Computational Simulation of the Thermal

Effects on Composite Slabs Under Fire

Conditions

Paulo A. G. Piloto, Carlos Balsa,

Fernando Ribeiro & Ronaldo Rigobello

Mathematics in Computer Science

ISSN 1661-8270

Math.Comput.Sci.
DOI 10.1007/s11786-020-00466-0

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
Nature Switzerland AG. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.

1 23
 Author's personal copy
Math.Comput.Sci.
https://doi.org/10.1007/s11786-020-00466-0 Mathematics in Computer Science

Computational Simulation of the Thermal Effects on

Composite Slabs Under Fire Conditions
Paulo A. G. Piloto · Carlos Balsa ·
Fernando Ribeiro · Ronaldo Rigobello

Received: 15 July 2019 / Revised: 13 January 2020 / Accepted: 6 February 2020

© Springer Nature Switzerland AG 2020

Abstract A computational model is presented to evaluate the thermal effects on composite slabs with still deck,
originated by standard fire exposure. Composite slabs with profiled steel deck are widely used in buildings which
require fire resistance. Computational simulations are of great importance in this field and consist of an alternative
to experimental fire tests that are expensive, time-consuming and require semi-specialized technical equipment.
However, computational simulations must be reliable and realistic. The resulting transient and non-linear thermal
problem is solved by the Finite Element Method in ANSYS and Matlab. The finite element models are three-
dimensional, full scale, and multi-domain. Additionally, the models also include an air gap between the steel deck
and the concrete part of the slab, in order to simulate the thermal effects induced by the debonding between the steel
deck and the concrete, verified in previous experimental investigations. The results of the numerical simulations
are validated against the results of experimental fire tests. The fire resistance of the composite slabs determined
computationally is also compared with simplified calculation methods available in standards.

1 Introduction

Composite slabs with profiled steel deck are slabs that use steel deck as a permanent formwork and reinforced
concrete placed on the top (see Fig. 1). This fact represents one of the main advantages of this building solution,
because it reduces the construction time and requires less concrete, providing slender slabs. The use of composite
slabs in buildings has become very popular since 1980. The overall depth usually varies between 100 and 170 mm.
The thickness of the steel deck normally varies from 0.7 mm to 1.2 mm and this part of the element is normally
galvanized to increase durability [10].

P. A. G. Piloto (B)· C. Balsa

Instituto Politécnico de Bragança, Campus Santa Apolónia, 5300-253 Bragança, Portugal
e-mail: ppiloto@ipb.pt
C. Balsa
e-mail: balsa@ipb.pt
F. Ribeiro · R. Rigobello
Campus Campo Mourão-Via Rosalina Maria dos Santos, 1233 CEP: 87301-899 Campo Mourão, PR, Brasil
e-mail: rigobello@utfpr.edu.br
R. Rigobello
e-mail: f.freire12@gmail.com
 Author's personal copy
P. A. G. Piloto et al.

Fig. 1 Typical layout of a

composite slab with
trapezoidal steel deck

Composite slabs plays an important role in preventing the spread of fire in buildings. Consequently, this building
element requires fire resistance in accordance to regulations and standards. The fire rating of this type of element
is then determined by standard fire tests, accounting for Load bearing (R), Integrity (E) and Thermal Insulation
(I). This means that for a floor slab to demonstrate fire resistance in a standard fire test, it must not only maintain
structural stability and integrity, but it must also provide thermal insulation, thus limiting the increase in temperature
on the unexposed surface [14]. This work is concerned with the thermal insulation criterion (I). According to this
criterion, which is specified in the standard EN 1994-1-2 [7], the fire resistance ti is the minimum time required
for an average temperature rise of 140 ◦ C, over the initial average temperature, or a maximum temperature rise of
180 ◦ C, over the initial average temperature, on the unexposed surface of the slab, when this element is subjected
to the standard fire ISO 834 [11] from below. This condition assumes a generalized fire in a comportment. This
means that the gas temperature is the same in all space point from bellow. This assumption is based on the flashover
condition, used according to EN 1991-1-2 [4].
There are three main ways to determine the fire rating of a composite slab. The first consists of submitting the slab
to experimental fire tests. The second consists of estimating the fire resistance based on the simplified calculation
method presented in the Annex D of the EN 1994-1-2 [7]. The third consists of simulating the experimental
fire tests through powerful computer programs. Computational simulations are of great importance in this field
because experimental fire tests are expensive, time-consuming and require semi-specialized technical equipment.
The simplified calculation method is based on studies that have been carried out a long time ago and are currently
outdated. As an alternative to experimental fire tests, computational simulations must be reliable and realistic to
describe physical phenomena.
In 1983, The European Convention for Constructional Steelwork ECCS [8], published some design rules applied
to the design of composite concrete slabs with a profiled steel deck, exposed to a standard fire. This document also
presents a resume of several experimental tests developed in different European testing laboratories. According to
this document, the explicit fire design calculation for composite slabs is not required when the fire requirements are
smaller or equal than 30 minutes. This rule should only be applied if the slab was safely designed for room tem-
perature. For the other cases, simple calculation formulae were presented in a basis of conservative approximations
for a safer design procedure. In this technical note, it is also assumed that if the insulation criterion is fulfilled, then
the integrity criterion is also fulfilled. The technical note identified the existence of the membrane effect when the
composite slab is relatively well attached to the boundary of the building structure.
In 1991, Hamerlinck [10] conducted a numerical and experimental study regarding the thermal and mechanical
behaviour of reinforced composite slabs under fire conditions. Both numerical models were experimentally validated
with loaded and unloaded tests. Due to the melting of the zinc layer and surface blackening, the resulting emissivity of
galvanized steel decks was calculated as temperature dependent. The testing programme took into consideration the
most important parameters for fire resistance and a new computer program was developed, enabling two dimensional
simulations at low computational cost (low time processing). It was concluded that the developed thermal model
provided satisfactory results, although not including three-dimensional thermal effects.
 Author's personal copy
Computational Simulation of Composite Slabs under Fire Conditions

