Journal of Advanced Concrete Technology Vol.

17, 419-433, July 2019 / Copyright © 2019 Japan Concrete Institute 419

Scientific paper

A Simplified Model to Predict Thermo-Hygral Behaviour and Explosive

Spalling of Concrete
Jin-Cheng Liu1* and Yao Zhang2

Received 6 March 2019, accepted 6 July 2019 doi:10.3151/jact.17.7.419

Abstract
Explosive spalling of concrete leads to premature failure of concrete structures under fire. Thus it is important to assess
explosive spalling risk of concrete. Thermo-hygral process in concrete is believed to be responsible for explosive
spalling of concrete. Currently, simple yet reliable models to predict thermo-hygral behaviour are still desirable. In this
study, a one-dimensional numerical model is established to predict moisture migration, pore pressure and explosive
spalling in concrete exposed to heating. The effectiveness of the model is validated against five sets of experimental
data. The experiments covered (a) moisture migration inside normal strength concrete (NSC); (b) pore pressure buildup
inside NSC; (c) pore pressure buildup inside high-performance concrete (HPC) containment vessel; (d) pore pressure
buildup inside HPC slab; and (e) one-time explosive spalling of concrete. The results predicted by the model match well
the experimental results. The model developed provides an effective and economical tool to assess explosive spalling of
concrete in addition to experimental methods.

1. Introduction binder, breakdown of calcareous aggregate, and

rehydration of lime. It often occurs at or after a material
Fire-induced spalling is a serious threat to concrete temperature above 700°C (Arioz 2009; Xing et al. 2011;
structures. It reduces member cross section, exposes Xiao et al. 2006). Among the three types of thermal
reinforcement to fire directly, consequently leading to spalling, thermo-hygral spalling, i.e., explosive spalling,
premature member failure (Shah and Sharma 2017; Han occurs much earlier in a fire than the two other types. As
et al. 2014; Umeki et al. 2016). This localized failure a result, explosive spalling poses a greater threat to fire
endangers the integrity of a structural system as a whole endurance of concrete members than the other two types
and may trigger progressive collapse in the worst-case of spalling (Sideris and Manita 2013). Therefore, devel-
scenario. Thus it is important to assess spalling risk of oping a model to predict explosive spalling of concrete
concrete and predict the time and location of concrete under fire is a priority.
spalling. A number of numerical models (Gawin et al. 2006;
Fire-induced concrete spalling can be classified into Dwaikat and Kodur 2009; Davie et al. 2012; Bary et al.
three categories based on the spalling mechanism: 2008; Zeiml et al. 2006; Khoury et al. 2002; Qian et al.
thermo-hygral spalling, thermo-mechanical spalling, 2005; Zhao et al. 2017) have been developed to predict
and thermo-chemical spalling (Liu et al. 2018). moisture transfer, pore pressure build up, and particu-
Thermo-hygral spalling, commonly referred to as explo- larly explosive spalling in concrete under fire. They
sive spalling, is caused by moisture clog and vapour provide a deep insight into complex physical and
pressure buildup in concrete (Monte et al. 2017; Heo et chemical processes behind explosive spalling phenome-
al. 2011; Ko et al. 2011; Raouffard and Nishiyama non from different perspectives. However, some of the
2015; Daungwilailuk et al. 2019). It often occurs at a models require complex material parameters, which are
material temperature below 320°C (Ju et al. 2017; Liu even more challenging to be measured experimentally.
and Tan 2018; Kanéma et al. 2011). Thermo-mechanical Besides, most of previous models were not fully vali-
spalling is induced by restrained thermal expansion in dated, i.e., they were only validated against pore pres-
cover concrete. It often occurs at a material temperature sure test results but not moisture distribution or explo-
within the range of 430°C to 660°C. Thermo-chemical sive spalling test results. Therefore, reliable and yet
spalling is due to loss of bond between aggregate and simple models that can predict moisture migration, pore
pressure build up and explosive spalling are still desir-
able.
1 This paper presents a simple one-dimensional model
Research Fellow, School of Civil and Environmental (1-D model) for predicting thermo-hygral behaviour and
Engineering, Nanyang Technological University, N1.1- explosive spalling of concrete. This 1-D model is devel-
B3-03, 50 Nanyang Avenue, 639798, Singapore. oped based on the work done by Ichikawa and England
*Corresponding author, E-mail: [email protected] (2004). The fundamental differences between the model
2
Research Fellow, School of Civil and Environmental proposed by them and the current one are summarized
Engineering, Nanyang Technological University, N1.1- below.
B3-03, 50 Nanyang Avenue, 639798, Singapore.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 420

