Professional Documents
Culture Documents
1 s2.0 S2352012419300967 Main
1 s2.0 S2352012419300967 Main
1 s2.0 S2352012419300967 Main
Structures
journal homepage: www.elsevier.com/locate/structures
School of Civil & Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore
Keywords: In this paper, the authors assess the robustness of steel-concrete composite floor systems subjected to a Corner
Progressive collapse Column (CC) removal scenario based on numerical simulations. First, a modelling method is verified by previous
Column removal actual test results conducted by the authors. Afterwards, using the reliable modelling method a composite floor
Composite floor systems system subjected to a corner column loss is simulated, producing the static load-displacement curve, the failure
Dynamic behaviour
mode and load-transfer mechanisms. These results are compared with those of composite floor systems under an
Numerical study
Internal Column (IC) removal scenario. Besides, the model is applied to simulate that the same composite floor
system experiences a sudden column removal under six levels of loads. The dynamic displacement-time histories
under all levels of loads are obtained, from which the loads and corresponding maximum deflections constitute
the dynamic load-deflection response. They are used to evaluate the dynamic load capacity and deformability of
this composite floor system subjected to a sudden CC removal scenario. Lastly, Dynamic Increase Factors (DIFs)
are obtained through comparing the dynamic load-deflection response with the quasi-static one, which are
further compared with DIFs under IC scenario.
⁎
Corresponding author.
E-mail address: fuqi0002@e.ntu.edu.sg (Q. Fu).
https://doi.org/10.1016/j.istruc.2019.06.003
Received 4 January 2019; Received in revised form 25 May 2019; Accepted 7 June 2019
Available online 17 June 2019
2352-0124/ © 2019 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
Restraint 155 30
500
65
65
C8
40
C5 Beam C1
75 75 92 93
Girder: H 200x100x5.5x8
Column:
Girder B2
Loading point Secondary beam
1000
H200x200x8x12
Girder
B8 B5
Inner beam:H150x75x5x7
Removed Fig. 3. Configurations of composite slabs.
1000
Column
Girder
C7 B7 B4 C2 the one under a CC removal scenario but was not successfully carried
Beam:H150x75x5x7 out. Therefore, there is still an absence of valid actual test results of
composite floor systems under a corner column removal scenario. Be-
Girder
1000
Girder
sides, although one composite floor system [4] was tested under a
B1
B6 B3 sudden column removal scenario, they did not investigate DIF and the
dynamic behaviour at large deformation. Hence, the authors aim to
Y 250 1500 250 conduct actual tests to investigate the static and dynamic behaviour of
1000
X composite floor systems subjected to a CC removal scenario, and to
C6 C4 C3 study DIF. Before the commencement of realistic 3D static and dynamic
tests, numerical predictions are carried out in this paper.
500
2. Verification of FE modelling method
500 2000 2000 500
revealed owing to the slightly damage of the structure. 2.1. Referenced test under IC removal scenario
Johnson et al. [7] tested a half-scale composite floor system sub-
jected to different column loss scenarios. During the test, the load ca- Considering the high cost of specimens and lab space constraint, the
pacity was unexpectedly low under the CC removal scenario, and the test specimen was one-third scaled, with 2 m by 2 m for each bay
specimen could not reach equilibrium when applying the first level of (Fig. 1) and 1.4 m for storey height. Fig. 1 shows the structural layout of
load. Consequently, the load-deflection curve was not obtained for the the specimen, with girders (primary beams) connecting to column
CC removal scenario. flanges in the y-direction and secondary beams (including beam and
Fu et al. [8,9] tested four composite floor systems under an IC re- inner beam in Fig. 1) to column webs or girder webs in the x-direction.
moval scenario. As a result, they have studied load-resisting mechan- H-shaped steel sections were used for the columns, girders and sec-
isms and the effects of slab aspect ratio, degree of composite action and ondary beams, indicated in Figs. 1 and 2 which shows the joint con-
boundary condition on the behaviour in composite floor systems sub- figurations. The girder-to-column connections adopted flush end plates
jected to an IC removal scenario. with bolts, while the beam-to-column and inner beam-to-girder joints
From the above literature review, the test in Johnson et al. [7] was employed double angle-cleat connections. The four numbers following
Column stub
Column stub 140 Secondary beam H200x200x8x12
H200x200x8x12 30 30 40 H150x75x5x7
M16 (8.8S) 10 60 8
26
25
50 20
25 50 25
Secondary beam
100
50
H150x75x5x7 Stiffener
t=4mm
50
40 45 5
25
End plate
20 50
t=8 mm
Double angle
Girder L90x56x5
60 8
H200x100x5.5x8
34
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
40
35
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
(a) The bottom bolt fracture at 0.73hg (b) The top bolt fracture at 1.16hg
cross-sections (Figs. 4 and 5). In this way, trapezoidal profiled com- capture more precisely the behaviour of ribs in the transverse direction,
posite slabs can be modelled by quadrilateral shell elements (S4R) one can refer to Izzuddin et al. [11] which introduced additional rib
(Fig. 4), which can be assigned with a composite section containing freedoms associated with hierarchic displacement fields for the ribs.
