Fire Safety Journal 89 (2017) 63–76

Contents lists available at ScienceDirect

Fire Safety Journal

journal homepage: www.elsevier.com/locate/firesaf

Experimental investigation on thermal and mechanical behaviour of MARK

composite floors exposed to standard fire
⁎
Guo-Qiang Lia,c, Nasi Zhangb, Jian Jiangc,
a
State Key Laboratory for Disaster Reduction in Civil Engineering, Tongji University, Shanghai 200092, China
b
Genex Systems, Newport News, VA 23606, USA
c
College of Civil Engineering, Tongji University, Shanghai 200092, China

A R T I C L E I N F O A BS T RAC T

Keywords: This paper presents experimental investigations on the thermal and mechanical behavior of composite floors
Fire resistance subjected to ISO standard fire. Four 5.2 m×3.7 m composite slabs are tested with different combinations of the
Experimental investigation presence of one unprotected secondary beam, direction of ribs, and location of the reinforcement. The
Composite floor experimental results show that the highest temperature in the reinforcements occurs during the cooling phase
Unprotected secondary beam
(30–50 °C increment after 10-min cooling). The temperature at the unexposed side of the slabs is below 100 °C
Direction of rib
up to 100-min heating, compared to the predicted fire resistance close to 90 mins from EC4. For the slabs
Location of reinforcement
EC4 without secondary beams, the cracks first occur around the boundaries of the slab, while for the slabs supported
by one unprotected secondary beam, concrete cracks first occur on the top of the slab above the beam due to the
negative bending moment, and later on develop around boundaries. Debonding is observed between the steel
deck and concrete slab. The secondary beam significantly impacts the deformation shape of tested slabs.
Although a large deflection, 1/20 of the span length, is reached in the tests, the composite slabs can still provide
sufficient load-bearing capacity due to membrane action. The occurrence of tensile membrane action is
confirmed by the measured tensile stress in the reinforcement and compressive stress in the concrete. A
comparison between measured and predicted fire resistance of the slabs indicates that EC4 calculations might
be used for the composite slabs beyond the specified geometry limit, and the prediction is conservative.

1. Introduction “compressive ring” at the boundary of the slab. Therefore, it is possible

Composite floor systems are commonly used in modern steel-
framed buildings. They consist of steel beams, steel decks, concrete
slabs, shear studs and reinforcement, as shown in Fig. 1. The composite
action between the steel beams and concrete slabs is achieved by
embedded shear studs. In the composite floor systems, the steel deck
can be taken as the bottom reinforcement when calculating the load
resistance at ambient temperature. In this way, the concrete can be
barely reinforced by a light anti-crack rebar or steel mesh. Another
advantage of composite slabs is that they allow to save construction
time since the steel deck is a permanent formwork. However, the
economy of composite floor systems is challenged by prescriptive fire-
resistant design provisions in the current building codes, which require
fire protection of the steel secondary beam. According to the observa-
tions in the large-scaled fire tests and in the real building fires, it is
found that the composite slab systems can bear the dead and live load
during the fire by “membrane action” mechanism, in which a “tensile
membrane” is formed at the center of the slab, and is supported by a
to remove the fire protection on the secondary beams due to the
enhancement of the fire resistance by the membrane action.
Many experimental studies have been conducted to investigate the
performance of composite floor systems in fire and the influencing
factors, as shown in Table 1. In 1989, an ECSC research project was
initiated at TNO (Netherland), in which 25 tests were performed to
study the thermal and structural behavior of composite slabs [15,6].
Twelve tests on two-dimensional thermal responses showed that the
geometry of the profiled steel deck greatly influence the temperature
distribution in the composite slab. Ten full-scale tests were performed
on simply supported, cantilever, and continuous slabs with a span of
3.2 m and different reinforcement ratios. The results showed that the
failure of composite slabs was controlled by the rupture of the
reinforcement. A low reinforcement ratio (typically 0.18%) led to an
early failure due to the insufficient bending capacity.
Motivated by the Broadgate Phase 8 fire and Churchill Plaza fire
occurred in UK in the early 1990s a total of 7 tests (Tests 1–6 in 1996
and Test 7 in 2003) were carried out on a full-scale eight-story steel-

⁎
Corresponding author.
E-mail address: jiangjian_0131@163.com (J. Jiang).

http://dx.doi.org/10.1016/j.firesaf.2017.02.009
Received 20 August 2015; Received in revised form 21 February 2017; Accepted 26 February 2017
0379-7112/ © 2017 Elsevier Ltd. All rights reserved.
G.-Q. Li et al. Fire Safety Journal 89 (2017) 63–76

