Engineering Structures 132 (2017) 87–97

Contents lists available at ScienceDirect

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

Three-dimensional progressive collapse analysis of reinforced concrete

frame structures subjected to sequential column removal
Amir Hossein Arshian ⇑, Guido Morgenthal
Institute of Modelling and Simulation of Structures, Bauhaus University Weimar, 99423 Weimar, Germany

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

Article history: In the last decade, a great care is exercised in progressive collapse analysis of structures to avoid the
Received 29 April 2016 catastrophic consequences of such a system-level problem. The majority of the previous research works
Revised 19 August 2016 dealt with the quantification of resisting mechanisms such as the compressive arching action using two-
Accepted 7 November 2016
dimensional frameworks. The three-dimensional (3D) studies are also limited to considering the initial
damage as instantaneous removal of one or simultaneous removal of multiple supporting elements.
This paper studies the 3D nonlinear dynamic response of reinforced concrete structures subjected to
Keywords:
sequential column removal scenarios. A sequential nonlinear time-history analysis algorithm alongside
Progressive collapse analysis
Sequential column removal
with a macro modeling approach is utilized to predict the dynamic redistribution of the gravity loads.
Nonlinear dynamic analysis The efficiency of such a numerical framework is verified through comparison of computational results
Reinforced concrete frames with the available experimental data from a past 3D half-scale test. Good agreement is observed for
Fiber-based modeling the global and for the local response quantities. Furthermore, a practical strengthening technique is
Compressive arching action applied into the computational model of the structural system for artificially activating the catenary
Catenary action mechanism. Analysis results show that strengthening of peripheral beams with externally bonded steel
plates significantly increases the rotational ductility at beam-sections and in turn, enables the damaged
structure to accommodate larger deformations. Finally, the influence of the removal sequence on the 3D
force redistribution mechanism is investigated. Permanent plastic deformations and maximum sectional
forces of a sequential removal scenario are found to be larger on average compared with those obtained
from an at-once removal scenario. It is demonstrated that the time-lag between the column removals
considerably affects the 3D redistribution of gravity loads, and shall not be neglected in case of consid-
ering an extreme initial damage.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
The catastrophic collapse such as those of the Ronan Point
apartment building in 1968, the Alfred P. Murrah Federal Building
in 1995 and the World Trade Center (WTC) towers on 11 Septem-
ber 2001 prompted the research communities to explicitly address
such a system-level problem. Progressive collapse indicates a situ-
ation where failure of primary elements spread from element to
element resulting in the collapse of the entire structure or a large
part of it [1]. Progressive collapse guidelines such as [2,3] propose
two general approaches of Direct Design and Indirect Design for
mitigating the risk of progressive collapse. Among the proposed
approaches, the threat independent alternate path (AP) method,
which is a direct design approach is of researchers’ interest. This
method investigates the ability of the structure in redistributing
⇑ Corresponding author.
E-mail address: amir.hossein.arshian@uni-weimar.de (A.H. Arshian).
of the gravity loads through bridging over the failed supporting
elements.
In the last decade, a great care is exercised in the alternate path
analysis of frame structures. Different numerical modeling
approaches were studied. One of the simplest practiced modeling
strategies is using elastic Bernoulli beam-column elements with
lumped rigid-plastic hinges [4,5]. Such a modeling approach
neglects the contribution of membrane actions of the bridging
structural members in resisting progressive collapse. However, it
has been well proved that the membrane actions significantly
increase the collapse load in laterally restrained RC members [6–
9]. In the last decade, the problem of compressive membrane
action and catenary action has been significantly under research.
However, the majority of previous experimental and numerical
studies were mainly focused on either laterally restrained RC
sub-assemblages [10–15] or 2D planar frames [16–18]. A two-
dimensional study provides a relatively simple and efficient solu-
tion to the collapse resisting mechanisms such as the flexural

