Engineering Structures 46 (2013) 392–406

Contents lists available at SciVerse ScienceDirect

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

Modified Bouc–Wen model for hysteresis behavior of RC beam–column joints

with limited transverse reinforcement
Piyali Sengupta ⇑, Bing Li
School of Civil and Environmental Engineering, Nanyang Technological University, Singapore 639798, Singapore

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

Article history: An analytical approach based on modified Bouc–Wen–Baber–Noori model has been proposed in this
Received 10 September 2011 paper for predicting the hysteresis behavior of reinforced concrete beam–column joints with limited
Revised 20 July 2012 transverse reinforcement. The analytical model presented in this research is able to capture the charac-
Accepted 1 August 2012
teristics of non-seismic detailed beam–column joints such as stiffness and strength degradation and
Available online 14 September 2012
pinching. Livermore Solver for Ordinary Differential Equations (LSODE) and Genetic Algorithm (GA) have
been employed to solve the differential equations and to execute systematic estimation of the parameters
Keywords:
associated with the model respectively. The analytical model has been calibrated with the experimental
Reinforced concrete
Beam-column joints
results of old fashioned interior and exterior beam–column joints obtained from the literature. In a bid to
Analytical modeling examine the influence of variation of each analytical parameter on the model, sensitivity analysis has
Hysteresis been performed. Thereafter, an extensive parametric study has been conducted to relate the physical
Degradation parameters of beam–column joints to the analytical model parameters. The upper and lower bounds
Pinching of the magnitude of the analytical model parameters have been proposed subsequently with a method
to identify the parameters for a specific beam–column joint depending on its physical parameters.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction deformation initiates after section cracks. Degrading Bilinear mod-

el by Imbeault and Neilson [9] is a peak-oriented hysteresis model
Reinforced concrete (RC) structures in regions of low to moder- in which stiffness changes only when the prior maximum is ex-
ate seismicity are designed for gravity loading, not complying with ceeded in any direction. For Q-Hysteresis model by Saidi and Sozen
the modern seismic design codes. RC moment resisting frames [10] and Otani Hysteresis model [11], the primary curve is a bilin-
have little or no transverse reinforcement in the beam–column ear curve with ascending post-yielding branches with inclusion of
joint regions which in turn may not perform adequately to with- stiffness degradation at unloading and load reversal. Hysteresis
stand earthquake-induced actions. In the Padang Earthquake Shear model by Ozcebe and Saatcioglu [12] is based on statistical
(2009), several RC buildings experienced extensive damage at the analysis of past experimental data. Alath and Kunnath model
beam–column joints, as shown in Fig. 1. Although substantial [13] simulates the joint shear deformation by a rotational spring
experimental research has been undertaken by several researchers with degrading hysteresis while Biddah and Ghobarah [14] model
[1–5] till date on RC non-seismic detailed beam–column joints, comprises of separate rotational springs for the joint shear and
there is still a dearth of analytical studies on prediction of their bond-slip deformation. Hwang and Lee model [15] predicts the
hysteresis behavior. shear strength of exterior beam–column joints for seismic resis-
Hysteresis models developed in the past includes Elasto-plastic tance based on softened strut-and-tie model. Youssef and Ghoba-
model by Veletsos and Newmark [6], where the primary force– rah joint element [16] has two diagonal translational springs and
deformation curve is represented by an elastic portion indicating twelve translational springs to simulate the joint shear deforma-
the cracked-section behavior with no incremental stiffness upon tions and bond-slip respectively. Lowes and Altoontash joint ele-
yielding and during unloading. Clough Degrading Stiffness model ment [17] consists of eight zero-length translational springs, a
[7] operates on a bilinear primary curve with ascending post yield- zero-length rotational spring and four zero-length shear springs
ing branches and stiffness degradation during load reversals. Tak- to simulate the bond-slip response of longitudinal reinforcement,
eda model [8] works on a trilinear primary curve representing the joint shear deformation and the interface shear deformation
uncracked, cracked and post-yielding stages where nonlinear respectively while simplified Altoontash joint element [18] has
four zero-length rotational springs at beam–column interfaces
and a rotational spring to simulate the member-end rotations
⇑ Corresponding author. Tel.: +65 96149455. due to bond-slip and the joint shear deformation. Shin and LaFave
E-mail address: piya0002@e.ntu.edu.sg (P. Sengupta).

0141-0296/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.08.003
 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406 393

Notations

m system mass e hysteretic energy

u relative displacement of mass with respect to base dis- fh hysteretic force
placement dm parameter that controls rate of strength degradation
t time dg parameter that controls rate of stiffness degradation
() derivatives with respect to time t f1 parameter that controls magnitude of initial drop in
c viscous damping coefficient slope, dz/du
z hysteretic displacement sgn() Signum function
a a weighing constant that represents relative participa- q parameter that sets pinching level as a fraction of ulti-
tion of linear and nonlinear terms, also known as rigid- mate hysteretic strength, zmax
ity ratio i.e. the ratio of final and initial tangent stiffness f2 parameter that causes the pinching region to spread
k stiffness fs parameter that measures total slip
ki initial tangent stiffness p parameter that controls the rate of initial drop in slope,
kf final tangent stiffness dz/du
FT total non-damping restoring force w parameter that contributes to amount of pinching
F(t) time dependent forcing function dw parameter specified for desired rate of pinching spread
n0 linear viscous damping ratio k parameter that controls rate of change of f2 with change
x0 pre-yield natural frequency of system of f1
f(t) mass normalized forcing function y vector containing the set of Ordinary Differential Equa-
b, c, n hysteresis shape parameter tions
m strength degradation parameter f vector-valued function of y and t
g stiffness degradation parameter ea0 root mean square error
A parameter that regulates the tangent stiffness N number of data points for input displacement function
h(z) pinching function |ea0| maximum root mean square error
zmax ultimate value of z i.e. z at dz/du = 0 fc0 concrete compressive cylinder strength
a0 magnitude of a at zero displacement Ag gross area of beam–column joint
Dmax maximum positive or negative displacement depending
on u > 0 or u < 0 respectively

