See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/335602342

Fire Performance of Concrete Flat Slabs

Conference Paper · June 2018

CITATIONS READS

2 300

4 authors:

Pasindu Weerasinghe Priyan Mendis

University of Moratuwa University of Melbourne
7 PUBLICATIONS   40 CITATIONS    440 PUBLICATIONS   7,743 CITATIONS   

SEE PROFILE SEE PROFILE

Kate T. Q. Nguyen Tuan Duc Ngo

RMIT University University of Melbourne
44 PUBLICATIONS   3,790 CITATIONS    374 PUBLICATIONS   12,157 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Lightweight geopolymer concrete View project

Material point method for solid and fluid mechanics problems View project

All content following this page was uploaded by Pasindu Weerasinghe on 04 September 2019.

The user has requested enhancement of the downloaded file.

 Structures in Fire
SiF’2018

10TH INTERNATIONAL CONFERENCE ON STRUCTURES IN

FIRE

BELFAST, UNITED KINGDOM

6-8 JUNE 2018

EDITORS
Ali Nadjai
Faris Ali
Jean-Marc Franssen
Olivier Vassart

ORGANISED BY
Ulster University
Faculty of Computing Engineering and Built Environment
Fire Safety Engineering Technology & Research Institute (FireSERT)
 SiF 2018– The 10th International Conference on Structures in Fire
FireSERT, Ulster University, Belfast, UK, June 6-8, 2018

FIRE PERFORMANCE OF CONCRETE FLAT SLABS

Pasindu Weerasinghe1, Priyan Mendis2, Kate Nguyen3, Tuan Ngo4

ABSTRACT
With the increasing use of concrete flat slabs especially in multi-storey buildings, fire resistance
requirements tend to be critical in the design. Current design guidelines are conservative as they are
based on isolated member tests which does not consider the continuous action of flat slabs in actual
structures. Issues experienced in the industry due to such design guidelines are discussed and the
need for more representative experimental and analytical models are highlighted. General purpose
FE program ABAQUS is used to model the thermo-mechanical behaviour of flat slabs in elevated
temperatures. Improvisations that can be used to avoid convergence issues and methods to reduce
the computational cost are presented. Thermal and mechanical response of the model is validated
with experimental results from a fire test on restrained concrete flat slabs. Further enhancements
that can be used to improve the accuracy and reliability of such models are discussed. The expected
outcome is to provide a reliable research base to improve the current building code and design
guidelines regarding fire design of concrete flat slabs and provide more accurate rational guidelines
for designers leading to significant cost savings. Preliminary findings from this project are
presented in the paper.

Keywords: Flat slabs, finite element modelling, thermo-mechanical analysis

1 INTRODUCTION
The use of flat slabs is becoming very popular in modern buildings because of their speed of
construction and use of cost-effective formwork. In the absence of beams, lower story heights can
be achieved which saves cost in vertical cladding, partition walls, mechanical systems and
plumbing especially in multi-story buildings. However, building code regulations regarding fire
design of flat slabs are based on research carried out a few decades ago. They are incomplete or
over conservative causing extra thick slabs in buildings.

With the increasing number of fire related disasters occurring around the world, more emphasis was
placed on the fire resistive design of structures. Compared to other materials, reinforced concrete
structures have comparatively performed better under fire because it is non-combustible and have a
low thermal conductivity. This could mean that the occupants will have more time to evacuate the
building before a total collapse. However, exposure to high temperatures can alter the thermal and
mechanical properties of concrete and steel which would reduce the capacity of the structural
elements. Ultimately it could damage the structure permanently causing harm to occupants and
resulting in a significant financial loss. In the context of flat slabs, the limited number of

1 PhD Candidate. Department of Infrastructure Engineering, The University of Melbourne, Melbourne, Australia.
e-mail: pthalpe@student.unimelb.edu.au
2 Professor. Department of Infrastructure Engineering, The University of Melbourne, Melbourne, Australia.