http://dx.doi.org/10.1016/j.engstruct.2016.11.018
0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.
88 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97
action, the compressive membrane action and the catenary action.
Nevertheless, apart from the ultimate-load bearing capacity of the
structural members, the 3D investigation of the force redistribu-
tion mechanism is also crucial in evaluating the collapse-
resistance of a given structural system. On the other hand, the
majority of 3D studies are limited to considering the initial damage
as instantaneous removal of one or simultaneous removal of mul-
tiple supporting elements, where little attention is paid to the
influence of the failure-sequence on the force redistribution
mechanism.
This paper aims to shed light on the influence of the removal
sequence on the 3D force redistribution of the gravity loads assum-
ing an extreme initial damage. A sequential nonlinear time-history
analysis algorithm alongside with a macro modeling approach is
utilized. Xiao et al.’s test [19] in which a 3D half-scale RC frame
structure is subjected to sequential removal of its four columns
is taken as the reference object. The efficiency of the applied
numerical framework is verified for the global as well as for the
local response quantities considering the entire range of the struc-
tural behavior from the elastic to the extreme plastic state, where
the structure is identified as prone to collapse. Furthermore, a
practical strengthening technique is applied into the computa-
tional model of the structural system for artificially activating
the catenary mechanism. Peripheral beams are strengthened with
externally bonded steel plates attached to the bottom of the beam-
sections. The load-bearing capacity enhancement due to the devel-
opment of catenary forces is comprehensively discussed. Finally,
the influence of the removal sequence on the 3D force redistribu-
tion mechanism is investigated through a comparative study,
where computational results are compared with an at-once
removal scenario.
2. Analysis methodology for sequential removal scenarios
2.1. Overview
In an alternate path analysis, the aim is to investigate the ability
of the structure in redistributing of the gravity loads to the undam-
aged bays. The proposed analytical approaches are typically using
the linear static, nonlinear static, and nonlinear dynamic analyses.
Although the linear static analysis approach is easy to be applied, it
may not properly reflect the nonlinearity and the dynamic effects
[20]. Usually, the nonlinear static procedure is preferred due to
its simplicity compared with the nonlinear dynamic analysis pro-
cedure. In a nonlinear static procedure, the dynamic effects will
be reflected using load increase factors such as those proposed
by [21,22]. Nevertheless, the proposed load increase factors are
mainly based on the single column removal assumption, which is
not valid for the case of extreme initial damage. Thus, to have a
clear understanding of force redistribution mechanism, particu-
larly for the extreme events, use of nonlinear dynamic analysis
approach is inevitable. As for considering an extreme initial dam-
age, the simultaneous removal of multiple beam-column elements
is already investigated by [23–25]. However, little attention is paid
to the influence of the removal-sequence on the 3D force redistri-
bution, where the redistribution of the gravity loads continuously
changes its direction following each column removal. Contrary to
the simultaneous removal strategy, the forces will be redistributed
in an unfavorable asymmetric manner considering the time-lag
between the column removals. That is, an at-once removal scenario
does not necessarily yield a conservative solution to the problem of
progressive collapse resistance for a given structural system, where
a limited number of beam-column elements are failed as the direct
or indirect consequence of an extreme loading scenario. This paper
utilizes a sequential nonlinear time-history analysis algorithm
using macro modeling approach for investigating the 3D force
redistribution mechanism for the sequential removal scenarios.
The proposed numerical framework can be also implemented in
robustness assessment of the structural systems. A comparable
numerical framework for robustness assessment of the structural
systems is the common used incremental-mass nonlinear dynamic
analysis [4,26,27]. In an incremental-mass nonlinear dynamic
analysis, the residual capacity of the damaged structure will be
obtained with virtually increasing the gravity loads in an incre-
mental time-history analysis. Although such an analysis approach
predicts the behavior of the damaged zone from the elastic state
to the dynamic instability state, the initial damage is limited to
the prescribed local failure. On the contrary, a sequential-
removal framework enables the analyst to robustly consider the
propagation of failures. Thus, predicting the collapse extension is
possible. Besides, there is no need to artificially increase the gravity
loads using such a framework, which is more realistic in terms of
structural loading.
2.2. Finite element modeling approach
Fiber-based modeling approach provides a reliable solution to
the distributed plasticity problem in frame elements. Such a mod-
eling technique has been widely used in predicting the ultimate
load-bearing capacity of frame elements considering the flexural-
axial deformation interactions [17,25,26,28]. As for the element-
formulation, the displacement-based element (DBE) and the
force-based element (FBE) formulations are typically implemented.
In the conventional DBE formulation, the solution is derived based
on discretization of the displacement fields of the element using
approximate interpolation functions. However, a FBE yields the
solution using exact force interpolation functions. Thus, a FBE pro-
vides a robust solution as only numerical integration error is
involved [29]. Nevertheless, facing extreme convergence problems
is very probable using FBEs, particularly in softening problems
such as progressive collapse analysis. Thus, a great care must be
exercised in the finite-element discretization as well as in the solu-
tion algorithm using FBE formulation. A comprehensive discussion
on the efficiency of using DBE and FBE formulations in progressive
collapse analysis of RC frame structures can be found in Arshian
et al. [26].
As for the planar elements, the nonlinear multi-layered model-
ing approach is widely used to account for the development of
nonlinear membrane forces [30–32]. In a multi-layered shell ele-
ment, using the plane-section assumption, strain at each layer will
be obtained according to the axial strain and curvature of the mid-
dle layer. Using corresponding constitutive material models, the
subsequent stresses at each layer will be calculated then. The pla-
nar constitutive model for concrete which is based on the concept
of damage mechanics and smeared crack model includes the non-
linear behavior of concrete such as cracking, crushing and aggre-
gate interlocking. The shear stiffness reduction following the
cracking of concrete layers is also accounted for using the so called
shear retention factor [33]. The nonlinear behavior of reinforcing
steels is also reflected in the uniaxial stress-strain relationship of
smeared steel layers.
2.3. Analysis methodology
The analysis procedure for a sequential column removal sce-
nario is schematically depicted in Fig. 1. Finding the equilibrium
forces (P 1 ; V 1 ; M 1 ) for the first removal through a finite element
analysis of the undamaged structure, the gravity loads G alongside
with the equilibrium forces are applied into a damaged finite ele-
ment model of the structural system. While the gravity loads
remain constant, the equilibrium forces drop to zero at t ¼ t1 , in
 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97 89

G
G

Load
P1 , V1 , M1
t1 t update the model
t = t2 restart the analysis
V1 M1 Δ11
P1

V2 M2 first removal
P2
Load G
G P1 , V1 , M1
t1 t
P2
t1 t2 t
t = t2 t >> t2
V1 M1 V2 M2 Δ11 V2 (t) M2 (t) Δ22
P1 P2 V2 Δ12
P2 (t)
t1 t2 t
M2 first removal second removal
t1 t2 t
Fig. 1. Time-history analysis procedure for sequential column removal scenarios: the first column is removed at t ¼ t 1 , and the second column is removed at t ¼ t 2 ; G is the
gravity load, P i  V i  M i are the sectional forces at the top of the removed columns, and Dij shows the permanent vertical displacement at the top of the ith column due to the
removal of the j column.

