Professional Documents
Culture Documents
Numerical Study of Progressive Collapse Resistance of RC Beam-Slab
Numerical Study of Progressive Collapse Resistance of RC Beam-Slab
Numerical Study of Progressive Collapse Resistance of RC Beam-Slab
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
A R T I C L E I N F O A B S T R A C T
Keywords: In this paper high fidelity solid-element-based numerical models are used to study the progressive collapse
Reinforced concrete resistance of reinforced concrete (RC) beam-slab substructures. This is a common type of system used in RC
Beam-slab substructures framed structures through monolithic construction. After careful calibration of model parameters, the numerical
Column removal scenario models are validated through comparisons with test data and further employed to investigate the load transfer
Progressive collapse
mechanisms and explore parameters affecting the robustness of RC beam-slab substructures under perimeter
Numerical simulations
column removal scenarios. The parameters studied include lateral restraint stiffness at the substructure
boundaries, loading schemes in testing, slab thickness and reinforcement detailing in slabs. The simulation
results show that the progressive collapse resistance of the RC beam-slab substructures is developed primarily
through compressive arch action of the longitudinal beams and flexural mechanism of the transverse beam at
small deflections and catenary action of beams as well as tensile membrane action (TMA) of slabs at large
deflections. The slabs enhance structural resistance through working compatibly with beams as L- and T-section
“composite beams” at hogging moment regions and developing TMA. If the slabs are just simply converted into
equivalent flanges of beams, the resistance at large deflections will be significantly underestimated.
⁎
Corresponding author.
E-mail address: yujun@hhu.edu.cn (J. Yu).
https://doi.org/10.1016/j.engstruct.2017.12.038
Received 7 October 2017; Received in revised form 20 December 2017; Accepted 20 December 2017
Available online 29 December 2017
0141-0296/ © 2017 Elsevier Ltd. All rights reserved.
J. Yu et al. Engineering Structures 159 (2018) 14–27
used to take account of dynamic effects. to the specimen, as demonstrated in Fig. 2. During the test, two hy-
Under gravity, provided that the supporting columns are adequately draulic actuators were respectively installed at the top and the bottom
strong, beams and slabs are the primary structural members to redis- of the middle column stub to generate a concentrated load to each
tribute loads after column removal. Previous experiments on the pro- specimen. The load was applied by monotonically increasing the dis-
gressive collapse of RC beam-slab substructures were mainly focused on placement of the middle column stub until the complete failure of the
the corner column [12,29] and interior column removal scenarios specimen.
[13,15]. The results show that slabs can significantly enhance the The parameters studied in the test program include: (1) slab thick-
progressive collapse resistance compared with skeleton beam-column ness, (2) beam depth, and (3) seismic design intensity. Note that both X-
substructures. This is because slabs can contribute flexural resistance and Y- beams had the same depth in each specimen. Geometric di-
and compressive membrane action at small deformation and tensile mensions of all the specimens are listed in Table 1. The reference beam-
membrane action (TMA) at large deformation based on proper slab specimen was SE1, and the companion beam-column substructure
boundary conditions. was BE1. On top of SE1, the slab of SE2 was thickened, and the beams of
Recently, the research on RC beam-slab substructures has been ex- SE3 were designed with a greater depth. Specimens SE1-3 were de-
tended to perimeter column removal scenarios [14,16,30]. Dat and Tan signed under seismic design intensity of 6 degree with peak ground
[14] tested three 1/3 scaled RC beam-slab substructures under a pe- acceleration (PGA) of 0.05 g, and SE4 under intensity of 8 degree with
nultimate-perimeter column removal scenario, and investigated the PGA of 0.2 g, resulting in larger beam reinforcement ratio. The slab top
effects of the amounts of beam stirrups in plastic hinge length, the as- reinforcement of SE1, 3 and 4 is shown in right side of Fig. 1, and the
pect ratios of slab panels and the amounts of slab longitudinal re- one in “Zones a and d” of SE2 was changed to ϕ6@140 and ϕ6@180,
inforcement above the removed column. Qian et al. [30] tested two 1/4 respectively. Moreover, the bottom slab reinforcement of all the beam-
scaled RC beam-slab substructures. In one specimen, a penultimate- slab specimens is ϕ6@200 in both x- and y-directions. For all the spe-
perimeter column removal scenario was considered, while in the other cimens, the concrete strength of standard concrete cubes was 43 MPa,
one, the simultaneous removal of a penultimate-perimeter column and equivalent to cylinder compressive strength of 34 MPa. The yield
a corner column was studied. The above two groups of results indicated strength and ultimate strength of the longitudinal reinforcement are
that despite no lateral restraints, the beam-slab substructures could still around 350 MPa and 500 MPa, respectively. The stirrups were made of
develop TMA to enhance progressive collapse resistance beyond the one ϕ4 with ultimate strength of 825 MPa. More model details about the
predicted by the yield-line theory. Lu et al. [16] tested four 1/3 scaled test program can refer to the paper [16].