e-mail: pamendis@unimelb.edu.au * Corresponding Author

3 Reserach Fellow. Department of Infrastructure Engineering, The University of Melbourne, Melbourne, Australia.

e-mail: kate.nguyen@unimelb.edu.au
4 Professor. Department of Infrastructure Engineering, The University of Melbourne, Melbourne, Australia.

e-mail: dtngo@unimelb.edu.au

125
 Fire performance of concrete flat slabs

experimental studies have shown that punching shear failure could be critical when exposed to fire
[1-5]. Interpretation of such experimental results and how well they represent the actual condition
of a flat slab in a building will be further discussed in this paper.

Fire design of structural elements is carried out in accordance with the guidance given in the
concrete design code of the respective country. Many of the modern day codes follow a
performance-based approach for fire design. Three main performance criteria considered are
stability, insulation and integrity. Fire design of flat slabs in Australian Concrete structures code AS
3600 [6] and other major codes is based on the European guidelines (EN 1992-1-2) [7] which
specify a minimum thickness and cover to reinforcement in order to satisfy the aforementioned
performance criteria. For fire ratings of 90 min and above it specifies a constant thickness (200mm).
This is a major problem for designers. For example, the National Construction Code of Australia
(NCC) [8] requires a minimum fire rating level (FRL) of 90 min for class A construction (Buildings
over 4 stories). Therefore, the flat slab thickness should be always higher than 200mm for multi-
storey buildings.

Another critical issue that has been raised due to this 200mm thickness limit is when there is a
requirement for a set down in slabs. Set downs are placed on slabs in bathroom areas to have a
proper drainage of water. Typical set down can be 30mm. However, even a 30mm set down in a
200mm thick flat slab would violate the concrete code and NCC requirement to achieve a 90 min
FRL [8]. Increasing the total slab thickness by 30mm would prove to be uneconomical when costs
accumulate for many floors throughout the height of a multi-storey building. Rather than having a
uniform 200mm minimum thickness throughout out the floor, it is more rational to propose certain
limits i.e. A minimum distance from columns where set downs are permitted, because only certain
areas of the floor slab will be critical under fire conditions. To make such decisions a
comprehensive research base is needed.

The main aim of the research conducted at the University of Melbourne by the authors is to conduct
a comprehensive study of the performance of concrete flat slabs under fire taking into account the
different boundary conditions and load cases. Limited studies conducted up to now are reviewed
and preliminary findings from this project related to modelling the behaviour of concrete flat slabs
under fire are presented in the paper.

2 EXPERIMENTAL STUDIES ON FIRE PERFORMANCE OF CONCRETE FLAT

SLABS
Only a limited number of experimental studies were conducted on flat slabs exposed to fire. Some
important tests will be discussed highlighting their significance and limitations.

Minimum thickness specification for fire design of concrete flat slabs in BS EN 1992-1-2 and AS
3600 is based on the tests conducted by Kordina [1]. The tests involved a simply supported slab
panel connected to single central column, heated from bottom surface with loading applied via a
ring to the slab panel to simulate uniformly distributed load (see Fig. 1.). This setup would not
allow for any catenary actions to be considered between columns. 10 slabs with thickness 200mm
and 4 slabs with thickness 150mm were tested and failure mechanism observed for all the cases was
punching shear failure. Time to failure ranged from 17 to 120mins for 200mm thick slabs and 8 to
180mins for 150mm thick slabs. With such a loading arrangement for a slab supported from a single
column it is clear that failure mechanism can only be punching shear after top reinforcement has
yielded. Membrane action of the slab which is there in an actual slab of a building would not be
taken into account as the slab is simply supported during the test.

126
 Concrete Structures

a) b)

Fig. 1. Test set-up used by Kordina [1] a) Sectional view; b) Plan view

However the normal failure mode for a one-way or two-way slab, subjected to fire from below, is
yielding of the bottom reinforcement due to heat that leads to moment redistribution to the support
and top reinforcement. In such a scenario a hinge will be formed at the support and slab will act like
a cantilever which tends to be a more ductile failure than punching shear failure.

