Journal of Constructional Steel Research 101 (2014) 192–206

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Nonlinear inelastic time-history analysis of three-dimensional semi-rigid

steel frames
Phu-Cuong Nguyen, Seung-Eock Kim ⁎
Department of Civil and Environmental Engineering, Sejong University, 98 Gunja-dong Gwangjin-gu, Seoul 143-747, Republic of Korea

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

Article history: This paper presents a simple, effective numerical procedure based on the beam–column method by using the
Received 2 July 2013 based-displacement finite element method for nonlinear inelastic time-history analysis of three-dimensional
Accepted 20 May 2014 semi-rigid steel frames. The nonlinear geometry effects are considered by using stability functions and the
Available online xxxx
geometric stiffness matrix. The inelasticity of material is considered by the gradual yielding of plastic hinges.
A space zero-length multi-spring element is proposed to simulate the nonlinear cyclic behavior of steel connec-
Keywords:
Stability functions
tions through the independent hardening model. The Hilber–Hughes–Taylor method combined with the
Refined plastic hinge Newton–Raphson balance iterative algorithm is adopted to solve the nonlinear equations of motion. The results
Semi-rigid connections of nonlinear responses for steel frames with fully rigid and linear semi-rigid connections compare well with those
3-D steel frames of previous studies and commercial SAP2000 software. Moreover, the results of steel frames with semi-rigid
Nonlinear time-history analysis connections shown in this paper can be used to calibrate similar frames in future studies.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Steel moment resisting frames have been extensively used in areas
of high seismic risk for low and mid-rise buildings due to their high duc-
tility. Conventional analysis and design of steel framed structures are
usually conducted under the assumption that the beam-to-column con-
nections are either fully rigid or frictionless pinned joints. However,
in the 1994 Northridge and 1995 Kobe earthquakes, these structures,
especially fully welded connections, were heavily and unexpectedly
damaged. Since then, over the last three decades, several experimen-
tal and analytical studies have been conducted to investigate the dy-
namic behavior of alternative connection types, including semi-rigid
connections.
The results of experimental studies showed that semi-rigid steel
frames feature the ductile and stable hysteretic behavior of the frames
when the connections are designed appropriately [1–4]. The energy is
dissipated through hysteretic loops of semi-rigid connections, which
are one of the important damping sources of structures. Those results
also showed that if the connection stiffness increases the base shear in-
creases but the lateral drift does not proportionally decrease. Base on
the impressive characteristics of semi-rigid connections, several mathe-
matic models were proposed to represent actual behavior of these
⁎ Corresponding author. Tel.: +82 2 3408 3004; fax: +82 2 3408 3906.
E-mail addresses: henycuong@gmail.com (P.-C. Nguyen), sekim@sejong.ac.kr
(S.-E. Kim).
connections. These models can be grouped into two categories: linear
connection models [5,6] and nonlinear connection models [7–10].
The linear connection models cannot properly predict the connection
behavior due to the stiffness of the connection being assumed to be con-
stant during the analysis procedure; while the moment–rotation curve
of connections is captured more exactly by the nonlinear connection
models which fitted well in the experimental curve.
In recent years, numerous analytical studies have been conducted to
investigate nonlinear inelastic dynamic behavior of semi-rigid frames
under dynamic and seismic loadings by Lui and Lopes [11], Awkar and
Lui [12], and Sekulovic and Nefovska [13], among others. In the studies
[11] and [12], the effects of geometric nonlinearity were considered by
using stability functions, the material nonlinearity was considered by
refined plastic hinge approach, and the connection nonlinearity was
considered by modifying the stiffness matrix of beam–column element
functions that account for finite rotation at the two ends of elements.
Sekulovic and Nefovska [13] presented a spring-in-series model con-
sidering both the material and connection nonlinearities, while the
geometric nonlinearity was considered by using the geometric stiffness
matrix obtained from approximate functions. To capture exactly the
effects of geometric nonlinearity, the members need to be divided into
several elements so that it consumes analysis time significantly. Though
these studies considered all three nonlinear components in the same
analysis, they are limited to planar semi-rigid steel frames. With the
following procedure, all limitations of the above-mentioned studies
are overcome by the integration of three nonlinearities in the complex
time-history analysis of three-dimensional frames.

http://dx.doi.org/10.1016/j.jcsr.2014.05.009
0143-974X/© 2014 Elsevier Ltd. All rights reserved.
 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206 193

This study aims to investigate the simultaneous effects of three non- rotational ones (y and z axis) have linear or nonlinear stiffness.
linear components on the dynamic behavior of different steel frame Coupling effects between the six springs of a connection are neglected.
types under various dynamic loadings. The geometric nonlinearity The relation between the incremental force vector {ΔFS} and dis-
caused by the interaction between the axial force and bending mo- placement vector {ΔUS} of the multi-spring element corresponding
ments is considered by the use of stability functions, which can accu- to six degrees of freedom is as follows:
rately capture the P–δ effect by simulating only one element per
member; the P–Δ effect is considered by the use of the geometric fΔF S g ¼ ½K S fΔU S g ð1Þ
stiffness matrix. The material nonlinearity is accounted for by using
the refined plastic hinge method [14], in which the Column Research
Council (CRC) tangent modulus concept is used to account for the 2 tra
3
Rx 0 0 0 0 0
gradual yielding due to residual stresses, while the gradual yielding 6 tra 7
6 0 Ry 0 0 0 0 7
due to flexure is represented by a parabolic function combined with 6 7
6 0 0 7
tra
0 Rz 0 0 7
the yielding surface proposed by Orbison [15]. An independent zero- ½K S  ¼ 6 ð2Þ
6 0 0 7
rot
length connection element with six different translational and rotational 6 0 0 Ry 0 7
6 7
4 0 0 5
rot
springs connecting two identical nodes is developed to simulate the 0 0 0 Rz
rot
beam-to-column connections. This is an efficient way because the 0 0 0 0 0 Rx
modification of the beam–column stiffness matrix considering the
semi-rigid connections is not necessary and the connection is ready to where [KS] is the diagonal tangent stiffness matrix for each multi-spring
integrate with any beam–column model. The independent hardening element, Rtra rot
n and Rn are the component stiffness for the translational
model is used for considering cyclic behavior of rotational springs and rotational springs with respect to the n axis (n = x, y, z).
through employing the static mathematic models (Kishi–Chen [7],
Richard–Abbott [8], and Chen–Lui [9]). 2.1.2. Semi-rigid connection models for rotational springs
This study is based on assumptions as follows: warping torsion and In this study, the Kishi–Chen three-parameter power model [7], the
axial shortening due to member curvature bending (bowing effects) Richard–Abbott four-parameter model [8], and the Chen–Lui exponen-
are ignored; lateral-torsion buckling of members is assumed to be tial model [9] are used to evaluate the nonlinear behavior of semi-rigid
prevented by adequate lateral braces; a compact W-section is assumed connections. The independent hardening model is used to predict the
so that the section can develop full plastic moment capacity without cyclic behavior of the connections.
local buckling; the connection element length is equal to zero; a possi- The Kishi–Chen model [7] is currently one of the most popular
ble joint degradation due to actions of cyclic loading is not considered. models used for semi-rigid connections since it needs only three param-
In the nonlinear time-history analysis, an incremental-iterative eters to capture the moment–rotation curve and always gives a positive
scheme based on the Hilbert–Hughes–Taylor method and the Newton– stiffness. The moment–rotation relationship of the connection is
Raphson method was developed for solving the nonlinear equations presented by Chen and Kishi as follows:
of motion. Viscous damping accounts for the use of Rayleigh damping
matrix. Several examples are presented to verify the accuracy and effi- Rki jθr j
M¼ ð3Þ
ciency of the proposed numerical procedure in predicting the nonlinear ½1 þ ðjθr j=θ0 Þn 1=n
dynamic response of three-dimensional framed structures with semi-
rigid connections. where M and θr are the moment and the rotation of the connection, n is
the shape parameter, θ0 is the reference plastic rotation, and Rki is the
2. Formulation initial connection stiffness.
Richard and Abbott proposed a four-parameter model [8]. The
2.1. Semi-rigid connection element moment–rotation relationship of the connection is defined by
 