RC beam-slab substructures under a middle-perimeter column removal
scenario, and the adequate lateral restraints were provided at the other 2.2. Finite element modeling
three edges of each specimen. Both compressive arch action and TMA
were evidently identified based on the load-deflection curves. More- To fully understand the load transfer mechanisms of RC beam-slab
over, it is confirmed that slabs could significantly enhance the pro- substructures subjected to perimeter column removal scenarios and the
gressive collapse resistance of structures. effect of relevant parameters, numerical investigation is employed with
Typically, the experiments on the progressive collapse resistance of commercial software Ls-dyna. Although only the quasi-static behavior
RC beam-slab substructures are expensive and thus the number of tests is concerned here, explicit finite element method (FEM) is used to avoid
is limited. Accordingly, the information is not adequate to showcase the the convergence difficulties of conventional implicit FEM. The numer-
load transfer mechanisms of beam-slab substructures against pro- ical models were built exactly the same as specimens SE1-SE4 and BE1,
gressive collapse and the effects of pertinent geometric and structural and validated by the corresponding experimental results. For example,
parameters. Moreover, perimeter columns are more likely to be re- the specimen and the reinforcement detailing as shown in Fig. 3 is
moved than the other columns due to accessibility. Therefore, in this identical to those of specimen SE1. The boundary beams are modeled as
paper high fidelity finite element (FE) numerical modeling is adopted to well to achieve the similar boundary conditions as the tests.
investigate the robustness of RC beam-slab substructures under peri-
meter column removal scenarios, which can provide the nonlinear be- 2.2.1. Element types
havior of the substructures in great detail. After the introduction and In the numerical modeling, as shown in Fig. 3, concrete part is
validation of the numerical modeling, the FE model is further used to modeled using 8-node solid elements with reducing integration scheme.
explore the load transfer mechanisms of beam-slab substructures There is only one integration point in each solid element. Based on the
against progressive collapse, the effects of loading schemes and the mesh sensitivity analysis, the mesh size of 12 ∼ 20 mm is used, in
axial restraint stiffness at substructure boundaries. Finally, the numer- which the smaller element size is used in the slab strips connected to the
ical models are employed to illustrate the contribution of beams and column stub and the larger element size is used in the other regions of
slabs to the progressive collapse resistance of the overall beam-slab RC beam-slab substructures, as shown in Fig. 3. The combination of
substructures, the effects of slab thickness and the slab reinforcement mixed element sizes is used to reduce the number of elements and in-
detailing on the substructure resistance. crease computational efficiency.
All the reinforcement is modeled using 2-node Hughes-Liu beam
2. Experiment and numerical validation on progressive collapse of elements with 2 × 2 Gauss quadrature integration at the cross-section.
RC beam-slab substructures As the beam element length is key to determine the computational time,
the element length of 20 mm to 25 mm is eventually chosen, which is
2.1. Brief of the experiments larger than the size of surrounding concrete elements.
Except the beam-column connections, where the development of
The FE numerical model is validated by the test data conducted by wide cracks can be clearly expected and the force transfer is mainly in
Lu et al. [16]. They tested four 1/3 scaled RC beam-slab substructures one direction, the bond between the reinforcement and concrete in the
with an additional companion beam-column substructure under a rest regions of a beam-slab substructure is assumed to be perfect. This is
perimeter middle column removal scenario. As shown in Fig. 1, each because the force transfer in the rest regions is complicated and varying
beam-slab specimen consisted of two-bay beam in longitudinal direc- in x- and y-directions, resulting in hard to predict the directions of
tion (i.e. X-beam), one-bay beam in transverse direction (i.e. Y-beam), cracks. Moreover, the slabs are typically thin, and the strain con-
two slab panels and three size-enlarged boundary beams. The three centration of reinforcement is not as severe as that in beams. Therefore,
boundary beams are fully fixed onto the concrete bases, which were perfect bond is tentatively assumed by using the keyword
anchored onto the strong floor, to provide adequate boundary restraints ∗CONSTRAINED_LAGRANGE_IN_SOLID in Ls-dyna to connect
15
J. Yu et al. Engineering Structures 159 (2018) 14–27
Boundary beam 3
550 370
Zone c
6@140
Boundary beam 1
1000
Y-beam
Slab panel
6@140 @120
6@200
Middle column
1000
Zone d
550
stub
Y
X-beam 85 550
X 550 470
Zone b Zone a
0.5
⎧ τbmax (s / s1) for s ⩽ s1
τb =
⎨= τbmax = 0.3 fc for s > s1 (1)
⎩
Lifting beams
Boundary where s1 = 0.1 mm and fc = compressive strength of standard cylin-
beam 2
ders, which is 34 MPa in the tests. Accordingly, τbmax = 1.75 MPa.
In Ls-dyna, the bond relationship used for Contact_1D is assumed to
Upper actuator be elastic-perfectly-plastic. After reaching the maximum shear stress,
Boundary
beam 1 the bond stress could decay following an exponential damage curve.