In 2002, Lim [16] published an analysis of the influence of both compressive membrane action on the behaviour
of restrained one-way flat slabs and tensile membrane action on unrestrained two-way composite and flat slabs.
In the first case, different boundary conditions were analysed to study and compare the various types of support
conditions in structures. In the second case, fire tests on axially unrestrained slabs and numerical modelling of the
tested slabs were performed. The main objectives of this research were to ascertain if tensile membrane forces could
avoid failure of simply supported two-way slabs during fire exposure and whether axial restraint could enhance
the fire resistance of one-way slabs. The numerical simulations were performed using shell and beam elements
on the non-linear finite element program SAFIR [9] and the validation of the developed numerical model was
accomplished based on experimental fire tests. In general, it was observed that the behaviour of one-way slabs in
fire conditions had great dependence on the support conditions and owing to tensile membrane action, two-way
slabs had better fire resistance than one-way slabs. Moreover, the results of modelling of the fire tests showed good
accordance with the experimental results, among other conclusions.
In 2008, an orthotropic slab finite element was implemented by Yu et al. [20] to perform numerical simulations
on composite slabs under fire conditions. The continuous part of the slab was represented by solid slab elements and
the ribbed part by special beam elements with uniaxial attributes. The thermal analyses were carried out on software
VULCAN and took into consideration the moisture evaporation in the concrete and the temperature-dependent
thermal properties of both steel and concrete. The model was validated against three fire tests and a parametric
study was conducted to evaluate the influence of the rib shapes on the thermal and structural performance of
fire exposed composite slabs. The results of the parameter study showed that the orthotropic model sensitively
considered the influence of the deck shape and a simple calculation method to estimate the fire resistance level was
proposed.
In 2017, Jiang et al. [13] from the National Institute of Standards and Technology (NIST) presented a numerical
study based on detailed and reduced-order models of heat transfer in composite slabs. Two-dimensional thermal
analyses were performed using the LS-DYNA finite-element software. The main objective of this research was to
develop a reduced-order modelling approach applicable for both thermal and structural analysis, in order to simplify
the analysis of the structural behaviour under fire conditions. Solid elements were used for the concrete slab and shell
elements for the steel deck. Both detailed and reduced-order models were validated against experimental tests and a
parametric study using the detailed model was conducted to evaluate the effect of some parameters on temperature
development such as thermal boundary conditions, thermal properties of materials and slab geometry. In addition,
the specific heat of the concrete was modified to better estimate the heat input in the web, and thereafter an equation
for the modification was suggested. In order to consider the effects of the change in emissivity of the galvanized
steel deck due to the melting of the zinc layer, a novel method to calculate the temperature-dependent emissivity
was proposed. Generally speaking, it was observed that satisfactory results did not require a great refinement of the
finite element mesh and temperatures at the unexposed side were mainly affected by the thickness of the concrete
topping. The results of the proposed model for emissivity showed better agreement with experimental results than
those calculated from the standard EN 1994-1-2 [7].
In 2018, an investigation of the thermal performance of composite slabs under standard fire conditions was
conducted by Prates [18]. The key objective of this study was to develop two-dimensional numerical models using
the software MATLAB and ANSYS in order to evaluate the fire resistance of different slab configurations according
to the insulation criterion. Several numerical simulations were performed with the aim of analysing the effect of
both concrete and steel deck thicknesses on the unexposed side temperature. Considering that the thermal behaviour
is not influenced by the mechanical behaviour, experimental fire tests were conducted on two unloaded samples.
Moreover, the results of numerical simulations were compared against results obtained with the experimental tests
as well as the simplified method given in EN 1994-1-2 [7] and NBR 14323 [2]. The fire resistance obtained from
the numerical models was considerably smaller than those measured on the experimental tests and a comparison
between the numerical and simplified method results evidenced that the design rules seem to be unsafe if a perfect
contact is considered for the advanced calculation method. According to the numerical results, a new and better
approach was proposed, considering a quadratic variation between the fire resistance and the effective thickness of
the composite slab.
 Author's personal copy
P. A. G. Piloto et al.

In the present work, four experimental fire tests are simulated computationally. Three-dimensional models are
presented to simulate the thermal effects of standard fire exposure on full composite slabs models are performed. The
full scale tests where simulated with three dimensional finite elements. The numerical models comprise different
physical domains with different thermal properties, corresponding to the components of the slab such as concrete,
steel deck, rebars and anti-crack mesh (see Fig. 1). In addition, an air gap with constant thickness is included between
the steel deck and concrete topping in order to simulate debonding effects. Effectively, previous studies mention the
separation between the steel deck and concrete during fire exposure, which increases the thermal resistance in this
interface (see for instance Li et al. [15]). In order to simulate the debonding effects, an alternative thermal model is
utilized, including an air gap with a constant thickness between the steel deck and concrete topping.
The resulting heat transfer problem is solved numerically through the Finite Element Method using two differ-
ent computational tools. The first model uses the software ANSYS, specially designed for solving multiphysics
problems. The second model uses the generic software Matlab combined with the Partial Differential Equation
PDE Toolbox. The simulations are validated with experimental results published by four different investigations,
namely Lim and Wade [16], Prates [18], Abdel-Halim et al. [1] and Hamerlinck [10]. The fire resistance criterion
for insulation (I) is evaluated for all the four different composite slabs, with different values of the air gap thickness,
comparing the results with the experimental tests as well as the simplified calculation method given in the standard
EN 1994-1-2 [7]. Additionally, a new formula based on the computational results is proposed for the calculation of
the fire resistance, depending on the effective thickness of the composite slab and the air gap thickness.