model [19] proposes the joint as rigid elements along the panel After exploring the above-mentioned analytical models, an utmost
edges with a rotational spring embedded in one hinge linking adja- effort has been undertaken to illustrate the hysteresis behavior of
cent rigid elements and two rotational springs at beam-joint inter- RC non-seismic detailed beam–column joints analytically based
faces to simulate the member-end rotations due to inelastic on modification of Bouc–Wen–Baber–Noori model [21,22]. The effi-
behavior of the beam longitudinal reinforcement and the plastic ciency of the proposed model is then verified by calibrating it with
hinge rotations due to inelastic behavior of the beam separately. the experimental results of interior and exterior beam–column
The analytical model proposed by Favvata et al. [20] assumes the joints, obtained from the literature. Sensitivity of the model due
exterior beam–column joint element as a zero length spring ele- to the variation of each analytical parameter has been investigated
ment which incorporates stiffness degradation and pinching effect and thereafter an extensive parametric study with varying joint
as special rules. physical parameters has been conducted to provide the approxi-
An analytical model of beam–column joint requires a force dis- mate range of magnitudes for the parameters and to determine
placement relationship capable of producing requisite strength an effective way to identify them for any beam–column joint
and stiffness degradation and pinching at all displacement levels. depending on its physical parameters.
This is a stringent requirement considering the numerous parame-
ters contributing to the hysteresis behavior of beam–column joints. 2. Proposed analytical hysteresis model

2.1. Background

With an objective of developing a generic, computationally effi-

cient and mathematically tractable hysteresis model, the basis for
this model is selected as the Bouc, Wen, Baber and Noori (BWBN)
model. Bouc suggested a versatile, smoothly varying hysteresis
model for a single-degree-of-freedom (SDOF) system under forced
vibration. Baber and Wen [22] extended the model to include stiff-
ness and strength degradation as a function of hysteretic energy
while Baber and Noori [21] incorporated pinching in the hysteresis
model. This paper modifies the original BWBN model accordingly
to assimilate hysteresis behavior of reinforced concrete substan-
dard beam–column joints.

2.2. Equation of motion and constitutive relations

Fig. 1. Severe damage to non-seismic detailed beam–column joints (Padang The equation of motion for a single-degree-of-freedom system
Earthquake, 2009). can be expressed as follows:
394 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406

€ þ cu_ þ F T ½uðtÞ; zðtÞ; t ¼ FðtÞ

mu ð1Þ

where u is the relative displacement of mass m with respect to

ground motion and dot () signifies the differential with respect to
time; c is the linear viscous damping coefficient; FT[u(t), z(t), t] is
the non-damping restoring force consisting of the linear restoring
force aku and the hysteretic restoring force (1  a)kz; a is stiffness
ratio i.e. the ratio of final asymptote tangent stiffness kf to initial
stiffness ki and F(t) is the time-dependent forcing function.
Dividing both sides of (1) by m, the following standard expres-
sion is obtained:

€ þ 2n0 x0 u_ þ ax20 u þ ð1  aÞx20 z ¼ f ðtÞ

u ð2Þ
pffiffiffiffiffiffiffiffiffiffiffi
where n0 is the linear damping ratio,
pffiffiffiffiffiffiffiffiffiffi
ffi c=2 ki =m; x0 is the pre-yield Fig. 2. Effect of constant and displacement based a in hysteresis loop.
system natural frequency, ki =m; f(t) is the mass-normalized forc-
ing function. 2.2.2. Hysteresis shape parameters
Hysteretic restoring force is a function of hysteretic displace- Three hysteresis parameters b, c and n and their interactions
ment z and thus the relationship between z and u is shown in determine the basic hysteresis shape. The absolute values of b
the following expression: and c inversely influence hysteretic stiffness and strength, as well
as smoothness of the hysteresis loops. During loading, parameter n
_ n1 z þ cujzj
Au_  mðbjujjzj _ nÞ controls sharpness of the transition from initial to asymptotic
z_ ¼ hðzÞ ð3Þ slope. For n = 1, the relationships between b and c and their effects
g
on hysteresis are described below and shown in Fig. 3.

where b, c and n are hysteretic shape parameters; A determines the bþc>0
tangent stiffness; m and g are the strength and stiffness degradation  Weak Softening
cb<0
parameters respectively and h(z) is the pinching function. 
bþc>0
For a non-pinching and non-degrading system, it is considered  Weak Softening on loading, mostly linear
cb¼0
that hysteresis is defined by a continuous function and hysteresis unloading
stiffness is always zero at local maximum or minimum. It is the 
bþc>cb
point on the load-slip curve where velocity changes its sign. Hence,  Strong Softening on loading and unloading,
cb>0
at an infinitesimal distance dz away from zmax, where velocity is narrow loop
close to but not equal to zero and z_ max  z_ 1 : 
bþc¼0
 Weak Hardening
cb<0
_ n1 z þ cujzj
z_ max  0 ¼ Au_  mðbjujjzj _ nÞ 
ð4Þ 0>bþc
 Strong Hardening
zmax ¼ fA=mðb þ cÞg1=n bþc>cb
Although inclusion of A grants increased versatility of the mod- 2.2.3. Hysteretic energy
el, this parameter is somewhat redundant as both hysteretic stiff- Hysteretic energy is an essential term to approximate strength
ness and hysteretic force, a function of hysteretic displacement, and stiffness degradation. The energy absorbed by the hysteretic
can be varied by the stiffness ratio a and the hysteresis shape element is the continuous integral of hysteretic force, fh over the
parameters b, c and n. Thus, A has been set to unity to remove total displacement u. The hysteretic energy is expressed as:
redundancy. Z uðtÞ Z uðtÞ
du
eðtÞ ¼ fh  du ¼ ð1  aÞx20 zðu; tÞ  du 
uð0Þ uð0Þ dt
2.2.1. Stiffness ratio or rigidity ratio Z t
Stiffness ratio or rigidity ratio a is the ratio of final asymptote ¼ ð1  aÞx20 _
zðu; tÞ  uðtÞ  dt ð6Þ
0
tangent stiffness to initial stiffness. The magnitude of a is 1 for a
linear system and 0 for a complete nonlinear system. In the original
BWBN model, a was considered to be of constant magnitude. How- 2.2.4. Strength and stiffness degradation
ever, based on experimental results of reinforced concrete beam– Strength and stiffness degradation parameters, m and g respec-
column joints under cyclic loading, it can be well perceived that tively, are the functions of total hysteretic energy as shown in
stiffness of the beam–column joints decreases after attaining a cer- the following expressions:
tain displacement and thus, stiffness ratio cannot be of constant mðeÞ ¼ 1 þ dm e and gðeÞ ¼ 1 þ dg e ð7Þ
magnitude. Therefore, a can be expressed as a function of Dmax:
where dm and dg are the constants specified for the desired rates of
a ¼ a0 eð0:1D maxÞ ð5Þ strength and stiffness degradation respectively at different dis-
placement levels. When the magnitudes of dm and dg are zero, the
Here Dmax is the absolute value of the maximum positive dis- structure does not degrade its strength and stiffness. Due to in-
placement and maximum negative displacement for u > 0 and crease in dg, both hysteretic force and hysteretic stiffness degrade
u < 0 respectively. Here a0 is the magnitude of a at zero displace- whereas increase in dm reduces the hysteretic force without chang-
ment, considered to be of constant magnitude. Pinching stiffness ing the hysteretic stiffness.
(minimum tangent stiffness of the curve where unloading finishes
and reloading begins) is around ax20 and the stiffness decreases 2.2.5. Pinching
with development of maximum displacement which proves that The expression for pinching function h(z) is as follows:
use of varying a is more accurate. The hysteresis loop with con- _ 2
=f22
stant and displacement-based a is shown in Fig. 2. hðzÞ ¼ 1  f1 eðzsgnðuÞqzmax Þ ð8Þ
 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406 395

