Finite Elements in Analysis and Design 60 (2012) 3548

Contents lists available at SciVerse ScienceDirect

Finite Elements in Analysis and Design

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

An efcient compound-element for potential progressive collapse analysis of

steel frames with semi-rigid connections
Hamid R. Valipour n, Mark Bradford
School of Civil and Environmental Engineering, The University of New South Wales, Sydney, NSW 2052, Australia

a r t i c l e i n f o abstract

Article history: In this paper, the formulation of a novel 1D frame compound-element for the materially and
Received 26 March 2012 geometrically non-linear analysis of steel frames with exible connections is outlined. The element
Received in revised form is formulated based on the force interpolation concept and the total secant stiffness approach, and
31 May 2012
implemented in a FORTRAN computer code. The accuracy and efciency of the formulation are veried
Accepted 31 May 2012
Available online 26 June 2012
through some numerical examples. For steel frames with bolted ush end-plate and extended end-
plate connections, a static and dynamic progressive collapse assessment based on the alternate load
Keywords: path (ALP) method is undertaken by employing the developed analytical tool and dynamic load factor
Alternate load path (DLF) is estimated. Furthermore, the implications of analyzing semi-rigid steel frames based on the
Force-based method
assumption of xed connections and the effects of the connection details on the global response of a
Non-linear analysis
frame during different progressive collapse scenarios are investigated.
Semi-rigid connection
Steel frame & 2012 Elsevier B.V. All rights reserved.
Total secant formulation

1. Introduction based models offer good versatility and accuracy which are
required for the detailed study of local effects, but they are very
Steel frames are usually designed based on the assumption of time-demanding from a computational point of view for the
either idealized rigid or pinned beam-to-column connections (or analysis of multiple storeymultiple bay frames with large
joints) [1]. It has been well-established that the majority of beam- numbers of degrees of freedom. Such demands on computational
to-column joints, however, do not show such idealized behavior. resources make the continuum-based FE modeling of large
Connections of this type are known exible (semi-rigid connec- structures inefcient and inapplicable. Discrete 1-D frame mod-
tions), and their rigorous inclusion in the analysis and design of els, however, are a good compromise between accuracy and
steel framed structures is fraught with difculty. efciency for predicting the global response of framed structures
Over the last two decades, a large number of studies have been [2,7,8,2125].
devoted to steel and steelconcrete composite frames with semi- Progressive collapse is an important issue in structural failure,
rigid connections [26]. Studies covering various aspects of steel and has been so since the well-reported partial collapse of the
and composite frames with exible connections, such as the Ronan point apartment building in London in 1968. Since this
global behavior of the frame including material and geometrical time, progressive collapse analysis has been the subject of much
non-linearities [710], theoretical and experimental investiga- research endeavor with regard to the global response of members
tions of behavioral models for the connections [1115], the global [2631], however, less attention has been paid to the effect of the
buckling and stability of steel frames [16,17] and analyses of stiffness and strength of the joints and their behavior on the
semi-rigid steel frames under cyclic, seismic and blast loads global response [3135]. One procedure for investigating the
[1820]. potential for progressive collapse is based on the so-called
For the non-linear nite element analysis of steel frames, two alternative path method (APM), which has been integrated into
different classes of model are available, viz. continuum-based several building codes [36,37], and in different forms has been
formulations and discrete 1-D frame elements, which have adopted by researchers for the numerical modeling of reinforced
different domains of applicability [4,11]. Non-linear continuum- concrete and steel frames [22,28,29,38,39]. In the APM approach,
one or more columns are assumed to fail and are removed from
the structural model with the remaining structure analyzed
n
Corresponding author. Tel.: 61 2 9385 6191; fax: 61 2 9385 6139. to determine whether other members (or the structure) will
E-mail address: [email protected] (H.R. Valipour). fail or not.

0168-874X/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.nel.2012.05.009
36 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548

In this paper, the force interpolation concept is employed to  T

rigid body modes) and dx er k is the section generalized
formulate an efcient non-linear compound-element, comprising strain vector.
of a frame element with nodal springs. The geometrical non-
linearity is taken into account by decomposing the element
2.2. Equilibrium equations and constitutive material law for steel
displacements into a rigid body rotation and deformations [40].
The effect of the transverse displacement on the axial strain is
Adopting the small slope assumption (siny tany y w0 ),
taken into account; however, the strains are taken to be small.
the equilibrium equations for the free body of Ax (Fig. 2) produce
The element formulation is implemented in a FORTRAN computer
code and the numerical tool developed is employed for static and NxVxyx Q 1 0, 4
dynamic progressive collapse assessment of steel frames with
bolted ush end-plate and extended end-plate connections and Nxyx Vx Q 2 Q 3 =l 0, 5
the dynamic load factor (DLF) is estimated. Furthermore, the
effects of connection details (i.e. the position of the bolts and the Mx Q 1 wx 1x=lQ 2 x=lQ 3 0, 6
end-plate thickness) on the global response of steel frames with
semi-rigid connections during scenarios of progressive collapse where N(x) and V(x) represent the section normal (axial) and
are investigated. tangential (shear) forces respectively, and y(x) is the section
rotation.
Rearranging Eqs. (4)(6), leads to the matrix representation
2. Element formulation Dx bx,wx, yxQ D x,
n
7

2.1. Compatibility equations where

" #
1 yx=l yx=l
Adopting the NavierBernoulli assumptions, section compat- bx,wx, yx , 8
wx x=l1 x=l
ibility requirements produce h iT
ex er yk, 1 Q Q1 Q2 Q3 denotes the nodal force vector (in the
 T
system without rigid body modes), Dx Nx Mx is the
where ex denotes the total strain at an integration point in the n
section internal force vector and D x is the section internal force
local xx direction (along the element axis; Fig. 1(a)), er is the n
vector due to the element load. It is noteworthy that D x is a
section axial strain, k is the total curvature of section and y is the function of the element rigid body rotation when gravity loads act
distance of the integration point (ber) from the mid-plane of the on the element.
element (Fig. 1(a)). Equilibrium across the section requires that
Fig. 1(a) shows a 2-node plane frame element AB with three hR iT
R
degrees of freedom at each node. Furthermore, at each nodal Dx O sx dA  O ysx dA , 9
point the element is attached to a rotational and a translational
spring, which represent the exural and axial stiffness of the where y is the distance of the integration point from the element
connections at this point respectively (Fig. 1(a)). The generalized mid-plane, sx is the total xx stress component at the monitoring
nodal displacement and force vectors (with rigid body modes points and O represents the cross-sectional domain.
included) are denoted by q and Q, respectively. Using the Decomposing the total strain ex into its elastic eex and inelastic
principle of virtual force and integration by parts for the simply epx components, the total stress and strain in the xx direction can
supported conguration shown in Fig. 1(b), the strain-deforma-
tion compatibility equation for the element (without nodal (x) = w'(x) = dw / dx
springs) is obtained as V (x)
y
Z l M (x) N (x)
T
q0 b x,wxdxdx, 2
0 Q2 Q3
Q1 w (x)
where x
" #
1 0 0
bx,wx , 3
wx=2 x=l1 x=l A x B
h iT l
0 00
q0 q1 q1 q02 q03 is the generalized nodal deformation
Fig. 2. Equilibrium in the simply supported conguration and free body diagram
vector of the frame element excluding the nodal springs (without of Ax, after deformation (system without rigid body modes).