2 Heat Transfer Problem

The heat transfer problem has to be solved in the multi-domain body corresponding to the composite slab, subjected
to the heat fluxes resulting from standard fire boundary conditions. The methodology used to model and numerically
determine the thermal effects on composite slabs subjected to standard fire conditions is outlined in this section.

2.1 Physical Domains

Three dimensional, transient heat transfer problem is to be solved on four different composite trapezoidal slabs.
Each of these slabs was experimentally tested: slab 1 was tested by Lim and Wade [16] (test number 4), slab 2 was
tested by Lucas Prates [18] (test number 1), slab 3 was tested by Abdel-Halim et al. [1] (test number 2) and slab 4
was tested by Hamerlinck [10]. The dimensions of the profiles of slabs 1, 2, 3 and 4 are presented in Figs. 2, 3, 4
and 5, respectively.
Each 3D model is a realistic representation of the physical model, corresponding to a composite slab. The model
considers the exact shape of the surfaces, with no simplification, and the full scale dimensions. For each slab, a
multiple sub-domains corresponding to 4 different components is presented: steel deck, concrete topping, rebars
and anti-crack mesh.
Slab 1 has a clear span of 3160 mm by 4160 mm and is composed by normal weight concrete with siliceous
aggregates with a moisture content of 5.6% by weight. The thickness of the steel deck is 0.75 mm and concrete
topping that varies from h 1 = 75 mm to h 1 + h 2 = 130 mm. It includes an anti-crack mesh with a diameter of 8.7
mm and spacing of 300 mm. For more details see Fig. 2. The initial temperature in all the components of the slab
1 is set to 13 ◦ C in agreement with the experimental test.
Slab 2 has a clear span of 985 mm by 916.8 mm and is composed by normal weight concrete with a moisture
content of 3.0% by weight. The thickness of the steel deck is 1.2 mm and the concrete topping varies from h 1 = 40
mm to h 1 + h 2 = 100 mm. It include 4 rebars with a diameter of 20 mm and an anti-crack mesh with a diameter
of 10 mm and spacing of 150 mm. For more details see Fig. 3. The initial temperature in all the components of the
slab 2 is set to 20 ◦ C in agreement with the experimental test.
 Author's personal copy
Computational Simulation of Composite Slabs under Fire Conditions

Fig. 2 Profile of slab 1: dimensions in millimetres (from [16])

Fig. 3 Profile of slab 2: dimensions in millimetres (from [18])

Fig. 4 Profile of slab 3:

dimensions in millimetres
(from [1])

Fig. 5 Profile of slab 4:

dimensions in millimetres
(from [10])

Slab 3 has a clear span of 900 mm by 1200 mm and is composed by a normal weight concrete. Prior the test, the
specimen was conditioned with the aim of reducing the moisture to air dry conditions. The thickness of the steel
deck is 1.0 mm and the concrete topping varies from h 1 = 60 mm to h 1 + h 2 = 110 mm. It includes 3 rebars with a
diameter of 16 mm and an anti-crack mesh with a diameter of 12 mm and spacing of 200 mm. For more details see
Fig. 4. The initial temperature in all the components of the slab 3 is set to 20 ◦ C in agreement with the experimental
test.
Slab 4 has a clear span of 660 mm by 3200 mm and is composed by normal weight concrete with a moisture
content of 3.5% by weight. The thickness of the steel deck is 0.75 mm and the concrete topping varies from h 1 = 70
mm to h 1 + h 2 = 143 mm. It includes 3 rebars with a diameter of 10 mm and an anti-crack mesh with a diameter
of 6 mm and spacing of 150 mm. For more details see Fig. 5. The initial temperature in all the components of the
slab 4 is set to 20 ◦ C in agreement with the experimental test.
 Author's personal copy
P. A. G. Piloto et al.

Fig. 6 Thermal properties of slab components

The materials that compose the physical sub-domains of the slabs are mainly carbon steel (steel deck, rebars and
anti-crack mesh) and concrete (concrete topping). Additionally, in some models an air gap with constant thickness is
included between the steel deck and concrete topping in order to simulate the separation between these components
during fire exposure, which increases the thermal resistance in this interface.

2.2 Transient-State Heat Conduction

The heat conduction inside the physical domain, corresponding to the slab, is mathematically modelled by the
energy conservation equation

∂T
ρ (T ) C p (T ) = ∇ · (λ (T ) ∇T ) , (1)
∂t

where T represents the temperature (◦ C), ρ(T ) is the specific mass (kg/m3),
  ) is the specific heat (J/kgK),
C p(T
λ(T ) is the thermal conductivity (W/mK), t is the time (s) and ∇ = ∂x , ∂ y , ∂z is the gradient. Equation (1), is
based on the heat flow balance, for the infinitesimal material volume, in each spatial direction.
Equation (1) is non-linear because the properties of the material are temperature dependent. The specific mass
ρ(T ), the specific heat C p(T ) and the thermal conductivity λ(T ) of the different materials used in the composite
slab vary with the temperature. The thermal properties of these materials are determined by standards used for
composite slabs [5–7].
 Author's personal copy
Computational Simulation of Composite Slabs under Fire Conditions

