Upload
independent
View
1
Download
0
Embed Size (px)
Citation preview
Computers and Geotechnics 31 (2004) 237–250
www.elsevier.com/locate/compgeo
Multiphase flow in deforming porous material
B.A. Schrefler *, F. Pesavento
Department of Structural and Transportation Engineering, University of Padova, via F. Marzolo 9, 35131 Padova, Italy
Abstract
Models for thermo-hydro-mechanical behaviour of saturated–unsaturated porous media are reviewed. The necessary balance
equations are derived using averaging theories. Constitutive equations are obtained using the Coleman–Noll procedure and ther-
modynamic equations for the model closure are introduced. A simplified form of the governing equations is then solved numerically
and the numerical properties are discussed. An example dealing with behaviour of concrete structures during tunnel fires concludes
the paper. The heat and mass transfer calculations in the tunnel needed as the input for the multiphase concrete model are also
shown. The behaviour of concrete under such situations, where very high temperatures are reached, can be satisfactorily simulated
only with an approach of the type presented here.
� 2004 Elsevier Ltd. All rights reserved.
Keywords: Multiphase; Interfaces; Concrete; High temperature; Damage; Fire; Tunnel; Radiation
1. Introduction
Multiphase porous media, i.e. porous media wherethe pores are filled by more than one fluid, are treated
here within the framework of averaging theories. In
particular the approach developed by Hassanizadeh
and Gray [1–5] is used. The isothermal case was shown
in Schrefler [6]. Here the approach is extended to non-
isothermal situations. It is recalled that interfaces be-
tween the constituents with their thermodynamic
properties are taken into account. In fact, fluids in aporous medium, such as gas and water, will remain
immiscible only if there are interfaces with non-zero
surface tension. If surface tension is zero, then capillary
pressure is zero too which means that the fluid pressure
will be equal all the time. Actually only averaging
theories such as that used here include explicitly in-
terfacial properties. As far as the constitutive rela-
tionships are concerned, limits to their form areobtained by systematic exploitation of the entropy in-
equality, following the procedure of Coleman and Noll
[7]. Particular attention will be paid to near equilibrium
results, which among others yield the well known laws
* Corresponding author. Tel.: +39-49-827-5611; fax: +39-49-827-
5604.
E-mail address: [email protected] (B.A. Schrefler).
0266-352X/$ - see front matter � 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2004.01.005
of Darcy, Fick and Fourier, augmented by the con-
tributions of the interfaces. From the general mathe-
matical model a simplified one is extracted, which willbe solved numerically. Particular attention is focused
here on the boundary conditions: for a realistic simu-
lation of heat and mass transfer problems in deforming
porous media boundary conditions of the third type
are needed which include convective mass transfer and
convective and radiative heat transfer. The theoretical
developments in this paper are necessarily short. For a
full development the reader is referred to the papers byHassanizadeh and Gray [1–4], to Lewis and Schrefler
[8] and to Schrefler [6,9]. An example dealing with
behaviour of concrete structures during tunnel fires
concludes the paper. The heat and mass transfer cal-
culations in the tunnel needed as the input for the
multiphase concrete model are also shown. The be-
haviour of concrete under such situations, where very
high temperatures are reached, can be satisfactorilysimulated only with an approach of the type presented
here.
2. Macroscopic balance equations
For sake of brevity the microscopic balance equa-
tions for the constituents of the porous medium are
sw sg
238 B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250
omitted here as well as the kinematics. They can be
found e.g. in Lewis and Schrefler [8]. In this section the
macroscopic balance equations for mass, linear mo-
mentum and energy as well as the entropy inequality
are given, which have been obtained for the bulkmaterial of the phases and for the interfaces by sys-
tematically applying the averaging procedures to the
microscopic balance equations as outlined by Hassan-
izadeh and Gray [5]. The balance equations listed be-
low have been specialised for a deforming porous
material, where heat transfer and flow of water (liquid
and vapour) and of dry air are taking place, Lewis and
Schrefler [8]. The constituents are assumed to be im-miscible except for dry air and vapour, and chemically
non-reacting. The mixture of dry air and vapour
(moist air) will be simply called gas in the following.
All fluids are in contact with the solid phase. Disso-
lution of air in water is here neglected. Stress is defined
as tension positive for the solid phase, while pore
pressure is defined as compressive positive for the
fluids. In the averaging procedure volume density pa-rameters gp (volume fractions) appear which are
expressed in terms of commonly used variables in
multiphase flow.
For solid phase, gs ¼ 1� n where n ¼ dvw þ dvgð Þ=dvis porosity and dvp is the volume of constituent p within
a R.E.V. (representative elementary volume); for water
gw ¼ nSw, where Sw ¼ dvw= dvw þ dvgð Þ is the degree of
water saturation and for gas gg ¼ nSg with Sg ¼dvg= dvw þ dvgð Þ the degree of gas saturation. It follows
immediately that Sw þ Sg ¼ 1.
There appears also the specific surfaces of the inter-
faces aab, where the Greek letters refer to the bulk
phases involved. The inclusion of interface phenomena
which at first sight appear to be of secondary interest,
allow to treat the dependence of phase properties on
interface properties. At macroscale the system is mod-elled as the superposition of six continua: three phases
and three interfaces. At every spatial point average or
macroscopic properties are defined for each continuum
and the continua interact and exchange properties. Two
sets of balance equations are needed. One set is for the
bulk phases, the second is for the interfaces. The equa-
tions are listed next:
for solid
Ds 1� nð Þqs
Dtþ 1ð � nÞqs div vs ¼ esgs þ esws; ð1Þ
for liquid water
DwnSwqw
Dtþ nSwqw div vw ¼ ewgw þ ewsw; ð2Þ
for gas
DgnSgqg
Dtþ nSgqg div vg ¼ egwg þ egsg: ð3Þ
The mass source terms on the r.h.s. of Eqs. (1)–(3)
correspond to exchange of mass with interfaces sepa-
rating individual phases (phase changes) and couple
these equations with the corresponding balance equa-
tions written for the interfaces. These last ones may bewritten as
DabaabCab
Dtþ aabCab divwab ¼ �eaab � ebab þ eabwgs: ð4Þ
The last term in Eq. (4) describes mass exchange of the
interfaces with their contact line. Since we have three
phases composing the medium, there is only one contact
line. This contact line does not have thermodynamic
properties.The momentum balance equations for the bulk pha-
ses are
for solid
1ð � nÞqs Dsvs
Dt� div 1ðð � nÞtsÞ � 1ð � nÞqsg ¼ Ts
sg þ Tssw;
ð5Þ
for water
nSwqw Dwvw
Dt� div nSwtwð Þ � nSwð Þqwg ¼ Tw
wg þ Twws; ð6Þ
for gas
nSgqg Dgvg
Dt� div nSgtg
� �� nSg� �
qgg ¼ Tggw þ Tg
gs; ð7Þ
where ta is the partial stress tensor which is symmetric.