y Q1 , q1
x
y
Rotational Q3 , q3
k1 Spring k2 Q , q A B
Q1 , q1 x 4 4
Q2 , q2 l0
A k1 Translational k2 B
Q3 , q3 Spring q2' q3"
q1 q2 q1' q1" q3

z Q2, q2 Q6 , q6
Q5 , q5 y
A x B
l

Fig. 1. (a) 2-node frame element AB in xy plane and (b) outline of the simply supported conguration (system without rigid body modes).
 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548 37

 (Fig. 1(b)). Several experimental and numerical test results on

e connection behavior have been reported in the literature
[3,42,43]. Accordingly, various analytical expressions in the form
of a piecewise linear, polynomial, exponential and B-spline func-
 tions have been proposed to represent the behavior of connec-
tions [14,44,45].

g
In the present formulation, the implementation of the equiva-

U n d in g
d in
lent spring model is easier if the generalized displacement of a

lo a
a
Lo
 spring q is expressed explicitly in terms of generalized force f. The
Ee= simplest option for such a case could be a bi-linear model which
( - p)
follows a kinematic hardening law (Fig. 4(a)). Alternatively, the
p   RambergOsgood model which was originally proposed by Ram-
berg and Osgood [46] and then standardised by Ang and Morris
Fig. 3. A typical uniaxial stressstrain relationship in a total secant framework.
[47] can be adopted, and for the unloading/reloading branch
Masings rule can be employed (Fig. 4(b)).
be related by Following the total secant concept and decomposing the total
sx Ee ex epx , 10 generalized displacement q to elastic qe and plastic qp compo-
nents, the generalized forcedisplacement relationship for the
where Ee is the elastic secant modulus of the theoretical unload- non-linear nodal springs can be expressed by
ing curve and is a function of the stress and strain components at
the integration point under consideration (Fig. 3). In this study, f kqqp , 19
the secant modulus is taken to be history independent.
where f is the generalized force and k is the elastic secant
Substituting Eqs. (1) and (10) into Eq. (9) gives
modulus of the unloading curve (see Fig. 4).
Dx ks xdx Dp x, 11 With regard to Fig. 1(b) and Eq. (19), the generalized load
displacement relationships for the nodal springs can be con-
where
densed into matrix form as
" R R #
Ee dA  O yEe dA
ks x RO R 2 12 qq0 Fsp Q qpr , 20
 O yEe dA O y Ee dA
h iT h iT
is the secant stiffness matrix of the section and where q q1 q2 q3 and Q Q 1 Q2 Q3 are the gen-
h R R iT eralized nodal deformation and force vector of the frame element
Dp x  O Ee epx dA O yEe epx dA 13 including nodal springs (without rigid body modes) respectively,
h iT
is the residual plastic force vector for the section. qpr qp1 qp2 qpy1 qpy2 is the plastic deformation vector of
The exibility matrix of the section fs(x) is obtained by the nodal spring system and Fsp represents the exibility matrix
inverting the section stiffness matrix, and then Eq. (11) is rear-
ranged as
f (M )
dx f s xfDxDp xg: 14
Using Eqs. (7) and (14), the section generalized strain vector fy
d(x) is related to Q , and then substituting the results into Eq. (2) k (k )
Rel ading

gives the relationship

ing

1
oa d
o
Unl

q0 FQ qp qn , 15
Rel ading
ing

qp q ( q )
oad

where
o
Unl

Z l
T
F b x,wxf s xbx,wx, yxdx 16
0 fy
is the exibility matrix of the simply supported conguration
(without rigid body modes), which is not typically symmetric
[41], f (M ) n1
Z l n2
T
qp  b x,wxf s xDp xdx 17 fy n3
0
Masing's rule
q -- qa f-- fa f-- fa n-1 n 1 < n 2< n 3
g

2 qy ( 2 fy )[ ( 2 fy ) ]
1+
din

=
is the nodal generalized plastic deformation vector excluding the
loa

Ramberg-Osgood model
Re

nodal springs, and q f f n-1

Z l qy = ( fy ) [1 + ( fy ) ]
T n
qn b x,wxf s xD xdx 18 qp qy 2qy q ( q )
g
din

0
loa
Un

is the nodal generalized deformation vector due to the

element loads.
Load reversal
2.3. Constitutive law for semi-rigid connections (qa , fa)

The rotational/horizontal stiffness of the connections can be Fig. 4. Model of semi-rigid connections: (a) bilinear with kinematic hardening and
modeled by assigning equivalent springs to the beam nodal points (b) RambergOsgood with Masings rule for unloading/reloading branch.
38 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548

of the nodal springs system given by and T* is the transformation matrix represented by
2 3 2 3
1
k
k12 0 0 t 11 t 12 0 t 11 t 12 0
6 1 7 6 7
6 1
07 Tn 4 t 21 t 22 1 t 21 t 22 0 5, 26
Fsp 6 0 ky1 7, 21
4 5 t 21 t 22 0 t 21 t 22 1
0 0 k1
y2
with
0
Substituting q from Eq. (15) into Eq. (20) then produces   (   )
1 q5 q2 2 3 q5 q2 2
q F Fsp Q qp qpr qn : 22 t 11 1 1 , 27a
2 l0 4 l0

 (   )
1 q5 q2 1 q5 q2 2
t 12 1 , 27b
2 l0 4 l0
3. Displacement interpolation along the element axis
 (   )
In displacement-based formulations, the deformed shape of 1 q5 q2 q q 2
t 21 1 5 2 27c
the element is obtained directly based on the nodal displacement l0 l0 l0
values and the adopted shape functions. In force-based formula-
and
tions however, more elaborate techniques are required [40,48]. In (   )
this study a composite Simpson integration method together with 1 1 q5 q2 2
t 22 1 : 27d
piecewise parabolic interpolation of the curvature are used to l0 3 l0
update the element geometry [49].
If the stiffness matrix of the simply supported conguration
including the nodal springs is denoted by K fF Fsp g1 , then Eq.
4. Rigid body motion and corresponding transformation (22) can be rewritten as
Q Kfqqp qpr qn g 28
The exibility formulation presented in the previous section
was derived in the element reference system without rigid body Multiplying Eq. (28) by TT and then substituting forq from Eq.
motion. Thus, a transformation is required to relate the deforma- (25) results in
tion and corresponding force vectors in the system without rigid
body modes to the system with rigid body modes. fTT KTn gq Q TT Kqp qpr Q n TT Kqn , 29
If the length of the nodal springs is taken to be negligible when and the dynamic form of this equation is
compared to that of the member, the transformation between the
force vectors of the systems with and without rigid body modes Mq Cq_ fTT KTn gq Q TT Kqp qpr Q n TT Kqn , 30
can be established [49] as where M and C are the mass and damping matrices of the element
Q TT Q , 23 and q_ and q are the velocity and acceleration vectors of the
 T element, respectively.
where Q Q 1 Q 2 Q 3 Q 4 Q 5 Q 6 denotes the nodal force
The recursive form of Eqs. (29) and (30) after time discretiza-
vector in the system with rigid body modes and T is the
tion (if required) and assembling at the structure level,
transformation matrix given by X T n X X T
2 3 fT KT gqi Qi fT Kqp qpr gi1
cosy siny 0 cosy siny 0 X n
6 siny cosy
1 siny 7
 cosl y 0 7, Q TT Kqn i1 , 31
T6 4
 l l l 5 24
 sinl y cosl y 0 siny
l
 cosy
l
1 can be solved by a nested direct iterative solution scheme [40]. In
1
the adopted recursive form, the right subscript i denotes the
where y tan q5 q2 =l0 q4 q1  is the rigid body rotation of iteration number.
the element.
With regard to Fig. 5, the geometric compatibility equation
q Tn q 25 5. Verication of formulation and analytical tool
h iT
is obtained [49], where q q1 q2 q3 q4 q5 q6 is the In the rst part of the verication, a 6-storey frame with the
nodal displacement vector in the system with rigid body modes geometry, section properties and loading shown in Fig. 6 is analyzed.
The steel behavior is assumed to be linear elasticperfectly plastic-
hardening, with a strain hardening modulus Esh 1 GPa, yield stress
q3 fy 300 MPa and elastic modulus Es 200 GPa. Further, the connec-
l B' tion behavior is linear elastic with a rotational stiffness of
te ky 3000 kN m/rad . The entire frame is modeled with just 41
l Sta
q2 ent Fina exibility-based frame elements, one for each member of the frame.
Elem
A composite Simpsons scheme with 17 integration points through
q5
the beam and 21 integration points through the column section depth
A'
q3 q6 (3 points over the thickness of each ange) and 21 points along the
q2
q1 q4 beam and 11 points along each column is used.
Element Initial State
The load versus vertical displacement of point A (see Fig. 6)
A B and the axial force at end C of span CD in the rst oor obtained
from the formulation developed in this paper are given in Fig. 7.
l0
The results obtained from the ABAQUS software using displace-
Fig. 5. Outline of the element and displacement (deformation) components at ment-based elements (B33 element) for three different meshes
initial and nal states. comprising of 59, 165 and 330 elements are also shown in Fig. 7.
 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548 39

