; while after the yield point, stresses dropped a little
Main mechanical properties of coupon specimens. and reached a short platform. With the increase of cycles, cyclic
Type Es fy fu2 fu el ð%Þ Failure mode hardening occurred and the expected yield phenomenon was not
ðMPaÞ ðMPaÞ ðMPaÞ ðMPaÞ observed at a certain point. In addition, the increase of stress be-
H1-1 210,000 469 418 565 10.48 Fracture came not obvious afterwards. The results of H7-1, H7-2 and H9-1
H1-2 185,000 461 416 560 12.56 Fracture demonstrated a correlation between the shapes of hysteresis loops
H1-3 208,000 467 410 580 14.07 Fracture and the strain amplitudes, showing that the center of the hystere-
Mean 201,000 466 415 568 12.37 Fracture
sis loop changed with the strain amplitude. The reloading stiffness
H2-1 189,000 496 – 621 7.26 Buckling
H2-2 191,000 494 – 601 6.94 Buckling of H8-1 was nearly identical to the unloading stiffness, and both
Mean 190,000 495 – 611 7.10 Buckling could be represented by the initial elastic modulus.
As shown in Fig. 6, all the specimens achieved full hysteresis
comparison to Chinese national standard (17%) [29]. As shown in loops, indicating good seismic performance and energy dissipation
Table 3, the fracture stress fu2 (415 MPa) of Q460D steel was capacity. It was evident that the behavior of the subsequent cycle
dropped by more than 27% compared with the maximum stress depended on stress, strain amplitude and cycle number of previous
fu (568 MPa). Yield strength and maximum strength values under cycles. Q460D high-strength steel exhibited similar cyclic charac-
compression were greater than those under tension, while the teristics as those of normal strength steel including the plastic, cyc-
strain of the maximum strength in compression was less than that lic hardening or softening, average stress relaxation, Bauschinger
in tension. The uniaxial compression tests were stopped prema- effect, etc., which were closely related to the plastic loading history
turely once any local buckling occurred (see Fig. 5b). In another and would gradually appear during the cyclic loading process [18].
words, the ductility of steel in compression was compromised by The hardening behavior of structural steel combined with isotropic
the buckling behavior, this behavior however corresponded to and kinematics hardening characteristics was also identified from
the response in structural level rather than that in material level. Fig. 6.
The difference between monotonic curve and hysteresis curve is Hysteresis skeleton curve as an envelope of stress–strain curves
shown in Fig. 5c. In cyclic loading scenario (H3-1), the hardening or could indicate the difference of material response under cyclic
softening effect caused the change of stress–strain relationship un- loading and monotonic loading. Ramberg–Osgood model [11]
til its stabilization. Although monotonic loading curve was often established a hysteresis skeleton curve and commonly used for
used in engineering calculations as steel constitutive relationship, cyclic loading condition as shown in Eq. (1). The Ramberg–Osgood
it was demonstrated that steel hysteresis and monotonic constitu- model was adopted to fit the experimental results for Q460D steel
tive curves were quite different (see Fig. 5c), for example a more (see Fig. 7) and the related parameters are given in Table 4.
obvious hardening phenomenon resulted from cyclic loading. The
damage accumulation by cyclic effect always leaded to a reduction   10
of steel ductility, and the rate of stress decrease after peak point De Dee Dep Dr Dr n
¼ þ ¼ þ ð1Þ
was significantly increased. 2 2 2 2E 2K 0
2.3. Hysteretic behavior Table 4 and Fig. 7 demonstrated a satisfactory performance of
Ramberg–Osgood model [11] in the fitting of hysteresis skeleton
Fig. 6 shows the hysteretic curves of Q460D steel under various curves; again an obvious difference can be seen in Fig. 7 between
cyclic loading conditions. H3-1–4 achieved the yield point in first the skeleton curve and the monotonic curve.

Fig. 5. Monotonic and cyclic loading curves of Q460D steel.