Column stub The constitutive law of shear stress τ and local slip s is given by
Table 1
Cross-sectional dimensions and reinforcement of the beams in the specimens (unit: mm) [16].
Specimen Beam depth Beam width Slab thickness X-direction (Top) X-direction (Bottom) Y-direction (Top) Y-direction (Bottom)
SE1 170 85 50 2ϕ6 + 1ϕ8 2ϕ6 2ϕ6 2ϕ8 + 1ϕ6 2ϕ8 2ϕ8
SE2 170 85 75 2ϕ6 + 1ϕ8 2ϕ6 2ϕ6 2ϕ8 + 1ϕ6 2ϕ8 2ϕ8
SE3 200 85 50 2ϕ6 + 1ϕ8 2ϕ6 2ϕ6 2ϕ8 + 1ϕ6 2ϕ8 2ϕ8
SE4 170 85 50 3ϕ10 2ϕ10 2ϕ10 2ϕ12 + 1ϕ10 2ϕ12 2ϕ12
BE1 170 85 N.A. 2ϕ6 + 1ϕ8 2ϕ6 2ϕ6 2ϕ8 + 1ϕ6 2ϕ8 2ϕ8
16
J. Yu et al. Engineering Structures 159 (2018) 14–27
-30
2.2.4. Material modeling
Continuous surface cap model (CSCM) is used for concrete material -35
properties. The failure surface of CSCM is defined by three-invariants -0.010 -0.008 -0.006 -0.004 -0.002 0.000 0.002 0.004
together with a cap hardening parameter [33]. CSCM can model da- Strain
mage-based softening and modulus reduction, shear dilation, shear
compaction, confinement effect and strain rate effect. Moreover, the Fig. 4. Unconfined uniaxial stress-strain relationship of concrete based on CSCM (Positive
stress for tension and negative stress for compression).
default model parameters converted from the input are for normal
strength concrete, with compressive strength ranging from 28 MPa to
58 MPa. strain should be provided. When the strain exceeds the ultimate elon-
The simplified version only requires the input parameters of un- gation ratio, the corresponding reinforcement element is to be deleted.
confined compressive strength fc, erosion criterion, maximum aggregate The detailed material properties of reinforcing bars with different dia-
size Ag and units. In the tests of SE1-SE4 and BE1, fc was 34 MPa. Since meters can refer to the paper [16].
the slabs of the substructures were very thin, Ag is taken as 5 mm. The With the default concrete model parameters, the numerical results
subroutine of the CSCM calculates the default model parameters as are stiffer and larger than the experimental results, as shown in Fig. 5. It
functions of fc. Ag is used for determining the fracture energy para- is noted that the compressive strength is reached at the strain of around
meters of CSCM in accordance with CEB-FIP concrete model code1990 0.001. However, normal strength concrete typically attains the com-
[34]. Based on the default CSCM, the unconfined uniaxial compressive pressive strength at the strain of 0.002. As a result, the elastic modulus
(UUC) and tensile (UUT) stress-strain relationship of concrete is shown of concrete can be reduced to fit the strain of 0.002 corresponding to
in Fig. 4. It can be seen that for each branch, the constitutive curve compressive strength, as shown in Fig. 4, and the rest parameters are
consists of a linear pre-peak part and a softening post-peak part. determined based on fc and Ag, which were introduced in the user
When the severe cracking and failure of concrete cause the concrete manual of CSCM [33]. This adjustment improves the numerical results
separation, the maximum principal strain εmax is a proper criterion for for BE1 and SE1 but still over-predicted.
erosion algorithm. By this criterion, the solid elements with concrete When it comes to over-prediction, the model developer suggested
properties will be deleted if εmax exceeds the principal strain at failure that the tensile fracture energy Gft, corresponding to the area under the
εf. The value of εf is mesh-dependent, and taken as 0.1 for element size softening constitutive curve of UUT, could be reduced to 80% of the
between 12 and 20 mm in accordance with the previous work on RC default one, and compressive fracture energy Gfc = 50Gft or shear
beam-column substructures under column removal scenarios [31]. fracture energy Gfs = 0.5Gft is a reasonable reduction as well if com-
A symmetric bilinear elastic-plastic model ∗Mat_Plastic_Kinematic pression or shear damage is evident [35]. Note that the default model
(MAT_003) is used for steel reinforcement. In this model, the elastic assumes Gfc = 100Gft and Gfs = Gft. Since the compressive failure of
modulus, yield strength, tangential modulus after yielding and failure concrete was severe at the beam ends and inclined cracks occurred at
17
J. Yu et al. Engineering Structures 159 (2018) 14–27
30
Fig. 7 shows that the numerical resistance of the three specimens
SE2-4 is in good agreement with the experimental ones. The essential
20 characteristics of the resistance curves, including the ascending, the
softening and the re-ascending part, are well captured. The difference of
the predicted and the experimental peak capacity is less than 10%. The
10
sudden reduction of structural resistance is attributed to the fracture of
reinforcing bars. As there are many reinforcement in the RC beam-slab
Numerical model
0 substructures, it is difficult to accurately predict each fracture of the bar
0 100 200 300 400 500 600 700 the at large deflection. However, Park [36] suggested that the deflec-
Vertical displacemnt of column stub (mm) tion equal to 10% the short span of a single-panel two-way slab can be
used to determine capacity of tensile membrane action. By analogy, the
(a) Specimen BE1 deflection (i.e. the vertical displacement of the column stub) equal to
10% the longitudinal span of the substructure can be used as the de-
120 flection limit. At this deflection, the prediction also agrees well with the
experimental results.