Fig. 7 Boundary conditions

Figure 6 depicts the evolution of the thermal properties along with temperature. The conductivity of steel
decreases as long as the temperature increases and the specific heat has a strong variation due to the allotropic
phase transformation. The density and the conductivity of concrete decrease with temperature, and for the second,
the upper limit have been selected for the simulations. The specific heat of concrete presents a peak value related
to 5.6% of moisture content for slab 1, 3% for slabs 2 and 3, and 3.5% for slab 4. Figure 6 also depicts the thermal
properties of air. These properties are temperature dependent and were used to simulate the insulating layer between
the steel deck and the bottom surface of the concrete topping.
Equation (1) is time-dependent because the heat flux on the boundary exposed to the fire change with time, i.e,
the thermal state of the slab is transient. The solution of Eq. 1 is required to determine the distribution of temperature
inside the physical domain along the time and, consequently, determine the fire resistance. To solve this problem,
the boundary conditions are required.

2.3 Boundary Conditions

The composite slab is subjected to three main boundary conditions. The lower part of the slab is exposed to fire;
the upper part (unexposed surface) is in contact with air, and the other surfaces are insulated. These conditions are
outlined in Fig. 7.
The boundary conditions in the lower part of the slab, correspond to heat transfer by convection and radiation,
given by the equation
 
λ (T ) ∇T.−
→
n = αc (T∞ − T ) + φεm  f σ T∞
4
− T4 (2)

where − →n the unitary vector normal to the external face, φ is the view factor, αc is the convection coefficient, εm is
the emissivity of the material, ε f emissivity of fire, σ the Stefan–Boltzmann constant and T∞ the gas temperature
of the fire compartment. This equation is based on the heat flow that enters to the exposed surface of the steel deck
(by conduction) and that arrives by heat transfer conditions (radiation and convections) from below, using the gas
bulk temperature.
In Eq. (2), the convection coefficient is αc = 25 W/m2 K on the exposed side. The emissivity of carbon steel is
m = 0.7 and the fire emissivity is ε f = 1, in agreement with the standards.
The view factor, φ, specified in the Eq. (2), quantifies the geometric relation between the surface emitting radiation
and the receiving surface. This parameter is dependent on the surfaces areas and orientations, as well as the distance
 Author's personal copy
P. A. G. Piloto et al.

between them. There are three different surfaces in the lower part of slab (see Fig. 7). The view factor of the lower
flange is given as φ = 1. The view factor of the web and upper flange of the steel deck are smaller than one, due
to the obstruction caused by the ribs of the steel deck. This view factor depends on the area and orientation of
the surface emitting radiation in relation to the surface receiving radiation, as well as the distance between them.
These equations are based on the crossed-string method developed by Hottel in the 1950s [21]. This method is also
incorporated in the standard EN 1994-1-2 [7]. The resulting equations for the upper flange (φup ) and web (φweb )
view factors are presented in Eqs. (3) and (4). These equations allow the calculation of the view factors as functions
of the parameters that describe the geometry of the slabs illustrated in Figs. 2, 3, 4 and 5.
     
2 2
l1 −l2 l1 −l2
h 22 + l3 + 2 − h 22 + 2
φup = (3)
l3
  2   2
l1 −l2 l1 −l2
h 22 + 2 + (l3 + l1 − l2 ) − h 22 + l3 + 2
φweb =    (4)
2
l1 −l2
2 h 22 + 2

This view factor depends on the area and orientation of the surface emitting radiation in relation to the surface
receiving radiation, as well as the distance between them.
In Eq. (2), the gas temperature of the fire compartment follows the standard fire curve ISO 834 (T∞ = TISO )
given by

TISO = 20 + 345 log10 (8t + 1) , (5)

where TISO is given in (◦ C) and t in minutes [11]. This standard specifies a test method for determining the fire
resistance of construction elements.
On the upper part of the slab there will be heat transfer from the slab to the above compartment. The standards
enables to consider the joint effect of convection and radiation on the unexposed side by the use of the convection
coefficient αc = 9 W/m2 K [6]. The boundary condition in the upper surface of the slab is given by Eq. (6),

λ (T ) ∇T.−
→
n = αc (T∞ − T ) (6)

where T∞ is the room temperature. This equation is based on the heat flow that exits by the unexposed surface of
the concrete topping (by conduction) and that flows by the heat transfer condition (convection), using the constant
room temperature for the gas bulk temperature.
The other four surfaces of the slab (front, back, left and right) are considered insulated, i.e, the boundary conditions
applied to these faces are given by Eq. (7),

λ (T ) ∇T.−
→
n =0 (7)

Equation (7) is based on the adiabatic condition, applied to the external lateral surfaces of the slab.

3 Numerical Solution Through Finite Element Method

The Eq. (1) is solved numerically through the Finite Element Method (FEM) on the three-dimensional domain.
Two 3-D models of each slab are generated separately in ANSYS and Matlab. These models are composed by
sub-domains which correspond to the different materials. For each sub-domain, a specific 3-D finite element is
used, according to the geometry of the sub-domain to be simulated. The spatial discretization of the domain give
 Author's personal copy
Computational Simulation of Composite Slabs under Fire Conditions

Fig. 8 Finite element meshes of slab 2 created in ANSYS

rise to a finite element mesh. The values of the temperatures at the nodes of the mesh are determined according to
the boundary conditions.
The general procedure of the finite element method for solving Eq. (1), based on the weak-form Galerkin model
and from the minimum condition for the weighted residual method, leads to the matrix formulation energy of
equation.