The r.h.s. terms in Eqs. (5)–(7) describe supply of
momentum from the interfaces, i.e. related to phase
changes. Analogous balance equations can be written
for the momentum of the three interfaces:
aabCab Dabwab
Dt� div aabsab
� �� aabCabgab
¼ � Taab
�þ eaabv
a;s�þ Tb
ab
�þ ebabv
b;s�
þ eaab�
þ ebab�wab;s þ sabwgs; ð8Þ
where sab is the surface stress tensor, which is also
symmetric.The last r.h.s. term corresponds to momentum supply
from the contact line �wgs� to the ab interface.
The energy balance equation for the bulk phases may
be written as follows
for solid
1ð � nÞqs DsEs
Dt� 1ð � nÞts : grad vs � div 1ðð � nÞqsÞ
� 1ð � nÞqshs ¼ Qs þ Qs ; ð9Þ
B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250 239
for water
nSwqw DwEw
Dt� nSwtw : grad vw � div nSwqwð Þ � nSwqwhw
¼ Qwws þ Qw
wg; ð10Þ
for gas
nSgqg DgEg
Dt� nSgtg : grad v
g � div nSgqg� �
� nSgqghg
¼ Qggs þ Qg
gw: ð11Þ
The source terms in Eqs. (9)–(11) describe supply of
heat to bulk phase from the interfaces, related to phase
changes. The energy balance equations for the three
interfaces read
aabCabDabEab
Dt�aabsab : gradwab�div aabqab
� ��aabCabhab
¼� Qaab
hþ Ta
ab � va;abþ eaab Ea;ab�
þ1=2 va;ab� �2�i
� Qbab
hþ Tb
ab � vb;abþ ebab Eb;ab�
þ1=2 vb;ab� �2�iþ Qab
wgs;
ð12Þ
where Ea;ab ¼ Ea � Eab.
The terms in square brackets in Eq. (12) describe theenergy supply from the bulk phase to the interface, energy
associated with momentum supply and energy related to
mass supply because of phase changes. The last r.h.s. term
is supply of heat to the interface from the contact line.
We assume that entropy fluxes are due solely to heat
input and the entropy external source terms are due only
to external energy sources. Thus, the entropy balance
may be expressed for the bulk phases as followsfor solid
1ð � nÞqs Dsks
Dt� div 1ð
�� nÞ q
s
hs
�� 1ð � nÞqs h
s
hs
¼ Ussg þ Us
sw þ Ks; ð13Þ
for water
nSwqw Dwkw
Dt� div nSw
qw
hw
� �� nSwqw h
w
hw
¼ Uwwg þ Uw
ws þ Kw; ð14Þ
for gas
nSgqg Dgkg
Dt� div nSg
qg
hg
� �� nSgqg h
g
hg
¼ Uggw þ Ug
gs þ Kg: ð15Þ
The two first terms in r.h.s. of Eqs. (13)–(15) describe
the entropy supply to the bulk phases from the inter-
faces, while the last one is the rate of net production of
entropy in the bulk phase.
Similarly, for the interfaces we have the following
three entropy balance equations
aabCab Dabkab
Dt� div aab
qab
hab
� �� aabCab h
ab
hab
¼ � Uaab
�þ eaabk
a;ab�� Ub
ab
�þ ebabk
b;ab�þ Uab
wgs þKab:
ð16Þ
The terms in parentheses in the r.h.s. of Eq. (16) are
supply of entropy from the interfaces and resulting from
mass supply (phase change), the last but one accounts
for entropy supply to the interface from the contact lineand the last one is the rate of net production of entropy
in the interface.
The terms related to exchange of mass, momentum,
energy and entropy between interfaces via the contact
lines must satisfy some restrictions, because the contact
lines do not possess any thermodynamic properties as
already stated. Thus, the following relations holdXab
eabwgs ¼ 0;
Xab
sabwgs
�þ eabwgsw
ab�¼ 0;
Xab
Qabwgs
hþ sabwgs � wab þ eabwgs Eab
�þ 1=2 wab
� �2�i ¼ 0;
Xab
Uabwgs
�þ eabwgsk
ab�¼ 0:
ð17Þ
3. The second law of thermodynamics
The balance laws must be supplemented with the
second law of thermodynamics, which states that forany process the rate of net entropy production must be
non-negative
K ¼ Ks þ Kw þ Kg þX
ab¼gs;gw;sw
Kab P 0; ð18Þ
where Kp is the rate of net production of entropy in the
bulk phases and interfaces and ab ¼ gs, gw, sw refer to
the interfaces between gas and solid, gas and water and
water and solid, respectively.
After introducing the balance laws into Eq. (18) and
substituting the internal energy by the Helmholtz free
energy, which is defined for the bulk phases as
Aa ¼ Ea � haka; a ¼ s;w; g ð19Þand for the interfaces as
Aab ¼ Eab � habkab; ab ¼ gw;ws; gs ð20Þan appropriate form of the entropy inequality (18) is
obtained [5]. The balance equations of mass, momentum
and energy must be supplemented by constitutive
240 B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250
equations describing the behaviour of individual phases.
In total there are 30 equations and the same number of
independent variables can be chosen as basic indepen-
dent fields. These field quantities or/and combinations
and space and time derivatives of them that are objectivecan enter as independent constitutive variables. Their
choice should be based on the expected behaviour of the
medium, as well as they should account macroscopically
for the microstructure due to the interfaces. For this last
reason for instance volume fractions, their gradients, the
specific surfaces of the interfaces and their gradients
may be added to the list of primary variables. This
augments accordingly the list of dependent variables toeliminate the ensuing equation deficit. The independent
variables are function of time and space. The list of in-
dependent variables chosen here is qa, Cab, va;s, wab;s, Ea,
ha, gradha, hab, gradhab, n, gradn, Sa, gradSa, aab,gradaab.
The remaining variables appearing in the balance
equations must be expressed in terms of the primary
unknowns and their derivatives. The equation deficit iseliminated by also requiring constitutive forms for some
of the time derivatives, (here of porosity, the degree of
water saturation and of specific surfaces of the interfaces
[5]) and by thermodynamic equilibrium equations. For
the list of dependent variables, for which constitutive
relations are needed, see [6].