100
Crack pattern of RC slabs at the failure stage cannot be directly
demonstrated because concrete model CSCM is unable to track cracks.
However, the cracks can be equivalently shown by effective plastic
80 strains [37]. Typically, wider cracks result in larger effective plastic
strains. For simplicity, only the crack pattern of SE1 is compared with
Load (kN)
the experimental one, with top and bottom of the slab shown in
60
Fig. 8(a) and (b), respectively. It is seen that the area with the dis-
tribution of extensive cracks and the direction of cracks agree well with
40 the experimental ones, respectively. Fig. 9(a) and (b) compare the
failure modes and crack patterns of longitudinal and transverse beams,
Experimental results
Results w/ default parameters respectively. The numerical results can effectively show the wide cracks
20 Results w/ reduced elastic modulus at the bar cut-off position with erosion of concrete elements in the
Results w/ reduced elastic modulus longitudinal beams. Also, the effective plastic strain can well indicate
& fracture enenrgy
the failure at the transverse beam end due to concrete crushing.
0
0 100 200 300 400 500 600 700 In summary, the agreement of the structural resistance curve and
the failure mode indicates the validity of the numerical model. The
Vertical displacement of column stub (mm)
model is then to be extended to investigate the load transfer mechan-
(b) Specimen SE1 isms of the RC beam-slab substructures and the effects of pertinent
parameters.
Fig. 5. Effect of concrete modeling parameters of CSCM on the numerical results.
18
J. Yu et al. Engineering Structures 159 (2018) 14–27
that at a small deflection (e.g. 28 mm) the large axial stress of SE1a is
mainly concentrated at the beam reinforcement and the adjacent slab
40 reinforcement in the column stub and the beam end regions. At the
column stub region, the top and the bottom reinforcement of the
longitudinal beam is in compression and tension, indicating the de-
20 velopment of sagging moment, whereas the stress of the reinforcement
in the transverse beam is quite low. At the far ends of longitudinal and
transverse beams, the top and bottom reinforcement is in tension and
0 compression, indicating the development of hogging moment. There-
0 100 200 300 400 500 600 700
fore, at this stage the structural resistance is primarily contributed by
Vertical displacement of column stub (mm) the beams through flexural resistance.
(a) Specimen SE2 The extensive cracks developed through the slab of specimen SE1 at
failure, as shown in Fig. 8, indicating a large proportion of the slab
120
Experimental results except two corners surrounded by boundary beams is mobilized into
Numerical results tensile membrane action (TMA) state. Fig. 10(b) demonstrates that
100 except the slab reinforcement in the two corners far from the perimeter
beam, most slab reinforcement (both at the top and bottom) are in
tension at the deflection of 626 mm, confirming the dominant devel-
80
opment of TMA.
Load (kN)
that the sections too near the column stub or the beam ends experience
element erosion at large deflection and fail to provide the internal
60 forces over the entire loading history.
For simplicity, specimens BE1, SE1 and SE3 are used for typical
40 demonstration, with numerical models designated as BE1a, SE1a and
SE3a, respectively. The comparison between BE1a and SE1a illustrates
20 the effects of the slabs on the load transfer mechanism of RC beams. The
comparison of SE1a and SE3a elucidates the effect of the beam depth.
0 Fig. 12 shows similar characteristics of axial force development at
0 100 200 300 400 500 600 cross-sections X-beam-L1 and -R1 of BE1a, SE1a and SE3a. The axial
Vertical displacment of column stub (mm) force initially is compression, indicating the mobilization of compres-
sive arch action (CAA), and then transitions into tension, reflecting the
(c) Specimen SE4 development of catenary action (CTA). Table 2 shows the critical states
Fig. 7. Comparisons of numerical and experimental results of RC beam-slab substructure. of the beam axial force in the development history. It should be pointed
out that CAA is a flexural mechanism of beams coupled with axial
compression, and CTA is a structural mechanism of beams with re-
column removal scenario, i.e., the load transfer mechanisms of RC
sistance provided by axial tension.