New Zealand, design approaches that allow the unprotected steel in

Fig. 1. Typical composite floor systems [3].
framed building at Cardington [1,18]. Different test configurations,
such as restrained beams, plane frames and corner compartments at
different locations of the building, were tested, and all the secondary
beams were unprotected. The highest fire temperature in the tests
exceeded 1000 °C. The composite floors sustained the load without any
collapse even if its deflection reached 1/20 span length. The membrane
action in the composite floor played an important role in the survival of
the frame. Extensive computer models were also built to simulate the
behavior of steel-framed buildings and the tensile membrane action of
the slabs exposed to fire [16,8,24,25,19,17,11].
To prove the existence of the membrane action, Bailey et al. [2]
performed a test on a 9.5 m×6.5 m composite floor at ambient
temperature. The steel deck was removed during the test to take into
account the fire effect. It showed that the failure load doubled that
calculated from classic yield line theory. The BRANZ (Building
Research Association of New Zealand) carried out a fire test on a
two-way simply supported Hi-bond composite slab (3.3 m×4.3 m)
[21,22]. The test was performed in a controlled furnace environment
(ISO834) in order to compare the results with the current simple
design method. The measured temperature at the bottom of the slab
was substantially lower than the numerical simulation results due to
the buckling of the steel deck and its debonding from the concrete slab.
Evidences of building behavior were also available in the large-scale
fire tests in Australia and Germany. The purpose of the Australian tests
(also known as William Street fire tests and Collins Street fire tests)
conducted by BHP was to assess the reliability of the existing sprinkler
system and the behavior of unprotected steel beams [7]. The tempera-
ture of unprotected steel beams and slabs remained low due to the fire
barrier effect of the suspended ceiling system. The fire tests were
conducted on a four-story steel-framed building in Germany. The test
results showed that the composite floor reached a maximum displace-
ment of 60 mm and retained its overall integrity. In both Australia and
multi-story steel-framed buildings have been developed.
As demonstrated in previous studies, the sizable extra load-bearing
capacity exhibited by composite slabs thanks to the tensile membrane
action allows to remove the fire protection of the secondary beams
underneath the slab, without harms to structural safety, something that
may be extended in principle to the whole structure of tall buildings.
Many factors have been pointed out to influence the resistance of
membrane action, such as the boundary restraints, beam-to-slab and
beam-to-girder connections, cooling phase of realistic fires, etc. Bailey
and Toh [4] conducted 22 small-scale fire tests on horizontally
unrestrained concrete slabs. The test results were used to validate the
design method proposed by Bailey [3] for predicting the membrane
action under fire conditions. It was found that the fracture of the
reinforcement across the shorter span governed the failure. Li et al.
[20] presented a theoretical model to calculate the membrane action, in
which the slab was divided into 5 parts (a center-elliptic part and four
rigid parts around) at the limit state. The equilibrium equations were
established by using the force and moment in discretized slab stripes.
This method was further developed to include both the geometric
continuity and equilibrium on the integral slab (no discretization) in
the calculation of the loading resistance of membrane action [27]. In
2008, CTICM (France) tested an 8.7 m×6.6 m composite slab in an ISO
834 standard fire [12]. It was intended to provide experimental
evidence about the behavior of composite steel and concrete floors
exposed to the standard temperature-time curve and to promote the
application of the design concept based on membrane action. In order
to investigate the fire resistance of connections between concrete slab
and steel members at the perimeter of the composite floor when
subjected to large deflections due to membrane action, another fire test
was carried out in the project of COSSFIRE [28]. Fike and Kodur [10]
presented experimental and numerical studies on steel beam-concrete
composite floors made of steel fiber reinforced concrete. The studies
showed that the fire resistance of composite slabs can be significantly
improved by the composite action of the beam-slab assembly and
tensile membrane action. Wellman et al. [26] tested the behavior of
thin composite floor systems (4 m×4.5 m) exposed to fire. Various
shear connections, fire scenarios, and fire protection scenarios of
secondary steel beams were considered. None of the shear studs and
beam-to-girder shear connections failed during the heating and cooling
phases of the tests. The conclusion was that removing the fire
protection of the interior beams in thin lightweight composite slabs
is not recommended. Guo and Bailey [13] conducted experimental
studies on the behavior of composite slabs during the heating and
cooling phases in real fires. The results showed that the behavior of

Table 1
Summary of previous experiments on composite floor systems.

References Slab Slab size (m) Type of decking Secondary beam Test load Fire Maximum deflection
(kN/m2) (mm)

TNO tests Simply supported 3.2×0.9 Prins PSV 73 NA 5.8 ISO 834 290
[15] Continuous 3.2×0.9 150
2
Cardington tests BS Corner 9.5×6.5 PMF CF70 Unprotected 5.4 Wood ribs 40 kg/m 428
[11,18] BRE Corner 9×6 269
Test 7 11×7 6.0 1000
BRANZ Two-way 4.3×3.3 Hibond NA 5.5 ISO 834 253
[21]
Purdue Tests Two-way 4.6×4 Vulcraft 1.5VLR Unprotected and 9 ASTM E119 with 250
[26] protected cooling
Manchester tests One-way rotational 6.45×1.2 PMF CF60 NA 3.85–11.7 Parametric fire 33–103
[13] restrained
FRACOF Two-way 8.7×6.7 COFRAPLUS 60 Unprotected 5.1 ISO834 460
[29]
COSSFIRE Two-way 9×6.7 COFRAPLUS 60 Unprotected 3.9 ISO834 550
[28]
CTU test Two-way 4.5×3 TR40/160 NA 1.8 ISO834 300
[5]

64
G.-Q. Li et al. Fire Safety Journal 89 (2017) 63–76

Table 2
Details of the tested slabs.

Specimen No. Arrangement of Concrete cover of reinforcement Direction of ribs Secondary beam Test load (kN/ Load Test duration
reinforcement (mm) m2) ratioa (min)

S−1 ϕ8@150 21 Parallel to long side Unprotected 18.4 0.6 75

S−2 ϕ 8@150 30 17.7 0.6 90
S−3 ϕ 8@150 30 Parallel to short N/A 8.8 0.6 100
S−4 ϕ 8@150 30 side 9.5 0.65 100

a
Load ratio: the ratio of applied load to the design load resistance

Fig. 2. In-plan view of the specimens (unit: mm).

in fire. It was proven that the thickness of the steel deck had a
significant influence on the fire resistance of the composite slab.
However, the strength of the concrete and the mesh size only played
a minor role. Based on the NIST investigation on the WTC7 collapse,
McAllister [23] proposed that the floor beam length and connection
type impact the structural response of the floor system more signifi-
cantly than the girder studs and girder framing.
In order to study the influence of the boundary conditions, slab
layout, reinforcement location, and the unprotected secondary beams
on the development of tensile membrane action, four composite slabs
were tested in the furnace of State Key Laboratory for Disaster
Reduction in Civil Engineering in Tongji University. Unprotected
secondary beams were placed at the mid-span of two out of the four
slabs. The concrete cover of the reinforcement at the top of the slabs
were 21 mm and 30 mm respectively, to investigate to what extent the
Fig. 3. Test setup of S-1 and S-2 (unit: mm).
location of the reinforcement affects load-bearing capacity of composite
slabs. In the tests, the steel decks were welded to the protected primary
beams by shear studs. The primary beams were sitting on the furnace
wall, and were considered as solid beams in fire. The direction of deck
ribs was designed along the longer span in two slabs with the
unprotected secondary beam, and along the shorter span in the other
two. The temperature in the reinforcement and on the top and bottom
of the slab, the strain in the reinforcement and concrete, the deflection
of slabs were measured and discussed. The test results were also
compared to the EC4 calculations [9].

