Engineering Structures 150 (2017) 520–536

Contents lists available at ScienceDirect

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

Investigations of tensile membrane action in beam-slab systems under

progressive collapse subject to different loading configurations and
boundary conditions
Anh Tuan Pham, Namyo Salim Lim, Kang Hai Tan ⇑
School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, 639798, Singapore

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

Article history: Although catenary action in beams and tensile membrane action in slabs are generally believed as high-
Received 16 May 2017 level analyses to mitigate progressive collapse in a reinforced concrete building, previous research studies
Revised 14 July 2017 did not clearly differentiate the contributions of the two mechanisms in combined three-dimensional
Accepted 16 July 2017
beam-slab systems. Besides, most of the recent experimental studies on column removal scenarios
focused on point load application as it is more difficult to apply the more realistic uniform distributed
loads in the laboratory. In this paper, numerical analyses were first employed to investigate the combined
Keywords:
effects of beams and slabs under both point load (idealised) and uniform distributed load (more realistic)
Progressive collapse
Catenary action
conditions. The results show that differences between these two loading methods not only affect overall
Tensile membrane action structural capacities, but also influence vertical deflections and failure modes. It is also observed that ten-
Concentrated load sile membrane action in slabs was less sensitive to boundary conditions compared to catenary action in
Uniform distributed load beams. Moreover, catenary action in beams which showed limited development in beam-slab structures
can be conservatively neglected. Besides, under uniform distributed loading condition, scenarios with dif-
ferent locations of column removal were numerically investigated, showing that the loss of a penultimate
column, rather than a corner column, could be the most critical case, contrary to conventional wisdom. In
fact, in a corner-column removal scenario, tensile membrane action can still be partially mobilised owing
to the presence of two stiff discontinuous edge beams. This phenomenon was well observed in a beam-
slab test conducted in this study.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction contribution of the slab system to overall structural resistance is

substantial. The neglect of slab behaviour when assessing progres-
Progressive collapse is defined as an accidental or man-made sive collapse potential can lead to an uneconomical design of beam
event starting from a local damage of a supporting component and column sections. Moreover, besides catenary action (CA) in
which may lead to disproportionate collapse of a large part or an beams, tensile membrane action (TMA) in slabs is considered as
entire building structure. From the last few decades, although the an upper-bound mechanism to enhance structural capacity at large
number of such events is very limited, once a progressive collapse deformation stage [3]. While the development of CA is normally
incident occurs, the consequences in terms of loss of human lives denoted by the change of axial force in the beam from compression
are dire and grave. After the collapse of the Ronan Point building to tension [4], mobilisation of TMA is indicated by the formation of
(1968), the concept of single column removal was developed and a tensile net at the middle of the slab and a peripheral compressive
is now generally accepted by the engineering community for pro- ring arching the tensile net [5,6].
gressive collapse analysis of buildings [1,2]. In a reinforced con- Regarding experimental investigations under quasi-static
crete (RC) structure, although beam-column frames have loading conditions, several tests have been conducted for RC struc-
significant importance in resisting the sudden column loss event, tures employing the single column loss assumption. Two types of
system are generally considered, including two-dimensional (2D)
beam-column structures [4,7–11] and three-dimensional (3D)
⇑ Corresponding author. beam-slab structures [12–15]. Studies on 2D skeletal frames
E-mail addresses: [email protected] (A.T. Pham), [email protected] demonstrate the contribution of compressive arch action (CAA)
(N.S. Lim), [email protected] (K.H. Tan).

http://dx.doi.org/10.1016/j.engstruct.2017.07.060
0141-0296/Ó 2017 Elsevier Ltd. All rights reserved.
 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536 521

and CA in mitigating collapse. However, most of these tests were paper, numerical models were developed and validated by one pla-
conducted by applying a point load at the middle joint following nar 2D beam-only test (P2), one 3D beam-only test (T2), and one
a displacement-controlled manner. Hence, the development of beam-slab structural test (S2). These specimens (P2, T2, and S2
CA under distributed loads, even though such loading condition from [13]) had the same centre-to-centre span of 1.5 m and the
is closer to reality, has not yet been conducted for any 2D struc- same beam section of 140x80 mm2. All the beams had continuous
tures. Compared to 2D structures, laboratory tests on 3D slab and longitudinal reinforcement consisting of top and bottom rebars.
beam-slab systems apply both concentrated loading (CL) Bottom reinforcement of the slabs was continuous throughout
[12,13,16] and uniform distributed loading (UDL) [14,15,17] meth- the slab in two directions. On the other hand, top reinforcement
ods. In most 3D tests, distributed loads on slabs are represented by was only provided at the edge regions. Fig. 1 illustrates the test
a multi-point loading system [6]. The 3D beam-slab tests, whether setup for T2 and S2 specimens. In the beam-slab tests from [13],
conducted under UDL or CL condition, showed enhancement in boundary conditions were simulated by connecting concrete col-
structural resistance beyond predicted yield-line capacity. How- umn stubs to steel circular hollow sections, which in turn were
ever, such studies did not clearly delineate the respective contribu- fixed to the strong floor. Removal of the middle column was simu-
tions of CA and TMA to overall structural resistance. Moreover, lated by gradually increasing the displacement of the middle joint
experimental investigations on the sensitivity of TMA to boundary using a hydraulic jack.
restraint conditions of slabs are constrained by costs and
laboratory space, and have not been comprehensively studied. In
addition, when assessing the vulnerability of a structure under 2.2. Numerical model
progressive collapse threats, the engineer is required to consider
various scenarios of column removal [1,2]. Among all the cases In this study, an explicit finite element software LS-Dyna [19]
considered, loss of a corner column is generally believed to be was used to simulate the RC member tests of P2, T2 and S2. Con-
the most critical scenario due to a lack of restraint from two adja- crete was simulated using 8-node solid elements with reduced
cent sides of the corner slab, as shown in the CL static tests of Qian integration scheme. Reinforcing bars were explicitly modelled by
and Li [12]. In their work, CA and TMA are conservatively neglected the 2-node Hughes-Liu beam element with 2
2 Gauss quadrature
for corner-column loss and only flexural mechanism is considered integration. Continuous surface cap model MAT_159 was
for both beam and slab members. Nonetheless, this paper shows employed to simulate the behaviour of concrete material. Although
that under UDL condition, corner column removal may not be element erosion is not a physical phenomenon for concrete mate-
the most critical case; instead, penultimate column removal may rial, this attribute allows modelling of spalling and separation of
be the governing scenarios. This finding agrees with a study on concrete under extremely high tensile force. In this study, the cri-
steel and composite grillage frames (without slabs) conducted by terion for erosion of elements was based on the maximum princi-
Stylianidis et al. [18]. pal strain value of 0.1. It was shown from a previous study [20] that
This study presents a numerical and experimental investigation MAT_159 model, together with the application of erosion criterion
on the effects of TMA in beam-slab systems under different bound- of maximum principal strain, can efficiently simulate actual
ary and loading conditions. In Section 2, simulations based on responses as well as damage modes of RC structures under both
detailed finite element method (FEM) were employed and quasi-static and blast conditions. MAT_159 can effectively capture
validated by a quasi-static test series, which included both the post-peak softening, shear dilation, confinement effect, and strain-
sub-assemblages (2D and 3D beam-only) and the 3D beam-slab rate hardening. An isotropic elastic–plastic material model ‘‘Mat
structures [13]. Comparison studies of structural responses subject Piecewise Linear Plasticity” (MAT_024) was used for steel material
to either CL or UDL conditions, contributions of CA and TMA in which was assumed to be identical in tension and compression.
beam-slab systems, and sensitivity of TMA to horizontal and rota- Two mesh sizes were applied for solid elements of beams, includ-
tional restraints, were carried out in Sections 3 and 4. Section 5 ing 10 mm for joint regions and 20 mm for the other non-critical
focused on the effects of different column removal locations on regions. Only one mesh size of 20
20
6 mm was used for the
progressive collapse resistance to identify the most critical sce- concrete slab. All rebar elements had a mesh size of 20 mm length.
nario for analysis and design. In Section 6, a beam-slab structural A sensitivity study showed that the adopted mesh size yielded rea-
test under corner-column removal applying UDL condition was sonably accurate results. Full models of the beam-only tests were
conducted to confirm the possibility of TMA mobilisation, com- developed, whereas only a quarter model of the beam-slab test
pared to a similar test under CL condition. Overall conclusions was simulated.
are presented in Section 7. Composite behaviour between reinforcement and concrete
material in the beams was simulated by applying the bond-slip
model in CEB 2010 [21] into Contact_1D function of LS-Dyna
2. Numerical models of beam-only and beam-slab structures [19]. Such an application improves the accuracy of simulations
under progressive collapse compared to actual tests and prevents premature fracture of
rebars in concrete due to localised stress concentration [20,22].
2.1. Quasi-static tests on RC structures under internal-column removal A comparison of the bond-slip response for deformed bars in
scenario beams between the CEB model and the proposed model using
LS-Dyna Contact_1D is presented in Fig. 2. Details of the
To investigate the beam-slab effect against progressive collapse, modelling procedure using Contac_1D keyword to consider
Qian et al. [13] conducted a series of quarter-scale quasi-static bond-slip behaviour can be found in Pham et al. [20]. For the
tests on beam-only and beam-slab systems under an internal col- reinforcement in slabs, since mild-steel bars with high ductility
umn loss scenario using CL method. The tests from [13] included were used in the tests [13], fracture of this reinforcement
2D beam-only specimens, 3D beam-only specimens (grillage), occurred much later than the high-yield deformed bars used in
and 3D beam-slab specimens. While the 2D and the 3D beam- the beams. As a result, simulating bond-slip behaviour for such
only specimens included double-span beams with a middle joint round bars would be complicated and yet not necessary. There-
and a column stub at each beam end, the beam-slab systems fore, to simplify the modelling and to save computational time,
included a 2
2-span panel with internal and edge beams plus perfect bonding was assumed between steel reinforcement and
slab extensions equal to a quarter of the beam span. In the present concrete slabs.
522 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536