Fig. 3. Possible hysteresis shapes for n = 1.

where f1 determines the severity of pinching or the magnitude of The hysteresis model Eqs. (10)–(12) can be rewritten based on
initial drop in slope dz/du and f1 varies from 0 to 1; f2 causes the Eq. (14) as follows:
pinching region to spread and q is a constant that sets the pinching
level as a fraction of zmax. Both f1 and f2 vary with hysteretic energy y_ 1 ¼ y2 ð15Þ
(Eq. (6)), as mentioned in the following equations:
y_ 2 ¼ 2n0 x0 y2  ax20 y1  ð1  aÞx20 y3 þ f ðtÞ ð16Þ
f1 ðeÞ ¼ fs f1  eðpeÞ g and f2 ðeÞ ¼ ðw þ dw Þðk þ f1 Þ ð9Þ
D 1=n 2 2 py 2
E
where p is a constant that contributes to the rate of initial drop in y_ 3 ¼ 1  fs ð1  epy4 Þeðy3 sgnðy2 Þqf1=ð1þdm y4 ÞðbþcÞg Þ =ðwþdw y4 Þ ½kþfs ð1e 4 Þ
slope; fs is the measure of total slip; w is a parameter that controls * +
n1 n
the amount of pinching; dw is a constant for the desired rate of y2  ð1 þ dm y4 Þðbjy2 jjy3 j y3 þ cy2 jy3 j Þ

pinching spread and k is a parameter that controls the rate of 1 þ dg y4
change of f2 with change of f1. ð17Þ

3. Solving procedure for proposed hysteresis model y_ 4 ¼ ð1  aÞx20 y2 y3 ð18Þ

The complete hysteresis model can be represented in its analyt- LSODE employs user-specified relative and absolute error con-
ical form as follows: trol. Satisfactory results have been obtained by turning off the rel-
ative error control and keeping the absolute error control at a
€ þ 2n0 x0 u_ þ ax20 u þ ð1  aÞx20 z ¼ f ðtÞ
u ð10Þ constant magnitude of 1012. Once the displacement function is
D E known and the parameters are estimated using a system identifica-
1=n 2
Þ =ðwþdw eÞ2 ½kþfs ð1epe Þ2
z_ ¼ 1  fs ð1  epe ÞeðzsgnðuÞqf1=ð1þd
_ m eÞðbþcÞg tion technique, LSODE can be used to work out these equations
* + without any difficulty.
_ n1 z þ cujzj
u_  ð1 þ dm eÞðbjujjzj _ nÞ
 ð11Þ
1 þ dg e 4. Estimation of parameters involved in hysteresis model
Z t
eðtÞ ¼ ð1  aÞx20 zðu; tÞ:uðtÞ _  dt ð12Þ The hysteretic force for input displacement cannot be computed
0
from the model until the analytical parameters are estimated prop-
Here all the notations carry their usual significances. In Eqs. (10)– erly and inserted in the solver subroutine. In order to minimize the
(12), all the derivatives appear in the first power and the variables difference between the experimental results and the model output
vary with time at highly different rates. Hence, the hysteresis model for a given input function, estimation of the analytical parameters
consists of a stiff set of Ordinary Differential Equations (ODE), which is requisite so that the hysteresis model can be practical and appli-
can be solved numerically by using Gear’s backward differential for- cable to a wide range of similar problems. Since the hysteresis
mulae. In the present research, Livermore Solver for Ordinary Dif- model is not only sensitive to the parameters, but also to the inter-
ferential Equations (LSODE) has been chosen for solving the ODEs action between them, it is almost impossible to identify the param-
involved in the proposed analytical model. LSODE, after determin- eters reasonably without a systematic search. Several methods
ing any problem to be comprising of a stiff set of ODEs, uses the [23–25] have been used by various researchers to carry out effi-
Gear Method for solving the equations. Moreover, the input dis- cient parameter estimation. In this study, a Genetic Algorithm
placement function required for computation, may not necessarily (GA) has been written in Visual Fortran to estimate the parameters
be continuous. Even discrete data points can be read from an exter- of the analytical model. The structure of GA is characterized by four
nal file to serve the purpose. nested loops [26–29]. The innermost loop (Loop 4) is the actual GA
LSODE requires the user to convert the system of ODEs into an that generates a population, checks solver (LSODE) calculations, as
array of first order ODEs. well as selects and mates the pairs to crossover and mutate. Solver
checking is necessary because the parameters are generated at ran-
dy dom. To prevent the GA from falsely recognizing the erroneous
¼ f ðt; yÞ ð13Þ
dt sums of squares as better fit, solver computation is checked after
where y is a vector containing the set of ODEs and f is a vector-val- each run. Loop 3 executes GA a user-specified number of times,
ued function of t and y. Subsequently, it can be written as each time with a different randomly chosen initial population.
Loop 2 progressively decreases or shrinks the parameter interval.
* y1 ðtÞ + * uðtÞ + GA is an adaptive algorithm in the sense that it is able to discover
y2 ðtÞ _
uðtÞ
¼ ð14Þ erroneous initial input ranges for the parameters. If a wrong inter-
y3 ðtÞ zðtÞ
val is specified and the optimal parameter lies outside the interval,
y4 ðtÞ eðtÞ the results tend to be clustered near the side of the interval that
should be readjusted. GA subsequently shifts the interval in the
396 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406