20 mm
150 mm
B B C D
5.0 m 6.0 m 5.0 m
w w w

240 mm
15 mm

20 mm
w w w

Beam Section

6 @ 3.0 m = 18.0 m
20 mm
200 mm w w w

Rotational Spring Rotational Spring

k = 3000 kNm/rad k = 3000 kNm/rad
320 mm

w w w
20 mm
Point A
20 mm

Removed
Column

Column Section
Fig. 6. Outline of the 6-storey frame, loading and geometry of the members.

Fig. 7. Level of load w versus (a) vertical displacement at point A and (b) axial force at end C of the span CD in the rst oor.

A B C D E
5.0 m 5.0 m 5.0 m 5.0 m
w w w w

w w w w
[email protected] m = 15.0 m

w w w

Rotational Spring Rotational Spring

My = 100 kNm My = 100 kNm
w w w w

Point B
Columns: 250UC72.9
Removed
Beams: 250UB31.4 Column

Fig. 8. Outline of the 5-storey frame, loading and geometry of the members.

It can be observed that the ultimate loading capacity of the frame obtained from the most rened ABAQUS model of wu 33.33 kN/m
predicted by the present formulation using 1 element per beam or using 330 elements. Moreover, it is observed that the member
column is wu 33.31 kN/m which is comparable with the results bowing action has a minor effect on the displacement response
40 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548

compared with its effect on the axial force of the beam (the bp
formulation without bowing effect leads to zero axial force in the g tp
beam). With the same level of accuracy, the total time for the analysis tf
with the exibility-based model was about one-fth of that with the pf
ABAQUS model with 330 elements, which is demonstrative of the End plate
efciency of the formulation. h tw
In the second part of the verication, a 5-storey frame with the
geometry, section properties and loading shown in Fig. 8 is analyzed.
The steel behavior is assumed to be linear elasticperfectly plastic- pf
tf
hardening, with a strain hardening modulus Esh 1 GPa, yield stress
fy 320 MPa and elastic modulus Es 200 GPa. Furthermore, the Ang db : bolt diameter
and Morris model [47] is adopted to model the behavior of the semi-
Fig. 10. Conguration of a typical ush end-plate connection.
rigid connections, with a yielding moment of My 100 kN m and
corresponding rotation of yy 2 mrad and the shape parameter is
taken n5. The entire frame is modeled with just 44 exibility-based [50] using the empirical relationships
frame elements, one for each member of the frame. A Simpsons 1:13 0:448 0:139 0:095 0:117 0:134 1:136 0:296
My 1:66g 0:033 db pf bp tp tf tw h fy 32
scheme with 15 integration points through the beam and 19
integration points through the column section depth (3 points over
thickness of each ange) and 19 points along the beam and 11 points yy 0:0019g 0:555 d0:231
b p2:938
f bp
0:499 0:563 0:08 0:485 1:102 0:062
tp tf tw h fy
along each column is used. 33
The load versus vertical displacement of point B (Fig. 8)
obtained from the formulation developed in this paper is given n 1:217
0:974 0:946 1:009 0:478 0:287 0:451 0:011 1:363
in Fig. 9. The results obtained from the ABAQUS software using 107 g 0:077 db pf bp tp tf tw h fy ,
displacement-based elements (B33) for the mesh including 128
34
elements are also shown in Fig. 9. It is observed that the ultimate
loading capacity of the frame predicted by the present formula- where fy is the yield stress of the material (in psi) and the
tion using 1 element per beam or column is wu 55.2 kN/m which denitions of other geometry-related parameters are given in
is comparable with the results obtained from the ABAQUS model Fig. 10 (all of the dimensions are in inches).
of wu 55.8 kN/m using 128 elements. It is noteworthy that a regression analysis of the results
obtained from continuum-based FE models have been used by
Abolmaali et al. [50] to develop these relationships. Furthermore,
Abolmaali et al. [50] have veried the accuracy of their FE models
with test results conducted and reported for ush end-plate
6. Applications for progressive collapse analysis connections.

Using the ALP method, the developed analytical tool can be

6.1. Pushdown test on a beamcolumn subassembly with ush end-
readily employed for the potential progressive collapse assess-
plate connections
ment of steel frames with ush end-plate and extended end-plate
connections. For the ush end-plate connection with high-
In order to show the accuracy of the adopted momentrotation
strength A-325 bolts (fy 570 MPa and fu 640 MPa), the char-
relationship (Fig. 4 and Eqs. (32)(34)) and to demonstrate the
acteristic moment My, characteristic rotation yy and rigidity
efciency of the developed compound frame element compared
parameter n required to construct the momentrotation curve
with more computationally-expensive continuum-based nite
based on a standardised Ramberg and Osgood model (Fig. 4) are
element models, a beamcolumn subassembly with ush end-
obtained from the relationships developed by Abolmaali et al.
plate connections is analyzed in this part. The geometry of the
subassembly and details of the ush end-plate connections are
shown in Fig. 11a. The steel yield strength is taken as fy 350 MPa
and elastic modulus as Es 200 GPa. A bilinear elasticplastic
hardening constitutive law with a hardening modulus of Esh
2 GPa is adopted for steel, whose ultimate tensile strength is
limited to fu 490 MPa.
A continuum-based nite element model was generated in
ANSYS using element SOLID185and the relative size of the
elements was limited to 8 mm. A plasticity-based constitutive
law based on von Mises yield criterion with isotropic hardening
was employed to capture the non-linear response of the materi-
als, and the interaction between the end-plate and column ange
was modeled with the node-to-node non-linear contact/target
elements of ANSYS. The bolts were treated as imaginary parts
within the end-plate and the column ange and the bolt preten-
sion element available in ANSYS was used to connect these two
imaginary parts.
Using the compound-element developed in this study, the
subassembly was modeled with 3 elements: one for each member
(i.e. beam and half of a column) of the subassembly. The nite
Fig. 9. Level of load w versus vertical displacement at point B on top of the element model of the subassembly within the formulation has
removed column. 6 DOFs and the characteristic parameters required to establish the
 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548 41