a) T 2 b) S2
Fig. 1. Test setup for T2 and S2 based on [13].

beam-slab models. Fig. 3(b) shows the simplified support in

Slip (mm) numerical simulations and Fig. 4 shows the three FEM models used
18 in the validations.
CEB model
16
14 Proposed model for Contact_1D
Bond stress (MPa)

2.4. Validation results for beam-only tests

12
10
Basically, structural responses of the beam-only systems from
8
both the actual tests and the simulations comprised three stages
6
of behaviour (Fig. 5). Stage I was from the beginning of the test
4
to the moment when the structure attained its first peak of load-
2
carrying capacity. Within this stage, flexural behaviour co-existed
0
0 2 4 6 8 10 12 14 with CAA. Stage II ran from the first peak to the instant before
the beam bottom rebars near the middle joint fractured, indicating
Fig. 2. Bond-slip models of CEB and Contact-1D for deform bars with 10 mm mobilisation of CA after vertical deflection had exceeded one beam
diameter. depth. In the last stage, residual structural capacity was con-
tributed by remaining top rebars of the double-span beam.
2.3. Boundary conditions Comparisons of applied load versus vertical deflection between
numerical and test results are presented in Fig. 5 for P2 and T2
In the actual tests [13], horizontal forces in the steel supports models, and they generally showed good agreement. In terms of
connected to the column stubs were not measured; instead, linear horizontal movement of the edge joints, numerical predictions
variable displacement transducers (LVDTs) were placed to measure provided a reasonable trend compared to actual test results
horizontal movements of the supports (Fig. 3(a)). The actual lateral (Fig. 5), which verified the modelling of boundary restraint condi-
movements of column stubs extrapolated from LVDT readings tions. Both the FEM and the test results showed an outward move-
were compared with numerical predictions. For simplicity, it was ment of the edge joints at the beginning stage due to CAA, and an
assumed that the connections between the column stubs and the inward displacement at the later part due to CA. Full CA was devel-
steel supports were rigid so that the beam-column joint was rota- oped for all the beam-only specimens.
tionally restrained. Besides, the horizontal restraint of the joint at The numerical simulations also matched well with the test
each beam end was simulated by two horizontal springs connected results in terms of the sequence of failure modes, which began
to the centre of the column stub. The stiffness of the horizontal with the fracture of the beam bottom rebars at the middle joint,
springs was based on the stiffness of the steel supports, which and finally the fracture of the beam top rebars near the end joint.
was roughly equal to 50,000 kN/m. The spring was assumed to Failure modes and damage patterns of FEM models are shown in
behave elastically in both the beam-only (2D and 3D) and the 3D Fig. 6 for 2D and 3D beam-only models, similar to those observed

column
stub

LVDT
P
steel
support KA N

Lspan
strong
floor

a) Measurement of horizontal b) Simplified supports at beam’s end

displacement
Fig. 3. Horizontal support of the system.
 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536 523

a) 2D beam-only model

b) 3D beam-only model c) 3D beam-slab model

Fig. 4. FEM models.