Annerel et al. [2], Salem et al. [3] and Ghoreishi et al. [4] have conducted similar fire tests on
isolated flat slabs. Test set-up used by Annerel et al. [2] is similar to that of Kordina and they have
tested four 250mm thick slabs (with added stirrups near the column face) under fire. Two slabs did
not fail at the total duration of 120 min and two slabs failed after 20 mins due to punching shear
failure. Salem et al. [3] have used an inverted specimen (see Fig. 2(b)) and heated from the tension
surface. Twelve specimens of 100mm thickness were tested and more attention was given to the
load deflection behaviour and failure load rather than fire resistance period. Ghoreishi et al. [4] have
used a similar inverted test arrangement and 6 slabs of 120 mm thickness (3 at ambient temperature
and 3 heated) were tested. However, the main difference of this test is that first the slabs were
heated up to a constant temperature and then load was applied until failure to find the punching
shear capacity. All the above tests have used a simply supported boundary condition along the
perimeter of the sample.

a) b) c)

Fig. 2. Support conditions, loading and heated surfaces of experiments by a) Annerel et al.[2] ;b) Salem et al. [3] and c)
Ghoreishi et al. [4]

A recent study conducted by Smith et al. [5, 9] has used a specially built steel frame to restrain the
slabs against lateral displacement and rotation as an attempt to represent the actual boundary
conditions of a continuous flat slab in a building. Results show that fire resistance of restrained
slabs is considerably higher than that of unrestrained slabs. An analytical study conducted using the
software SAFIR [10] has also shown that the fire resistance of flat slabs is significantly improved
due to restraint conditions provided by adjacent slab panels and membrane action. Therefore, Smith
et al. test results [5, 9] were chosen to model and validate the behaviour of flat slabs when exposed
to fire.

127
 Fire performance of concrete flat slabs

3 MODELLING OF BEHAVIOUR OF CONCRETE FLAT SLABS SUBJECTED TO

FIRE

3.1 General details

The tests conducted by Smith et al. which involved restrained support conditions were used to
develop the finite element model of the flat slab. The study consisted of a total of 15 slab specimens
(50, 75 and 100mm thick) which were tested at ambient temperature and elevated temperature
under restrained and unrestrained support conditions. Details of the experiment can be found in
Smith et al. [5, 9]. Fig. 3 shows the specimen size, loading arrangement, fire exposure and support
conditions along with the reinforcement arrangement of flat slab specimens used in the test. 100mm
thick, 0.8% reinforced, restrained slab specimen was selected to illustrate the modelling technique.

Fig. 3. Test set up used by Smith et al. [5, 9] a) Support conditions, loading and fire exposure; b) Reinforcement
arrangement (Specimen size – 1.4m x 1.4m x 0.1m, 0.25m x 0.25m x 0.1m column at the middle)

3.2 Finite element modelling of behaviour under fire

The widely used finite element program ABAQUS was chosen to model the fire behaviour of slabs.
The concrete damage plasticity model was selected to represent concrete as previous studies have
validated the use of this model for flat slabs in ambient conditions and other concrete elements
subjected to elevated temperatures. Convergence was reported as a common issue related to
modelling behaviour under fire as it contains both material nonlinearities and geometric
nonlinearities [11]. Temperature dependency of material properties make it even more difficulty to
converge on a solution. As a remedy in standard implicit modelling, the time increment of an
iteration can be made very small. However, the computational costs of such models are very high as
the usual duration of the applied fire in an analysis can be 1 to 4 hours. Therefore, an explicit
dynamic solver was used in this analysis to simulate the quasi-static behaviour with some
improvisations. The techniques used will be discussed in detail in the following sections.