5.25 m 5.25 m
Pushdown

250UC72.9
250UC72.9
3.25 m 250UB25.7 250UB25.7

8 mm
d b = 20 mm (bolt diameter)
80 10 mm
60
Flush end
248 mm

250UB25.7 plate
Web
stiffener
60

Connection details

5.25 m
Element-2

Rotational Rotational
P/2
Spring Spring
3.25 m

Element-3
Element-1

M y = 55.7 kN.m
Characteristic parameters for
moment-rotation relationship y = 0.00325

n = 8.3
Fig. 11. (a) Geometry of the subassembly and details of ush end-plate connections and (b) idealized FE model used in conjunction with compound frame element
formulation.

RambergOsgood momentrotation relationship for the rota-

tional springs are obtained from Eqs. (32) to (34) and given in
Fig. 11b. The integrals are estimated by a composite Simpsons
integration scheme with 15 integration points through the sec-
tion depth (3 points over the thickness of each ange) and the
distance between the longitudinal integration points is limited to
150 mm. The load versus mid-span displacement of the beam for
the pushdown test is shown in Fig. 12. It is observed that the
results obtained from the developed compound-element based on
exible behavior of the connections correlates well with the
results of the more expensive continuum-based nite element
model whereas assumption of rigid behavior for the connections
leads to a dramatic overestimation of both the stiffness and
strength.
In addition, the response obtained from the ANSYS model
assuming a rigid (non-deformable) panel zone is shown in Fig. 12,
which is in close agreement with the other results based on a
Fig. 12. Load P versus mid-span displacement for the subassembly.
deformable panel zone and demonstrates the negligible effect of
the panel zone deformation on the global response in this
example. The deformed shape and contours of the equivalent Dual Core computer with a 2.8 GHz processor and running Windows
plastic strain at deection of 486 mm (P65 kN) are shown in XP. This demonstrates the superior efciency of the formulation.
Fig. 13. It is observable that yielding of the material and the
subsequent plastic deformation are limited to the end-plate and 6.2. Progressive collapse assessment of multi-storey frames (ush-
no plastic hinge has formed within the beam (Fig. 13b), which is end plate connections)
in agreement with the compound frame model using semi-rigid
connections. In the frame model with rigid connections, however, In this example, the potential progressive collapse of a three
a plastic hinge forms at such a deection level. and a seven-storey steel frame for two different scenarios based
For the compound frame model, the analysis time needed to on the removal of a corner or a middle column on the ground oor
construct the equilibrium path with 15 steps was 2.1 s on an Intel is studied. The geometry of the frame and details of the ush
42 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548

1.505
Local
yielding of
end-plate

Z MX
X Y

MN

MX

-485.588 -377.345 -269.102 -160.86 -52.617 0 .042972 .085945 .128917 MN .17189

-431.467 -323.224 -214.981 -106.738 1. .021486 .064459 .107431 .150404 .193

Fig. 13. (a) Deformed shape of the subassembly and (b) contour of equivalent plastic strain at load level P 65 kN.

[email protected] m = 20.0 m d b = 24 mm (bolt diameter)

70

60
248 mm
n @ 3.0 m

60
n : number All beams: All columns: 18 mm
of storeys 250UB25.7 250UC72.9
Flush end
A Beam-1 B Beam-2 C Beam-3 plate

Removed in Removed in Removed in

250UB25.7
scenario-1 scenario-2 scenario-3

Connection details
Fig. 14. (a) Geometry of frame and (b) details of ush end-plate connections.

end-plate connections are shown in Fig. 14. The steel yield and the nodal forces on the top of the column to be removed were
strength is fy 350 MPa (Steel Grade350) and the elastic modulus obtained. Next, the column was removed and the nodal forces at
is taken Es 200 GPa. A bilinear elasticplastic hardening consti- the top of the column imposed on the top joints of the removed
tutive law with a hardening modulus of Esh 2 GPa is adopted for column with opposite sign. In the ensuing analysis, these nodal
the steel and its ultimate tensile strength is limited to forces were removed within 0.01 s and the dynamic response of
fu 490 MPa. The characteristic parameters obtained from Eqs. the structure was obtained.
(32) to (34) to establish the momentrotation relationship are For the dynamic analysis, the Newmark scheme with variable
My 72 kN m, yy 0.0022 rad and n7.7. time steps was used (maximum time step of 2 ms) and Rayleigh
All of the frames are subjected to a permanent action (dead damping, with a mass (a) and initial stiffness multiplier (b) of
load) of G (DL) 12 kN/m (including the self-weight of structure), a b 0.001, were adopted. The time history of the vertical
an imposed action (live load) of Q (LL)16 kN/m and a minimum displacements for points A and B within different scenarios are
uniform lateral wind load of W2 kN/m over the frame height. shown in Figs. 15 and 16 and the time history of the bending
All of the members have been designed according to the moment and axial force at the right end of Beam-1 and Beam-2
Australian Standard AS4100-1998 [51] with lateraltorsional corresponding with scenario-1 and scenario-2 are shown in
buckling being prevented, and the bolts were designed for Figs. 1720, respectively.
combinations of tension and shear assuming that their threads It is observable that the assumption of rigid behavior for the
do not intercept the shear planes. It is noteworthy that the ush end plate connections can lead to an underestimation of the
connections in this part have been designed for bending moments maximum and permanent deection as well as the axial force in
and shear forces corresponding to the full plastic bending the beams up to 10%, and consequently underestimation of the
moment capacity of the beams (Mp 112 kN m) which is higher potential for progressive collapse of the frame. With regard to the
than the maximum bending moment in the most critical load bending moment, however, both the assumptions of rigid or
combination (Mmax 80 kN m). exible behavior for the ush end plate connections more or less
With regard to the GSA provisions [36], the load combination have led to the same results (viz. Figs. 17 and 18).
considered for potential progressive collapse scenarios is Comparing the static and dynamic responses of the structure
G 0.25Q (DL 0.25LL). First the structure was analyzed statically demonstrates the importance of the dynamic analysis for the
 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548 43

Time (sec) Time (sec)

0 4 8 12 16 20 0 4 8 12 16 20
0 0

Vertical displacement at B (mm)

Vertical displacement at A (mm)
-100 -100
-200 Rigid (linear static, DL+0.25LL) Rigid (linear static, DL+0.25LL)
Rigid (nonlinear static, DL+0.25LL) -200 Rigid (nonlinear static, DL+0.25LL)
-300 Rigid (nonlinear dynamic, DL+0.25LL) Rigid (nonlinear dynamic, DL+0.25LL)
Flexible (linear static, DL+0.25LL) Flexible (linear static, DL+0.25LL)
Flexible (nonlinear static, DL+0.25LL) -300
Flexible (nonlinear static, DL+0.25LL)
-400 Flexible (nonlinear dynamic, DL+0.25LL)
Flexible (nonlinear dynamic, DL+0.25LL)

-500 -400

-600 -500

Fig. 15. History of vertical displacement for three-storey frames with ush end-plate connections: (a) scenario-1 and (b) scenario-2.