2.1.1. Element modeling Rki −Rkp jθr j
An independent zero-length multi-spring element with three trans- M¼    þ Rkp jθr j ð4Þ
ðR −R Þjθ jn 1n
lational and three rotational springs was developed to simulate a gen- 1 þ  ki M kp r 
0
eral connection in 3-D-framed analysis. Contact or gap elements are
different from the multi-spring element at their physical nature. The
multi-spring element connects two nodes with coincident coordinates where M and θr are the moment and the rotation of the connection, n is
as shown in Fig. 1. In the scope of this study, the translational and the parameter defining the shape, Rki is the initial connection stiffness,
torsional springs have linear stiffness and fully rigid, while the two Rkp is the strain-hardening stiffness and M0 is the reference moment.

y y

dy θy dy θy

i dx θx Beam-column i dx θx x
j element j
dz θz dz θz

z Connection element z Connection element

Fig. 1. A space connection element model with zero-length.
194 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206

Lui and Chen [9] proposed the following exponential model: form of member basic force and deformation relationship of three-
dimensional (3-D) beam–column element can be expressed as
X
n  
jθ j
− r
M ¼ M0 þ C j 1− exp 2jα þ Rkf jθr j ð5Þ 2 3
EA
j¼1
6 0 0 0 0 07
6 L 7
6 EIy EI y 7
8 9 6 0 S1y S2y 0 0 0 78
in which M and |θr| are the moment and the absolute value of the rota- 6 L L 7 δ 9
>
> P > 6 7> >
tional deformation of the connection, α is the scaling factor, Rkf is >
>
>
>
> 6 7>
> >
> θyA >
>
>
> M >
yA > 6 EIy EI y 7>
> >
>
the strain-hardening stiffness of the connection, M0 is the initial mo- >
<M > = 6 07 >
< >
6 0 S2y
L
S1y
L
0 0 7 θyB =
ment, Cj is the curve-fitting coefficient, and n is the number of terms yB
¼6 7 ð7Þ
> > 6 EI z EI z 7> >
>
> M >
> 6 07 > θzA >
considered. >
>
zA
> 6 0 0 0 S1z S2z 7>
> >
>
>
> M > 6> L L 7>
> θzB >
> >
: zB > ; 6 7: >
;
T 6 EI z EI z 7 ϕ
2.1.3. Cyclic behavior of rotational springs 6 0 0 0 S2z S1z 07 7
6 L L
The independent hardening model shown in Fig. 2 is used to repre- 6 7
4 5
sent for the cyclic behavior of semi-rigid connections because of its sim- GJ
0 0 0 0 0
ple application [16]. The virgin M − θr relationship is defined by the L
connection models in Eqs. (3)–(5). The instantaneous tangent stiffness
of the connections is determined by taking derivative of Eqs. (3)–(5). where P, MyA, MyB, MzA, MzB, and T are incremental axial force, end mo-
Hysteretic behavior of semi-rigid connections is as follows: ments with respect to y and z axes, and torsion, respectively; δ, θyA,
θyB, θzA, θzB, and ϕ are the incremental axial displacement, the joint
1) If a connection is initially loaded, M ⋅ ΔM is positive and the M − θr rotations, and the angle of twist; A, Iy, Iz, J and L are area, moment of
curve follows the line OA with the initial stiffness Rki shown in Fig. 2. inertia with respect to y and z axes, torsional constant, and length
The instantaneous tangent stiffness will be Rkt ¼ ddM . of beam–column element; E and G are elastic and shear modulus of
jθr j
material; S1n and S2n are the stability functions with respect to the n
2) At point A, if the connection is unloaded, M ⋅ ΔM is negative and
axis (n = y, z) given by Chen and Lui [17] as
the M − θr curve goes back along the line ABC with the initial stiff-
ness Rki.
8 pffiffiffiffiffiffi
pffiffiffiffiffiffi
pffiffiffiffiffiffi
3) At point C, if the connection is continuously unloaded, M ⋅ ΔM is > π ρn sin π ρn −π2 ρn cos π ρn
>
>


positive and the M − θr curve follows the line CD with the initial >
< 2−2 cos πpffiffiffiffiffi
ffi pffiffiffiffiffiffi pffiffiffiffiffiffi if P b 0
ρn −π ρn sin π ρn
stiffness Rki followed by the tangent stiffness Rkt. S1n ¼
ffi
pffiffiffiffiffiffi ð8aÞ
> π2 ρ cosh πpffiffiffiffiffi
>
pffiffiffiffiffiffi
ρ −π ρn sinh π ρn
4) At point D, if the connection is reloaded, M ⋅ ΔM is negative and the >
> n

pffiffiffiffiffinffi
pffiffiffiffiffiffi if P N0
: pffiffiffiffiffiffi
M − θr curve follows the straight line DE with the initial stiffness Rki. 2−2 cosh π ρn þ π ρn sinh π ρn
5) At point E, if the connection is continuously reloaded, the M − θr
curve follows the line EF which is similar to the line OA. 8 pffiffiffiffiffiffi
pffiffiffiffiffiffi
>
>
2
π ρn −π ρn sin π ρn
6) At point F, the connection shows a similar curve to steps 1)–5). >
>
ffi pffiffiffiffiffiffi
pffiffiffiffiffiffi if P b 0
< 2−2 cos π pffiffiffiffiffi
ρn −π ρn sin π ρn
S2n ¼ pffiffiffiffiffiffi
pffiffiffiffiffiffi ð8bÞ
>
> π ρn sinh π ρn −π 2 ρn
2.2. Nonlinear beam–column element >
>
pffiffiffiffiffiffi pffiffiffiffiffiffi
pffiffiffiffiffiffi if P N0
:
2−2 cosh π ρn þ π ρn sin π ρn
2.2.1. Stability functions accounting for second-order effects
To capture the effect of the interaction between axial force and
where ρn = |P|/(π2EIn/L2).
bending moment, the stability functions are used to minimize modeling
and solution time. Generally only one element per member is needed to
2.2.2. Refined plastic hinge model accounting for inelastic effects
capture the P − δ effect accurately. From Kim et al. [14], the incremental
The material nonlinearity includes gradual yielding of steel associ-
ated with residual stresses and flexure. The gradual yielding due to re-
sidual stresses is considered by utilizing the Column Research Council
Moment, M (CRC) tangent modulus concept Et, while the gradual yielding due
to flexure is represented by a parabolic function. The relationship be-
F A G
tween the basic force and deformation of 3-D beam–column is modi-
fied to account for the inelastic effects as
Rkt
8 9 2 38 9
Rkt B >
> P > > δ >
>M >
> >
>
Et A=L 0 0 0 0 >
> >
Rki Rki >
> yA >
> 6 0 kiiy kijy 0 0 > θyA >
7> >
>
>
<M > 6 7> >
= 6 0 kijy kjjy 0 0 7< θyB =
yB
¼6 7 ð9Þ
> M > 6 0 0 0 kiiz kijz 7> θzA >
>
> zA >
> 6 >
7> >
>
E C > > 4 5> >
>
>
> M > > 0 0 0 kijz kjjz > θzB >
>
: >
O : zB > ; 0 0 0 0 GJ=L ϕ
;
Relative rotation,θr T

Rkt
Rki where

!
S22
 Et I y
kiiy ¼ ηA S1 − 1−ηB ð10aÞ
S1 L
D

Et I y
kijy ¼ ηA ηB S2 ð10bÞ
Fig. 2. The independent hardening model. L
 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206 195

