You are on page 1of 9

Experimental and Numerical Investigation of Fire Effect on GFRP Sheets

Used in Strengthening RC Structures Considering Anisotropic

Properties of Composite Materials
A.G. Chegini1, H. Nikopour2, M. Nehdi2, J. Akbari1
1. Department of Mechanical Engineering, Sharif University of Technology, Tehran, Iran.
2. Department of Civil Engineering, University of Western Ontario, London, Canada.

Abstract: In this study a model is presented to predict the residual strength of composite laminates exposed
to the heat flux of fire. This model calculates the number of the damaged laminas considering the charred
thickness of the composite laminate and then predicts the overall residual strength of the laminate analyzing
each lamina. Charred laminas no longer have their initial properties due to decomposition of their polymer
matrix. This model can obtain the damaged thickness by introducing a damage temperature, and solving the
energy equation for the thermochemical response of the laminate. The thickness of the charred layer is
measured for the laminate after exposure to different heat fluxes and times. The residual strength of the
laminate is investigated under tensile, buckling and bending loads. Comparisons are made between the
predicted results of thermal and mechanical models and those obtained experimentally. Good agreement
was observed between the results of mathematical and experimental models.
Keywords: High temperature Fire damage; Composite laminated; Thermal decomposition; Residual strength.

glass fiber-reinforced plates exposed to fire

(Gibson et al. 1994).
This paper considers a similar approach to the
Mouritz model, but takes into consideration the
laminated composites with anisotropic properties
(Bakaiyan et al. 2005). The model analyzes the
properties of each lamina of the laminate, which
includes the lamina thermal decomposition. When
the thermal decomposition had occurred the
lamina is assumed to have failed. The damage
region includes some laminas with changed
mechanical properties and other laminas are intact
having their original strength (Figure 1). The main
assumption of the model is that the damaged
laminate consists of two main regions; the char
and intact regions (Figure 2). The char region in
the fire-damaged composite has a uniform
thickness with a negligible change in the
mechanical properties because of the thermal
decomposition of the polymer matrix.

1. Introduction
Fiber-reinforced polymer composites are used
increasingly in the construction of marine
structures, building structures, thermal protection
of aerospace vehicles and etc used. Despite the
advantages of these composites the main concern
with the use of these materials is their high
flammability and poor fire resistance. Most
composite materials with a polymer matrix ignite
when exposed to a high temperature fire and large
amounts of heat, smoke and fumes are released
due to the thermal decomposition of the organic
constituents while the mechanical properties of the
composite are severely degraded.
An experimental fire study by Pering et al. revealed
that most mechanical properties of composites
were changed after exposure to fire, (Pering et al.
1981). Mouritz and Mathys proposed a simple
analytical model in order to calculate the post-fire
mechanical properties of burnt composites. In their
model, the post-fire properties were determined by
combining the properties of the unburnt and char
regions using a rule-of-mixtures formulation to give
the bulk stiffness and failure strength of the firedamaged composite. This model was applied to
predict the post-fire tension, compression and
flexure properties of burnt fiber-reinforced
thermoset polymer (FRP) composites (Mathys et
al. 2004). Looyeh and Gibson and their colleagues
determined the thermal response of laminated

2. Thermal modeling
A number of researchers have developed
mathematical and numerical approaches to predict
the thermal response of polymer matrix composite
decomposition. A model for the thermal response
of composite materials, validated against standard
fire-test methods, was reported by Looyeh (Looyeh
et al. 1998).The present thermal model is based on

the model developed by Looyeh and Bettess and

Gibson for offshore applications. Hence, the onedimensional heat transfer in a laminate plate,
undergoing thermochemical decomposition with no
surface recession is given by a non-linear, secondorder partial differential equation as follows:

opaque. In the absence of the radiation heat

exchange between surfaces, a simplified approach
for representing the radiation heat transfer is an
equivalent convection boundary condition where
the non-linearity is considered through a
temperature-dependent convection coefficient. The
function g(t), presented in Eq. 4, is therefore
defined as:
[7] g (t ) = ( hr + hc ) (T T )

[1] c p T = k T m& g c pg T r (Q + h h g )
x 2

Where hc is the average free convection heat

transfer coefficient (W/m2 K) and is evaluated as:

Here, introducing the endothermic decomposition

with the mass flow effect terms and assuming that
decomposing composite material and the resultant
Subject to the initial conditions:
T ( x, t ) = T , = 0 , c p = c p 0 , k = k 0
c pg = c pg 0 , m& g = 0 , s = s 0
And the following boundary conditions:
[3] k T = q
for x = 0, t > 0
[4] k ( T ) = g (t ) for x = L, t > 0
where T is the temperature (K), Cp is the laminate
specific heat (Jkg-1K-1), Cpg is the gas specific heat
(Jkg-1K-1), k is the laminate thermal conductivity
(Wm-1K-1) and s is the surface emissivity given in