beam-slab substructures against progressive collapse. However, the
The comparison between Fig. 12(a) and (b) suggests that the pre-
monitoring of internal forces or global stress distribution in beam-slab
sence of slabs significantly weakens CAA of longitudinal beams near the
substructures during testing is very difficult. On the contrast, the
column stub (e.g. peak axial compression of 140 kN reduced to 25 kN),
19
J. Yu et al. Engineering Structures 159 (2018) 14–27
and makes the beams evolve into CTA at a much smaller deflection and Fig. 13 shows the development of the shear forces at the cross-
develop larger ultimate axial tension. This suggests that near the sections near the column stub. All the shear forces are contributed to
column stub region of the beam-slab substructure, CTA is much more the structural resistance. It is seen that the shear forces of the long-
dominant in longitudinal direction. The comparison of itudinal beams in the three specimens increase first and then sig-
Fig. 12(b) and (c) shows that a larger beam depth in SE3a increases the nificantly decrease after attaining the peak value. The great reduction
peak compression of longitudinal beams, but has insignificant effect on in shear forces of the longitudinal beams is mainly due to the transition
the development of axial tension. This is because the amount of steel into axial tension as well as concrete crushing at compression zone of
reinforcement in SE1a and SE3a is the same. the far beam ends.
In the overall loading history, the transverse beam in BE1a is in Fig. 13(a) shows that without the slabs, the shear forces of the
slight tension with maximum value of 5 kN due to short of adequate longitudinal beams are much larger than that of the transverse beam.
axial restraints. In comparison, the presence of slabs in SE1a and SE3a Consequently, the decrease of structural resistance coincides with the
makes the transverse beam mainly in compression with peak value of decrease of the shear force of the longitudinal beams at small deflec-
around 20 kN due to the composite effect of T-beam section, but tions. Moreover, at large deflections, the increase of structural re-
eventually the axial force in the transverse beams of SE1a and SE3a sistance albeit the reduction of shear force suggests the great con-
decreases to around zero. tribution from CTA of the longitudinal beams.
A F
B E
Slab and beam C D
Slab and beam cracks merged at bar curtailment point cracks merged
20
J. Yu et al. Engineering Structures 159 (2018) 14–27
21
J. Yu et al. Engineering Structures 159 (2018) 14–27
Fig. 14(b) shows that for the longitudinal beam, the sections near
Resistance (kN)
the column stub and the far ends are mainly in sagging and hogging
0 0
moment, respectively. However, after attaining the peaks or deflecting
one beam depth, the moments at the sections X-beam-R1, -R2 and -R3
-50 -25 keep decreasing as increasing the deflection. Due to the composite ef-
X-beam-L1 fect, sections throughout most of the transverse beam are subjected to a
X-beam-R1
Y-beam-1 hogging moment.
Resistance
-100 -50 Combining Fig. 14(a) and (b), it is observed that after the deflection
of one beam depth, at the central region near the column stub, the
bending moment is small, and the axial tension is dominant in long-
-150 -75 itudinal direction. At the beam ends adjacent to the boundaries, the
0 100 200 300 400 500 600 hogging moment is prevailing even at very large deflection (e.g.
Vertical displacement of column stub (mm) 600 mm). In particular, the composite effect between slabs and beams
(a) Specimen BE1a enhance the contribution of transverse beam to the structural resistance
of beam-slab substructure.
80 120
X-beam-L1 4. Effect of boundary and loading approaches on progressive
X-beam-R1
Y-beam-1 collapse resistance
60 90
Resistance
Beam axial force (kN)
40 60 substructure
For brevity, models BE1a and SE1a are used for investigating the
40 60
Resistance (kN)
22
J. Yu et al. Engineering Structures 159 (2018) 14–27
Table 2
Critical states of the beam axial force at cross-section X-beam-R1.
Specimens Peak axial compression Displacement at the onset of axial tension (mm) Peak axial tension
stiffness, the first peak capacity of BE1a and SE1a decreases due to performance of beam-slab substructures but takes more costs in prac-
weakened CAA, and the structural resistance of BE1a and SE1a at large tice.
deflection stay almost the same. When there are no lateral restraints,
the structural resistance of BE1a significantly decreases due to the ab- 5. Effect of structural parameters on progressive collapse
sence of CTA at large deflection, whereas the structural resistance of resistance
SE1a only drops around 25% at the deflection of 400 mm. The reason is
that laterally unrestrained slabs can still develop TMA with a self- 5.1. Effect of slab
equilibrium mechanism [14]. In a word, the development of TMA of RC
beam-slab substructures is insensitive to the lateral restraint stiffness at The research on RC beam-slab substructures against progressive
the boundaries. When the relative lateral restraint stiffness is not less collapse illustrates the load transfer mechanism of the substructures
than 0.1, RC beam-slab substructures show similar structural resistance after a column removal, and also demonstrates the contribution of the
at large deflections. beams and slabs to the progressive collapse resistance, as shown in
Fig. 18. The ratio the beam-column substructure resistance to the cor-
4.2. Effect of loading approaches responding beam-slab substructure resistance is used for estimating the
contributions from the beams to the entire substructure. Through this
For quasi-static tests on RC structures against progressive collapse, way, when the deflection is less than 30 mm, the resistance contribution
there are three types of loading schemes, including concentrated of the beams can reach around 50–60%, but with increasing the de-
loading, equivalent uniformly distributed loading (UDL) and preloading flection, the contribution of the beams keeps decreasing until the onset
UDL combined with unloading at a removed column. With concentrated of catenary action. After the deflection of 30 mm, the resistance con-
loading, a monotonic load is directly applied at the top of the column tribution of the beams ranges from 30% to 50%.