direction of clustering and starts over, which is the task of Loop 1. with experimental results in order to verify the effectiveness of the
One of the significant benefits of using GA is that the interval selec- proposed approach. The interior and exterior beam–column joint
tion for each parameter does not affect the end result, but it can specimens with limited transverse reinforcement tested under cyc-
make a significant difference in the CPU time needed to reach lic loading have been selected from literature in such a way that a
the ultimate solution. Although GA takes longer time to converge wide range of variation is covered with respect to the joint aspect
than the calculus-based techniques, but a trend can be recognized ratio, application of column axial load, plain or deformed reinforc-
relatively faster and quick insight can be gained regarding the ing bars, reinforcing bar layout and the grade of concrete and steel.
problem at hand. However, the retrofitted beam–column joints or the beam–column
joints with transverse beam or with slab have been kept out of the
scope of this research.
5. Calibration of analytical model with experimental results
After selection of the specimens to be calibrated, their load
deformation data are retrieved to estimate the analytical model
To check the appropriateness of selection of the pinching func-
parameters for each specimen using Genetic Algorithm, where
tion and the accuracy of the solver and algorithm, the hysteresis
the stiffness ratio and pinching function are defined based on Sec-
model after parameter identification has been calibrated with the
tion 2. Then, from the analytical parameters estimated for interior
experimental results of reinforced concrete (RC) interior and exte-
and exterior beam–column joint specimens, analytical shear force
rior beam–column joints with limited transverse reinforcement
versus horizontal deflection plots can be obtained using LSODE.
obtained from the literature.
The entire process has been summarized in the form of a flowchart
RC non-seismic detailed interior beam–column joint specimens,
in Fig. 4. Comparison between the experimental and analytical
Unit O1 by Hakuto et al. [1], Units 1 and 2 tested by Liu et al. [2],
shear force-horizontal deflection plots of lightly reinforced con-
Units PEER-1450, PEER-2250, CD15-1450, CD30-1450 and CD30-
crete interior and exterior beam–column joint specimens are pre-
2250 tested by Walker [3] and PEER-0995 and PEER-4150 tested
sented in Figs. 5 and 6 respectively. In order to maintain a level
by Alire [4] and exterior beam–column joint specimens Units O6
of accuracy for all the specimens, the analytical parameters have
and O7 tested by Hakuto et al. [1], Units EJ2 and EJ3 tested by
been estimated such that the correlation coefficient of the compar-
Liu et al. [2] and Units 1, 2, 3, 4, 5 and 6 tested by Pantelides
ison plots remains 0.98 for all of them.
et al. [5] have been selected for calibration of the analytical model

6. Model sensitivity to parameter variations

In pursuance of judging the sensitivity of the hysteresis model

to the variation of associated analytical parameters, a numerical
example has been deduced from previous section. Calibration of
the analytical model with the experimental result of Unit O1 by
Hakuto et al. [1] at a level of their correlation coefficient as 0.98,
yields the following parameter magnitudes.

a0 ¼ 0:025; x0 ¼ 2:5; n0 ¼ 0:02; b ¼ 0:05; c ¼ 0:01;

n ¼ 1:01; dm ¼ 0:00005; dg ¼ 0:0005; fs ¼ 0:93;
p ¼ 0:08; w ¼ 0:8; dw ¼ 0:11; q ¼ 0:03; k ¼ 0:1
Sensitivity of the hysteresis loop has been investigated by
changing the magnitude of each analytical parameter individually
one after another, while other parameters are kept constant in
magnitude. Fig. 7a–j displays the influences of variations of the
analytical parameters a0, x0, n0, b, n, dm, dg, fs, q, p, w, dw and k on
the hysteresis loop correspondingly. The analytical parameters,
stiffness ratio a0 and system natural frequency x0 with their
changing magnitudes, affect the stiffness of the hysteresis loop,
as depicted in Fig. 7a and b. Altering the magnitude of the linear
damping ratio n0, while keeping all other parameters constant,
does not produce any significant changes in the hysteresis loop,
as shown in Fig. 7c. Fig. 7d–f illustrates how the magnitudes of hys-
teresis shape parameters b, c and n can individually influence the
shapes of the hysteresis loops. From Fig. 7g and h, it can be under-
stood that with increase or decrease of degradation parameters dm
and dg, structure experiences more or less degradations respec-
tively. The remaining parameters, being pinching parameters only
control the pinching of the hysteresis loop. fs controls the amount
of total slip in the hysteresis loop as observed in Fig. 7i. When this
parameter is of zero magnitude, no slip can be observed in the hys-
teresis loop whereas with an increase in its magnitude, greater slip
is found in the loop. As q is a constant that sets the pinching level,
the influence of its varying magnitude can significantly deviate
pinching as shown in Fig. 7j. Variation of p contributes to the rate
of drop in the slope, as illustrated in Fig. 7k. Moreover, with an in-
Fig. 4. Flow chart of the entire process to resolve hysteresis behavior of lightly crease or decrease in w, the amount of pinching behaves propor-
reinforced concrete beam–column joints. tionately as depicted in Fig. 7l and due to increase in the
 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406 397

Fig. 5. Experimental and analytical shear force versus horizontal deflection plots of reinforced concrete interior beam–column joints with limited transverse reinforcement.
398 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406

Fig. 6. Experimental and analytical shear force versus horizontal deflection plots of reinforced concrete exterior beam–column joints with limited transverse reinforcement.
 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406 399

Fig. 7. Sensitivity of hysteresis loop to variation of each model parameter.

400 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406

magnitude of dw, the pinching region spreads as shown in Fig. 7m.

