You are on page 1of 18

International Journal of Thermal Sciences 185 (2023) 108063

Contents lists available at ScienceDirect

International Journal of Thermal Sciences


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

Combined transient 3D simulation and experimental methods to assess a


slow heat-up curve used to dry a refractory concrete
J. Juárez Trujillo a, b, *, J.A. Castro b, M.D.M. Innocentini c, F. Vernilli a
a
Universidade de São Paulo, EEL/DEMAR, Lorena, 12600-970, Brazil
b
Universidade Federal Fluminense, EEIMVR/PPGEM, Volta Redonda, 27255-125, Brazil
c
Universidade de Ribeirão Preto, CCET, Ribeirânia, 14096900, Brazil

A R T I C L E I N F O A B S T R A C T

Keywords: Explosive spalling is a major concern when drying refractory concretes. For safety, empirically defined over­
Refractory drying estimated slow heat-up curves are used in industry. Polypropylene fibres reduce spalling; however, they are
Heat-up curve inconvenient for certain castables due to properties deterioration, such as slag penetration. There is neither a
Moisture migration
generally accepted methodology to define monolithic refractories dry-out schedules nor agreement about a
Pore pressure
Permeability
mathematical model to predict spalling phenomena. This paper presents original numerical and experimental
results of castable drying investigation using slow heating. Modifications in experimental methods were made to
reproduce refractory lining boundary conditions. A conservative fully implicit finite volume method in three-
dimensions was developed to predict pressure and moisture migration. The influence of the methods used to
characterize the properties needed for the mathematical model was discussed. The combined analyses suggest
that the first holding time used was not productive for drying; however, its effect on curing must be reviewed.

1. Introduction be fast but cautiously due to the risk of cracks that would become hot
spots, source of leaks and reduced life span, or explosions with over­
In industrial installations, large refractory monoliths are used to whelming losses. Consequently, experimental, and computational re­
produce and convey hot fluids while protecting an external steel shell sources are needed to predict concrete’s hydro thermomechanical
that holds them. These complex, large scale structures must be moulded, behaviour and define efficient dry-out schedules.
cured, and dried within an ample range of temperatures, usually above Research on HPC has the objective of preventing structural failure in
550 ◦ C [1]. Drying out new refractory linings is a time-consuming pro­ case of accidental fire. Hence, the investigations usually apply fast heat-
cess that uses heat to remove the mass of free and chemically combined up curves, like ISO-834 [10,11]. However, the cementitious systems
water remained in the concrete after moulding and curing. It is intended used in industrial applications are exposed to different thermal condi­
to optimize the drying operation to reduce downtime, energy con­ tions; therefore, they require to be studied from additional approaches.
sumption and pollution. The problem is that concrete heating presents Refractory linings are normally dried with progressive heat by one
the risk of cracks or explosive spalling. Therefore, the thermal stresses surface, having the opposite side in contact with an impermeable steel
must be controlled while providing time for the water vapour to escape shell exposed to environmental conditions. Recent works have contrib­
without raising the internal pressures above the tensile strength of re­ uted with different experimental setups and improvements in the in­
fractory concrete [2]. In heated high-performance concrete (HPC), the struments used to measure pore pressures [2,4,10–19]; however, there is
addition of polypropylene fibres has shown to be efficient to mitigate neither standard methods nor commercial devices nor gauges to mea­
spalling [3,4]; nonetheless, in the case of some refractory concretes, fi­ sure vapour pore pressure and evaluate the hydro thermomechanical
bres have shown to be detrimental for important properties like corro­ behaviour in heated concrete [20]. Sophisticated experiments have been
sion resistance, mechanical strength, flowability and slag penetration proposed by means of nuclear magnetic resonance, neutron radiography
[5–7]. Currently, the heat-up curves are defined by refractory manu­ and tomography for detailed visualization of the moisture migration and
facturers from empirical experience [2,8,9]. Any change is subject to its obstruction, representing valuable resources to understand the in­
hesitation as it involves safety, quality, and costs because drying should ternal mechanisms during heating [21–24]. The image experiments

* Corresponding author. UFF/EEIMVR, Av. dos Trabalhadores 420, 27255-125, V. R., RJ, Brazil.
E-mail address: jorgejuareztrujillo@id.uff.br (J. Juárez Trujillo).

https://doi.org/10.1016/j.ijthermalsci.2022.108063
Received 9 March 2022; Received in revised form 12 October 2022; Accepted 24 November 2022
1290-0729/© 2022 Elsevier Masson SAS. All rights reserved.
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

have been useful to shown how gauges inside concrete influence the reactions of CAC to propose holding times at 150◦ , 300◦ and 565 ◦ C with
measurements [25]; although limitations arise in the image analyses due the use of internal pressure measurements to define the duration of each
to the reduced size of the sample, and the high cost. Other experimental one [40].
arrangements have presented reproduction difficulties, even to replicate In this research, a phenomenological model was implemented using
the explosive condition that has been considered a complex and not the FVM in 3D to analyse transient moisture migration during the drying
well-interpreted phenomenon [15,26,27]. Therefore, solutions to these of a conventional, hydraulic bonding high-alumina castable. As simu­
problems are needed, as internal pressure is pointed out to explain the lated results of temperature, pressure, and mass loss must be validated,
explosive spalling and quantitative data is relevant to validate numerical an experimental setup was developed to dry a castable slab reproducing
models. To date, there is no consensus for the industrial use of neither the specific conditions of refractory lining, heating by one face and with
experimental methodologies nor mathematical models. But the most the opposite face joined to an impermeable surface. The objective is to
important is that few works in the literature provide details of the explore strategies to identify the most convenient practices that lead to
problems and errors to prevent their repetition [2,16]. the prediction of the drying behaviour of complex geometries to eval­
The mathematical models proposed in the literature are complex. uate heat-up curves. As the phenomenon is considerably intricate, and as
Strongly non-linear parameters are involved, such as desorption iso­ previous investigations have shown the distinguishable influence of
therms, permeability, and thermal conductivity. Those parameters are bound water on the drying rate curve [41], the proposed idea is to
function of several variables and require specific methods to charac­ perform the analyses by steps, isolating the drying stages [42] to
terize the properties involved. The models are also controversial, either distinguish the effect of heating on the different types of water in a
considering one or several phases with distinction between liquid water, prismatic specimen. For that, this research was focused on the complete
vapour, and air, balancing those species separately or together, with removal of only the free water. Therefore, we evaluated transient and
arguments over the capillary pressure effect and pores supersaturation stationary states of a slow heat-up curve used in some industries to dry a
[2,28–31]. Moreira et al. performed a direct comparison of different specific refractory castable with a holding time at 110 ◦ C and a
theoretical approaches, concluding that the single-phase model is a maximum temperature of 300 ◦ C.
convenient election to model refractory drying in practical applications
[32]. The election of the most convenient numerical method for the 2. Materials and experimental methods
solution is also a matter of discussion. The finite element method (FEM)
has been traditionally adopted to investigate the behaviour of structural 2.1. Experimental setup
concrete under thermal stress [8,28,33,34]. Dal Pont et al. pioneered the
use of the finite volume method (FVM) for this kind of problem A device was developed to dry calcium aluminate concrete slabs 300
obtaining a qualitative description of the complex phenomena [18]. × 300 × 120 mm through controlled upper face radiating heat, based on
Meftah et al. proposed a FVM-FEM comparison with two algorithms Kalifa’s et al. experimental setup [10,11]. Modifications were made to
obtaining similar numerical results, with one algorithm slightly faster reproduce the boundary conditions of a refractory lining by coupling the
[35]. In contrast, Fey et al. used the finite difference method (FDM) in an bottom face with a steel shell 4 mm thick which also served as a support
extensive investigation of refractory concrete drying for optimized to hold measuring probes in vertical position, Fig. 1 (a). As previous tests
pressure-driven heat-up curves [2,12,36]. The increased effort needed to made with smaller samples confirmed, in agreement with Jansson et al.
solve models with added dimensions is also a matter of discussion. To [16], that the size of the specimen influences the results; thus, the
investigate the drying behaviour of a furnace’s two walls edge, Fey et al. specimen height of this setup was selected to match the thickness of a
evolved from a 1D to a 2D model [2,36]. Yet, for more complex geom­ blast furnace blow pipe refractory lining.
etries, such as three walls’ corners, a 3D model would be more suitable. To reduce the loose of thermal energy by the lateral faces, 100 mm
Meftah et al. [37] emphasized the need for 3D modelling to treat con­ thick alumina silicate insulating refractory bricks were used, with high
crete heterogeneity, obtaining a proper description of the spatial fluc­ porosity (50–60%), low density (800 kg/m3) and low thermal conduc­
tuations in the solution fields. Dauti et al. [27] developed 3D simulations tivity (0.29 W/mK). Additional 150 mm fiberglass insulation above the
and compared with neutron tomography images, showing how useful 5000 W radiating elements and by the sides of the furnace proved to be
those tools are to better understand the behaviour of HPC under thermal important to achieve the temperatures and heating rates required. Two
stress. Nevertheless, in the case of castables drying, a model and its upper chimneys for natural air and vapour circulation were made with
numerical solution will be satisfactory only when it is finally able to cuts in some bricks of the lateral insulation, so that the heating chamber
effectively define the time in which a certain geometry of a specific type can be sealed or opened for exchange with the environment; likewise,
of concrete has lost its free and bound water and predict the time and any condensation from the specimen may come out freely, Fig. 1 (b). A
spatial position in which a crack or explosive spalling would occur or proportional integral derivative controller was fitted to program heat-up
not, as a result of a peculiar heat-up curve. Important insights have been curves for drying experiments, it was connected to an alumina-coated K-
obtained, but more investigation is still needed. type thermocouple placed between the radiating elements and the
Moreover, HPC and refractory concretes have different binders and specimen.
hydration products that influence their final properties and spalling
behaviour, besides their differences in particle size distribution. In the 2.2. Pressure – temperature probes
case of HPC the most important binder is Portland cement, whose main
hydration phases affecting the hydro thermomechanical behaviour are Six simultaneous temperature-pressure probes were manufactured
cement calcium silicate hydrate (CSH) and calcium hydroxide (CH). In using a seamless A269 stainless steel tube (outer diameter 9.53 mm,
the case of refractory castables, the main binder is calcium aluminate inner diameter 6.53 mm). To prevent the entrance of concrete into the
cement (CAC) whose hydration reactions produce CAH10, C2AHx, C3AH6 probes during casting, sintered steel tablet filters of ballistic shape with
and variated amounts of AH3 depending on the Ca/Al ratio of clinker. diameter 7.2 mm, height 4.0 mm, and 50 μm pore size were inserted at
The effect of temperature during hydration of CAC has been broadly one end of the tubes. At the other end, a “T” connection was linked to a
studied, reporting that different Ca–Al phases can be formed, depending high sensitivity piezoresistive pressure transducer MPX5999D (capacity
on temperature during hydration [38,39]. Notwithstanding, the effect of from 0 to 1.0 MPa) and to a K-type thermocouple inserted into the tube
phase composition after hydration and during the first heat-up has in contact with the filter to obtain temperature and pressure measure­
limited investigation. It may be expected that the different phases pro­ ments in the same coordinates, Fig. 2 (a) and (b). As the thermocouples
duced after hydration can result in different behaviour during heating. have a diameter of 3 mm, once inserted into the tubes they left an air
For castables dying, Hips and Brown considered the dehydration volume gap of 10 ml. To reduce air negative effect on vapour pressure