!
S22
 Et I y where
kjjy ¼ ηB S1 − 1−ηA ð10cÞ
S1 L
2
kiiy kjjy −kijy þ kiiy Asz GL
! C iiy ¼ ð15aÞ
kiiy þ kjjy þ 2kijy þ Asz GL
S
 Et I z
2
kiiz ¼ ηA S3 − 4 1−ηB ð10dÞ
S3 L

−kiiy kjjy þ k2ijy þ kijy Asz GL

EI C ijy ¼ ð15bÞ
kijz ¼ ηA ηB S4 t z ð10eÞ kiiy þ kjjy þ 2kijy þ Asz GL
L

!
S24
 Et I z 2
kjjz ¼ ηB S3 − 1−ηA ð10fÞ kiiy kjjy −kijy þ kjjy Asz GL
S3 L C jjy ¼ ð15cÞ
kiiy þ kjjy þ 2kijy þ Asz GL

From Chen and Lui [17], the CRC tangent modulus Et accounting for
the effects of residual stresses and gradual yielding due to axial force
kiiz kjjz −k2ijz þ kiiz Asy GL
is written as C iiz ¼ ð15dÞ
kiiz þ kjjz þ 2kijz þ Asy GL

Et ¼ 1:0E for P ≤0:5P y ð11aÞ

2
−kiiz kjjz þ kijz þ kijz Asy GL
! C ijz ¼ ð15eÞ
P P kiiz þ kjjz þ 2kijz þ Asy GL
Et ¼ 4 1− E for P N0:5P y ð11bÞ
Py Py

The terms ηA and ηB are scalar parameters that allow for gradual in- kiiz kjjz −k2ijz þ kjjz Asy GL
C jjz ¼ ð15fÞ
elastic stiffness reduction of the element associated with plastification kiiz þ kjjz þ 2kijz þ Asy GL
at ends A and B, respectively. These terms are equal to 1.0 when the
element is elastic, and zero when a plastic hinge is formed. The param-
eter η is assumed to vary according to the parabolic function as where Asy and Asz are the shear areas with respect to the y and z axes,
respectively.
η ¼ 1:0 for α ≤ 0:5 ð11cÞ
2.2.4. Element stiffness matrix accounting for P − Δ effect
The incremental end forces and displacements used in Eq. (14) are
η ¼ 4α ð1−α Þ for α N0:5 ð11dÞ
shown in Fig. 3a. The sign convention for the positive directions of ele-
ment end forces and displacements of a frame member is shown in
where α is a force-state parameter predicted by the magnitude of the Fig. 3b. By comparing the two figures, we can express the equilibrium
axial force and bending moment at the element end. The term α can and kinematic relationships in symbolic form as
be expressed in yield surface proposed by Orbison [15] as
T
2 2 4 2 2 6 2 4 2 f f n g ¼ ½T 612 f f e g ð16Þ
α ¼ 1:15p þ mz þ my þ 3:67p mz þ 3:0p my þ 4:65mz my ð12Þ

where fde g ¼ ½T 612 fdL g ð17Þ

p ¼ P=P y ð13aÞ

y
my ¼ My =M py for weak axis ð13bÞ
MyA,θyA MyB,θyB
mz ¼ Mz =Mpz for strong axis ð13cÞ T,φ P,δ L P, δ T,φ x

where Py, Myp, and Mzp are squash load, and plastic moment capacity of MzA,θzA MzB ,θzB
the cross-section about the y and z axes, respectively.
z (a)
2.2.3. Shear deformation effect y
To account for the shear deformation effect, the member basic force
and deformation relationship of 3-D beam–column is modified as r5, d 5 r11, d 11
follows:
r2, d 2 r8, d 8
2 3
8 9 Et A r4, d 4 r1, d 1 L r7, d 7 r10, d 10 x
> P > 6 L 0 0 0 0 0 78 δ 9
>
> > 6 7>
> >
> MyA >
> >
> 6 0 C iiy C ijy 0 0 07 >θ >
> >
>
>
> > 7>
> yA >>
<M > = 6
6 0 C ijy 07 < = r3, d 3 r9, d 9
C jjy 0 0 7 θ
¼6
yB
6 0 7
yB
ð14Þ r6, d 6
>
> M zA >
> 6 0 0 C iiz C ijz 0 7>> θzA >> r12, d 12
>
> >
> >
>θ > >
>
>
> M >
zB >
6
> 6 0 0 0 C ijz C jjz 07 7>
> zB >
: >
;
z (b)
: ; 4 GJ 5 ϕ
T 0 0 0 0 0
L Fig. 3. Element end force and displacement notations of a beam element.
196 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206

where {fn} and {dL} are the incremental end force and displacement Newton–Raphson method is proposed for the numerical integration of
vectors of a beam–column member expressed as the nonlinear equation of motion because the HHT method possesses
unconditional stability and second-order accuracy. In addition, it can in-
T
f f n g ¼ f r n1 rn2 r n3 r n4 r n5 r n6 r n7 r n8 r n9 r n10 r n11 r n12 g duce numerical damping in the nonlinear solution which is not possible
ð18Þ with the regular Newmark method [19]. The incremental equation of
motion of a structure can be modified as

T n o n o n o
fdL g ¼ f d1 d2 d3 d4 d5 d6 d7 d8 d9 d10 d11 d12 g ð19Þ ½M ΔD ˙˙
tþΔt
þ ð1 þ α Þ½C L  ΔḊ
tþΔt
þ ð1 þ α Þ½K T  ΔD
tþΔt
¼…
n o n n o ð27Þ
tþΔt t t
and [T]6× 12 is a transformation matrix written as ΔF ext þ α ½C L  ΔḊ g þ α ½K T  ΔD

2 3 n o n o
−1 0 0 0 0 0 1 0 0 0 0 0
where ΔḊ˙ , ΔḊ , and {ΔD} are the vectors of incremental accelera-
6
6 0 0 −1=L 0 1 0 0 0 1=L 0 0 077
6 0 0 −1=L 0 0 0 0 0 1=L 0 1 077 tion, velocity, and displacement, respectively; [M], [CL], and [KT] are
½T 612 ¼6
6
6 0 1=L 0 0 0 1 0 −1=L 0 0 0 077 mass, damping, and tangent stiffness matrices, respectively; {ΔFext} is
4 0 1=L 0 0 0 0 0 −1=L 0 0 0 15 the external incremental load vector; superscripts t and t + Δt are
0 0 0 1 0 0 0 0 0 −1 0 0 used to distinguish the values at time t and t + Δt. The viscous damping
ð20Þ matrix [CL] can be defined as Rayleigh damping matrix [20]:

Using the transformation matrix by equilibrium and kinematic rela- ½C L  ¼ α M ½M þ βK ½K L  ð28Þ
tions, the force–displacement relationship of a beam–column member
may be written as where αM and βK are mass- and stiffness-proportional damping factors,
respectively. In the scope of this paper, these coefficients are assumed to
f f n g ¼ ½K n fdL g ð21Þ be constant in nonlinear dynamic time-history analysis. [KL] is the last-
committed stiffness matrix. The last-committed stiffness matrix is the
where [Kn] is the element stiffness matrix expressed as tangent stiffness matrix at the starting time of the new time step, [KL]
is unchanged during the unbalanced iterative procedure at each time
T step. If both modes are assumed to have the same damping ratio ξ, then
½K n 1212 ¼ ½T 612 ½K e 66 ½T 612 ð22Þ
2ω1 ω2 2
αM ¼ ξ βK ¼ ξ ð29Þ
Eq. (21) is used to enforce no side-sway in the member. If the mem- ω1 þ ω2 ω1 þ ω2
ber is permitted to sway, additional axial and shear forces will be in-
duced in the member. We can relate these additional axial and shear where ω1 and ω2 are the natural frequencies of the first and second
forces due to a member sway to the member end displacements as modes of the considered frame following decisive direction of dynamic
n o h i loading.
f g ¼ K g fdL g ð23Þ

where
" # Previous time step
h i ½K s  − ½K s 
Kg ¼ T ð24Þ
1212 − ½K s  ½K s 
Form the tangent stiffness matrix
in which