In the model by Ichikawa and England (2004), liquid 2. Numerical modelling

water is compressible. However, in the proposed
spalling model, liquid water is assumed to be incom- 2.1 Discretization of model
pressible, which indicates the degree of saturation of A one-dimensional numerical model is established to
concrete should not exceed 100% (Gawin et al. 2006). predict moisture transfer and pore pressure build up
The permeability and tensile strength of concrete are across the concrete section under elevated temperature.
taken as a function of temperature in the proposed To achieve this, the concrete section is discretized into a
model rather than constant values in Ichikawa and finite number of nodes as shown in Fig. 2. Total depth
England (2004). The presented model incorporates the of the 1-D concrete section is assumed to be L. The
effect of silica fume on proportions of concrete compo- variable x represents the depth of a section from the fire-
nents and the effect of drying on moisture distribution. exposed side of concrete, with x=0 representing the fire-
Water released patterns for chemically-bound water in exposed face and x=L the unexposed face.
the two models are also different.
The developed model is validated against five sets of 2.2 Modelling of concrete components
experimental results: (a) moisture migration inside It is worth noting that concrete is a porous material con-
NSC; (b) pore pressure buildup inside NSC; (c) pore sisting of unhydrated cement, hydrated cement, aggre-
pressure buildup inside a HPC containment vessel; (d) gates, chemically-bound water, gel water, free water and
pore pressure buildup inside a HPC slab; and (e) one- empty pores. The corresponding masses and volumes of
time explosive spalling of concrete. From the five com- aforementioned components in a unit volume of con-
parison studies, it is shown that the proposed model can crete are determined using the formulas given in Ap-
predict with reasonable accuracy the moisture migration, pendix I. The model proposed by Ichikawa (2000) ap-
pore pressure development, the first occurrence of plies the method developed by Powers and Brownyard
spalling time and approximation of spalling location (1946) to calculate the above quantities of concrete
across the 1-D specimens. components. However, the formulas proposed by the
The developed 1-D model can be used in a wide above authors are mainly for cement-based systems. In
range of concrete structures, such as tunnels, cooling this study, the formulas are extended to incorporate ce-
towers, walls, slabs and containment vessels, as shown ment-silica fume blended systems based on the work of
in Fig. 1. It provides an alternative means to assess ex- the above authors and that of Atlassi (1995). The model
plosive spalling risk of concrete under fire in addition to can then be used for predictions of spalling behaviour of
experimental studies. The numerical method is cost- HPC.
effective and time-saving, obviating the need to conduct
sophisticated finite element modelling, which requires 2.3 Governing equations
much more resources and many more input parameters When concrete is under fire, heat transfer and moisture
that often cannot be measured experimentally. The pro- migration occur inside concrete simultaneously. The 1-
posed model can also be used to customize explosive- D governing heat transfer equation is given by,
spalling-free concrete mixes in the preliminary stage of
concrete mix design. Future efforts can be devoted to ∂T ∂ ⎛ ∂T ⎞
ρc = ⎜λ ⎟ (1)
develop a performance-based spalling risk assessment ∂t ∂x ⎝ ∂x ⎠
method to be incorporated in the framework of perform-
ance-based fire safety design of concrete structures. where ρ is concrete density, c is specific heat of concrete,
λ is thermal conductivity of concrete, T is temperature
of concrete and t is time from the start of heating.
Moisture transfer within the concrete section is domi-

Fig. 1 Applications of 1D thermo-hygral model. Fig. 2 Discretization of concrete section.

J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 421

nated by two flows, viz., diffusion-based flow governed determined considering the effect of environmental
by Fick’s law and pressure-induced flow governed by conditioning, which is influenced by curing age and
Darcy’s law. When concrete is under rapid heating, dif- environmental relative humidity (RH). The RHs at dif-
fusion-based flow is negligible. The change in moisture ferent times and distances from x=0 are calculated first,
content should be equal to the sum of the change in using the equation proposed by Parrott (1988). The ini-
moisture content due to pressure-induced flow and the tial relative moisture content distribution can then be
gain in moisture content due to dehydration of concrete. determined based on the relationship between RH and
Based on conservation of mass of water, the moisture moisture content proposed by Jiang and Yuan (2013).
transfer equation is given by Ichikawa (2000) as, Finally, together with the respective mass of free water
and gel water determined in Section 2.2, the initial dis-
∂m fw ∂ ⎛ ρ fw ∂p ⎞ ∂ tributions of free water and gel water can be determined.
= ⎜k ⎟ + ( mgw − r + mcw− r ) (2)
∂t ∂x ⎜⎝ η fw ∂x ⎟⎠ ∂t Distribution of chemically-bound water is assumed to
be uniform inside concrete. Pore pressure distribution is
where mfw, k, p, ρfw, ηfw, mgw-r and mcw-r represent mass uniform inside concrete and is set to 0 initially.
of free water, permeability of concrete, pressure, density,
dynamic viscosity of free water, mass of released gel 2.5 Boundary conditions
water and mass of released chemically-bound water, The boundary condition for heat transfer analysis is as
respectively. in Eq. (3) for the fire-exposed face, and for the unex-
When one face of concrete section is under heating, posed face, it is as in Eq. (4);
there are three zones forming across the depth as shown
in Fig. 3, namely, dry zone, wet zone and saturated zone. −k
∂T
∂x
( ) (
= h f T f − T x = 0 + σε 0 T f4 − T x4= 0 ) (3)
When the calculated moisture content at a node ex-
ceeds the saturated liquid water content, the excess
moisture content is driven to adjacent nodes towards the
unexposed side. When the calculated moisture content is
−k
∂T
∂x
( ) (
= he T x = L − Te + σε L T x4= L − Te4 ) (4)

negative at a node in the dry zone, the moisture content where hf and he are the respective heat transfer coeffi-
is redistributed linearly from the node to heat-exposed cients of the fire-exposed and unexposed faces; ε0 and εL
node based on conservation of mass of moisture content. are surface emissivities at x=0 and x=L, respectively;
Linear interpolation is used to determine the moisture T|x=0 and T|x=L are temperatures at x=0 and x=L, respec-
content and pore pressure at the location in between two tively; σ is Stefan-Boltzmann constant.
adjacent nodes. The boundary condition for moisture transfer analysis
is as in Eqs. (5) and (6) for the fire-exposed face, and
2.4 Initial conditions for the unexposed face, it is as in Eqs. (7) and (8);
A uniform temperature (room temperature) distribution
is taken as the initial condition of heat transfer analysis. m fw x = 0 = 0 (5)
The initial distribution of free water and gel water is
p x=0 = 0 (6)

p fw x = L = φe ρ g ,T V p x = L (7)

p x = L = φe psvp ,Ta (8)

where φe is the RH of the environment at the unexposed

face, ρg,T is the density of dry saturated water vapour at
the temperature of T, Vp|x=L is the pore volume at x=L,
and psvp,Ta is the saturated vapour pressure at ambient
temperature.