Helmholtz free energy for the bulk phases is assumed
to have the following functional form which is particu-larly simple, but sufficient for our purpose
Aw ¼ Aw qw; hw; Swð Þ; ð21Þ
Ag ¼ Ag qg; hg; Sg� �
; ð22Þ
As ¼ As qs; hs;Es; Swð Þ; ð23Þ
and for the interfaces
Aab ¼ Aab Cab; hab; aab; Sw� �
;
where ab ¼ gw;ws; gs: ð24Þ
All remaining dependent variables are allowed to
depend on the complete set of independent variables
given above. Note that the assumptions (21)–(24) departfrom the principle of equipresence.
According to the principle of admissibility, the con-
stitutive postulates relating dependent to independent
variables must not violate the balance laws and the en-
tropy inequality. These requirements are satisfied using
the procedure proposed by Coleman and Noll [7]. The
rather lengthy transformations of the entropy inequality
necessary are omitted here. They can be found in [5,9].Following the procedure of Coleman–Noll the fol-
lowing non-equilibrium results are obtained:
ka ¼ � oAa
oha; a ¼ w; g; s; ð25Þ
kab ¼ � oAab
ohab; ab ¼ gw;ws; gs; ð26Þ
tw ¼ �pwI; ð27Þ
tg ¼ �pgI; ð28Þ
ts ¼ tse � psI; ð29Þ
where tse ¼ qs Fsð ÞT � oAs
oEs � Fs, is the effective stress tensor
of the solid phase, and ps qs; hs;Es; Swð Þ ¼ qsð Þ2 oAs
oqs the
thermodynamic pressure of the solid phase,
sab ¼ cabI; ð30Þwhere cab is the surface tension.
Some additional information can be obtained, whenexamining the system under consideration at equilib-
rium state where there is no relative movement of phases
and interfaces, degree of saturation with water (thus also
with gas) and porosity are constant, the phases and in-
terfaces have the same uniform temperature. At these
conditions the total rate of entropy production K equals
to zero, i.e. reaches its minimum value. Thus the nec-
essary and sufficient conditions for K to be at minimumat equilibrium are
oKozk
� �eq
¼ 0; k ¼ 1; . . . ; 46 ð31Þ
and
o2Kozk ozm
� �eq
be positive semi-definite;
k;m ¼ 1; . . . ; 46: ð32Þ
Application of restriction (31) to the entropy in-
equality allows to obtain the following main relationsvalid at equilibrium
psð Þeq ¼ Sgpg þ Swpw; ð33Þ
pc ¼ pgð � pwÞeq; ð34Þ
qsð Þeq ¼ qwð Þeq ¼ qgð Þeq ¼ 0; ð35Þ
qgwð Þeq ¼ qwsð Þeq ¼ qgsð Þeq ¼ 0: ð36Þ
The definition of capillary pressure, Eq. (34), shows
that pc depends on independent variables as follows
pc ¼ pc Sw; n; awg; aws; h; qw; qg;Cwg;Cws;Cgsð Þ: ð37Þ
At equilibrium, pc is given by Eq. (34). The capillary
pressure is commonly assumed to be a function of
Sw; qw, and qg; n and h. Eq. (37) shows that also inter-
facial areas and surface densities play a role. Cases
where surface densities may change and affect capillary
pressure are when surfactants are present and changethe character of the interfaces.
B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250 241
Eqs. (35) and (36) indicate that at equilibrium there is
no heat transfer within the phases and interfaces, what is
true for a wide class of practical problems, where a state
of ‘‘local thermal equilibrium’’ or, more general, ‘‘local
thermodynamic equilibrium’’ is assumed. Finally, atequilibrium the Gibbs free energy per unit mass for each
phase and interface will be equal.
For subsequent discussions attention is restricted
more to the mechanical aspect of the problem and less to
its thermal aspect. It is assumed that temperature dif-
ferences between phases at a macroscopic point are
negligible also near equilibrium which is acceptable for a
large class of problems. For these cases, a state of localthermal equilibrium prevails such that all phases and
interfaces will have the same temperature h at a point,
which although may still vary in space.
If the system is considered ‘‘near’’ equilibrium some
additional simplifications may be obtained, as far as
the constitutive functions are concerned. In such situ-
ations a linear dependence of constitutive functions
(describing thermodynamic flows) on primary variables(thermodynamic potentials) may be postulated. Some
of these linear relations are widely used in practice, like
for example Darcy�s law or Fick�s law for fluid flow
and Fourier�s law. For a complete description of the
procedure of linearization see [9]. The fact that the
constitutive assumptions are to be linearized does not
influence any conclusion that may be drawn concerning
the equilibrium state of the system (i.e. only the dy-namic state will be influenced, Gray and Hassanizadeh
[10,11]).
As pointed out in Section 1, for the model closure
state equations are still needed. Moist air (gas) in the
pore system is assumed to be a perfect mixture of two
ideal gases, dry air and water vapour. The equation of
perfect gas is hence valid
pga ¼ qgahR=Ma pgw ¼ qgwhR=Mw; ð38Þ
qg ¼ qga þ qgw; pg ¼ pga þ pgw;
Mg ¼qgw
qg
1
Mw
�þ qga
qg
1
Ma
��1
;ð39Þ
where R is the universal gas constant and Ma the molar
mass of the constituents.
The equation of state of water is of the formpw ¼ pw qw; T ; Swð Þ [10,11].
4. Heat and mass transfer in deforming partially saturated
geomaterials
From the equations listed in the previous sections a
model for heat and mass transfer in partially saturatedgeomaterials will now be established. The system of
governing equations of this model will then be solved
numerically. It is assumed that the system is near equi-
librium as explained in the previous sections. Compared
to the general theory outlined in the previous sections
the model is simple, for instance it does not consider the
balance equations for the interfaces. However comparedto the models currently in use in the geomechanics
community, it is rather advanced. It may be considered
as a first step in the effort to obtain such thermody-
namically based numerical models. Its complexity can
then be augmented at will following the equations given
in the theoretical section.
4.1. Simplified field equations
The model is built in the following way. From the
linearised equations Fick�s law and Fourier�s law are
chosen neglecting interfacial terms. As far as Darcy�sequation is concerned, the linear momentum balance
equation for the fluid phases is used directly with an
appropriate constitutive assumption for the momentum
exchange term in the form
gaqa ta ¼ �Ragavsa þ pa gradga; ð40Þwhere Ra is given by
Ka ¼ ga Raa
� ��1 ¼ k
lpqa; ga; Tð Þ: ð41Þ
These equations are supplemented by the mass bal-
ance equations for solid (1), water (2), vapour and gas
(3), the sum of the linear momentum balance equation
for the constituents (5)–(7), and the sum of the energy
balance equation for the constituents (9)–(11), which has
been transformed into an enthalpy balance (see [8]).