stub [16]. With an equivalent UDL, a loading tree is used to apply 12 In progressive collapse analysis of framed structures, it is quite often
equal concentrated loads at the top surface of the beam-slab sub- to convert slabs as effective flanges of beams [40,41]. Consequently,
structure [14]. With the last scheme, a UDL is preloaded until the load model SE1a is further simplified into a beam-slab substructure (desig-
is balanced, and thereafter one column is unloaded to simulate a re- nated as BE1b) with composite T-Section and L-Section in the trans-
moved column until a new balance is achieved at a certain deflection verse and longitudinal beams, respectively, as shown in the inset of
[38]. The last scheme is the most realistic but can only provide the Fig. 18. In accordance with Chinese Code for Design of Concrete
applied load and the corresponding deflection. In comparison, the first Structures [42], the width of the flange in the T- and L-Sections is
two simplified schemes are more easily to derive structural resistance. 666 mm and 333 mm, respectively. Fig. 18 demonstrates that the
For an equivalent UDL, the loading tree made of steel members is structural resistance of BE1b has the same characteristics as those of
explicitly modeled using rigid materials, as demonstrated in the inset of BE1a, such as the softening resistance after the peak capacity and the
Fig. 17. More modeling details can refer to the paper [39]. The beauty re-ascending resistance at catenary mechanism. As expected, the
of using equivalent UDL rather than UDL is that displacement-con- structural resistance of BE1b is between BE1a and SE1a, indicating that
trolled loading can still be used, so that softening resistance can be the slab strips enhance the overall structural resistance compared with
captured if any. For the third loading scheme, a rigid pad is modeled BE1a. Moreover, compared with SE1a, the resistance contribution of
and fixed beneath the column stub and a contact surface is set between BE1b mainly varies from 60% to 70%; when the deflection is less than
the pad and the stub. After a given UDL is applied and balanced, the pad 30 mm, BE1b provides 70–90% resistance of the substructure. There-
is slowly moved downwards in vertical direction until the separation fore, converting slabs into beam flanges is feasible for structural ana-
with the stub and further the stoppage of the stub movement. The lysis at small deflection but too conservative at large deflection.
eventual displacement of the stub is the ultimate deflection for the
given UDL. For simplicity, only five given UDL are used in total herein. 5.2. Effect of slab thickness
Fig. 17 shows that the concentrated load and the equivalent UDL
schemes mobilize the structural resistance with similar characteristics In the reference tests, the slab thickness of SE2 (75 mm) was 50%
(i.e. the ascending, softening and re-ascending resistance) but the greater than that of SE1 (50 mm), and the amount of reinforcement
concentrated loading scheme significantly underestimates the struc- used in SE2 was 43% more than the one in SE1. The comparison of the
tural resistance, around 1/3.5 to 1/4.5 the one under the UDL scheme. experimental results of them, as shown in Fig. 19, indicates that the slab
Fig. 17 suggests that the equivalent UDL scheme provides the similar thickness has little effect on the stiffness of ascending and the first peak
load-deflection pairs as the third loading scheme, but the deflection is resistance at beam mechanism stage, whereas with increasing the de-
slightly smaller for a given UDL at large deformation stage. This is flection, the difference of the resistance of SE1 and SE2 stays around
because with increasing deflection the loading points near the peri- 10 kN. Prior to the first bar fracture of SE2 at the deflection of 366 mm,
meter beam, which deforms more than the other regions, slightly take the resistance of SE2 was 13% larger than that of SE1. The insignificant
less load than the other points. effect of the slab thickness and the corresponding reinforcement is
In a word, the concentrated loading scheme is unable to fully ex- partly attributed to the concentrated loading scheme, with which only
plore the structural resistance of beam-slab substructures, but is still the reinforcement adjacent to the beams are fully mobilized.