2.6 Permeability model

When concrete is under rapid heating, moisture inside
the section cannot escape within a short time. Perme-
ability of concrete is highly influenced by moisture con-
tent of concrete. The initial free water together with
released water tends to reduce the permeability of con-
crete. A decrease in permeability of concrete has been
observed between 175°C and 275°C (Klingsch 2014).
The surface cracking of concrete was observed to occur
Fig. 3 Thermo-hygral process in concrete under fire. at about 290°C (Lin et al. 1996). Cracks significantly
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 422

increase the permeability of concrete (Yildirim et al. plosive spalling. Furthermore, keeping increasing per-
2015). The critical point of water is about 374°C, at meability beyond 374°C tends to induce numerical in-
which the phase boundary between liquid water and stability and oscillation. Therefore, the normalized per-
water vapour terminates. The pressure of water in- meability model of concrete as shown in Fig. 4 is
creases with temperature. At 374°C, the pressure of wa- adopted for all 5 case studies. The permeability of con-
ter reaches its maximum limit, about 22 MPa. De- crete is constant from room temperature to 290°C for
velopment of water vapour pressure in concrete is an simplification. From 290°C to 374°C, the permeability
important source of concrete damage and permeability increases log-linearly to 1000 times of its initial value.
increase. Greathead (1986) showed that the permeability From 374°C onwards, the permeability remains constant.
of concrete drying at 400°C increased more than 1000 Although the permeability still increases beyond 374°C,
times. It is reasonable to assume that permeability in- further increasing permeability has little influence on
creased 1000 times at 374°C. Above 374°C, the the pore pressure dissipation. Hence, the assumption is
permeability is assumed to be constant. It is acceptable reasonable.
in this simplified spalling model, as the pore pressure in
concrete peaks at a lower temperature. Therefore, it will 2.7 Water-release pattern
not influence prediction of peak pore pressure and ex- There are three forms of water in concrete, i.e. free wa-
ter, gel water and chemically-bound water. The free wa-
ter is the water that can move freely inside concrete.
Chemically-bound water is the water consumed in the
formation of hydration products. Gel water is the water
Relative permeability

adsorbed onto the surfaces and in the interlayer spaces

of layered gel products during the hydration process.
Gel water and chemically-bound water are released
gradually into concrete pores and become free water as
temperature increases as illustrated in Fig. 5. The re-
lease of gel water follows the pattern specified by
Ichikawa (2000), i.e., linear pattern in association with
temperature from 20°C to 105°C. The release of gel
water is completed at 105°C. The release pattern of
chemically-bound water in association with temperature
Temperature (°C) is shown in Fig. 6 with different water-release patterns
Fig. 4 Normalized permeability model. proposed by different researchers (Bazant and Kaplan

Fig. 5 Illustration of water-release of concrete at elevated temperature.

Fig. 6 Chemically-bound water release pattern of concrete at elevated temperature.

J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 423

1996; Ichikawa 2000; Hilsdorf 1967) and experimental of temperature, and temperature-dependent tensile
data (Tan and Ng 2006). It can be seen that the proposed strength reduction factor in Eurocode 2 (CEN 2004) is
water release pattern matches well with the experimen- adopted.
tal data.
2.10 Flowchart
2.8 Pressure and dynamic viscosity The simplified flowchart of the thermo-hygral model is
The pore pressure in the wet zone and saturated zone is shown in Fig. 7. Firstly, the concrete section is repre-
calculated using the equation developed by Saul and sented by a finite number of nodes as shown in Fig. 2
Wagner (1987), while that in the dry zone follows the and time domain is discretized into a finite number of
equation by Saul and Wagner (1989). The dynamic vis- time steps. Given the fire curve and thermal properties
cosity of water is calculated using the equation devel- of concrete, the temperature profiles of concrete section
oped by Sengers (1986). at different time steps can be determined following Eq.
(1). Subsequently, moisture transfer analysis is con-
2.9 Explosive spalling criterion ducted. Based on the input of concrete mix design pa-
Explosive spalling of concrete under fire is deemed to rameters, concrete density, environmental RH and cur-
occur when the pore pressure-induced tensile stress ex- ing time of concrete, the initial distribution of moisture
ceeds the corresponding high-temperature tensile is determined. Given the temperature profile across the
strength. The pore pressure-induced tensile stress is de- section, the amount of released gel water and chemi-
termined using the following equation (Ichikawa and cally-bound water can be determined according to Sec-
England 2004), tion 2.7. Given the initial pore pressure in Section 2.5,
the moisture content is updated using Eq. (2). The pore
σ t = βt p (9) pressure is updated following Section 2.8. The above
steps are repeated for subsequent time steps until the
where βt is the scaling factor, taken as 1 in this model
end of heating. After obtaining the pore pressure pro-
(Zeiml et al. 2006).
files of concrete section at different time steps, the ten-
In this study, tensile strength of concrete is a function

Time and space domain

discretization

Heat transfer analysis Input thermal

Fire curve
∂T ∂ ⎛ ∂T ⎞ properties of
input ρc = ⎜λ ⎟ concrete
∂t ∂x ⎝ ∂x ⎠

Determined
Moisture released water
transfer amount
analysis
Update of Input concrete mix,
∂m fw ∂ ⎛ ρ fw ∂p ⎞ moisture concrete density
= ⎜k ⎟ content and RH
∂t ∂x ⎜⎝ η fw ∂x ⎟⎠
∂
+ ( mgw− r + mcw− r ) Update of
∂t pore
pressure

Tensile stress High

induced by temperature
pore pressure concrete tensile
t strength ft(T)

No No spalling
t>=ft(T)
occurs

Yes
Spalling
occurs
Fig. 7 Simplified flowchart of thermo-hygral model.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 424

Table 2 Model parameters for numerical analysis.

Time interval 1s Thermal conductivity 2 Wm-1K-1
No. of nodes 25 Specific heat 900 Jkg-1K-1
Initial permeability 8.0E-18 m-2 hf 0 Wm-2K-1
Concrete density 2460 kg/m3 he 0 Wm-2K-1
Concrete age at testing 365 days ε0 0.7
Environmental RH 97% εL 0.3