2 3 Form the last-committed damping matrix

0 a −b 0 0 0
6 a
6 c 0 0 0 077
Note: only at the first iteration
6 −b 0 c 0 0 077
½K s  ¼ 6 ð25Þ
6 0
6 0 0 0 0 077
4 0 0 0 0 0 05 Solve for the increment displacement
Next iteration
Current time step

0 0 0 0 0 0

and Update element force

MzA þ MzB M yA þ MyB P

a¼ ; b¼ ; c¼ ð26Þ
L2 L2 L Calculate the residual force R

By combining Eqs. (21) and (23), we obtain the stiffness matrix

[K]t = [Kn] + [Kg] that includes the terms representing axial and flexural No Check convergence of R?
actions simultaneously.
Yes
3. Nonlinear dynamic analysis algorithm Update structural response D, D, D

3.1. Formulations of the proposed algorithm

Next time step
An algorithm based on combination of the Hilber–Hughes–Taylor
(HHT) method [18] (also known as the alpha-method) and the Fig. 4. A flow chart of the proposed program.
 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206 197

0.9 50 kN 100 kN 50 kN
δ
0.6 0.5F(t)
E = 205 x 10 6 kN/m²
0.3 All of members: W8x48 σ y = 235 MPa
Acceleration (g)

3.0 m
50 kN 100 kN 50 kN
0.0 Lumped mass:
0 5 10 15 20 25 30 = 10.2 T
-0.3 F(t) = 5.1 T

-0.6 Connection "C" F(t)

3.0 m
-0.9 ψ ψ 100 kN

-1.2 ψ = 1/438
Time (s) 0.5
a) El Centro 2 x 4.0 = 8.0 m
time, t (s)

0.9
Fig. 6. A two-story 2-D steel frame.
0.6

0.3
Acceleration (g)

Transforming Eqs. (30) and (31), the incremental velocity and accel-
eration at the first iteration of each time step can be written as
0.0
0 5 10 15 20 25
n o   n o
-0.3 γ n tþΔt o γ n t o γ
Δt Ḋ˙
tþΔt t
ΔḊ ¼ ΔD − Ḋ þ 1− ð32Þ
β  Δt β 2β
-0.6

-0.9 n o 1 n tþΔt o 1 n to 1 n ˙ to
ΔḊ˙
tþΔt
¼ ΔD − Ḋ − Ḋ ð33Þ
β  Δt 2 β  Δt 2β
-1.2
Time (s)
b) Northridge Substituting Eqs. (32) and (33) into Eq. (27), the incremental dis-
0.9 placement can be calculated from
h in o n o
0.6 ^ ΔDtþΔt ¼ Δ F^
K ð34Þ

0.3 h i n o
Acceleration (g)

^ and Δ F^ are the effective stiffness matrix and incrementally

where K
0.0
0 10 20 30 40 effective force vector, respectively, given as
-0.3 h i
^ ¼ ð1 þ α Þ½K  þ ð1 þ α Þ γ 1
K T ½C  þ ½M ð35Þ
-0.6 β  Δt L β  Δt 2

-0.9
n o n o n o  
1 n to 1 n ˙ to
Δ F^ ¼ ΔF
tþΔt t
-1.2 þ α ½K T  ΔD þ ½M  Ḋ þ Ḋ
Time (s) β  Δt 2β
   n o o ð36Þ
c) San Fernando γ n to γ n
Δt Ḋ˙ þ α ΔḊ
t t
þ½C L  ð1 þ α Þ Ḋ −ð1 þ α Þ 1−
β 2β
Fig. 5. Earthquake records.

The residual forces in each time step can be eliminated by using the
Newton–Raphson iterative procedure. At the first iteration of each time
Using Newmark's approximate equations in standard form [19] as: step, the total displacement, velocity, and acceleration at the time t + Δt
n o n o n o 1  n o n o
−β Δt Ḋ˙ þ β  Δt Ḋ˙
tþΔt t t 2 t 2 tþΔt
D ¼ D þ Δt Ḋ þ ð30Þ
2 Table 2
Peak displacements (mm) of the two-story 2-D steel frame.
n n o n o n o
g ¼ Ḋ þ ð1−γÞΔt Ḋ˙ þ γ  Δt Ḋ˙
tþΔt t t tþΔt Frame type Max/min Analysis type Present Chan and Chui SAP2000
Ḋ ð31Þ
Rigid Max Elastic 98.82 93.25 98.46
Inelastic 137.37 125.25 135.55
Min Elastic −81.96 −77.01 −81.69
Inelastic −2.89 0.34 −6.90
Table 1 Linear semi-rigid Max Elastic 120.45 116.38 121.68
Peak ground acceleration and its corresponding time step of the earthquake records. Inelastic 159.47 151.42 161.02
Min Elastic −109.22 −108.90 −112.39
Earthquake PGA Time
Inelastic −22.18 −9.82 −16.84
(g) step (s)
Nonlinear Max Elastic 140.73 138.60
El Centro (1940) (Array, #9, USGS Station 117) 0.319 0.020 semi-rigid Inelastic 184.75 181.10
Northridge (1994) (Simi Valley-Katherine, 090, USC Station 90055) 0.640 0.010 Min Elastic −84.12 −82.10
San Fernando (1971) (Pacoima Dam, 254, CDMG Station 279) 1.160 0.010 Inelastic 4.62 16.02
198 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206

0.15 Proposed For the second and subsequent iterations of each time step, the struc-
Chan and Chui tural system is solved under the effect of the residual force vector {R} as
0.10 SAP2000
h i n o
^
Lateral displacement (m)

tþΔt
K δΔD ¼ fRgk ð40Þ
k kþ1
0.05 h i
^ and the residual force vector {R}k
where the effective stiffness matrix K
k
0.00 are calculated at the unbalanced iterative step k, respectively, as follows
0.0 0.5 1.0 1.5 2.0 2.5 3.0
h i γ 1
^ ¼ ð1 þ α Þ½K  þ ð1 þ α Þ
K ½C  þ ½M  ð41Þ
-0.05 T k
β  Δt L
k β  Δt 2
n o
-0.10 tþΔt
fRgk ¼ F ext −f F int gk −f F dam gk −f F ine gk ð42Þ

-0.15 Time (s)

a) Rigid connections 0.18 Proposed, refined plasic hinge
Chan and Chui, lumped plasic hinge
Proposed 0.15 SAP2000, FEMA356 - plasic hinge
0.15

Lateral displacement (m)

Chan and Chui
SAP2000 0.12
0.10
Lateral displacement (m)

0.09
0.05
0.06
0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.03

-0.05 0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0
-0.10 -0.03 Time (s)
a) Rigid connections
-0.15 Time (s)
b) Linear semi-rigid connections 0.18 Proposed, refined plasic hinge
Chan and Chui, lumped plasic hinge
0.15 0.15 SAP2000, FEMA356 - plasic hinge
Lateral displacement (m)

Proposed
Chan and Chui 0.12
0.10
Lateral displacement (m)

0.09
0.05
0.06

0.00 0.03
0.0 0.5 1.0 1.5 2.0 2.5 3.0
-0.05 0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0
-0.03 Time (s)
-0.10
b) Linear semi-rigid connections
-0.15 Time (s)
0.21
c) Nonlinear semi-rigid connections Proposed, refined plasic hinge
Chan and Chui, lumped plasic hinge
0.18
Lateral displacement (m)

Fig. 7. Second-order elastic responses of the two-story 2-D steel frame.