2. Test setup

2.1. Description of the specimens

Fig. 4. Overview of the test setup.

composite slabs depended on the heating rate, the highest temperature
reached and the cooling rate. The highest temperature on the un-
exposed side of the slab and in the reinforcement mesh was reached
during the cooling phase of the fire. A numerical study was followed
[14] to investigate the influence of the thickness of the steel deck,
strength of concrete and mesh size on the behavior of composite slabs
Four 5.2 m×3.7 m composite slabs were tested in the
4.5 m×3 m×2.2 m furnace at Tongji University, and denoted as S-1,
S-2, S-3 and S-4. The bottom surface of the slabs was exposed to fire,
with the upper surface exposed to the ambient environment. The
geometric properties of the four specimens are shown in Table 2. The
composite slabs, S-1 to S-4, consisted of trapezoidal steel deck, normal
weight concrete and anti-crack reinforcement mesh. S-1 and S-2 had
steel deck ribs placed parallel to the long side of the slab and an

65
G.-Q. Li et al.
Fig. 5. Details of cross-section of the slabs (unit: mm).
Fig. 6. Steel deck-beam connection.
unprotected secondary beam in the middle of the long span, while the
ribs in S-3 and S-4 were placed along the short side without any
secondary beam. The illustrations of the two types of floor systems are
shown in Fig. 2. The concrete cover of the reinforcement was 21 mm
for the S-1 and 30 mm for the rest of slabs. The specimens S-3 and S-4
had the same dimensions but very slight difference in the loading ratio.
The steel deck and secondary beams (if applicable) of all specimens
were unprotected, while all the primary beams were protected. The test
setup of S-1 and S-2 is shown in Fig. 3. The composite slab was
connected to the primary and secondary beams by shear studs. The
primary beams were connected to the base beams of the furnace by 116
bolts to ensure no vertical or horizontal displacements occurring on the
primary beams. Fig. 4 shows an overview of the test setup in the
laboratory.
The thickness of the steel decking was 1 mm and its configuration is
depicted in Fig. 5. The total depth of the slabs was 146 mm, and the
concrete thickness above ribs was 70 mm. An anti-crack reinforcement
mesh, consisting of 8 mm diameter bars at a spacing of 150 mm in two
directions, was placed at the top of the concrete. A pair of shear studs
(16 mm in diameter, 125 mm in height, and 80 mm apart) was welded
Fig. 7. Details of cross-section of the steel beams (unit: mm).
66
Fire Safety Journal 89 (2017) 63–76
vertically at each lower flange of the steel deck, as shown in Fig. 6. The
cross sections of hot-rolled I-type primary and secondary beams are
shown in Fig. 7. Grade 10.9 M16 high strength friction bolts were used
for beam-to-beam connection as shown in Figs. 8 and 9. The end of the
reinforcement penetrated the concrete and was extended for 150 mm
to simulate the boundary condition observed in Cardington fire tests
[18]: concrete crushing and reinforcement rupture above the primary
beams due to negative bending moment (Fig. 10).
2.2. Material properties
Five 150 mm cubic concrete blocks and three reinforcement speci-
mens were tested for each slab, and the material properties are shown
in Table 3. The characteristic yield strength of steel deck, primary and
secondary beam were 270 MPa, 345 MPa, and 235 MPa respectively.
2.3. Loading system
The tested slabs were loaded at 24 points to simulate uniformly
distributed load, as shown in Fig. 11. The load was gradually applied at
a 10% increment until the designed test load before the fire started, and
remained constant during the heating and cooling phases of the fire
test. Note that the composite slab exposed to the fire will undergoe
large deflections and a sort of elliptical deflection forms in the central
part of the slab. Consequently, designing and assembling a loading
system able to apply the same load in each of the 24 loading points was
no easy matter, as all points were subjected to sizable deflections and
rotations, which should not affect the applied load. A novel loading
system was, therefore, designed and built (Fig. 12).
The loading system was composed of one hydraulic jack, one
transfer beam 1 (TB1), two transfer beams 2 (TB2), and four triangular
loading plates. First, the hydraulic jack was connected to TB1 by a
bearing, and the load in the hydraulic jack was evenly distributed to the
two ends of TB1. Then, TB1 was supported by two TB2s at the mid-
point of TB2. In order to allow the rotation between TB1 and TB2, a
G.-Q. Li et al. Fire Safety Journal 89 (2017) 63–76

Fig. 8. Details of the beam-to-beam connections (unit: mm).

Fig. 9. Secondary beam-primary beam connection.