G. Shi et al. / Engineering Structures 37 (2012) 1–13 5

Fig. 6. Hysteretic loading curves of typical Q460D coupon specimens.

Table 5 summarizes the main mechanical properties of the and the fracture strain was more than 45%. The level of cross-section
tested specimens from all the loading scenarios. The Chinese seis- reduction also highly indicated good ductility of structural steel.
mic design code [32] prescribes that the structure fails when its Local buckling was one failure mode deliberately avoided in this
load carrying capacity drops by 15%, accordingly Table 5 also pre- study, which caused the degradation of the strength and stiffness
sents the stress fu1 (as 85% of the strength fu) and the correspond- in macro-curve [31] and corresponded to a failure mode in structural
ing strain eu1. It can be seen that the peak point and the point with level rather than that in material level.
15% capacity reduction after cyclic loading are quite different from Electron microscope scanning tests were carried out for speci-
those under monotonic loading (see Fig. 6c and Table 5). The strain men fracture surfaces. It can be seen from micro-failure modes in
eu1 of Q460D is reduced by maximum 24% after cyclic loading (H5- Fig. 8b that all the three specimens showed a certain amount of
1). It therefore can be concluded that the ductility of steel material round and ellipse dimples. This phenomenon suggested a ductile
decreases significantly after cyclic loading, and low cycle fatigue failure according to [33].
could induce damage accumulation and structure deterioration,
therefore result in structural failure in advance.
3. Cyclic constitutive model

2.4. Failure modes The simplified constitutive model for high-strength steel under
uniaxial cyclic loading consists of three parts according to [10]:
Fig. 8a demonstrated two typical forms of failure modes in macro monotonic loading curve, cyclic skeleton curve and hysteresis cri-
level – fracture in tension and buckling in compression. The series of terion. These three parts can basically describe the cyclic behavior
specimens did not show significant sound when fracture occurred and hardening characteristic of steel.
6 G. Shi et al. / Engineering Structures 37 (2012) 1–13

Fig. 7. Comparison of cyclic skeleton curves using Ramberg-Osgood model to experimental results.
G. Shi et al. / Engineering Structures 37 (2012) 1–13 7