0.15
+ Δt
are updated based on the incremental displacement {ΔDt } as 0.12
follows:
0.09
n o n o n o
tþΔt t tþΔt
D ¼ D þ ΔD ð37Þ 0.06

0.03
n o  γ
 n o  
γ n to γ n tþΔt o 0.00
Δt Ḋ˙ þ 1−
tþΔt t
Ḋ ¼ 1− Ḋ þ ΔD ð38Þ 0.0 0.5 1.0 1.5 2.0 2.5 3.0
2β β β  Δt
Time (s)
n o   c) Nonlinear semi-rigid connections
1 n ˙ to 1 n to 1 n tþΔt o
Ḋ˙
tþΔt
¼ 1− Ḋ − Ḋ þ ΔD ð39Þ
2β β:Δt β:Δt 2 Fig. 8. Second-order inelastic responses of the two-story 2-D steel frame.
 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206 199

250 P W5x16
Moment (kN.m)
200

5.5 ft (66 in)

W5x16

W5x16
150

100
Connection "C"
50 2P W5x16
0

5.5 ft (66 in)

-0.006 0 0.006 0.012 0.018 0.024 0.03
-50

W5x16

W5x16
-100 E = 29,000 ksi
σy = 36 ksi
-150 Rotation (rad)
Proposed, rigid con. Chan and Chui, rigid con.
Proposed, linear con. Chan and Chui, linear con.
Proposed, nonlinear con. Chan and Chui, nonlinear con. 9 ft (108 in)
a) Elastic responses
a) Configuration
200
Moment (kN.m)

5
150 4
3

Lateral load, 2P (kip)

100 2
1
50
0
0 1 2 3 4 5
0 -1
-0.006 0 0.006 0.012 0.018 -2
-50 -3
-4
-100
-5 Load cycle
-150 Rotation (rad)
Proposed, rigid con. Chan and Chui, rigid con.
b) Lateral load history
Proposed, linear con. Chan and Chui, linear con. 200
Proposed, nonlinear con. Chan and Chui, nonlinear con.
175
b) Inelastic responses
Connection Moment (kip.in)

150
Fig. 9. Hysteretic loops at the connection C for various analyses of the two-story 2-D steel
frame. 125

100 Upper bound of 14 tests - Stelmack et al.

Lower bound of 14 tests - Stelmack et al.
where {Ftext+ Δt} is the total external force vector; the inertial force vector 75 Trilinear model - Stelmack et al.
{Fine}k, the damping force vector {Fdam}k, and the updated internal Richard-Abbott model - Proposed
50
force vector {Fint}k at the unbalanced iterative step k are respectively
defined as: 25
n o 0
f F ine gk ¼ ½M  Ḋ˙
tþΔt
ð43Þ 0.00 0.01 0.02 0.03 0.04 0.05
k
Connection Rotation (rad)
n
tþαΔt
o c) Moment-rotation curves of 1/2-in.angle connections
f F dam gk ¼ ½C L  Ḋ ð44Þ
k
Fig. 10. An experimental two-story 2-D steel frame.

n n o o
tþαΔt
f F int gk ¼ F int D ð45Þ
k
Once the convergence criterion is satisfied, the structural response is
updated for the next time step as
where the displacement vector and velocity vector at the intermediate
time are given by
n o n o n o
tþΔt tþΔt tþΔt
ΔD ¼ ΔD þ δΔD ð48Þ
n o n o n o kþ1 k kþ1
tþαΔt tþΔt t
D ¼ ð1 þ α Þ D −α D ð46Þ
k k

n o n o n o n o n o n o n o
tþαΔt tþΔt t tþΔt tþΔt t tþΔt
Ḋ ¼ ð1 þ α Þ Ḋ −α Ḋ ð47Þ D ¼ D ¼ D þ ΔD ð49Þ
k k kþ1 kþ1
200 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206

200 200
150 150
100 100

2.2 m
Moment (kip.in)
Moment (kip.in)

50 50 All members
0 0 W8x31
-50 -50
-100 -100

1.76 m
-150 -150
Stelmack's Experiment Stelmack's Experiment
-200 Stelmack's Analysis -200 Stelmack's Analysis
Proposed Analysis Proposed Analysis
-250 -250 Z
-0.02 -0.01 0.00 0.01 0.02 -0.02 -0.01 0.00 0.01 0.02
Rotation (rad) Rotation (rad) Y
Cycle 3 Cycle 4 m
X 2.5
200 3.0 m
150
Fig. 12. A two-story 3-D steel frame.
100
Moment (kip.in)

50
3.2. Application of the proposed algorithm
0
As indicated in [21], the HHT method will possess the unconditional
-50 
stability and second-order accuracy when α ∈ − 13 ; 0 and
-100
2
-150 1 ð1−α Þ
Stelmack's Experiment γ¼ −α β¼ ð52Þ
2 4
-200 Stelmack's Analysis
Proposed Analysis
-250 With the smaller value of α, the more numerical damping is induced
-0.04 -0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 0.04 in the solution, it is necessary to obtain a convergent solution for some
Rotation (rad) complicated nonlinear problems. Especially if the choice α = 0 leads
Cycle 5 to the average acceleration method of the Newmark family (γ = 0.5,
a) Comparing with Stelmack et al.’s results β = 0.25) with zero numerical damping.

200 4. Numerical verification

150
A computer program written in Fortran programming is developed
100 based on the above-mentioned algorithm. The flow chart of the
Moment (kip.in)

proposed program for the application of the HHT method and the
50
Newton–Raphson method is illustrated in Fig. 4. In this study, the inte-
0 gration coefficients of the HHT method applied for the proposed pro-
gram are α = 0, γ = 0.5 and β = 0.25. Three earthquake records
-50 of the El Centro, the Northridge, and the San Fernando as shown in
-100 Fig. 5 are used as ground excitation in dynamic analysis. Their peak
ground accelerations and time steps are listed in Table 1. Several numer-
-150 Proposed Analysis ical examples are presented and discussed to verify the accuracy and
efficiency of the proposed program in predicting the nonlinear response
-200
-0.04 -0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 0.04 of semi-rigid steel frame structures subjected to dynamic loadings.
Rotation (rad) For verification purposes, the predictions obtained from the proposed
program are compared with available results reported in the literature,
b) Proposed analysis and those generated by SAP2000. It should be noted that SAP2000
did not provide a semi-rigid element accounting for nonlinear effects
Fig. 11. Moment–rotation loops at the connection C of the Stelmack experimental frame. of beam-to-column connection, whereas the proposed element can
consider these effects.

n o n o   n o  
γ γ n to γ n tþΔt o Table 3
Δt Ḋ˙ þ 1−
tþΔt tþΔt t
Ḋ ¼ Ḋ ¼ 1− Ḋ þ ΔD
kþ1 2β β β  Δt kþ1 Comparison of first two natural periods (s) along the applied earthquake direction of the
two-story 3-D steel frame.
ð50Þ
Frame type Mode SAP2000 Present Diff. (%)

n o n o   Rigid 1 0.9849 0.9849 0.00

1 n ˙ to 1 n to 1 n tþΔt o
Ḋ˙ ¼ Ḋ˙ 2 0.2949 0.2949 0.00
tþΔt tþΔt
¼ 1− Ḋ − Ḋ þ ΔD
kþ1 2β β  Δt β  Δt 2 kþ1 Semi-rigid 1 1.2128 1.2126 −0.02
2 0.3109 0.3109 0.00
ð51Þ
 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206 201

150 Proposed, rigid con. 150 Proposed, linear semi-rigid

SAP2000, rigid con. SAP2000, linear semi-rigid

Lateral displacement (mm)

Lateral displacement (mm)
Proposed, nonlinear semi-rigid
100 100

50 50

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
-50 -50

-100 -100

-150 -150 Time (s)

Time (s)
a) Second-order elastic responses of the rigid frame b) Second-order elastic responses of the semi-rigid frame
150 Proposed, rigid con. 150 Proposed, linear semi-rigid
SAP2000, rigid con. SAP2000, linear semi-rigid
Lateral displacement (mm)

Proposed, nonlinear semi-rigid

Lateral displacement (mm)

100 100

50 50

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
-50 -50

-100 -100

-150 -150 Time (s)

Time (s)
c) Second-order inelastic responses of the rigid frame d) Second-order inelastic responses of the semi-rigid frame
Fig. 13. Nonlinear time-history responses of the two-story 3-D steel frame subjected to the El Centro earthquake.