2
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

Fig. 1. (a) Device scheme showing internal components, depths of pressure-


temperature measuring points and the bottom steel shell. (b) Furnace with
upper radiating heat, controlling thermocouple, insulation, and chimney.

measurement, a paraffinic oil with boiling point above 200 ◦ C was used
to serve as a pressure transfer medium to the transducers; it was left a
free volume for oil expansion, as proposed by Li et al. [17], and probes
were hold in vertical position to prevent oil drain off. The calibration of
instruments was confirmed before and after experiments. The detach­
able mould used to prepare the specimen was a tight assembly of
stainless-steel plates 4 mm thick with exact measurements to obtain the
prismatic specimen size. The plates were covered with mineral grease to
facilitate unmoulding after curing. To obtain unilateral thermal flow in
the probes, they were mounted at the bottom of the mould within a 100
mm2 central area of the steel plate with depths of 20, 30, 40, 50, 70 and
100 mm in relation to heated surface, Fig. 2 (c).
One of the main problems detected for pressure measuring was the
vapour leakage along the interface between the concrete and the steel of
the probes because the lack of adhesion which result in surrounding
cracks or air bubbles, as observed in Ref. [25]; the following modifica­
Fig. 2. Probes for simultaneous pressure-temperature measurement. (a) Inter­
tions were performed: nal arrangement. (b) Probes with knurled surface and fastening O-rings; (c)
Mould assembled and sealed with probes located in a central squared area. (d)
1. The tube surface was knurled to improve the adhesion between steel Concrete moulding covered with PVC for hydration. (e) Specimen unmoulded
and concrete Fig. 2 (b); with sensors and wires connections underneath.
2. Steel O-rings were welded to the probes to prevent movements that
break the adhesion. A twisted one was located 20 mm below the top 2.3. Specimen preparation
of the tube to prevent rotational movements and a flat one at the
bottom, upheld by the steel plate, to prevent vertical movements Table 1 shows the composition, aggregate size, and water/binder
Fig. 2 (a, b); ratios of the high-alumina castable. For the specimen preparation, the
3. All joints were sealed with high-temperature resistant silicone Fig. 2 mass of anhydrous castable was 24.2 kg and the water added to the mix
(c). was 2.8 kg (10.3 wt %). It was not possible to vibrate the specimen due
to false contact problems that vibration was causing in the electronic
sensors. The specimen was weighted before and after curing and heating

3
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

Table 1 released by dehydration; J is the mass flux (kg/m2s) proportional to the


High alumina castable composition. vapour pressure gradient ∇P (Pa/m); g is the gravitational acceleration; t
Compound wt % is time, and a = a(h,T) is the concrete relative permeability to humidity,
Al2O3 73.0 dependent on temperature and the water properties as liquid or vapour.
SiO2 13.0 See appendix A and B for nomenclature, values, units, and parametric
Fe2O3 1.3
equations.
CaO 9.3
Binder wt % Equation (3) is the energy balance, which considers the main en­
CAC (60% Al2O3; 40% CaO) 23 thalpies of the system. Equation (4) is the Fourier’s law for heat con­
Aggregate maximum size mm duction generated by the temperature gradient. The heat convection
Mullite 10.0 results from the mass of vapour transported through the porous medium.
Ratios
Water/Solid 11.6 ∂T ∂W
Water/Binder 50.3 ρC − Ca = − ∇ • q + Cw J • ∇T (3)
∂t ∂t

q = − λ∇T (4)
to register the total mass loss, Fig. 2 (e).
where ρ is concrete solid matrix density, C is its isobaric heat capacity,
2.3.1. Concrete curing Ca = Ca(T) is the free water latent heat of vaporization, Cw is water
The specimen was covered with polyvinyl chloride (PVC) film during isobaric heat capacity, q is the heat flux (J/m2s) proportional to the
48 h for hydration at room conditions, 29 ◦ C and 56% RH; see Fig. 2 (d). temperature gradient ∇T (K/m), λ is the concrete thermal conductivity;
Afterwards, the cover and the mould lateral plates were removed, see appendix A and B. This simplified model considers concrete as a
leaving the concrete in contact with the bottom steel shell that was also porous medium partially filled with humidity, without considering the
holding the probes. More samples were prepared with the same condi­ air mass neither the latent heat of dehydration.
tions to characterize physical properties of recently cured concrete, i.e., By coupling the mass and energy balance equations, a system of two
in “green” condition. parabolic PDE with four state variables (Wd, W, T, P) is obtained. Wd, is
calculated at each node and iteration according to equation B.10 in
2.4. Heat-up curve (HUC) appendix B. As W is a thermodynamic function of temperature T and
pressure P, its variation over time, ∂W/∂t, can be substituted in both
The aim of the experiment was to analyse the drying behaviour to PDEs using its variation over pressure, ∂W/∂P, and its variation over
remove only the free water. Therefore, it was used a slow heat-up curve temperature, ∂W/∂T, whose values can be calculated by central differ­
currently used in industry to avoid cracks and risks of explosive spalling. ences approximation. Therefore, two unknown state variables, temper­
It was programmed a first heating rate of 20 ◦ C/h to reach 110 ◦ C at the ature T and pressure P, remain to be calculated with two PDEs. Details
specimen’s upper face with a holding time of 10 h. Thereafter, a heating about the method for conditioning equations (1) and (3) for the numeric
rate of 30 ◦ C/h to reach the final temperature of 300 ◦ C at the specimen’s solution, according to Refs. [8,28,43,44,46], can be found in Appendix
upper face. The final temperature was held until reaching thermal and C.
hydraulic stationary state. Thereafter, the specimen was kept inside the
furnace for natural cooling. The thick red line in Fig. 5 shows the heat-up
curve (HUC). 3.2. Discretization method

3. Mathematical model and numerical methods The domain was divided in non-overlapping regular control volumes
with a representative node at the centre, as shown in Fig. 3. The in­
The mathematical model is based on a system of two coupled para­ terfaces of the control volumes were placed midway between adjacent
bolic partial differential equations (PDE) proposed by Bazant [28], re­ nodes in a cartesian 3D stencil forming a rectangular grid point cluster
ported as a practical option for technological applications [43,44]. It
considers the flow of energy and a simplified one-fluid flow of mass, i.e.,
moisture (liquid water and vapour) in a porous medium. The perme­
ability is a function of humidity and temperature for the microstructural
evolution that follows the dehydration reactions. A fully implicit FVM
was implemented to convert the PDEs into systems of linear equations
solved by iterative procedures [45]. The solution were the scalar fields
of temperature T, pore vapour pressure P, and moisture content W over
time in a 3D domain standing for the concrete specimen.

3.1. Conservation equations

Equation (1) is the mass balance, which includes a mass source term
for the release of chemically bound water, triggered by heat during
dehydration. Equation (2) is the Darcy’s law for the transport of mois­
ture generated by the pressure gradient inside the concrete pores.
∂W ∂Wd
= − ∇•J+ (1)
∂t ∂t
a
J= − ∇P (2)
g
Fig. 3. Control volume rectangular stencil for coordinates (x, y, z), showing the
where W = W(h,T) is the free water content dependent on the pores’ node P, with consecutive nodes W, E, T, B, N, S and the interfaces w, e, t, b, n, s.
relative humidity h, and temperature; Wd = Wd (T) is the bound water Here, the distance between nodes is the same for all directions δx = δy = δz.

4
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

for the relationships between volumes at top, bottom, west, east, north,
and south. The upper case and lower case in Fig. 3 are used to differ­
entiate the consecutive nodes from the control volume interfaces in the
mesh, respectively. The central node in the stencil may stand for pres­
sure, temperature, or moisture content. To obtain the solution fields,
equations (1) and (3) were integrated over the control volume and over
time interval from t to Δt considering the principles of mass and energy
conservation within the boundary conditions shown in Fig. 4.
For the spatial discretization, the two conditioned PDEs (see Eq. (C7)
and C11 in appendix C) were integrated on each control volume shown
in Fig. 3. A piecewise-linear profile was assumed with linear interpola­
tion functions for the state variables P and T between the nodes in the
grid to calculate the fluxes at the control volume interfaces. For the time
discretization it was adopted a fully implicit scheme to define the vari­
able behaviour on each time step according [45]. With this strategy, two Fig. 5. Heat-up curve (HUC) programmed in the electric furnace and behaviour
systems of discretized linear algebraic equations were developed for the of measured internal temperatures over time at different depths in the pris­
nodal values of temperature and pressure. The systems of equations matic specimen.
were solved alternately with a staggered procedure, calculating first the
temperatures field, then that result was used to calculate the pressures 3.3. Boundary conditions
field; thereafter, a time step advanced to compute a next iteration. A
line-by-line procedure was used combining the Gauss-Seidel with the 3.3.1. Mass transport boundary conditions
Thomas algorithms with relaxation through inertia in a top-to-bottom The boundary conditions reproduced the experimental setup, as
sweep direction to transmit the temperatures of the experiment shown in Fig. 4. Concrete surfaces were assumed as permeable bound­
heat-up curve into the domain at each iteration. The solution implies aries, except the lower surface which is joint to the steel shell forming an
that for this highly nonlinear problem, the conservation of mass and impermeable boundary where moisture flux disappears with a Neumann
energy magnitudes are satisfied in every control volume at every time type boundary condition [43], according to Eq. (5):
step. That is the advantage of the FVM, featured for ensuring exact mass
J • n=0 (5)
balance [18,28,35]. Implementations of this method to obtain the dis­
cretized equations for 1D and 2D porous thermo-hydral models can be where n is the orthogonal normal unit vector of the surface.
found in Refs. [18,35,47,48]. For more details about this development in The upper surface was in contact with the atmosphere, the lateral
3D, the reader is referred to Refs. [46,49,50]. surfaces were in contact with the insulating refractory bricks of high
porosity, where vapour was considered to flow freely. Thus, in all those
surfaces P equals the environmental pressure Penv. The vapour exchange
between the surfaces and the environment is not instantaneous, it was

Fig. 4. Discretized mesh with the boundary conditions scheme for a concrete slab heated by the upper face with lateral thermal insulation, and impermeable surface
at the bottom. A control volume is highlighted in the centre denoting its consecutive nodes.