Further, for the fluid stresses, the solid pressure and thecapillary pressure the respective equilibrium values are
taken, Eqs. (27), (28), (33) and (34). The effective stress
is assumed in the form:
t ¼ 1ð � nÞtse � I Sgpg�
þ Swpw�: ð42Þ
Finally the state equations for water and gas, Eqs.
(38) and (39) of Section 3 are used and the capillary
pressure saturation relationship (37) (with a much sim-
pler functional dependence suggested by experimentalobservations pc ¼ pc Sw; hð Þ). Actually, the inverse of this
function is used because of the choice of the primary
variables.
Vapour pressure has to be expressed as function of
the relevant primary variables capillary pressure and
temperature. This is obtained by means of the Kelvin–
Laplace equation, which is a thermodynamic relation at
equilibrium (see [12,13]).
pgw
pgws¼ exp
pcMw
qwRh
� �; ð43Þ
where the vapour saturation pressure pgws is obtained
from the Clausius–Clapeyron equation.
242 B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250
Some rearrangements of these equations (see [8]) yield
the following field equations. The macroscopic mass
balance equations are:
for the solid phase
1� nqs
Dsqs
Dt�Dsn
Dtþ 1ð � nÞdiv vs ¼ 0; ð44Þ
for dry air
� nDsSwDt
�bs 1ð � nÞSgDsTDt
þ Sgdivvsþ Sgn
qga
Ds
DtMa
hRpga
� �
� 1
qgadiv qgMaMw
M2g
Dg gradpga
pg
� �" #þ 1
qgadiv nSgqgavgs
� �¼ 0; ð45Þ
for the water species, i.e. liquid water and vapour to-
gether
n qwð � qgwÞDsSwDt
� bswg
DsTDt
þ qgwSg�
þ qwSw�
� div vs þ nqwSwKw
Dspw
Dtþ Sgn
Ds
DtMw
hRpgw
� �
� div qg MaMw
M2g
Dg gradpgw
pg
� �" #
þ div qgw kkrg
lg½
� gradpg þ qg gð � as � agsÞ�
�
þ div qw kkrw
lw½
� gradpw þ qw gð � as � awsÞ�
�¼ 0; ð46Þ
where
bswg ¼ bs 1ð � nÞ Sgqgw�
þ qwSw�þ nbwq
wSw; ð47Þ
with bp the thermal expansion coefficients.
The linear momentum balance equation for fluids is
gpvps ¼ kkrp
l½ � gradpp þ qp gð � as � apsÞ�; ð48Þ
and for the multiphase medium
� qas � nSwqw aws½ þ vws � grad vw� � nSgqg
� ags½ þ vgs � grad vg� þ div tþ qg ¼ 0: ð49Þ
Finally the enthalpy balance for the multiphase
medium may be written as
qCp
� �eff
oTot
þ qwCwp v
w�
þ qgCgpv
g�� gradT
� div veff gradTð Þ ¼ � _mDHvap; ð50Þ
where
qCp
� �eff
¼ qsCsp þ qwC
wp þ qgC
gp ;
veff ¼ vs þ vw þ vg;
DHvap ¼ H gw � Hw;
ð51Þ
with Hp the specific enthalpy and Cpp the heat capacity.
4.2. Initial and boundary conditions
The initial conditions specify the full fields of gas
pressure, capillary or water pressure, temperature, dis-
placements and velocities
pg ¼ pg0; pc ¼ pc0; T ¼ T0; u ¼ u0; _u ¼ _u0; at t ¼ t0:
ð52ÞThe boundary conditions are formulated for vol-
ume-averaged quantities, which are continuous fields.
Thus they concern an arbitrary boundary of the anal-
ysed space domain which corresponds to the ‘‘macro-
scopic’’ external surface of the porous body coinciding
with its fixed boundary considered during volumeaveraging.
The boundary conditions can be imposed values on
Cp or fluxes on Cqp, where the boundary C ¼ Cp [ Cq
p.
The imposed values on the boundary for gas pressure,
capillary or water pressure, temperature and displace-
ments are
pg ¼ pg on Cg; pc ¼ pc on Cc;
T ¼ T on CT ; u ¼ u on Cu:ð53Þ
The volume averaged flux boundary conditions for
water species and dry air mass balance equations and
the energy conservation equation, to be imposed at the
interface between the porous media and the surrounding
fluid are as follows
qga�vg�
� qg�vwv�� n ¼ qga on Cq
g;
qwv�vg�
þ qw�vw þ qg�vwv�� n
¼ bc qwv�
� qwv1�þ qwv þ qw on Cq
c ;
� qw�vwDhvap�
� veff rT�� n
¼ ac Tð � T1Þ þ r0e T 4�
� T 41�þ qT on Cq
T ; ð54Þ
where n is the unit vector, perpendicular to the surface
of the porous medium, pointing toward the surrounding
gas, qgw1 and T1 are, respectively, the mass concentration
of water vapour and temperature in the undisturbed gas
phase distant from the interface, ac, bc, r0 and e are
convective heat and mass transfer coefficients, the
Boltzmann constant and the emissivity, while qga, qwv, qw
and qT are the imposed dry air flux, imposed vapour
flux, imposed liquid flux and imposed heat flux,
respectively.
The convective term on the r.h.s. of the last of
Eq. (54) corresponds to Newton�s law of cooling and
describes the conditions occurring in most practical sit-
uations at the interface between a porous medium and
the surrounding fluid (air in this case).The traction boundary conditions for the displace-
ment field are
B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250 243
t � n ¼ ^t on Cqu; ð55Þ
where ^t is the imposed traction.
4.3. Numerical solution
The governing equations of the model are discretisedin space by means of the finite element method [14,15].
The unknown variables are expressed in terms of their
nodal values as,
pg tð Þ ¼ Np �pg tð Þ; pc tð Þ ¼ Np �p
c tð Þ;T tð Þ ¼ Nt
�T tð Þ; u tð Þ ¼ Nu �u tð Þ:ð56Þ
The variational or weak form of the model equations,
including also the ones required to complete the model,
was obtained in [8,16] by means of Galerkin�s method(weighted residuals), and can be written in the following
concise discretised matrix form,
Cij xð Þ oxot
þ Kij xð Þx ¼ f i xð Þ; ð57Þ
with
Kij ¼
Kgg Kgc Kgt 0
Kcg Kcc Kct 0
Ktg Ktc Ktt 0
Kug Kuc Kut Kuu
26664
37775;
Cij ¼
Cgg Cgc Cgt Cgu
0 Ccc Cct Ccu
0 Ctc Ctt Ctu
0 0 0 0
26664
37775; f i ¼
fg
fc
ft
fu
8>>><>>>:
9>>>=>>>;: ð58Þ
The time discretization is accomplished through afully implicit finite difference scheme (backward differ-
ence),
Wi xnþ1ð Þ ¼ Cij xnþ1ð Þ xnþ1 � xn
Dtþ Kij xnþ1ð Þxnþ1
� f i xnþ1ð Þ ¼ 0; ð59Þ
where superscript �i� (i ¼ g; c; t; u) denotes the state var-
iable, n is the time step number and Dt the time step
length.