feasible to represent the essential load transfer mechanisms against To re-evaluate the effect the slab thickness, the equivalent UDL
progressive collapse. In comparison, the equivalent UDL is a more re- loading scheme is employed. The comparison of the numerical results
presentative loading scheme to demonstrate progressive collapse as shown in Fig. 19 suggests that the difference of the resistance is
23
J. Yu et al. Engineering Structures 159 (2018) 14–27
120
Section: X-beam-L1
16 Section: X-beam-R1 48 100 Deflection of one beam depth
Section: Y-beam-1
Resistance 96
Resistance (kN)
Shear force (kN)
Resistance (kN)
72
-100
8 24
48
-200
4 12 24
X-beam-R1 Y-beam-1
X-beam-R2 Y-beam-2
-300 X-beam-R3 Y-beam-3
Resistance
0 0 0
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
Vertical displacement of column stub (mm) Vertical displacement of column stub (mm)
80
Resistance (kN)
0
15 72
-5 60
10 48
-10 40
5 24
-15 20
X-beam-R1 Y-beam-1
X-beam-R2 Y-beam-2
X-beam-R3 Y-beam-3
Resistance
0 0 -20 0
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
Vertical displacement of column stub (mm) Vertical displacement of column stub (mm)
80 The design guideline UFC 4-023-03 requires that the ties used to
Shear force (kN)
20
Resistance (kN)
24
J. Yu et al. Engineering Structures 159 (2018) 14–27
RC slab
Steel
plate
Rigid plates
Restraint
springs
30 240
Load (kN)
20
160 Concentrated loading
Equivalent UDL w/ 12 points
UDL w/ unloading at removed
10 column
80
0
0 100 200 300 400 500 600 700
0
Vertical displacement of column stub (mm) 0 100 200 300 400 500 600 700
(a) Model BE1a Vertical displacement of the column stub (mm)
120
Restrained by boundary beams Fig. 17. Effect of loading approach on the structural resistance of RC beam-slab sub-
Rigid structure.
ks=100kN/mm (ra=1.0)
100
ks=50kN/mm (ra=0.5)
120 1.0
ks=10kN/mm (ra=0.1)
80 ks=0
96
60 Numerical model to beam-slab assembley
72 0.6
Load (kN)
40
20 48 0.4
25
J. Yu et al. Engineering Structures 159 (2018) 14–27
200 500 include lateral restraint stiffness at the substructure boundaries, loading
schemes in testing, slab thickness and reinforcement detailing in slabs.
The main conclusions drawn from the numerical investigations are as
160 400
follows:
75
Acknowledgement
50 125
SE1a under point loading The authors would like to acknowledge the final support by the
25 SE1b under point loading
SE1a under equivalent UDL National Natural Science Foundation of China (No. 51408189); Natural
SE1b under equivalent UDL Science Foundation of Jiangsu Province of China (No. BK20140855);
0 0
0 100 200 300 400 500 600 700 and the Fundamental Research Funds for the Central Universities (No.
Vertical displacement of column stub (mm) 2017B41814).
Fig. 20. Effect of additional continuous top slab reinforcement along the perimeter beam.
26
J. Yu et al. Engineering Structures 159 (2018) 14–27
References [21] Xiao Y, Kunnath S, Li FW, Zhao YB, Lew HS, Bao Y. Collapse test of three-story half-
scale reinforced concrete frame building. ACI Struct J 2015;112:429–38.
[22] Yu J, Rinder T, Stolz A, Tan KH, Riedel W. Dynamic progressive collapse of an RC
[1] American Society of Civil Engineers (ASCE). Minimum Design Loads for Buildings assemblage induced by contact detonation. J Struct Eng 2014;140:04014014.
and Other Structures. ASCE/SEI 7–10. Reston (VA). Reston (VA): American Society [23] Pham AT, Tan KH. Experimental study on dynamic responses of reinforced concrete
of Civil Engineers; 2010. frames under sudden column removal applying concentrated loading. Eng Struct
[2] European Committee for Standardization. EN 1991-1-7: Eurocode 1 - Actions on 2017;139:31–45.
structures - Part 1-7: General actions - Accidental actions. Brussels: CEN; 2006. [24] Izzuddin BA, Vlassis AG, Elghazouli AY, Nethercot DA. Progressive collapse of
[3] General Services Administration (GSA). Progressive collapse analysis and design multi-storey buildings due to sudden column loss – Part I: simplified assessment
guidelines for new federal office buildings and major modernization projects; 2003. framework. Eng Struct 2008;30:1308–18.
[4] Department of Defense (DOD). UFC 4–023-03, Design of Buildings to Resist [25] Li Y, Lu X, Guan H, Ye L. Progressive collapse resistance demand of reinforced
Progressive Collapse. Including Change 2–1 June 2013; 2013. concrete frames under catenary mechanism. ACI Struct J 2014;111:1225–34.
[5] Qian K, Li B. Performance of three-dimensional reinforced concrete beam-column [26] Tsai M-H, You Z-K. Experimental evaluation of inelastic dynamic amplification
substructures under loss of a corner column scenario. J Struct Eng factors for progressive collapse analysis under sudden support loss. Mech Res
2013;139:584–94. Commun 2012;40:56–62.
[6] Yu J, Tan KH. Structural behavior of reinforced concrete beam-column sub- [27] Tsai M-H. An analytical methodology for the dynamic amplification factor in pro-
assemblages under a middle column removal scenario. J Struct Eng gressive collapse evaluation of building structures. Mech Res Commun
2013;139:233–50. 2010;37:61–6.