Table 4 stage, yield platform stage, yield hardening stage and second yield
Parameters used in modeling of cyclic hardening. platform stage. In order to consider the yielding and hardening
Type Es K0 n0 behavior during a monotonic loading, a quadratic parabola was
ðMPaÞ ðMPaÞ used to fit the shape of the hardening stage. An example was given
H3-1 202,000 677.20 0.0541 in Fig. 9 for tensile loading and the mathematical expression was
H3-2 201,000 728.55 0.0869 specified in Eq. (2).
H3-3 203,000 770.48 0.0904
8
H3-4 203,000 827.43 0.1143 > Es e ðe  ey Þ
H4-1 201,000 662.71 0.0515
>
>
>
>
H6-1 201,000 807.94 0.1001 < fy ðey  e  k1 ey Þ
H7-1 201,000 678.37 0.0623 r¼ Es ð1k3 Þ
 2 ð2Þ
>
> k3 fy þ e  k2 ey ðk1 ey  e  k2 ey Þ
H7-2 204,000 673.92 0.0614 >
> ey ðk2 k1 Þ2
Mean 202,000 736.01 0.0776 >
:
fu ðe  k2 ey Þ
Eq. (2) was used to fit the experimental results of Q460D steel
Table 5 (see Fig. 10a), the resulting values for the parameters k1, k2 and
Summary of major material characteristics from experiments. k3 are 5, 46, and 1.2, respectively. The corresponding modeling
Type fy fu el ð%Þ fu1 eu1 ð%Þ Np En curves were compared with the experimental data in Fig. 10a, indi-
ðMPaÞ ðMPaÞ ðMPaÞ ðMPaÞ cating a satisfied fitting based on the selected parameters.
H1-1 469 565 10.48 476 40.88 Monotonic 848,043
H1-2 461 560 12.56 476 38.84 Monotonic 803,315 3.2. Hysteresis skeleton curve
H1-3 467 580 14.07 493 42.09 Monotonic 894,172
H2-1 496 621 7.26 – – Monotonic 146,957
H2-2 493 601 6.93 – – Monotonic 152,806
It was shown that Ramberg–Osgood model [11] fitted well with
H3-1 476 576 10.58 489 33.54 16 3,348,976 test data in Section 2.3. However, in this model, stress is the inde-
H3-2 466 580 8.88 492 32.90 8 1,652,834 pendent variable and strain is given as a function of stress, and it
H3-3 480 575 8.84 482 36.59 7 1,490,138 is difficult to obtain an explicit analytical expression of stress as a
H3-4 479 596 7.22 505 33.06 10 737,320
function of strain. This fact significantly increases calculation work
H4-1 473 562 11.56 476 34.36 15 3,263,592
H5-1 460 562 13.22 478 33.00 11 1,533,270 when a strain value is given (for example measured from experi-
H5-2 464 582 8.91 491 31.48 20 2,190,953 ments) to initiate the program. According to reference [10], a more
H6-1 474 571 7.28 487 32.38 6 1,072,960 efficient two-stage cyclic skeleton curve is proposed in the follow-
H7-1 482 575 7.696 488 36.03 14 2,104,424 ing and its performance is evaluated by the comparison with the
H7-2 485 565 4.85 478 15.92 12 1,571,350
H8-1 461 569 17.22 485 38.32 8 263,409
experimental results in Fig. 10b. In the proposed curve, the first
H9-1 467 555 2.32 472 34.08 7 1,064,171 stage is elastic stage before yield and the second stage is the cyclic
hardening stage, as given in Eq. (3). Parameters a, b0, b1, b2 are the
fitting parameters. Considering that a = 7.123, b0 = 2.301, b1 = 0.745
and b2 = 0.0004 (determined by data fitting), the resulting model-
3.1. Monotonic loading curve ing curve compares well with the experimental curves in Fig. 10b,
where the cyclic hardening stage of the hysteresis skeleton curve
Monotonic loading curve adopted the form of quadratic plastic can be well described. In order to illustrate a clear hardening rela-
flow model proposed by Esmaeily and Xiao [26], where stress– tionship, normalized stress (r ~ = r=ry ) and strain (~e = e=ey ) are used
strain relationship of high-strength steel was divided into elastic in this model as shown in Eq. (3).

1) Fracture 2) Buckling
(a) Failure modes in macro scale

60μm

1) H1-3 Monotonic loading 2) H3-1 Gradual amplitude 3) H5-2 Equal amplitude


cyclic loading cyclic loading
(b) Failure modes in micro scale

Fig. 8. Observation of failure modes.


8 G. Shi et al. / Engineering Structures 37 (2012) 1–13

d2 ¼ ðEs  Ek Þðe  e0 Þ ð6Þ


The proportional coefficient g as defined in Fig. 11a was
obtained by fitting with the experimental data according to refer-
ence [10] and considering that 0 6 g 6 1 (see Fig. 11a).
In case that the strain amplitude is quite small, the reloading
curve may exhibit excessive softening according to Eq. (4) (for
example, the reloading curves of first two cycles are quite different
from the last two cycles). Therefore, g of reloading curve needs to
be correlated to the strain amplitude (ep  e0 ). Based on the exper-
imental phenomenon, g was expressed in Eq. (7) with two stages
according to the value of strain amplitude,
8  
< 1:048  eeee0 þ0:05 ðep  e0  P 0:04Þ
> 0:05
p 0
g¼   ð7Þ
>
: 1:074  ee0:08 0
ðep  e0  6 0:04Þ
ep e0 þ0:08