5
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

adopted that the normal flow is proportional to the pressure difference, 4. Experimental results analysis
according to Eq. (6):
4.1. Measured temperatures
J • n = Bw (P − Penv ) (6)

where, Bw is the surface vapour transfer coefficient that depends on air Fig. 5 shows the heat-up curve and temperatures at different depths
movement or the partial insulation in contact with lateral faces [28]. See over time. T20 is the thermocouple at 20 mm, T30 is at 30 mm and so on.
table A3 in appendix A for values and units. The first 10-h holding time at 110 ◦ C in the upper face was not efficient
for drying, i.e., that time did not promote the elimination of free water.
3.3.2. Heat transport boundary conditions That is because the temperatures inside the specimen stayed below the
The radiation condition is described by Stefan-Boltzmann’s law in boiling point, at around 70 ◦ C. These internal temperatures were
Eq. (7). beneficial only for the hydration reactions of concrete curing [38]. Had
( ) it been kept for a longer period, temperatures would have stabilized in
qr • n = εc σSB T 4 − THUC
4
(7) this range, remaining free water. On the other hand, if this holding time
is to be reduced or eliminated, the influence of total real curing time on
where n is the normal unit vector of the boundary surface, qr is the heat concrete properties should be evaluated.
flux by radiation, THUC is the radiating elements’ temperature according In the second and last holding time, the time needed to reach steady-
to the heat-up curve, T is the concrete temperature, σSB is the Stefan- state temperatures in all depths was investigated. The heat-up curve was
Boltzmann constant, and εc is the heat emissivity of the concrete sur­ programmed to reach 300 ◦ C at 21 h and kept stable. About 8 h after that
face [28]; see table A3. were necessary for all internal temperatures to become stable, at around
The lateral surfaces of the specimen were considered thermally t = 30 h; with T20 = T30 = 220 ◦ C; T40 = T50 = 200 ◦ C; T70 = 175 ◦ C,
insulated because of the low thermal conductivity of the insulating re­ and T100 = 160 ◦ C. Those internal temperatures were indeed efficient to
fractory bricks, with a Neumann type boundary condition [43], ac­ dry, i.e., to boil the free water and trigger the transport mechanisms to
cording to Eq. (8): remove the steam from the pores. In the transient trajectory of the
curves, a deflection can be observed at around 20 h, better noticed in
q • n=0 (8)
T100; when temperatures are reaching boiling point. This deviation is
The slab’s bottom face was in contact with a 4 mm thick steel plate explained by the higher energy demand of the latent heat of vapor­
with a considerable thermal conductivity λs (see table A3 for value) and ization. This agrees with the findings reported by Refs. [2,10,13,14,52].
subsequently with environmental air using the Newton’s law of cooling, In this experiment, the frequency of the data gathering permitted to
Eq. (9): distinguish those small temperature variations. In the steady-state, T20
and T30 remained 80 ◦ C below the heat-up curve; however, they
qc • n = BT (T − Tenv ) (9)
reached the oil’s boiling temperature. This could result in false pressure
where T is the slab’s surface temperature. It was considered a surface measurements. This condition brings a restriction for this type of
free convection and BT is the heat transfer coefficient [28]; see table A3. experimental arrangement. The thermocouple in the deepest position is
Since the specimen is in the absence of a vacuum, radiation and heat an important reference for concrete drying research. Mainly in this
conduction occur simultaneously, so the total heat flux at the heated investigation in which, unlike other projects in the literature, the spec­
face is obtained by adding Eq. (7) and Eq. (9) with Tenv = THUC, i.e., imen bottom was considered as a composed impermeable boundary
following the heat-up curve used in the experiment. condition because the 4 mm thick steel plate. A natural cooling stage
begun at 40 h with the specimen kept inside the furnace until all ther­
3.4. Initial values and parameters mocouples reached room temperature, which took about 24 more hours.

A set of concrete and fluid properties were necessary as input pa­ 4.2. Measured pore pressure
rameters. Appendix A summarizes the values of the parameters used and
appendix B the parametric equations to obtain them. The intrinsic Obtaining reproducible pore pressure measurements is a challenging
permeability, κ, was measured from cured concrete samples at room task [2,16,28,53], they are laborious and time-consuming experiments;
temperature using the method proposed by Innocentini et al. [51]. The however, they are crucial for validating numerical simulations. To
initial relative permeability a0, its evolution a(h,T), the desorption iso­ capture the pressure variations reported here, it was necessary to
therms W(h,T) and the water latent heat of vaporization Ca(T) were perform several, repetitive previous unsuccessful tests, many adjust­
calculated according to Refs. [28,43]. The function for bound water ments with state-of-the-art piezoresistive pressure sensors and frequent
dehydration Wd was obtained by thermogravimetric analysis (TGA) readings in time intervals of 2 min during a 40-h lasting experiment.
from “green” concrete samples. The thermal conductivity λ, was initially Fig. 6 shows the experimental pressures over time; P20 is the probe
assumed as constant [43], using the value reported by manufacturer. at 20 mm depth and so on. All sensors, except P70, began to sense
The isobaric heat capacity of concrete C was assumed constant, as re­ pressures when t ≈ 15 h, and detected a pressure peak at t ≈ 18 h. The
ported by Ref. [8]. Apparent density and porosity were measured from maximum pressure value registered was 17.2 × 10− 3 MPa in P20, which
dried concrete samples. Environmental temperature Tenv and pressure is a pressure peak significantly lower than those reported in the litera­
Penv were measured from local conditions. It was supposed an initial pore ture and below the saturation pressure, despite the slow heating rate
relative humidity h0 = 90%, and a uniform initial field of temperatures used. The peaks momentum coincides with water boiling point and the
in equilibrium T0 = Tenv; therefore, according to Eq. B.1, the initial pore deflection observed in the temperature curves shown in Fig. 5. The
pressure in all nodes was P0 = 3595 Pa. The mass of anhydrous cement curves of the other pressure sensors also showed a bell-shape behaviour,
Wc and the initial water content W0 were the same as the experiment. which is a characteristic found in literature [10,15,19,53]. Another
Concrete physical properties for heat and mass fluxes were assumed distinctive feature to notice is the second peak detected by all sensors at
isotropic. Water properties were taken from steam tables. t ≈ 21 h, except failed P70. Other authors have noticed this second peak
behaviour [2,8,19]; it is assumed that the cause of this second peak is
related to the chemically bound water release.
Before concrete casting, all sensors were calibrated and tested. The
possible reason for P70 fail could had been a not detected leakage or a
faulty electronic connection between the sensor, the wires, and the

6
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

sensors was remarkable.


Regarding the measuring instruments, the use of oil as a transfer
medium has a limitation that may become a source of error. Specialized
silicone oils typically boil at around 300 ◦ C, but depending on supplier,
some react at lower temperatures. Some works in the literature report its
use without mentioning its boiling/firing point or thermal expansion. In
this research, although some probes attained the oil’s boiling point, it
was not when the pressure peaks were detected; therefore, measure­
ments were not influenced by oil expansion. The importance of sealing
the system was confirmed. A high temperature silicone sealant was used
in the area where probes exit the concrete; however, it was not enough
to prevent vapour leaks. The negative influence of inserted instruments
to accumulate air bubbles or cracks along them, since moulding and
curing, is aggravated by the lack of adhesion between the concrete and
Fig. 6. Measured pressures over time at different depths in the specimen. Ps(T) the polished stainless steel tube surface. This condition is even worse if
is the vapour saturation pressure corresponding to the temperature of
the surface is unintentionally polluted with oil. Any minor movement
measuring point T20 along the time.
causes the materials to detach, opening a route for vapour to escape or
air to enter. The solution to improve adhesion was by modifying the tube
capturing data hardware, since the beginning of the drying. Because no surface texture with knurling, after that modification, pressure readings
pressure was sensed in only this probe, in any moment of this experi­ were possible. The welded O-rings also helped to avoid circular or lon­
ment. Then, this sensor’s data should not be considered, but it is shown gitudinal movements that could break the adhesion; nevertheless, the O-
here as an example of the possible errors that may occur. The highest- ring located 20 mm below the top of the tube could be substituted by a
pressure peaks were sensed in the experiment only when boiling point fastener at the bottom to also avoid discontinuity and air bubbles in the
was reached internally; albeit low values, the behaviour was consistent mass of concrete around the measuring point. As probes extremely
in all the sensors working properly. This was confirmed by repeating the invasive introduce errors in the system, reduction in tubes diameter is
experiment and observing the same behaviour. It is assumed that slow suggested. Moreover, these modifications must still be tested at faster
heating is best suited to attain higher pressures since fast heating in­ heating rates and higher temperatures for future work, highlighting the
duces thermal damage. However, this assumption may not be general­ importance of including explosion-proof protection, as proposed by
ized, as results depend on several conditions, since sample preparation. Auvray et al. [54].
The curing time and conditions used in this research contrast to previous
experiments, e.g., Kalifa et al. [10] kept the specimens for 6 months in a 4.3. Measured mass loss
sealed bag, Bangi et al. [4] and Choe et al. [14] cured in water bath at
20 ◦ C for 28 days. Thus, the differences regarding the initial pore relative During curing there was no water loss, this was confirmed by
humidity and desorption isotherms could be markedly different. measuring the total mass of the specimen before and after curing. The
The total amount of time that sensors detected positive pressures was total mass was measured just before heating and after cooling. The
considerable. The bell-shaped curve in P20 lasted about 10 h, the specimen mass that entered the furnace was 27.0 kg. The specimen mass
pressure rose from 0 to 17.2 kPa in 3 h and then drop to zero in 7 h, that came out of the furnace was 25.1 kg, so that it lost 1.9 kg. Therefore,
including the second peak. The other sensors were detecting pressures inside the specimen, 0.9 kg of water remained (3.7 wt %), as shown in
for about 6 h in total, including the second peak. This means that there Fig. 7.
was not a sudden drop of pressure caused by any crack or leak in the As T100 reached a stationary state at 160 ◦ C and kept stable for more
system and the length of the bell-shaped curves was the time vapour than 10 h, it is believed that all the free water came out and probably
needed to exit the concrete. It is important to notice here that during the some bound water also. The remaining water is assumed to be exclu­
10-h holding time at 110 ◦ C in the upper face, there was not detected any sively bound water that would need higher temperatures to get out of
pressure that may cause structural risk to the concrete. Therefore, ac­ the specimen; for this reason, the concrete cannot yet be considered
cording to the result of this experiment, for this concrete composition, totally dried.
thickness, and preparation method, the holding time of 10 h at a tem­
perature of 110 ◦ C on the heated face can be inferred as an unproductive 4.4. Concrete physical behaviour analysis
time for the drying operation. Simultaneous peaks at different depths for
one face heated specimen coincide with some results reported in the When furnace temperature reached 300 ◦ C, water dripping was seen.
literature [11,19]. However, this contrasts some others in which sensors It is assumed that it was the moisture transported from the inside out
with different depths detected peaks at considerably different times [4, that condensed on the surface and slipped down through the sides of the
10,52,53]. Although it is not determinant to make direct comparisons specimen. Similar behaviour was reported by Refs. [11,27]. Specimen
between experiments in literature, peaks intensity and timing differ­ dimensions were measured before and after drying, finding neither
ences are due to several factors. The different heating speeds and dis­ dimensional changes nor trace of spalling, cracking nor any other
tance between probes affect timing. Differences in concrete damage by naked eye. Therefore, pressure leaks due to cracks were
composition, preparation, and cure greatly influence the hydraulic discarded. Irregular zones in the external shape of the specimen were
properties of concrete as a porous media. For example, on the one hand, perceived, Fig. 2 (e), which are due to the lack of vibration. After
permeability is affected by vibration, on the other hand, thermal con­ removing the specimen from the furnace, samples were taken by cutting
ductivity, and desorption isotherms are influenced by initial humidity. sections from different regions of the volume to characterize physical
All of them are critical for this kind of problem and should be measured properties of dried concrete; no internal cracks were found.
and reported. Besides that, the boundary conditions used in this study
which have important influence in the fluids mechanics are also 5. Numerical results and discussion
different. Therefore, to this point, it is not possible to define the degree
of accuracy of the pore pressure measurements obtained; future analyses In this section, the numerical solutions are compared with experi­
with more repeated experiments are needed. In this study, the consistent mental measurements at the corresponding coordinates and times. A
record of a pressure peak and the similar behaviour in five out of six Fortran 90 code was developed to solve the discretized linear equations.