250 Proposed, rigid con. 250 Proposed, linear semi-rigid

SAP2000, rigid con. SAP2000, linear semi-rigid
200 200
Lateral displacement (mm)
Lateral displacement (mm)

Proposed, nonlinear semi-rigid

150 150
100 100
50 50
0 0
0 5 10 15 20 25 0 5 10 15 20 25
-50 -50
-100 -100
-150 -150
-200 -200
-250 Time (s) -250 Time (s)
a) Second-order elastic responses of the rigid frame b) Second-order elastic responses of the semi-rigid frame
250 Proposed, rigid con. 250 Proposed, linear semi-rigid
SAP2000, rigid con. SAP2000, linear semi-rigid
200 200
Lateral displacement (mm)

Lateral displacement (mm)

Proposed, nonlinear semi-rigid

150 150
100 100
50 50
0 0
0 5 10 15 20 25 0 5 10 15 20 25
-50 -50
-100 -100
-150 -150
-200 -200
-250 Time (s) -250 Time (s)
c) Second-order inelastic responses of the rigid frame d) Second-order inelastic responses of the semi-rigid frame
Fig. 14. Nonlinear time-history responses of the two-story 3-D steel frame subjected to the Northridge earthquake.
202 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206

4.1. Two-story 2-D steel frame Table 4

Parameters of semi-rigid connections follow the Kishi–Chen model.

A single-bay two-story 2-D steel frame with flexible beam-to- Beam section Bending-axis Mu (kN·m) Rki (kN·m/rad) n
column connections was studied by Chan and Chui [22]. The geometry
W12 × 87 Strong-axis 300 160,503.2 1.57
and loading of the frame are given in Fig. 6. All the frame members are Weak-axis 300 52,267.75 1.57
W8 × 48 with Young's modulus E of 205 × 106 kN/m2 and the yielding W12 × 53 Strong-axis 300 92,185.09 1.57
stress of the steel material σy of 235 MPa is considered. An initial geo- Weak-axis 300 20,776.82 1.57
W12 × 26 Strong-axis 200 44,247.8 0.86
metric imperfection of column ψ of 1/438 is considered. The vertical
Weak-axis 200 3752.54 0.86
static loads are applied on the frame to consider the second-order ef-
fects followed by the horizontal forces applied suddenly at each floor
during 0.5 s, as shown in Fig. 6. The lumped masses of 5.1 and proposed element are in good agreement with those predicted by
10.2 tons are modeled at the top of columns and the middle of the Chan and Chui [22], and those results match well with those of
beams, respectively. A time step Δt of 0.001 s is chosen and viscous SAP2000. In the case of the inelastic responses as shown in Fig. 8, it
damping of structure is ignored in the dynamic analysis. The four pa- can be recognized that the main source of the differences is due to the
rameters of the Richard–Abbott model for semi-rigid connections are: modeling of material nonlinearity. The proposed program employs the
Rki = 23, 000 kN ⋅ m/rad, Rkp = 70 kN ⋅ m/rad, Mo = 180 kN ⋅ m, and refined plastic hinge method based on stability functions, whereas
n = 1.6. Table 2 shows the peak displacements generated by the pro- Chan and Chui's research and SAP2000 use the lumped plastic hinge
posed program, Chan and Chui [22], and SAP2000. As shown in Fig. 7, method. The moment–rotation curves at connection C are also plotted
it can be seen that the results of the elastic responses obtained by the in Fig. 9.

Y 4.2. Experimental two-story 2-D steel frame

W12x26 W12x26 This numerical example is to verify the proposed connection

modeling with the results of the experimental test and numerical anal-
ysis of Stelmack et al. (1986) [23]. The geometry and applied loads of the
test frame are shown in Fig. 10a and b. All members are W5 × 16 A36
W12x53

W12x87

W12x53
7.315 m

steel. Young's modulus of the material is 29,000 ksi. 1/2-in angle con-
nections were used in the test, and fourteen monotonic moment–rota-
tion test results of such connections were reported by Stelmack et al.
[23]. The upper and lower bounds obtained from the test data are
shown in Fig. 10c. The trilinear model was utilized by Stelmack et al.
W12x26 W12x26
X [23] while the Richard–Abbott model is used in the present study, as
7.315 m 7.315 m plotted in Fig. 10c. The four parameters of the Richard–Abbott model
are determined by a curve-fitting procedure as follows: Rki =
20, 000 kip ⋅ in/rad, Rkp = 800 kip ⋅ in/rad, Mo = 150 kip ⋅ in, and
a) Plan view n = 1.5. The lateral load cycles begin with loads applied to the first
story of ± 1 kip and increased by 1 kip increments up to ± 5 kips, as
Z
shown in Fig. 10b; the second-story load is a half of the first-story load.
Joint C The experimental and analytical results of the moment–rotation
Lumped mass: loops at the connection C are separately compared for each load cycle
Node A as shown in Fig. 11a. Fig. 11b shows the moment–rotation loops of the
= 128.42 kN.sec²/m
proposed analysis during the load cycles. It can be seen that the results
W10x60

W10x60

= 256.84 kN.sec²/m
predicted by the proposed analysis using the Richard–Abbott model
: Semi-rigid joint
W10x60
W10x60

are relatively close to the experimental results and smoother than the
results of Stelmack et al. [23] using the trilinear model. It may be con-
cluded that the proposed program can generate acceptably accurate
H = 6 x 3.658 m = 21.948 m

and smooth moment–rotation curves in predicting the nonlinear be-

havior of steel frames with semi-rigid connections.

4.3. Two-story 3-D steel frame

The next example is a two-story 3-D steel frame as shown in Fig. 12.
W12x120
W12x87

W12x87

The nonlinear inelastic dynamic response of two-story 3-D steel frames

with various connection types (fully rigid, linear semi-rigid, and non-
W12x120

W12x87
W12x87

linear semi-rigid connections) subjected to two different earthquakes

(El Centro and Northridge) shown in Fig. 5 was studied. This example

Table 5
Y Comparison of first two natural periods (s) along the applied earthquake direction of the
on
oti six-story 3-D steel frame.
n dm X
ou Frame type Mode SAP2000 Present Diff. (%)
Gr
Rigid 1 5.5385 5.5386 +0.002
b) Perspective view 2 1.9958 1.9959 +0.005
Semi-rigid 1 6.0846 6.0797 −0.08
2 2.1340 2.1338 −0.009
Fig. 15. A six-story 3-D steel frame.
 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206 203