a short period of time, representing the instantaneous loss of the
first column. Variation of the equilibrium forces at the top of the
second column (P 2 ; V 2 ; M 2 ) is recorded during the first time-
history analysis (t ¼ ½0  t 2 ). In the next step, the gravity loads
alongside with the equilibrium forces for the two removed col-
umns are applied into the updated damaged-model of the struc-
ture, and the analysis is repeated (t ¼ ½0  t1 ), where the
equilibrium forces for the second removal drop to zero at t ¼ t2 .
This procedure shall continue till the structural system reaches
to a global dynamic instability state, where the structure is prone
to collapse. Apart from the column removal strategy, the force
redistribution due to the failure of beam-sections resulting from
the crushing of concrete fibers or rupturing of reinforcing steels
can be also reflected in the analysis procedure.
3. Reference object
3.1. Description of the test-frame
The reference object is a 3-bay  3-bay, 3-story, half-scale rein-
forced concrete frame, which has been taken from Xiao et al.’s test
[19] at the structural laboratory of Hunan University in China. The
structural system is a moment resisting frame with two way solid
slab system, which is meeting the requirements of seismic design.
During Xiao et al.’s test, the frame is subjected to sequential
removal of A1, B1, D3, and D2 columns. Columns were removed
one after another using hydrogen guns. The plan view of the model
frame, loaded regions (hatched areas), elevation view and cross-
section properties are shown in Fig. 2. Material properties for con-
0 order P-Delta geometric transformation is also used for the
crete compressive strength f c , yielding strength f y and ultimate
strength f u of rebars are provided in Table 1.
3.2. Numerical model
A 3D computational model of the reference object is developed
in the object-oriented, open-source software framework, OpenSees
[34]. Beam-column elements associated with the damaged bays
are modeled using FBEs while DBEs are used for those of in undam-
aged bays. Each longitudinal and transverse beam-span is dis-
cretized into eight and six beam-column elements, respectively.
Four Gauss-Lobatto integration points are assigned to each
beam-element. The Gauss-Lobatto integration provides an integra-
tion point at each end of the element, and is the most common
integration type for the nonlinear finite element analysis of frame
elements [35]. As for the constitutive models, the uniaxial Popovics
concrete material model [36] is used for both unconfined and con-
fined concrete fibers considering tensile strength of concrete with
an exponential decay. The uniaxial ‘Reinforcing Steel’ material
model according to the work of Chang and Mander [37] is also
accepted for the stress-strain relationship of steel fibers. Using
‘MinMax’ material object, ultimate strain values are introduced
assuming that fibers will fail (stress and tangent will be zero) by
reaching that strain threshold. Slabs are modeled using the four
nodded ‘ShellMITC4’ element alongside with a nonlinear multi-
layered modeling approach [32]. Each slab-panel is discretized into
number of shell elements affine to the number of beam elements in
the longitudinal and transverse directions, where each shell ele-
ment consists of eight layers (two layers for top and bottom con-
crete covers, four smeared layers of top and bottom reinforcing
steels in orthogonal directions and two concrete layers between
the reinforcing steel layers). Fig. 3 schematically shows the compu-
tational model of the reference object.
The geometrical nonlinearity is accounted for considering
‘Corotational’ formulation for beam-column elements in the dam-
aged bays, where large deformations are expected. The second-
beam-column elements associated with the undamaged bays.
Damping is also introduced into the structural system using a clas-
sical mass-stiffness proportional damping ratio of n ¼ 0:05 for the
two fundamental modes of vibration associated with the damaged
bays.
As for the solution algorithm, the common used ‘Newton-
Raphson’ algorithm is used. The Nemark-Beta method using
90 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97

1 2 3 4 225 4 #12 225 4 #12

13710
4570 4570 4570
355 355 3 #12
225x355 225x355 225x355
D 3 #12 4 #12
C1 C2 C2 C1 Ends Mids
3 #12 3 #10
225x355

225x355
3050
280 280
225x355 225x355
C 3 #10 3 #12
C3 C4 C4 C3 200 200
200x280

200x280

225x355
9150 3050 Ends Mids

225x355 225x355 350

B 20 #14
C3 C4 C4 C3 350
200x280

200x280

225x355
3050
(C1 & C4)
225x355 225x355
A 330
C1 C2 C2 C1
16 #14
330
1830
(C2)

300
1830
16 #14
300

2280
(C3)

115 # 8 @ 200
4570 4570 4570
13710
1 2 3 4 RC Slab

Fig. 2. Structural plan, elevation view and sectional properties of frame model: hatched areas are loaded with concrete blocks (20 kN/m2), cross marks show removed
columns, sequence of column removal is A1-B1-D3-D2, Ends and Mids in the reinforcing layout stand for end beam sections and middle beam sections, reproduced after Xiao
et al. [19].

Table 1
Material properties of the test-frame as taken from Xiao et al. [19].

Concrete properties Steel properties

0
Location ID f c [MPa] ID / [mm] f y [MPa] f u [MPa]

1st-S Co 33.28 #8 8 453.0 657.3

2nd-S BM-S 29.76 #10 10 432.7 649.1
2nd-S Co, 3rd-S BM-S 33.84 #12 12 485.8 640.7
3rd-S Co, 4th-S BM-S 29.36 #14 14 456.8 627.4

i-S: story ID.

Co: columns, BM-S: beams and slabs.

average acceleration method (c ¼ 0:5; b ¼ 0:25) is also used as for
the numerical integrator. The convergence tolerance is initially
set to 108 using the ‘Norm Displacement Increment’ convergence
test. Numerical convergence is typically difficult to achieve in non-
linear analysis of RC structures. Hence, a robust combination of
solution algorithms such as ‘Krylov-Newton’ [38] and ‘Broyden’
[39] is used to overcome convergence problems. To simulate the
collapse, including extreme physical and geometrical nonlineari-
ties, several solution schemes are practiced in an iterative manner.
A combination of solvers (Krylov-Newton and Broyden) with dif-
ferent convergence tolerances (107 ; 106 ; 105 , and 104 ) is used
to ensure that the ultimate convergence failure is probably hap-
pening due to global dynamic instability of the structural system.
In those converged steps associated to the use of large convergence
 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97 91

(a) (c)

Z Y

X σc
X
Concrete layer
Integration Points
εc
concere fibres

•
•

•
•

••
••

••
••
σs

•
•

•
•

•
•

•
•
(b) (d)
εs
Layer of smeared rebar steel fibres

Fig. 3. Finite element model of the analyzed frame: (a) plan view of the computational model with the location of removed columns (cross marks), (b) a typical multi-layered
shell element for modeling of reinforced concrete slabs, (c) elevation view of longitudinal frames with internal nodes, and (d) a typical fiber-based beam element with four
integration points, columns are modeled as one element with seven integration points.