Vertical Displacement (mm) Vertical Displacement (mm)

60 24 Horizontal disp (mm) 100 20

50 20 80 16

Horizontal disp (mm)

Applied load (kN)
Applied load (kN)

40 16 60 12
30 12
40 8
20 8
20 4
10 4
0 0 0 0
0 100 200 300 0 50 100 150 200 250 300
-10 -4 -20 -4
Applied load - Test Applied load - FEM Applied load - test Applied load - FEM
Horizontal disp. - Test Horizontal disp. - FEM Horizontal disp. - test Horizontal disp. - FEM
a) P 2 b ) T2
Fig. 5. Test validation of P2 and T2.

rebars
fracture

a) P2

rebars fracture

b) T2
Fig. 6. Damage patterns.
524 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536

from testing [13]. In the FEM models, the damage and failure of Vertical displacement (mm)
concrete were simulated through the damage index and element 80
erosion. A damage index of 0 denoted the virgin stage, while a
value of 1 indicated total failure of concrete. 60 120

Horizontal reaction (kN)

Applied load (kN)
40
2.5. Validation results for beam-slab tests 70

20
Numerical predictions from S2 achieved good agreement with
20
test results in terms of load-carrying capacity (Fig. 7(a)). At large
0
deformation state, the formations of a tensile net at the centre 0 50 100 150 200 250 300
and a peripheral compressive ring surrounding the tensile region -30
-20
were well replicated in the one-quarter numerical model ((Fig. 7
(b)). The model also successfully simulated the failure modes in -40 -80
the tests, i.e. punching shear failure around the middle column
CL - Applied load UDL - Applied load
and fracture of internal-beam bottom rebars near the middle col-
umn. That is to say, the developed FEM models could represent CL - Horizontal reaction UDL - Horizontal reaction
the actual behaviour of 3D RC beam-slab systems.
Fig. 8. UDL vs. CL for beam-only structures.

3. Investigations of structural responses comparing between CL

and UDL conditions
moment hinge M 0b near the end joint (Fig. 10). Structural capacities
Having validated the FEM models for 2D and 3D tests, they were under CL (PCL ) and UDL (PUDL ) are defined as Eqs. (1) and (2) show-
then used to investigate the differences in structural behaviour ing that P UDL ¼ 2P CL .
subjected to either CL or UDL conditions. The study included three
Pð2LÞ 2ðM b þ M 0b Þ
levels of structural system, viz. beam-only, slab-only, and beam- ¼ ðM b þ M0b Þ ! Pcl ¼ ð1Þ
slab models. 4 L

qð2LÞ2 4ðM b þ M 0b Þ
3.1. Beam-only structure ¼ ðM b þ M0b Þ ! Pudl ¼ qð2LÞ ¼ ð2Þ
8 L
The beam-only model of specimen P2 was employed to com- where Mb and M0b are the plastic bending moment capacities at the
pare structural behaviours between CL and UDL methods, and middle and the end joints, respectively. They can be calculated
results are presented in Fig. 8. It clearly shows that, at the same based on Eq. (3) proposed by Paulay and Priestley [23].
vertical displacement, total load from the UDL case was twice that
0
from the CL condition, while both cases provided identical results M ¼ ðd  d ÞAtss f y ð3Þ
for horizontal reactions. However, maximum vertical deflection 0
of the middle joint from the UDL case was only half of that of where d and d are the distances from the extreme compression
the CL case. With regard to failure modes, the model under UDL fibre of concrete to the centre of tension and compression reinforce-
experienced fracture of top rebars near the end joint (Fig. 9), ment, respectively; Atss and f y are the cross-sectional area and yield

whereas the model under CL and test P2 showed fracture of bottom strength of tension longitudinal rebars, respectively. Eq. (3) is appli-
rebars near the middle joint and top rebars near the end joint cable for under-reinforced sections neglecting the effect of beam
(Fig. 6(a)). As the horizontal reaction had not yet changed to ten- axial forces.
sion in the UDL case, mobilisation of CA was very limited compared
to the CL case. 3.2. Slab-only structure
The relationship between structural capacities of the two load-
ing cases can be explained by plastic-hinge theory (neglecting axial To compare the behaviour between CL and UDL conditions for
compression force effect on plastic moment resistance), assuming the 3D slab-only structure, a FEM model was constructed similar
a collapse mechanism of the double-span beam which includes to S2 model, except that all internal beams were removed. Hence,
positive-moment hinge M b near the middle joint and negative- the RC slab was considered as the only element within the region

Displacement (mm)
200
Compressive
ring
150
Applied load (kN)

100
Tensile net
50 S2 - Test
S2 - FEM Punching
0 shear
0 50 100 150 200 250

a) Structural response b) Damage patterns

Fig. 7. Test validation results of S2.

 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536 525

Top rebars fracture

Fig. 9. Damage patterns of P2 under UDL condition.

a) CL case b ) UDL cas e

Fig. 10. Moment redistributions of sub-assemblages (idealised).

of the removed column. A comparison of structural responses Pslab

cl ¼ 2pðm0s þ ms Þ ¼ 6:28ðm0s þ ms Þ ð4Þ
between the two loading cases is presented in Fig. 11(a). Basically,
the response of the CL case had a much smaller central deflection 2 0
Pslab
udl ¼ wu :l ¼ 24ðms þ ms Þ ð5Þ
than the UDL case. From the initial stage to the first peak, the
capacity of the UDL case was approximately four times that of where m0s and ms are the negative and positive moment capacities
the CL case. This relationship can be explained by employing the per unit width of the slab, respectively; wu and l are the maximum
yield-line theory at small deformations. Under the UDL condition, distributed load and the clear span of the slab, respectively. m0s and
at the yielding of reinforcement in the square slab, the yield-line ms can be simply calculated based on Eq. (6) from Park and Gamble
pattern includes four triangular segments, whereas under the CL [3].
condition the slab shows a circular pattern of yield-line (Fig. 12).  
According to Park and Gamble [3], for a square slab under fixed 0:59As f y
ms ¼ As f y ds  ð6Þ
boundary condition and with equal reinforcement in orthogonal f ck
directions, total applied load on the slab at yielding for the CL
where As and f y are the reinforcement area per unit length and the
(P slab slab
cl ) and the UDL (P udl ) conditions can be calculated by using
yield strength of tensile rebars, respectively; ds is the effective
Eqs. (4) and (5), respectively. It is shown that from these two equa-
depth of the slab and f ck is the concrete cylinder strength.
tions Pslab slab
udl  3:82P cl , which explains why the slab resistance under After the first stage of flexural and compressive membrane
UDL case was nearly four times that of the CL case. Nonetheless, if action (CMA) in the slabs, TMA was more efficiently mobilised in
the slab had an aspect ratio larger than one, or had different rebar the UDL case whereas development of TMA was limited in the CL
arrangements in two orthogonal directions, then this value would case due to potential premature collapse. Concerning failure
be different. modes, the governing damage of the CL condition was punching

Displacement (mm) Displacement (mm)

400 600
350 500
Applied load (kN)
Applied load (kN)

300
400
250
200 300
150 200
UDL UDL
100
CL x 4 100 CL x 3
50
0 0
0 50 100 150 200 0 50 100 150 200 250
a) Slab-only structures b) Beam-slab structures
Fig. 11. Structural responses between CL and UDL conditions.
526 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536

l l

Fig. 12. Yield-line patterns.