C(T )Ṫ + K (T )T = F (8)

where C is the capacitance matrix, Ṫ is the vector of time derivatives of the temperatures, K the conductivity
matrix and F the vector of the thermal loads (see, for instance, [12]). The capacitance and conductivity matrices are
temperature dependent because they include the contributions of the thermal properties. The vector of the thermal
loads F includes the boundary conditions. The solution of the first order non-linear system of ordinary differential
equations given by Eq. (8), given T (t0 ) = T 0 and the boundary conditions,
 enables to determine the temperature
at each node of the finite element mesh over the time interval t0 , t f . The method of solving this equation differs
from ANSYS to Matlab.

3.1 Computational Solution with ANSYS

ANSYS is a well known commercial software suitable for structural, fluid and thermal problems. All the models
where generated with ANSYS Mechanical APDL 18.2, a 3D finite element mesh composed by sub-domains
corresponding to different materials, namely the concrete topping, the steel deck, the steel rebars and the steel
mesh. Additionally, in some finite element models, an air gap is included between the steel deck and the concrete.
For each material, a specific 3D finite element is used, according to the geometry of the sub-domain to be simulated.
Three different types of finite elements are used: SHELL131, SOLID70 and LINK33.
The SHELL131 element has four nodes with up to 32 degrees of freedom per node, depending on the number
of layers. The SOLID70 element presents eight nodes with a single degree of freedom (temperature) at each node.
This finite element is used to model the concrete topping and, in some cases, the air gap. The LINK33 element has
two nodes with a single degree of freedom (temperature) per node. The LINK33 element is used to model the steel
reinforcement, that is, the anti-crack mesh and the rebars. All these elements use linear interpolating functions (for
details see [3]). The meshes of slab 2 created with these elements are presented in Fig. 8.
The built-in ANSYS
 solution
 method for Eq. (8) is based on the approximation of the time derivative by finite
difference Ṫ ≈ T n+1 − T n /Δt, where T n is the temperature at time tn and Δt = tn+1 − tn , jointly with
T ≈ (1 − θ ) T n + θ T n+1 , for 0 < θ < 1. Replacing these formulae in Eq. (8), the non-linear system of algebraic
equations

  T n+1 − T n      
C T n+1 + θ K T n+1 T n+1 = F T n+1 − (1 − θ ) K T n+1 T n . (9)
Δt
 Author's personal copy
P. A. G. Piloto et al.

Fig. 9 Finite element meshes of slab 3 created in Matlab

Fig. 10 Matlab mesh

convergence study

is obtained. The solution of this system, with an optimized parameter θ , enables to determine the temperature field
inside the composite slab at time tn+1 . Equation (9) is solved iteratively by Newton–Raphson method at each time
step. The convergence criterion is based on the reduction of the residual of the system (9) above a tolerance times
a reference value. In this work, a tolerance of 10−3 and a minimum reference value of 10−6 are considered.

3.2 Computational Solution with Matlab

The Matlab (R2019a) Partial Differential Toolbox was used to create a computational model of the composite slab
under fire conditions. The finite element model of the slab created with Matlab is simpler than the one created with
ANSYS, because it uses only one type of finite element. Tetrahedral finite elements are defined by four nodes (at the
corners) using linear interpolation functions. The finite element mesh is composed by either two or three different
sub-domains. One model considers only the steel deck and the concrete topping, and the other considers the steel
deck, the air gap and the concrete topping (see Fig. 9).
The maximum mesh edge length Hmax varies from simulations to simulations, usually standing between 0.015
and 0.02 m. These values have been selected according to convergence tests using different mesh sizes. Figure 10
presents the convergence of results with the mesh refinement, for the temperature at one point on the unexposed
surface of a composite slab, after 60 min of fire exposure.
In Matlab, the Eq. (8) is converted in

C(T )Ṫ = F̄ (T ) (10)

where F̄ (T ) = F − K (T ). The solution of Eq. (10) is made by the built-in function ode15s [19]. The algorithm
implemented in this function is based on the discretization of the time derivative by numerical differentiation
formulas (NDFs) of orders 1–5. The order of accuracy of the solution can be explicitly controlled through the
absolute or relative tolerance parameters. In this work a value of 10−2 is set to the absolute tolerance, and 10−1 is
set to the relative tolerance. In each time step, the non-linear system of algebraic equation is solved through Newton–
 Author's personal copy
Computational Simulation of Composite Slabs under Fire Conditions

Table 1 Coefficients for determination of the fire resistance for NWC [7]
a0 (min) a1 (min/mm) a2 (min) a3 (min/mm) a4 (min.mm) a5 (min)

−28.8 1.55 −12.6 0.33 −735 48.0

Raphson method whose stopping criterion can be monitored by means of the maximal number of iterations, set to
15, and the residual tolerance, set to 10−2 (for more details see [17]).

4 Simplified Calculation Method

The simplified calculation method is also presented in order to compare the results with the ones obtained with
computational simulations. The Annex D of the EN 1994-1-2 (Eurocode 4 [7]) presents a simplified method for
the calculation of the fire resistance of unprotected composite slabs exposed to fire from below, when using the
standard fire curve ISO 834, given by Eq. (5). The fire resistance (ti ) with respect to thermal insulation criterion
should be determined according to the Eq. (11).

A 1 A1
ti = a0 + a1 h 1 + a2 φup + a3 + a4 + a5 (11)
Lr l3 lr l 3

where A/L r , the rib geometry factor of the slab, should be determined as follows:

A h 2 ((l1 + l2 ) /2)
= (12)
Lr l2 + 2 h 22 + ((l1 − l2 ) /2)2

In addition to the geometric parameters of the slab, previously defined, the fire resistance also depends on partial
factors (ai ). Table 1 presents these factors for normal weight concrete (NWC).