tolerance such as 104 , the global and the local response quantities
are checked to ensure that no numerical error is developing in the
solution.
The computational model of the structural system is subjected
to sequential removal of A1, B1, D3, and D2 columns using the
sequential algorithm depicted in Fig. 1. The time-lag between the
column removals is considered large enough letting the structure
to find a new equilibrium path, where the dynamic effects are
gone. Based on a trial analysis, a time-lag of 1:2 seconds is found
to be large enough for excluding the dynamic effects from the pre-
vious column removal as well as for justifying the overall compu-
tational costs.
3.3. Finite element analysis results
3.3.1. Vertical displacements and global force redistribution
Vertical displacement-time histories of the second floor joints
above the removed columns are compared with those of Xiao
et al.’s test [19] in Figs. 4 and 5. A reasonable agreement can be
seen in global displacements, which demonstrates that the utilized
numerical framework can robustly estimate the global dynamic
behavior of the structure. Numerical results are slightly deviated
from the experimental records of Xiao et al.’s test [19] in the first
three column removals and particularly after the peak response,
where the structure is responded mainly in the elastic fashion. It
is well-known that the use of Rayleigh damping is only a rough
approximation for expressing the energy dissipation in a structural
system, and can result in unrealistic damping forces [40]. Further-
more, the actual behavior of beam-column joints cannot be
reflected using rigid beam-column connections. Apart from the
modeling issues, the uncertainty in material properties, geometri-
cal properties, mechanical properties and loading is another source
of the observed disagreements [26,41,42]. The contour plot of per-
manent vertical displacements at the second floor is also depicted
in Fig. 6 for each removal-sequence. The white color in the contour
plots shows the upward movement of the floor-joints. It can be
seen that with consideration of the time-lag between the column
removals, the direction of the deflection in some regions conscien-
tiously changes, which results in multiple change in the direction
of the curvature in the slabs.
Fig. 7 also shows how the gravity load is redistributed. The aver-
age of relative increase in the axial forces of the first story columns
shows the relative severity of each removal-sequence in terms of
global force redistribution. The least severe case is found as
removal of A1 column, where the structure responded mainly in
the elastic fashion. Following the removal of A1 column, an average

0.0

−6.0
(B1)

−12.0 (A1)
Δ[mm]

−18.0

−24.0
(B1) (A1)
−30.0
0.0 0.4 0.8 1.2 1.6 2.0 2.4 1.2 1.6 2.0 2.4
t[s] t[s]
Fig. 4. Vertical displacement D time-histories at the top of the A1-column (left) and B1-column (right) following the removal of A1 and B1 columns (Phase I): A1 column is
removed at t ¼ 0, and B1 column is removed at t ¼ 1:2, numerical results (thick —), and test results (thin —) with a cut-off in time domain as taken from Xiao et al.’s test [19].
92 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97

0.0
(D2) (D3)
−14.0

−28.0
Δ[mm]

−42.0

−56.0
(D2) (D3)
−70.0
2.4 2.8 3.2 3.6 4.0 4.4 4.8 3.6 4.0 4.4 4.8
t[s] t[s]
Fig. 5. Vertical displacement D time-histories at the top of the D3-column (left) and D2-column (right) following the removal of D3 and D2 columns (Phase II, A1 and B1
columns are already removed): D3 column is removed at t ¼ 2:4, and D2 column is removed at t ¼ 3:6, numerical results (thick —), and test results (thin —) with a cut-off in
time domain as taken from Xiao et al.’s test [19].

Δ[mm]
+3.00

− 6.00

− 15.0

− 24.0
D
− 33.0
C − 42.0

B − 51.0

A − 60.0
1 2 3 4 1 2 3 4

Fig. 6. Contour plot of permanent vertical displacements D at the second floor following the sequential removal of: A1 column (top-left), B1 column (bottom-left), D3 column
(top-right), and D2 column (bottom-right).

240.0 increase of 78% in the axial forces of the remaining first story col-
umns is observed for this case. Although the double-span effect is
Increase in axial forces [%]

200.0 found as the common characteristic for B1 and D3 column

removals, the boundary condition for the B1 column removal is
160.0 more severe compared with that of D3 column removal due to
the removal of A1 column. Thus, B1 column removal is identified
120.0 as the second most extreme scenario. Besides, the average of rela-
tive increase in the axial forces of neighboring columns for each
80.0 removal-sequence found to be much larger compared with that
of considering all remaining columns. That is, the gravity loads
40.0 are mainly redistributed to the neighboring columns, which is in
agreement with the concentration of global deformations in the
0.0 vicinity of removed columns as depicted in the displacement con-
A1 B1 D3 D2
tour plots.
Column removal ID
Fig. 7. Average relative increase in the axial forces of first story columns: ( ) shows
the average increase in the axial forces of neighboring columns, and (j) shows the 3.3.2. Sectional forces, local plastification and global failure
average increase in the axial forces of all remaining first-story columns. Following the removal of A1 and B1 columns the structure
mainly responded in elastic fashion, which is in agreement with
the experimental observations from Xiao et al.’s test [19]. Axial
force-time history and moment-curvature relationships for end
increase of 4% is observed in the axial forces of the remaining col- sections of the longitudinal beam A1-A2 at the second floor are
umns in the first story. The most extreme case is associated with shown in Fig. 8. Moment-curvature relationships show slight non-
the removal of D2 column, where large plastic deformations linear behavior, which is due to cracking of concrete fibers. Bend-
occurred due to the triple-span effect alongside with an asymmet- ing moments exceeded the yielding capacity M y based on
ric force redistribution in the plane of frame ‘D’. An average conventional plastic analysis that disregards axial force effects.
 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97 93

150.0 80.0

0.0 40.0 (B1)

(A1)
−150.0 0.0

ΔM [kNm]
N [kN]

−300.0 −40.0 S1 S2

−450.0 −80.0

−600.0 −120.0

−750.0 −160.0
0.0 0.4 0.8 1.2 1.6 2.0 2.4 −0.020 −0.015 −0.010 −0.005 0.000 0.005 0.010
t[s] Δφ[1/m]

Fig. 8. Axial force N time-history (left) and moment-curvature relationship (right) for section S1 (thin —) and section S2 (thick —) following the removal of A1 and B1
columns: A1 column is removed at t ¼ 0:0 and B1 column is removed at t ¼ 1:2, bending moment M and curvature / variations are shown with respect to the undamaged
state.