7
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

Fig. 7. Reduction of water content sowing the remaining bound water in the specimen dried with a slow one face heating up to 300 ◦ C, as shown in Fig. 5.

The control volumes were refined to form 37.5 × 37.5 × 5.0 mm prisms 5.1. Simulated temperatures
for a total of 2025 nodes placed at the centre of each control volume in
the slab’s domain. For the next simulations, the initial and boundary conditions of the
To validate the computational code, comparisons were made with experimental setup of Fig. 1 were used, they are resumed in Table 2. The
the experimental measurements reported by Fey et al. [2] to dry a parameters needed are listed in Appendix A.
similar refractory concrete with the same specimen size and a heat-up Fig. 9 (a) shows the simulated temperatures; T20sim is the node at
curve with an analogous holding time. A simulation was made repro­ 20 mm, T30sim is at 30 mm and so on; these profiles can be compared
ducing their boundary conditions with a first heating rate of 75 ◦ C/h to with Fig. 5. A direct comparison between the simulated and measured
reach 200 ◦ C at the specimen’s upper face. Subsequently, a slow heating temperatures for the nearest and farthest thermocouple in relation to the
rate of 5 ◦ C/h during approximately 6 h. Thereafter, a heating rate of heated surface is in Fig. 9 (b) and (c) respectively. Measured and pre­
75 ◦ C/h to the final temperature of 700 ◦ C. Fig. 8 (a) and (b) point out dicted temperatures show similar behaviour during the transient state
the experimental results obtained by Fey et al. and the curves obtained (seen in Fig. 5, from 0 < t < 30 h). Some sections presented better
with the numerical implementation of this study, showing a reasonable proximity, e.g., T20sim with T20exp from 0 < t < 14 h in Fig. 9 (b).
agreement along the time. It must be noticed that the permeability and In the second heating ramp, starting at t = 14 h, the predicted tem­
thermal conductivity were adjusted to find the best match with the peratures show a faster response, if compared to the gradual increment
experimental measurements. of the temperatures measured in the experiment. This difference is due
to the real thermal conductivity variations with temperature. From t >
30 h, the simulated temperatures did not reach a steady state but a
tendency to equalize the plateau at 300 ◦ C; different from the experi­
mental temperatures whose stability started at t = 30 h. This is explained
by the fact that the model reproduced the energy conservation assuming
zero loss in the lateral surfaces and a small loss at the bottom, as shown
in the boundary conditions in Fig. 4; by contrast, the experimental re­
sults showed that there was a larger energy loss. Part of the unconsid­
ered energy loss in the system was due to the air circulation promoted by
the chimneys and a millimetric air gap between the specimen and the
lateral insulating bricks. That gap was there because the specimen was
inserted and removed from below the furnace. Fibreglass was added to
cover the gap and avoid air circulation; however, the energy loss was

Table 2
3D model initial and boundary conditions.
Space and time discretization
Slab dimensions 300 × 300 × 120 mm
Calculated time 40 h
Space step δx = δy = 37.5; δz = 5.0
mm
Time step Δt = 0.25 s
Initial conditions
T (x, y, z, t = 0) T0 = 302.15 K
P (x, y, z, t = 0) P0 = 3595 Pa
W (x, y, z, t = 0) W0 = 258.98 kg/m3
Boundary conditions
T (x, y, z = 120, t) Rad. + Conv.: q • n = εc σSB (T4 − THUC
4
)+
BT (T − THUC )
T(x|x=300
y=300 Adiabatic: q • n = 0;
x=0 , y|y=0 ,
z, t)
T (x, y, z = 0, t) Convection: q • n = BT (T − Tenv );
P (x, y, z = 120, t) Permeable: J • n = Bw (P − Penv )
P(x|x=300
y=300 Permeable: J • n = Bw (P − Penv );
Fig. 8. Comparison of measured (Exp) and simulated (Sim) results of (a) x=0 ,y|y=0 ,z,
t)
Temperatures and (b) Pressures. The marked points were obtained from the
P (x, y, z = 0, t) Impermeable: J • n = 0;
experimental results originally measured by Fey et al. [2].

8
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

Fig. 10. Temperature field in the specimen at t = 25 h. The coloured scale


reports the intensity of simulated temperature (◦ C) for the rectangular control
volumes in the domain.

surface in contact with the steel plate is perceived, as a result of a ver­


tical unidirectional thermal flow. At the bottom, the thermal flow
continued unidirectional through the steel shell that has a thermal
conductivity considerably larger than the specimen’s thermal conduc­
tivity (see tables A.1 and A.3 in appendix A).

5.2. Simulated pore pressures

Fig. 11 (a) shows the simulated pressures in the nodes at the same
coordinates of the probes. The predicted maximum pressure was 2.6
MPa with the highest value in the bottom node P100 at t = 36 h. It is
important to notice that these numerical results are first-hand with no
adjustments in any parameter to match the experimental results. To be
aware of the effect caused by parameter adjustments, another simulation
was performed modifying only one parameter. Instead of using a con­
stant concrete thermal conductivity, a function of temperature obtained
from a similar concrete was used (equation B.11 in appendix B). Fig. 11
Fig. 9. (a) Simulated temperatures over time in nodes at the same coordinates
(b) shows that in this other simulation the predicted pressure peak was
of thermocouples. (b) Comparison of simulated and experimental temperature
profiles at 20 mm depth from heated surface. (c) Comparison at 100 mm depth. less than the half of the previous one. It reached 1.1 MPa, 6 h before, at
time t = 30 h, with a narrower bell shape and it presented a small second
peak of 0.75 MPa in the descending part of the curve, at t = 32 h. The
higher than assumed. This is evidence of the strong influence of
thermal conductivity as a function of temperature resulted in values five
boundary conditions. Further experiments with adjusted experimental
times higher, modifying the speed of the thermal processes.
and numerical boundary conditions are required. To avoid lateral losses,
Theoretically, the pressure starts to rise when the evaporation front
an option is to use aluminium foils glued with epoxy resin in the lateral
passes through the respective finite volume. Fig. 11 (a) and (b) show
surfaces, as reported by Refs. [13,21,22], although its influence was
how the pressure started to rise at t = 15 h. If Fig. 9 is reviewed, this
found minimal in small samples [25], more tests are needed in larger
happens when T20sim = 100 ◦ C. In the experimental measurements, the
specimens.
pressure started to rise at the same time, but with different intensity, as
The model was not sensitive to the deflection in the temperature
seen in Fig. 11 (c), where the numerical results of the second simulation
curves caused by water vaporization, as seen in the experiment at t = 20
are compared with the experimental results in P20 and P100. The
h, better noticed in T100 exp, Fig. 9 (c). Although it was used a
experimental peak reached a maximum 17.2 × 10− 3 MPa in the node
formulation for the water latent heat of vaporization as a function of
P20, which is closer to heated face, at t = 18 h. A detailed analysis in
temperature (see equation B.12 in appendix B), it did not reproduce the
Fig. 11 (a) reveals that P20 > P100 when t < 23 h; but when t > 23 h they
deflection observed close to the boiling temperature. One reason for the
invert, remaining P100 > P20, a similar behaviour can be perceived in
difference is that concrete and water isobaric heat capacities were
Fig. 11 (b). This contrasts with the experimental results in Fig. 11 (c),
assumed as constants.
where P20 remained higher along the time. Samples may present
explosive spalling close to heated face, or, when they accumulate energy
5.1.1. Temperature field
in the core, a catastrophic explosion occurs [56,57]. In this study, any
Fig. 10 illustrates the simulated temperatures distribution in the
pore pressure attained excessive values, e.g., the saturation pressure was
concrete domain at t = 25 h; this is a moment before the physical
not reached at any time. If there are not high pressures, that is because
experiment reached the thermal steady-state. The radiant heating is
the pressured steam flowed away, and the pressure dropped. This may
evident at the upper surface with no thermal loos in the sides because of
happen during a long period of time, as needed for the steam to find the
the thermal insulation condition in the lateral surfaces. A uniform
way out through the pores, or in a fast pressure drop originated by a
temperature gradient with a uniform pattern towards the colder lower
sudden crack in the structure that lets the steam escape.

9
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

Fig. 11. Predicted pressures at different depths inside the domain over time.
(a) Profile with constant thermal conductivity λ = 1.34 W/mK; (b) Profile with
variable thermal conductivity λ = λ(T), see Eq. B.11 in appendix B. The vapour
saturation pressure curve, Ps(T), was calculated with Eq. B.2 using the simu­
lated temperature T20 along the time; (c) Comparison of experimentally
measured and numerically simulated pressures.