sile stress generated by pore pressure is compared with Table 1 Mix proportion of NSC by weight (Van der Heijden
the temperature-dependent tensile strength of concrete. et al. 2012).
For a given time step, if the tensile stress at a certain CEM I 32.5 R 1
location exceeds the tensile strength of concrete at that Sand (0.125–0.250 mm) 0.363
temperature, explosive spalling is assumed to occur. If, Sand (0.250–0.500 mm) 0.62
throughout the heating, the tensile stress always remains Sand (0.500–1 mm) 0.62
below the tensile strength, then there is no spalling. Sand (1–2 mm) 0.723
Sand (2–4 mm) 1.086
gravel (4–8 mm) 1.754
3. Validation water 0.5
Water/cement ratio (w/c) 0.5
3.1 Moisture migration inside NSC Aggregate/cement ratio (a/c) 5.166
Van der Heijden et al. (2012) developed a nuclear mag-
netic resonance (NMR) set-up to measure the moisture
The specimen was pressed in a PTFE holder so that the
transport in heated NSC. The schematic diagram of the
moisture transport was one-dimensional. Heat flow was
NMR set-up is shown in Fig. 8. One concrete cylinder
also confined to one dimension by insulating the speci-
specimen (φ80×100 mm) was prepared for the study.
men using mineral wool. One face of the cylindrical
sample was heated with a radiative heat flux of ap-
proximately 12 kW·m-2. The moisture profiles of the
sample were recorded every 8.5 minutes.
The mix proportion of the concrete used in the ex-
periment is given in Table 1. Based on the formulation
in Appendix I, the volumetric proportions of different
components of hydrated concrete are shown in Fig. 9.
The model parameters are summarized in Table 2.
Based on the concrete age at testing and environmental
RH, moisture distribution along the cylinder height is
determined (Fig. 10), which serves as the initial condi-
tion of moisture transfer analysis.
Figure 11 compares predicted and measured moisture
profiles in the concrete cylinder at different heating
Fig. 8 Schematic diagram of the NMR apparatus [after times (8.5 min. to 76.5 min. at intervals of 8.5 min.).
Van der Heijden et al. (2012)]. The predicted results capture important features of
moisture migration in the concrete specimen. Moisture
in the hot region migrates deep into the cold region be-
cause of pressure-induced flow. As a result, the moisture
content in the hot region decreases gradually. When the
moisture content becomes very small, a dry zone forms.
The dry zone increases as heating continues. In the mid-

Fig. 9 Volumetric proportions of components of the NSC Fig. 10 Moisture distribution along the height of the NSC
given in Table 1. cylinder.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 425

dle of the specimen, a saturated zone forms. The forma- Table 3 Mix proportion of NSC by weight (Kalifa et al.
tion of saturated zone is known as a 'moisture clogging' 2000).
phenomenon (Van der Heijden et al. 2012) where the Cement CPJ-CEM II 32.5 1
concrete pores are filled with water. The saturated zone Silico-calcareous sand (Seine 0/4 mm) 1.146
moves towards the unexposed face as heating continues. Calcareous sand (Boulonnais 0/5 mm) 1.146
The predicted results are in good agreement with the Calcareous aggregates (Boulonnais 5/12.5 mm) 1.469
measured results. Calcareous aggregates (Boulonnais 12.5/20 mm) 1.469
Water 0.5
3.2 Pressure buildup inside NSC Water/cement ratio (w/c) 0.5
Aggregate/cement ratio (a/c) 5.23
Kalifa et al. (2000) conducted a test to measure pore
pressure development inside a NSC slab at high tem-
perature. The experimental set-up is shown in Fig. 12.
As shown in the figure, the NSC slab was subjected to
600°C on one face and to 20°C on the opposite face.
The four lateral faces of the slab were insulated with
ceramic blocks. Pore pressure gages were installed at
five locations (10, 20, 30, 40 and 50 mm from heated
face) to record the pore pressure time history during
heating.
The mix proportion of the concrete used in the ex-
periment is given in Table 3. Based on the formulation
in Appendix I, the volumetric proportions of different
components of hydrated concrete are shown in Fig. 13.
The model input parameters are summarized in Table
4. Moisture distribution inside the NSC slab is deter-
mined as shown in Fig. 14. Fig. 12 The experimental apparatus [after Kalifa et al.
(2000)].

(a) 8.5 min. (b) 17 min. (c) 25.5 min.

(d) 34 min. (e) 42.5 min. (f) 51 min.

(g) 59.5 min. (h) 68 min. (i) 76.5 min.

Fig. 11 Moisture profiles of the concrete cylinder at different times.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 426

Figure 15 compares predicted and measured pore Table 4 Model parameters for numerical analysis.
pressure histories at locations 10, 20, 30, 40 and 50 mm Thermal
away from the heated face of the NSC slab. Both the Time interval 1s 2 Wm-1K-1
conductivity
predicted and measured pore pressures increase to a Specific
No. of nodes 26 900 Jkg-1K-1
peak value at first, and then decrease gradually to a neg- heat
ligible value. In the pre-peak stage, the predicted curves Initial
4.0E-18 m2 hf 25 Wm-2K-1
match well with measured curves at the five locations. permeability
At 10, 20 and 30 mm, the predicted peak pore pressures Concrete density 2400 kg/m3 he 25 Wm-2K-1
are slightly lower than the measured results. At 40 and Concrete age at
180 days ε0 0.25
testing
50 mm, the predicted peak pore pressures are slightly
Environmental RH 80% εL 0.39
higher than the test results. The time corresponding to
predicted peak pore pressures matches well with that
corresponding to measured peak pore pressures at the the measured pore pressure all the way. This difference
five locations. In the post-peak stage, the predicted peak is possibly because cracks occurred after 2 hours of
pore pressures at 10, 20, 30 and 40 mm are greater than heating. The cracks provide additional pathways that
the measured counterparts in the beginning, and became vapour can go through to the outside, thus leading to a
lower than the measured counterparts later; the pre- decrease in pore pressure. The agreement between pre-
dicted peak pore pressure at 50 mm were greater than dicted and measured results is qualitatively good in gen-
eral.

3.3 Pressure buildup inside HPC

(1) Reactor containment under accident condition
Shekarchi et al. (2003) conducted a test to study the
hygro-thermal behaviour of a HPC inner wall of a nu-
clear power station under a severe accident condition. A

Fig. 13 Volumetric proportions of components of the

NSC given in Table 3. Fig. 14 Moisture distribution along the NSC slab depth.

