You are on page 1of 14

Engineering Structures 149 (2017) 147–160

Contents lists available at ScienceDirect

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

Damage assessment for reinforced concrete frames subject to


progressive collapse
Jian Weng a,⇑, Chi King Lee b, Kang Hai Tan a, Namyo Salim Lim a
a
School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, 639798, Singapore
b
School of Engineering and Information Technology, University of New South Wales, Canberra, Northcott Drive, Campbell, ACT 2600, Australia

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

Article history: Due to increasing threats from terrorism, progressive collapse modeling is gaining popularity with the
Available online 13 August 2016 objective of simulating the collapse process of the whole or partial structural system. In order to conduct
an accurate progressive collapse modeling, one needs to identify damaged members and to trace the
Keywords: propagation of damage. Hence, good damage assessment criteria are vital to the modeling and analysis.
Progressive collapse modeling Only with reliable damage assessment criteria, realistic collapse mechanisms of structures can be simu-
Damage assessment lated. It can then provide useful guidance for a better and more economic structural design against pro-
Reinforced concrete frame
gressive collapse. This paper presents a set of damage assessment criteria that can be easily implemented
Axial-shear-flexural interaction
for progressive collapse analysis of reinforced concrete (RC) frames. Flexural, shear and axial damage cri-
teria for RC members are separately proposed incorporating axial-shear-flexural interactions in the anal-
ysis. Three scaled moment-resisting RC frame tests were conducted to validate the proposed flexural and
axial damage criteria. Three shear-dominant damaged tests were also modeled to assess the proposed
shear damage criteria. The results obtained show that the proposed damage assessment criteria are effec-
tive and reliable for progressive collapse analysis of RC frames.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction developed for beam-column finite elements for progressive col-


lapse analysis of RC frames is needed.
Heavy losses of human life and property in progressive collapse So far, many damage assessments for progressive collapse have
events [1–3] have great economic and political consequences in been presented and discussed in the literature, including the
many countries. While there are methodologies [4,5] to enhance assessment of ultimate collapse strength [6], ultimate ductility
structural resistance to progressive collapse of buildings, there is capacity [7,8] as well as energy flow [9] and stability degradation
a general lack in systematic evaluation of structural behavior dur- [10] of building structures. These assessment approaches basically
ing the whole process of progressive collapse. Towards this end, identify the global structural resistance to progressive collapse
one attractive way is to perform numerical modeling of progressive without accounting for the detailed damage propagation within
collapse for full-scale structures, whereby useful guidance can be the structure, especially when it undergoes the progression of col-
obtained to improve structural performance against progressive lapse. Other damage assessments were proposed for structural
collapse. To conduct an accurate progressive collapse modeling, members, based on reduction of cross-sectional stiffness [11] or
one always needs to identify the degree of damage of the structure strain energy dissipation [12]. The former ignores structural soft-
and to trace the propagation of damage during progressive col- ening under large deformations and the latter requires much more
lapse. In addition, since it is not practical to carry out a full analysis computational effort in determination of damage parameters. In
of a building using 3D solid elements, the more cost-effective Refs. [13–15], more damage assessment criteria were developed
approach based on beam-column elements is popular for progres- in different ways for RC structures. However, they were not aimed
sive collapse analysis of RC frames. Hence, there is a need to quan- for progressive collapse analysis, and hence, catenary action as
tify the damage state of structural members of RC frames with well as the ultimate state of collapse under extreme deformations
regard to flexural, shear and axial effects and any combinations has not been considered.
thereof. Towards this end, a set of good damage assessment criteria For RC framed structures, only limited publications [16,17] have
been reported so far on comprehensive damage assessments for
⇑ Corresponding author. progressive collapse analysis. Kaewkulchai [17] proposed a hinge
E-mail address: jweng004@e.ntu.edu.sg (J. Weng). damage parameter to quantify the damage of frame structures in

http://dx.doi.org/10.1016/j.engstruct.2016.07.038
0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.
148 J. Weng et al. / Engineering Structures 149 (2017) 147–160