shear around the removed column location (Fig. 13(a)). Since the model of specimen S2. Numerical results show that, in terms of
structure under the CL case behaved like a flat slab, punching shear load-carrying capacity, the UDL case was nearly three times larger
would limit the maximum displacement, hindering mobilisation of than the CL case as shown in Fig. 11(b). Considering the difference
TMA. On the other hand, the slab under UDL could experience in load-carrying capacity, predicted structural capacity for the UDL
greater deflection, leading to a higher capacity at the final stage case is twice that of the CL case in beam-only system (Section 3.1),
compared to the first peak of CMA. Failure mode of the UDL case or nearly four times that in slab-only system (Section 3.2). For this
included occurrences of full-depth cracks and fracture of rebars particular RC design, the maximum capacities of the beam-only
near the slab edges (Fig. 13(b)). structure (T2) based on plastic-hinge theory under UDL and CL
Fig. 14 shows the spread of reinforcement yielding in the slab at conditions (Eqs. (1) and (2)) are 96 kN and 48 kN, respectively,
a vertical deflection of 55 mm, or one slab thickness. For the CL whereas the capacities of the slab-only structure (S2 without inter-
case, the top rebars which were already yielded concentrated only nal beams) based on yield-line theory under UDL and CL conditions
within one quarter of the slab edges. However, for the UDL case, (Eqs. (4) and (5)) are 85 kN and 23 kN, respectively. Therefore,
most of the top rebars along the slab edges had already yielded when combining the beam-only and the slab-only models
at this deflection. Similarly for the bottom-rebar layer in the slab, together, the maximum capacity of the system under UDL condi-
yielding in the CL case only congregated around the central col- tion (181 kN) is approximately 2.6 times that under CL condition
umn, whereas yielding reinforcement in the UDL case developed (71 kN). The UDL case also had a higher maximum displacement.
along the slab diagonal and extended to nearly one quarter of Failure mode of the beam-slab structure at the final stage is shown
the area. That is to say, for the same central deflection, both the in Fig. 15 and will be further discussed in Section 3.4.
top and the bottom reinforcement was more mobilised in the
UDL case compared to the CL case.
3.4. Discussions on the combined effects of beams and slabs under both
CL and UDL conditions
3.3. Beam-slab structure
Under ‘‘CL condition” at the internal column, responses of the
The last level of comparison between CL and UDL cases involved three modelling levels from Sections 3.1-3.3 are shown in Fig. 16
the full 3D beam-slab system, which employed the numerical (a). For the beam-slab model, not only did its structural capacity

a) CL b) UD L

Fig. 13. Failure modes of slab-only structures.

 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536 527

Fig. 14. Yielding of slab-only rebars at 55 mm deflection.

a) CL b) UDL

Fig. 15. Failure modes of beam-slab structures.

Displacement (mm) Displacement

180 Beam-slab 600
160 Slab-only
Beam-only
500
140
Applied load (kN)

Applied load (kN)

120 400
100
300
80
60 200 Beam-slab
40 Beam-only
100
20 Slab-only
0 0
0 100 200 300 0 50 100 150 200 250
a ) CL b) UDL
Fig. 16. Effects of beam and slab.

increase compared to the slab-only model, the maximum deflec- compared with both the beam-slab and the slab-only models. In
tion also increased because punching shear occurred later due to terms of failure modes, the beam-slab model failed when full-
shear strength contribution from internal beams. Additionally, its depth cracks in the slab occurred near the edge beams and the
maximum deflection was similar to the displacement of the top rebars of internal beams fractured near the columns (Fig. 15
beam-only structure at the time when the beam bottom rebars (b)). That is to say, the failure mode of the beam-slab system under
started fracturing. Besides punching shear failure, the final peak UDL was also a combination of failure modes from both the beam-
of the beam-slab model was denoted by fracturing of bottom only and the slab-only models. Moreover, the top rebars in the
rebars in internal beams (Fig. 15(a)). In other words, the failure internal beams from the beam-slab system fractured later com-
mode of the beam-slab system at the final stage under CL condition pared to those from the beam-only system, due to the T-beam
was a culmination of failure modes from both the beam-only and effect in the beam-slab system. TMA, which was effectively mobi-
the slab-only models. Up to the instant before the occurrence of lised in the slab-only system, was well developed in the beam-slab
failure, CA in internal beams of the beam-slab model had not been model. Under UDL condition, after the first peak of CMA, response
significantly mobilised as the maximum vertical deflection was curves of the slab-only and the beam-slab models were parallel to
only slightly larger than one beam depth (Fig. 16(a)). Nevertheless, each other. The difference between these two curves was 110 kN,
TMA had been significantly mobilised after the displacement has which was comparable to the capacity of the beam-only structure
exceeded one slab thickness. based on plastic-hinge theory (96 kN). This means that the
Similar to the CL case, FEM results from all the three structural strength contribution of internal beams in the beam-slab system
levels under ‘‘UDL condition” are shown in Fig. 16(b). Clearly, the at large deflection state arose primarily from flexural response.
maximum displacement of the beam-only model was the smallest Enhancement from CA in the internal beams was rather limited.
528 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536

4. Effect of boundary stiffness on the mobilisation of TMA Displacement (mm)

600
4.1. Horizontal restraint
500
To study the sensitivity of TMA to horizontal restraint condi-

Applied load (kN)

tions, the beam-slab model was analysed with several values of 400
horizontal spring stiffness, including zero stiffness, perfectly rigid,
and 50,000 kN/m (semi-rigid) under both CL and UDL conditions. 300
Values of applied load and horizontal reaction in beams versus
central displacement (at the removed column location) are pre- 200
sented in Fig. 17. For the first peak of CMA, applied load from the
rigid-restraint case was larger than that from the zero-restraint 100
case by 17% and 25% with regard to CL and UDL cases, respectively.
However, there was almost no difference in the load-carrying 0
0 50 100 150 200
capacity for the last peak before the models failed under CL or
UDL-Slab extension CL-Slab extension
UDL cases. Suffice to say, the stiffness of horizontal restraint had
some effect on CMA, but little or no effect on TMA. In fact, the UDL-No slab extension CL-No slab extension
formation of the peripheral compressive ring had in some way
Fig. 18. Effects of rotational restraint on beam-slab structures.
provided sufficient restraint for the mobilisation of tensile
membrane forces at the centre. Moreover, the behaviour of the
beam-slab system did not benefit from CA of interior beams. Under developed later compared to the original model. This observation
CL condition, horizontal reaction from internal beams was always agreed well with Section 3.2 since stresses of top reinforcement
in compression showing no development of CA. But under UDL at the slab edges under the UDL case could reach yielding faster
condition, CA in internal beams was developed as horizontal and more uniformly than the CL case. Hence, it is concluded that
reaction switches from compression to tension. Nonetheless, this beam-slab models under UDL condition were more influenced by
mobilisation was somewhat limited due to small deflections of rotational stiffness of restraint condition compared to the models
the slab. under CL condition. However, since this conclusion is made based
on simulation results from FEM models developed in this study,
there is a need to conduct experimental work on actual beam-
4.2. Rotational restraint
slab specimens under different loading configurations to generalise
this finding.
To investigate the effect of rotational restraint at the slab edges,
the extended parts of the slab in the beam-slab model were deleted
to reduce rotational stiffness at the slab perimeter. Both CL and 5. Effect of location of removed column on structural response
UDL cases were considered. Results from the revised model were
compared with those from the original model which includes the To date, most studies comment that the removal of a corner
slab extension (Fig. 18). It is shown that under the CL condition, column is the most critical scenario for progressive collapse due
there was almost no difference in terms of structural response to a lack of restraints from two adjacent sides of the removed
between the two models. Nonetheless, for the UDL case, the overall column. Besides, penultimate columns which are next to the
load-carrying capacity of the revised model (without slab corner column are considered as the second most critical scenario
extension) was smaller than that of the original model (with slab for progressive collapse analysis. Experimental studies [14,15]
extension) for both the first peak of CMA and the latter part of show that the removal of the penultimate column leads to large
TMA. Additionally, TMA in the model without slab extension was inward movements of adjacent columns, which in turn may