also aims at verifying the accuracy of the beam–column element in
predicting the nonlinear effects of 3-D framed structures. The section
of all members was W8 × 31. Young's modulus and Poisson ratio of
the material are 200 GPa and 0.3 respectively. A steel yield stress of
500 MPa was assumed for the nonlinear inelastic time-history analysis
of the frame. The Chen–Lui exponential model was used for the
flush end plate connection of semi-rigid joints. The parameters of
the model are: Rki = 12,340.198 kN·m/rad; Rkf = 108.924 kN·m/rad;
Mo = 0.0 kN·m; α = 0.00031783; C1 = −28.286; C2 = 573.189; C3 =
−3433.98; C4 = 8511.3; C5 = −9362.567; and C6 = 3832.899 (unit
of Ci is kN·m) [24]. The connection stiffness about the weak-axis
of the sections was assumed to be one fifth of the stiffness about
the strong-axis. The masses lumped at the framed nodes were assumed
to be 50 Ns2/mm. The earthquake excitations were applied in the
X-direction. The first two periods corresponding to the frame types
are compared with those of SAP2000 listed in Table 3, these periods
are used to predict viscous damping of the structure with damping
ratio ξ of 0.05.
In the numerical modeling, only one element per member is used in
the proposed program. Since the beam element provided by SAP2000
cannot capture accurately the second-order effects if only one element
per member is used, the member needs to be divided into many ele-
ments to capture accurately the second-order effects. In this study, six
elements per member are automatically used in the modeling of the
framed structure in SAP2000. Moreover, SAP2000 does not also include
a beam-to-column connection element considering the nonlinearity of
the moment–rotation relationship.
The displacement responses at the roof of the rigid and semi-rigid
frames generated by the proposed program and SAP2000 are shown
in Figs. 13 and 14 for the El Centro and Northridge earthquakes, respec-
tively. It can be observed that the obtained results are identical with
those of SAP2000 in all cases of nonlinear elastic analysis. In the case
of the nonlinear semi-rigid frame, the response shows the displace-
ment drift due to permanent rotational deformation at connections
as shown in Fig. 14b. In nonlinear inelastic analysis, there is a slight
difference between the proposed results and those of SAP2000 be-
cause the steel yielding model and the structural viscous damping
are estimated differently. In this numerical analysis, SAP2000 is ap-
plied to the lumped plastic hinge model following the FEMA 356 spec-
ification and the initial structural stiffness matrix is used to calculate
the damping matrix, whereas the proposed program uses the refined
plastic hinge model and the damping matrix of the structure is
recalculated at each time step based on the tangent stiffness matrix
of the structure.
4.4. Six-story 3-D steel frame
A six-story 3-D steel frame with semi-rigid connections is plotted in
Fig. 15. Static problem of this frame is solved by Chiorean [25], Ngo-Huu
et al. [26], and Nguyen and Kim [27]. The nonlinear elastic dynamic be-
havior was investigated by Nguyen and Kim [28]. The present study in-
vestigated the nonlinear dynamic behavior of the frame subjected to
the El Centro and San Fernando earthquakes, these earthquake records
are shown in Fig. 5. The elastic modulus and shear modulus are
206,850 MPa and 79,293 MPa, respectively. A yield stress of 250 MPa is
assumed for the nonlinear inelastic dynamic analysis of the frame. Three
parameters of the Kishi–Chen model for the semi-rigid joints are listed
in Table 4 [25]. The lumped masses of 128.42 and 256.84 kN·s2/m are
transferred from the uniform floor pressure of 9.6 kN/m2, and they are
assigned at the frame nodes as shown in Fig. 15b. The earthquake exci-
tations are applied in the Y-direction. Viscous damping is considered by
using the Rayleigh damping matrix with a damping ratio ξ of 0.05,
where the first two natural periods of the frames corresponding to the
earthquake direction of Y are compared in Table 5. It can be seen that

0.40 Proposed, rigid con.

0.40
0.30 SAP2000, rigid con. 0.30
Lateral displacement (m)
Lateral displacement (m)

0.20 0.20
0.10 0.10
0.00 0.00
0 5 10 15 20 25 30 0 5 10 15 20 25 30
-0.10 -0.10
-0.20 -0.20
-0.30 -0.30
Proposed, linear semi-rigid
-0.40 -0.40 SAP2000, linear semi-rigid
Proposed, nonlinear semi-rigid
-0.50 -0.50
Time (s) Time (s)
a) Second-order elastic responses of the rigid frame b) Second-order elastic responses of the semi-rigid frame
0.40 Proposed, rigid con. 0.40 Proposed, linear semi-rigid
SAP2000, rigid con. SAP2000, linear semi-rigid
0.30 0.30
Proposed, nonlinear semi-rigid
Lateral displacement (m)
Lateral displacement (m)

0.20 0.20
0.10 0.10
0.00 0.00
0 5 10 15 20 25 30 0 5 10 15 20 25 30
-0.10 -0.10
-0.20 -0.20
-0.30 -0.30
-0.40 -0.40
-0.50 Time (s) -0.50 Time (s)
c) Second-order inelastic responses of the rigid frame d) Second-order inelastic responses of the semi-rigid frame
Fig. 16. Nonlinear time-history responses of the six-story 3-D steel frame subjected to the El Centro earthquake.
204 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206

0.20 Proposed, rigid con.

0.20 Proposed, linear semi-rigid
SAP2000, linear semi-rigid
0.15 SAP2000, rigid con. 0.15
Proposed, nonlinear semi-rigid

Lateral displacement (m)

Lateral displacement (m)

0.10 0.10
0.05 0.05
0.00 0.00
0 10 20 30 40 0 10 20 30 40
-0.05 -0.05
-0.10 -0.10
-0.15 -0.15
-0.20 -0.20
-0.25 Time (s) -0.25 Time (s)
a) Second-order elastic responses of the rigid frame b) Second-order elastic responses of the semi-rigid frame

0.20 Proposed, rigid con. 0.20 Proposed, linear semi-rigid

SAP2000, rigid con. SAP2000, linear semi-rigid

0.15 0.15
Proposed, nonlinear semi-rigid
Lateral displacement (m)

Lateral displacement (m)

0.10 0.10
0.05 0.05
0.00 0.00
0 10 20 30 40 0 10 20 30 40
-0.05 -0.05
-0.10 -0.10
-0.15 -0.15
-0.20 -0.20
-0.25 Time (s) -0.25 Time (s)
c) Second-order inelastic responses of the rigid frame d) Second-order inelastic responses of the semi-rigid frame
Fig. 17. Nonlinear time-history responses of the six-story 3-D steel frame subjected to the San Fernando earthquake.

the obtained results by modal analysis of the proposed program and Table 6
SAP2000 are nearly identical. Comparison of peak displacements (mm) at node A of the six-story 3-D steel frame.
Figs. 17 and 16 show the lateral displacement responses at A-node
Earthquakes Max/min Frame type–analysis Displacement (mm) Error (%)
appropriating to the rigid, linear semi-rigid, and nonlinear semi-rigid
type
frames under the El Centro and San Fernando earthquakes, respectively. SAP2000 Present
The results of peak displacement are compared in Table 6. It can be El Centro Max RC–NE 248.19 249.05 +0.35
observed that a strong agreement of displacements predicted by the RC–NI 220.17 222.08 +0.87
proposed program and SAP2000 is obtained in the second-order elastic LC–NE 338.22 342.49 +1.26
LC–NI 238.07 237.55 −0.22
analysis. Since using a different material-nonlinearity approach, the dis- NC–NE 251.10
placement responses of the proposed program and SAP 2000 have a NC–NI 248.59
small shift in the second-order inelastic analysis. Figs. 18 and 19 show Min RC–NE −437.36 −437.15 +0.05
the hysteretic loops of the strong-axis rotational spring at the connec- RC–NI −380.00 −368.04 +3.15
LC–NE −442.89 −442.78 −0.02
tion C corresponding to various earthquakes, respectively, which cannot
LC–NI −419.30 −428.28 −2.14
be obtained by SAP2000. NC–NE −331.82
The peak displacement of the nonlinear semi-rigid frame is less than NC–NI −375.80
that of both the rigid and linear semi-rigid frames because of hysteretic San Fernando Max RC–NE 160.48 160.30 −0.11
damping of the nonlinear semi-rigid connections. It can be concluded RC–NI 79.00 79.04 +0.05
LC–NE 152.69 152.61 −0.05
that the beam-to-column connections play a significant role in the LC–NI 73.40 73.45 +0.07
earthquake resistant design. The accurate prediction of real behavior NC–NE 73.29
of framed structures leads to both safer and more economic design. NC–NI 73.29
Using the same personal computer configuration (AMD Phenom II X4 Min RC–NE −162.30 −162.06 +0.15
RC–NI −216.41 −205.80 +4.90
955 Processor, 3.2 GHz; 4.00 GB RAM), the analysis time of the proposed
LC–NE −125.38 −125.36 +0.02
program and SAP2000 for the nonlinear inelastic time-history analysis of LC–NI −203.02 −193.83 +4.53
the six-story linear semi-rigid frame subjected to San Fernando earth- NC–NE −78.32
quake, which is the problem having the longest analysis time among NC–NI −80.94
all cases, are 3 min 54 s and 4 min 50 s, respectively. This result demon- Note: Rigid connections (RC), linear semi-rigid connections (LC), nonlinear semi-rigid
strates the high computational efficiency of the proposed program. connections (NC); nonlinear elastic analysis (NE), nonlinear inelastic analysis (NI).
 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206 205