And hr is the equivalent temperature-dependent

non-linear convection coefficient (W/m2 K)
expressed as:



1/ 2



= 0.825 + 0.387Ra1 / 6 / 1 + (0.492 / Pr)9 / 16

5 / 27

The solution of Eq. 1, Eq. 5 and Eq. 6 with suitable

boundary conditions provides the basis for a 1-D
thermal response model for the fire behavior of
composite materials. An implicit finite difference
scheme has been used to solve Eq. 1, with the
stated initial and boundary conditions.
3. Mechanical modeling
The classical form of the constitutive relations, in
the linear elastic beam and plate theory is used for
investigating the bending, tensile and buckling
strength of the plate as shown in Figure 4. The
beam is subjected to a lateral load in the x-z plane,
in the z direction, such that bending occurs and the
constitutive equations for beam are:

Where A=8.709248x106 is the Pyrolysis constant

(1/s), Ea= 1.29520x105 is the activation energy
(kJ/kg mol), R is the Universal gas constant (8.314
kJ/kg mol K), and n=0.75 is the order of
decomposition reaction. The laminate density is
(kgm-3) and the subscript rf denotes the final resin
The energy transferred by convection due to the
gaseous reaction products diffusing away from the
reaction front can be expressed as the product of
the gas enthalpy and mass flux. Assuming that all
of the gaseous reaction products immediately
diffuse away from the reaction front, then by
conservation of mass, the mass flux of gas must
equal the rate of material decomposition. Thus, the
mass flux and hence the effect of convection of the
reaction products can be calculated from the
knowledge of the reaction rate as expressed in Eq.
5. That is:


The Nusselt number Nu, is obtained based on an

empirical correlation for a vertical plate and is valid
for all Rayleigh (Ra) and Prandtl (Pr) numbers
(Looyeh et al. 1998) .The correlation is:

= A r 0 [( r rf ) r 0 ]n Exp (

[6] m& g = (

hr = s T 2 + T2 T + T


Table 1. The heat of decomposition is taken as

Q=3.5169x106 (Jkg-1).
The chemical reaction may be conveniently
expressed using an Arrhenius rate expression
assuming no expansion of the material as:

hc = k air Nu / l


N x A11
M x B11





Substituting Eq. 11 into equilibrium equations

results in the following differential equation for the
bending of the beam:

B11 4 w
) = q (x)
A11 x4

[12] b (D11

The stresses in each lamina are:

[13] [ x ]k = z [Q11 ]k [ x ]
Where the boundary conditions for the simple
support are given as:
[14] w = 0 , M b = 0


While the beam is loaded in the axial direction (the

x direction) the tensile force is:

The laminates experience both free convection

and radiation boundary conditions on its
unexposed surface (Figure 3). It is assumed that
the unexposed surface is diffuse, grey and

[15] P = b ( A11

B11 du 0
D11 dx

And stresses in every ply can be determined by:


[16] [ x ]k = [Q11 ]k [ x 0 ]
Furthermore, if the slender composite beam is
loaded in uniaxial compression then the beam
failure generally occurs by buckling, so the failure
mode of buckling is considered. The differential
equation of buckling for the beam is:
[17] b ( A B11 ) d w = P d w
D11 dx
dx 2
In the equation, there is a coupling between the inplane load and the lateral deflection. The buckling
mode shapes for the composite beam which is
simply supported at each end is given by:

[18] w( x) = An sin nx
n =1
and the critical values of load are:
2 2
[19] Pcr = n b ( D11 B11 )
The Tsai-Hill criterion theory was applied in plane
stress state (Gibson et al. 1994) to investigate the
failure criteria as shown below:

[20] 1 1 2 + 2 + 12 = 1
Y2 S2
The mechanical properties of laminates are given
in Table 2.