Fig. 11. Arrangement of the loading points.

between TB2 and the triangular plate, a steel hemisphere was welded to
the bottom of TB2 at each end. These steel hemispheres were put onto
the smaller circular holes/slots located at the center of the triangular
plates. The triangular loading plate was then supported at its vertices
on the slab. Three hemispheres were welded at the vertices to allow
rotations. A 150 mm by 150 mm by 10 mm steel plate was inserted
between the hemisphere (on the triangular loading plate) and the
concrete slab at each loading point to avoid the stress concentration.
The deformation and stress distribution in the triangular loading plate
was examined using Solid 92 element in the finite element software
ANSYS, as shown in Fig. 13. The figure shows that in most of the area
of the plate the stress level was lower than 50% of the strength at
yielding. The maximum deformation was less than 3 mm, which
Fig. 10. Extension of the reinforcement along the sides of the slab.
satisfied the test requirement. A total of two sets of the described
loading systems were used in the tests (one for half of the slab). The
Table 3
performance of the proposed loading systems at large deflections is
Material properties of concrete and reinforcement.
shown in Fig. 14. It was found that the loading points remained in the
Test No. Concrete Reinforcement initial position at the end of the tests.
strength (MPa)
Yield Ultimate fy/fu Ultimate
strength fy strength fu strain (%) 2.4. Instrumentation
(MPa) (MPa)
A total of 9 large-stroke LVDTs, 13 thermocouples, 26 steel strain
S−1 26.1 579.1 632.1 0.92 33.3
S−2 21.0 531.9 604.9 0.88 36.0 gauges, and 20 concrete strain gauges were installed in each specimen.
S−3 22.37 557.0 661.3 0.84 31.3 The LVDTs were located at the center, quarter span, and the sides of
S−4 22.87 the slab (on the top surface) to monitor the displacement profile
(shown in Fig. 15). The thermocouples were attached to the bottom
flange of the steel deck, to the top surface of the slab, and to the
piece of steel tube was welded to each end of TB1 to form a point
reinforcement embedded in the concrete to record the temperature at
contact. Two steel rods were welded on TB2 beside the contact point to
the exposed and unexposed surface of the slab, and in the reinforce-
prevent the unexpected lateral sliding of TB1. To allow the rotation
ment mesh as shown in Fig. 16. The steel strain gauges were installed

67
G.-Q. Li et al. Fire Safety Journal 89 (2017) 63–76

Fig. 12. Details of the loading system.

Fig. 13. Stresses and displacements in the triangular loading plate (unit: MPa).

in the embedded reinforcement mesh to observe the tensile stress in

the rebar. The concrete strain gauges were glued on the top surface of
the slab at the sides and corners to observe the compressive stress in
the compressive ring. Note that the operating temperature of the strain Fig. 15. Arrangement of the LVDTs (unit: mm).

Fig. 14. Loading system in the test.

68
G.-Q. Li et al.
Fig. 16. Arrangement of the thermocouples (unit: mm).
Fig. 17. Arrangement of the steel strain gauges (unit: mm).
Fig. 18. Arrangement of the concrete strain gauges (unit: mm).
Fig. 19. Comparison of ISO fire and the average furnace temperature.
gauges is 10–60 °C. Accordingly, all the data on the temperature out of
this range were ignored. The locations of the steel and concrete strain
gauges are shown in Figs. 17 and 18, respectively.
69
Fire Safety Journal 89 (2017) 63–76
Fig. 20. Comparison of the temperature at the bottom surface of the slabs.
Fig. 21. Comparison of the temperature on the top surface of the slabs.
Fig. 22. Comparison of the average temperature in the reinforcement.
2.5. Test procedure
The test protocol is as follows:
Phase I: loading at ambient temperature. The slabs were loaded in
10 even steps, until the designed testing load (in Table 2) was applied.
Phase II: heating process. ISO 834 standard fire was adopted. The
fire durations for the four slabs were 75 min, 90 min, 90 min, and
100 min, respectively.
G.-Q. Li et al. Fire Safety Journal 89 (2017) 63–76

Table 4
Comparison of the temperature in the steel deck and reinforcement during heating and cooling phases.

Test No. Steel Deck Reinforcement

End of heating Maximum temperature End of heating Maximum temperature

T (°C) Time (min) Tmax (°C) Time (min) T (°C) Time (min) Tmax (°C) Time (min)

S−1 876 75 876 75 166 75 202 101

S−2 900 90 923 99 358 90 394 102
S−3 866 100 879 105 363 100 406 110
S−4 937 100 952 103 335 100 370 112

3. Test results

3.1. Temperature in slabs

Fig. 23. Comparison of the temperature in the reinforcement at the center of the slabs.
Phase III: cooling process. The furnace was cooled via natural air
circulation, until the residual deformation of the slab was stabilized.
Data acquisition remained operative during the entire cooling phase.
Phase IV: unloading. The slabs were unloaded in 5 even steps.
The targeted ISO 834 standard fire heating curve and the average
furnace temperatures for all tests are compared in Fig. 19. The
recorded furnace temperatures agree well with the ISO fire during
the heating phase. A cooling phase followed the heating phase under
the air circulation of the furnace, and the furnace temperature reduced
to 200 °C after 105 min cooling for S-1, and 300 °C after 80–90 min
cooling for S-2 to S-4.
The temperature at the bottom of the steel deck and on the top
surface of the four slabs are shown in Figs. 20 and 21, respectively. It is
observed that the temperature at the bottom of the slabs (above the
steel deck) is about 100 °C lower than the furnace temperature due to
the isolation of the steel deck. The temperature on the unexposed top
surface is less than 100 °C for all slabs due to the isolation of the
composite slabs and the evaporation of the remaining free water in the
concrete. The difference in the temperature at the unexposed surface
(Fig. 21) is due to the presence of the secondary beam. For S-1 and S-2,
cracks first developed at the center of the slab due to the negative
bending moment above the secondary beam, while for S-3 and S-4,
cracks first developed at the boundaries of the slab. The occurrence of

Fig. 24. Residual crack patterns in the slabs.