dynamic progressive collapse. It was assumed that hinge failures at RC frames in progressive collapse without resorting to 3D solid ele-
the member-ends would cause disconnections from adjacent ments which inevitably associate with complex failure criteria, as
members. This led to a conservative assessment for progressive the use of 3D solid elements with failure criteria often leads to
collapse resistance due to the omission of catenary action in complicated model updating requirements and numerical conver-
beams. Accordingly, the prediction for the next phase of collapse gence difficulties when one or more of the elements have failed.
could differ from the real collapse situation. Talaat [16] defined
axial and flexural damage indices of cross-sections for a fiber 2. Flexural, shear and axial damage criteria for RC frames
model, and zero-length elements with a shear model were intro-
duced to assess the extent of shear damage. In his study, the 2.1. Overview of proposed damage criteria
flexural-axial and shear-axial interactions were considered sepa-
rately while flexural-shear interaction was ignored. For simulating the behavior of RC beam/column members, a
In general, the propagation of damage is dependent on the flex- fiber beam element formulation is often preferred to obtain rea-
ural, shear and axial interactions in progressive collapse. Realistic sonable accuracy with acceptable computational cost [19,20] com-
collapse mechanisms of a structure are closely associated with pared to 3D solid formulation. In such formulation, a reasonably
the combined flexural, shear and axial failures within a member detailed strain or stress profile can be described by beam elements
or a substructure. This could be illustrated by the examples in using a fiber model at the section level. Nonlinear behavior at the
Fig. 1. Fig. 1(a) shows that a structure contains plastic hinges but material level is characterized through a fiber discretization of the
is still stable, because complete and continuous boundary cross-section at integration points of each element. Fig. 2 shows a
restraints are maintained to mobilize the catenary action at the typical fiber beam element in an RC member. In this study, based
large deformation stage after undergoing transient instability due on this fiber beam model, flexural, shear and axial damages of an
to the loss of most flexural resistance under the plastic hinge RC member are quantified by introducing appropriate damage
mechanism. In Fig. 1(b), the structure subjected to a combination indices in terms of flexural rigidity, shear and axial strain energy
of flexural, shear and axial failures may experience collapse mech- dissipation of the cross-section. For different kinds of damages,
anisms. Therefore, in order to trace and identify the progression of the effect from flexural/shear/axial coupling is also considered.
collapse, a set of criteria to reasonably assess flexural, shear and As shown in Fig. 3, an overall framework for the proposed damage
axial damages and failures is essential. assessment scheme is summarized, which provides the basis for
This paper aims at establishing such a set of damage assessment identification of element failure and member removal due to
criteria for progressive collapse analysis of RC frame structures. collapse.
Based on a validated fiber-discretized finite beam element formu-
lation [18], flexural, shear and axial damage indices are separately 2.2. Flexural damage assessment
proposed with consideration of flexural, shear and axial coupling
during the loading process. Flexural failure of RC cross-sections is 2.2.1. Flexural damage with axial-flexural interaction
indicated by the complete loss of flexural rigidity and shear/axial Flexural damage of RC frames is quantified by a reduction factor
failures are associated with all material fibers across a cross- for flexural rigidity. Furthermore, the effect from axial-flexural
section attaining their ultimate tensile/compressive strain. Three interaction will also be taken into account. Flexural damage index
RC moment-resisting frames were tested to investigate structural of a cross-section is defined as
behavior in progressive collapse and to verify the proposed flexu-
ral/axial damage criteria. The results show an accurate capture of EIr
Df ¼ 1  ð1Þ
critical damage characteristics (cracking/crushing of concrete, frac- EI
ture/buckling of rebar and final failure/collapse) of RC frames dur- where EI and EIr are the pristine and the residual flexural rigidity of
ing deformation, even at the late stage of catenary action. In the cross-section and are defined in Eqs. (2) and (3), respectively.
addition, three shear-dominant damaged tests were studied to
X X   X
ei  e0 ei
demonstrate the validity of the proposed shear damage criteria. EI ¼ Mi =/i ¼ Ei ei Ai yi ¼ Ei Ai y2i
The main novelty of this work is to develop a systematic dam- i i
y i i
ei  e0
age quantification scheme that could be easily implemented and X Pi y2
¼ i
ð2Þ
applied in progressive collapse analysis of RC frames. The assess-
i
ei  e0
ment criteria proposed separately for flexural, shear and axial
damages are able to predict realistic collapse mechanisms by con-
X ei X P i ð1  Di Þy2
sidering the combined actions of flexural, shear and axial failures, EIr ¼ Ei ð1  Di ÞAi y2i ¼ i
ð3Þ
whereby providing an appropriate means to trace the progression i
ei  e0 i
ei  e0
of collapse. One additional advantage of the proposed approach is
In Eqs. (2) and (3), Pi ¼ Ei Ai ei is the axial force. Mi and /i are the
that such simple and yet effective damage assessment criteria
bending moment and curvature of fiber i of the cross-section;
allow engineers to achieve a satisfactory prediction for the propa-
Ei ; ei and Ai are the elastic modulus, normal strain and area of fiber
gation of different damages (flexural, shear and axial damages) of
i; e0 is the normal strain at the centroid of the cross-section; yi is the

Fig. 1. Typical failure or collapse modes for frame structures.


J. Weng et al. / Engineering Structures 149 (2017) 147–160 149

which are aligned with the direction of principle stress. In this


study, h is set to be 45° for simplicity so that no complicated
analysis to find h is needed. For confined concrete, the variation of
normal strain ecV is determined by force equilibrium as
Es esV As
ecV ¼ ð5Þ
Ec Ac0
where Ec and Ac0 are the elastic modulus and area of confined
concrete.
The flexural damage of fiber i considering shear effect is
obtained by superimposing esVi and ecVi onto the normal strain ei
in Eqs. (2) and (3). Flexural failure of a cross-section occurs when
Fig. 2. Fiber beam element model. its flexural damage index Df attains unity. Since flexural failure
of RC structures depends mainly on concrete crushing while
ductility performance is closely related to tensile yielding of
reinforcement, it is assumed that a complete flexural failure of a
cross-section will take place once all the extreme concrete fibers
in compression crush in that cross-section and all the steel bars
in tension are yielded.