5 Validation of the Computational Model

The Figs. 11, 12, 13 and 14 illustrate the temperature development (numerical and experimental) at different selected
points as well as the average and maximum temperatures at the unexposed surface of the slab for the four validation
models. In these results, ANS δa and MAT δa stands for ANSYS and Matlab computational models with and air
gap thickness equal to δa . If δa is 0 it means that the contact between the steel deck and the concrete topping is
perfect.
From the results of slab 1 (Fig. 11), a reasonable agreement is observed between the furnace temperature and
the standard fire curve ISO 834. For both perfect contact models ANS 0 and MAT 0, a considerable difference to
experimental results (EXPT) can be observed in the first stages of heating. The numerical results from MATLAB are
slightly closer to experimental results than results from ANSYS. The air gap models ANS 1 and MAT 1 (inclusion of
an air gap with 1 mm between the concrete and the steel deck) provide a good agreement to measured temperatures
for all selected points. With respect to the unexposed surface, a big difference between the results of the perfect
contact models and the air gap models is verified. For the models ANS 1 and MAT 1, the agreement between the
computed and measured temperatures is very good for both average and maximum temperatures until 76 minutes.
After that, some differences are noticed, probably due to experimental deviations such as the furnace temperature,
for example. In general, the results of the air gap models ANS 1 and MAT 1 are closer to each other in comparison
to the perfect contact models ANS 0 and MAT 0. Small differences between the results of ANSYS and MATLAB
 Author's personal copy
P. A. G. Piloto et al.

Fig. 11 Numerical and experimental results for slab 1—Points 1, 2 and 3 at distance 20, 70 and 130 mm from the top

Fig. 12 Numerical and experimental results for slab 2—Points 1, 2 and 3 at distance 20, 15 and 15 mm from the top

models are observed, with exception of the average and maximum temperatures on the unexposed surface for the
perfect contact models ANS 0 and MAT 0.
Regarding the slab 2 (Fig. 12), the furnace temperature is very close to the standard fire curve ISO 834, although
presenting a small deviation in the first minutes of fire exposure. For the perfect contact models ANS 0 and MAT
0, the temperatures are higher than experimental results (EXPT) throughout the entire duration of the test. The
results of the air gap models ANS 3 and MAT 3 (inclusion of an air gap with 3 mm) are considerably smaller than
the perfect contact models and present satisfactory agreement with measured temperatures, principally for points 2
and average temperature at the unexposed side. For points 1 and 3, the results of the air gap models present good
agreement with measured temperatures until the first 42 min of fire. Some differences are observed between the
results of ANSYS and MATLAB models, being greater for the perfect contact models (ANS 0 and MAT 0).
For slab 3 (Fig. 13), it is noteworthy that the furnace temperature is under the standard fire curve ISO 834, which
was used for the numerical simulations. Consequently, the numerical results should be higher than experimental
results. For the perfect contact models ANS 0 and MAT 0, a good agreement with experimental results (EXPT) is
observed for point 1 until the first 44 min of fire. After that, better results are obtained using the air gap models ANS
2 and MAT 2 (inclusion of an air gap with 2 mm). This is because during a fire, the debonding of the steel deck from
concrete occurs after a period of time. A delay in the rate of temperature increase is observed in the last minutes
of fire on curve ANS 0 on P2. It can be justified by the coincidence of two different materials (steel and concrete)
on the same node and the abrupt change in the specific heat of steel for temperatures between 700 ◦ C and 800 ◦ C,
see Fig. 6a). In general, the results of the air gap models present a better agreement with measured temperatures in
comparison to the results of the perfect contact models. For the selected points, models ANS 0 and MAT 2 present
 Author's personal copy
Computational Simulation of Composite Slabs under Fire Conditions

Fig. 13 Numerical and experimental results for slab 3—Points 1, 2 and 3 at distance 50, 85 and 70 mm from the top

Fig. 14 Numerical and experimental results for slab 4—Points 1, 2 and 3 at distance 20, 74 and 123 mm from the top

better agreement with experimental results when compared to models MAT 0 and ANS 2, respectively. Considerable
differences between ANSYS and MATLAB models are evident, mainly for curves ANS 0 on P2 and MAT 0 P2.
On the other hand, the results of these models are very close to each other on the unexposed surface.
From the results of the slab 4 (Fig. 14), it can be concluded that the temperature development on the selected
points is quite similar between the experimental (EXPT) and the perfect contact model (NUM 0) at the first minutes
of heating. For temperatures over 100 ◦ C, the model with the air gap of 1 mm (NUM 1) present good approximation
to the experimental results for point 2. However, for the points 1 and 3, the perfect contact model presents better
agreement with measured temperatures at the last minutes of heating. For the unexposed surface, the maximum and
average temperature curves are very close for all the models. Better agreement with the experimental results can be
noticed using the model with the air gap.
The following tables present the results obtained for the fire resistance according to the insulation criterion with
respect to the average temperature rise (t Ave) and the maximum temperature rise (t Max) at the unexposed surface
for each slab.
The results obtained with the air gap models underestimate the fire resistance for slab 1, with a relative error of
9.2% for ANSYS and 9.9% for MATLAB. Better approximation to experimental results is observed for the standard
EN 1994-1-2 provisions, with a relative error of 6.7% (see Table 2).
With respect to the slab 2, a good agreement between the fire resistance obtained with the air gap models and
experimental results is achieved, resulting in a relative error of 1.0% for ANSYS and 3.4% for MATLAB. An
underestimated value is obtained using the EN 1994-1-2 provisions, with a relative error of 38.9% (see Table 3).
 Author's personal copy
P. A. G. Piloto et al.