70
G.-Q. Li et al.
Fig. 25. Cracks along the sides of the slabs.
Fig. 26. Cracks on the side of S-1.
Fig. 27. Cracks at the corner of S-1.
the crack tremendously encourages the local heat transfer, and
increases the temperature on the top surface at the center.
Fig. 22 shows the comparison of average temperatures in the
reinforcement of S-1 to S-4. It was found that the reinforcement
temperature of S-1 was lower than that of the other slabs. This was
mainly due to the shorter heating period (75 mins) of S1, compared to
about 100 mins for the other slabs. A big temperature gap was
observed between S-1 and other slabs after 75-min heating. The lower
temperature in S1 was also due to the smaller furnace temperature
(about 20 °C~70 °C lower) as shown in Fig. 19. In addition, the
reinforcement in S1 had a larger distance to the fire-exposed surface
of the slab (a distance of 21 mm to the top surface of the slab) and thus
a lower temperature in it, compared to a distance of 30 mm for the
other three tested slabs. Besides, the grading of the coarse aggregate,
71
Fire Safety Journal 89 (2017) 63–76
the sand pockets and the free water filling the voids are all factors
which may affect the temperature of the reinforcement.
Table 4 compares the maximum temperature in the steel deck and
in the reinforcement during the heating and cooling phases of the fire.
It was observed that the temperatures in the reinforcement reached
their maximum values during the cooling phase. The delay on the
occurrence of the highest temperature is more significant for the
reinforcement than the concrete. The temperatures in the reinforce-
ment continue to rise in the cooling phase for a period of 10 mins for S-
2, S-3, S-4 and 25 mins for S-1, and the increment is about 40 °C. This
indicates that there might be a potential failure after the peak
temperature has been reached at the end of the heating phase even if
the slab survives the heating phase of the fire. The post-peak of the
temperature in the reinforcement during the cooling phase should be
properly considered when determining the fire resistance of slabs
through membrane action.
Fig. 23 compares the temperature in the reinforcement at the center
of the slab. The temperatures measured in S-1 and S-2 are lower than
those in S-3 and S-4 due to the presence of the secondary beam. The
shielding effect of the secondary beams may lead to a temperature
reduction of 50–100 °C in the reinforcement.
3.2. Cracks on concrete
The crack patterns past the tests are shown in Fig. 24. The crack
pattern of S-3 and S-4 is very similar to that of S-1 and S-2, except that
S-1 and S-2 have cracks aligned with the short sides above the
secondary beam due to the negative bending moment at the beginning
of the test. In all slabs, the elliptical cracks were developed at the
boundary of the slab, which indicates the formation of the tensile
membrane action. The concrete rings comprised between the quasi-
elliptical cracks and the concrete outside the cracked elliptical zone
formed the "compressive ring" which allows the membrane action to be
developed. Highly-visible cracks were found along the long and short
sides of the slabs (as shown in Fig. 25). Cracks are also formed on the
side due to the bending and pull-out trend of the steel bars (shown in
Fig. 26). However, no significant bond-slip were observed between the
steel bars and the concrete. After the test, cracks were observed at some
corners (Fig. 27). The debonding between the steel deck and the
concrete slab is shown in Fig. 28, which may have significant influence
on the heat transfer through the depth of the slab. No collapse was
found in the tests, which indicates that the load-bearing capacity of the
slabs at high temperatures benefits from membrane action.
G.-Q. Li et al.
Fig. 28. Debonding between the steel deck and the concrete.
Fig. 29. Comparison of the strain of the reinforcement in the middle of the slabs.
Fig. 30. Comparison of the strain of concrete along the sides of the slabs.
3.3. Strain of the reinforcement and concrete
Figs. 29 and 30 show the typical strain of the reinforcement in the
middle of the slabs and that of the concrete at the boundary of the
slabs, respectively. Note that the critical temperature of strain gauges is
60 °C, and thus the strains measured after 50 min are not presented. It
is found that there was tensile strain in the reinforcement and
compressive strain in the concrete at the boundary, which indicates
the development of the tensile membrane action in the slabs.
72
Fire Safety Journal 89 (2017) 63–76
Numerical simulations are, therefore, required to further investigate
the distribution of the strain along the bars, at different points of the
slab, in order to let one draw comprehensive conclusions.
3.4. Deformation of the slabs
The deformation of S-1 and S-4 after unloading are shown in
Fig. 31. Large deflections were found at the quarter-span of S-1 and S-2
due to the presence of the secondary beam, while S-3 and S-4 (without
secondary beam) exhibited deflections in a parabolic shape. It indicates
that the appearance of the secondary beam has significant influence on
the deformation shape of the composite slab.
In the tests S-1 and S-2, the secondary beam experienced large
deflection as shown in Fig. 32. The unprotected secondary beam of S-2
underwent obvious global buckling (Fig. 32b). Based on the records,
the maximum temperature difference in the furnace besides the
secondary beam was about 200 °C, and might be a reason for the
torsional buckling. The eccentric loading on the beam (when the slab
underwent large deflections) may also contribute to its torsion. The
occurrence of global buckling of S-2 indicates that the unprotected
secondary beam may lose its strength at high temperatures and thus
has little effect on the tensile membrane action. It is suggested that it is
unnecessary to consider the secondary beam for determining the load-
bearing capacity of composite slabs at elevated temperatures.
Numerical simulations are needed to further explore the reasons for
triggering the failure of secondary beams.
The time history of vertical displacements at the center of the four
slabs are compared in Fig. 33. Note that S-1 experienced larger
G.-Q. Li et al. Fire Safety Journal 89 (2017) 63–76

Fig. 31. Deflection of the slabs after test.

Fig. 32. Deformation of the secondary beams.