2.3. Shear damage assessment

2.3.1. Bilinear shear model with flexural and axial effects


A bilinear shear model is developed to describe the shear
behavior of RC structures, taking into account of flexural and axial
effects. As shown in Fig. 5, the key parameters V y , K, K h , cy and cu
are needed to be defined. V y is the vertical shear strength of cross-
section; K and K h are the shear rigidity of cross-section before and
after shear yielding; cy and cu are the yield and ultimate shear
strain of cross-section, respectively.
By using a variable angle truss model [21] for RC structures
under shear and considering that shear distortion distributes uni-
Fig. 3. An overall view of damage assessment scheme.
formly along a shear-dominant damaged element, the combined
shear-flexural mechanism in that element can be expressed by
distance from the center of fiber i to the neutral plane of the cross- the relationship between shear force V and shear strain c as
section. In fact, the term Pi ð1  Di Þ in Eq. (3) gives the residual axial follows
force of fiber i. Furthermore, the flexural damage index Di describes V K cs þ K cf
the degradation of tangent modulus of fiber i. For concrete fibers in c¼ ¼ V ð6Þ
K K cs  K cf
compression, Di is given by Di ¼ ðEi  Eri Þ=Ei where Eri represents
the residual tangent modulus, while for concrete fibers in tension, where K cs and K cf are shear rigidity contributions from shear and
Eri is respectively set as the initial elastic modulus and zero before flexural actions, deriving from virtual work principle in accordance
and after cracking. For steel fibers, the damage index is defined as with the variable angle truss model. The expressions [21] are
Di ¼ 0 for damage state before yielding and Di ¼ 1 after yielding. Z 1
Ec Ag cot2 a
K cs ¼ 2 2
dx ð7Þ
2.2.2. Shear effect on flexural damage
0
qsw n þ 2ð1 þ x
1 2  cot2 aÞ þ 2½1 þ ð1  xÞ2  cot2 a
In the uniaxial constitutive laws adopted in the present formu-
lation, shear is independent of bending. But a general shear analy- Es Asl tan2 a
K cf ¼ ð8Þ
sis [22] implied that shear force will increase tensile strain of f
longitudinal reinforcement when shear cracks occur. As shown in
where cot a ¼ a=h which is the shear span to the height of cross-
Fig. 4, additional tensile strain esV of longitudinal steel bars due
section; Ag is the gross area of cross-section; Asl is the area of longi-
to shear can be expressed [22] as
tudinal steel bar; qsw refers to transverse reinforcement ratio;
0:5V cot h n ¼ Es =Ec which is the modulus ratio of steel to concrete;
esV ¼ ð4Þ
E s As
where V is the shear force of cross-section; Es and As are the elastic
modulus and area of steel; h is the inclination angle of shear cracks

Fig. 4. Shear effect on flexural damage. Fig. 5. A bilinear shear model.


150 J. Weng et al. / Engineering Structures 149 (2017) 147–160

f ¼ 0:5704 which is a constant coefficient. Two-point Gaussian inte- 2.4. Axial damage assessment
gration is employed to solve Eq. (7). For a shear-critical damaged
element, yielding of transverse reinforcement should occur after Under large deformation of RC beams in progressive collapse,
peak shear. Hence, the post-yield shear rigidity can be predicted catenary action may be mobilized when sufficient restraints are
by the yielding of transverse reinforcement. Substituting the post- maintained at both beam-ends. In this case, axial damage of
yield (hardening) modulus Eh of transverse reinforcement into Eq. cross-sections dominates the eventual failure of the beam. For RC
(7) will lead to the shear rigidity K cs . As Eh is usually small com- columns, axial failure can be caused by either the fracture of tensile
pared to the pre-yield modulus, the value of K cs is small. Thus, from reinforcement or the buckling of compressive reinforcement. These
Eq. (6) the post-yield shear rigidity would be expressed as two scenarios for RC beams or columns should be considered in
K K axial damage assessment.
K h ¼ K ccssþKcf which is also a small value, as shown in Fig. 5.
cf
The shear strength V y suggested by Sezen et al. [23] is adopted
2.4.1. Axial damage index
as
With fiber-discretized beam element formulation, the axial
0 qffiffiffiffi 1
0 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
damage index of a cross-section is defined by summing up the
f yw Asw d qffiffiffiffi 0:5 f d u
0 B c u P C axial damage of each fiber in the cross-section as
Vy ¼ k f c Ag þ k @ t1 þ qffiffiffiffi A  0:8Ag ð9Þ
s a 0 P
0:5 f A c g Dai Ai
Da ¼ P ð15Þ
Ai
0
where k is defined by displacement ductility; f c is the compressive
strength of concrete; d is the depth of cross-section; a is the shear Pai
Dai ¼ ð16Þ
span; P is the axial load (positive for compression and negative Pamax
for tension); s is the transverse reinforcement spacing; f yw and Asw
where Dai is the axial damage index of fiber i which is determined
are the yield strength and area of transverse reinforcement,
from the axial strain energy dissipation. The term
respectively. Re
Pai ¼ ð 0 i rðeÞdeÞ represents the current axial strain energy norm
1=2
The yield shear strain can then be calculated from
Re
of fiber i; ei is the normal strain of fiber i; Pamax ¼ ð 0 u rðeÞdeÞ is
1=2

Vy the maximum axial strain energy norm of the fiber. For concrete, eu
cy ¼ ð10Þ
K refers to the cracking strain and ultimate compressive strain when
it is subjected to tension and compression, respectively; for steel
and the ultimate shear strain cu is obtained by [24]
bars, eu is the fracture strain in tension, or the buckling strain
! ebuckling in compression. Note that Eqs. (15) and (16) are aimed to
my
cu ¼ 4  12 0  cy ð11Þ reflect the axial damage at the stage both before and after the cate-
fc nary action. In general, for a beam member, before the catenary
action takes place (i.e. in the initial flexural stage and the compres-
where coefficient my is given as [25] sive arch action stage) axial resistance is mainly due to concrete
qffiffiffiffi compression, so Eqs. (15) and (16) will be able to represent the
my 0
¼ 0:25 f c þ qsw  f yw ð12Þ degree of axial damage of the beam. If eventually the beam is under
catenary action, some sections of the beam must have undergone
From this model, it can be found that the flexural effect on shear severe flexural/shear damage. However, at other sections where
behavior of the element is incorporated into the shear rigidity K flexural/shear damage is not severe, the actual damage should be
with K cf , while the yield shear strain cy varies with shear strength mainly linked to steel rebar damage only. In this case, since con-
V y which is affected by axial load P. crete fiber will contribute most area of the cross section, the dam-
age index will be conservative for those sections. In general, when
assessing the overall damage of the beam as a whole, Eqs. (15)
2.3.2. Shear damage index
and (16) are still acceptable in practice and could truly reflect the
Based on the above shear model, shear damage index Ds is
overall damage of the whole beam while they may overestimate
introduced to assess the damage growth of cross-section in terms
the damage in some sections.
of shear strain energy dissipation as
Excessive compression in an RC member may cause buckling of
Ps reinforcement determined by spacing between stirrups and it in
Ds ¼ ð13Þ turns marks the deformation limit of the member. The buckling
Psmax
strain of reinforcement can be determined based on the equilib-
R c 1=2 rium of bending moment of reinforcement cross-section [26] as
where Ps ¼ ð sðcÞdcÞ is the current shear strain energy norm
0
R cu
4p2 I
2
and Psmax ¼ ð 0 sðcÞdcÞ
1=2
is the maximum shear strain energy bkL
ebuckling ¼ þ 6 ey ð17Þ
norm. The current shear strain c  is determined by calculating 2
L A EA
weighted average for the fiber shear strain using the following
where ey is the yield strain of longitudinal reinforcement; E, A and I
equation.
are the elastic modulus, the cross-sectional area and the second
P moment of longitudinal reinforcement, respectively; b is a constant
cA
c ¼ P i i ð14Þ and is equal to 0.0875; the terms k and L are given by k ¼ K=s and
Ai
L ¼ ð4p2 Eh I=bkÞ where Eh is the hardening modulus of reinforce-
1=4