Time (sec) Time (sec)

0 4 8 12 16 20 0 4 8 12 16 20
0 0

Vertical displacement at B (mm)

Vertical displacement at A (mm)

-100
-100
-200 Rigid (linear static, DL+0.25LL)
Rigid (linear static, DL+0.25LL)
Rigid (nonlinear static, DL+0.25LL) -200
Rigid (nonlinear static, DL+0.25LL)
-300 Rigid (nonlinear dynamic, DL+0.25LL) Rigid (nonlinear dynamic, DL+0.25LL)
Flexible (linear static, DL+0.25LL) Flexible (linear static, DL+0.25LL)
Flexible (nonlinear static, DL+0.25LL) -300
-400 Flexible (nonlinear static, DL+0.25LL)
Flexible (nonlinear dynamic, DL+0.25LL) Flexible (nonlinear dynamic, DL+0.25LL)

-500 -400

-600 -500

Fig. 16. History of vertical displacement for seven-storey frames with ush end-plate connections: (a) scenario-1 and (b) scenario-2.

Time (sec) Time (sec)

0 4 8 12 16 20 0 4 8 12 16 20
0 0
Bending moment (kN.m)
Bending moment (kN.m)

-25 -25
Rigid (linear static, DL+0.25LL)
Rigid (linear static, DL+0.25LL) Rigid (nonlinear static, DL+0.25LL)
-50 Rigid (nonlinear static, DL+0.25LL) -50 Rigid (nonlinear dynamic, DL+0.25LL)
Rigid (nonlinear dynamic, DL+0.25LL) Flexible (linear static, DL+0.25LL)
Flexible (linear static, DL+0.25LL) Flexible (nonlinear static, DL+0.25LL)
-75 Flexible (nonlinear static, DL+0.25LL) -75 Flexible (nonlinear dynamic, DL+0.25LL)
Flexible (nonlinear dynamic, DL+0.25LL)
-100 -100

-125 -125

-150 -150

Fig. 17. History of bending moment at right end of Beam-1 and -2 for three-storey frames with ush end-plate connections: (a) scenario-1 and (b) scenario-2.

Time (sec) Time (sec)

0 4 8 12 16 20 0 4 8 12 16 20
0 0
Rigid (linear static, DL+0.25LL) Rigid (linear static, DL+0.25LL)
Bending moment (kN.m)

Rigid (nonlinear static, DL+0.25LL)

Bending moment (kN.m)

Rigid (nonlinear static, DL+0.25LL)

-25 Rigid (nonlinear dynamic, DL+0.25LL) -25 Rigid (nonlinear dynamic, DL+0.25LL)
Flexible (linear static, DL+0.25LL) Flexible (linear static, DL+0.25LL)
-50 Flexible (nonlinear static, DL+0.25LL) -50 Flexible (nonlinear static, DL+0.25LL)
Flexible (nonlinear dynamic, DL+0.25LL) Flexible (nonlinear dynamic, DL+0.25LL)

-75 -75

-100 -100

-125 -125

-150 -150

Fig. 18. History of bending moment at right end of Beam-1 and -2 for seven-storey frames with ush end-plate connections: (a) scenario-1 and (b) scenario-2.
44 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548

70 100

Axial compressive force (kN)

60

Axial tensile force (kN)

80
50 Rigid (linear static, DL+0.25LL)
Rigid (nonlinear static, DL+0.25LL)
40 60 Rigid (nonlinear dynamic, DL+0.25LL)
Flexible (linear static, DL+0.25LL)
30 Flexible (nonlinear static, DL+0.25LL)
Rigid (linear static, DL+0.25LL) 40 Flexible (nonlinear dynamic, DL+0.25LL)
Rigid (nonlinear static, DL+0.25LL)
20 Rigid (nonlinear dynamic, DL+0.25LL)
Flexible (linear static, DL+0.25LL) 20
10 Flexible (nonlinear static, DL+0.25LL)
Flexible (nonlinear dynamic, DL+0.25LL)
0 0
0 4 8 12 16 20 0 4 8 12 16 20
Time (sec) Time (sec)

Fig. 19. History of axial force at right end of Beam-1 and Beam-2 for three-storey frames with ush end-plate connections: (a) scenario-1 and (b) scenario-2.

70 100
Axial compressive force (kN)

60

Axial tensile force (kN)

80
50 Rigid (linear static, DL+0.25LL)
Rigid (nonlinear static, DL+0.25LL)
40 60 Rigid (nonlinear dynamic, DL+0.25LL)
Flexible (linear static, DL+0.25LL)
30 Flexible (nonlinear static, DL+0.25LL)
40 Flexible (nonlinear dynamic, DL+0.25LL)
Rigid (linear static, DL+0.25LL)
20 Rigid (nonlinear static, DL+0.25LL)
Rigid (nonlinear dynamic, DL+0.25LL) 20
10 Flexible (linear static, DL+0.25LL)
Flexible (nonlinear static, DL+0.25LL)
Flexible (nonlinear dynamic, DL+0.25LL)
0 0
0 4 8 12 16 20 0 4 8 12 16 20
Time (sec) Time (sec)

Fig. 20. History of axial force at right end of Beam-1 and Beam-2 for seven-storey frames with ush end-plate connections: (a) scenario-1 and (b) scenario-2.

progressive collapse assessment with the dynamic load factors Table 1

(DLF) varying between 1.79 and 1.88 for the models with exible Maximum displacement obtained from different types of analysis and dynamic
load factor (DLF) within different scenarios (ush end plates assumed to be
connection behavior which is consistent with the value of DLF2
exible).
proposed by some standards based on linear analysis [35]. The
dynamic load factors (DLF) for models using rigid connection Maximum displacement from different DLF for Non-
behavior, however, vary between 2.85 and 3.31, which is larger analyses (mm) linear analysis
than the value of DLF 2 proposed by some standards based on
Linear static Non-linear analysis 3- 7-
linear analysis results [35]. The maximum displacements obtained analysis storey storey
from different types of analysis and the dynamic load factors (DLF)
within different scenarios based on assumptions of exible and rigid 3- 7- Static Dynamic
connection behavior are reported in Tables 1 and 2, respectively. storey storey
3- 7- 3- 7-
With regard to Tables 1 and 2, it can be concluded the rigidity
storey storey storey storey
of the ush end plate connection and the adopted assumption for
modeling connection behavior has a major effect on the DLF and Scenario- 110.5 114.1 308.2 309.7 550.5 554.3 1.79 1.79
assumption of xity could lead to overestimation of DLF. 1
In the second part of this example, assuming that the connec- Scenario- 78.5 80.6 261.9 265.3 491.6 499.7 1.88 1.88
2
tions are designed for the maximum bending moment with the
most critical load combination (Mmax 80 kN m), the thickness of
the end plate and the diameter of the bolts are reduced to moment (critical load combination) improve the performance of
tp 14 mm and db 20 mm and the characteristic parameters for the frame during progressive collapse scenarios by reducing the
the connection are calculated as My 57 kN m, yy 0.0025 rad and maximum displacement responses.
n 7.2. Accordingly, in Fig. 21 the non-linear response of the
3-storey frame with connections designed for the maximum
bending moment (connection-m) is compared with a frame with 6.3. Progressive collapse assessment of 3-storey frame (extended end
its connections designed for the plastic bending moment plate connections)
(connection-p). It is observed that the assumption of full xity
for ush end plates designed for the maximum bending moment In this example, the potential progressive collapse of a three-
can lead to an underestimation of the maximum and permanent storey frame (with extended end-plate connections) for two
deection, as well as axial force in the beams, of up to 18%. different scenarios based on the removal of a corner or a middle
Moreover, it is concluded that designing the ush end plates for column on the ground oor is studied. The geometry of the frame
the plastic bending moment of beams instead of the maximum and member sizes are given in Fig. 14a and details of the extended
 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548 45