Concrete damage plasticity (CDP) model in ABAQUS uses concepts of isotropic damaged
elasticity in combination with isotropic tensile and compressive plasticity to represent the inelastic
behaviour of concrete [12]. The main two failure mechanisms assumed are tensile cracking and
compressive crushing in concrete. More details can be found in [13]. General parameters considered
for the CDP model are the dilation angle (ψ) 30o , the eccentricity 0.1, the stress ration (fbo/fco) 1.16,
the shape factor (Kc) 0.667 and the viscosity parameter 0.005 which have been used in other similar
studies as well [11, 13, 14].

3.3 Material data and other parameters used in FE model

Stress-strain behaviour of concrete in compression and tension at elevated temperatures were
calculated in accordance with the guidelines given in Table 3.1 of EN 1992-1-2 [7]. Concrete
compressive strength (fcu) and tensile strength (fck,t) were determined experimentally as 51 MPa and
4MPa respectively [9]. Based on that a linear elastic behaviour was assumed up to 0.4fcu for
compressive stress strain behaviour of concrete. It was followed by a parabolic variation until it
reaches the ultimate strength according to the relationship given in [7]. The softening behaviour of
concrete is represented through a linear descending branch.

128
 Concrete Structures

For the stress-strain response of concrete in tension, a linear elastic behaviour was assumed up to its
tensile strength followed by a bilinear softening response. The area under the softening branch was
determined based on the fracture energy (Gf) which is assumed to be 0.25 N/mm for this model.
More details of this method can be found in [13, 15].

The stress-strain relationship for reinforcing steel at elevated temperatures was determined
according to Table 3.2b of EN 1992-1-2 [7] having a yield strength of 549MPa and an elastic
modulus of 200GPa [9].

Thermal properties of concrete (specific heat and conductivity) were also derived from relationships
given in EN 1992-1-2 [7] for siliceous concrete [9]. Lower limit of thermal conductivity was
selected as it gives more realistic temperatures for concrete structures than the upper limit [7].
Thermal elongation for siliceous concrete was determined in accordance with section 3.3.1
[7].Thermal properties of steel were not taken into account as the element types used in the FE
model for reinforcing steel did not participate in the heat transfer analysis. Wang [16] has showed
that the temperature across the depth with and without reinforcing bars was very similar and
therefore the effect from steel for the heat transfer can be negligible. However, the temperature of
the reinforcement can be indirectly found by relating to the temperature of the concrete at the same
depth.

Based on the symmetricity of loading and boundary conditions, only a quarter of the slab (0.7 m x
0.7m x 0.1m) was modelled to save the computational time. C3D8RT (An 8-node thermally
coupled brick) elements were used to model concrete whereas reinforcement was modelled using
T3D2 (A 2-node linear 3-D truss.) elements. Truss elements representing reinforcement were
embedded using the embedded region constraint assuming a perfect bond between concrete and
reinforcement. This assumption is made in several other studies [13, 17] as well and it has given
accurate results. Slab specimen in the experiment was restrained along the perimeter against both
translation and rotation. (See Fig. 3). In the model the bottom edge was restrained, and axial spring
elements were used along the side surfaces to represent the lateral restrained (see Fig. 4). A similar
approach was used by Genikomsou et al. [18] to investigate the compressive membrane action in
flat slabs. The advantage of using spring supports in this particular case is further explained under
the results section.
Bottom support
z-y symmetry plane (Uz = 0)
(Ux = 0)
Spring support
(z-x face)
P

Rigid body
constraint

Spring support
z-x symmetry plane (z-y face)
(Uy = 0)

Bottom support
(Uz = 0)

Fig. 4. Boundary conditions, loading and constraints used in the FE model

129
 Fire performance of concrete flat slabs

3.4 Analysis steps

Analysis is carried out in two different steps. First, loading was applied using a dynamic explicit
step. Although ABAQUS/Explicit uses a dynamic solution procedure it can be used for quasi-static
analysis with a low rate of loading [13]. In the experiment, the slab was casted along with a short
column where the loading was applied through a hydraulic jack to the column. However, in the
analysis only the slab was modelled and to simulate the effect of the column, a rigid body constraint
was applied to the location of loading. (See Fig. 4). Then, dynamic explicit coupled temperature
displacement step in ABAQUS was used to carry out the thermo-mechanical analysis. In this step,
the load (P) applied in the first step was kept constant while the temperature measured in the
experiment was applied to the top surface of the slab. The use of an explicit solver avoids the
convergence issues encountered in a standard static implicit analysis.