where ci and Ai are respectively the shear strain and area of fiber i. ment, K and s are the effective axial stiffness and spacing of trans-
The term Ds is set as unity once c
 is greater than cu when shear fail- verse reinforcement.
ure occurs. Considering the irreversibility of shear damage, it is For RC structures, due to confinement of surrounding concrete,
assumed that unloading of shear forces during load redistribution it is assumed that the reinforcement will not buckle before con-
process will not reduce shear damage. crete crushing. Hence, the buckling strain ebuckling will be given by
J. Weng et al. / Engineering Structures 149 (2017) 147–160 151

maxfecu ; ebuckling g, where ecu is the ultimate compressive strain of model. Towards this end, it is assumed that axial failure will have
concrete around the reinforcement. occurred in a member if all its top and bottom rows of rebars have
fractured, regardless whether the fracture is at the same cross-
2.4.2. Axial failure criteria section or not. After such an RC member is identified, the cross-
It should be mentioned that based on the axial damage criteria section with the maximum axial damage index is considered to
given in Section 2.4.1, axial failure of an RC member should only have failed axially. The key steps to identify such axial failure mode
occur when the axial damage index attaining unity (i.e. Da = 1 are listed below.
which indicates that all the concrete and the steel fibers have failed
at that cross-section) at one cross-section of the member. How- (1) At each load step, all the axial damage indices at the integra-
ever, it has been observed in many practical cases that axial failure, tion cross-sections of the elements for each member are
or almost complete loss of stiffness, of an RC beam/column could tracked to check if all the top rebars or the bottom rebars
occur when the axial damage indices at two or more different have fractured.
cross-sections in the member are close to but still lower than 1 (2) If it is found that for a given member both the top rebars and
(i.e. Da < 1). A common example of such case is shown in Fig. 6. the bottom rebars have fractured (the fractured sections can
In Fig. 6, the RC beam has practically failed due to the complete be different), the member is then considered to fail axially.
fracture of tensile rebars in the member at 3 different cross- The failed cross-section of the member is at where the axial
sections, but the axial damage indices at these locations have not damage index is the maximum.
reached (but close to) unity as the fracture of reinforcement occurs
at different cross-sections. Such phenomenon can often be 3. Validation of the proposed damage assessment criteria
observed at the last stage of catenary action of RC beams under a
missing column scenario, when all the tensile rebars at both the In order to validate the proposed flexural and axial damage
mid-span (bottom rebars) and the beam-end (top rebars) have assessment criteria, an experimental study with three scaled RC
fractured by axial tension. In this case, the stiffness of the member frames was conducted under a middle column removal scenario
will only be a very small fraction of the stiffness of the undamaged and compared with numerical models. Since the specimens were
member and the member can be considered to have failed com- specially designed as moment-resisting frames, shear damage
pletely in practice. In addition, such conditions may also cause was not evidently observed during the tests. Therefore, three
numerical instability or convergence problems during progressive shear-dominant damaged tests from the literature [27] were mod-
collapse analysis. Hence, to ensure the stability of the numerical eled to verify the proposed shear damage criteria.
model and to avoid non-conservative prediction (very weak mem-
bers hanging with very little stiffness), it is necessary to identify 3.1. RC moment-resisting frames under a middle column removal
such axial failure case in an RC member and remove it from the scenario

3.1.1. Experimental program


The RC moment-resisting frame was designed based on EC2
[28], with dead loads and imposed loads equal to 5.0 kN/m2 and
7.1 kN/m2, respectively. To investigate the structural mechanism
of the frame under progressive collapse, beam-column sub-
assemblages for the middle-bay and the end-bay were extracted
at the locations of the contra-flexural points of bending moments
as shown in Fig. 7. The test program studied the influence of three
types of boundary effect on RC sub-assemblages with middle col-
umn removal scenario.
Accordingly, three one-third scaled specimens were designed
and tested under a middle column removal scenario. Full Restraint
Fig. 6. Axial failure due to rebar fracture. (FR) and Full Restraint-Seismic (FR-S) specimens from the middle-

Fig. 7. Locations of test specimens in the prototype structures.