In this study, the only similitude between numerical and experi­


mental results is the bell-shaped behaviour which is observed repeatedly
in the literature. However, by comparing the numerical results of Fig. 11
(a) and (b) with the numerical and/or experimental results of previous
authors [2,4,9–11,14,53,55], where similar pressure magnitudes are Fig. 12. Pore pressure evolution over space and time. The coloured scale re­
reported, it can be inferred that the experimental pressure measure­ ports the intensity of simulated pressure (Pa) in the rectangular control volumes
ments of this study are not convincing. Nevertheless, it is important to of the domain. (a) pressure field in a 25% section of the domain at t = 15 h. (b)
consider that the mathematical model implemented to obtain the nu­ pressure field in a 50% section at t = 25 h. (c) pressure field in a 25% section at
merical results conveys important simplifications, and, although inter­ t = 30 h.
esting results are shown, they cannot be considered definitive. The lack
of exact reproduction of the boundary conditions, that brought a larger direction to the lower volumes.
energy loss than assumed, contributed to the indecisive match between In Fig. 12(b), the temperature in all the volumes had surpassed the
the experimental and the numerical results in this study. Therefore, for boiling point and the predicted higher pressures moved to the centre;
the prediction of pressures and eventual explosions, accurate replication therefore, the simulation reproduced the pressure migration as theo­
of the boundary conditions and satisfactory comparisons of numeric retically expected. Fig. 11 (c) shows the advance of pressure to the lower
results with a large set of reproducible experiments are necessary. and lateral volumes unveiling a low-pressure area at the bottom centre
due to a higher water saturation resulting from the impermeable
5.2.1. Pressure field boundary condition. It is also seen an increase in the lateral moisture
The sequence of 3D images in Fig. 12 shows the simulated pore transfer as the lateral sides would be the natural way to release pressure.
pressure evolution over space and time for different sections of the Reaching the boiling temperature in the system does not guarantee
domain. In Fig. 12 (a), the upper representative volumes were reaching that all the free water has been removed. A transitory period is necessary
the boiling temperature, and as expected, the higher vapour pressures for the steam to flow, triggered by pore pressures, and leave the porous
were predicted to be close to the heating source with a gradient in medium. Free liquid water can be considered to cease being present in

10
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

the domain when all the nodes reach thermal equilibrium above 100 ◦ C
without subsequent mass-loss. Furthermore, drying can be considered
complete only when both, physically adsorbed and chemically bound
water, are totally released, and that can occur when temperatures reach
around 550 ◦ C [28].

5.3. Simulated moisture content

Fig. 13 shows the simulated water content profile. All the finite
volumes were integrated to analyse the total water mass-loss starting
from W0 used to prepare the concrete mix. The graph shows a slow mass
variation from 0 < t < 15 h; if collated with Fig. 9, a steep mass-loss is
predicted only after internal temperatures exceed 100 ◦ C, when t > 15 h.
Conversely, if Fig. 13 is collated with Fig. 11 (a), the highest pore
pressure is reached at t = 36 h. That is, the highest pressure was attained Fig. 14. Prediction of chemically bound water release in the domain over time.
when free water amount was reaching a minimum. A reason for that is
found in Fig. 14, where a pronounced increase of bound water release is direction of the vapour flow begins a multidirectional trajectory that
predicted around that time. Therefore, the model predicted a significant will subsequently affect the finite volumes of other areas in the domain.
effect on pressures caused by the release and evaporation of bound The impermeable bottom surface forces the vapour to flow to the sides,
water. This result comes from the combination of the desorption iso­ but, as it is a porous media which also restricts the flow, part of it ac­
therms and the TGA interpolated function that predicted the release of cumulates in the centre. The transient patterns of the sections shown in
bound water at a temperature above boiling point, i.e., a violent Figs. 12 and 15 appear because of the vapour accumulation and vapour
expansion of water that suddenly appeared. Figure B1 in appendix B turbulence in different directions. This complex pattern is only possible
presents the TGA results showing that the steepest mass-loss takes place to unveil with a transient 3D model. The drying mechanism starts with
when the temperature exceeds 225 ◦ C; if Fig. 9 is reviewed, this internal the evaporation of free water, increasing the vapour content in the finite
temperature is predicted at t = 35 h. However, in the experiment the volumes, later by the vapour pressure resulting from the restriction of
specimen did not reach that temperature, as seen in Fig. 5. the porous media. Pressure increment forces the vapour to flow. Part of
In this research, the mass of the experimental specimen was the vapour exits the domain and part flows toward the colder internal
measured only before and after heating, Fey et al. [2] did not measure volumes. Finally, a font of moisture appears in every finite volume due
the mass considering that a 1D model is unsuitable for mass loss com­ to the bound water release, repeating the cycle of flow forced by
parisons. Nonetheless, to define the precise time in which all the free pressure.
water was removed, and for better numerical validation, the most The model reproduced the blocking exerted by the impermeable
convenient practice would have been to measure the mass variation surface in the xy-plane on the vertical moisture flow, forcing it to
continuously and perform transient comparisons in 3D. permeate horizontally along the steel shell in a transversal direction to
the slab’s lateral surfaces, see Fig. 14 (b). The effect of water saturation
5.3.1. Moisture content field on pore pressure development becomes clear if Figs. 14 (c) and Fig 11 (c)
Fig. 15 shows the simulated evolution of moisture distribution over are collated to notice that the large liquid saturation in the lower central
space and time for different sections of the domain. The results predict finite volumes creates a barrier of condensed moisture so that the vol­
how, although the thermal flow perceived in Fig. 10 was unidirectional, umes with higher pore pressure are located outside that zone. The
due to the boundary conditions, the moisture has multidirectional saturated areas become a thermal barrier since moisture consumes a
migration to the upper and lateral surfaces to gradually exit the domain. large amount of thermal energy to evaporate, reducing the flow of heat
This can be perceived thanks to the analysis in 3D. Initially, in the upper and delaying the development of pressure.
region, the drying front advances parallel to the heated surface, Fig. 15 From a qualitative point of view, the 3D simulation replicated the
(a). The subsequent drying front by the sides results from the lateral moisture content behaviour. The advance of the drying front within the
boundary condition that permeates moisture through the porous bricks, domain is in accordance with reality. These images can be compared
Fig. 15 (b). This is not the case for the lower coldest region and the with state-of-the-art experimental procedures, where neutron radiog­
central volumes, as delayed moisture loss is observed due to the raphy or neutron tomography have been used, like those performed by
impermeable surface, Fig. 15 (c). In the finite volumes at the bottom, the Refs. [21,22]. As dehydration decomposes concrete, the model coupled
the mechanisms of heat and moisture flow with the simulation of the
microstructure evolution performed through the increase of perme­
ability as a function of temperature and humidity, according to Eq. B.4
in Appendix B.
Collating Fig. 12 with Fig. 15 permits to notice that in Fig. 12 (b) and
(c), the maximum pressure is no longer in the upper volumes, the
pressure distribution was inverted since at t = 15 h the water content in
the upper volumes is minimal, as they are already dried, Fig. 15 (b) and
(c). As the drying front advances, the pressure is transported to a lower
central region.
Behind the drying front, a fraction of the evaporated water condenses
when it finds lower temperatures. That condensation increases the
saturation again and decreases the permeability to a critical point until it
blocks the pores and prevents the scape of subsequent steam. This is
known as the “moisture clog”. If temperatures and pressures continue
increasing faster than the vapour release, the probability of explosive
Fig. 13. Prediction of water mass-loss in the domain over time.

11
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

the removal of the chemically bound water at higher temperatures.

5.4. Influence of the methods used to characterize properties

The methods to define the necessary parameters may bring signifi­


cant variations in the simulations. In sensitivity analyses Fey et al. [2]
and Moreira et al. [44] concluded that permeability, thermal conduc­
tivity, and desorption isotherms, in that order, had the biggest impact on
results. However, it is necessary to consider the effect of other param­
eters because of the differences in composition, preparation, cure, age,
geometry, and boundary conditions.
In this research, the measured initial intrinsic permeability had a
significant standard deviation among samples and the resulting value
can be considered high for this type of concrete. That is the result of a
non-vibrated specimen, which highlights the importance of controlled
vibration to prepare the samples. The relative permeability as a function
of temperature and humidity determined by an empirical formula for
generic concretes [28] gave pressure values comparable to those found
in the literature; albeit, different to the measured pressures. Initial
intrinsic permeability can be measured using a Cembureau per­
meameter; however, measuring the evolution of intrinsic permeability
in recently cured CAC concrete while it is being heated, and modelling it,
becomes a complex endeavour [58]. Several models are proposed in the
literature to describe permeability as a function of temperature, hu­
midity, pressure, damage, tortuosity, and the consideration of the
Klinkenberg and inertial effects [9,33]. A convenient practice would be
to characterize this parameter with the method at high temperatures
proposed by Innocentini et al. [59], from “green” concrete.
Thermal conductivity variation was initially assumed slight and
disregarded, as originally supposed by Gong and Mujumdar [60];
however, the simulations shown in Fig. 11 (b) proved a keen effect,
alerting that a function of temperature is more representative. Accord­
ing to Bazant and Jirásek [28] the thermal conductivity varies consid­
erably as a function of temperature and humidity and can be expressed
as proposed by Gawin et al. [61]. Fey et al. [2] obtained this
double-dependent parameter from “green” concrete, resulting four times
higher than the value reported by the manufacturer. Further simulations
and comparisons are necessary using functions of temperature and hu­
midity obtained by characterizing “green” samples of the specific cast­
able with the hot-wire method.
It was confirmed that one of the most challenging model components
are the desorption isotherms. The generic formulation for temperatures
above 100 ◦ C adopted from Ref. [28] provided moisture content results
qualitatively in accordance with reality. Fey et al. [2] used a sorption
isotherm obtained at 20 ◦ C with the aid of dynamic vapour sorption
(DVS) and reported a big effect on maximum pressure curve.
Baroghel-Bounny [62] made available a useful isotherms database and
Fig. 15. Moisture content evolution over space and time. The coloured scale Davie et al. [63] proposed an alternative formulation. Notwithstanding,
reports the intensity of simulated water concentration (kg/m3) in the rectan­ the difficulties arise from the need for a formulation for the specific
gular control volumes of the domain. (a) moisture content field in a 40% section
concrete at high temperatures. The use of neutron image analyses [25,
of the domain at t = 15 h. (b) moisture content in a 15% section at t = 25 h. (c)
64] is a promising alternative to improve the numerical results. As a
moisture content in a 20% section at t = 30 h.
consensus, this parameter represents an opportunity for higher research.
The simulation was performed assuming the initial value of pore
spalling increases. relative humidity h0 = 90%, as originally supposed by Gong and
During heating up to 300 ◦ C, the experimentally measured, and the Mujumdar to model refractory concrete drying [8,43,60]. Nonetheless,
numerically calculated pore pressures were low. Those pressures in the this value must be evaluated to accurately represent the specific “green”
pores do not represent any risk to create a stress that could overpass the castable. The overestimation of this parameter may lead to pressure
tensile strength σT reported by manufacturer for this specific concrete results higher than reality. Refractory concretes curing condition and
(see table A1). After heating, no traces of cracks were found. The heat-up initial pore saturation contrasts with the case of aged HPC in environ­
curve started to be effective to dry only after 20 h of heating. For this mental conditions, or the case of specimens in water bath or longer
specific concrete, the empirical practice of holding time at 100 ◦ C in the curing times. Chemical balance analysis of hydration and dehydration
heating face represents an important waste of time and energy. reactions at different temperatures would be useful to estimate the
Notwithstanding, further research is needed to apply modifications in initial pore relative humidity and bound water amount that would be
the industry for the effect of this holding time in curing and because the expected to stay or leave at different stages of curing and drying.
presented model is not yet considering the utterly mechanical aspects Scanning electron microscopy of “green” microstructure and neutron
and the addition of stresses resulting from the next drying step, which is tomography could be helpful resources.