Total duration of the test is 4 hours; 2 hours of heating followed by 2 hours of cooling. For the
heating step the convective coefficients were taken as 25 W/m2K and 9 W/m2K for exposed surface
and unexposed surface respectively [7]. In the cooling step a convective coefficient of 9 W/m2K
was assigned for all the surfaces. Concrete emissivity was taken as 0.7 to simulate the radiative heat
flux in both steps.

Generally an explicit step is particularly well-suited to simulate brief transient dynamic events [12].
However, it can be used for an analysis consisting of a long duration with some improvisations.
Explicit time integration algorithm uses a very small time increment. It can be increased by either
artificially increasing the mass of the structure (mass scaling) or by artificially increasing the
loading speed (load factoring) [14]. Mass scaling can cause large inertia forces which can deviate
the solution from quasi-static state to dynamic state. Therefore, load factoring method is applied in
this model. It should be noted that load factoring in this study refers to the thermal load. Although
the duration of heating is 2 hours, it has been scaled down to 2 seconds in the coupled temperature
displacement step with modifications to the units involving time. i.e. conductivity, specific heat,
coefficient of convection and Stefan-Boltzmann constant for radiation.

4 RESULTS AND DISCUSSION

4.1 Thermal Response

Thermocouples (TCs) in the experiment were placed 200 mm away from the column surface across
the depth (at 10, 33, 65, 82 and 101mm). TC which was used to measure the exposed surface
temperature was located 10mm inside the slab depth [9]. Therefore, measured temperatures were
extrapolated to calculate the applied temperature to the exposed surface. Nodal temperature results
from 200 mm away from the column surface across the depth were compared with the measured
temperature values. It should be noted that temperatures at 20mm intervals across the depth were
determined using the temperature profiles at 10min intervals from Smith [9]. Fig. 5. shows the
temperature variation across the slab depth during heating and cooling phases of the experiment and
model. Modelled temperatures are in good agreement with the measured data.

4.2 Structural Response

Deflection criteria were considered to validate the structural response of the FE model. Downward
displacement was determined using two methods in the experiment, by employing a displacement
transducer (LP) connected to the loading rod which goes through the column and using digital
image correlation (DIC). However, it was concluded that the DIC measurements were more
accurate than LP as localised crushing of the concrete around the column stub and extension of the
loading rod due to the elevated temperature can artificially increase the LP measurements [9].

130
 Concrete Structures

Therefore, DIC measurements were taken to compare the modelled displacements with
experimental displacements.

Fig. 5. Experimental and modelled temperature variation across the slab depth

Fig. 6. Experimental and modelled deflection of the slab during heating phase

Deflections only during the heating phase was considered at this stage as material properties during
the cooling phase undergoes changes after experiencing elevated temperatures. The incorporation of
residual material properties is being further studied by the authors at present. The results are
presented in Fig. 6. It should be highlighted that when the perimeter is fully fixed the modelled
deflections are much lower than the measured deflections. Releasing the restraints increase the
deflection, but makes the model too relaxed. Therefore, spring supports were used as discussed
before and a spring stiffness of 30,000 N/mm used in the model as suggested by [18] to represent
the membrane action due to lateral restraints. With the use of spring supports a close match between
measured and modelled deflections have been achieved. Further studies are underway to evaluate
and validate the damage and failure criteria predicted by the model.