152 J. Weng et al. / Engineering Structures 149 (2017) 147–160

bay frame were designed with normal and seismic detailing to tional angles of the beam could be determined by linear interpola-
investigate the effect of closer stirrup spacing on progressive col- tion [29]. Thus, equilibrium relationships of the frame as a free
lapse resistance. Partial Restraint (PR) specimen was extracted body could be obtained.
from an end-bay frame and designed with normal detailing. Rein-
forcement details of the three specimens are shown in Fig. 8. For
3.1.2. Numerical modeling of test specimens
concrete material, the compressive strength is f c ¼ 30 MPa and
pffiffiffiffi Fiber-based finite element models adopting the proposed dam-
the tensile strength is calculated as f t ¼ 0:56 f c [25]. For steel age assessment criteria were employed to simulate the failure of
material, the yield and ultimate strength used are f sy ¼ 505 MPa all three specimens. The modified Kent-Park model was adopted
and f su ¼ 605 MPa, respectively. to provide tensile and compressive constitutive laws for concrete
Fig. 9 shows the test setup. To simulate gravity loads from the fibers, where the hysteretic behavior was predicted based on the
above stories of the extracted frame, constant axial forces were rules by Spacone et al. [31]. In addition, a bilinear elastic plastic
applied on top of the two side columns. To investigate the behavior strain-hardening model [32] was employed for steel in both ten-
of the frame during progressive collapse when the middle column sion and compression. Fig. 11(a) shows the numerical model for
has been removed, an actuator was applied at the top of the middle Specimens FR and FR-S and in Fig. 11(b) Spring 4 was removed
joint until the specimen completely failed [29]. Horizontal steel for the Specimen PR in order to simulate the effect of an end bay.
rods at both sides of the frame provided axial restraints for the 3-node Total Lagrangian (TL) Timoshenko beam element formula-
FR and FR-S specimens (Fig. 9(a)) so that the full restraint condi- tion [18] was employed to capture the nonlinear response of RC
tions of an internal frame (Fig. 7(a)) are reproduced. Tension- frames under large deformations. Beam-column elements of differ-
compression load cells were installed to monitor horizontal reac- ent lengths were employed in different parts of the structure in
tions during the test. In addition, load pins were installed at the accordance with longitudinal reinforcement ratio and stirrup spac-
base of both columns to measure support reactions. Note that for ing. Horizontal restraints from four tension-compression load cells
the PR specimen, one horizontal support was removed (Fig. 9(b)) were modeled by four bilinear springs. The properties of each
so that the boundary condition effect of the end-bay frame (Fig. 7 spring in Table 1 were determined from the relationship of reac-
(b)) was reproduced accordingly. tion forces and displacements measured by each load cell. To
The instrumentation scheme shown in Fig. 10 was used to cal- model the steel plate connections between the top of the column
culate internal forces of cross-section in the specimens. Six linear and the horizontal load cell, two pure steel elements were inserted
variable differential transformers (LV1 to LV6) were arranged to as Elements 23 and 24 (Elements 22 and 23 for Specimen PR), as
measure vertical displacements along the beam, upon which, rota- shown in Fig. 11. A concentrated load was applied at the mid-

Fig. 8. Reinforcement details of test specimens.


J. Weng et al. / Engineering Structures 149 (2017) 147–160 153

Fig. 9. Support and boundary conditions of test setup.

Fig. 10. Layout of instrumentation.

Fig. 11. Finite element models for Specimen (a) FR/FR-S; and (b) PR.

span between elements 15 and 16 until the frame failed com- of both columns to simulate gravity loads from above stories and
0
pletely. Prior to this load, two point loads of 0:3f c Ag (i.e. the axial their values are kept constant during the whole test [29]. Under
compression ratio is assumed to be 0.3) were applied at the top extreme loading, generalized displacement control method [30]
154 J. Weng et al. / Engineering Structures 149 (2017) 147–160

Table 1 3.1.3. Validation of flexural and axial damage criteria


Bilinear spring properties. Before verifying the proposed damage assessment criteria, gen-
Specimen Spring Property Legend eral results from numerical analysis are compared with those from
K1 (N/mm) Gap (mm) K2 (N/mm) experimental study to show the accuracy and validity of the pre-
FR 1 636 2.4 18,355 sent modeling. Note that reasonably accurate but not perfect pre-
2 244 3.1 27,677 dictions of the test results are expected. As in any numerical
3 337 3.7 10,566 model, it is impossible to consider all factors that may affect the
4 455 2.6 85,010
FR-S 1 2715 2.2 20,641
modeling accuracy. For example, in the present numerical model,
2 716 4.5 23,667 factors such as splicing of reinforcement and discontinuity of
3 160 7.8 101,894 cracking are not considered. However, it is expected that if a rea-
4 2073 3.9 4080 sonably accurate numerical model is available, a reliable and effec-
PR 1 1359 2.9 27,860
tive damage assessment scheme should be able to give useful
2 396 1.9 5917
3 2695 1.7 27,606 indication to the damage status of structural members of the RC
frames.

3.1.3.1. Comparison between numerical analysis and experimental


with superior convergence performance through limit points and results. Fig. 12 shows the predicted and measured applied forces
snap-back points was applied to ensure numerical stability in the and displacements relationships at the mid-span for Specimens
large deformation analysis. FR, FR-S and PR. In general, good agreement can be found between
As severe damages were expected to occur at both the mid-span the two sets of results in terms of peak resistance of test frames at
and the two beam-ends, four Gauss-point cross-sections were both the flexural and catenary action phases. It demonstrates the
selected to trace the crucial damage propagation as shown in validity of the proposed damage assessment criteria.
Fig. 11, in which, ‘L1’ and ‘L2’ represent the cross-sections near As shown in Fig. 12(a) and (b), the first sudden drop of the
the beam-end and the mid-span of the left-side beam; ‘R1’ and applied force for Specimens FR and FR-S was caused by the fracture
‘R2’ are the cross-sections near the beam-end and the mid-span of bottom steel bars at the mid-span. Such occurrences only result
of the right-side beam. in minor deviation between the numerical models and actual