Displacement (mm) Displacement (mm)

200 400 600 400
500
Horizontal reaction (kN)

150 300
Horizontal reaction (kN)

300
400
Applied load (kN)

Applied load (kN)

100 200 300 200

200
50 100 100
100
0 0 0 0
0 50 100 150 200 0 50 100 150 200 250
-100
-50 -100 -100
-200
-100 -200 -300 -200
Applied load - Rigid restraint Applied load - Rigid restraint
Applied load - k=5e4kN/m Applied load - k=5e4kN/m
Applied load - Zero restraint Applied load - Zero restraint
Horizontal reaction - Rigid restraint Horizontal reaction - Rigid restraint
Horizontal reaction - k=5e4kN/m Horizontal reaction - k=5e4kN/m
Horizontal reaction - Zero restraint Horizontal reaction - Zero restraint

a) CL b) UDL

Fig. 17. Effects of horizontal restraint stiffness on beam-slab structures.

 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536 529

precipitate progressive collapse. However, due to cost constraints, Displacement (mm)

there is no experimental work for beam-slab structures that sys- 70
tematically consider all the possible scenarios of missing column.
60
In this section, several scenarios of column loss were numerically
investigated using the validated beam-slab models from Section 2 50

UDL (kN/m2)
(Fig. 19). For more realistic modelling, UDL was applied to the slab
40
area affected by the column removal [2]. Five scenarios were con-
sidered including internal-column, corner-column, edge-column, 30
internal
penultimate-internal (PI) and penultimate-external (PE) column edge
20
removal cases. The internal-column case with the greatest struc- corner
tural capacity was used as the baseline for comparison. The rela- 10 PI
PE
tionship between displacement of the joint above the removed
0
column and structural capacity per unit area from the affected slab
0 50 100 150 200 250
was used as a comparison basis among the five different column
removal scenarios. Fig. 20. Structural responses from different column loss scenarios.

5.1. Corner-column removal

In the internal-column case, TMA in the centre of the slab was
The comparison results in Fig. 20 show that the corner-column surrounded by the formation of a peripheral compressive ring
case was not the one with the smallest structural capacity per unit (Fig. 22(a)). On the other hand, for the corner-column case, the
area. In fact, the response of the corner-column scenario was com- presence of edge beams DA and DC at discontinuous sides
parable with the internal-column and the edge-column cases. (Fig. 22(b)) helped to reduce the slab deformation at these edges.
Although the effect of CMA on the corner-column scenario was As a result, a peripheral compressive ring could still be partially
unnoticeable due to a lack of horizontal restraints, the final capac- formed and limited TMA could be mobilised.
ity as well as the maximum central deflection was comparable to
internal- and edge-column cases. Indeed, it was the PI-column 5.2. Penultimate-column removals
and the PE-column scenarios that sustained the lowest structural
capacities. Therefore, it is concluded that the corner-column loss Ultimate capacities from the PI- and the PE-column cases were
was not necessarily the most critical scenario. This is explained the smallest among all the five scenarios (Fig. 20). Besides, struc-
through the yield-line formation in Fig. 21. In this figure, the tural capacity per unit area of the PE-column case was slightly
yielding zones of top and bottom rebars at a deflection of one slab greater than that of the PI-column case, although the load from
thickness for the internal- and the corner-column cases are respec- the PI-column case could be redistributed to a larger area. These
tively shown. It also illustrates the formation of positive- and observations can be explained by the presence of some discontin-
negative-moment yield lines for the two cases. For the internal- uous slab edges in the PI- or the PE-column cases which provided
column case, bottom-rebar yielding zone concentrated near the hogging-moment resistance. Since the boundary beams at these
slab centre and along the slab diagonal, whereas for the corner- discontinuous sides did not have relatively strong rotational
column case, yielding developed towards the edge columns restraints compared to those from the continuous sides, slab
besides the diagonal region. It is shown that the two edge beams hogging-moment resistances at these regions were affected.
DA and DC (Fig. 21(b)) provided noticeable strength (based on their Fig. 23 shows the yielding zones of top- and bottom-layer rebars
stiffness) to the slab area so that additional positive yield lines at a deflection of one slab thickness for the PI- and the
(AO and CO) could be formed and therefore increased the PE-column cases. The yielding of top reinforcement at
structural capacity per unit area. discontinuous edges was rather limited compared to that at

Fig. 19. Progressive collapse scenarios of multi-storey buildings.

530 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536

//

//
//

//
//

//
//
// // // // //

/
//

//
//

//
//

//
//

//
//

//
//

//
//

//
//

//
//

//
//
//

//

//
//
//

//

//
//

//
//

//
//

//

//
//
//

//

/
//
//

//

//
//
//

//

//
//

//

//
//
//

// //

//
//
//

//

//
//
//

// //

//
//
//

a) Internal-column case b) Corner-column case

Fig. 21. Rebar yielding at 55 mm deflection of internal- and corner-column cases.

Fig. 22. TMA in slabs.

internal continuous edges and was mostly concentrated at the of 160 mm, at which both the PI- and the PE-column cases almost
areas above the peripheral columns. Therefore, hogging moment attained their ultimate resistances, the inward movement of the AP
resistance as well as tensile effect from the slab top reinforcement columns from the PE- and the PI-column cases were 11 and 17
could not be effectively mobilised at those areas. This phenomenon mm, respectively. With such large horizontal deflections, the AP
did not exist in the internal- and the corner-column cases (Fig. 21). columns could have failed prematurely. In summary, the PE- and
Besides, in the PI-column model, one half of hogging-moment the PI-column models were more vulnerable compared to the
edges was discontinuous, whereas in the PE-column model it other cases such as the corner- or the edge-column scenarios due
was only one quarter. As a result, structural resistance from the to smaller structural capacity and the danger of pulling-in of AP
PE-column model was generally greater than the PI-column model. columns. Comparing between the PE- and the PI-column cases,
When the vertical deflection in the PI- or the PE-column case although the PE-column case had a greater structural response, it
became relatively large, the peripheral columns which were next also generated a larger movement for its AP column, which in turn
to the removed column, so-called the affected peripheral (AP) col- made it more critical to progressive collapse. In fact, the previous
umns, would experience excessive inward horizontal displace- beam-slab tests for PE-column removal [15,17] showed very lim-
ments. Again, this phenomenon did not occur under the edge- ited mobilisation of TMA.
and the corner-column removal scenarios. Fig. 23(a) and (b) show The conclusion in this study showing that corner-column
the positions of the AP columns as well as their inward movements removal was not necessary the most critical scenario agrees with
in the respective PI- and PE-column cases. At a vertical deflection the assessment of composite frame structures from Stylianidis
 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536 531