The equation set (59) is solved by means of a
monolithic Newton–Raphson type iterative procedure[8,16]:
Wi xknþ1
� �¼ � oWi
ox
����Xknþ1
Dxknþ1;
xkþ1nþ1 ¼ xk
nþ1 þ Dxknþ1; ð60Þ
where k is the iteration index and the Jacobian matrix is
defined by
oWi
ox
����xknþ1
¼
oWg
o�pgoWg
o�pcoWg
o�T
oWg
o�uoWc
o�pgoWc
o�pcoWc
o�T
oWc
o�uoWt
o�pgoWt
o�pcoWt
o�T
oWt
o�uoWu
o�pgoWu
o�pcoWu
o�T
oWu
o�u
266666666664
377777777775
����������������x¼xk
nþ1
: ð61Þ
4.4. Constitutive model for thermo-chemical and mechan-
ical deterioration of concrete at high temperature
During heating, concrete at high temperature is ex-
posed to complicated physical and chemical transfor-
mations [17,18], causing changes of its inner structure,
what has also a sensible influence on the materialproperties. From a practical point of view, among the
most important macroscopic consequences of these
processes are concrete dehydration and crack develop-
ment, resulting in a significant decrease of mechanical
properties of concrete at high temperatures. The mate-
rial stress – strain behaviour in such conditions is highly
non-linear and depends not only on the history of me-
chanical loading during heating, but also on the tem-perature and dehydration degree [18].
For these reasons a parameter describing the degree
of the thermally induced material deterioration process
advancement has been introduced, similarly as done by
Gerard et al. [19] and Nechnech et al. [20]. It is called
thermo-chemical damage, V , because it accounts for
changes of material stiffness, both due to thermally in-
duced micro-cracks, caused mainly by stresses at micro-and meso-level, (e.g. resulting from different thermal
expansion coefficients of cement paste and aggregate,
and from local increase of dehydration products� vol-ume), and due to decrease of concrete strength proper-
ties caused by the dehydration process (thus related to
the Cdehydr value).
The thermo-chemical damage parameter, V , is de-
fined in terms of the experimentally determined evolu-tion of Young�s modulus of mechanically undamaged
material (i.e. heated to a given temperature, without any
additional mechanical load), E0, expressed as a function
of temperature,
V ¼ 1� E0ðT ÞE0ðTaÞ
; ð62Þ
where Ta ¼ 20 �C is room temperature.
Mechanical damage of concrete is considered fol-
lowing the scalar isotropic model by Mazars [21,22]. In
this model, the damaged material at given temperature,
T , is supposed to behave elastically and to remain iso-
tropic. Its Young�s modulus at this temperature, EðT Þ,can be obtained from the value of mechanically un-
damaged material at the same temperature, E0ðT Þ
244 B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250
[21,22], and mechanical damage parameter, d, being a
measure of cracks� volume density in the material,
EðT Þ ¼ 1ð � dÞE0ðT Þ: ð63ÞA total effect of the mechanical and thermo-chemical
damages, to which the material is exposed at the same
time, is multiplicative, i.e. the total damage parameter,
D, is defined by the following formula,
D ¼ 1� EðT ÞE0ðTaÞ
¼ 1� EðT ÞE0ðT Þ
E0ðT ÞE0ðTaÞ
¼ 1� 1ð � dÞ � 1ð � V Þ: ð64Þ
Therefore, the classical effective stress concept, is
modified to take into account both the mechanical and
thermo-chemical damage, so a further reduction of re-
sistant section area due to thermo-chemical degradation
is added to that caused by the mechanical damage, i.e.
the section reduction by cracking:
~r ¼ rS~S¼ r
1� dð Þ 1� Vð Þ ; ð65Þ
where S and ~S mean total and resistant area of thedamaged material, r is the tensor of nominal stress and~r the tensor of ‘‘modified’’ effective stress (in the sense of
Mazars [21,22]). This formulation of damage model is
implemented by means of a non-local technique. For
further details, see [23,24].
5. Behaviour of concrete in tunnel fires
We extend here the procedure outlined in Sections
2–4 to deal with the behaviour of concrete walls and
ceilings in case of tunnel fires. The recent fires which
occurred in major European tunnels (Channel [25,26],
Mont-Blanc, Great Belt Link, Tauern, St.Gotthard,
etc.) emphasise the serious hazards they represent and
the impact they may have on human beings, economyand repair costs. Extensive and heavy damage in the
concrete elements were observed. The structural damage
can be attributed to two main factors, namely, spalling
of concrete and excessive temperatures attained in both
concrete and steel components leading to serious loss of
load bearing capacity. The temperature field inside the
wall depends on the incident heat fluxes (convective and
radiative), related to location and type of fire, air-flowrate through the tunnel and smoke properties.
The numerical simulation of such complex situations
requires advanced numerical models which allow to
simulate the fire scenario in the tunnel and the real
physical behaviour of the concrete structure in such se-
vere conditions. The behaviour of concrete can be sat-
isfactorily simulated only with a multiphase porous
media approach such as that used in this paper. In case ofhigh temperatures the following mass transport mecha-
nisms take place in concrete: for capillary water we have
darcian type flow and the thermodynamic force is the
water pressure gradient; for adsorbed water there is dif-
fusive flow and the force is the water concentration
gradient. For chemically adsorbed water there is no
transport, but it acts as source or sink term. Water va-pour presents both darcian flow under gas pressure
gradient and diffusive flow under water vapour concen-
tration gradient. Similarly dry air presents darcian flow,
also under gas pressure gradient and diffusive flow under
dry air concentration gradient. All these phenomena can
be modelled by the equations shown in Sections 3 and 4.
Further there are phase changes of chemical and physical
nature: dehydration, evaporation and desorption whichare endothermic processes, while hydration, condensa-
tion and adsorption are exothermic ones. All these re-
quire appropriate exchange terms in the balance
equations, which have been introduced in [13,16,24,27].
As far as the fire scenario is concerned, this represents a
boundary condition for the concrete wall.
Prediction of the thermal field within a tunnel in
presence of fire is quite complex, since combustion givesway to a combination of heat transfer phenomena (by
convection, radiation and conduction) with strong in-
teractions among them. As well known, in the reaction
zone chemical energy is converted into internal energy of
the products thus leading to convective movements of
the fluid, as well as short wave radiation; at the same
time mutual radiation takes place among the walls and
to the other parts of the fluid, according to their tem-peratures and radiative properties. A complete proce-
dure for the computer simulation of such complex
phenomena is described in detail in [28]. Here, our at-
tention is focused on the combustion and thermal ra-
diation models which have been applied in the numerical
example presented in this work.