However, no yielding is observed at any beam-section, which is in
agreement with the experimental observations from Xiao et al.’s
test [19]. Such a large yielding capacity is due to the considerable
compressive forces developed at the fixed end of bridging beams. It
seems that the out-of-plane stiffness of perimeter frames namely,
frame ‘1’ and frame ‘A’, is enough for restraining the axial elonga-
tion of bridging beams and in turn, activating the compressive
arching action at the free end (A1 joint) of these beams. At the
A1-end (section S1) of A1-A2 beam, and following the removal of
B1 column, development of compressive forces made an abrupt
change within the moment-curvature relationship of the beam-
section avoiding yielding of bottom reinforcing steels. Further-
more, changing of bending moments’ direction at the face of A1-
column resulted in double curvature in the longitudinal and trans-
verse beams and in turn, redistribution of the gravity loads due to
the so-called Vierendeel action.
Fig. 9 shows the strain-time histories of top layer reinforcing
steels at A2-end (sections S2 and S4) of beam A1-A2. The numerical
tensile strains in these bars are in good agreement with the exper-
imental records of Xiao et al.’s test [19] for section S2 at the second
floor. However, the computational model is overestimating the
tensile strains at the fourth floor probably because of considering
a rigid connection between the beams and the columns.
A similar trend is observed in the global and in the local
response quantities following the removal of D3 column. Although
no yielding is observed following the removal of D3 column, con-
crete cracking is extensively propagated due to sequential removal
of three columns. On the contrary, and following the removal of D2
column, large plastic deformations occurred in the damaged zone
of D1-C1-C4-D4. During Xiao et al.’s test [19], frame ‘D’ is collapsed
with a delay of approximately 15 min following the removal of D2
column. As for the numerical studies, development of severe plas-
tic deformations at several beam-column sections in the damaged
zone also demonstrates the collapse-vulnerability of frame ‘D’.
Besides, a subsequent incremental-mass nonlinear dynamic analy-
sis of the damaged zone revealed that the residual capacity is only
14% of the applied gravity loads, which is negligible considering
existing uncertainties in loading and structural properties. The
computational results are discussed for the following time-
domain:
 Column A1 is removed at t ¼ 0:0 s.
 Column B1 is removed at t ¼ 1:2 s.
 Column D3 is removed at t ¼ 2:4 s.
 Column D2 is removed at t ¼ 3:6 s.
 t ¼ 3:6 s to t ¼ 5:2 s, a period for stabilizing the vibration.
 From t ¼ 5:2 s, the structure is subjected to an incremental-
mass nonlinear dynamic analysis, where the gravity loads asso-
ciated with the damaged bays are only increased.
During the incremental-mass analysis, transverse beam C2-D2
failed due to failure of top and bottom reinforcing steels at C2-
end. The longitudinal beam D1-D2 also lost load-bearing capacity
due to excessive yielding of top and bottom reinforcing steels

0.0020 0.0020

(B1)
0.0015 0.0015
S4
ε[-]

ε[-]

0.0010 0.0010 S2

0.0005 0.0005

0.0000 0.0000
1.2 1.6 2.0 2.4 1.2 1.6 2.0 2.4
t[s] t[s]
Fig. 9. Strain e time-histories of top reinforcing steels at section S2 (left) and section S4 (right) following the removal of B1 column at t ¼ 1:2 (A1 column is already removed):
numerical results (thick —), and test results (thin —) as taken from Xiao et al.’s test [19], ey ¼ 0:0024.
94 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97

alongside with crushing of concrete fibers at D1-end. Moment- 0.0

curvature relationships of these beam-sections are shown in
Fig. 10. Eventually, failure of bottom reinforcing steels at C3-end
of C3-D3 beam resulted in a full formation of collapse mechanism.
Fig. 11 also demonstrates that the utilized numerical framework −150.0
can robustly predict the ultimate deformation of frame ‘D’ at the Effective Depth

Δ[mm]
collapse-state.

−300.0
4. Activating of the catenary response

It is well-known that developing of tensile membrane forces in

laterally restrained RC components at large deflections can consid- −450.0
erably increase the collapse load and in turn, contributes to the D3
D1 D2 D4
structural robustness [16,44,43]. Following the removal of side col-
umns, large plastic deformations occurred in beam-column sec- Fig. 11. Vertical displacement D of frame ‘D’ for the second floor at the collapse
tions. Vertical displacements exceeded one-depth of longitudinal state: numerical results (—), and test results (–  –) as taken from Xiao et al.’s test
beams, and frame ‘D’ collapsed due to failure of several beam sec- [19].
tions, where no contribution is observed from the tensile catenary
mechanism. It seems that neither the available rotational ductility
in the longitudinal beams nor the out-of-plane stiffness of frames
‘1’ and ‘4’ are adequate for activating the catenary response in
the longitudinal beams.
ts ≤ 4 mm ts ≤ 12 mm
A practical strengthening technique is applied to the computa-
tional model of the test-frame for artificially activating the cate- tg ≤ 2 mm tg ≤ 2 mm
nary mechanism. Peripheral beams are strengthened with steel bs ≥ 50 mm bs ≥ 80 mm
plates attached to the bottom of beam-sections at all floor-levels.
Fig. 12 shows the recommended dimension limits for the steel
plates [45]. In this study, steel plates of 200 mm width
(bs ¼ 200 mm) and 8 mm thick (t s ¼ 8 mm) are assumed to be
anchored (using high-strength steel bolts) to the peripheral RC
beam-sections, where a perfect bond is assumed between the con-
crete section and the steel plate.
The strengthened structure is subjected to the same sequential tg tg
ts ts
removal-scenario. Following the removal of the columns, the resid- epoxy
ual capacity of the damaged structure is obtained using an steel plate
incremental-mass nonlinear dynamic analysis. Computational metallic anchor
results showed that the use of externally bonded steel plates
bs bs
enables the longitudinal beams to go further into the catenary
phase. Development of tensile membrane forces alongside with Fig. 12. Recommended dimensions limits for strengthening of RC beams with steel
the displacement-time history of joint D2 is depicted in Fig. 13. plates for two cases of without and with metallic anchors, after [45].