(a) 10 mm (b) 20 mm (c) 30 mm

(d) 40 mm (e) 50 mm
Fig. 15 Measured and predicted pore pressure histories at different depths of the NSC slab from fire-exposed face.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 427

Table 5 Mix proportion of HPC by weight (Shekarchi et

al. 2003). Table 6 Model parameters for numerical analysis.
CEM II 42.5 cement 1 Thermal
Time interval 60 s 2.3 Wm-1K-1
Calcareous sand 0/5 2.940 conductivity
Fine gravel 5/12.5 1.195 Specific
No. of nodes 35 900 Jkg-1K-1
Gravel 12.5/25 3.064 heat
Calcareous fillers 0.214 Initial
1.0E-18 m2 hf 60 Wm-2K-1
Silica fume 0.152 permeability
water 0.605 Concrete density 2450 kg/m3 he 0 Wm-2K-1
Superplasticizer 0.034 Concrete age at
200 days ε0 0.9
Retarding admixture 0.004 testing
Water/cement ratio (w/c) 0.605 Environmental RH 80% εL 0.0
Aggregate/cement ratio (a/c) 7.41 sisted of a linear temperature rise from 20°C to 200°C,
Silica fume/cement ratio (s/c) 0.152
and a steam pressure increasing to 13 bars in 24 h. Then
the temperature and steam pressure were maintained for
sketch of the experimental apparatus is shown in Fig. 16. another 24 h before cooling.
A cylindrical specimen with a thickness of 1.3 m (the The mix proportion of the HPC used in the experi-
same as in the nuclear containment) was used. The ment is given in Table 5. Based on the formulation in
specimen was vertically put into place and laterally Appendix I, the volumetric proportions of different
sealed by resin and wrapped with thermal insulation to components of hydrated concrete are shown in Fig. 17.
ensure 1-D moisture migration and heat transfer. One In order to simulate the experiment of Shekarchi et al.
face of the cylindrical specimen was subjected to the (2003), the cylindrical specimen is represented by a 1-D
severe accident condition. The accident condition con- model of length L = 1.3 m. The pressure at the hot face
follows the saturated vapour pressure curve. The model
parameters are summarized in Table 6. Moisture distri-
bution inside the HPC slab is determined as shown in
Fig. 18.
Figure 19 compares measured and predicted pressure
distributions in the cylindrical specimen at 24 h and 48
h. The pressure is greatest at the fire-exposed face and
decreases with increasing distance x from that face. At
the far end, the pressure is negligible. The predicted
results simulate the test results reasonably well.

Fig. 16 Sketch of experimental apparatus [after Shekar-

chi et al. (2003)].

Fig. 18 Moisture distribution along the height of the cy-

lindrical specimen.

Fig. 17 Volumetric proportions of components of the Fig. 19 Measured and predicted pressure profiles at 24 h
HPC given in Table 5. and 48 h.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 428

Table 7 Mix proportion of HPC by weight (Kalifa et al. Table 8 Model parameters for numerical analysis.
2001). Thermal
Time interval 1s 2.3 Wm-1K-1
Cement CPA CEM I 52.5 PM CP 1 conductivity
Specific
Silico-calcareous sand (Seine 0/4 mm) 1.058 No. of nodes 26 1000 Jkg-1K-1
heat
Calcareous sand (Boulonnais 0/5 mm) 1.041 Initial
Calcareous aggregates (Boulonnais 5/12.5 mm) 1.176 6.0E-19 m2 hf 25 Wm-2K-1
permeability
Calcareous aggregates (Boulonnais 12.5/20 mm) 1.352 Concrete density 2500 kg/m3 he 25 Wm-2K-1
Condensed silica fume 0.1 Concrete age at
90 days ε0 0.25
testing
water 0.3 Environmental RH 80% εL 0.3
Superplasticizer Chryso GT 0.033
Water/cement ratio (w/c) 0.3 peak values are reached, both predicted and measured
Aggregate/cement ratio (a/c) 5.23 pore pressures decrease with time. The predicted peak
Silica fume/cement ratio (s/c) 0.1 pore pressures match well with the measured peak pore
pressures at 20, 30 and 40 mm. At 50 mm, the measured
peak pore pressure is lower than the prediction possibly
(2) Concrete slab under high temperature due to occurrence of micro-cracks around the pressure
Kalifa et al. (2001) measured pore pressure develop- gage. Steel pressure gauge could be thermally incom-
ment inside HPC at high temperature. Two 120-mm patible with the surrounding concrete, and similarly for
thick HPC slabs were prepared for pore pressure meas- cement paste and aggregate. These incompatibility prob-
urements. The same experimental set-up (Fig. 12) as lems may induce micro-cracks in concrete. The pre-
described in section 3.2 was adopted. The pore pressure dicted time corresponding to peak pore pressures is also
was measured at five locations (10, 20, 30, 40 and 50 in good agreement with the time corresponding to
mm from heated face). measured peak pore pressures. With increasing distance
The mix proportion of the concrete used in the ex- from the fire-exposed face, the time corresponding to
periment is given in Table 7. Based on the formulation peak pore pressure increases. The numerical model in
in Appendix I, the volumetric proportions of different general captures well the phenomenon of pore pressure
components of hydrated concrete are shown in Fig. 20. development inside the HPC slabs under heating.
The model parameters are summarized in Table 8.
Moisture distribution inside the HPC slab is determined 3.4 Prediction of explosive spalling
as shown in Fig. 21. Ozawa et al. (2012) used acoustic emission method to
Figure 22 compares predicted and measured pore detect the time of explosive spalling of concrete. The
pressure histories at 10, 20, 30, 40 and 50 mm away heating test set-up is shown in Fig. 23. A concrete panel
from the heated face of the HPC slab. The peak pore with a thickness of 100 mm was subject to heating at the
pressure in the HPC slab is predicted to be higher than bottom face. The gas temperature at the heat-exposed
that in the NSC slab (Fig. 15), which agrees with the face was increased at a rate of 20°C/min. A pressure
consensus that HPC is more prone to explosive spalling gage was embedded in the panel at a depth of 8 mm
than NSC (Yermak et al. 2017; Lee et al. 2012; Kuroiwa from the heated face to measure internal pore pressure.
et al. 2007). Both the predicted and measured pore pres- The mix proportion of the concrete used in the ex-
sures increase at first at the five locations. When the periment is given in Table 9. Based on the formulation
in Appendix I, the volumetric proportions of different
components of hydrated concrete are shown in Fig. 24.
The model input parameters are summarized in Table
10. The tensile strength of concrete is estimated to be
1/15 of its compressive strength. Moisture distribution
inside the concrete panel is determined as shown in Fig. 25.