Fig. 11b summarizes the above hysteresis criterion, where the


Fig. 9. Monotonic and hysteretic models of steel stress–strain relationship.
stress and strain couple of the start point on tension/compression
( reloading curve for the i-th loop is denoted as (r0,it/c, e,it/c) on the x-
~e ð~e  1Þ axis, the end point (ep,it/c, rp,it/c) is on the curve and points to the
r~ ¼ ð~eþaÞ ð3Þ
ðb0 þb1 ð~eþaÞþb2 ð~eþaÞ2 Þ
ð~e  1Þ previous peak point in the same direction (tension or compres-
sion). The unloading point is denoted as (ru,it/c, eu,it/c). Firstly, the
point goes along the monotonic curve until reaches the first
3.3. Hysteresis criteria unloading point (ru,1t, eu,1t), then elastic unloading (according to
elastic modulus Es) occurs until the first reloading point in com-
Hysteresis criterion of high-strength structural steel mainly in- pression (r0,1c, e0,1c). The reloading process is described by the
cludes four components: the first time loading curve, the unload- reloading criteria in Fig. 11a, pointing to the first peak point yc
ing curve, the reloading direction and the reloading curve. Those (rp,1c, ep,1c). After the peak point, the curve goes along the hystere-
four components are defined below: (1) The first time loading of sis skeleton curve until unloading point (ru,1c, eu,1c), which is also
steel goes along the uniaxial monotonic loading curve including the peak point (rp,2c, ep,2c) in the next cycle. The second reloading
tension and compression. (2) The unloading curve of steel follows process ranges from point (r0,2t, e0,2t) to point (rp,2t, ep,2t) and
linear elasticity line and the unloading stiffness is represented by ep,2t is equal to eu,1t. After the point (rp,2t, ep,2t), the curve follows
the initial elastic modulus Es. (3) For the reloading direction of the hysteresis skeleton curve until the point (ru,2t, eu,2t), which is
steel, it was suggested by Légeron et al. [27] that the reloading the same as the previous process.
curve developed towards the previous peak point in the same
direction. This approach, as called ‘‘peak-orientation model’’, has 4. Model implementation
its simplicity and been widely used in constitutive models [6,20].
(4) The reloading curve of steel is according to reference [10]. 4.1. Define material constitutive model in ABAQUS
The corresponding stiffness, as the slope of the reloading curve,
is between the unloading stiffness and the tangent stiffness and In seismic design, beam element model integrated with the
determined by the starting point (r0,e0) to the end point (ep,rp) fiber beam method is often used for its efficiency and convenience
using ‘‘peak-orientation criteria’’ shown in Fig. 11. In this way, [7], while an accurate uniaxial constitutive model has to be defined
Eq. (4) is formulated to identify stress points on the reloading in the fiber beam method.
curve, where d1 and d2 are defined by Eq. (5) and Eq. (6) and cor- The proposed constitutive model for high strength steel sub-
responds to the distances shown in Fig. 11a. jected to uniaxial cyclic loading was formulated as a user-defined
(
r ¼ d1  gd2 ¼ Es ðe  e0 Þ þ r0  gðEs  Ek Þðe  e0 Þ material using FORTRAN and ABAQUS UMAT (User-defined Mate-
rp r0 ð4Þ rial Mechanical Behavior) [35]. ABAQUS main program called
Ek ¼ ep e0 UMAT subroutine at each material integration point, before each
where load increment. In this way, the element stiffness matrix [Ke] was
integrated with the Jacobian matrix [D] calculated by calling
d1 ¼ Es ðe  e0 Þ ð5Þ UMAT. All the element stiffness matrices then integrated the global

Fig. 10. Comparison of experimental curves with modeling results proposed.