Fig. 33. Deflection at the center of the slabs. Fig. 34. Deflection at various measuring points of S-1.

deflections during the heating phase than S-2 due to higher tempera-
ture of steel deck (about 50 °C higher as shown in Fig. 20) and
premature debonding at its lower flange (Fig. 28). However, S-1 had
a smaller maximum deflection because of the shorter heating period
and lower rebar temperature at the end of the test. The effect of steel
deck on the deflection of slabs is limited since it was at very high
temperatures and subjected to severe material degradation.
A continuous increment of displacements was observed at the
beginning of the cooling phase for all tests, as shown in Fig. 33, which
leads to maximum deflections of 180 mm, 190 mm, 150 mm and
150 mm for S-1 to S-4 respectively. This is mainly due to the further
temperature increment (30–50 °C) in the reinforcement after 10 min
cooling. The larger central displacement of S-1 and S-2 is caused by the
higher applied load (Table 2). In addition, the better fire protection
provided by the concrete ribs on the short span reinforcement in S-3
and S-4 may also have a slight contribution. As for the membrane
action, the reinforcement aligned with the short sides gives a major
contribution to the bearing capacity, compared with the minor role of
the reinforcement aligned with the long sides. This is due to the fact
that the same central deflection induces larger strains and curvatures in
the bars aligned with the short sides. The lower temperature of the
reinforcement along the short span of S-3 and S-4 resulted in smaller
deflections, compared to S-1 and S-2.
Fig. 34 compares the deflection at various locations on S-1. It shows
the deflection at Point 2 (refer to Fig. 15) is significantly smaller than
the other two points due to the presence of the secondary beams. It
again proves the significant influence of the secondary beam on the
deflection shape of the slabs.

73
G.-Q. Li et al. Fire Safety Journal 89 (2017) 63–76

Table 5
Comparison of the vertical displacements of four slabs during heating and cooling.

Specimen No. Ambient displacement (mm) End of heating Maximum values Residual displacement (3-h cooling) (mm)

δ (mm) T (min) δ (mm) δ/L T (min)

S−1 2.32 166 75 182 1/20 82 155

S−2 1.63 140 90 191 1/19 109 163
S−3 27.1 145 100 151 1/25 108 115
S−4 24.2 145 100 151 1/25 107 108

resistance of the tested slabs was thus calculated as 93 min. As shown

Fig. 35. Comparison on the temperature of lower flange of steel deck between EC4 and
tests.
Table 5 shows the deflections at ambient temperature and at the
end of heating, and the maximum and residual deflections in 3 h
cooling. S-3 and S-4 had a larger displacement under the initial loads at
ambient temperature due to the absence of the secondary beam. The
maximum deflection occurred at about 10 min after the cooling phase
started. The ratios of the maximum deflection to the length of the short
span (3.7 m) for the four slabs are 1/20, 1/19, 1/25, and 1/25
respectively, which are larger than the commonly accepted failure
criterion of deflection, 1/30 of span length. This indicates a sufficient
load-bearing capacity of composite slabs even when the deflection is
larger than the criterion. Larger residual deflections were observed in
S-1 and S−2 due to the larger applied load.
4. Comparison with EC4 calculations
In EC4, the fire resistance of composite slabs is determined by Eq.
(1) which is based on the thermal insulation criterion as the average
temperature increase reaches 140 °C and maximum temperature rising
reaches 180 °C. It depends on the thickness of the upper flat portion of
the slab (h1=70 mm in this case), view factor (Φup=0.75), rib geometry
factor (A/Lr=43 mm) and width of the upper flange of steel deck
(l3=142 mm). For normal weight concrete, the coefficients a0~ a5 were
taken as a0=−28.8(min), a1=1.55 (min/mm), a2=−12.6(min),
a3=0.33(min/mm), a4=−735(mm·min), and a5=48(min). The fire
in Fig. 21, the temperatures on the unexposed side of S1~S4 were lower
than 100 °C after the fire duration of 90 min, 100 min, and 100 min,
and the temperature increase was much smaller than the insulation
criteria of EC4. Therefore, the tested slabs satisfied the fire resistance
requirement of EC4.
A 1 A 1
ti = a 0 +a1 ∙h1+a2 ∙Φup + a3 ∙ + a ∙ + a5 ∙ ∙
Lr 4 l3 L r l3 (1)
EC4 also provides the methods to calculate the temperature at the
lower flange, web and upper flange of composite slabs, as well as the
temperature in the reinforcements. Fig. 35 compares the temperature
measured from the tests with the values predicted by EC4. It is found
that the EC4 predictions agree well with the test measurements. Based
on the temperature of the steel decking, the sagging moment resistance
of composite slabs can be determined as given in EC4:
n n
∑ Ai ks,T,i fy,i + αslab ∑ Aj kc,T, j fc, j = 0
i =1 j =1 (2)
n n
MT = ∑ Ai zi ks,T,i fy,i + αslab ∑ Aj zj kc,T, j fc, j
i =1 j =1 (3)
where Ai and Aj are the area of steel component i and concrete
component j, respectively; fy, i and fc,j are the characteristic strength
of steel and concrete, respectively; ks,T, i and kc,T,j are the reduction
factors of material strength, respectively; zi and zj are the distances
from the plastic neutral axis to the centroid of the elemental area Ai and
Aj, respectively.
The plastic neutral axis of the tested slabs was first determined by
Eq. (2). It was calculated by the equilibrium of tensile forces in the
reinforcement and steel deck, and compressive forces in the concrete at
the upper region of the slab. A coefficient αslab=0.85 was assumed for
the compression stress block of concrete at the top of slab. The
measured temperature of the steel components was used to determine
their residual strength, and the compressive strength at ambient
temperature was used for the concrete since their temperatures were
below 250 °C (EC4). The design moment resistance MT was determined
by Eq. (3) where the strength reduction factors at elevated tempera-
tures referred to EC4. The load-induced bending moment Mq at the
mid-span can be determined by Mq =ql2/8. For a strip of the slab with a
width of 344 mm (one rib part with a width of 202 mm plus a flat part
with a width of 142 mm), its mid-span bending moment Mq was
calculated as about 5.5 kN m (l=2.6 m for Slabs S-1 and S-2; l =3.7 m

Table 6
Moment resistances of the tested slabs from bending and membrane action calculated by EC4.