Nowadays, several combustion models are available
for fire simulation in enclosures: among them, the so-called ‘‘volumetric heat source (VHS)’’ [29], appears to
be quite simple when compared to other methods (e.g.
the ‘‘Eddy break-up’’ or the ‘‘presumed probability
density function (prePDF)’’ models [30]), specifically
developed to take into account some chemical aspects of
the reactions. VSH method, instead, does not consider
any chemical reaction and simply assumes a given spa-
tial and temporal distribution of the equivalent heatgeneration H per unit volume:
H ¼ f ðr; tÞ; ð66Þwhere r is coordinate vector, a t – time. Usually, H is
assumed to be constant throughout a given volume Vrepresenting the zone of reaction and therefore, being
QðtÞ the overall heat generation:
H ¼ QðtÞ=V : ð67ÞDespite of its simplicity, this model appears to be
reliable enough for enclosure fires, as shown by the
B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250 245
experimental validations reported in [29], and can
therefore be applied with satisfactory confidence.
Thermal radiation inside a tunnel with fire is quite
complex since participating media (including smoke,
soot, liquid droplets, ash and dust particles) are usu-ally present in a wide range of temperatures and
compositions. Again this presents a boundary condi-
tion for the structural problem as indicated in the third
of Eq. (54). The ‘‘Monte Carlo’’ statistical method is
used to obtain this flux [31]. In principle, a point in the
domain of the function is selected at random and,
sampling from the probability distributions that model
the physical process, the mapping of this to a pointin the co-domain is calculated. After many repetitions,
the correspondence between areas in the domain and
the co-domain may be described statistically with in-
creasing confidence.
In this regard, a computer model named SMOKE 2.1
has been specifically written for tunnel fires, as a de-
velopment of previous works in the field of radiative
heat exchanges with non-participating medium [32–34].Basically, the space enclosed by the tunnel as well as
its surfaces are discretized into a suitable number of
elements. For each of them, assuming the temperature
to be known, the radiative emission can be determined
as follows, being r0 the Stefan–Boltzmann coefficient.
For surface elements with area A, total emissivity e andabsolute temperature TS, the well known Stefan–Boltz-
mann�s law leads to the following expression for thepower emitted by the surface:
qr ¼ r0eAT 4S : ð68Þ
Instead, for fluid elements with volume V and
absolute temperature Tf the radiative emission can be
approximated as follows [35]:
qr � 4r0eVT 4f ; ð69Þ
where e is an equivalent emissivity depending on
temperature and chemical composition of the fluid, as
well as on the so-called ‘‘mean beam length’’ L of the
element:
e ¼ f ðTf ; L; cÞ; ð70Þwhere c means concentration vector describing compo-sition of gasses, taking into account the possible pres-
ence of solid and liquid particles.
According to Eckert [36], the mean beam length can
be approximated as
L � 3:6V =A: ð71ÞAssuming a lambertian behaviour for both gas and
surfaces, the application of the ‘‘Monte Carlo’’ ap-
proach to the problem under consideration can be car-
ried out by generating at the centre of each element(surfaces or volumes) a certain number N of rays with
randomly selected direction and the following initial
power qb:
qb ¼ qr=N : ð72ÞThen, each ray is traced throughout the tunnel in
order to locate its interception with a wall and the
resulting segment is subdivided into intervals of suit-able length Dx. At each increment Dx the power of the
ray is progressively reduced, because its energy is
partly absorbed and partly scattered. Being ðqbÞx the
ray power at a given location x, the absorbed com-
ponent ðqaÞDx, to be added to the heat generation of
the fluid, will be:
ðqaÞDx ¼ aDxðqbÞx; ð73Þwhere aDx is the absorption coefficient of the fluid over
Dx. Instead, the scattered component, to be assigned toone of several auxiliary radiant sources located along
the path, will be:
ðqaÞDx ¼ qDxðqbÞx; ð74Þwhere qDx is the scattering coefficient of the fluid over
Dx. For the application of this procedure both aDx andqDx must be known.
Unfortunately, information on scattering phenomena
in tunnel fires is limited and qDx must be determined by areasonable assumption.
Also the evaluation of aDx is quite complex, espe-
cially because the selective behaviour of most fluids
makes aDx not constant. However, in tunnel fires where
significant stratification usually occurs and the ab-
sorbing fluid is confined in the hot upper layer, an
estimation of aDx of the smoke can be obtained as
follows. For an uniform volume with ‘‘mean beamlength’’ L, the overall absorption coefficient aL is given
in [35] as a function of its chemical composition, its
temperature Tf and the temperature T0 of the emitting
source:
aL ¼ f ðTf ; T0; cÞ: ð75ÞTherefore, it can be shown that a reasonable esti-
mation of aDx, for the smoke with average temperature
at a given distance from the main source, is
aDx ¼ 1� ð1� aLþDxÞ=ð1� aLÞ; ð76Þ
where in this case L is the overall length of the ray path
across the smoke (starting from the main source).
When the ray reaches a wall, absorption and reflec-
tion will be treated in accordance with the surface ra-
diative properties and the reflected component will beemitted by an additional auxiliary source placed on the
wall. Then, the same process is applied to each auxiliary
source, until the rays leave the system or their powers
become negligible. By repeating this procedure several
times with different sets of rays initiated at the main
sources, the radiative behaviour of the system becomes
increasingly clear and the radiation absorbed by the
various elements (surfaces or volumes) can consequentlybe determined.
Fig. 2. Discretization used in 2D calculation of the top stressed section
of the tunnel (section of the fire). Points indicate the tunnel zones
considered in the Figs. 3–8.
Table 1
Properties for ordinary concrete at 20 �C
Parameter Symbol Value
Porosity n [)] 0.1
Intrinsic permeability K[m2] 10�18
Apparent density q [kg/m3] 2585
Specific heat C [J/kgK] 940
Thermal conductivity v [W/mK] 1.67
Young�s modulus E [GPa] 33
Poisson�s coefficient m [)] 0.20
Compression strength fc [MPa] 30
246 B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250
5.1. Case study
The model described in the previous paragraphs has
been applied to a fire scenario for which some experi-
mental measurements were available [37].Since the measurements already provided the tem-
peratures of the fluid in several points inside the tunnel,
Fig. 1, CFD computations were not necessary.