Fig. 12. Relationships of the applied force and displacement for Specimen (a) FR; (b) FR-S; and (c) PR.
J. Weng et al. / Engineering Structures 149 (2017) 147–160 155

experiments. For the second drop, it is found that consistent frac- as can be seen in the actual tests. Nevertheless, the top rebar frac-
ture of the top bars for Specimen FR occurred at the mid-span, ture of both cases implies the eventual failure of the structure due
but there are differences in the fracture locations of the top bars to axial tension at catenary action stage. As the difference only sur-
of Specimen FR-S for the experiment and the model. The actual faced at the last stage, it did not affect the accuracy of predictions.
fracture occurred near Section R1 while the model predicted its The behavior of Specimen PR with one less horizontal restraint is
occurrence at Element 14. This difference may result from the different from FR and FR-S. Catenary action could not further
finite element assumption of continuous displacement field which develop after the fracture of bottom bars at the mid-span. Numer-
could lead to deviations in predictions of localized fracture of rein- ical model agrees well with the test results for Specimen PR, as
forcement where large cracking and spalling of concrete take place, shown in Fig. 12(c).

Fig. 13. Relationships of bending moment/flexural damage index Df to displacement.


156 J. Weng et al. / Engineering Structures 149 (2017) 147–160

In summary, although there are some differences in the predic- Sections ‘R1’ and ‘R2’ are for Specimens FR and PR. The relation-
tion of the location of rebar fracture forwards the end stage, load- ships between the bending moment/flexural damage index and
displacement curves for both flexural and catenary action stages of the mid-span displacement for the test frames are shown in
each specimen are in good agreement with test results. Hence, it Fig. 13. From the moment-displacement relationships of FR and
can be concluded that the numerical model can provide good FR-S, it can be seen that consistently good results were predicted
inputs for the validation of the proposed damage assessment by the two numerical models, showing effective and reliable pre-
criteria. dictions for flexural damage, especially for the yield displacement
and the peak flexural resistance. A slightly larger error of predicted
3.1.3.2. Flexural damage identification. The beam-end and mid-span peak bending moment is found in Specimen PR. This is probably
cross-sections of the more severely damaged beam for each speci- due to non-symmetric boundary conditions.
men are chosen to demonstrate the proposed damage assessment The flexural damage index Df is used to identify the flexural
criteria. Sections ‘L1’ and ‘L2’ are selected for Specimen FR-S and damage state of beam sections, as labeled in Fig. 13. The first sud-

Fig. 14. Relationships of axial force/axial damage index Da to displacement.


J. Weng et al. / Engineering Structures 149 (2017) 147–160 157

den increase of Df reflects the cracking of extreme concrete fibers for the tested frames. For Specimen FR and FR-S, consistent trends
in tension. The tensile stiffness of concrete fibers reached zero of axial force resulting from tests and numerical models can be
when cracks occurred, leading to a sharp reduction of cross- seen, while relatively larger difference for PR is observed in
sectional flexural stiffness. The second sudden increase of Df was Fig. 14(e) and (f). The reason is that the column joint near Sec-
caused by rebar yielding, which fits well with the yield point tion R1 was severely damaged during the test due to the non-
obtained from test results. Once the rebar has yielded, the value symmetric boundary conditions, but in the numerical model, no
of Df shot up from about 0.4 to 0.8 or even beyond. This implies special joint elements were employed to capture joint damage
a close correlation between rebar yielding and flexural damage effects. Fig. 15 shows the eventual damage configurations of the
mechanism. After that, Df attains unity and this corresponds to joint near Section R1 of the three tests conducted. It can be seen
eventual crushing of extreme concrete fibers which marks the flex- that more severe damages occurred for Specimen PR than Speci-
ural failure of the cross-section. Then a plastic hinge forms at the mens FR and FR-S due to non-symmetric test setup of PR. In
corresponding cross-section reflected by the constant value of essence, when damages propagate in a joint, their effects on lap
Df ¼ 1. Therefore, the proposed flexural damage criterion is able length between rebars and concrete as well as dowel action inside
to predict well the yielding of rebars and the cracking and crushing the joint may lead to relatively inaccurate prediction of axial forces
of concrete, providing an effective trace of flexural damage propa- near the joint. Hence, developing effective joint elements to prop-
gation. In addition, flexural failure after rebar yielding will be more erly predict the behavior of RC joints during progressive collapse is
reasonable than the conventional prediction for plastic hinges an important research topic, but it is outside the scope of the pre-
using rebar yielding, which mostly occurred before attaining the sent study. Suffice to say, the proposed axial failure criteria indi-
peak flexural resistance, as shown in Fig. 13. cated by fiber elements attaining the ultimate strain of rebars
provide reliable predictions for the ultimate state of collapse.
Similar to flexural damage curve, two sharp growths of axial
3.1.3.3. Axial damage identification. Fig. 14 shows the relationships damage index Da can be found in each axial damage curve
of the axial force/axial damage index to the mid-span displacement (Fig. 14). The first sudden increase corresponds to the cracking of

Fig. 15. Damage patterns of the joint near Section R1 for Specimen (a) FR; (b) FR-S; and (c) PR.

Fig. 16. The ultimate state of collapse for Specimens.