350 350

250 250

150 150

Moment (kN.m)
Moment (kN.m)

50 50

-50 -50

-150 -150

-250 -250

-350 -350
-0.012 -0.008 -0.004 0 0.004 0.008 0.012 0.016 -0.012 -0.008 -0.004 0 0.004 0.008 0.012 0.016
Rotation (rad) Rotation (rad)
a) Second order elastic analysis a) Secondorder elastic analysis
350
350
250
250
150

Moment (kN.m)
150
Moment (kN.m)

50
50
-50
-50
-150
-150
-250
-250
-350
-350 -0.012 -0.008 -0.004 0 0.004 0.008 0.012 0.016
-0.012 -0.008 -0.004 0 0.004 0.008 0.012 0.016
Rotation (rad)
Rotation (rad)
b) Second-order inelastic analysis
b) Second-order inelastic analysis
Fig. 19. Hysteretic loops of a strong-axis rotational spring at the connection C for various
Fig. 18. Hysteretic loops of a strong-axis rotational spring at the connection C for various analyses of the six-story 3-D steel frame subjected to the San Fernando earthquake.
analyses of the six-story 3-D steel frame subjected to the El Centro earthquake.

FEA software packages such as NASTRAN, ABAQUS, ANSYS, etc. which

5. Summary and conclusions
A simple and effective numerical procedure for the nonlinear
dynamic time-history analysis of three-dimensional steel frames con-
sidering geometry, material, and connection nonlinearities at the same
time has been presented. Three main resources of damping of steel
frame structures taken into account in the proposed program are:
(1) hysteretic damping due to inelastic material; (2) structural viscous
damping employing Rayleigh damping; (3) hysteretic damping due to
nonlinear beam-to-column connections. An effective integration algo-
rithm was developed for solving the equation of motion of framed struc-
tures based on the combination of the Hilber–Hughes–Taylor method
and the Newton Raphson method. The proposed program was verified
for accuracy and computational efficiency through several numerical
examples with various dynamic loadings. It is also capable of accurately
predicting natural periods of framed structures. The significant differ-
ence between the proposed program and SAP2000 is that SAP2000
cannot simulate the nonlinear behavior of semi-rigid connections
whereas the proposed program can, as illustrated in Section 4.4. More-
over, by using only one element per member, the proposed program can
accurately capture the second-order effects while the beam–column
members of SAP2000 need to be divided into several elements. This pro-
cedure saves computer resources, so it reduces computational time.
Thus, it can be concluded that the proposed program can effectively be
used for practical design in predicting nonlinear inelastic behavior of
3-D semi-rigid steel framed structures subjected to static and dynamic
loadings instead of using the time consuming and costly commercial
cannot accurately consider the nonlinear effect of semi-rigid connec-
tions by a simple and efficient manner.
Acknowledgments
This work was supported by a grant from the Human Resources
Development program of the Korea Institute of Energy Technology
Evaluation & Planning (KETEP) funded by the Korea government
Ministry of Knowledge Economy (No. 20124030200050) and by
the National Research Foundation (NRF) of Korea grant funded by the
Korea government (MEST) (No. 2011-0030847).
References
[1] Azizinamini A, Radziminski JB. Static and cyclic performance of semirigid steel
beam-to-column connections. J Struct Eng-ASCE 1989;115:2979–99.
[2] Nader MN, Astaneh A. Dynamic behavior of flexible, semirigid and rigid steel frames.
J Constr Steel Res 1991;18:179–92.
[3] Elnashai AS, Elghazouli AY. Seismic behaviour of semi-rigid steel frames. J Constr
Steel Res 1994;29:149–74.
[4] Elnashai AS, Elghazouli AY, Denesh-Ashtiani FA. Response of semirigid steel frames
to cyclic and earthquake loads. J Struct Eng-ASCE 1998;124:857–67.
[5] Yu CH, Shanmugam NE. Stability of semi-rigid space frames. Comput Struct
1988;28:85–91.
[6] Xu L. The buckling loads of unbraced PR frames under non-proportional loading. J
Constr Steel Res 2002;58:443–65.
[7] Chen WF, Kishi N. Semirigid steel beam-to-column connections—data-base and
modeling. J Struct Eng-ASCE 1989;115:105–19.
[8] Richard RM, Abbott BJ. Versatile elastic–plastic stress–strain formula. J Eng Mech
Div-ASCE 1975;101:511–5.
206 P.-C. Nguyen, S.-E. Kim / Journal of Constructional Steel Research 101 (2014) 192–206
[9] Lui EM, Chen WF. Analysis and behavior of flexibly-jointed frames. Eng Struct
1986;8:107–18.
[10] Ramberg W, Osgood WR. Description of stress–strain curves by three parameters, in.
Washington DC: National Advisory Committee for Aeronautics; 1943.
[11] Lui EM, Lopes A. Dynamic analysis and response of semirigid frames. Eng Struct
1997;19:644–54.
[12] Awkar JC, Lui EM. Seismic analysis and response of multistory semirigid frames. Eng
Struct 1999;21:425–41.
[13] Sekulovic M, Nefovska-Danilovic M. Contribution to transient analysis of inelastic
steel frames with semi-rigid connections. Eng Struct 2008;30:976–89.
[14] Kim S-E, Park M-H, Choi S-H. Direct design of three-dimensional frames using
practical advanced analysis. Eng Struct 2001;23:1491–502.
[15] Orbison JG, McGuire W, Abel JF. Yield surface applications in nonlinear steel frame
analysis. Comput Methods Appl Mech Eng 1982;33:557–73.
[16] Chen WF, Saleeb AF. Uniaxial behavior and modeling in plasticity. West Lafayette:
Purdue University; 1982.
[17] Chen WF, Lui EM. Structural stability: theory and implementation. New York:
Elsevier; 1987.
[18] Hilber HM, Hughes TJR, Taylor RL. Improved numerical dissipation for time integra-
tion algorithms in structural dynamics. Earthq Eng Struct Dyn 1977;5:283–92.
[19] Newmark NM. A method of computation for structural dynamic. J Eng Mech
Div-ASCE 1959;85:67–94.
[20] Chopra AK. Dynamics of structures: theory and applications to earthquake engi-
neering. Upper Saddle River, New Jersey: Pearson Prentice Hall; 2007 07458.
[21] Hughes TJR. The finite element method: linear static and dynamic finite element
analysis. Dover Publications; 2000.
[22] Chan SL, Chui PPT. Nonlinear static and cyclic analysis of steel frames with semi-rigid
connections. Elsevier; 2000.
[23] Stelmack TW, Marley MJ, Gerstle KH. Analysis and Tests of flexibly connected steel
frames. J Struct Eng-ASCE 1986;112:1573–88.
[24] Ostrander JR. An Experimental Investigation of End-plate Connections. Saskatoon,
Saskatchewan, Canada: University of Saskatchewan; 1970.
[25] Chiorean CG. A computer method for nonlinear inelastic analysis of 3D semi-rigid
steel frameworks. Eng Struct 2009;31:3016–33.
[26] Ngo-Huu C, Nguyen PC, Kim SE. Second-order plastic-hinge analysis of space semi-
rigid steel frames. Thin-Walled Struct 2012;60:98–104.
[27] Nguyen PC, Kim SE. An advanced analysis method for three-dimensional steel
frames with semi-rigid connections. Finite Elem Anal Des 2014;80:23–32.
[28] Nguyen PC, Kim SE. Nonlinear elastic dynamic analysis of space steel frames with
semi-rigid connections. J Constr Steel Res 2013;84:72–81.