4. Experimental work
Specimen plates for fire testing were manufactured by hand lay-up and the vacuum bagging process.
Laminates were bagged at room temperature and cured in an autoclave according to the manufacturers
The materials used are specified in Table 3. The details of the laminated test plates are given in
Table 4. Experimental fire tests were done using a radiation heater. The test specimens were vertically
exposed to the heater in an area of 50 250 mm as shown in Figure 5. The temperature of each specimen
was measured using a PhotoTemp MX6 photographic infrared thermometer for measuring the temperature
of the hot and cold faces. The output of each measurement was continuously monitored throughout the test
runs and recorded using a computer data logging system. The tests were run for several times at different
heat exposures. The cold face temperature had reached at least 250C and the hot face temperature had
exceeded 400C. The tests were terminated before full penetration of the plates occurred. The specimens
were exposed to three heat fluxes, namely 20, 25 and 30, kw/m2. Buckled specimens were exposed to
20kw/m2 after seven heating periods from 90 to 195 seconds. Tension specimens were exposed to 20,
kw/m2 after three heating periods from 150 to 300 seconds and exposed to 25, kw/m2 after seven heating
times from 90 to 170 seconds. Flexure specimens were exposed to 25kw/m2 after eight heating periods from
180 to 840 seconds and exposed to 30kw/m2 after three heating periods from 240 to 300 seconds.
The post-fire tensile, buckling and flexural properties of the burnt composites were measured at room
temperature. The tensile failure strength of the composites was evaluated according to the ASTM tensile test
specification using specimens that were 25 mm wide and with a gauge length of 250 mm. Buckling tests
were performed at room temperature on slender composite beams that were 250 mm long and 20 mm wide.
The ends of the beams were simply supported. The Euler buckling strength was taken to be the applied
stress at which the specimen began to buckle. The flexural strength of the burnt composites was determined
at room temperature using the ASTM three-point bending procedure. The specimens were 250 mm long and
10 mm wide, and were loaded at a cross-head speed of 10 mm/min in 1/2 point loading until failure. The
flexural tests were performed with the burnt surface placed against the load point so it experienced bendinginduced compression. All tests were performed on glass/epoxy specimens. Table 5 shows some
characteristics of mechanical test specimens and their standard procedures.
Table 1. Physical properties of Epoxy resin and fiber glass
-1 -1
Specific heat at 287 T 363 cpr (Jkg K )
-1 -1
Specific heat at 363 T 393 cpr (Jkg K )
Specific heat at

393 T 1373


1050 + 0.8 ( T - T )
1050 + C pv + 0.8 ( T - T )

1050 + 0.8 ( T - T )


cpr (Jkg K )

Gas specific heat at 287 T 1373 cpg (Jkg K )



Thermal conductivity at

287 T 363 kr (Wm-1K-1)

Thermal conductivity at

363 T 393
393 T 473

Thermal conductivity at

2386.5 + 1.05 ( T T )
0.19 f ( m v0 ) 1.356e 04 ( T T )



0.19 f (mv ) 1.356e 04 ( T T )



0.19 1.356e 04 ( T T )

kr (Wm K )
kr (Wm K )

Thermal conductivity at 473 T 1373 kr (Wm K )



287 T 873 s

0.19 + 2e 04 ( T T )

Isobar specific heat at 287 T 1373 Cpfr (J/kg K )

0.755 + 2.5 10 4 ( T T )
760 + 3.88e- 02( T - T )

Thermal conductivity at 287 T 1373 kfr (W/m K)

1.09 2.05e 04 (T T )

Surface emissivity at

Table 2. Mechanical properties of laminate

Volume Fraction
Tensile Strength (MPa)
Transverse Tensile Strength (MPa)
Compressive Strength (MPa)
Transverse Compressive Strength (MPa)
Shear Strength (MPa)


Table 3. Mechanical properties of fiber and matrix

Shear Module
Poisson Ratio
Modulus (GPa)

Strength (MPa)

Fiber Glass
Type E





Epoxy Resin





Table 4. Characteristics of laminates




Fraction (vf)

[(0 / 90 / 0)]S


[(0 / 90) 4 ]S


Table 5. Characteristics of specimens and test standard procedures

Test Standard
Mechanical Test
Plate No.
Dimension ( mm mm )
ASTM D 3039
25 250
---20 250
10 250

Figure 1. Schematic diagram of two main regions of laminate after fire damage.

Figure 2. Different properties of two main regions of laminate after fire damage.

Figure 3. Vertical GRP laminate subject to fire from one side.

Figure 4. Geometry of laminate by layers and loading.

Figure 5. Specimens subject to heat flux

5. Results and Discussion

In Figure 6 the predicted temperatures by the thermal model are plotted as a function of time for the two
sides of the plate versus experimental data. Good agreement between the predicted temperatures and the
experimental data is shown. The maximum temperature is 375 oC for the front surface and 326 oC for the
back side at 25, kw/m2. The maximum temperatures at the final time for these two sides of specimens are
400 oC and 352 oC, respectively. Figures 7 and 8 show the temperatures of the two sides of the plate versus
the heating time for flexure and buckling specimens subjected to 25kw/m2 for a maximum of 840 seconds.
The temperature prediction at the initial time indicates a relatively poor agreement between resulsts of the
mechanical model and experimental data, but this agreement became better with time.

Figure 6. Temperature versus time for tensile test specimen at 25 kw/m2.