A change in k also affects the amount and spread of pinching in the
hysteresis loop as demonstrated in Fig. 7n. In brief, system proper-

Root Mean Square Error

ties (a0, x0, n0) and hysteresis shape parameters (b, c, n) control α0
the skeleton of hysteresis loops; degradation parameters (dm, dg) β
γ
determine strength and stiffness deteriorations and pinching
ν
parameters (fs, q, p, w, dw, k) govern the slip and pinching magni- θ
tude and pinching spread. However, after varying each parameter π
up to a definite range and computing the error occurred due to ζσ
ψ
each variation, sensitive ranking of each parameter can be easily δψ
deduced. λ
Let [Y] be the hysteretic force for a given input function and the δν
δη
magnitudes of the analytical parameters are estimated as:

a0 ¼ 0:025; x0 ¼ 2:5; n0 ¼ 0:02; b ¼ 0:05; c ¼ 0:01;

n ¼ 1:1; dm ¼ 0:00005; dg ¼ 0:0005; fs ¼ 0:9; q ¼ 0:03; Percentage Variation from Estimated Parameter
p ¼ 0:08; w ¼ 0:8; dw ¼ 0:11; k ¼ 0:1: Magnitude at middle

Fig. 8. Spider diagram of root mean square error versus percentage variation of
Then, each parameter, excluding x0 and n0, is varied from 10% to each parameter.
+10% of its original magnitude. In this sensitivity ranking determi-
nation, the system natural frequency x0 and the linear viscous
final response is relatively less. Therefore, providing narrower
damping ratio n0 have been excluded due to the fact that x0 of
ranges for less sensitive parameters can increase simplicity in the
any structure is invariable and for a given x0, variation of n0 does
procedure without affecting the quality of the results.
not affect the hysteresis loop. Therefore, an attempt has been made
to relate x0 with the physical parameter of the beam–column joint
and fix a range for n0 in the next section. Now, if due to the variation 7. Parametric study
of a parameter, say a0, the hysteretic force becomes [Y0 ], then the
root mean square error ea0 will be as follows: An extensive parametric study has been conducted in this
research in order to relate the physical parameters of the beam–col-
* +1=2
XN umn joints with the hysteresis model parameters. Reinforced
0 2
ea0 ¼ ðY  Y Þ ð19Þ concrete interior and exterior beam–column joints with limited
i¼1

Here N is the number of data points for input displacement func- Table 2
tion. The maximum error related to the variation of a0, termed as General features of the UC-Win/WCOMD model.
jea0 j can be obtained by the following expression. Type of Joint Interior and exterior joints

jea0 j ¼ maximumðea0 Þ ð20Þ Joint aspect ratio 1.67 (column cross sectional depth: 300 mm
and beam depth: 500 mm)
The maximum root mean square error associated with each Total height of the 2900 mm
parameter variation is summarized in Table 1. The parameter with beam–column joint
Total span of the 3500 mm for interior joint and 1900 mm for
the highest magnitude of maximum root mean square error is
beam–column joint exterior joint
ranked as 1 based on its sensitivity. By plotting the root mean Concrete grade fc0 ¼ 40 MPa
square error for any parameter within the range of its variation, Steel grade fy = 350 MPa and deformed bars
a Spider diagram is obtained as shown in Fig. 8. Column longitudinal 2%
Parameter sensitivity analysis is vital when dealing with the reinforcement ratio
Beam longitudinal 2%
system identification techniques. A sensitive parameter when
reinforcement ratio
deviated from its sought-after magnitude will show reasonable er- Displacement history 1–5% drift ratio
ror. Thus, by changing the magnitude of sensitive parameters, bet- Axial load ratio 0
ter correlation can be achieved. On the contrary, a less sensitive
parameter, even when it is fluctuated from its sought-after magni-
tude, can produce a reasonable response as its contribution to the
Table 3
Estimated parameters for interior and exterior beam–
Table 1 column joints.
Parameter sensitivity ranking.
Parameter Interior joint Exterior joint
Parameter Maximum root Rank
a0 0.03 0.01
mean square error
x0 2.85 1.85
a0 1.15 7 n0 0.02 0.02
b 5.08 3 b 0.05 0.05
c 0.95 8 c 0.01 0.01
n 10.1 2 n 1.01 1.01
fs 52.15 1 fs 0.91 0.91
q 0.65 11 q 0.03 0.03
p 0.27 12 p 0.08 0.05
w 3.09 4 w 0.8 0.8
dw 0.81 9 dw 0.11 0.15
k 0.78 10 k 0.1 0.1
dm 1.34 6 dm 0.00005 0.00006
dg 2.90 5 dg 0.0005 0.0008
 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406 401

Table 4 7.1. Effect of joint aspect ratio

Key factors varied in parametric study.

Factors No. Description The joint aspect ratio is defined as the ratio of beam depth and
Type of joint Interior and exterior joints column cross-sectional depth. From the simulation results, it has
Joint aspect ratio 10 (a) Varying column depth: 1, 1.11, been observed that the joint aspect ratio plays a pivotal role in
1.25, 1.43, 1.67 determining the joint shear strength. In this parametric study,
(b) Varying beam depth: 1, 1.17, 1.33, the joint aspect ratio is changed in two ways, first, by keeping
1.5, 1.67
the beam depth constant and varying the column cross-sectional
Concrete grade 3 fc0 ¼ 20; 30 and 40 MPa
Steel grade 2 fy = 250 MPa for plain bars, depth and secondly, by keeping the column cross-sectional depth
fy = 350 MPa for deformed bars constant and varying the beam depth. Joint shear strength de-
Column longitudinal 5 1.0%, 1.5%, 2%, 2.5%, 3% creases with decrease in the column cross-sectional depth whereas
reinforcement ratio
it increases with decrease in the beam depth. As a consequence,
Beam longitudinal 5 1.0%, 1.5%, 2%, 2.5%, 3%
reinforcement ratio
the system parameters (a0, x0) associated with the hysteresis
Axial load ratio 6 0, 0.1, 0.15, 0.2, 0.25, 0.3 model also suffer perturbation. Figs. 9 and 10 show changes in
these two parameter magnitudes due to changes in the joint aspect
ratio by varying the column cross-sectional depth and the beam
depth respectively.
transverse reinforcement have been modeled in UC-win/WCOMD
[30] based on the principal features depicted in Table 2. The 7.2. Effect of concrete and steel grades
analytical parameters estimated for the interior and exterior
beam–column joints are summarized in Table 3. Next, the In non-seismic designed reinforced concrete buildings, high
structural performance of reinforced concrete non-ductile beam– strength concrete is generally not used for construction. As per
column joints has been investigated by varying the key factors based the old practice, concrete with compressive cylinder strengths fc0
on Table 4. Thus, the structural responses of the beam–column joint of 20 MPa, 30 MPa and 40 MPa has been considered for this para-
models are obtained from simulation and the analytical hysteresis metric study. With decrease in the concrete compressive strength,
parameters are estimated using Genetic Algorithm accordingly. joint shear strength deteriorates along with reduction in the hys-
From the estimation, with the change of each joint physical teresis model parameters a0 and x0 as presented in Fig. 11. The
parameter, the affected mathematical parameters can be recognized influence of usage of deformed and plain round bars as longitudi-
and subsequently, the magnitudes of concerned analytical parame- nal reinforcement has been investigated when the same specimen
ters are plotted against that joint physical parameter to elucidate has been modeled once with deformed bars having yield strength
their inter-relation. of 350 MPa and thereafter with plain bars of yield strength

Interior Joint Interior Joint

Exterior Joint Exterior Joint

α0
ω0

Joint Aspect Ratio Joint Aspect Ratio

Fig. 9. Effect of joint aspect ratio (varying column depth) on model parameters.