various materials at different layers. Hence, reinforcement is defined as In the modelling, bilinear stress-strain relationship is adopted to
one part of a shell element section. The interactions between re- represent uniaxial material property for the steel reinforcement and
inforcement and concrete, such as bond slip and dowel action, can be profiled decking, of which fracture is controlled by ductile damage
simulated because tension stiffening of concrete is specified by means of criteria. The FE models in this study incorporate concrete with damaged
a post-failure stress-strain relation. Considering that the steel decking plasticity model, which assumes concrete with scalar (isotropic) da-
would delaminate from the RC slab at very small deformation, shell mage. One can find all the input information for materials from Fu [10].
elements of RC slab and steel decking are separately built ignoring bond Additionally, beam elements (B31) are used to connect composite slabs
but incorporating contact behaviour between them. For both “strong” to steel girders and beams (Fig. 4) such as shear studs in test specimens.
and “weak” sections, the steel decking layers are defined at the bottom The determination of properties of the “shear studs” has been in-
of the sections. Hence, the steel decking layer continues in the rib di- troduced by Fu [10] in detail.
rection but discontinues at the boundaries between the strong and the Subsequently, the authors conduct numerical simulations to com-
week strips in the direction normal to the ribs (transverse direction). To pare with the respective counterparts in the test.
36
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
(a) Failure of angles and steel decking (b) Force development of components in the
in the test beam-to-removed column connection
S-1 S-3 in the girder-to-removed-column connection, while Fig. 7 (c) plots the
1 component force versus displacement of the removed column. In the
Adjacent bay test, the bottom- and top-row components of the sagging moment
500
500
y Beam B4 C2
(Fig. 8 (b)). Similarly, the model can also predict failure of the top- and
2
middle-row bolts at the hogging end at a correct range of displacement,
Removed column S-2 as shown in Fig. 8 (a) and (b).
x 3000 750
Moreover, in the test, the fracturing process of angles in the beam-
Fig. 10. Structural layout and boundary conditions. to-removed column connection starting at 300 mm (1.4hb where hb is
the depth of composite beam including slab thickness, hb = 215 mm) to
400 mm (1.86hb) in the test is predicted by descending forces of the
2.3. Comparisons between numerical predictions and test results joint components commencing at 256 mm (1.19hb) to 407 mm (1.89hb),
as shown in Fig. 9 (a) and (b). Lastly, fracture of the steel decking at the
In this section, the load-deflection curves and failure modes are first final stage is also captured by the model (Fig. 9 (a) and (c)).
compared between numerical predictions and test results to validate the After the numerical comparison, the modelling method shows the
FE modelling before conducting further parametric analyses. Fig. 6 capability to predict the overall load-displacement response and cap-
shows that the FE model can predict the overall load-deflection re- ture the failure mode of structural components. The verified model is
sponse well. then used to study the corner column removal effects on composite
Fig. 7 (a) and (b) show the photographs of the fractured components floor systems under a corner column removal scenario.
37
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
Restraint
Restraint
Free edge
Free edge
Removed column
Restraint
Displacement
Removed column
Fixed Fixed
Fig. 10 shows that the structural layout of the analysed floor system
which is supported by three columns with the corner column removed.
It should be noted that steel member sections, joint configurations, slab
details and restraint properties remain the same as those of the tested
specimen [10] introduced in Section 2.
300 To model the boundary conditions, the three columns are fixed at
their feet and the floor is restrained at the two inner edges with the
250
other two edges remaining free (Figs. 10 and 11). Specifically, the re-
Vertical load (kN)
200 straints are simulated by axial springs with very high stiffness values in
150 the horizontal direction (Figs. 10 and 11) and rigid restraints in the
UDL vertical direction in the model. According to [10], the stiffness values of
100
12-point Springs S-1 and S-3 in Fig. 10 are assigned with 230 kN/mm and those
50 of Springs S-2 and S-4 are set to 36 kN/mm.