[7] Yu J, Tan KH. Experimental and numerical investigation on progressive collapse [28] Yu J, Yin C, Guo Y. Nonlinear SDOF model for progressive collapse responses of
resistance of reinforced concrete beam column sub-assemblages. Eng Struct structures with consideration of viscous damping. J Eng Mech 2017;143:04017108.
2013;55:90–106. [29] Qian K, Li B. Slab effects on response of reinforced concrete substructures after loss
[8] Kang S-B, Tan KH. Behaviour of precast concrete beam–column sub-assemblages of corner column. ACI Struct J 2012;109:845–55.
subject to column removal. Eng Struct 2015;93:85–96. [30] Qian K, Li B, Zhang Z. Influence of multicolumn removal on the behavior of RC
[9] Yu J, Tan KH. Structural behavior of reinforced concrete frames subjected to pro- floors. J Struct Eng 2016;142:04016006.
gressive collapse. ACI Struct J 2017;114:63–74. [31] Pham AT, Tan KH, Yu J. Numerical investigations on static and dynamic responses
[10] FarhangVesali N, Valipour H, Samali B, Foster S. Development of arching action in of reinforced concrete sub-assemblages under progressive collapse. Eng Struct
longitudinally-restrained reinforced concrete beams. Constr Build Mater 2016.
2013;47:7–19. [32] fib. fib Model Code for Concrete Structures 2010. Ernst & Sohn; 2013.
[11] Lin Kaiqi, Li Yi, Lu Xinzheng, Guan H. Effects of seismic and progressive collase [33] Murray YD. Users manual for LS-DYNA concrete material model 159. Colorado:
designs on the vulnerability of RC frame structures. J Perform Constr Facil 2016. APTEK, Inc.; 2007. p. 92.
04016079-1-11. [34] CEB. CEB-FIP model code 1990. Thomas Telford; 1991.
[12] Kai Q, Li B. Dynamic performance of RC beam-column substructures under the [35] Murray YD, Abu-Odeh A, Bligh R. Evaluation of Ls-dyna concrete material model
scenario of the loss of a corner column—experimental results. Eng Struct 159. APTEK, Inc.; Texas Transportation Institute; 2007. p. 209.
2012;42:154–67. [36] Park R. The ultimate strength of uniformly loaded laterally restrained rectangular
[13] Pham XD, Tan KH. Experimental study of beam–slab substructures subjected to a two-way concrete slabs. University of Bristol; 1964.
penultimate-internal column loss. Eng Struct 2013;55:2–15. [37] Lew HS, Bao Y, Sadek F, Main JA, Pujol S, Sozen MA. An Experimental and com-
[14] Dat PX, Tan KH. Experimental response of beam-slab substructures subject to pe- putational study of reinforced concrete assemblies under a column removal sce-
nultimate-external column removal. J Struct Eng 2015;141:04014170. nario. NIST Technical Note 1720. National Institute of Standards and Technology,
[15] Qian K, Li B, Ma JX. Load-carrying mechanism to resist progressive collapse of RC U.S. Department of Commerce; 2011. p. 106.
buildings. J Struct Eng 2015;141:04014107. [38] Yi W-J, Zhang F-Z, Kunnath S. Progressive collapse performance of RC flat plate
[16] Lu X, Lin K, Li Y, Guan H, Ren P, Zhou Y. Experimental investigation of RC beam- frame structures. J Struct Eng 2014;140:04014048.
slab substructures against progressive collapse subject to an edge-column-removal [39] Yu J, Luo L, Ge C. Numerical investigation on structural behavior of RC beam-slab
scenario. Eng Struct 2017;149:91–103. assemblies under an exterior column removal scenario; 2017.
[17] Yi WJ, He QF, Xiao Y, Kunnath SK. Experimental study on progressive collapse- [40] Li Y, Lu X, Guan H, Ren P, Qian L. Probability-based progressive collapse-resistant
resistant behavior of reinforced concrete frame structures. ACI Struct J assessment for reinforced concrete frame structures. Adv Struct Eng
2008;105:433–9. 2016;19:1723–35.
[18] Shan S, Li S, Xu S, Xie L. Experimental study on the progressive collapse perfor- [41] Sasani M. Response of a reinforced concrete infilled-frame structure to removal of
mance of RC frames with infill walls. Eng Struct 2016;111:80–92. two adjacent columns. Eng Struct 2008;30:2478–91.
[19] Qian K, Li B. Effects of masonry infill wall on the performance of RC frames to resist [42] Ministry of Housing and Urban-Rural Development of the People's Republic of
progressive collapse. J Struct Eng 2017;143:04017118. China (MOHURD). Code for design of concrete structures. GB50010-2010. Beijing,
[20] Shan S, Li S, Zhai C, Xie L. Nonlinear dynamic analysis on the PC response of an RC China; 2010.
frame with perforated infill walls. In: S JG, editor. Structures congress 2017.
Denver: ASCE; 2017. p. 119–29.
27