Interior Joint Interior Joint

Exterior Joint Exterior Joint

ω0 α0

Joint Aspect Ratio Joint Aspect Ratio

Fig. 10. Effect of joint aspect ratio (varying beam depth) on model parameters.
402 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406

Interior Joint Interior Joint

Exterior Joint Exterior Joint

ω0 α0

Grade of Concrete (MPa) Grade of Concrete (MPa)

Fig. 11. Effect of grade of concrete on model parameters.

Interior Joint Interior Joint

Exterior Joint Exterior Joint

ω0 α0

Grade of Steel (MPa) Grade of Steel (MPa)

Interior Joint Interior Joint

Exterior Joint Exterior Joint

ζs q

Grade of Steel (MPa) Grade of Steel (MPa)

Interior Joint
Exterior Joint

Grade of Steel (MPa)

Fig. 12. Effect of plain/deformed bar on model parameters.

250 MPa as longitudinal reinforcement. The model with plain bar 7.3. Effect of column axial load
produces reduced shear strength, but higher slip and profound
pinching due to severe slippage of longitudinal reinforcement bars. The column axial load plays a significant role in the hysteresis
Pinching parameters (fS, q, w) and the system parameters (a0, x0) behavior of beam–column joints. From the simulation results, it
also experience significant changes in their magnitudes accord- has been observed that the effect of the column axial load is more
ingly as depicted in Fig. 12. prominent in the exterior beam–column joints than the interior
 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406 403

beam–column connections. Due to the presence of the column ax- ied from 0 to 0.1, 0.15, 0.2, 0.25 and 0.3. The influences of different
ial load, the joint strength is enhanced, but with increase in the col- levels of the column axial load ratio on the system properties
umn axial load ratio, more degradation and pinching are observed. (a0, x0), degradation parameters (dm, dg) and pinching parameters
In this parametric study the column axial load ratio has been var- (fs, q, w) have been portrayed in Fig. 13.

Interior Joint Interior Joint

Exterior Joint Exterior Joint

ω0 α0

Column Axial Load Ratio Column Axial Load Ratio

Interior Joint Interior Joint

Exterior Joint Exterior Joint

ζs q

Column Axial Load Ratio Column Axial Load Ratio

Interior Joint Interior Joint

Exterior Joint Exterior Joint

ψ δν

Column Axial Load Ratio Column Axial Load Ratio

Interior Joint
Exterior Joint

δη

Column Axial Load Ratio

Fig. 13. Effect of column axial load ratio on model parameters.

404 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406

Interior Joint Interior Joint

Exterior Joint Exterior Joint

α0
ω0

Column Longitudinal Reinforcement Ratio (%) Column Longitudinal Reinforcement Ratio (%)

Fig. 14. Effect of column longitudinal reinforcement ratio on model parameters.

Interior Joint Interior Joint

Exterior Joint Exterior Joint

ω0 α0

Beam Longitudinal Reinforcement Ratio (%) Beam Longitudinal Reinforcement Ratio (%)

Fig. 15. Effect of beam longitudinal reinforcement ratio on model parameters.

7.4. Effect of column and beam longitudinal reinforcement

Table 5
Upper and lower bounds of model parameters.
In order to investigate the effect of column longitudinal rein-
Parameter Lower bound Upper Bound
forcement ratio, reinforced concrete beam–column joint models
have been built in UC-win/WCOMD with variation of the column a0 0.005 0.07
longitudinal reinforcements, while keeping remaining key features x0 0.5 6.0
n0 0.01 0.03
of the model unchanged. The simulation results indicate that b 0.04 0.06
change in the column longitudinal reinforcement does not exhibit c 0.02 0.01
any influence on the joint shear strength and as a result, the hyster- n 1.0 1.1
esis model parameters will also remain unchanged as shown in fs 0.85 0.98
q 0.03 0.08
Fig. 14. Similarly, to check the possible influence of the beam longi-
p 0.04 0.09
tudinal reinforcement ratio, reinforced concrete beam–column w 0.75 0.95
joints have been modeled by varying the beam longitudinal dw 0.1 0.2
reinforcement only. The simulation results show that with increase k 0.1 0.12
in the beam longitudinal reinforcement ratio from 1.0 to 2.0, joint dm 0.00005 0.00008
dg 0.0005 0.002
shear strength increases around 30%. With alteration of the beam
longitudinal reinforcement ratio, the hysteresis model parameters
a0 and x0 varies as presented in Fig. 15.
The entire parametric study has been undertaken to realize the sponse under a specific displacement history can be attained
hysteresis behavior of a wide range of beam–column joints and the using any suitable solver. Two examples have been included in
relationship between the joint physical parameters and the math- the Appendix on the selection process of the analytical parameters
ematical parameters associated with the analytical model. The for the interior beam–column joint Unit C2 and the exterior beam–
only aim underlying this is to provide a comprehensive range of column joint Unit L1 by Pampanin et al. [31] based on their phys-
the analytical parameters to enable the model to be user-friendly. ical parameters according to Figs. 9–15. The experimental versus
Though it is not feasible to include parameter magnitudes for all analytical load–deflection plots for Unit C2 and Unit L1 have been
the cases involved in the parametric study, the generalized upper shown in Fig. 16.
and lower bounds for each of the parameters are tabulated in Ta-
ble 5. From this table and the figures denoting the influence of 8. Conclusions
the physical parameters on the analytical parameters, the user
can gain a quick impression of the magnitudes of the model This paper presents an analytical approach to predict the hys-
parameters for a wide variety of beam–column joints. As soon as teresis behavior of non-seismic detailed reinforced concrete beam
the approximate parameter sets are identified for a definite rein- column joints based on modified Bouc–Wen hysteresis model.
forced concrete non-ductile beam–column joint, its hysteresis re- LSODE and Genetic Algorithm (GA) have been opted for solving
 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406 405

Experimental Experimental
Analytical Analytical

Shear Force (kN)

Shear Force (kN)

Specimen C2 Specimen L1
(Pampanin) (Pampanin)

Horizontal Displacement (mm) Horizontal Displacement (mm)