Table 2 Fire resistance of slab 1: experimental, numerical and analytical results

Model ANS 0 MAT 0 ANS 1 MAT 1 Exper EN 1994-1-2 (ti )

t Ave (min) 65.6 76.7 97.0 97.7 102.7 98.5

t Max (min) 62.6 71.4 93.3 92.5 103.0

Table 3 Fire resistance of slab 2: experimental, numerical and analytical results

Model ANS 0 MAT 0 ANS 3 MAT 3 Exper EN 1994-1-2 (ti )

t Ave (min) 28.3 29.7 62.8 62.3 62.2 38.0

t Max (min) 27.0 26.5 65.9 60.1 −

Table 4 Fire resistance of slab 3: experimental, numerical and analytical results

Model ANS 0 MAT 0 ANS 2 MAT 2 Exper EN 1994-1-2 (ti )

t Ave (min) 49.4 51.4 82.6 82.9 80.0 73.9

t Max (min) 44.1 46.5 78.3 79.1 −

Table 5 Fire resistance of slab 4: experimental, numerical and analytical results

Model ANS 0 MAT 0 AN S1 M AT 1 E x per EN 1994-1-2 (ti )

t Ave (min) 75.6 73.6 93.5 91.6 88.2 106.5

t Max (min) 84.8 82.0 105.1 102.5 102.1

Regarding the slab 3, the results evidence that the fire resistance obtained using the air gap models is slightly
underestimated, with a relative error of 2.1% for ANSYS and 1.1% for MATLAB. In addition, a good agreement
between EN 1994-1-2 calculations and experimental data is also observed, resulting in a relative error of 7.6% (see
Table 4).
Concerning the slab 4, the results show that the air gap model slightly overestimated the fire resistance, with a
relative error of 6% for ANSYS and 3.9% for Matlab. The EN 1994-1-2 provisions overestimated the fire endurance,
providing an unsafe result with a relative error of 20.7% (Table 5).
In experimental fire tests, the air layer which appears due to debonding of the steel deck from concrete does not
have constant thickness. This means that the air gap is 0 mm thick at the beginning of the test (perfect contact) and
as the specimen is heated, the steel deck separates slowly from the concrete, hence increasing the air gap thickness.
By means of simplification, this work assumes that the air gap has constant thickness throughout the whole duration
of the simulations. The thicknesses of the air layer used in the air gap model for the four slabs (1 mm, 3 mm and
2 mm) were determined from a parametric analysis, selecting the values that best fit with the experimental data for
each slab.

6 Parametric Analysis

A parametric study is conducted in order to determine the influence of the concrete and air gap thickness was
developed to determine the fire resistance of composite slabs according to the thermal insulation criterion. A total
of 40 numerical simulations have been carried out, of which 32 performed in ANSYS, using the perfect contact
model and the air gap model; and 8 performed in MATLAB, using the air gap model.
 Author's personal copy
Computational Simulation of Composite Slabs under Fire Conditions

Table 6 Investigated parameters of the parametric study

h 1 (mm) δa (mm)

ANSYS 50, 60, 70, 80, 90, 100, 110, 120 0, 1, 2, 3

MATLAB 50, 60, 70, 80, 90, 100, 110, 120 3

Fig. 15 New proposal for

the fire resistance (I) of
composite slabs

The O Feliz H60 steel deck profile (slab 2) has been selected to perform the numerical analyses. Furthermore,
some parameters have been fixed, namely the rebar diameter (8 mm) and the anti-crack mesh diameter (6 mm) and
spacing (100 mm). Table 6 presents the ranges of investigated parameters of this study.
Figure 15 presents the results for the fire resistance (I) obtained through the computational models (ANSYS
and MATLAB); experimental fire tests (EXPT); and simplified calculation methods presented in Sect. 4. The
experimental results were published in 1983 by the European Convention for Constructional Steelwork (ECCS)
[8]. These results are given as function of the effective thickness h eff . The EN 1994-1-2 [7] states that the effective
thickness of a composite slab h eff should be calculated according to

⎧  
⎪
⎪ h 1 + 0.5h 2 ll11 +l 2
if h 1 / h 2 ≤ 1.5
⎨ +l3
h eff = (13)
⎪
⎪   
⎩h 1 + 0.75 l1 +l2 if h 1 / h 2 > 1.5.
1 l1 +l3

A model with quadratic dependence between the fire resistance t f and the effective thickness h eff is chosen to
fit the results of the numerical simulations. A linear relationship between the effective thickness h eff of the slab
and the thickness of the air gap δa is considered in order to take into account the increase of fire resistance. The
proposed new simplified model for the fire resistance of composite slabs is given in Eq. 14.

t f = 0.006h 2eff + 0.113h eff − 5.807 + (0.142h eff + 2.467) δa (14)

where t f , the estimation of the fire resistance of the slab, is given in minutes, and h eff and δa are given in millimetres;
with the following limits: 70 ≤ h eff ≤ 150 mm and 0 ≤ δa ≤ 3 mm. Figure 15 illustrates the results obtained with
the application of Eq. (14) for different air gap thicknesses. In this figure t f 0 corresponds to the results of Eq. (14)
obtained with perfect contact (δa = 0 mm), t f 1 corresponds to δa = 1 mm, t f 2 corresponds to δa = 2 mm and t f 3
corresponds to δa = 3 mm.
Analysing the results in Fig. 15, a good agreement is observed between the results of the new proposal and the
numerical results. A satisfactory conformity between the results for fire resistance (I) of the European standard (ti )
 Author's personal copy
P. A. G. Piloto et al.

from Eq. (11) and the numerical simulations was obtained using an air gap thickness of 2 mm (t f 2 ). Good agreement
between experimental and numerical results was obtained considering an air gap thickness of 3 mm (t f 3 ).