158 J. Weng et al. / Engineering Structures 149 (2017) 147–160

concrete in tension which leads to axial tensile failure of concrete (Section 2.4.2), Section L2 is predicted to attain axial failure with
fibers. The second jump is due to gradual yielding of steel bars, a greater axial damage index. Compared with the test results, the
which results in a rapid growth of fiber strain. For each plot in ultimate failure of Specimen FR-S occurred at mid-span with
Fig. 14, it is found that the value of axial damage index at rebar almost complete fracture of reinforcement over the cross-section,
yielding (around 0.4) is consistently smaller than the flexural dam- although the section has not been severed completely as shown
age index at yielding (around 0.8), which implies that the ultimate in Fig. 16(b). It indicates good agreement of axial failure between
capacity of a cross-section is eventually dominated by axial the experiment and the simulation for Specimen FR-S. Generally,
damage. it has been observed from Fig. 16 that good predictions on the ulti-
As shown in Fig. 14(b), the predicted axial failure occurred at mate state of collapse are achieved for both specimens. For Speci-
Section R2 for Specimen FR when all rebars have fractured at this men PR, the maximum axial damage index attained is less than 1.0
section. Axial failure of Specimen FR is marked by the entire failure since the test was terminated at a nearly buckling state of the right
of the mid-span section, as shown in Fig. 16(a). This demonstrates column due to safety concern and displacement constraint of the
a good prediction of axial failure for Specimen FR based on present test rig. In addition, it should also be noted that similar trends
axial failure criteria. Fig. 14(d) shows that the axial failure of Spec- for the variation of axial damage index can be found when com-
imen FR-S occurred at Section L2 with a sudden jump of Da to pared with Specimen FR and FR-S.
unity. This is because the axial failure took place when all the steel In short, the proposed axial damage criteria are able to trace
bars fractured in the left beam while the fractured sections (Sec- axial damage propagation of RC members beyond the stage of cate-
tion in Element 14 and Section L2) are different for the top and bot- nary action until the eventual axial failure of the structure. The
tom bars. According to the proposed axial failure criteria yielding and complete fracture of rebars, as well as the final col-

Table 2
Geometrical and material properties for shear tests.

Specimen Legend Section (mm2) L (mm) Strength (MPa) Reinforcement


Concrete Longitudinal bar Transverse bar Longitudinal Transverse
SBV1 300  300 570 39.4 566.0 514.0 12;12 2;6@80
SBV2 660
SBV3 750

Fig. 17. Shear damage assessment for Specimen (a) SBV1; (b) SBV2; and (c) SBV3.
J. Weng et al. / Engineering Structures 149 (2017) 147–160 159