−90.0 (D2)

−75.0
Section S5
−60.0
ΔM [kNm]

Section S6
−45.0
−30.0
−15.0

0.0
0.00 −0.07 −0.14 −0.21 −0.28 −0.35
S5
Δφ[1/m] (D2) (D3)

Section S5 Section S6

S6

(D2)

Fig. 10. Moment-curvature relationships of section S5 and section S6 following the removal of D2 column (A1, B1, and D3 columns are already removed): bending moment M
and curvature / variations are shown with respect to the undamaged state, ‘O’ is showing the permanent state following the instantaneous removal of D2 column, and ‘⁄’ is
showing the failure state in the subsequent incremental-mass analysis, the corresponding damaged photos are taken from Xiao et al.’s test [19].
 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97 95

0.0 600.0

Start of catenary action

−200.0 300.0 Δ=340 mm
Δ[mm]

N [kN]
(D2)
−400.0 0.0 (D1)

Effective Depth
−600.0 −300.0
S7
(D2) (D3)
−800.0 −600.0
3.60 4.80 6.00 7.20 3.60 4.80 6.00 7.20
t[s] t[s]

Fig. 13. Vertical displacement D time-history at the top of the D2 column (left) and axial force N time-history of section S7 (right) following the removal of D2 column at
t ¼ 3:6 (A1, B1, and D3 columns are already removed): original test-frame (thick —), peripheral beams strengthened with steel-plates of 8 mm in thickness attached to the
bottom of the beam-sections (thin —).

Tensile catenary forces started to develop at a large deflection of Table 2

340 mm, which is almost one-depth of beam-sections Influence of the removal sequence on permanent vertical displacements at the top of
the removed columns.
(Hbeam ¼ 355 mm). Accommodating of such a large deformation is
due to significant increase in the rotational ductility of beam- Joint ID DSi a [mm] DSe b [mm] ½ðDSi  DSe Þ=DSe  [%]
sections provided by the strengthening steel plates. Fig. 14 com- A1 21.91 20.60 +6.40
pares the ductility of the original frame with the strengthened B1 15.18 14.00 +8.43
one at local and global levels. D3 38.74 39.73 2.50
D2 53.10 55.70 4.70
The residual load-bearing capacity of the strengthened frame is
a
found as 60% of the applied gravity loads. Hence, use of strength- Simultaneous removal.
b
ening steel plates enhanced the ultimate load-bearing capacity by Sequential removal.
46% (the original frame had a 14% residual capacity) which is
partly due to the increase in the flexural capacity of peripheral
beams, and partly because of the development of large tensile cate- Analysis results showed that in a sequential removal strategy,
nary forces. beam-column sections may undergo cycles of loading and unload-
ing due to a continuous change in the direction of the force redis-
tribution. Fig. 15 explains such a phenomenon, which is a
5. Influence of the removal sequence characteristic of sequential removal scenarios. Furthermore, it
might seems at first glance that simultaneous removal of columns
This section aims to study the influence of the removal (only when limited number of columns are to be removed)
sequence on the 3D force redistribution mechanism. The computa- imposes larger plastic deformations to the beam-column sections.
tional model of the test-frame is subjected to simultaneous (at- Nevertheless, studying 54 critically loaded beam sections showed
once) removal of the columns neglecting the time-lag between that the maximum sectional forces as well as the plastic deforma-
the removals. Permanent vertical displacements at the top of the tions are larger on average in case of sequential removal. This is
removed columns are compared for the studied removal strategies due to the fact that in a sequential removal scenario, the axial
in Table 2. Although the permanent vertical displacements are forces of neighboring columns (subsequent removals) are already
found comparable, the 3D redistribution of the gravity loads is increased due to the redistribution of the gravity loads from the
quite different. previous removal. That is, each subsequent removal imposes larger
dynamic effects to the damaged structure. Besides, the sequential
50.0 removal of side columns resulted in an unfavorable asymmetric
force redistribution of forces in the plane of frame ‘D’, which is
40.0 not the case for the simultaneous removal of these columns.

30.0
6. Summary and conclusion
μ[-]