end-plate connections (designed for plastic bending moment constitutive law based on von Mises yield criterion with isotropic
capacity of the beam) are shown in Fig. 22. hardening rule. The momentrotation of the connection obtained
The material properties and loading types (i.e. dead, live and from the ANSYS model is shown in Fig. 23.
wind) and magnitudes are as given in previous example (Section
6.2). All members and connections have been designed according 150 22 mm
to the Australian Standard AS4100-1998 [51] with lateral 70
torsional buckling being prevented, and the bolts were designed

35

35
for combinations of tension and shear assuming that their threads

40

40
do not intercept the shear planes. The load combination G 0.25Q
Extended end
(DL 0.25LL) and analysis procedure for potential progressive

400 mm
248 mm
plate
collapse scenarios are as proposed by GSA [36] and explained in db = 16 mm (bolt diameter)
previous example (Section 6.2).
250UB25.7
The momentrotation behavior of extended end-plate connection

40

40
was captured using a continuum-based nite element model gener-

35

35
ated in ANSYS. The connection was modeled by element SOLID185
with the maximum element size of 8 mm and plasticity-based Cross-section Side-view elevation

Fig. 22. Geometry and details of extended end-plate connections.

Table 2
Maximum displacement obtained from different types of analysis and dynamic 150
load factor (DLF) within different scenarios (connections assumed to be rigid).

Bending moment (kN.m)

120
Maximum displacement from different DLF for non-
analyses (mm) linear analysis
90
Linear static Non-linear analysis 3- 7-
analysis storey storey
60
3- 7- Static Dynamic
storey storey
3- 7- 3- 7-
30 ANSYS model
storey storey storey storey
Ramberg-Osgood
0
Scenario- 90.3 93.3 174.2 174.5 496.8 500.1 2.85 2.87 0 10 20 30 40 50 60
1
Scenario- 63.8 66.0 135.2 134.8 445.4 446.0 3.29 3.31 x 10-3 (Rad)
2
Fig. 23. The moment versus rotation of extended end-plate connection.

Time (sec) Time (sec)

0 4 8 12 16 20 0 4 8 12 16 20
0 0
Vertical displacement at A (mm)

Vertical displacement at B (mm)

-100 -100

-200 -200
Rigid (Dynamic) Rigid (Dynamic)
-300 Flexible (Dynamic, connection-m) -300 Flexible (Dynamic, connection-m)

Flexible (Dynamic, connection-p) Flexible (Dynamic, connection-p)

-400 -400

-500 -500

-600 -600

70 120
Axial compressive force (kN)

60
Axial tensile force (kN)

100
50
80
40 Rigid (Dynamic)
60
30 Flexible (Dynamic, connection-m)

Rigid (Dynamic) 40 Flexible (Dynamic, connection-p)

20
Flexible (Dynamic, connection-m)
10 20
Flexible (Dynamic, connection-p)
0 0
0 4 8 12 16 20 0 4 8 12 16 20
Time (sec) Time (sec)

Fig. 21. History of response for three-storey frames: (a) vertical displacement at point A within scenario-1, (b) vertical displacement at point B within scenario-2, (c) axial
force at the right end of Beam-1 within scenario-1 and (d) axial force at the right end of Beam-2 within scenario-2 (frame with ush end plate connection).
46 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548

For the extended end-plate connection the characteristic plate connections more or less have led to the same results (viz.
moment My, characteristic rotation yy and rigidity parameter n Fig. 25). The maximum displacements obtained from different
required to construct the momentrotation curve based on a types of analysis and the dynamic load factors (DLF) within
standardised Ramberg and Osgood model (Fig. 4) can be obtained different scenarios for the three-story frame with extended end
from the physically-based mathematical model developed by Lee plate are reported in Table 3 and the DLF varying between 1.75
and Melchers [52]. It is noteworthy that Lee and Melcherss [52] and 2.02 for the models with exible extended end plate connec-
model can take account of end-plate and column ange deforma- tions which is consistent with the value of DLF2 proposed by
tion, shear deformation and post-buckling stiffness of panel zone some standards [35]. However, the assumption of rigid connec-
and elongation of bolts as well as their associated failure modes. tion leads to overestimation of DLF as shown in Table 1 and
The characteristic parameters obtained from the equations devel- discussed in previous example.
oped by Lee and Melchers [52] to establish the momentrotation With regard to Tables 2 and 3, it can be concluded the rigidity
relationship are My 87 kN m, yy 0.0025 rad and n 6.9, and the of the extended end-plate connection and the adopted assump-
corresponding RambergOsgood momentrotation is shown in tion for modeling connection behavior has a major effect on
Fig. 23 which shows reasonably good correlation with moment the DLF and assumption of xity could lead to overestimation
rotation predicted by ANSYS nite element model. Furthermore, of DLF.
the ultimate loading capacity of the connection captured by
the ANSYS nite element model is associated with bolt yield-
ing which is consistent with Lee and Melcherss [52] model 7. Conclusions
prediction.
The time history of the vertical displacements for points A and A formulation for a novel exibility-based frame element for
B within scenario-1 and -2 are shown in Fig. 24 and the time the non-linear analysis of steel frames with semi-rigid connec-
history of the bending moment and axial force at the right end of tions in the framework of a total secant stiffness matrix has been
Beam-1 and Beam-2 corresponding with scenario-1 and scenario- derived. The formulation takes account of physical and geome-
2 are shown in Figs. 25 and 26, respectively. It is observable that trical non-linearities as well as non-linearity of the frame con-
the assumption of rigid behavior for the extended end plate nections. The formulation has the capability to capture the
connections can lead to an underestimation of the maximum bowing action and large displacements of the element accurately,
and permanent deection as well as the axial force in the beams that plays a signicant role within the context of extreme loading
(up to 9% overestimation in the case considered), and conse- scenarios such as progressive collapse analysis. Furthermore, the
quently underestimation of the potential for progressive collapse proposed formulation and solution scheme offers a better stabi-
of the frame. With regard to the bending moment, however, both lity compared with existing NewtonRaphson family of solution
the assumptions of rigid or exible behavior for the extended end schemes.