G. Shi et al. / Engineering Structures 37 (2012) 1–13 9

Start

ABAQUS Pre-processing

Input Es , fy and other parameters of steel

Solution, P incremental steps starting

Strain, strain increment and the current


stress of integration points are given No satisfaction,
change the strain
increment

UMAT and Jacobian


Proposed matrix [D] and updated stress Check the
are obtained and state Balance
model
equation
variables are stored
Satisfaction

Element stiffness matrix [Ke] is


obtained through the section Current incremental
integration points integrating step ends
next incremental
step starts
Integrating global stiffness matrix [K]

Fig. 12. Flow chart for programing of model proposed in ABAQUS.

curves were presented in Fig. 13 and good agreements can be


found with all the experimental scenarios.

4.3. Validation in structural level

The developed constitutive model can be easily implemented


into a finite element program to describe structural responses
under cyclic loading. Experiments were carried out by Li et al.
[37] on a steel beam to column system and the details of specimen
dimensions and material properties were shown in Fig. 14a. The
steel beam and column was modeled in ABAQUS where the mate-
Fig. 11. Reloading and hysteresis criteria developed in [10]. rial was defined according to the proposed constitutive model. The
boundary condition was specified based on the experimental set-
up, i.e. two pin supports for the ends of column and a free end of
stiffness matrix [K] of the structure. Subjected to a current load
the beam (see Fig. 14a). The beam-to-column joint was considered
increment DP, [K] was identified when the resulting structural load
to be rigid because of its small shear deformation.
converged to the applied load, then the next incremental step was
The resulting load displacement curve at the beam free end was
initiated. The program flow charts and process of UMAT in ABAQUS
shown in Fig. 14b with a comparison to the experimental results,
were shown in Fig. 12.
where a good agreement was found. In addition, such a numerical
model was analyzed again based on the material bilinear constitu-
4.2. Validation in material level tive relationship and the calculated load displacement curve is
compared with experimental results in Fig. 14c. Through a compar-
The comparisons of proposed model and experimental data of ison of Fig. 14b and c, a superior performance was found in Fig. 14b
high-strength steel in this paper are shown in Fig. 13a and a good for the modeling curve using the proposed constitutive model,
agreement can be found. especially in the final stages of cyclic loading.
Experimental results of high strength steel under cyclic loading A composite beam to steel column system was further selected
from Peter et al. [34] were adopted to verify the applicability and to demonstrate the performance of the proposed constitutive mod-
accuracy of the proposed model. In this study, HT440 steel as used el in structural level. The experimental scenario was introduced by
in structural steel construction in Japan and A709 steel as devel- Su et al. [38]. As shown in Fig. 15a, the structure was subjected to
oped through joint efforts of the American Iron and Steel Institute, cyclic loading at the two ends of composite beam (see the beam
were included. Details of specimen dimensions, constraints condi- section in Fig. 15a). The composite beam and steel column were
tion and material properties are shown in Table 6 and Fig. 13b. modeled in ABAQUS using the method of fiber beam element,
Parameters a, b0, b1, b2 of hysteresis skeleton were determined in and the beam-to-column joint was treated as rigid region. The
the hysteresis curves: for HPS485, the parameters were the same boundary condition was defined according to the experimental set-
as those used in current study; because of more cyclic hardening, up (see Fig. 15a). The proposed constitutive model and bilinear
the parameters were determined as a = 5.796, b0 = 2.486, model were implemented into the finite element program and
b1 = 0.608 and b2 = 0.00046 for HT440. The resulting stress–strain the modeling results were compared with the experimental
10 G. Shi et al. / Engineering Structures 37 (2012) 1–13

Fig. 13. Comparison of proposed modeling results with experimental data from current study and literature in material level.

proposed model were in a better agreement with experiments


Table 6 for both loading and reloading processes. The Proposed model pre-
Main mechanical properties of specimens presented in [34].
dicted well the structural hysteresis behavior.
Type Es fy fu
ðMPaÞ ðMPaÞ ðMPaÞ
4.4. Full-scale frame analysis
HPS485 201,300 503 590
HT440 208,200 501 688
In order to examine the feasibility of the proposed model for the
static and seismic analysis of steel frame, a two-floor high-strength
steel frame model was established in Fig. 16 for quasi static analysis
measurements in Fig. 15b and c, respectively. As evidenced in the and a ten-floor frame model was established in Fig. 17 for nonlinear
comparison of these two figs, the results calculated from the time history analysis. The parameters used in Eqs. (2)–(7) for Q460D