20.0 This paper studied the three-dimensional (3D) force redistribu-

10.0
0.0
W/O W/Spl−8
Fig. 14. Influence of strengthening of peripheral beams on local and global ductility
l levels: ( ) shows the curvature ductility (lu =ly ) at section S5, and (j) shows the
global-displacement ductility (Du =Dy ) considering the vertical displacement at the
top of the D2 column, ‘W/O’ stands for the original frame, and ‘W/Spl-8’ stands for
the case of strengthening peripheral beams with 8 mm steel plates, see Fig. 10 for
the intended locations.
tion mechanism for sequential column removal scenarios. Xiao
et al.’s test [19] in which a 3D half-scale reinforced concrete (RC)
frame structure is subjected to sequential removal of its four col-
umns is taken as the reference object. Computational model of
the test-frame is built using a macro modeling approach, where
beam-column elements are modeled using stiffness-based and
flexibility-based element formulations. Slabs are also modeled
using quadrilateral multi-layered shell elements. The computa-
tional model is analyzed using a sequential nonlinear time-
history analysis. Based on a trial analysis, a time-lag of 1:2 s is
found to be large enough for excluding the dynamic effects from
96 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97
60.0
15.0
N [kN]
90.0
165.0
240.0
0.00 1.60 3.20
t[s]
Fig. 15. Axial force N time-history of column D1 in the first story (left) and bending moment M time-history of section S7 (right): (thick —) is showing the sequential removal
scenario (removing of A1, B1, D3, and D2 columns one-after-another), and (thin —) is showing the simultaneous removal of the columns, see Fig. 13 for the given beam-
column sections.
the previous column removal as well as for justifying the overall
computational costs. Following the removal of four columns, the
computational model of the structure is subjected to an
incremental-mass nonlinear dynamic analysis, where the gravity
loads associated with the damaged bays are only increased. The
obtained numerical results such as the global deformations, the
sectional forces, and the local plastifications were justifying the
experimental measurements and observations from Xiao et al.’s
test [19]. Following the removal of the last column (column D2),
and during the incremental-mass nonlinear dynamic analysis, the
associated longitudinal frame moved downward and vertical dis-
placements exceeded one-depth of beam-sections. The test-frame
collapsed due to the failure of several beam-sections, where the
failure pattern is identified quite similar to that of reported in Xiao
et al.’s test [19]. Such a good agreement demonstrates that the
applied numerical framework can robustly predict the entire range
of structural behavior from the elastic to the extreme plastic state,
where the structure is prone to collapse. Major local resisting
mechanisms are identified as flexural action and compressive
membrane action in beam-slab elements associated with the dam-
aged bays. As for the global resisting mechanism, the Vierendeel
frame action redistributed the gravity loads to the undamaged
bays.
During the failure process, no contribution is observed from the
catenary mechanism. Thus, a practical strengthening technique is
applied into the computational model of the test-frame for artifi-
cially activating the catenary response. Analysis results showed
that use of externally bonded steel plates significantly increases
the rotational ductility at beam-sections and in turn, enables the
damaged structure to accommodate larger deformations. Conse-
quently, considerable tensile membrane forces are developed at
beam-sections, where the longitudinal frame is failed at a very
large deflection of 680 mm (1.92 of beam-depth). The residual
load-bearing capacity of the strengthened frame is found as 60%
of the applied gravity loads. Hence, use of strengthening steel
plates enhanced the ultimate load-bearing capacity by 46% (the
original frame had a 14% residual capacity) which is partly due
to the increase in the flexural capacity of peripheral beams, and
partly because of the development of large tensile catenary forces.
Finally, the influence of the removal sequence on the 3D force
redistribution mechanism is investigated through a comparative
study, where computational results are compared with an at-
once removal scenario. Permanent plastic deformations and maxi-
mum sectional forces in the critically loaded beam-column sec-
tions are found to be larger on average for the sequential
removal strategy. Beam-column sections underwent cycles of load-
40.0
0.0
M [kNm]
40.0
80.0
120.0
4.80 0.00 1.60 3.20 4.80
t[s]
ing and unloading, which is due to the continuous change in the
direction of the redistributed gravity loads, and is the characteristic
of a sequential removal strategy. Furthermore, it is demonstrated
that in a sequential removal scenario, the gravity loads will be
redistributed in an unfavorable asymmetric pattern, which
imposes larger plastic deformations to the beam-column section
in the damaged bays. It seems that an at-once removal analysis
is only acceptable for those extreme initial damages with a negli-
gible time-lag (few milliseconds) between the removals. Generally,
the time-lag between the removals has a great impact on the 3D
force redistribution, and shall not be neglected in evaluating the
collapse resistance of a given structural system subjected to an
extreme initial damage.
References
[1] ASCE. Minimum design loads for buildings and other structures. ASCE/SEI 7;
2005.
[2] DoD. Unified Facilities Criteria (UFC) 4-023-03: design of structures to resist
progressive collapse. United States Department of Defense (DoD); 2010.
[3] GSA. Alternate path analysis and design guidelines for progressive collapse
resistance. The US General Services Administration; 2013.
[4] Tsai M-H, Lin B-H. Investigation of progressive collapse resistance and inelastic
response for an earthquake-resistant rc building subjected to column failure.
Eng Struct 2008;30(12):3619–28.
[5] Kokot S, Anthoine A, Negro P, Solomos G. Static and dynamic analysis of a
reinforced concrete flat slab frame building for progressive collapse. Eng Struct
2012;40:205–17.
[6] Ockleston A. Load tests on a three storey reinforced concrete building in
johannesburg. Struct Eng 1955;33(10):304–22.
[7] Park R. Ultimate strength of rectangular concrete slabs under short-term