Test No. Fire duration Temperature of Mq (kNm) Distance of neutral axis to the top MT (kN·m)
(min) reinforcement (°C) surface of slabs (mm)
Bending Membrane of Membrane of Total
decking reinforcement

S−1 75 166 5.41 7 2.71 2.32 4.49 9.52

S−2 90 358 5.20 6.5 2.36 1.62 3.68 7.65
S−3 100 363 5.24 6.2 2.21 1.53 3.77 7.51
S−4 100 335 5.65 6.3 2.22 1.53 3.85 7.60

74
G.-Q. Li et al.
for Slabs S-3 and S-4). The temperatures of the lower flange, web and
upper flange of the deck were assumed the same. The critical
temperature of the deck was thus determined as about 600 °C from
the equilibrium of MT=Mq. This critical temperature was equivalent to
a fire resistance of about 40 mins. This calculated fire resistance is
much smaller than the practical fire duration of the tested slabs without
collapse (over 90 mins). The enhancement of the load-bearing capacity
of the tested slabs can be attributed to the tensile membrane action.
The moment resistances of the tested slabs at elevated temperatures
were determined and listed in Table 6. The moment resistance from
bending was determined by Eq. (3) using the measured temperature of
steel deck and reinforcement in the tests. The moment resistances from
the membrane action of the deck and reinforcement were calculated
from the product of their forces and maximum deflection of the slab.
Table 6 shows that the total mid-span moment resistance MT of the
tested slabs at the end of heating was larger than the load-induced
moment Mq. This indicates that the composite slabs had a fire
resistance of at least 90 mins without any fire protection. The mem-
brane effect of the reinforcement contributed nearly 50% of the
moment resistance. The contribution from the steel decking through
membrane action was smaller than that from the bending, and reduced
significantly as temperature increased.
In addition, a composite slab can be treated as an equivalent flat
concrete slab by using effective depth heff. For slabs having the same
lower flange and upper flange width, which is also the case in this
study, EC4 suggests that the heff should be determined as the sum of
the thickness of the upper continuous flat part and one half the height
of the rib. The heff of the tested slabs was thus calculated to be 108 mm.
EC4 also specifies the minimum effective thickness requirement as a
function of the standard fire resistance. For a fire resistance of
120 mins, the required minimum heff is 100 mm, which is smaller
than 108 mm. Therefore the tested slab should have a fire resistance of
120 mins at least. This seems reasonable since the temperature at the
top surface of the tested slabs was lower than 100 °C for a fire duration
of 100 mins.
The calculation of fire resistance of composite slabs in EC4 applies
to the slabs with a specified range on the width of the top of the rib
(80–155 mm), width of the lower flange (32–132 mm) and width of the
upper flange (40–115 mm), respectively (Table D.7 in EC4). These
geometric limitations are drawn from the parametric studies by Both
[6]. The corresponding parameters for the tested slabs are 202 mm,
142 mm and 142 mm, respectively, which are out of the specified range
in EC4. This indicates that EC4 formulas might be used for the
composite slabs beyond those specified in EC4, and the prediction is
conservative.
5. Conclusions
Four full-scale tests in fire were carried out on composite floor
systems supported by one unprotected secondary beam or not. The
thermal field, the deflections, the crack patterns and the strains in both
the reinforcement and the concrete were investigated. The results
presented and discussed in this paper allow to draw the following
conclusions:
(1) The highest temperatures of reinforcement and steel deck occurred
during the cooling phase. The increment of the temperature in
reinforcement is 30–50 °C after 10 mins of cooling. A potential
failure might occur during the cooling phase although the slabs
survives the heating process.
(2) The location of the reinforcement had significant influence on its
temperature. The shielding effect of the secondary beam led to a
temperature reduction of 50–100 °C in the reinforcement above
the beam.
(3) For the fire insulation criterion, the temperature on the unexposed
side of all the tested slabs was less than 100 °C for a heating
75
Fire Safety Journal 89 (2017) 63–76
duration up to 100 mins, compared to a calculated fire resistance
of 93 mins based on EC4. This indicates a conservative prediction
of fire resistance by EC4. For the load-bearing criterion, the tensile
membrane action contributed to nearly 50% of the load-bearing
capacity of the tested slabs, which led to a fire resistance over
100 mins. Based on the EC4, the calculated fire resistance is
40 mins when the membrane action is not considered.
(4) For slabs with the secondary beam, concrete cracks were first
observed at mid-span above the beam due to negative bending
moments. For all tests, concrete cracks developed at the boundary
of slabs due to membrane action. Some cracks also extended to the
side surface of slabs. Significant debonding of the steel deck and
the concrete was observed, which can influence the heat transfer
and the load bearing capacity of the slab.
(5) The occurrence of membrane actions was confirmed by the tensile
stress measured in the reinforcements and primary compressive
stress in the concrete.
(6) Large deflections were found at the quarter-span of slabs with a
secondary beam in the mid-span. A large deflection, about span/