Fig. 23. Rebar yielding at 55 mm deflection of PI- and PE-column cases.

et al. [18]. While in the current study, structural capacity in the Displacement (mm)
corner-column case was enhanced by limited development of 70
TMA, the contribution from grillage system is the one investigated
60
in [18] since slabs are not considered.
50

UDL (kN/m2)
5.3. Analyses of progressive collapse using 3D beam-slab-column 40
models
30 internal
To get a close-to-reality response of structures under several edge
20
missing column cases, 3D beam-slab-column models were corner
employed in which the beam-slab models were updated by adding 10 PI
the peripheral columns to the slab edges. Boundary conditions of PE
these columns were simplified by using pin and roller supports 0
0 50 100 150 200 250
at their mid-height. All the columns had the same dimension of
200
200 mm2 and a same reinforcement of 4T13 (high strength Fig. 25. Structural responses using beam-slab-column models.
deformed bars). Before the UDL load was applied to slab area, the
peripheral columns were loaded by axial forces with a load ratio 6. Beam-slab test under corner-column removal
of 0.2. All the five column loss scenarios were considered. Fig. 24
illustrates the beam-slab-column model applied for the internal- To study the possibility of TMA mobilisation in the corner-
column case. Results of this simulation series are presented in column scenario mentioned in Section 5.1, a beam-slab test
Fig. 25 showing that PI- and PE-column cases still had smallest labelled as S-COR-UDL was conducted in this study. The corner-
structural capacities. This observation agrees with the results of column specimen was tested under UDL condition, and employed
beam-slab models in Fig. 20. Nevertheless, there was no premature similar geometry, reinforcement design, as well as boundary con-
failure of AP columns. In fact, the likelihood of AP columns being dition from a previous test named S-COR conducted under CL
prematurely failed due to inward movement relied on the level method [24].
of axial forces applied on these columns. As a result, the use of such
complex beam-slab-column models is necessary for assessing PI- 6.1. Specimen design and test setup
and PE-column scenarios. For internal-, edge- and corner-column
cases, results from beam-slab-column models (Fig. 25) are rela- The beam-slab specimen was a 2/5 scaled-down model from a
tively similar to those from beam-slab models (Fig. 20) indicating 6-storey, 6
4 bay prototype building. It had equal span of 2.4 m
that the modelling of columns is not necessary for those scenarios. in both directions and was isotropically reinforced for gravity loads
410
410

375 1500 1500 375

Fig. 24. Beam-slab-column model for internal-column case.

532 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536

in accordance to Eurocode 2 [25]. The specimen consisted of than 180 mm, both the beams and the slab would have gone into
180 mm deep by 100 mm wide beams with a span length of tensile mechanism. At the first stage, vertical load increased lin-
2.4 m and 180 mm square column stubs. The slab thickness was early from 0 kN up to 45 kN (at about 10 mm MJD). Subsequently,
80 mm whereas rotational restraints from adjacent slabs were rep- the load increased non-linearly with subsequent crack develop-
resented by thicker slab portion (120 mm) extending 240 mm ment in the slab and the beams. Initially, cracks were observed
beyond the perimeter beams. The slab bottom reinforcement was at the beams and the slab near the corner joint and the beam-
continuous, whereas the slab top reinforcement was only placed end regions, as well as exterior faces of primary columns (C-L
at edge regions. Reinforcement detailing of the specimen is shown and C-T). But as the specimen deflected further, more cracks were
in Fig. 26. developed near the middle and the bottom regions of the column
The concrete compressive cylinder strength was 32 MPa. The but only a few at the beam-end or perimeter of the slab. After
yield strength of high-strength deformed bars (T10) and mild- attaining the first peak of 88 kN at a deformation of 97mm (i.e. half
steel round bars (R6) were 507 and 400 MPa, respectively. Dis- beam depth) as shown in Stage I of Fig. 28, a slight drop in the
placement transducers were located along the beams and the slab structural capacity was observed, before showing a second trend
to measure vertical deflections. Strain gauges were mounted of load increase with increasing of vertical displacement (Stage
extensively along the primary beams and on the slab to capture II). Such an increase denoted the mobilisation of some tensile
the development of different load-carrying mechanisms. A loading mechanisms in the beam-slab structure. On the other hand, in
scheme consisting of twelve contact points was employed for S- the test of S-COR (CL condition), such a second peak of load-
COR-UDL to simulate actual UDL condition (Fig. 27(a)), compared carrying capacity was not observed after the structure attended
with CL method employed in the previous test (Fig. 27(b)). Such its first peak of 21 kN at 166 mm displacement. When the load sus-
a multi-point loading configuration has been successfully used in tained by S-COR-UDL became significantly large, the vertical and
previous UDL tests [14,15,17]. Primary columns C-L and C-T were rotational restraints provided by the self-weight of the extended
fixed to steel supports which in turn were bolted to the strong floor slab were inadequate, inducing large bending moment at the col-
(similar to specimen S2 [13]). On the other hand, secondary col- umns. As the primary columns were not strengthened or over-
umn C-LT was connected to a pin-based circular hollow steel col- designed, plastic hinges were developed at the bottom exterior
umn oriented at 45° direction (Fig. 26). faces of C-L and C-T columns (Fig. 29(a)). This led to lifting up of
the slab region supported by the perimeter beams and hence neg-
6.2. Experimental results ative yield line was not fully formed at the perimeter region of the
slab. As the specimen deformed, the applied load kept increasing
FigS. 28and 29 show the load–displacement relationship and and reached the maximum load of 105 kN at a displacement of
the crack patterns of S-COR-UDL, respectively, compared with 270 mm. Subsequently, localised failure due to severe damages,
those of the CL test [24]. The development of crack patterns was i.e. opening of cracks and concrete crushing at the column plastic
divided into two stages, i.e. when the corner-column displacement hinges were observed, leading to a gradual reduction in load. The
was smaller or larger than one beam depth (180 mm), indicated by test was stopped at the deflection of 400 mm when the load was
black and red lines, respectively. Generally, CA in beams and TMA reduced to 85 kN due to safety concern (fracture of column rein-
in slabs were mobilised at one cross-sectional depth of beams and forcement was undesirable as it might lead to sudden collapse of
slabs, respectively. Hence, it was assumed that at deflection larger the structure).