0 Firstly, the model is applied with static 12-point loads and
0 100 200 300 Uniformly Distributed Loads (UDL) (Fig. 11 (a) and (b)), respectively.
Deflection of removed column (mm) The load-deflection curves and failure modes under the two loading
scenarios are discussed and compared in Section 4.1. Secondly, using
Fig. 13. Comparison of load-deflection curves between UDL and 12-point the dynamic solver in Abaqus the model is analysed for simulating the
loading scenarios.
sudden corner column removal. Specifically, a UDL is applied on the
38
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
Fig. 14. Distribution of in-plane principal strain in steel decking at the peak load point [12].
0 4. Simulation results
Component force (kN)
C1 C5 C1
C5
Failure Failure
B2 B2
B4 B4
39
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
C5
C1
Failure
C2
Fig. 17. Distribution in-plane principal strain in steel decking when slab collapses.
and IC removal scenarios are different from each other. For the latter, it 4.1.3. Reactions at column feet
was the complete failure of joint components at the girder-to-removed The force in restraining spring S-1 with a greater stiffness (230 kN/
column and beam-to-removed column connections that govern the mm) is significantly smaller than that in S-4 with a smaller stiffness
collapse of the floor system, when joint components at the hogging (36 kN/mm). The reason is that column C5 possesses greater horizontal
moment ends of the double-span girder and the double-span beam were stiffness in B2-direction compared with that of C2 in B4-direction. As a
also severely damaged [8]. For IC removal scenario, CA was mobilised result, C5 resists most of the horizontal force transferred from girder B2,
in the double-span girder and the double-span beam, and TMA was thus reducing horizontal force transferred to S-1. In comparison, most
mobilised in the composite slab [8]. Consequently, composite floor of the reaction force in line with B4 is sustained by S-2 instead of C2, as
systems under an IC removal scenario can sustain greater loads and shown in Fig. 22, where RC5,y indicates that the horizontal reaction
have better ductility compared with those under a CC removal scenario, force at C5 is in the positive y-direction while RC2,x at C2 is in the
as shown in Fig. 18. It should be noted that the numerical modelling positive x-direction.
information under CC and IC scenarios shown in Fig. 18 is the same, the However, the numerical result shows that the force in S-1 is smaller
only difference lies in the column removal location. than that in S-2 (Fig. 21), which runs contrary to the observation in the
floor systems with an internal column missing [9] where Compressive
4.1.2. Horizontal deformations and restraining forces Arch Action (CAA) was mobilised in the double-span girder and the
Fig. 19 shows the deflected shapes of the floor system when the double-span beam. For specimens in Fu et al. [9], the much greater CAA
vertical displacement of the removed column reached around 50 mm, in girder B2 produced greater outward horizontal force compared to
100 mm, 150 mm and 200 mm. It can be seen that the deformed shape that in beam B4, so that the horizontal reaction at the end of B2 (cor-
of the girder and the beam were caused by cantilever action which responding to S-1 in the present model) is greater than that of B4
played an important role in transferring the applied load to adjacent (corresponding to S-2), although C5 was placed in a better orientation
bays. Hence, horizontal deformations and restraining forces at the far to resist horizontal force compared to C2.
ends of girder B2 (Point 1 in Fig. 10) and beam B4 (Point 2 in Fig. 10), Fig. 23 shows the development of reactions at the feet of column C2
as well as the reaction forces at the associated column feet (C5 and C2, and C5, where RC2,V and RC5,V represent the respective vertical reaction
respectively), are discussed in this section. Fig. 20 shows the develop- forces at C2 and C5; MC5,ma denotes the bending moment at the feet of
ment of horizontal deformations (H-1 and H-2) at respective Points 1 C5 about major-axis of the cross-section while MC2,mi denotes the
and 2. At the corresponding same locations, the development of forces bending moment at the feet of C2 about minor-axis of the cross-section.
in the horizontal restraining springs (S-1 and S-2) are shown in Fig. 21. Positive bending moments mean that the fibre in the flange which is
40
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
Fig. 19. Deformed shape of the floor system at different loading stages.
farther to the removed column is in tension. It can be seen that C5 Since the load in C5 is mainly transferred from girder B2 and the as-
connecting with girder B2 sustains a greater vertical load and a bending sociated effective composite slab flange, the decrease in load should be
moment than C2 before the deflection at the removed column reaches caused by failures of the slab and joint at the hogging end of girder B2.
around 150 mm. Actually, the vertical load and the bending moment at Hence, the strength of composite girder B2 at the hogging end governs
C5 decrease rapidly from a displacement of around 76 mm at the re- the maximum load capacity of the composite floor system after losing
moved column, synchronising with the development of applied load. the corner column.