As far as thermal radiation was concerned, the fol-
lowing assumptions were made. The fluid was divided
into two layers, the lower one (up to 3.17 m from the
road) made of fresh air at ambient temperature, the
upper one (from 3.17 m up to the vault) consisting ofsmoke with 20% CO2, 10% H2O (percentages by vol-
ume) and a scattering coefficient (for short wave only)
qDx ¼ 0:006, with Dx ¼ 0:1 m.
Considering the plume of fire at a temperature
T0 ¼ 1200 �C and emitting a radiative power qr ¼ 2:5MW, the absorbed components at the walls (the quan-
tities absorbed by the fluid are of no use in this case)
were calculated with the assumed values of reflectanceq ¼ 0:25 for vault (of the dark grey colour) and q ¼ 0:15for the asphalt road pavement [38].
Additionally, also the mutual radiation between the
walls and smoke was calculated assuming the walls as
black bodies. Smoke was calculated assuming the walls
as black bodies. The convective heat transfer was esti-
mated from the correlations given in [39]. Heat transfer
was estimated from the correlations given in [39]. Heattransfer coefficient was equal to 27 W/m2 K in the zone
affected by the plume.
At this point the thermo-structural model was applied
and the temperatures and stresses were calculated (see
Fig. 2 for the discretization employed in the calculations).
The tunnel has been built of an ordinary concrete
whose properties are reported in Table 1.
It has been assumed that the fire acted with maximumenergy for a time span of 7.5 min maintaining stationary
Fig. 1. Schematic representation of the longitudinal and cross sections of the
the instrumented sections at 0, 20 and 50 m from the flames.
conditions for fire. In the real fire test, the firemen began
to extinguish the fire after 5 min. The extra 2.5 min have
been simulated to analyse a further evolution of the
hygro-thermo-mechanical behaviour of the concrete
vault, assuming that the heat fluxes are the same as
during the first 5 min of fire. The resulting distributions
of temperature, relative humidity, vapour pressure, as
tunnel, with the position of fire, smoke and maximum temperatures in
Fig. 4. Relative humidity distributions at time stations t ¼ 5 min and
t ¼ 7:5 min (a) and time histories at three points of tunnel located at
different heights above the pavement.
B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250 247
well as mechanical-, thermo-chemical- and total damage
at three locations: A, B, C of the central tunnel cross-
section (see Fig. 2), located at different heights above the
tunnel pavement, h ¼ 6:03, 5.15 and 1.5 m, for two time
stations, t ¼ 5 min (the end of the real fire test) andt ¼ 7:5 min (the end of the simulations), are shown in
Figs. 3(a), 4(a), 5(a), 6(a), 7(a) and 8(a). Because of the
short time of exposition and the total thermal power
involved during the test, only a superficial 6-cm layer of
the vault has been affected by the phenomenon and
shown in the graphs. In Figs. 3(b), 4(b), 5(b), 6(b), 7(b)
and 8(b), the time evolutions of the analysed physical
quantities in the points lying 8 mm from the vault�ssurface (except of the temperatures, where the surface
values are shown) in the locations A, B and C are pre-
sented. It is worth to underline that two of them, i.e. A
and B are situated in upper part of the tunnel, filled with
the smoke, what influenced significantly the heat radia-
tion and the resulting temperature distribution (Fig. 3).
It is possible to note the sharp desaturation process af-
fecting the layer of concrete close to the heated surface(Fig. 4) and the thermo-diffusion inside the pores of the
material. A significant increase of the vapour pressure in
this layer is observed, Fig. 5.
No visible vault deterioration has been observed
during the experimental test. However, the numerical
calculations showed that because of the temperatures
reached, a considerable thermo-chemical damage is
Fig. 5. Vapour pressure distributions at time stations t ¼ 5 min and
t ¼ 7:5 min (a) and time histories at three points of tunnel located at
different heights above the pavement.
Fig. 3. Temperature distributions at time stations t ¼ 5 min and t ¼ 7:5
min (a) and time histories at three points of tunnel located at different
heights above the pavement.
Fig. 6. Thermo-chemical damage parameter distributions at time sta-
tions t ¼ 5 min and t ¼ 7:5 min (a) and time histories at three points of
tunnel located at different heights above the pavement.
Fig. 7. Mechanical damage parameter distributions at time stations
t ¼ 5 min and t ¼ 7:5 min (a) and time histories at three points of
tunnel located at different heights above the pavement.
Fig. 8. Total damage parameter distributions at time stations t ¼ 5 min
and t ¼ 7:5 min (a) and time histories at three points of tunnel located
at different heights above the pavement.
248 B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250
present in the layer of concrete directly exposed to fire,
in the three considered locations, Fig. 7(a). This is
mainly due to the high temperatures reached, about 630
K after 5 min and 690 K after 7.5 min (Fig. 3), whichinduce thermo-chemical changes in the microstructure
of the material in terms of porosity and permeability.
Also mechanical features of concrete are affected by
these processes (i.e. dehydration and cracking) and this
leads to a damaging of material. A thermal dilatation of
the surface vault layer, while the internal part remained
at the initial temperature, caused a considerable me-
chanical damage of the surface layer of the vault, Fig. 6,what contributes to the total concrete damage, Fig. 8.
In presence of a mechanical and thermo-chemical
damage of the concrete, i.e. when the thermal front
deeply enters into the wall causing differential thermal
dilatation for a larger depth of material, significant
values of pore vapour pressure can produce a separation
of superficial layers of concrete, phenomenon known as
spalling [16,23,24,27]. With the increase of the fire du-ration, a risk of this phenomenon significantly increases.
6. Conclusions
In this paper a thermodynamically consistent model
for heat and mass transfer in deforming porous media,
B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250 249
including phase changes, has been developed. Hybrid
mixture theory has been used for this purpose and in-
terface properties have been included. From this theo-
retical model a simplified numerical model has been
extracted. This model is used for simulating the behav-iour of heated concrete in tunnel fires. Significant phe-
nomena such as spalling and thermal diffusion have
been evidenced. This application shows, together with
other not reported here, that the model is extremely
versatile.
Acknowledgements
The authors are grateful to Prof. P. Brunello from the
Dipartimento di Costruzione dell�Architettura, Univer-
sit�a IUAV (Venice, Italy), for his valuable contribution.
This work has been partially funded by the UE project‘‘UPTUN – Cost-Effective, Sustainable and Innovative
Upgrading Methods for Fire Safety in Existing Tunnels’’.
References
[1] Hassanizadeh M, Gray WG. General conservation equations for
multiphase systems. 1. Averaging procedure. Adv Water Resour
1979;2:131–44.
[2] Hassanizadeh M, Gray WG. General conservation equations for
multiphase systems. 2. Mass, momenta, energy and entropy
equations. Adv Water Resour 1979;2:191–203.
[3] Hassanizadeh M, Gray WG. General conservation equations for
multiphase systems. 3. Constitutive theory for porous media flow.