20, was reached in the test without any collapse, which indicates a
sufficient load-bearing capacity of composite floor systems when
exposed to fire.
(7) Compared with the test results, the prediction of EC4 on the fire
resistance and temperature of steel deck was conservative. The
application of EC4 equations can be extended to other geometric
configuration of composite slabs.
Acknowledgements
The work presented in this paper was supported by the National
Natural Science Foundation of China with grant 51120185001 and
51408418.
References
[1] C.G. Bailey, T. Lennon, D.B. Moore, The behaviour of full-scale steel framed
buildings subject to compartment fires, Struct. Eng. 77 (8) (1999) 15–21.
[2] C.G. Bailey, D.S. White, D.B. Moore, The tensile membrane action of unrestrained
composite slabs simulated under fire conditions, Eng. Struct. 22 (2000)
1583–1595.
[3] C.G. Bailey, Efficient arrangement of reinforcement for membrane behaviour of
composite floor slabs in fire conditions, J. Constr. Steel Res. 59 (2003) 931–949.
[4] C.G. Bailey, W.S. Toh, Behaviour of concrete floor slabs at ambient and elevated
temperatures, Fire Saf. J. 42 (2007) 425–436.
[5] J. Bednar, F. Wald, J. Vodicka, A. Kohoutkova, Experiments on membrane action
of composite floors with steel fiber reinforced concrete slab exposed to fire, Fire Saf.
J. 59 (2013) 111–121.
[6] C. Both, The fire resistance of composite steel-concrete slabs (Ph.D. dissertation),
Delft University of Technology, Delft, Netherlands, 1998.
[7] BS, The behavior of multi-storey steel framed buildings in fire. Reports of British
Steel, Swinder Technology Center, 1999.
[8] A.Y. Elghazouli, B.A. Izzuddin, A.J. Richardson, Numerical modelling of the
structural fire behaviour of composite buildings, Fire Saf. J. 35 (2000) 279–297.
[9] EN 1994-1-2, Eurocode 4 Design of composite steel and concrete structures: Part 1.
2: General rules, Structural fire design, European Committee for Standardisation,
Brussels, 2005.
[10] R. Fike, V. Kodur, Enhancing the fire resistance of composite floor assemblies
through the use of steel fiber reinforced concrete, Eng. Struct. 33 (2011)
2870–2878.
[11] S. Foster, M. Chladna´, C. Hsieh, I. Burgess, R. Plank, Thermal and structural
behaviour of a full-scale composite building subject to a severe compartment fire,
Fire Saf. J. 42 (2007) 183–199.
[12] FRACOF, Fire resistance assessment of partially protected composite floors
(FRACOF): Engineering background, Technical Report of European Coal and Steel
Community (ECSC), 2010.
[13] S. Guo, C.G. Bailey, Experimental behaviour of composite slabs during the heating
and cooling fire stages, Eng. Struct. 33 (2011) 563–571.
[14] S. Guo, Experimental and numerical study on restrained composite slab during
heating and cooling, J. Constr. Steel Res. 69 (2012) 95–105.
[15] R. Hamerlinck, L. Twilt, J. Stark, A numerical model for fire-exposed composite
steel/concrete slabs. in: Proceedings of the tenth International Specialty
Conference on Cold-formed Steel Structures, USA, 115-130, 1990.
[16] Z. Huang, I.W. Burgess, R.J. Plank, Effective stiffness modelling of composite
concrete slabs in fire, Eng. Struct. 22 (9) (2000) 1133–1144.
[17] B.A. Izzuddin, X.Y. Tao, A.Y. Elghazouli, Realistic modelling of composite and R/C
G.-Q. Li et al.
floor slabs under extreme loading–Part I: analytical method, J. Struct. Eng. 130
(12) (2004) 1972–1984.
[18] B.R. Kirby, British steel technical European fire test program design, construction
and results, Fire, static and dynamic tests of building structures, London, 1997.
[19] S. Lamont, A.S. Usmani, M. Gillie, Behaviour of a small composite steel frame
structure in a “long-cool” and a “short-hot” fire, Fire Saf. J. 39 (2004) 327–357.
[20] G.Q. Li, S.X. Guo, H.S. Zhou, Modeling of membrane action in floor slabs subjected
to fire, Eng. Struct. 29 (6) (2007) 880–887.
[21] L.C.S. Lim, A. Buchanan, P. Moss, J.M. Franssen, Numerical modelling of two-way
reinforced concrete slabs in fire, Eng. Struct. 26 (2004) 1081–1091.
[22] L. Lim, C. Wade, Experimental fire tests of two-way concrete slabs (Fire
Engineering Research Report 02/12), University of Canterbury and BRANZ Ltd,
New Zealand, 2002.
[23] T.P. McAllister, Sensitivity of composite floor system response at elevated
temperatures to structural features, Eng. Struct. 58 (2014) 115–128.
[24] A.M. Sanad, S. Lamont, A.S. Usmani, J.M. Rotter, Structural behaviour in a fire
76
Fire Safety Journal 89 (2017) 63–76
compartment under different heating regimes—part 1 (slab thermal gradients), Fire
Saf. J. 35 (2000) 99–116.
[25] A.M. Sanad, S. Lamont, A.S. Usmani, J.M. Rotter, Structural behaviour in a fire
compartment under different heating regimes—part 2 (slab mean temperatures),
Fire Saf. J. 35 (2000) 117–130.
[26] E.I. Wellman, A.H. Varma, R. Fike, V. Kodur, Experimental evaluation of thin
composite floor assemblies under fire loading, J. Struct. Eng. 137 (9) (2011)
1002–1016.
[27] N.S. Zhang, G.Q. Li, A new method to analyze the membrane action of composite
floor slabs in fire condition, Fire Technol. 46 (2009) 3–18.
[28] B. Zhao, M. Roosefid, A. Breunese, Connections of steel and composite structures
under natural fire conditions (COSSFIRE), Technical Report of Research Found for
Cool & Steel (FRCS), 2011.
[29] B. Zhao, M. Roosefid, O. Vassart, Full scale test of a steel and concrete composite
floor exposed to ISO fire. Proceedings of 5th International Conference on
Structures in Fire, Singapore, 2008.