Fig. 14. Comparison of proposed modeling results with experimental data from [34] for a steel beam to steel column system using beam element method.
G. Shi et al. / Engineering Structures 37 (2012) 1–13 11

Fig. 15. Comparison of proposed modeling results with experimental data from [35] for a composite beam to steel column system using fiber element method.

Fig. 16. Comparison analysis between proposed model and bilinear model under cyclic loading for a two-floor high strength steel frame.

Fig. 17. Comparison analysis between proposed model and bilinear model under seismic loading for a ten-floor high strength steel frame.
12 G. Shi et al. / Engineering Structures 37 (2012) 1–13

steel were inputted into the program developed in Section 4.1. For is found with experimental results. The constitutive model
the two-floor frame model, the geometries and loading pattern were used in the structural nonlinear time-history analysis affects
specified in Fig. 16a, according to design requirements of Chinese obviously the resulting structural seismic responses.
national standards [32,36]. Calculations were carried out based on (4) The proposed high-strength structural steel uniaxial consti-
both bilinear kinematic hardening finite element model and the tutive model is developed as user-defined material based on
model proposed in this paper, respectively, comparative results the user subroutine interface UMAT in ABAQUS. Accuracy
were shown in Fig. 16b. It was found from the figures that the pro- and applicability of proposed models are verified by typical
posed model and bilinear Kinematic hardening were different experiments respectively in material and member levels. It
mainly on two aspects: (1) the phenomenon of cyclic hardening is also demonstrated that that the proposed model can be
was more obviously illustrated by the proposed constitutive model; conveniently implemented in nonlinear time history analy-
and (2) a plump hysteresis curve could be avoided in the reloading sis of steel system under earthquake and provide more accu-
process using the proposed model. rate results of seismic analysis.
For the ten-floor frame model, the height of each story was
3200 mm and each span length was specified as 6450 mm. Beam
and column sizes were shown in Fig. 17a, where the size bottom
column section increased according to the actual case. The first Acknowledgment
step was to apply vertical gravity loads and the load combination
of superimposed dead and live loads on the structure according This work was supported by the Open Fund of State Key Labo-
to [36]. For seismic analysis, the superimposed dead loads should ratory for Disaster Prevention in Civil Engineering (Grant No.:
be imposed as mass points on beams. The second step was to input SLDRCE08-TS-01).
ground motion acceleration to the structure in horizontal direction
according to Wenchuan earthquake wave and then elastic-plastic References
time-history analysis was initiated.
In order to compare the structural behavior under severe earth- [1] Shi G, Wang YQ, Shi YJ. Behavior of high strength steel column under axial
quake with high geometrical and material nonlinearity, the seismic compression. J Build Struct 2009;30(2):92–7 [in Chinese].
[2] Ricles JM, Sause R, Green PS. High-strength steel: implications of material and
peak acceleration (PGA) was adjusted to 1.0g. From the floor shift geometric characteristics on inelastic flexural behavior. Eng Struct 1998;20(4–
distribution (Fig. 17b, c and d), it was found that the results of 6):323–35.
the proposed model and bilinear Kinematic hardening were differ- [3] National Natural Science Foundation Committee. Subject development
strategy research report – construction, environment and civil engineering II.
ent (the maximum difference of xxx was more than 100%), indicat-
Beijing: Science Press; 2006 [in Chinese].
ing that constitutive model of high-strength structural steel under [4] Masatoshi K. Extremely low cycle fatigue life prediction based on a new
cyclic loading could significantly affect the seismic behavior of the cumulative fatigue damage model. Int J Fatigue 2001;24(6):699–703.
structure. Through these examples, the importance of an accurate [5] Xue L. A unified expression for low cycle fatigue and extremely low cycle
fatigue and its implication for monotonic loading. Int J Fatigue 2008;30(10–
constitutive model was demonstrated in the structural seismic 11):1691–8.
design and analysis. [6] Nip KH, Gardner L, Davies CM, et al. Extremely low cycle fatigue tests on