Time (sec) Time (sec)

0 4 8 12 16 0 4 8 12 16
0 0
Vertical displacement at A (mm)

Vertical displacement at B (mm)

-100 -100
Rigid (linear static, DL+0.25LL)
-200 Rigid (linear static, DL+0.25LL) -200 Rigid (nonlinear static, DL+0.25LL)
Rigid (nonlinear static, DL+0.25LL)
Rigid (nonlinear dynamic, DL+0.25LL) Rigid (nonlinear dynamic, DL+0.25LL)
-300 -300
Flexible (linear static, DL+0.25LL) Flexible (linear static, DL+0.25LL)
Flexible (nonlinear static, DL+0.25LL) Flexible (nonlinear static, DL+0.25LL)
-400 Flexible (nonlinear dynamic, DL+0.25LL) -400 Flexible (nonlinear dynamic, DL+0.25LL)

-500 -500

-600 -600

Fig. 24. History of vertical displacement for three-storey frames with extended end-plate: (a) scenario-1 and (b) scenario-2.

Time (sec) Time (sec)

0 4 8 12 16 0 4 8 12 16
0 0
Bending moment (kN.m)

Bending moment (kN.m)

-25 -25
Rigid (linear static, DL+0.25LL)
Rigid (linear static, DL+0.25LL) Rigid (nonlinear static, DL+0.25LL)
-50 Rigid (nonlinear static, DL+0.25LL) -50 Rigid (nonlinear dynamic, DL+0.25LL)
Rigid (nonlinear dynamic, DL+0.25LL) Flexible (linear static, DL+0.25LL)
Flexible (linear static, DL+0.25LL) Flexible (nonlinear static, DL+0.25LL)
-75 Flexible (nonlinear static, DL+0.25LL) -75 Flexible (nonlinear dynamic, DL+0.25LL)
Flexible (nonlinear dynamic, DL+0.25LL)
-100 -100

-125 -125

-150 -150

Fig. 25. History of bending moment at right end of Beam-1 and -2 for three-storey frames with extended end-plate: (a) scenario-1 and (b) scenario-2.
 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548 47

80 100

Axial compressive force (kN)

70

Axial tensile force (kN)

80
60
Rigid (linear static, DL+0.25LL)
50 Rigid (nonlinear static, DL+0.25LL)
60 Rigid (nonlinear dynamic, DL+0.25LL)
40 Flexible (linear static, DL+0.25LL)
Flexible (nonlinear static, DL+0.25LL)
30 Rigid (linear static, DL+0.25LL) 40 Flexible (nonlinear dynamic, DL+0.25LL)
Rigid (nonlinear static, DL+0.25LL)
20 Rigid (nonlinear dynamic, DL+0.25LL)
Flexible (linear static, DL+0.25LL) 20
10 Flexible (nonlinear static, DL+0.25LL)
Flexible (nonlinear dynamic, DL+0.25LL)
0 0
0 4 8 12 16 0 4 8 12 16
Time (sec) Time (sec)
Fig. 26. History of axial force at right end of Beam-1 and Beam-2 for three-storey frames with extended end-plate: (a) scenario-1 and (b) scenario-2.

Table 3 [3] M. Ivanyi, Full-scale tests of steel frames with semi-rigid connections, Eng.
Maximum displacement obtained from different types of analysis and dynamic load Struct. 22 (2) (2000) 168179.
factor (DLF) within different scenarios (extended end plate assumed to be exible). [4] P.C.G. Da S. Vellasco, S.a.L. De Andrade, J.G.S. Da Silva, L.R.O. De Lima, O. Brito
Jr, A parametric analysis of steel and composite portal frames with semi-rigid
connections, Eng. Struct. 28 (4) (2006) 543556.
Maximum displacement from DLF for non-linear
[5] A. Shrih, A. Rahman, K.S. Al-Jabri, Finite element analyses of ush end-plate
different analyses (mm) analysis
connections between steel beams and columns at elevated temperatures,
Adv. Struct. Eng. 12 (3) (2009) 311324.
Linear static Non-linear [6] X.-M. Jiang, H. Chen, J.Y.R. Liew, Spread-of-plasticity analysis of three-
analysis analysis dimensional steel frames, J. Constr. Steel Res. 58 (2) (2002) 193212.
[7] E.M. Lui, W.F. Chen, Steel frame analysis with exible joints, J. Constr. Steel
Static Dynamic Res. 8 (1987) 161202.
[8] M. Sekulovic, R. Salatic, Nonlinear analysis of frames with exible connec-
Scenario- 105.6 296.2 516.9 1.75 tions, Comput. Struct. 79 (11) (2001) 10971107.
1 [9] J.Y.R. Liew, C.H. Yu, Y.H. Ng, N.E. Shanmugam, Testing of semi-rigid unbraced
Scenario- 75.7 256.1 518.4 2.02 frames for calibration of second-order inelastic analysis, J. Constr. Steel Res.
2 41 (23) (1997) 159195.
[10] A. Zona, G. Ranzi, Finite element models for nonlinear analysis of steel
concrete composite beams with partial interaction in combined bending and
shear, Finite Elem. Anal. Des. 47 (2) (2011) 98118.
The accuracy and efciency of the formulation and of the
[11] M.R. Mohamadi-Shooreh, M. Mod, Parametric analyses on the initial
adopted models to capture the global response of the steel frames stiffness of ush end-plate splice connections using FEM, J. Constr. Steel
including the strength and stiffness of the connections was Res. 64 (10) (2008) 11291141.
veried through numerical examples. [12] G. Shi, Y. Shi, Y. Wang, M.A. Bradford, Numerical simulation of steel
pretensioned bolted end-plate connections of different types and details,
The implications of different assumptions for the analysis and Eng. Struct. 30 (10) (2008) 26772686.
design of connections in steel frames with ush end plate and [13] W.F. Chen, N. Kishi, Semi-rigid steel beam-to-column connections: data base
extended end plate during different progressive collapse scenarios and modeling, J. Struct. Eng. ASCE 115 (1) (1989) 105119.
[14] N. Kishi, W.F. Chen, Momentrotation relations of semirigid connections with
were studied. In the cases studied in this paper, it was observed angles, J. Struct. Eng. ASCE 116 (7) (1990) 18131834.
that the assumption of full xity for ush end plate and extended [15] S.M. Choi, S.D. Hong, Y.S. Kim, Modeling analytical momentrotation curves
end plates can lead to an underestimation of the maximum and of semi-rigid connections for CFT square columns and steel beams, Adv.
Struct. Eng. 9 (5) (2006) 697706.
permanent deection as well as the axial force in the beams of up
[16] G.E. Mageirou, C.J. Gantes, Buckling strength of multi-story sway, non-sway
to 18%, depending on the method adopted for designing the and partially-sway frames with semi-rigid connections, J. Constr. Steel Res.
connections. Moreover, it was concluded that designing the ush 62 (9) (2006) 893905.
end plates for the plastic bending moment of the beams instead of [17] L. Xu, Y. Liu, Story stability of semi-braced steel frames, J. Constr. Steel Res. 58
(4) (2002) 467491.
the maximum bending moment (corresponding to the critical load [18] Z. Fu, K. Ohi, K. Takanashi, X. Lin, Seismic behavior of steel frames with semi-
combination) can improve the global performance of the structure rigid connections and braces, J. Constr. Steel Res. 46 (13) (1998) 440441.
during progressive collapse scenarios. [19] J.G.S. Da Silva, L.R.O. De Lima, P.C.G. Da S. Vellasco, S.a.L. De Andrade, R.A. De
Castro, Nonlinear dynamic analysis of steel portal frames with semi-rigid
Signicantly, for the cases studied in this paper it was shown
connections, Eng. Struct. 30 (9) (2008) 25662579.
that the dynamic load factors (DLF) vary between 1.75 and 2.02 [20] G.S. Urgessa, T. Arciszewski, Blast response comparison of multiple steel
for non-linear analysis (in steel frames with ush end plate and frame connections, Finite Elem. Anal. Des. 47 (7) (2011) 668675.
[21] C.G. Chiorean, A computer method for nonlinear inelastic analysis of 3D
extended end plate connections), which is consistent with the
semi-rigid steel frameworks, Eng. Struct. 31 (12) (2009) 30163033.
value of DLF2 proposed by some standards based on linear [22] T. Kim, J. Kim, Progressive collapse-resisting capacity of steel moment frames
analysis results [35]. Furthermore, it was concluded that the considering panel zone deformation, Adv. Struct. Eng. 12 (2) (2009) 231240.
assumption of full xity for ush end plate and extended end [23] J.Y.R. Liew, W.F. Chen, H. Chen, Advanced inelastic analysis of frame
structures, J. Constr. Steel Res. 55 (1) (2000) 245265.
plate connections has a major effect on the DLF. [24] R.L. Taylor, F.C. Filippou, A. Saritas, F. Auricchio, A mixed nite element
method for beam and frame problems, Comput. Mech. 31 (12 SPEC) (2003)
192203.
References [25] A. Kidarsa, M.H. Scott, C.C. Higgins, Analysis of moving loads using force-
based nite elements, Finite Elem. Anal. Des. 44 (4) (2008) 214224.
[26] A.G. Vlassis, B.A. Izzuddin, A.Y. Elghazouli, D.A. Nethercot, Progressive
[1] N.S. Trahair, M.A. Bradford, D.A. Nethercot, L. Gardner, The Behaviour and collapse of multi-storey buildings due to sudden column loss-Part II:
Design of Steel Structures to EC3, Taylor and Francis, London, 2008. Application, Eng. Struct. 30 (5) (2008) 14241438.
[2] E. Bayo, J.M. Cabrero, B. Gil, An effective component-based method to model [27] D.E. Grierson, L. Xu, Y. Liu, Progressive-failure analysis of buildings subjected
semi-rigid connections for the global analysis of steel and composite to abnormal loading, Comput.-Aided Civil Infrastruct. Eng. 20 (3) (2005)
structures, Eng. Struct. 28 (1) (2006) 97108. 155171.
48 H.R. Valipour, M. Bradford / Finite Elements in Analysis and Design 60 (2012) 3548