5 CONCLUDING REMARKS
• Current design guidelines regarding fire design of concrete flat slabs are based on isolated member
tests which do not consider the continuous action and thus neglects effects such as membrane action
which could improve the fire resistance.

131
 Fire performance of concrete flat slabs

• General purpose FE program ABAQUS was successfully used to simulate the thermal and
mechanical response of concrete flat slabs exposed to fire.
• The use of ABAQUS/Explicit solver with coupled-temperature displacement analysis
procedure is beneficial in eliminating the convergence issues arising as a result of added
nonlinearities due to temperature dependant material properties.
• Computational cost for the explicit analysis can be significantly reduced by reducing the
step time with modifications to time dependant thermal properties of materials.
• Modelled thermo-mechanical behaviour and the experimental results are in good agreement,
which endorses the accuracy of the model.
• Such FE models can be used to study the effect of different support conditions, load
combinations, changes in the slab thickness etc. to fire resistance level.

REFERENCES
1. Kordina, K., Flat slabs under fire–Redistribution of the internal forces and punching tests. Institut
für Baustoffe, Massivbau und Brandschutz, Technische Universität Braunschweig, Germany,
CEN/TC, 1993. 250: p. 1-2.
2. Annerel, E., L. Lu, and L. Taerwe, Punching shear tests on flat concrete slabs exposed to fire. Fire
Safety Journal, 2013. 57: p. 83-95.
3. Salem, H., et al., Punching shear strength of reinforced concrete flat slabs subjected to fire on their
tension sides. HBRC Journal, 2012. 8(1): p. 36-46.
4. Ghoreishi, M., A. Bagchi, and M.A. Sultan, Punching shear behavior of concrete flat slabs in
elevated temperature and fire. Advances in Structural Engineering, 2015. 18(5): p. 659-674.
5. Smith, H.K., T.J. Stratford, and L.A. Bisby. Punching shear of restrained reinforced concrete slabs
under fire conditions. in 8th International Conference on Structures in Fire. 2014.
6. Standards Australia, AS 3600 - Concrete structures. 2009.
7. European committee for standardization, Eurocode 2: Design of concrete structures - Part 1-2:
General rules -Structural fire design. 2004.
8. The Australian Building Codes Board, National Construction Code: Volume 1. 2016.
9. Smith, H.K.M., Punching shear of flat reinforced-concrete slabs under fire conditions. 2016, The
University of Edinburgh.
10. Moss, P.J., et al., The fire behaviour of multi-bay, two-way reinforced concrete slabs. Engineering
Structures, 2008. 30(12): p. 3566-3573.
11. George, S.J. and Y. Tian, Structural performance of reinforced concrete flat plate buildings
subjected to fire. International Journal of Concrete Structures and Materials, 2012. 6(2): p. 111-121.
12. Dassault Systèmes, ABAQUS version 6.14 documentation. 2014.
13. Genikomsou, A.S. and M.A. Polak, Finite element analysis of punching shear of concrete slabs
using damaged plasticity model in ABAQUS. Engineering Structures, 2015. 98: p. 38-48.
14. Albrifkani, S. and Y.C. Wang, Explicit modelling of large deflection behaviour of restrained
reinforced concrete beams in fire. Engineering Structures, 2016. 121: p. 97-119.
15. Ellobody, E. and C.G. Bailey, Modelling of unbonded post-tensioned concrete slabs under fire
conditions. Fire Safety Journal, 2009. 44(2): p. 159-167.
16. Wang, G., Performance of Reinforced Concrete Flat Slabs Exposed to Fire. 2006: Department of
Civil Engineering, University of Canterbury.
17. Kodur, V. and A. Agrawal, An approach for evaluating residual capacity of reinforced concrete
beams exposed to fire. Engineering Structures, 2016. 110: p. 293-306.
18. Genikomsou, A.S. and M.A. Polak, 3D finite element investigation of the compressive membrane
action effect in reinforced concrete flat slabs. Engineering Structures, 2017. 136: p. 233-244.

132

View publication stats