Fig. 20 Volumetric proportions of components of the

HPC given in Table 7. Fig. 21 Moisture distribution along the HPC slab depth.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 429

Table 9 Mix proportion of concrete by weight (Ozawa et Table 10 Model parameters for numerical analysis.
al. 2012). Thermal
Time interval 1s 2.2Wm-1K-1
Cement 1 conductivity
Fine aggregate 1.909 Specific
No. of nodes 25 1000 Jkg-1K-1
Coarse aggregate 2.409 heat
Water 0.3 Initial
1.0E-18m2 hf 25 Wm-2K-1
Admixture 0.05 permeability
Water/cement ratio (w/c) 0.3 Concrete density 2480 kg/m3 he 25 Wm-2K-1
Aggregate/cement ratio (a/c) 4.32 Concrete drying
118 days ε0 0.25
Compressive strength (MPa) 83.5 time
Environmental RH 40% εL 0.3
Tensile strength 5.6 MPa

(a) 10 mm (b) 20 mm (c) 30 mm

(d) 40 mm (e) 50 mm
Fig. 22 Measured and predicted pore pressure histories at different depths of the HPC slab from fire-exposed face.

Fig. 23 Heating test apparatus [after Ozawa et al. (2012)].

Fig. 25 Moisture distribution along the concrete panel

depth.

Fig. 24 Volumetric proportions of components of the

concrete given in Table 9.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 430

Figure 26 gives the predicted concrete panel thickness across the heated face. The minimum spalling depth was
as a function of heating time. The concrete panel thick- 0 mm and the maximum 12 mm. Fig. 28 shows the
ness reduces as concrete spalls. The concrete panel measured and predicted pore pressure histories at a
thickness after spalling is calculated by subtracting depth of 8 mm from the heated face. Both the measured
spalling depth from concrete panel thickness before and predicted pore pressures increase slowly in the first
spalling occurs. The predicted spalling time and spalling 35 minutes, and after that, they increased rapidly to the
depth of concrete panel are about 41 min. and 8.4 mm, maximum. At the time of explosive spalling, the meas-
respectively. The predicted pore pressure-induced ten- ured and the predicted pore pressures began to decrease.
sile stress across the panel depth at the spalling time is The prediction of explosive spalling in general agrees
plotted together with instantaneous tensile strength pro- well with the test results.
file of concrete in Fig. 27. In the heating test, explosive
spalling of concrete occurred at about 43 minutes. The 4. Discussion
spalling depths of the concrete panel were uneven
Though the simplified model predicts moisture move-
ment, pressure buildup and explosive spalling in con-
crete panels well qualitatively, there are some limita-
tions. These limitations are summarized below. Future
work can be done to improve these aspects.
(1) The model does not take into account the influence
of thermal-gradient induced stress, external load and
restraint. Thermal-gradient induced stress may in-
duce deformation incompatibility between cement
paste and aggregate due to shrinkage of cement
paste and dilation of aggregate. The deformation in-
compatibility will induce micro-cracks in concrete,
and increase permeability of concrete. If the micro-
cracks initiate before reaching critical explosive
spalling temperature, it tends to reduce the likeli-
Fig. 26 Comparison of predicted and measured time and
hood of explosive spalling. Similarly, if the external
location of explosive spalling.
load and restraint induces cracks in concrete mem-
bers before the critical temperature, the likelihood of
explosive spalling will also be reduced. In view of
these, the proposed model could be conservative.
(2) Explosive spalling of concrete can be one-off or
progressive. Progressive spalling of concrete in-
volves dynamic, instantaneous change in heat trans-
fer and moisture transfer boundary conditions. This
phenomenon is very complex and difficult to model
at this stage. Currently, the proposed model is not
able to consider progressive spalling.

5. Conclusion
A simple one-dimensional thermo-hygral model is pro-
Fig. 27 Pressure-induced tensile stress across the con-
posed in this study. The predictions of the thermo-hygral
crete panel thickness at spalling time.
model are compared with five different sets of
experimental data. The predictions are in relatively sat-
isfactory agreement with the measured results.
The proposed model predicts realistically the move-
ment of moisture inside the heated concrete. Both the
model and experiment show that moisture moves gradu-
ally from hot region to cold region as heating continues.
The phenomenon of 'moisture clog' is also captured by
the model.
The developed model predicts reasonably well the
pore pressure buildup in concrete. For rapid uniform
heating of one face of a concrete slab, both predicted
Fig. 28 Measured and predicted pore pressure histories and measured pore pressure initially increase to the peak
at a depth of 8 mm from heated face.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 431