12
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

The effect of bound water release was perceived in the simulation 6. Conclusions
performed in this research, although its resulting effect on pressures is
debatable. The prediction of a fast increment of pressure when bound 1) The definition of safe and efficient dry-out schedules to install
water is released was also perceived by Fey et al. [2], considering it an monolithic refractories is a scientific and technological challenge.
accuracy fail. According to Mindeguia et al. [11] a small amount of This research tested combined experimental and numerical meth­
bound water does not amplify the build-up of pressure neither repre­ odologies aimed to evaluate a heat-up curve currently used to dry a
sents a risk of explosion, and Fey et al. [2] affirms that bound water high-alumina castable through the analysis of pore pressures, inter­
mostly affects the drying time and disregarding it will not modify nal temperatures, and mass loss behaviour. Details of some diffi­
maximum pressures unless released at temperatures close to saturation culties and sources of error are presented as well as suggestions to
temperature. Hence, it is convenient to obtain this parameter from the overcome them.
specific castable with the appropriate methodology. It must be consid­ 2) For the castable prepared in this research, the 10-h holding time at
ered that the characterization of bound water release is performed using 110 ◦ C on the heated surface was inefficient to remove the free water,
considerably small samples in an equilibrium state; when heating a therefore it misspends time and energy. However, as the internal
larger volume in different conditions the equilibrium is attained in temperatures reached are beneficial for hydration reactions (around
different times and places for the reactions that release bound water. 70 ◦ C), further evaluations of the real total curing time and final
Besides that, the desorption isotherm determines the ultimate behaviour properties of the concrete should be performed, before suggesting
of the bound water recently released. This confirms the need for more any change. The second ramp and the plateau at 300 ◦ C on the heated
research regarding the effect of chemical reactions during drying and the surface shows to be effective to remove the free water. After 10 h
determination of more accurate specific models of desorption isotherms. with T100 at 160 ◦ C, only one-third of the water originally used in
Therefore, for bound water release input data, the most convenient the mix remained, which is believed to be exclusively chemically
practice is to perform representative TGA from concrete samples bound water. Nevertheless, for the sake of energy efficiency, a
without free water, i.e., previously dried at 100 ◦ C, and not from “green” unique heating rate of 30 ◦ C/h with only one holding time at 250 ◦ C
samples. It is also convenient to ensure the presence of aggregates and would have been enough to remove the free water from the 120 mm
matrix in the samples. The late is applicable to characterize the concrete thick slab of this specific concrete.
latent heat of dehydration through differential scanning calorimetry 3) From a quantitative point of view, the experimental method to
(DSC). Conversely, for concrete heat capacity as function of temperature measure temperatures is the most reliable result, followed by mass
and humidity, to obtain representative DSC analysis, they should be variation measurement. The numerical validations can be reliably
performed with “green” samples. Appropriate TGA characterization performed using those two variables initially to predict the heat and
together with thermal conductivity are also necessary for the confir­ mass loss behaviour. In laboratory practice, reaching the boiling
mation of the second pressure peak that was detected in the experiment point in the thermocouple close to the cold side, along with stability
and predicted in the model. in mass variation, can help to define the time in which free water
In this research, the amount of air and its effect was assumed removal is complete.
negligible. That might be expected from a “green” castable, especially if 4) It was possible to sense pore pressures with the use of oil and knurled
the mix has been vibrated and the air bubbles removed. Fey et al. [2,12, probes; however, measured pressures resulted considerably lower in
36] implemented an advanced model considering the air mass in the comparison to simulated pressures. Although the sensors were cali­
balance equations, resulting in a minimal contribution to the pore brated before the experiment, and the numeric methodology is mass-
pressures. Nevertheless, some experimental works in the literature re­ conservative, the pressure data obtained in this investigation is
ported pressure measurements higher than vapour saturation and neither enough to draw definitive inferences nor suitable to control
explained that the reason was the air mass expansion [10,11]. In this drying in a technological application.
work such effect was not perceived, as pressures remained below vapour 5) The numerical results in 3D of this research are a valuable resource to
saturation. To confirm the effect of air on the drying of refractory con­ visualize the transport phenomena, although they can still only be
crete linings, a future step is to apply multiphase models existing in the used as a rough guide. Yet, this experience showed that by combining
literature, like [2,29,31,34,37,65,66], and perform analyses in 3D with the experimental and numerical methodologies, more comprehen­
the specific boundary conditions at higher temperatures. sive and useful insights can be obtained to evaluate the current heat-
Regarding the use of a specific numeric method to solve the PDEs, up curves used in industry.
there are arguments about the convenience of using one or another. For
example, there are questions about the use of FVM because the non- Funding sources
linearity of the problem is stronger than the non-conservativity of
FEM. It must be considered that the two methods handle strong non- This study was financed in part by the Conselho Nacional de
linearities in a similar form, with approximations using the tangents of Desenvolvimento Científico e Tecnológico – CNPq (grant number:
the interpolation functions for the state variables at each node. How­ 140142/2017–7), by Fundação de Apoio à Ciência, Tecnologia e Edu­
ever, the FVM, besides handling non-linearities, adds a treatment for cação - FACTE and by the Coordenação de Aperfeiçoamento de Pessoal
mass conservation in the grid, which is important if an abrupt moving de Nível Superior - Brasil (CAPES) - Finance Code 001. No one of the
interface between oversaturated and dried concrete occurs. In those funding sources was involved in the study design, collection, analysis,
cases, the local mass conservation is guaranteed by using FVM. For and interpretation of data, neither in the writing report nor the decision
example, in the accidental cases of fire with rapid heating of a structural to submit the article for publication.
column, or an abrupt thermal load due to cooling failure in a refractory
wall. FVM is recommended for complex geometries, and for interfacing Declaration of competing interest
coatings with layers of different castables and properties, as frequently
found in refractory linings. For example, to analyse the drying of the The authors declare that they have no known competing financial
blast furnace running channels, as intended for future research. The FEM interests or personal relationships that could have appeared to influence
and FDM can also provide reliable results in those cases, but with special the work reported in this paper.
care, additional derivations, using the lowest-order finite elements and
implicit time integration. Data availability

Data will be made available on request.

13
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

Acknowledgement development of the experimental device, and to Gabriel Antonio Mar­


sola, Leonardo Augusto Damasceno and Isli Samara Flauzino from
The authors wish to thank Bento Ferreira and Leandro Kellermann de Universidade de Ribeirão Preto for their valuable support to measure
Oliveira from EEL-USP for their valuable contributions in the concrete permeability.

Appendix A. Parameters needed for the simulations


Table A.1
Parameters from concrete physical properties.

Description Nomenclature Value Units


Initial pore relative humidity h0 90 %
Pore relative humidity evolution h(T,P) Eq. B.1 %
Intrinsic permeability κ 1.45 x 10-15 m2
Initial relative permeability a0 Eq. B.3 m/s
Relative permeability evolution a(h,T) Eq. B.4 m/s
Free water content W(h,T) Eq. B.8 kg/m3
Chemically bound water released Wd (T) Eq. B.10 kg/m3
Mass of anhydrous CAC (23 wt %) Wc 504.23 kg/m3
Thermal conductivity as a constant λ 1.34 W/m K
Thermal conductivity as a variable λ(T) Eq. B.11 W/m K
Isobaric heat capacity C 1100 J/kg K
Apparent solid matrix density ρ 2380 kg/m3
Apparent porosity n 23.17 %
Tensile strength σT 11.5 MPa

Table A.2
Parameters from fluid properties (water liquid or vapour).

Description Nomenclature Value Units


Vapour saturation pressure Ps(T) Eq. B.2 Pa
Isobaric heat capacity Cw 4182 J/kg K
Latent heat of vaporization Ca(T) Eq. B.12 J/kg
Density, at 25 ◦ C ρa 997.05 kg/m3
Viscosity, at 25 ◦ C ηa 890 × 10− 6 N s/m2
6
Vapour viscosity, at 99,6 ◦ C ηv 282.7 × 10− N s/m2

Table A.3
Parameters for boundary conditions.

Description Nomenclature Value Units


Environmental temperature Tenv 302.15 K
Environmental relative humidity (RH) henv 56 %
Environmental pressure Penv 101,325 Pa
Gravitational acceleration g 9.81 m/s2
Convection vapour transference Bw 1 × 10− 6 s/m
Surface convection heat transference BT 1 W/m2K
8
Stefan-Boltzmann constant σSB 5.67 × 10− W/m2K4
Concrete surface heat emissivity εc 0.9 –
Steel plate thermal conductivity λs 50.2 W/m K

Appendix B. Parametric equations

Pore relative humidity [28,44,60]

P
h(T, P) = × 100% (B.1)
Ps (T)

where, P is pore vapour pressure (Pa), and Ps(T) is the saturation vapour pressure (Pa) dependent of absolute temperature T (K) according to Antoine’s
function:

14
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

⎧ ( )

⎪ 11.9515(T− 373.15 )


⎨ Penv e
T− 39.724
(T < 373.15 K);
Ps (T) = . (B.2)

⎪ ( )

⎩ 12.1074(T− 373.15 )
T− 28.665
Penv e ( T ≥ 373.15 K);

Initial Relative permeability [60]

ρa g
a0 = κ (B.3)
ηa

Variable Relative permeability [8,28]

{
a0 f1 (h, T) f2 (T)(T ≤ Ttr );
a(h, T) = (B.4)
a0 f2 (Ttr ) f3 (T)(T > Ttr );

where,


⎪ 1.2893 − 0.01357 (T − 273)

⎪ + 0.01357(T − 273) − 0.2893(h < 1);


⎨ 1 + [4(1 − h)]4
.
f1 (h, T) = (B.5)



⎪ .



1; (h ≥ 1)
[ ( )]
2700 1 1
T0 − T
f2 (T) = exp (B.6)
[ ]
T− Ttr

(B.7)
0.881+0.214(T− Ttr )
f3 (T) = exp

with T and T0 = Tenv (K), and the transition temperature for high permeability Ttr = 368.15 K.

Free water content [8,28]

⎧ )1/ (

⎪ W0 m(T)

⎪ h Wc
(h ≤ 0.96);

⎪ Wc



⎪ .


⎨ W2 − W1
W(h, T) = W1 + (h − 0.96)
1.04 − 0.96
(0.96 < h < 1.04); (B.8)



⎪ .



⎪ [ ( )]

⎪ (T − 273)2

⎪ W 0.037(h − 1.04) + 0.3335 1 − (h ≥ 1.04);
⎩ c
3.6 ∗ 105

Where, W1 and W2 are the water content calculated at a given temperature when h = 0.96 and h = 1.04, respectively; and m(T) is a temperature-
dependent empirical coefficient:
1
m(T) = 0.04 + / (B.9)
1 + (T − 263)2 27370

where T is absolute temperature.

Chemically bound water released



⎪ 0.0017 ∗ T̂ − 0.0171(0 < T
̂ < 200 ◦ C)


⎨ .
Wd (T) = 0.0407 ∗ T̂ − 7.5143(200 ≤ T̂ < 400◦ C) (B.10)



⎪ .
⎩ ̂ + 6.5253(400 ≤ T̂ ≥ 800◦ C)
0.0025 ∗ T

15
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

Where T
̂ is temperature in ◦ C.