[28] B.A. Izzuddin, A.G. Vlassis, A.Y. Elghazouli, D.A. Nethercot, Progressive [40] H.R. Valipour, S.J. Foster, A total secant exibility-based formulation for
collapse of multi-storey buildings due to sudden column lossPart I: frame elements with physical and geometrical nonlinearities, Finite Elem.
Simplied assessment framework, Eng. Struct. 30 (5) (2008) 13081318. Anal. Des. 46 (3) (2010) 288297.
[29] K. Khandelwal, S. El-Tawil, S.K. Kunnath, H.S. Lew, Macromodel-based [41] A. Neuenhofer, F.C. Filippou, Geometrically nonlinear exibility-based frame
simulation of progressive collapse: steel frame structures, J. Struct. Eng. nite element, J. Struct. Eng. 124 (6) (1998) 704711.
134 (7) (2008) 10701078. [42] A.R. Kukreti, T.M. Murray, A. Abolmaali, End-plate connection moment
[30] C.-H. Lee, S. Kim, K.-H. Han, K. Lee, Simplied nonlinear progressive collapse rotation relationship, J. Constr. Steel Res. 8 (1987) 137157.
analysis of welded steel moment frames, J. Constr. Steel Res. 65 (5) (2009) [43] A.R. Kukreti, A. Abolmaali, Momentrotation hysteresis behavior of top and
11301137. seat angle steel frame connections, ASCE J. Struct. Eng. 125 (8) (1999)
[31] J. Lim, T. Krauthammer, Progressive collapse analyses of 2D steel-framed 810820.
structures with different connection models, Eng. J. 43 (3) (2006) 201215. [44] E. Attiogbe, G. Morris, Momentrotation functions for steel connections, J.
[32] J.E. Karns, D.L. Houghton, J.-K. Hong, J. Kim, Behavior of varied steel frame Struct. Eng. ASCE 117 (6) (1991) 17031718.
connection types subjected to air blast, debris impact, and/or post-blast [45] R.M. Richard, B.J. Abbott, Versatile elasticplastic stressstrain formula, ASCE
progressive collapse load conditions, Austin, TX, United States, 2009, pp. J. Eng. Mech. 101 (EM4) (1975) 511515.
18681877. [46] W. Ramberg, W.R. Osgood, Description of Stressstrain Curves by Three
[33] T. Kim, J. Kim, Collapse analysis of steel moment frames with various seismic Parameters, Report no. 902, National Advisory Committee for Aeronautics,
connections, J. Constr. Steel Res. 65 (6) (2009) 13161322. Washington, DC, 1943.
[34] A.G. Vlassis, B.A. Izzuddin, A.Y. Elghazouli, D.A. Nethercot, Progressive [47] K.M. Ang, G.A. Morris, Analysis of three-dimensional frames with exible
collapse of multi-storey buildings due to failed oor impact, Eng. Struct. 31 beam-column connections, Can. J. Civil Eng. 11 (1984) 245254.
(7) (2009) 15221534. [48] I. Carol, J. Murcia, Nonlinear time-dependent analysis of planar frames using
[35] C.H. Yu, J.Y.R. Liew, N.E. Shanmugam, Y.H. Ng, Collapse behaviour of sway an exact formulation. II. Computer implementation for R.C. structures and
frames with end-plate connections, J. Constr. Steel Res. 48 (23) (1998) examples, Comput. Struct. 33 (1) (1989) 89102.
169188. [49] H.R. Valipour, Nonlinear Analysis of Reinforced Concrete Frames Under
[36] GSA, Progressive Collapse Analysis and Design Guidelines for New Service Extreme Loadings, Ph.D. Thesis, School of Civil and Environmental Engineer-
Administration, Washington, DC, 2003. ing, The University of New South Wales, Sydney, Australia, 2009.
[37] Dod, Design of Building to Resist Progressive Collapse, Washington, DC, 2005. [50] A. Abolmaali, J.H. Matthys, M. Farooqi, Y. Choi, Development of moment
[38] J. Kim, T. Kim, Assessment of progressive collapse-resisting capacity of steel rotation model equations for ush end-plate connections, J. Constr. Steel Res.
moment frames, J. Constr. Steel Res. 65 (1) (2009) 169179. 61 (2005) 15951612.
[39] J. Kim, D. An, Evaluation of progressive collapse potential of steel moment [51] Standardsaustralia, AS4100 Steel Structures, SA, Sydney, 1998.
frames considering catenary action, Struct. Des. Tall Spec. Build. 18 (4) (2009) [52] Y.L. Lee, R.E. Melchers, Momentrotation curves for bolted connections, ASCE
455465. J. Struct. Eng. 112 (3) (1986) 615635.