uniform loading with edges restrained against lateral movement. ICE
proceedings, vol. 28. Thomas Telford; 1964. p. 125–50.
[8] Rankin G, Long A. Arching action strength enhancement in laterally-restrained
slab strips. Proc ICE-Struct Build 1997;122(4):461–7.
[9] Taylor SE, Rankin B, Cleland DJ, Kirkpatrick J. Serviceability of bridge deck slabs
with arching action. ACI Struct J 2007;104(1).
[10] Su Y, Tian Y, Song X. Progressive collapse resistance of axially-restrained frame
beams. ACI Struct J 2009;106(5).
[11] Wang G, Wang Q-x, Li Z-J. Membrane action in lateral restraint reinforced
concrete slabs. J Central South Univ Technol 2011;18:550–7.
[12] Beeby A, Fathibitaraf F. Membrane effect in the reinforced concrete framesa
proposal for change in the design of frames structures. Eng Struct 2001;23
(1):82–93.
[13] He Q, Yi W. Experimental study on collapse-resistant behavior of rc beam-
column sub-structure considering catenary action. In: The 14th world
conference on earthquake engineering. p. 12–7. October.
[14] Choi H, Kim J. Progressive collapse-resisting capacity of rc beam-column sub-
assemblage. Mag Concr Res 2011;63(4):297–310.
[15] Kai Q, Li B. Dynamic performance of rc beam-column substructures under the
scenario of the loss of a corner column experimental results. Eng Struct
2012;42:154–67.
[16] Yi W-J, He Q-F, Xiao Y, Kunnath SK. Experimental study on progressive
collapse-resistant behavior of reinforced concrete frame structures. ACI Struct
J 2008;105(4).
 A.H. Arshian, G. Morgenthal / Engineering Structures 132 (2017) 87–97
[17] Yu J, Tan K-H. Experimental and numerical investigation on progressive
collapse resistance of reinforced concrete beam column sub-assemblages. Eng
Struct 2013;55:90–106.
[18] Stinger SM, Orton SL. Experimental evaluation of disproportionate collapse
resistance in reinforced concrete frames. ACI Struct J 2013;110(3).
[19] Xiao Y, Kunnath S, Li F, Zhao Y, Lew H, Bao Y. Collapse test of three-story half-
scale reinforced concrete frame building. ACI Struct J 2015;112(4):429.
[20] Powell G. Progressive collapse: case studies using nonlinear analysis. In:
Structures congress proceedings. ASCE; 2005. p. 1–14.
[21] Marchand K, McKay A, Stevens DJ. Development and application of linear and
non-linear static approaches in ufc 4-023-03. In: Structures congress 2009:
don’t mess with structural engineers: expanding our role. ASCE; 2009. p. 1–10.
[22] McKay A, Marchand K, Diaz M. Alternate path method in progressive collapse
analysis: variation of dynamic and nonlinear load increase factors. Pract Period
Struct Des Construct 2012;17(4):152–60.
[23] Sasani M. Response of a reinforced concrete infilled-frame structure to
removal of two adjacent columns. Eng Struct 2008;30(9):2478–91.
[24] Sasani M, Kazemi A, Sagiroglu S, Forest S. Progressive collapse resistance of an
actual 11-story structure subjected to severe initial damage. J Struct Eng
2011;137(9):893–902.
[25] Brunesi E, Nascimbene R. Extreme response of reinforced concrete buildings
through fiber force-based finite element analysis. Eng Struct 2014;69:206–15.
[26] Arshian AH, Morgenthal G, Narayanan S. Influence of modelling strategies on
uncertainty propagation in the alternate path mechanism of reinforced
concrete framed structures. Eng Struct 2016;110:36–47.
[27] Brunesi E, Nascimbene R, Parisi F, Augenti N. Progressive collapse fragility of
reinforced concrete framed structures through incremental dynamic analysis.
Eng Struct 2015;104:65–79.
[28] FarhangVesali N, Valipour H, Samali B, Foster S. Development of arching action
in longitudinally-restrained reinforced concrete beams. Construct Build Mater
2013;47:7–19.
[29] Neuenhofer A, Filippou FC. Evaluation of nonlinear frame finite-element
models. J Struct Eng 1997;123(7):958–66.
[30] Dat PX, Hai TK. Membrane actions of rc slabs in mitigating progressive collapse
of building structures. Eng Struct 2013;55:107–15.
[31] Keyvani L, Sasani M, Mirzaei Y. Compressive membrane action in progressive
collapse resistance of rc flat plates. Eng Struct 2014;59:554–64.
97
[32] Lu X, Xie L, Guan H, Huang Y, Lu X. A shear wall element for nonlinear seismic
analysis of super-tall buildings using opensees. Finite Elem Anal Des
2015;98:14–25.
[33] Rots J, Nauta P, Kuster G, Blaauwendraad J. Smeared crack approach and
fracture localization in concrete. HERON 1985;30(1).
[34] McKenna F, Fenves G, Scott M, et al. Open system for earthquake engineering
simulation. Berkeley, CA: University of California; 2009.
[35] Scott M, Fenves G. Plastic hinge integration methods for force-based beam
column elements. J Struct Eng 2006;132(2):244–52.
[36] Popovics S. A numerical approach to the complete stress-strain curve of
concrete. Cem Concr Res 1973;3(5):583–99.
[37] Chang G, Mander J. Seismic energy based fatigue damage analysis of bridge
columns: Part 1 – Evaluation of seismic capacity. NCEER technical report No.
NCEER-94-0006. Buffalo, New York: State University of New York; 1994.
[38] Scott MH, Fenves GL. Krylov subspace accelerated newton algorithm:
application to dynamic progressive collapse simulation of frames. J Struct
Eng 2009;136(5):473–80.
[39] Broyden CG. A class of methods for solving nonlinear simultaneous equations.
Math Comput 1965;19(92):577–93.
[40] Hall JF. Problems encountered from the use (or misuse) of rayleigh damping.
Earthq Eng Struct Dynam 2006;35(5):525–45.
[41] Arshian AH, Morgenthal G. Alternate path sensitivity analysis of reinforced
concrete frames using first-order second-moment method. IABSE symposium
report, vol. 103. International Association for Bridge and Structural
Engineering; 2015. p. 170–7.
[42] Jung B, Morgenthal G, Xu D. Integrative sensitivity analysis applied to semi-
integral concrete bridges. J Bridge Eng 2014.
[43] Sasani M, Kropelnicki J. Progressive collapse analysis of an rc structure. Struct
Des Tall Special Build 2008;17(4):757–71.
[44] Gouverneur D, Caspeele R, Taerwe L. Experimental investigation of the load–
displacement behaviour under catenary action in a restrained reinforced
concrete slab strip. Eng Struct 2013;49:1007–16.
[45] Gomes A, Appleton J. Strengthening design of concrete beams by addition of
steel plates. In: Arantes e Oliveira E, Bento J, editors. Proceedings of the
international conference on enhancement and promotion of computational
methods in engineering and science (EPMESC VII). Macau: Elsevier; 1999. p.
657–66.