Figure 7. Temperature versus time for flexure specimen at 25 kw/m2.

Figure 8. Temperature versus time for buckle specimen at 25kw/m2.

Figure 9 shows the effect of the heat-exposure time on the thickness of the char in the laminate for thick
specimens (flexure specimens). When the plate was exposed to the higher heat flux, the char layer formed in
all the composite materials due to the thermal degradation of the polymer matrix. The char layer usually
started to grow as the material ignited and the burning started. It was observed that the thickness of the char
layer grew rapidly with the heating time till the composite reached the char time and then it became constant.
The formation of char in the epoxy composites occurred when the surface was heated to approximately 380
C, which is about the temperature at which the epoxy used in these composites began to undergo
significant thermal decomposition.
The post-fire strengths were calculated using the mechanical model which only requires the original
mechanical properties and strength of the un-burnt laminate and the number of damaged layers considering
the thickness of the char layer at a given heating time. It is shown in Figure 10 that the mechanical strengths
decrease due to the increase of the char thickness. The results indicate a good agreement between
theoretical prediction and experimental results. In most cases the agreement between the theoretical and
measured post-fire properties is within 10%. This reveals that the post-fire mechanical properties of
composites can be accurately determined using models that require the knowledge of measured parameters;
namely the original mechanical properties of the composite and the thickness of the char layer. In addition, it
indicates that the char is the dominant type of fire damage that degrades the mechanical properties. Despite
not considering these types of damage, the models were still accurate which proves that char is mostly
responsible for the reduction to the post-fire properties.

Figure 9. Effect of heat-exposure time on the thickness of the char in the thick laminate

Figure 10. Comparison mechanical model and

experiment in flexure failure load at 25 kw/m2

6. Conclusions
A theromechanical model for investing behavior of GFRP sheet under fire effect is presented and validated
using an with an experimental setup for a laminate composite material exposed to fire. A good agreement
between theoretical and experimental data was obtained. The model predicts more accurate results at longer
heating time. The simulations and experiments are preformed at three different loading conditions; tensile,
flexure and buckling specimens. It is shown that the tensile prediction is much closer to the experimental
data than others. In addition, it is indicated that the post-fire flexure properties of fiber-reinforced thermoset
composites decrease by being exposed to a fire. This reduction is mainly due to the thermal degradation and
combustion of the polymer matrix, which leads to the formation of a char region. The post-fire properties of
uniformly burnt composites can be determined using the proposed model. The model was validated using
three mechanical strength data for a glass/epoxy composite that had been exposed to fire at heat fluxes. The
thickness of the char layer increase rapidly with higher heat flux which caused the lowering of the mechanical
properties of the composites.
7. Nomenclature

m& g


pre-exponential factor
extensional stiffness matrix
bending-stretching coupling matrix
flexural stiffness matrix
laminate specific heat
gas specific heat
activation energy
laminate enthalpy
gas enthalpy
average free convection heat
transfer coefficient
equivalent temperature-dependent
nonlinear convection coefficient
laminate thermal conductivity
plate height
stress couples
gas mass flux




order of decomposition reaction

stress resultants
Nusselt number

Prandtl number
heat of decomposition
elasticity tensor components
net heat flux at the front surface
gas constant
Rayleigh number
shear strength
ambient temperature
spatial coordinate
fiber direction strength
transverse strength
out of plane direction
mid-surface strain
laminate density
Stefan-Boltzmann constant
stress tensor

8. References

Pering G.A., Farrell P.V., Springer G.S., 1981, Degradation of tensile and shear properties of composites
exposed to fire or high temperature, Environmental Effects on Composite Materials, Technomic
Publishing, Vol. 14, No. 1, pp. 145-159.
Gardiner C.P., Mathys Z., Mouritz A.P., 2004, Post-fire structural properties of burnt GRP plates, Marine
Structures, Vol. 17, No. 1, pp. 53-73.
Looyeh M.R.E., Bettess P., 1998, A finite element model for the fire-performance of GRP plates including
variable thermal properties, Finite Elements in Analysis and Design, Vol. 30, No. 4,pp. 313-324.
Dodds N., Gibson A.G., Dewhurst D., Davies J.M., 2000, Fire behaviour of composite laminates,
Composites: Part A, Vol. 31, No 7,pp. 689-702.
H. Bakaiyan, H. R. Daghyani, May 2005, Residual Strength Analysis of Composite Plates under Fire, The
13th annual (International) Conference of Mechanical Engineering. Isfahan, Iran, pp. 34-45
Callister W.D., Jr., 1994, Materials science and engineering an introduction, John Wiley & Sons, Inc. Third
Edition. New York
Gibson R.F., 1994, Principles of composite material mechanics, Mc Graw-Hill, New York