41
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
0
0 50 100 150 200 250 300 sudden removal of the corner column under a total load of 160 kN is
force (kN)
-10 similar to that under quasi-static scenario (Fig. 14). Under this level of
load, the crucial components at the hogging end of B2 do not fail but
-20
have plastic deformation, as shown in Fig. 26 (a). However, under a
-30 free-fall load of 180 kN, these components fracture, and the forces de-
velop in a similar manner to those under quasi-static loading scenario,
-40 as shown in Figs. 15 and 26.
Deflection of removed column (mm)
flection curve (Fig. 27), which is in good agreement with the pseudo-
0 50 100 150 200 250 300 static response calculated based on energy method [13]. It should be
-10 noted that the pseudo-static response has been calculated to its max-
imum value. However, it does not give any indications of dynamic ul-
timate loads or deformations as investigated by Fu et al. [14]. A value
-20 of DIF is the quotient of a static load and the corresponding dynamic
load on the same deformation basis. As shown in Fig. 28, deflections are
-30 normalised as rotations (θpra/θyb) of the primary member (girder). The
RC5,y RC2,x
terms θpra and θyb indicate the plastic and yield rotations of the girders,
respectively. It can be seen that the DIF for the CC removal scenario
-40 varies from around 1.8 to 1.5 with increasing deflection, which is
Deflection of removed column (mm)
greater than the corresponding DIF under an IC removal scenario.
Fig. 22. Development of horizontal reaction forces at column C2 and C5. However, this finding from the numerical predictions requires further
verifications by actual tests.
150
Horizontal reaction force (kN)
20
RC5,v
Reaction moment (kN·m)
120
RC2,v 15
90
10
60
5 MC5,ma MC2,mi
30
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Deflection of removed column (mm) Deflection of removed column (mm)
42
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
10 Time (s)
0
Deflection at removed
30 kN 1 1.25 1.5 1.75 2
Deflection at removed
-10 1 1.25 1.5 1.75 2 -100
column (mm)
60 kN -200
column (mm)
-30 90 kN -300 180 kN
120 kN -400
-50 150 kN -500
160 kN -600
-70 Time (s)
C5 C1
C2
Fig. 25. In-plane principal strain distribution of steel decking under a load of 160 kN at 1.14 s (first peak deflection) [12].
0 0
Component Force (kN)
Component force (kN)
-300 -300
Deflection of the removed column (mm) Deflection of the removed column (mm)
43
Q. Fu and K.-H. Tan Structures 21 (2019) 33–44
280 Acknowledgments
240
Vertical load (kN)
1.8
concrete composite floor system subjected to column removal scenarios. J Struct
Eng 2016;142(2):04015133.
1.6 [8] Fu QN, Tan KH, Zhou XH, Yang B. Load-resisting mechanisms of 3D composite floor
systems under internal column-removal scenario. Eng Struct 2017;148:357–72.
1.4 [9] Fu QN, Tan KH, Zhou XH, Yang B. Three-dimensional composite floor systems under
column-removal scenarios. J Struct Eng 2018;144(10).
1.2 [10] Fu QN. Structural behaviour of composite floor systems under column removal
scenario (PhD thesis). Nanyang Technological University; 2018.
1 [11] Izzuddin BA, Tao XY, Elghazouli AY. Realistic modeling of composite and re-
0 1 2 3 4 5 6 7 inforced concrete floor slabs under extreme loading. I: analytical method. J Struct
Eng 2004;130(12):1972–84.
Normalised rotation of girder (θpra/θyb) [12] Fu QN, Tan KH. Numerical study on steel-concrete composite floor systems under
corner column removal scenario. 12th international conference on advances in
Fig. 28. DIF under IC and CC scenario. steel-concrete composite structures (ASCCS 2018). 2018. p. 897–904. València,
Spain.
[13] Izzuddin BA, Vlassis AG, Elghazouli AY, Nethercot DA. Progressive collapse of
girder, beam and slab. Consequently, composite floor systems under a multi-storey buildings due to sudden column loss — part I: simplified assessment
framework. Eng Struct 2008;30(5):1308–18.
corner column removal scenario fail in a less ductile manner compared [14] Fu QN, Tan KH, Zhou XH, Yang B. Numerical simulations on three-dimensional
with those under an internal column removal scenario. Furthermore, composite structural systems against progressive collapse. J Constr Steel Res
the dynamic increase factor for corner column removal scenario is 2017;135:125–36.
greater than that under internal column removal scenario.
44