Adv Water Resour 1980;3:25–40.
[4] Hassanizzadeh M. Derivation of basic equations of mass trans-
port in porous media, Part 2 Generalized Darcy�s and Fick�s law.Adv Water Resour 1986;9:207–22.
[5] Hassanizadeh M, Gray WG. Mechanics and thermodynamics of
multiphase flow in porous media including interphase transport.
Adv Water Resour 1990;13:169–86.
[6] Schrefler BA. Thermodynamics of saturated–unsaturated porous
materials and quantitative solutions: the isothermal case. Pro-
ceedings of Second European Conference on Computational
Mechanics, 2001.
[7] Coleman BD, Noll W. The thermodynamics of elastic materials
with heat conduction and viscosity. Arch Ration Mech Anal
1963;13:168–78.
[8] Lewis RW, Schrefler BA. The finite element method in the static
and dynamic deformation and consolidation of porous media.
Chichester: J. Wiley & Sons Ltd; 1998.
[9] Schrefler BA. Mechanics and thermodynamics of saturated–
unsaturated porous materials and quantitative solutions. Appl
Mech Rev 2002;55(4):351–88.
[10] Gray WG, Hassanizadeh M. Paradoxes and realities in unsatu-
rated flow theory. Water Resour Res 1991;27:1847–54.
[11] Gray WG, Hassanizadeh M. Unsaturated flow theory includ-
ing interfacial phenomena. Water Resour Res 1991;27:
1855–63.
[12] Baggio P, Bonacina C, Schrefler BA. Some considerations on
modelling heat and mass transfer in porous media. Transport
Porous Media 1997;28:233–81.
[13] Gawin D, Pesavento F, Schrefler BA. Modelling of hygro-thermal
behaviour and damage of concrete at temperature above the
critical point of water. Int J Numer Anal Meth Geomech
2002;26:537–62.
[14] Zienkiewicz OC, Taylor RL. The finite element method, vol. 1:
The basis. Oxford: Butterworth-Heinemann; 2000.
[15] Zienkiewicz OC, Taylor RL. The finite element method, vol. 2:
Solid mechanics. Oxford: Butterworth-Heinemann; 2000.
[16] Pesavento F. Non-linear modelling of concrete as multiphase
porous material in high temperature conditions. Ph.D. thesis,
University of Padova, Padova, 2000.
[17] Bazant ZP, Kaplan MF. Concrete at high temperatures: material
properties and mathematical models. Harlow: Longman;
1996.
[18] Khoury GA, Majorana CE, Pesavento F, Schrefler BA. Modelling
of heated concrete. Mag Concr Res 2002;54:1–25.
[19] Gerard B, Pijaudier-Cabot J, Laborderie C. Coupled diffusion-
damage modelling and the implications on failure due to strain
localisation. Int J Solids Struct 1998;35:4107–20.
[20] Nechnech W, Reynouard JM, Meftah F. On modelling of thermo-
mechanical concrete for the finite element analysis of structures
submitted to elevated temperatures. In: de Borst R, Mazars J,
Pijaudier-Cabot G, van Mier JGM, editors. Proceedings of the
fracture mechanics of concrete structures. Lisse: Swets & Zeitlin-
ger; 2001. p. 271–8.
[21] Mazars J. Application de la mecanique de l� endommagement au
comportament non lineaire et la rupture du beton de structure.
Th�ese de Doctorat d� Etat, L.M.T., Universite de Paris, France,
1984.
[22] Mazars J. Description of the behaviour of composite concretes
under complex loadings through continuum damage mechanics.
In: Lamb JP, editor. Proceedings of the Tenth US National
Congress of Applied Mechanics. ASME, 1989.
[23] Gawin D, Pesavento F, Schrefler BA. Simulation of damage –
permeability coupling in hygro-thermo-mechanical analysis of
concrete at high temperature. Commun Numer Meth Engng
2002;18:113–9.
[24] Gawin D, Pesavento F, Schrefler BA. Modelling of hygro-thermal
behaviour of concrete at high temperature with thermo-chemical
and mechanical material degradation. Comput Methods Appl
Mech Engrg 2003;192:1731–71.
[25] Ulm FJ, Acker P, L�evy M. The ‘‘chunnel’’ fire. II. Analysis of
concrete damage. J Eng Mech Trans ASME 1999;125(3):283–9.
[26] Ulm FJ, Coussy O, Bazant ZP. The ‘‘chunnel’’ fire. I. Chemo-
plastic softening in rapidly heated concrete. J Eng Mech Trans
ASME 1999;125(3):272–82.
[27] Gawin D, Majorana CE, Schrefler BA. Numerical analysis of
hygro-thermic behaviour and damage of concrete at high temper-
ature. Mech Cohesive-Frictional Mater 1999;4:37–74.
[28] Schrefler BA, Brunello P, Gawin D, Majorana CE, Pesavento F.
Concrete at high temperature with application to tunnel fire.
Computat Mech 2002;29:43–51.
[29] Xue H, Ho JC, Cheng YM. Comparison of different combustion
models in enclosure fire simulation. Fire Safety J 2001;36(1):37–
54.
[30] Goldin GM, Menon SA. A comparison of scalar PDF turbulent
combustion models. Combust Flame 1998;113(3):442–53.
[31] Rubenstein RY. Simulation and the Monte Carlo method. New
York: J. Wiley & Sons Ltd; 1981.
[32] Brunello P, Zecchin R. A Monte-Carlo approach for the design
of high temperature heating panels. Proceedings of the Inter-
national Symposium Energy Conservation in the Built Envi-
ronment 1993, CIB Publication 152, Stuttgart, IRB-Verlag,
Germany.
[33] Brunello P. Solar pressure evaluation on large reflectors for space
applications. Commun Appl Numer Meth 1993;9(10):787–
95.
250 B.A. Schrefler, F. Pesavento / Computers and Geotechnics 31 (2004) 237–250
[34] Brunello P, Gnesotto F, Majorana CE, Schrefler BA. Thermal
analysis of a graphite first wall for fusion experiment RFX.
Commun Appl Numer Meth 1988;4:647–56.
[35] Holman JP. Heat transfer. New York: McGraw-Hill; 1976.
[36] Eckert ERG, Drake RM. Analysis of heat and mass transfer. New
York: McGraw-Hill; 1972.
[37] Ministry of Internal Affairs. Fire test in highway tunnel, (in
Italian), Corpo Nazionale dei Vigili del Fuoco, Roma, 2000.
[38] IES 1993. Lighting Handbook. IES of North America, New York,
1993.
[39] Karlsson B, Quintiere JG. Enclosure fire dynamics. Boca Raton:
CRC Press; 2000.