value and then diminish gradually. The HPC is predicted through a multi-phase porous media model of
to have a higher peak pore pressure than NSC due to concrete.” Computer Methods in Applied Mechanics
lower permeability. This prediction is also in agreement and Engineering, 195(41), 5707-5729.
with widely-accepted opinion that HPC is more suscep- Greathead, R., (1986). “Permeability of concrete
tible to spalling than NSC. containing blast furnace slag as affected by
The thermo-hygral model is also capable of predict- temperature, moisture, and time.” Thesis (PhD).
ing explosive spalling of concrete under fire, i.e., when University of London.
and where explosive spalling would occur. This capabil- Han, L. H., Tan, Q. H. and Song, T. Y., (2014). “Fire
ity can be incorporated into fire resistance analysis of performance of steel reinforced concrete columns.”
concrete members to get more realistic modelling. Journal of Structural Engineering, 141(4), 04014128.
The numerical model provides a cost-effective and Heo, Y. S., Sanjayan, J. G., Han, C. G. and Han, M. C.,
time-saving method to assess explosive spalling risk (2011). “Construction application of fibre/mesh
under fire. It can also be used to tailor concrete mixes method for protecting concrete columns in fire.”
that are free from explosive spalling under fire. Current Construction and Building Materials, 25(6), 2928-
model is only applicable for prediction of the first-time 2938.
explosive spalling. But explosive spalling of concrete Hilsdorf, H., (1967). “A method to estimate the water
under fire typically occurs in a progressive manner. content of concrete shields.” Nuclear Engineering
There is a need to develop a model capable of predicting and Design, 6(3), 251-263.
progressive explosive spalling of concrete under fire. Ichikawa, Y., (2000). “Prediction of pore pressures, heat
and moisture transfer leading to spalling of concrete
Acknowledgement during fire.” Thesis (PhD). Imperial College London.
Discussions with Professor George L England and Pro- Ichikawa, Y. and England, G., (2004). “Prediction of
fessor Tan Kang Hai are greatly acknowledged. moisture migration and pore pressure build-up in
concrete at high temperatures.” Nuclear Engineering
References and Design, 228(1), 245-259.
Arioz, O., (2009). “Retained properties of concrete Jiang, J. and Yuan, Y., (2013). “Relationship of moisture
exposed to high temperatures: size effect.” Fire and content with temperature and relative humidity in
Materials, 33(5), 211-222. concrete.” Magazine of Concrete Research, 65(11),
Atlassi, E. H., (1995). “Nonevaporable water and degree 685-692.
of cement hydration in silica fume-cement systems.” Ju, Y., Tian, K., Liu, H., Reinhardt, H. W. and Wang, L.,
ACI Special Publication, 153, 703-718. (2017). “Experimental investigation of the effect of
Bary, B., Ranc, G., Durand, S. and Carpentier, O., silica fume on the thermal spalling of reactive powder
(2008). “A coupled thermo-hydro-mechanical- concrete.” Construction and Building Materials, 155,
damage model for concrete subjected to moderate 571-583.
temperatures.” International Journal of Heat and Kalifa, P., Chene, G. and Galle, C., (2001). “High-
Mass Transfer, 51(11), 2847-2862. temperature behaviour of HPC with polypropylene
Bazant, Z. P. and Kaplan, M. F., (1996). “Concrete at fibres: From spalling to microstructure.” Cement and
high temperatures: material properties and Concrete Research, 31(10), 1487-1499.
mathematical models.” Harlow, UK: Longman. Kalifa, P., Menneteau, F. D. and Quenard, D., (2000).
CEN, (2004). “Eurocode 2: Design of concrete “Spalling and pore pressure in HPC at high
structures - parts 1-2: general rules-structural fire temperatures.” Cement and Concrete Research,
design.” Brussels: Comitée Européen De 30(12), 1915-1927.
Normalisation. Kanéma, M., Pliya, P., Noumowé, A. and Gallias, J.,
Daungwilailuk, T., Kitagawa, T., Bui, P. T., Ogawa, Y. (2011). “Spalling, thermal, and hydrous behavior of
and Kawai, K., (2019). “Penetration of moisture, CO2, ordinary and high-strength concrete subjected to
and Cl ions in concrete after exposure to high elevated temperature.” Journal of Materials in Civil
temperature.” Journal of Advanced Concrete Engineering, 23(7), 921-930.
Technology, 17(1), 1-15. Khoury, G., Majorana, C., Pesavento, F. and Schrefler,
Davie, C., Pearce, C. and Bićanić, N., (2012). “Aspects B., (2002). “Modelling of heated concrete.” Magazine
of permeability in modelling of concrete exposed to of Concrete Research, 54(2), 77-101.
high temperatures.” Transport in Porous Media, 95(3), Klingsch, E., (2014). “Explosive spalling of concrete in
627-646. fire.” Thesis (PhD). ETH-Zürich.
Dwaikat, M. B. and Kodur, V., (2009). “Hydrothermal Ko, J., Ryu, D. and Noguchi, T., (2011). “The spalling
model for predicting fire-induced spalling in concrete mechanism of high-strength concrete under fire.”
structural systems.” Fire Safety Journal, 44(3), 425- Magazine of Concrete Research, 63(5), 357-370.
434. Kuroiwa, S., Inoue, Y., Fujioka, K. and William, A.,
Gawin, D., Pesavento, F. and Schrefler, B., (2006). (2007). “Performance confirmation tests on C100
“Towards prediction of the thermal spalling risk concrete in Dubai, UAE.” Journal of Advanced
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 432

Concrete Technology, 5(2), 171-180. conductivity of water substance.” Journal of Physical