100 2300 100 corner Beam-column 200

column A 650 A
650 B

200
240
290

180
C-LT
C-L
45°
R6@100

C C
720

300
R6@90 R6@110 R6@90
720 720 2220
3T10
R6@150 R6@100 A B A
4T10 4T6
4 1
2400

80

80

2 2T10
3 R6@150
180

180

180
R6@150
R6@150

2T10 2T10
720

100 100 180

A-A B-B C-C
Corner column Column-stub supports
C-T
2400 290

1 Main Slab (thickness=80 mm)

1125

1125

2 Extension Slab (thickness=120 mm) I-section tube-section

3 Bottom Slab reinforcement

4 Top Slab reinforcement C-L and C-T C-LT

Fig. 26. Specimen design.

 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536 533

C-L C-T
12-point
loading point
C-L C-T
load

L T L T

corner column
corner column

a) S-COR-UDL b) S-COR [24]

Fig. 27. Test setup.

120 Stage I Stage II (a). These positive yield lines contributed to flexural capacity and
enabled the development of compressive effects at the slab
Total applied load (kN)

100 perimeter region (Fig. 29(a)) as the slab deflected. This compres-
sive arching supported TMA in the central region. On the other
80
hand, there was no sign of positive yield lines occurring at the bot-
60 S-COR-UDL tom surface of S-COR (Fig. 30(b)). Instead, a failure line due to con-
S-COR crete crushing was observed at the slab soffit, developing from
40 column C-L to column C-T. That is to say, the change of loading
Peak configuration from CL to UDL in the corner-column scenario not
20
only increased the load-carrying capacity of the structure signifi-
0 cantly, but also changed the formation of both positive and nega-
0 100 200 300 400 tive yield lines in the slab.
Displacement (mm) The deflection profile of the slab along diagonal line D1 is plot-
ted in Fig. 31(a) for S-COR-UDL. The deflection profile at the slab
Fig. 28. Load-displacement responses.
central region had a parabolic shape which indicated membrane
action. This phenomenon was not observed in the CL case of S-
COR (Fig. 31(b)). Strain readings of slab reinforcement along diag-
During both the CL and UDL tests, it was difficult to trace the onal line D1 were also obtained and plotted in Fig. 32. It is shown
exact development of cracks under the slabs. Due to safety reason, that the strains of bottom rebars at all locations were under ten-
access to the bottom of the slabs was prohibited. Therefore, cracks sion (Fig. 32(a)) as they were located along the positive yield line.
at the bottom surface were only drawn after testing (Fig. 30). At the In TMA mechanism, compressive force at the slab perimeter region
slab soffit of S-COR-UDL, formation of positive yield lines started supported the development of tensile net at the centre of the slab.
occurring at about 40 mm displacement (0.5 slab thickness), and This explained why the strains were generally greater nearer to the
progressively developed from the corner joint and C-LT column middle joint, and smaller nearer to the slab perimeter and C-LT col-
towards the slab centre. A complete diagonal line was formed at umn. For the top rebars which were located along D1 (Fig. 32(b)),
about 100 mm displacement (1.25 slab depth). After testing, paral- tensile strains were also observed for most of the test duration
lel positive yield lines were observed and marked as seen in Fig. 30 although the rebars did not actually reach yielding.

C-L C-LT C-L

C-LT

Plastic
Plastic
hinges
hinges
D1
D1

C-T C-T
Corner column Corner column Interior
face Exterior
face

Plastic hinges Plastic hinges

a) S-COR-UDL b) S-COR [24]

Fig. 29. Crack patterns at slab top surface.

534 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536

corner joint
C-L C-L
Positive corner joint
yield lines Rotate and
pulled-in about
the line
Left beam
concrete
Transverse
crushing
beam
C-T

C-T C-LT
C-LT

(a) S-COR-UDL (b) S-COR [24]

Fig. 30. Crack patterns at slab bottom surface.

corner V-1 V-2 V-3 C-LT corner V-1 V-4 V-5 C-LT
column column
0 600 1200 1800 2400 3000 3600 0 600 1200 1800 2400 3000 3600
0 0
72mm
Vertical displacement (mm)

Vertical displacement (mm)

90mm
-100 144mm
-100
180mm
-200 C-LT
-200 C-LT 276mm
270mm V-5
V-3 -300
-300 408mm V-4
360mm V-2
-400 V-1
V-1 D1
-400 D1
400mm -500
corner column
540mm
corner column
-500 -600
Location along slab diagonal (mm) Location along slab diagonal (mm)

a) S-COR-UDL b ) S -C OR

Fig. 31. Deflection profile of slab along D1.

Displacement (mm) Displacement (mm)

10000 2500
B4 B7
9000
B16 2000 T4 T2
8000 B5
B10
B7 D1
7000 1500
B4 B5 T2
Strain (µ)

6000 D1
Strain (µ)

B16
5000 1000 T4
4000 B10
3000 yielding 500
2000
0
1000 0 100 200 300 400
0 -500
0 100 200 300 400
a) Bottom rebars b) Top rebars
Fig. 32. Strain values in slab reinforcement along D1 of specimen S-COR-UDL.

6.3. Presence of TMA in the corner-column tests patterns (Fig. 29(a) and Fig. 30(a)), as well as tensile strains in
the slab reinforcement along line D1 (Fig. 32). The development
The absence of TMA in S-COR (point load) was clearly indicated of positive yield lines contributed to flexural capacity, as well as
from the decreasing load–displacement curve after attainment of facilitated the formation of compressive arching at the slab
flexural capacity (Fig. 28), as well as from localised bending along perimeter, which in return anchored tensile net in the central
negative yield lines (Fig. 29(b)). That is to say, S-COR solely relied region. In general, under UDL condition, the load capacity of the
on flexural capacity, i.e. yield lines in the slab and plastic hinges beam-slab structure could increase to nearly five times of that
in the beams. On the other hand, TMA development was possible under point load.
in the beam-slab specimen subject to UDL (S-COR-UDL), which From the observations of crack patterns at the top and the
was confirmed from the load-displacement curve (Fig. 28), crack bottom surfaces of the corner specimen, together with measured
 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536 535

Negative yield line TMA was less affected by this parameter. This finding was observed
Cracks at top surface in numerical studies for both loading cases. On the other hand, the
response of the beam-slab structure under UDL was dependent on
Tensile action at slab edges
rotational restraint of the slab edges.
Tensile action at slab centre Under the UDL condition, among different scenarios of column
Compressive zones at slab edges loss, the corner column case was not necessarily the most critical
one. If the boundary beams at the slab discontinuous edges were
C-LT stiff enough, they could help to develop limited TMA in the slab
C-L
even under unrestrained conditions from two sides of the removed
column. Instead, the penultimate column removal scenarios could
be more vulnerable to progressive collapse due to weak rotational
restraint of discontinuous edges which affected the stress develop-
ment of the slab’s top reinforcement on these boundaries. More-
1

over, excessive inward movement of AP columns next to the

D

removed one (which did not occur in the edge- or the corner-
column loss) made the penultimate-column removal scenarios
more critical.
The corner-column test under UDL conducted in this study
showed partial formations of both the compressive ring at the slab
edges and the tensile net at the slab centre, which indicated TMA
development. Compared to the test under CL condition, structural
capacity of the UDL test was much greater and it included a second
phase of load increase due to TMA. Last but not least, the test
Corner C-T demonstrated that the change of loading method from CL to UDL
also changed the crack patterns and formations of yield lines.
Fig. 33. Partial mobilisation of TMA in S-COR-UDL.