Fig. 16. Experimental and analytical shear force versus horizontal deflection plots of interior and exterior beam–column joint specimens C2 and L1 with the parameters
estimated in Appendix.

the analytical model equations and estimation of the parameters  Column longitudinal reinforcement ratio
associated with the equations respectively. The efficiency of the
¼ ½ð6  50:3Þ=ð330  200Þ  100% ¼ 0:46%
analytical model and the accuracy of the solver and the algorithm
have been proved by the strong correlations between the experi- Beam longitudinal reinforcement ratio
mental and the analytical shear force-horizontal deflection plots ¼ ½ð4  50:3 þ 3  113Þ=ð330  200Þ  100% ¼ 0:82%
for lightly reinforced concrete interior and exterior beam–column
joint specimens from the literature. The sensitivity of the proposed .  Column axial load ratio = 0.08.
model to the variation of its analytical parameters has been inves-
tigated and the sensitivity ranking of the parameters have been The parameter magnitudes for the interior beam–column joint
finalized. Less sensitive parameters can be kept constant in order with physical characteristics as mentioned in Table 2 are as
to keep the model simple without causing much error to the re- follows:
sponse. The influence of the joint physical parameters, such as
the joint aspect ratio, concrete compressive cylinder strength, plain a0 ¼ 0:03; x0 ¼ 2:85; n0 ¼ 0:02; b ¼ 0:05; c ¼ 0:01;
or deformed bars for longitudinal reinforcement, the column and
beam longitudinal reinforcement ratio, the column axial load ratio, n ¼ 1:01; dm ¼ 0:00005; dg ¼ 0:0005; fs ¼ 0:91; q ¼ 0:03;
on the model parameters has been examined meticulously based p ¼ 0:08; w ¼ 0:8; dw ¼ 0:11; k ¼ 0:1
on the extensive parametric study. Moreover, the approximate
upper and lower bounds for the model parameters have been spec- Step 1: Effect of joint aspect ratio
ified, from which the user can identify the approximate magni- In the present problem, joint aspect ratio is 1.65 with beam
tudes of the parameters for any beam–column joint depending depth 330 mm and column cross-sectional depth 200 mm. Hence
on its physical parameters. In this simplified approach, the analyt- from Figs. 9 and 10 by linear interpolation, it can be calculated that
ical parameters can be estimated instantly without using any sys- a0 = 0.02, x0 = 2.35 when remaining physical parameters of the
tem identification tool and the hysteresis behavior of the beam– interior joint remain unaltered.
column joint can be accomplished using the estimated parameters Step 2: Effect of grade of concrete and steel
with the help of any efficient solver. For concrete compressive cylinder strength 23.9 MPa and plain
reinforcement bar, based on Figs. 11 and 12 respectively, modified
Appendix A magnitudes of the parameters are:
a0 ¼ 0:01; x0 ¼ 1:60; fs ¼ 0:95; q ¼ 0:05; w ¼ 0:905
An illustration on determination of parameter magnitudes for
an interior beam–column joint specimen Unit C2 and an exterior Step 3: Effect of longitudinal reinforcement ratio
beam–column joint specimen Unit L1 [29] is shown hereunder. Column longitudinal reinforcement ratio does not influence the
From the sample parameter set, analytical parameters for both analytical parameters. But due to shift of beam longitudinal rein-
joints will be calculated based on the parametric study conducted forcement ratio from 2% to 0.82%, magnitudes of the parameters
in this research. a0 and x0 will experience little change according to Fig. 15 as
a0 = 0.008, x0 = 1.17.
(a) The physical characteristics of Unit C2 by Pampanin et al. Step 4: Effect of column axial load ratio
[29] are as follows: According to Fig. 13, due to presence of column axial load
 Joint aspect ratio = 1.65 (beam depth 330 mm and 0:08fc0 Ag , the final parameter magnitudes of Unit C2 are as follows:
column cross-sectional depth 200 mm).
 Concrete compressive cylinder strength of the specimen
a0 ¼ 0:009; x0 ¼ 1:22; n ¼ 0:02; b ¼ 0:05; c ¼ 0:01;
fc0 ¼ 23:9 MPa. n ¼ 1:01; dm ¼ 0:000052; dg ¼ 0:00062; fs ¼ 0:966;
 Average yield strength of steel for longitudinal reinforce-
ment bars = 365.75 MPa (plain bar). q ¼ 0:058; p ¼ 0:08; w ¼ 0:934; dw ¼ 0:11; k ¼ 0:1
406 P. Sengupta, B. Li / Engineering Structures 46 (2013) 392–406