Lee, J. H., Sohn, Y. S. and Lee, S. H., (2012). “Fire and Chemical Reference Data, 15(4).
resistance of hybrid fibre-reinforced, ultra-high- Shah, A. H. and Sharma, U., (2017). “Fire resistance
strength concrete columns with compressive strength and spalling performance of confined concrete
from 120 to 200 MPa.” Magazine of Concrete columns.” Construction and Building Materials, 156,
Research, 64(6), 539-550. 161-174.
Lin, W. M., Lin, T. and Powers-Couche, L., (1996). Shekarchi, M., Debicki, G., Billard, Y. and Coudert, L.,
“Microstructures of fire-damaged concrete.” ACI (2003). “Heat and mass transfer of high performance
Materials Journal, 93(3), 199-205. concrete for reactor containment under severe
Liu, J. C. and Tan, K. H., (2018). “Fire resistance of accident conditions.” Fire Technology, 39(1), 63-71.
ultra-high performance strain hardening cementitious Sideris, K. and Manita, P., (2013). “Residual mechanical
composite: residual mechanical properties and characteristics and spalling resistance of fiber
spalling resistance.” Cement and Concrete reinforced self-compacting concretes exposed to
Composites, 89, 62-75. elevated temperatures.” Construction and Building
Liu, J. C., Tan, K. H. and Yao, Y., (2018). “A new Materials, 41, 296-302.
perspective on nature of fire-induced spalling in Tan, T. and Ng, L., (2006). “Predicting non-evaporable
concrete.” Construction and Building Materials, 184, water loss of cement paste at elevated temperature.”
581-590. Advances in Cement Research, 18(2), 83-89.
Mills, R., (1966). “Factors influencing cessation of Umeki, Y., Sawada, S., Mitsugi, S., Maenaka, T. and
hydration in water cured cement pastes.” Highway Takiguchi, K., (2016). “Outline of guidelines for
Research Board Special Report 90. maintenance and management of structures in nuclear
Monte, F. L., Lombardi, F., Felicetti, R. and Lualdi, M., facilities.” Journal of Advanced Concrete Technology,
(2017). “Ground-penetrating radar monitoring of 14(10), 643-663.
concrete at high temperature.” Construction and Van Der Heijden, G., Pel, L. and Adan, O., (2012). “Fire
Building Materials, 151, 881-888. spalling of concrete, as studied by NMR.” Cement
Ozawa, M., Uchida, S., Kamada, T. and Morimoto, H., and Concrete Research, 42(2), 265-271.
(2012). “Study of mechanisms of explosive spalling Xiao, J., Xie, M. and Zhang, C., (2006). “Residual
in high-strength concrete at high temperatures using compressive behaviour of pre-heated high-
acoustic emission.” Construction and Building performance concrete with blast furnace slag.” Fire
Materials, 37, 621-628. Safety Journal, 41(2), 91-98.
Parrott, L., (1988). “Moisture profiles in drying Xing, Z., Beaucour, A. L., Hebert, R., Noumowe, A. and
concrete.” Advances in Cement Research, 1(3), 164- Ledesert, B., (2011). “Influence of the nature of
170. aggregates on the behaviour of concrete subjected to
Powers, T. C. and Brownyard, T. L., (1946). “Studies of elevated temperature.” Cement and Concrete
the physical properties of hardened portland cement Research, 41(4), 392-402.
paste.” ACI Journal, 43(9), 249-336. Yermak, N., Pliya, P., Beaucour, A. L., Simon, A. and
Qian, C., Wang, H., Sun, W., Guo, Z. and Stroeven, P., Noumowé, A., (2017). “Influence of steel and/or
(2005). “Numerical calculations of vapour pressure in polypropylene fibres on the behaviour of concrete at
concrete exposed to fire.” Magazine of Concrete high temperature: spalling, transfer and mechanical
Research, 57(3), 179-184. properties.” Construction and Building Materials,
Raouffard, M. M. and Nishiyama, M., (2015). “Fire 132, 240-250.
resistance of reinforced concrete frames subjected to Yildirim, G., Sahmaran, M., Balcikanli, M., Ozbay, E.
service load, part 1: experimental study.” Journal of and Lachemi, M., (2015). “Influence of cracking and
Advanced Concrete Technology, 13(12), 554-563. healing on the gas permeability of cementitious
Saul, A. and Wagner, W., (1987). “International composites.” Construction and Building Materials,
equations for the saturation properties of ordinary 85, 217-226.
water substance.” Journal of Physical and Chemical Zeiml, M., Leithner, D., Lackner, R. and Mang, H. A.,
Reference Data, 16(4), 893-901. (2006). “How do polypropylene fibers improve the
Saul, A. and Wagner, W., (1989). “A fundamental spalling behavior of in-situ concrete?” Cement and
equation for water covering the range from the Concrete research, 36(5), 929-942.
melting line to 1273 K at pressures up to 25 000 Zhao, J., Zheng, J. J., Peng, G. F. and Sun, P. S., (2017).
MPa.” Journal of Physical and Chemical Reference “Spalling and cracking modelling of high-
Data, 18(4), 1537-1564. performance concrete exposed to elevated tempera-
Sengers, J., (1986). “Improved international tures.” Magazine of Concrete Research, 69(24), 1276-
formulations for the viscosity and thermal 1287.
J.-C. Liu and Y. Zhang / Journal of Advanced Concrete Technology Vol. 17, 419-433 2019 433

AppendixⅠ Formulations to determine αu: degree of hydration of cement. If no measured

the quantities of concrete components degree of hydration is given, the equation (A.14)
developed by Mills (1966) is used to calculate the
If w/c ≥ 0.44αu, value.
αp: degree of pozzolanic reaction of silica fume. If no
mcw = 0.23α u mcem (A.1) measured degree of pozzolanic reaction is given,
the value is assumed to be equal to αu.
mgw = 0.93α u mcw + 0.34α pη s ( s / c)mcem (A.2) mfw: mass of free water per unit volume of concrete,
kg/m3.
m fw = ( w / c − 0.44α u )mcem − mgw (A.3) mgw: mass of gel water per unit volume of concrete,
kg/m3.
If 0.23αu ≤ w/c< 0.44αu,
mcw: mass of chemically-bound water per unit volume
mcw = 0.23α u mcem (A.4) of concrete, kg/m3.
mcem: mass of cement per unit volume of concrete,
mgw = ( w / c − 0.23α h )mcem (A.5) kg/m3.
ηs: percentage of pure amorphous silica in silica fume
m fw = 0 (A.6) = 90%, typically.
Vfw: volume of free water per unit volume of concrete,
If w/c < 0.23αu, m3/m3.
Vgw: volume of gel water per unit volume of concrete,
mcw = ( w / c)mcem (A.7) m3/m3.
Vcw: volume of chemically-bound water per unit vol-
mgw = 0 (A.8) ume of concrete, m3/m3.
Vp: pore volume per unit volume of concrete, m3/m3.
m fw = 0 (A.9) ρcem: density of cement = 3150 kg/m3.
ρw: density of water = 1000 kg/m3.
Vcw = mcw / ρ cw (A.10) ρgw: density of gel water = 1110 kg/m3.
ρcw: density of chemically-bound water = 1220 kg/m3.
Vgw = mgw / ρ gw (A.11)
ρa: density of aggregate, = 2600 kg/m3 if not speci-
fied.
V fw = m fw / ρ w (A.12)
s/c: silica fume to cement mass ratio.
a/c: aggregate to cement mass ratio.
V p = 1 − (a / c / ρ a + 1/ ρcem + Vcw + Vgw ) (A.13) w/c: water to cement mass ratio.

1.031w / c
αu = (A.14)
0.194 + w / c