lapse status can be identified based on the present axial assess- uate the initiation of plastic hinges. The axial damage criteria are
ment procedure. This gives useful indication to the damage state capable of assessing the ultimate capacity of RC members after
of RC members after flexural action and provides reasonable evi- attaining flexural failure while the sub-assemblages were undergo-
dence to identify collapse mechanisms under very large ing catenary action. Furthermore, a failure criterion to identify
deformations. realistic axial failure of RC members with rebar fracture is pro-
posed and it is capable of predicting realistic collapse modes. For
shear damage analysis, three shear-dominant damage tests were
3.2. Validation of shear damage criteria
simulated. The results obtained show that the proposed shear
damage criteria are able to simulate shear damage and predict
Since the three specimens in Section 3.1 were designed as
shear failure of RC structures with acceptable accuracy.
moment-resisting frames, shear damage is not critical and shear
As the proposed damage assessment scheme can effectively
failure is avoided. Therefore, in this section, three shear tests
identify the flexural, shear and axial failures of RC structures and
SBV1, SBV2 and SBV3 from the literature [27] were modeled to
can be easily implemented with a cost-effective fiber beam-
demonstrate the validity of the proposed shear damage criteria.
column finite element formulation, it should provide a good basis
All the specimens were designed as cantilever short column struc-
to identify collapse mechanisms in progressive collapse modeling.
tures for investigation of shear-dominant damage during the tests.
Towards this end, one possible immediate extension of the present
Geometrical and material properties for each specimen are given in
work is to combine proposed damage assessment scheme with an
Table 2. Three equal-length finite beam elements, which were cor-
effective algorithm to trace the progression of collapse for RC
responding to the minimum numbers of element needed to model
frames so that an automatic and practical progressive collapse pre-
the test results with reasonable accuracy, were employed along
diction tool could be developed.
each column. A horizontal load was applied at the top of each spec-
imen until final failure of the structure.
Fig. 17(a)–(c) shows the shear force-drift ratio curves for the
References
tested structures from numerical analysis and test results. Consis-
tent peak shear force for each specimen can be found between [1] Seffen KA. Progressive collapse of the World Trade Center: simple analysis. J
experiment and simulation, showing good predictions for shear Eng Mech 2008;134:125–32.
resistance of the specimens. The predicted shear yielding occurred [2] Pearson C, Delatte N. Ronan point apartment tower collapse and its effect on
building codes. J Perform Constr Facil 2005;19:172–7.
in a strongly nonlinear shear state before attaining peak shear. This [3] Georgakopoulos PJ. An overview of progressive collapse in structural systems
implies an acceptable degree of accuracy of the present modeling. MEng Thesis. Massachusetts Institute of Technology; 2005.
As seen in Fig. 17, the proposed bilinear shear model has provided [4] Department of Defence (DOD). Design of buildings to resist progressive
collapse. Unified Facilities Criteria (UFC) 4-023-03; 2010.
reasonable predictions of shear behavior for each specimen, which [5] General Services Administration (GSA). Progressive collapse analysis and
gives good inputs for shear damage assessment. design guidelines for new federal office buildings and major modernization
The relationships between the shear damage index Ds and the projects; 2003.
[6] Underwood JM, Sobey AJ, Blake JIR, Ajit Shenoi R. Ultimate collapse strength
drift ratio for the structures are also shown in Fig. 17. The value assessment of damaged steel-plated structures. Eng Struct 2012;38:1–10.
of Ds hovers around 0.7 for each test when shear yielding occurs [7] Dat PX, Tan KH, Yu J. A simplified approach to assess progressive collapse
and it continues to unity upon reaching the ultimate shear strain. resistance of reinforced concrete framed structures. Eng Struct
2015;101:45–57.
Adopting a bilinear shear model, shear strain before yielding is lin- [8] Izzuddin BA, Vlassis AG, Elghazouli AY, Nethercot DA. Progressive collapse of
early proportional to shear force. Hence, shear damage index multi-storey buildings due to sudden column loss — Part I: Simplified
defined by the concept of shear strain energy norm is increased assessment framework. Eng Struct 2008;30:1308–18.
[9] Szyniszewski S, Krauthammer T. Energy flow in progressive collapse of steel
in proportion. Shear failure from modeling has been underesti-
framed buildings. Eng Struct 2012;42:142–53.
mated for all three specimens in terms of shear ductility calculated [10] Schafer BW, Bajpai P. Stability degradation and redundancy in damaged
from Eq. (11), indicating conservative predictions of shear failure structures. Eng Struct 2005;27:1642–51.
for these tests based on present shear damage criteria. However, [11] Sanches F, Venturini WS. Damage modeling of reinforced concrete beams. Adv
Eng Softw 2007;38:538–46.
considering brittle property of shear, shear failure would usually [12] Saetta A, Scotta R, Vitaliani R. Coupled environmental-mechanical damage
occur soon after shear yielding. Therefore, the proposed damage model of RC structures. J Eng Mech 1999;125:930–40.
criteria give reasonable predictions of shear failure of these tested [13] Bui QB, Mommessin M, Perrotin P, Plassiard JP, Ple O. Assessing local-scale
damage in reinforced concrete frame structures using dynamic measurements.
specimens. Eng Struct 2014;79:22–31.
[14] Wahab MMA, De Roeck G, Peeters B. Parameterization of damage in reinforced
concrete structures using model updating. J Sound Vib 1999;228:717–30.
4. Conclusions [15] Cipollina A, Lopezinojosa A, Florezlopez J. A simplified damage mechanics
approach to nonlinear analysis of frames. Comput Struct 1995;54:1113–26.
[16] Talaat M. Computational modeling of progressive collapse in reinforced
A set of damage assessment criteria to separately monitor flex- concrete frame structures PhD Thesis. Berkeley: University of California; 2007.
ural, shear and axial damage propagations is proposed for progres- [17] Kaewkulchai G. Dynamic progressive collapse of frame structures PhD
sive collapse modeling of RC frames. The proposed damage criteria Thesis. Taxus: University of Austin; 2003.
[18] Dvorkin EN, Onte E, Oliver J. On a non-linear formulation for curved
consider the damage effects from axial-shear-flexural coupling in Timoshenko beam elements considering large displacement/rotation
progressive collapse. In addition, a numerical study was conducted increments. Int J Numer Meth Eng 1988;26:1597–613.
and compared with the test results obtained on three scaled RC [19] Miha J, Boštjan B, Adnan I. Failure analysis of reinforced concrete frames by
beam finite element that combines damage, plasticity and embedded
frames. Comparisons of test results with models show reasonable discontinuity. Eng Struct 2014;75:507–27.
agreement. Clearly, provided that reasonable modeling results [20] de Felice G. Assessment of the load-carrying capacity of multi-span masonry
are available, the proposed damage criteria are able to capture arch bridges using fibre beam elements. Eng Struct 2009;31:1634–47.
[21] Kim JH, Mander JB. Truss modeling of reinforced concrete shear-flexure
damage characteristics of the tested moment-resisting frames
behavior MCEER-99-0005; 1999.
and specimens with dominant shear behavior. The proposed flexu- [22] Collins MP, Mitchell D, Adebar P, Vecchio FJ. A general shear design method.
ral damage criteria predict well the critical damage characteristics ACI Struct J 1996;93:36–45.
(cracking and crushing of concrete, yielding of rebar etc.) of RC [23] Sezen H. Shear deformation model for reinforced concrete columns. Struct Eng
Mech 2008;28:39–52.
frames at the stage of flexural action. Moreover, flexural failure is [24] Gerin M, Adebar P. Accounting for shear in seismic analysis of concrete
identified after rebar yielding. This provides a better means to eval- structures. In: Proceeding of the 13th world conf. on earthquake engineering.
160 J. Weng et al. / Engineering Structures 149 (2017) 147–160

[25] ACI318-05. Building code requirements for structural concrete. American (Fédération internationale du béton/International Federation for Structural
Concrete Institute; 2005. Concrete) Symposium, Copenhagen, Denmark; 2015.
[26] Pantazopoulou SJ. Detailing for reinforcement stability in RC members. J Struct [30] Yang YB, Shieh MS. Solution method for nonlinear problems with multiple
Eng 1998;124:623–32. critical-points. AIAA J 1990;28:2110–6.
[27] Martinelli L. Modeling shear-flexure interaction in reinforced concrete [31] Spacone E, Filippou FC, Taucer FF. Fiber beam-column model for Nonlinear
elements subjected to cyclic lateral loading. ACI Struct J 2008;105:675–84. analysis of RC frames: Part II: applications. Earthq Eng Struct D 1996;25
[28] Eurocode 2: Design of concrete structures EN 1992-1-1; 2004. (7):727–42.
[29] Lim NS, Lee CK, Tan KH. Experimental study on 2-D RC frame with middle [32] Owen D, Hinton E. Finite element in plasticity, theory and
column removed under progressive collapse. In: Proceeding of the Fib practice. Swansea: Pineridge Press; 1980.

You might also like