(b) The physical characteristics of Unit L1 by Pampanin et al. [3] Walker SG. Seismic performance of existing reinforced concrete beam–column
joints. MS thesis. University of Washington; 2001. 308 pp.
[29] are as follows:
[4] Alire DA. Seismic evaluation of existing unconfined reinforced concrete beam–
 Joint aspect ratio = 1.65 (beam depth 330 mm and column joints. MS thesis. University of Washington; 2002. 291 pp.
column cross-sectional depth 200 mm). [5] Pantelides CP, Hansen J, Nadauld JD, Reaveley LD. Assessment of reinforced
 Concrete compressive cylinder strength of the speci- concrete building exterior joints with substandard details. PEER report 2002/
18. University of California, Berkley; May 2002.
men fc0 ¼ 23:9 MPa. [6] Veletsos AS, Newmark NM, Chelapati CV. Deformation spectra for elastic and
 Average yield strength of steel for longitudinal rein- elastoplastic systems subjected to ground shock and earthquake motions. In:
forcement bars = 365.75 MPa (plain bar). Proceedings, 3rd world conference on earthquake engineering, vol. II; 1965. p.
663–82.
 Column longitudinal reinforcement ratio [7] Clough RW, Johnston SB. Effect of stiffness degradation on earthquake ductility
requirements. In: Proceedings, second Japan national conference on
¼ ½ð6  50:3Þ=ð330  200Þ  100% ¼ 0:46% earthquake engineering; 1966. p. 227–32.
[8] Takeda T, Sozen MA, Nielson NN. Reinforced concrete response to simulated
Beam longitudinal reinforcement ratio
earthquakes. J Struct Div, ASCE 1970;96:2557–73 (ST 12).
¼ ½ð4  50:3 þ 4  113Þ=ð330  200Þ  100% ¼ 0:99% [9] Imbeault FA, Nielsen NN. Effect of degrading stiffness on the response of
. multistory frames subjected to earthquakes. In: Proceedings, fifth world
conference on earthquake engineering; 1973. p. 1756–65.
 Column axial load ratio = 0.
[10] Saidi M, Sozen MA. Simple and complex models for nonlinear seismic response
of reinforced concrete structures. A report to the national science foundation.
The parameter magnitudes for an exterior beam–column joint
University of Illinois at Urbana Campaign; August, 1979.
with physical characteristics as mentioned in Table 2 are as [11] Otani S. Nonlinear dynamic analysis of reinforced concrete building structures.
follows: Can J Civil Eng 1980;7:333–44.
[12] Ozcebe G, Saatcioglu M. Hysteresis shear models for reinforced concrete
a0 ¼ 0:01; x0 ¼ 1:85; n0 ¼ 0:02; b ¼ 0:05; c ¼ 0:01; members. J Eng Mech, ASCE 1989;115(1):132–48.
[13] Alath S, Kunnath SK. Modelling inelastic shear deformations in RC beam–
n ¼ 1:01; dm ¼ 0:00006; dg ¼ 0:0008; fs ¼ 0:91; q ¼ 0:03; column joints. In: Engineering mechanics proceedings of 10th conference, May
p ¼ 0:05; w ¼ 0:8; dw ¼ 0:15; k ¼ 0:1 21–24, University of Colorado at Boulder, Boulder, Colorado, ASCE, New York,
vol. 2; 1995, 822–5 pp.
Step 1: Effect of joint aspect ratio [14] Biddah A, Ghobarah A. Modelling of shear deformation and bond slip in
reinforced concrete joints. Struct Eng Mech 1999;7(4):413–32.
Here, joint aspect ratio is 1.65 with beam depth 330 mm and [15] Hwang SJ, Lee HJ. Analytical model for predicting shear-strengths of exterior
column cross-sectional depth 200 mm. Hence from Figs. 9 and 10 reinforced concrete beam–column joints for seismic resistance. ACI Struct J
by linear interpolation, it can be calculated that a0 = 0.01, 1999;96(5):846–57.
[16] Youssef M, Ghobarah A. Modelling of RC beam column joints and structural
x0 = 1.49, when remaining physical parameters of the exterior walls. J Earthq Eng 2001;5(1):93–111.
joint remain unchanged. [17] Lowes LN, Altoontash A. Modelling reinforced-concrete beam–column joints
Step 2: Effect of grade of concrete and steel subjected to cyclic loading. ASCE J Struct Eng 2003;129(12):1686–97.
[18] Altoontash A. Simulation and damage models for performance assessment of
For concrete compressive cylinder strength 23.9 MPa and plain reinforced concrete beam–column joints. PhD thesis. Department of Civil and
reinforcement bar, based on Figs. 11 and 12 respectively, the mod- Environmental Engineering, Stanford University, Stanford, California; 2004.
ified parameter magnitudes are: [19] Shin M, LaFave JM. Testing and modelling for cyclic joint shear deformation in
RC beam–column connections. In: Proceedings of 13th world conference on
a0 ¼ 0:006; x0 ¼ 1:01; fs ¼ 0:93; q ¼ 0:05; w ¼ 0:9 earthquake engineering, August 1–6, 2004, Vancouver, BC, Canada, paper no
0301.
Step 3: Effect of longitudinal reinforcement ratio [20] Favvata MJ, Izzuddin BA, Karayannis CG. Modelling exterior beam–column
Due to change in beam longitudinal reinforcement ratio from joints for seismic analysis of RC frame structures. J Earthq Eng Struct Dynam
2008;37:1527–48.
2% to 0.99% according to Fig. 15, magnitudes of a0 and x0 will be [21] Baber TT, Noori MN. Random vibration of degrading pinching systems. J Eng
changed into a0 = 0.005, x0 = 0.17. Mech, ASCE 1985;111(8):1010–26.
Step 4: Effect of column axial load ratio [22] Baber TT, Wen YK. Random vibrations of hysteretic degrading systems. J Eng
Mech, ASCE 1981;107(6):1069–87.
According to Fig. 13, due to presence of column axial load [23] Kunnath SK, Mander JB, Fang L. Parameter identification for degrading and
0:08fc0 Ag , the final parameter magnitudes of Unit L1 are as follows: pinched hysteretic structural concrete system. Eng Struct 1997;19(3):224–32.
[24] Hurtado JE, Barbat AH. Equivalent linearization of the Bouc–Wen hysteresis
a0 ¼ 0:005; x0 ¼ 0:77; n0 ¼ 0:02; b ¼ 0:05; c ¼ 0:01; n ¼ 1:01; model. Eng Struct 2000;22(9):1121–32.
[25] Li Z, Albermani F, Chan RWK, Kitipornchai S. Pinching hysteresis response of
dm ¼ 0:00006; dg ¼ 0:0008; fs ¼ 0:93; q ¼ 0:05; p ¼ 0:05; w ¼ 0:9;
yielding shear panel device. Eng Struct 2011;33(3):993–1000.
dw ¼ 0:15; k ¼ 0:1 [26] Goldberg DE. Genetic algorithm in search, optimization and machine
learning. New York: Addison, Wesley; 1989.
The experimental and analytical hysteresis loops for Unit C2 [27] Lybanon M, Messa KC. Genetic algorithm model fitting. Practical handbook of
genetic algorithms. In: Chambers LD, editor. Complex coding systems, vol.
and Unit L1, are shown in Fig. 16 which depicts the accuracy of
III. New York: CRC Press; 1999.
the estimated parameters. [28] Heine CP. Simulated response of degrading hysteretic joints with slack
behavior. PhD thesis. Virginia Polytechnic Institute and State University;
References 2001. 286 pp.
[29] Xu J. Development of a general dynamic hysteretic light-frame structure
model and study on the torsional behavior of open-front light-frame
[1] Hakuto S, Park R, Tanaka H. Seismic load tests on interior and exterior structures. PhD thesis. Washington State University; 2006. 296 pp.
beam–column joints with substandard reinforcing details. ACI Struct J 2000; [30] UC-Win/WCOMD [version 2] User’s Manual.
97(1):11–25. [31] Pampanin S, Calvi GM, Moratti M. Seismic behaviour of RC beam–column
[2] Liu A, Carr AJ, Park R. Seismic assessment and retrofit of pre-1970s reinforced joints designed for gravity only. In: 12th European conference on earthquake
concrete frame structures. Research report 2002-1. Christchurch: University of engineering, paper no 726, London; 2002.
Canterbury; April 2002.