Fig. B.1. TGA (green) and DSC (brown) curves from cured concrete samples.

Variable thermal conductivity

λ(T) = 11.5 − 0.029 T + 3.88x10− 5 T 2 − 1.89x10− 8 T 3 (B.11)


With T in K; obtained using the parallel hot-wire method ISO 8894–2 from a similar high-alumina castable.

Water latent heat of evaporation [61]

{
2.672x105 (Tcr − T)0.38 (T ≤ Tcr )
Ca (T) = (B.12)
0(T > Tcr )
With T in K, and the critical water temperature Tcr = 647.15 K.

Appendix C. Partial Differential Equations conditioning

Mass conservation equation [8,28]

Substituting equation (2) in equation (1),


( )
∂W a ∂Wd
= ∇ • ∇P + (C.1)
∂t g ∂t
As W is a thermodynamic function of T and P, that is, W = W(P,T), it can be written
∂W ∂W ∂P ∂W ∂T
= + (C.2)
∂t ∂P ∂t ∂T ∂t
Using
from equation (C.2) in equation (C.1),
∂W
∂t
( )
∂W ∂P ∂W ∂T a ∂Wd
+ = ∇ • ∇P + (C.3)
∂P ∂t ∂T ∂t g ∂t
For practicality, coefficients A1 and A3 are defined as
⃒ ⃒
∂W ⃒⃒ ∂W ⃒⃒
A1 = ; A = (C.4)
∂P ⃒T ∂T ⃒P
3

∂W
∂Pand ∂∂W
T
can be calculated by central difference approximation to obtain step n+1, by
( )
∂W(P, T) W(P + ΔP, T) − W(P − ΔP, T)
≅ (C.5)
∂P T 2ΔP
( )
∂W(P, T) W(P, T + ΔT) − W(P, T − ΔT)
≅ (C.6)
∂T P 2ΔT

Employing (C.4) in equation (C.3) and rearranging, the pressure transient term is isolated as
⎛ ⎞
∂P a/g
⎜ ⎟ 1 ∂Wd A3 ∂T
=∇ • ⎝ ∇P⎠ + − (C.7)
∂t A1 A1 ∂t A1 ∂ t

16
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

Energy conservation equation [8,28]

Substituting equation (4) in equation (3),


( )
∂T ∂W a
ρC − Ca = ∇ • (λ∇T) − Cw ∇P • ∇T (C.8)
∂t ∂t g

Using from equation (C.2) in equation (C.8) and rearranging,


∂W
∂t
( ) ( )
∂W ∂P ∂W ∂T a
− Ca + ρC − C a = ∇ • (λ∇T) − Cw ∇P • ∇T (C.9)
∂P ∂t ∂T ∂t g
For practicality, coefficients A2 and A4 are defined as
⃒ ⃒
∂W ⃒ ∂W ⃒
A2 = − Ca ⃒⃒ ; A4 = ρC − Ca ⃒⃒ (C.10)
∂P T ∂T P
Using (C.5) and (C.6), substituting (C.10) in (C.9) and rearranging, the temperature transient term is isolated as
( ) ( )
∂T λ Cw a A2 ∂P
=∇ • ∇T − ∇P • ∇T − (C.11)
∂t A4 A4 g A4 ∂ t

References high temperatures, Int. J. Heat Mass Tran. 47 (2004) 135–147, https://doi.org/
10.1016/S0017-9310(03)00381-8.
[19] L.T. Phan, Pore pressure and explosive spalling in concrete, Mater. Struct. Constr.
[1] M.D.M. Innocentini, F.A. Cardoso, A.E.M. Paiva, V.C. Pandolfelli, Dewatering
41 (2008) 1623–1632, https://doi.org/10.1617/s11527-008-9353-2.
refractory castables, Am. Ceram. Soc. Bull. 83 (2004) 9101–9108. www.
[20] K. Krzemień, I. Hager, Assessment of concrete susceptibility to fire spalling: a
ceramicbulletin.org.
report on the state-of-the-art in testing procedures, in: Procedia Eng., Elsevier Ltd,
[2] K.G. Fey, I. Riehl, R. Wulf, U. Gross, Experimental and numerical investigation of
2015, pp. 285–292, https://doi.org/10.1016/j.proeng.2015.06.149.
the first heat-up of refractory concrete, Int. J. Therm. Sci. 100 (2016) 108–125,
[21] N. Toropovs, F. Lo Monte, M. Wyrzykowski, B. Weber, G. Sahmenko, P. Vontobel,
https://doi.org/10.1016/j.ijthermalsci.2015.09.010.
R. Felicetti, P. Lura, Real-time measurements of temperature, pressure and
[3] P. Kalifa, G. Chéné, C. Gallé, High-temperature behaviour of HPC with
moisture profiles in High-Performance Concrete exposed to high temperatures
polypropylene fibres - from spalling to microstructure, Cement Concr. Res. 31
during neutron radiography imaging, Cement Concr. Res. 68 (2015) 166–173,
(2001) 1487–1499, https://doi.org/10.1016/S0008-8846(01)00596-8.
https://doi.org/10.1016/j.cemconres.2014.11.003.
[4] M.R. Bangi, T. Horiguchi, Effect of fibre type and geometry on maximum pore
[22] D. Dauti, A. Tengattini, S. Dal Pont, N. Toropovs, M. Briffaut, B. Weber, Analysis of
pressures in fibre-reinforced high strength concrete at elevated temperatures,
moisture migration in concrete at high temperature through in-situ neutron
Cement Concr. Res. 42 (2012) 459–466, https://doi.org/10.1016/j.
tomography, Cement Concr. Res. 111 (2018) 41–55, https://doi.org/10.1016/j.
cemconres.2011.11.014.
cemconres.2018.06.010.
[5] R. Salomão, V.C. Pandolfelli, Polypropylene fibers and their effects on processing
[23] A.J. Barakat, L. Pel, O. Krause, O.C.G. Adan, Direct observation of the moisture
refractory castables, Int. J. Appl. Ceram. Technol. 4 (2007) 496–502, https://doi.
distribution in calcium aluminate cement and hydratable alumina-bonded
org/10.1111/j.1744-7402.2007.02170.x.
castables during first-drying: an NMR study, J. Am. Ceram. Soc. 103 (2019) 1–13,
[6] A.P. Luz, M.H. Moreira, C. Wöhrmeyer, C. Parr, V.C. Pandolfelli, Drying behavior
https://doi.org/10.1111/jace.16814.
optimization of dense refractory castables by adding a permeability enhancing
[24] M.H. Moreira, S. Dal Pont, A. Tengattini, A.P. Luz, V.C. Pandolfelli, Experimental
active compound, Ceram. Int. 45 (2019) 9048–9060, https://doi.org/10.1016/j.
proof of moisture clog through neutron tomography in a porous medium under
ceramint.2019.01.242.
truly one-directional drying, J. Am. Ceram. Soc. 105 (2022) 3534–3543, https://
[7] V.S. Pinto, D.S. Fini, V.C. Miguel, V.C. Pandolfelli, M.H. Moreira, T. Venâncio, A.
doi.org/10.1111/jace.18297.
P. Luz, Fast drying of high-alumina MgO-bonded refractory castables, Ceram. Int.
[25] D. Dauti, A. Tengattini, S.D. Pont, N. Toropovs, M. Briffaut, B. Weber, Some
46 (2020) 11137–11148, https://doi.org/10.1016/j.ceramint.2020.01.134.
observations on testing conditions of high-temperature experiments on concrete:
[8] Z.X. Gong, A.S. Mujumdar, Review of R&D in drying of refractories, Dry. Technol.
an insight from neutron tomography, Transport Porous Media 132 (2020)
25 (2007) 1917–1925, https://doi.org/10.1080/07373930701727200.
299–310, https://doi.org/10.1007/s11242-020-01392-2.
[9] G. Palmer, J. Cobos, J. Millard, T. Howes, The accelerated drying of refractory
[26] J. Zhao, J.J. Zheng, G.F. Peng, K. Van Breugel, A meso-level investigation into the
concrete - Part 2 numerical modelling, Refract. Worldforum. 6 (2014) 89–97.
explosive spalling mechanism of high-performance concrete under fire exposure,
[10] P. Kalifa, F.D. Menneteau, D. Quenard, Spalling and pore pressure in HPC at high
Cement Concr. Res. 65 (2014) 64–75, https://doi.org/10.1016/j.
temperatures, Cement Concr. Res. 30 (2000) 1915–1927, https://doi.org/10.1016/
cemconres.2014.07.010.
S0008-8846(00)00384-7.
[27] D. Dauti, S. Dal Pont, M. Briffaut, B. Weber, Modeling of 3D moisture distribution
[11] J.C. Mindeguia, P. Pimienta, A. Noumowé, M. Kanema, Temperature, pore pressure
in heated concrete: from continuum towards mesoscopic approach, Int. J. Heat
and mass variation of concrete subjected to high temperature - experimental and
Mass Tran. 134 (2019) 1137–1152, https://doi.org/10.1016/j.
numerical discussion on spalling risk, Cement Concr. Res. 40 (2010) 477–487,
ijheatmasstransfer.2019.02.017.
https://doi.org/10.1016/j.cemconres.2009.10.011.
[28] Z.P. Bažant, M. Jirásek, Creep and Hygrothermal Effects in Concrete Structures,
[12] K.G. Fey, I. Riehl, R. Wulf, U. Gross, Pressure driven heat-up curves – a numerical
Springer, Dordrecht, The Netherlands, 2018, https://doi.org/10.1007/978-94-
and experimental investigation, Int. J. Therm. Sci. 113 (2017) 1–9, https://doi.
024-1138-6.
org/10.1016/j.ijthermalsci.2016.09.013.
[29] A. Millard, P. Pimienta, Modelling of Concrete Behaviour at High Temperature :
[13] R. Felicetti, F. Lo Monte, P. Pimienta, A new test method to study the influence of
State-Of-The-Art Report of the RILEM Technical Committee 227-HPB, Springer,
pore pressure on fracture behaviour of concrete during heating, Cement Concr.
2019. http://www.springer.com/series/8780.
Res. 94 (2017) 13–23, https://doi.org/10.1016/j.cemconres.2017.01.002.
[30] H.T. Vu, E. Tsotsas, Mass and heat transport models for analysis of the drying
[14] G. Choe, G. Kim, M. Yoon, E. Hwang, J. Nam, N. Guncunski, Effect of moisture
process in porous media: a review and numerical implementation, Int. J. Chem.
migration and water vapor pressure build-up with the heating rate on concrete
Eng. 2018 (2018), https://doi.org/10.1155/2018/9456418.
spalling type, Cement Concr. Res. 116 (2019) 1–10, https://doi.org/10.1016/j.
[31] H.T. Vu, E. Tsotsas, A framework and numerical solution of the drying process in
cemconres.2018.10.021.
porous media by using a continuous model, Int. J. Chem. Eng. 2019 (2019),
[15] S. Dal Pont, H. Colina, A. Dupas, A. Ehrlacher, An experimental relationship
https://doi.org/10.1155/2019/9043670.
between complete liquid saturation and violent damage in concrete submitted to
[32] M.H. Moreira, S. Dal Pont, R.F. Ausas, T.M. Cunha, A.P. Luz, V.C. Pandolfelli,
high temperature, Mag. Concr. Res. 57 (2005) 455–461, https://doi.org/10.1680/
Direct comparison of multi and single-phase models depicting the drying process of
macr.2005.57.8.455.
refractory castables, Open Ceram 6 (2021), https://doi.org/10.1016/j.
[16] R. Jansson, Fire Spalling of Concrete, Theoretical and Experimental Studies, KTH
oceram.2021.100111.
Royal Institute of Technology, 2013.
[33] D. Gawin, F. Pesavento, B.A. Schrefler, Modelling of hygro-thermal behaviour of
[17] Y. Li, D. Zhang, K.H. Tan, On measuring techniques of pore pressure in concrete at
concrete at high temperature with thermo-chemical and mechanical material
elevated temperature, Cem. Concr. Compos. 114 (2020), https://doi.org/10.1016/
degradation, Comput. Methods Appl. Mech. Eng. 192 (2003) 1731–1771, https://
j.cemconcomp.2020.103737.
doi.org/10.1016/S0045-7825(03)00200-7.
[18] S. Dal Pont, A. Ehrlacher, Numerical and experimental analysis of chemical
dehydration, heat and mass transfers in a concrete hollow cylinder submitted to