Acknowledgement
tensile strains of reinforcement along diagonal line D1, partial
mobilisation of TMA was proposed as shown in Fig. 33. A compres-
The authors gratefully acknowledge the funding entitled as
sive ring was partially formed, concentrating near the corner col-
‘‘Development of a Design Guideline and Analytical Tool to Miti-
umn and near C-LT column. It anchored tensile forces towards
gate Progressive Collapse of Buildings against Explosive Effects”,
the slab central region. On the other hand, at the slab centre along
which is provided by Ministry of Home Affairs - Singapore.
line D1, tensile forces were also created from yielding stresses of
bottom rebars, heading towards columns C-L and C-T. They repre-
sented the tensile net generated in TMA. Such a partial mobilisa- References
tion of TMA was similar to that observed from numerical
predictions (Fig. 22(b)). As the columns were not over-designed, [1] General Services Administration. Progressive collapse analysis and design
guidelines for new federal office buildings and major modernization projects,
plastic hinges were developed at the columns instead of at the Washington, D.C.; 2003.
beams (Fig. 29(a)), resulting in twisting of secondary beams and [2] Department of Defense. Design of buildings to resist progressive collapse,
hindering the formation of negative yield lines along the slab Unified Facilities Criteria (UFC) 4-023-03. Washington, D.C.; 2013.
[3] Park R, Gamble WL. Reinforced concrete slabs. John Wiley & Sons; 2000.
perimeter. In the event of successful development of negative
[4] Yu J, Tan KH. Experimental and numerical investigation on progressive
yield-lines, the load capacity may be even higher, not to mention collapse resistance of reinforced concrete beam column sub-assemblages.
additional contribution from TMA. Eng Struct 2013;55:90–106.
[5] Bailey CG. Membrane action of unrestrained lightly reinforced concrete slabs
at large displacements. Eng Struct 2001;23:470–83.
[6] Foster S, Bailey C, Burgess I, Plank R. Experimental behaviour of concrete floor
7. Conclusions slabs at large displacements. Eng Struct 2004;26:1231–47.
[7] Yi WJ, He QF, Xiao Y, Kunnath SK. Experimental study on progressive collapse-
In this paper, numerical models were employed to investigate resistant behavior of reinforced concrete frame structures. ACI Struct J
2008;105.
structural responses under UDL, compared to CL case which is pre- [8] Sadek F, Main JA, Lew HS, Bao Y. Testing and analysis of steel and concrete
ferred in laboratory tests due to expediency in test setup. It was beam-column assemblies under a column removal scenario. J. Struct. Eng.
shown that, under the single column removal scenario, the 2011;137:881–92.
[9] Yu J, Tan KH. Structural behavior of RC beam-column subassemblages under a
beam-only and the beam-slab structures behaved differently when middle column removal scenario. J. Struct. Eng. 2013;139:233–50.
subjected to either UDL or CL case. While CA in beam-only systems [10] Lim NS, Lee CK, Tan KH. Experimental studies on 2-D RC frame with middle
had significant effect under the CL condition, its development column removed under progressive collapse. In: Proceedings of fib symposium
2015. Copenhagen, Denmark; 2015.
under the UDL case was moderate because the top longitudinal [11] Pham AT, Tan KH. Experimental study on dynamic responses of reinforced
reinforcement fractured at smaller central deflection. For the concrete frames under sudden column removal applying concentrated
beam-slab behaviour, the fracture of reinforcement in internal loading. Eng Struct 2017;139:31–45.
[12] Qian K, Li B. Slab effects on response of reinforced concrete substructures after
beams could be used as an indicator of the load-carrying capacity loss of corner column. ACI Struct J 2012;109.
for both CL and UDL conditions. The combined effects of beams [13] Qian K, Li B, Ma JX. Load-carrying mechanism to resist progressive collapse of
and slabs could be considered by separately taking into account RC buildings. J Struct Eng 2014;141:04014107.
[14] Pham XD, Tan KH. Experimental study of beam–slab substructures subjected
the behaviour of the beam-only and the slab-only structures under
to a penultimate-internal column loss. Eng Struct 2013;55:2–15.
both types of loading method. However, the contribution of inter- [15] Pham XD, Tan KH. Experimental response of beam-slab substructures subject
nal beams was limited to flexure and CAA only. The development to penultimate-external column removal. J Struct Eng 2015;141.
of beam’s CA in the beam-slab structure was limited due to small [16] Lu X, Lin K, Li Y, Guan H, Ren P, Zhou Y. Experimental investigation of RC beam-
slab substructures against progressive collapse subject to an edge-column-
value of vertical deflection corresponding to beam rebar fracturing. removal scenario. Eng Struct 2016. http://dx.doi.org/10.1016/j.
While CMA was influenced by the stiffness of horizontal restraints, engstruct.2016.07.039.
536 A.T. Pham et al. / Engineering Structures 150 (2017) 520–536

[17] Qian K, Li B, Zhang Z. Influence of multicolumn removal on the behavior of RC [21] CEB-FIP Model Code. Design of concrete structures, Fédération Internationale
floors. J Struct Eng 2016;142(04016006):1–13. du Béton fib/International Federation for Structural Concrete; 2010.
[18] Stylianidis P, Nethercot D, Izzuddin B, Elghazouli A. Robustness assessment of [22] Bao Y, Lew HS, Kunnath SK. Modeling of reinforced concrete assemblies under
frame structures using simplified beam and grillage models. Eng Struct column-removal scenario. J Struct Eng 2014;140:04013026.
2016;115:78–95. [23] Paulay T, Priestley M. Seismic design of reinforced concrete and masonry
[19] Hallquist JO. LS-DYNA keyword user’s manual version 971. Livermore buildings. Wiley, New York; 1992.
Software Technology Corporation; 2007. [24] Lim NS. Systematic study on reinforced concrete structures under progressive
[20] Pham AT, Tan KH, Yu J. Numerical investigations on static and collapse. Singapore: Nanyang Technological University; 2017.
dynamic responses of reinforced concrete sub-assemblages under [25] European Committee for Standardization. Eurocode 2: Design of concrete
progressive collapse. Eng Struct 2016. http://dx.doi.org/10.1016/j.engstruct. structures - General rules and rules for buildings, 1992-1-1:
2016.07.042. 2004. Brussels: CEN; 2004.