7 Final Considerations

This study presented a discussion about the results of three-dimensional thermal analyses performed on ANSYS
and MATLAB for four different composite slabs. Full scale and realistic computational models have been developed
to simulate the evolution of the temperature inside the composite slabs submitted to standard fire conditions. The
fire resistance according to thermal insulation criterion (I) was evaluated and compared to the experimental results
and the simplified calculation method of Eurocode 4 as well. With the aim of simulating the effects of debonding
of the steel deck from concrete, an alternative thermal model was used, in which an insulating layer (air gap) with
a constant thickness was introduced between the concrete and steel deck.
For the experimental results of the four slabs, a plateau at about 100 ◦ C (due to moisture evaporation) should be
highlighted, consisting of a decrease in the rate of temperature increase. The results of the numerical simulations
do not present this pronounced plateau, probably because localized moisture concentrations in the tests were higher
than the uniform moisture content introduced in the thermal models for each slab.
The maximum temperature rise criterion was decisive for the fire resistance according to thermal insulation
criterion for most of the simulations. With respect to the experimental results, the average temperature rise criterion
governed the fire resistance. The Eurocode 4 provisions underestimated the fire resistance for all the slabs, and
for slab 2 in particular, a considerable difference was observed. The perfect contact models underestimate the fire
resistance. Therefore, it is evident that the air gap models provide much better results for fire resistance from the
thermal insulation standpoint when compared to the perfect contact models, reducing the temperature rise on the
selected points and unexposed surface as well.
In some cases, the results of ANSYS and MATLAB models presented noticeable differences for the same
situation, that is, for a same point and type of model (perfect contact or air gap model). These differences can be
justified by the presence of the steel components within the concrete layer (rebar and mesh) in the ANSYS model
which are not included in the MATLAB model. Although not significantly affecting the fire resistance (I), these
components can affect the temperature development at particular points.
Based on the numerical results, a new formula has been proposed for the calculation of the fire resistance,
depending on the effective thickness of the composite slab and on the air gap thickness. This proposed model
presents good agreement with numerical results and considers parameters which are not included in the current
calculation rules of standards.

References

1. Abdel-Halim, M.A.H., Hakmi, M.R., O’Leary, D.C.: Fire resistance of composite floor slabs using a model fire test facility. Eng.
Struct. 21, 176–182 (1999)
2. ABNT: NBR 14323 - structural fire design of steel and composite steel and concrete structures for buildings [in Portuguese] (2013)
3. ANSYS: ANSYS Mechanical APDL, 18.2, help system, Thermal Analysis Guide, ANSYS, Inc
4. CEN: EN 1991-1-2: Actions on structures - part 1-2: General actions - action on structures exposed to fire (2002)
5. CEN: EN 1993-1-2: Design of steel structures. part 1-2: General rules - structural fire design eurocode (2002)
6. CEN: EN 1992-1-2: Design of concrete structures. part 1-2: General rules - structural fire design (2004)
7. CEN: EN 1994-1-2: Design of composite steel and concrete structures. part 1-2: General rules - structural fire design (2005)
8. ECCS: Calculation of the fire resistance of composite concrete slabs with profiled steel sheet exposed to the standard fire (1983)
9. Franssen, J.M.: SAFIR - A thermal/structural program modelling structures under fire. In: NASCC 2003 (2003)
10. Hamerlinck, A.F.: The behaviour of fire-exposed composite steel/concrete slabs. Ph.D. Thesis, Eindhoven University of Technology
(1991)
11. ISO: International standard ISO 834 - fire-resistance tests: Elements of building construction (1975)
12. Reddy, J.N., G, D.: The Finite Element Method in Heat Transfer and Fluid Dynamics, 3rd edn. CRC Press, Boca Raton (2010)
 Author's personal copy
Computational Simulation of Composite Slabs under Fire Conditions

13. Jiang, J., Main, J.A., Sadek, F.H., Weigand, J.M.: Numerical modeling and analysis of heat transfer in composite slabs with profiled
steel decking. Tech. Rep. (2017)
14. Jiang, J., Pintar, A., Weigand, J.M., Main, J.A., Sadek, F.: Improved calculation method for insulation-based fire resistance of
composite slabs. Fire Saf. J. 105, 144–153 (2019)
15. Li, G.Q., Zhang, N., Jiang, J.: Experimental investigation on thermal and mechanical behaviour of composite floors exposed to
standard fire. Fire Saf. J. 89, 63–76 (2017)
16. Lim, L., Wade, C.: Experimental fire tests of two-way concrete slabs—fire engineering research report 02/12. Tech. Rep., University
of Canterbury (2002)
17. MathWorks: Partial Differential Equation Toolbox User’s Guide, Heat Transfer Problem with Temperature-Dependent Properties,
The MathWorks, Inc
18. Prates, L.M.S.: Numerical simulation of the fire behaviour of composite strucures (slabs) [in Portuguese]. Master’s Thesis, Politech-
nic Institute of Bragan ça (2018)
19. Shampine, L.F., Reichelt, M.W.: The MATLAB ODE suite. SIAM J. Sci. Comput. 18(1), 1–22 (1997)
20. Yu, X., Huang, Z., Burgess, I., Plank, R.: Nonlinear analysis of orthotropic composite slabs in fire. Eng. Struct. 30(1), 67–80 (2008)
21. Yunus A. Cengel, A.J.G.: Heat and Mass Transfer: Fundamentals and Applications. McGraw-Hill Education, Europe (2014)

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.