17
J. Juárez Trujillo et al. International Journal of Thermal Sciences 185 (2023) 108063

[34] C.T. Davie, C.J. Pearce, N. Bićanić, A fully generalised, coupled, multi-phase, J. Mater. Res. Technol. 9 (2020) 6001–6013, https://doi.org/10.1016/j.
hygro-thermo-mechanical model for concrete, Mater. Struct. Constr. 43 (2010) jmrt.2020.04.004.
13–33, https://doi.org/10.1617/s11527-010-9591-y. [51] M.D.M. Innocentini, P. Sepulveda, F.S. Ortega, Permeability, in: M. Scheffler,
[35] F. Meftah, S. Dal Pont, Staggered finite volume modeling of transport phenomena P. Colombo (Eds.), Cell. Ceram. Struct. Manuf. Prop. Appl., Wiley-VCH, 2005,
in porous materials with convective boundary conditions, Transport Porous Media pp. 313–341.
82 (2010) 275–298, https://doi.org/10.1007/s11242-009-9422-1. [52] P. Meunier, J.C. Mindeguia, P. Pimienta, Mass, temperature and pressure
[36] K.G. Fey, I. Riehl, R. Wulf, U. Gross, First heat-up of 1D multi-layer walls and 2D measurements during the dry out of refractory castables, in: Int. Colloq. Refract.
geometries consisting of refractory concrete, Int. J. Therm. Sci. 116 (2017) Refract. Metall., Aachen, Germany, 2008, pp. 95–98.
159–171, https://doi.org/10.1016/j.ijthermalsci.2016.11.021. [53] S. Dal Pont, B.A. Schrefler, A. Ehrlacher, Experimental and finite element analysis
[37] F. Meftah, S. Dal Pont, B.A. Schrefler, A three-dimensional staggered finite element of a hollow cylinder submitted to high temperatures, Mater. Struct. Constr. 38
approach for random parametric modeling of thermo-hygral coupled phenomena (2005) 681–690, https://doi.org/10.1617/14167.
in porous media, Int. J. Numer. Anal. Methods GeoMech. 36 (2012) 574–596, [54] J.M. Auvray, C. Zetterström, C. Wohrmeyer, H. Fryda, C. Parr, Eychenne-Baron,
https://doi.org/10.1002/nag.1017. Dry-out simulation of castables containing calcium aluminate cement under
[38] F.A. Cardoso, M.D.M. Innocentini, M.M. Akiyoshi, V.C. Pandolfelli, Effect of curing hydrothermal conditions, in: UNITECR 2013, Canada, Victoria, 2013, pp. 1–9.
conditions on the properties of ultra-low cement refractory castables, Refract. [55] D. Gawin, F. Pesavento, B.A. Schrefler, What physical phenomena can be neglected
Appl. News. 9 (2004) 12–16, https://doi.org/10.1016/S0360-1323(01)00078-6. when modelling concrete at high temperature? A comparative study. Part 1:
[39] A. Koehler, J. Neubauer, F. Goetz-Neunhoeffer, Phase changes during the drying of physical phenomena and mathematical model, Int. J. Solid Struct. 48 (2011)
calcium aluminate cement bond castables – the influence of curing and drying 1927–1944, https://doi.org/10.1016/j.ijsolstr.2011.03.004.
conditions, Cemento 7 (2022), 100020, https://doi.org/10.1016/j. [56] N. Rotschi, Empa - Communication - Concrete Explosion, 2019. https://www.
cement.2021.100020. empa.ch/web/s604/concrete-explosion. (Accessed 28 August 2021). accessed.
[40] D.L. Hipps, J.J. Brown, Internal pressure measurements for control of explosive [57] R. Zanettin, Empa - Communication - Fire Resistant Concrete, 2015. https://www.
spalling in refractory castables, Am. Ceram. Soc. Bull. 63 (1984) 905–910. empa.ch/web/s604/fire-resistant-concrete. (Accessed 28 August 2021). accessed.
[41] E. Tsotsas, E.U. Schlünder, Vacuum contact drying of mechanically agitated beds: [58] S. Dal Pont, B.A. Schrefler, A. Ehrlacher, Intrinsic permeability evolution in high
the influence of hygroscopic behaviour on the drying rate curve, Chem. Eng. temperature concrete: an experimental and numerical analysis, Transport Porous
Process 21 (1987) 199–208, https://doi.org/10.1016/0255-2701(87)80017-X. Media 60 (2005) 43–74, https://doi.org/10.1007/s11242-004-3252-y.
[42] M.D.M. Innocentini, F.A. Cardoso, M.M. Akiyoshi, V.C. Pandolfelli, Drying stages [59] M.D.M. Innocentini, M.G. Silva, B.A. Menegazzo, V.C. Pandolfelli, Permeability of
during the heating of high-alumina, ultra-low-cement refractory castables, J. Am. refractory castables at high temperatures, J. Am. Ceram. Soc. 84 (2001) 645–647,
Ceram. Soc. 86 (2003) 1146–1148, https://doi.org/10.1111/j.1151-2916.2003. https://doi.org/10.1111/j.1151-2916.2001.tb00715.x.
tb03438.x. [60] Z.X. Gong, A.S. Mujumdar, Development of drying schedules for one-side-heating
[43] Z.X. Gong, A.S. Mujumdar, The influence of an impermeable surface on pore steam drying of refractory concrete slabs based on a finite element model, J. Am. Ceram.
pressure during drying of refractory concrete slabs, Int. J. Heat Mass Tran. 38 Soc. 79 (1996) 1649–1658.
(1995) 1297–1303, https://doi.org/10.1016/0017-9310(94)00239-R. [61] D. Gawin, C.E. Majorana, B.A. Schrefler, Numerical analysis of hygro-thermal
[44] M.H. Moreira, R.F. Ausas, S. Dal Pont, P.I. Pelissari, A.P. Luz, V.C. Pandolfelli, behaviour and damage of concrete at high temperature, Mech. Cohesive-Frict.
Towards a single-phase mixed formulation of refractory castables and structural Mater. 4 (1999) 37–74, https://doi.org/10.1002/(SICI)1099-1484, 199901)4:
concrete at high temperatures, Int. J. Heat Mass Tran. 171 (2021), https://doi.org/ 1<37::AID-CFM58>3.0.CO;2-S.
10.1016/j.ijheatmasstransfer.2021.121064. [62] V. Baroghel-Bouny, Water vapour sorption experiments on hardened cementitious
[45] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Taylor and Francis Inc., materials. Part I: essential tool for analysis of hygral behaviour and its relation to
1980, https://doi.org/10.1201/9781482234213-6. pore structure, Cement Concr. Res. 37 (2007) 414–437, https://doi.org/10.1016/j.
[46] J.J. Trujillo, Contributions for the determination of monolithic refractories drying cemconres.2006.11.019.
schedules aided by experiments and numerical simulation, Universidade de São [63] C.T. Davie, C.J. Pearce, K. Kukla, N. Bićanić, Modelling of transport processes in
Paulo (2020), https://doi.org/10.11606/T.97.2020.tde-05082021-175041. concrete exposed to elevated temperatures – an alternative formulation for
[47] S. Sungsoontorn, P. Rattanadecho, W. Pakdee, One-dimensional model of heat and sorption isotherms, Cement Concr. Res. 106 (2018) 144–154, https://doi.org/
mass transports and pressure built up in unsaturated porous materials subjected to 10.1016/j.cemconres.2018.01.012.
microwave energy, Dry. Technol. 29 (2011) 189–204, https://doi.org/10.1080/ [64] D. Dauti, S. Dal Pont, B. Weber, M. Briffaut, N. Toropovs, M. Wyrzykowski,
07373937.2010.483029. G. Sciumé, Modeling concrete exposed to high temperature: impact of dehydration
[48] Z.P. Bažant, G. Zi, Decontamination of radionuclides from concrete by microwave and retention curves on moisture migration, Int. J. Numer. Anal. Methods
heating. II: computations, J. Eng. Mech. 129 (2003) 777–792, https://doi.org/ GeoMech. 42 (2018) 1–15, https://doi.org/10.1002/nag.2802.
10.1061/(asce)0733-9399, 2003)129:7(785. [65] S. Dal Pont, S. Durand, B.A. Schrefler, A multiphase thermo-hydro-mechanical
[49] J.A. De Castro, H. Nogami, J.I. Yagi, Three-dimensional multiphase mathematical model for concrete at high temperatures-Finite element implementation and
modeling of the blast furnace based on the multifluid model, ISIJ Int. 42 (2002) validation under LOCA load, Nucl. Eng. Des. 237 (2007) 2137–2150, https://doi.
44–52, https://doi.org/10.2355/isijinternational.42.44. org/10.1016/j.nucengdes.2007.03.047.
[50] J.A. de Castro, L.M. da Silva, G.A. de Medeiros, E.M. de Oliveira, H. Nogami, [66] M. Mainguy, O. Coussy, V. Baroghel-Bouny, Role of air pressure in drying of
Analysis of a compact iron ore sintering process based on agglomerated biochar weakly permeable materials, J. Eng. Mech. 127 (2001) 582–592, https://doi.org/
and gaseous fuels using a 3D multiphase multicomponent mathematical model, 10.1061/(asce)0733-9399 (2001)127:6(582).

18

You might also like