structural carbon steel and stainless steel. J Constr Steel Res
2010;66(1):96–110.
[7] Menegotto M, Pinto PE. Method of analysis for cyclically loaded reinforced
5. Conclusions concrete plane frames including changes in geometry and non-elastic behavior
of elements under combined normal force and bending. In: Proceedings, IABSE
symposium on resistance and ultimate deformability of structures acted on by
A series of experiments were conducted on high-strength steel well defined repeated loads. Lisbon; 1973. p. 15–22.
Q460D subjected to different loading patterns, including mono- [8] Hodge PG. The theory of piecewise linear isotropic plasticity. In: Proceedings,
tonic loading and cyclic loading. A simplified constitutive model colloquium on deformation and flow of soild. Madrid, Spain; 1955. p. 147–69.
[9] Prager W. The theory of plasticity: a survey of recent achievement. In:
of high-strength structural steel under unaxial cyclic loading was
Proceedings of the institute of mechanical engineers, vol. 169. London,
proposed including monotonic loading curve, hysteresis skeleton England; 1955. p. 41–57.
curve and hysteresis criterion. The constitutive model was vali- [10] Shi YJ, Wang M, Wang YQ. Experimental and constitutive model study of
structural steel under cyclic loading. J Constr Steel Res 2011;67(8):1185–97.
dated in both material and structural levels and further integrated
[11] Ramberg W, Osgood WR. Description of stress–strain curves by three
into structural nonlinear time-history analysis to predict seismic parameters. National Advisory Committee for Aeronautics, TN 902; 1943.
responses of steel frames made using the high-strength steel. [12] Krawinkler H, Zohrei M, Bahman LI, et al. Recommendations for experimental
Based on this work, the following conclusions can be drawn: studies on the seismic behavior of steel components and materials. Stanford,
CA: Earthquake Engineering Center, Department of Civil and Environmental
Engineering; 1983. p. 61.
(1) Specimens failed still in a ductile manner under a number of [13] Nathaniel GC, Krawinkler H. Uniaxial cyclic stress–strain behavior of [J]. J Eng
cyclic loadings, which indicates high-strength structural Mech 1985;111(9):1105–20.
[14] Chaboche JL. Time independent constitutive theories for cyclic plasticity. Int J
steel Q460D owns good ductility and energy dissipation Plast 1986;2(2):149–88.
capacity. Compared with normal strength steel, Q460D [15] Chaboche JL. Constitutive equations for cyclic plasticity and cylic
high-strength steel exhibit similar cyclic characteristics such viscoplasticity. Int J Plast 1989;5:247–302.
[16] Atkan AE, Karlson BI, Sozen MA. Stress–strain relationships of reinforceing
as plasticity, cyclic hardening or softening, average stress bars subjected to large strain reversals. Civil engineering studies, SRS No. 397,
relaxation and Bauschinger effect. University of Illinois, Urbana-Champaign, Ill; 1973.
(2) Cyclic loops and the strain amplitudes significantly affect the [17] Wang XM, Xu B, Shen JM. A constitutive model of reinforcing steel under
reversed loading. J Build Struct 1992;13(6):41–7 [in Chinese].
material ductility. The fracture strain is reduced and necking
[18] Dong YT, Zhang YC. Cyclic plasticity constitutive model of structural steel [J]. J
occurs in advance because of cyclic loading, which indicates Harbin Univ Civil Eng Archit 1993;26(5):106–112 [in Chinese].
that damages are accumulated during cyclic loading. There- [19] Shen C, Tanaka Y, Mizuno E, Usami T. A two-surface model for steels with yield
plateau. J Struct Eng/Earthquake Eng JSCE 1992;8(4):179–88.
fore the fracture strain of steel under cyclic loading cannot
[20] Usami T, Gao S, Ge H. Elastoplastic analysis of steel members and frames
be directly determined by that from monotonic loading. subjected to cyclic loading. Eng Struct 2000;22(2):135–45.
(3) The responses of steel under cyclic loading and monotonic [21] Maria AC, Luigi M, Vincenzo P. Plastic design and seismic response of knee
loading are different. The proposed model in this paper can braced frames. Adv Steel Construct 2009;5(3):343–65.
[22] Nogueiro P, Silva LS, Bento R, Simoes R. Numerical Implementation and
describe the responses of high-strength steel under both calibration of a hysteretic model with pinching for the cyclic response of steel
cyclic loading and monotonic loading and a good agreement joints. Adv Steel Construct 2007;3(1):459–84.
G. Shi et al. / Engineering Structures 37 (2012) 1–13 13

[23] Nip KH, Gardner L, Elghazouli AY. Cyclic testing and numerical modelling of [31] Lee PS, Noh HC. Inelastic buckling behavior of steel members under reversed
carbon steel and stainless steel tubular bracing members. Eng Struct cyclic loading. Eng Struct 2010;32(9):2579–95.
2010;32(2):424–41. [32] Ministry of Housing and Urban-Rural Development of the People’s Republic of
[24] Chen CC, Li JM, Weng CC. Experimental behavior and strength of concrete- China. GB50011-2001 Code for seismic design of buildings. Beijing: China
encased composite beam-columns with T-shaped steel section under cyclic Architecture and Building Press; 2001 [in Chinese].
loading. J Constr Steel Res 2005;61(7):863–1006. [33] Hull D. Fractography: observing, measuring, and interpreting fracture surface
[25] Part R, Kent DC, Sampson RA. Reinforced concrete members with cyclic topography. Cambridge University Press; 1999.
loading. J Sturct Div ASCE 1972;98(ST7):1341–60. [34] Peter D, Ahmad MI, Ian GB. Cyclic response of plate steels under large inelastic
[26] Esmaeily A, Xiao Y. Behavior of reinforced concrete columns under variable strains. J Constr Steel Res 2007;63(2):156–64.
axial loads: analysis. ACI Struct J 2005;102(5):736–44. [35] ABAQUS. Analysis user’s manual I_V. Version 6.9. USA: ABAQUS, Inc., Dassault
[27] Légeron F, Paultre P, Mazars J. Damage mechanics modeling of nonlinear Systèmes; 2009.
seismic behavior of concrete structures. J Struct Eng ASCE 2005;131(6):946– [36] Ministry of Housing and Urban-Rural Development of the People’s Republic of
55. China. GB50009-2001 Load code for design of building structures. Beijing:
[28] Garcia R, Hajirasouliha I, Pilakoutas K. Seismic behaviour of deficient RC China Architecture and Building Press; 2001 [in Chinese].
frames strengthened with CFRP composites. Eng Struct 2010;32(10):3075–85. [37] Shi YJ, Li ZF, Chen H. Experimental research on cyclic behavior of new types of
[29] Ministry of Housing and Urban-Rural Development of the People’s Republic of beam column connections in highrise steel frames. J Build Struct
China. GB/T 1591-2008 High strength low alloy structural steels. Beijing: 2002;23(3):2–7 [in Chinese].
China Architecture and Building Press; 2008 [in Chinese]. [38] Shi YJ, Su D, Wang YQ. An experimental study on the seismic performance of
[30] ASTM, Annual Book of ASTM Standards. American Society for Testing and beam-column joints in steel frames with the effect of concrete slabs
Materials. Philadelphia, PA; 1998. considered. China Civil Eng J 2006;39(9):26–31 [in Chinese].

You might also like