Professional Documents
Culture Documents
A 2D Model For Coupled Heat Air and Moisture TR - 2016 - International Journal
A 2D Model For Coupled Heat Air and Moisture TR - 2016 - International Journal
A 2D Model For Coupled Heat Air and Moisture TR - 2016 - International Journal
A 2D model for coupled heat, air, and moisture transfer through porous
media in contact with air channels
Clément Belleudy a,b,⇑, Monika Woloszyn a, Marx Chhay a, Matthieu Cosnier b
a
LOCIE, CNRS UMR 5271, Université Savoie Mont Blanc, Campus scientifique Savoie Technolac, Bâtiment Helios, Avenue du Lac Léman, 73376 Le Bourget du Lac, France
b
Centre Scientifique et Technique du Bâtiment (CSTB), 24 rue Joseph Fourier, 38400 Saint Martin d’Hères, France
a r t i c l e i n f o a b s t r a c t
Article history: Detailed modelling of air leakage paths through complex building wall assemblies is a challenging task. It
Received 27 April 2015 requires transient modelling of diffusion and advection phenomena through fluid and solid domains,
Received in revised form 1 December 2015 including porous materials and air channels. In this article, the development of a numerical model
Accepted 10 December 2015
coupling heat air and moisture transfers (commonly called HAM transfers) is presented. The model is able
Available online 29 December 2015
to deal with non-isothermal air flow through complex 2D (two-dimensional) geometries, combining air
channels and porous media, air permeable or not. A stepwise 1D validation of the model is achieved
Keywords:
with numerical benchmarks. The model is then tested on a 2D air leakage configuration subjected to
HAM model
Moisture
infiltration and exfiltration scenarii.
Transient Ó 2015 Elsevier Ltd. All rights reserved.
Porous media
Air leakage
Air channels
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.12.030
0017-9310/Ó 2015 Elsevier Ltd. All rights reserved.
454 C. Belleudy et al. / International Journal of Heat and Mass Transfer 95 (2016) 453–465
material were achieved by [11,12], using relative humidity and (also called ‘‘conjugate approach”) except near bends where vortex
suction pressure as moisture potential, respectively. Further work effects cannot be captured. The line source approach has been
was done on realistic 2D and 3D geometries [13,14], with air enter- recently used to assess the effect of streaming air between timber
ing uniformly through one side of the building assembly. beam and masonry [17].
To deal with coupled HAM transfers in both porous media and An alternative way to capture the effect of air flow on the
air channels, different modelling strategies are reported in the lit- hygrothermal field is to add a transient moisture source in the
erature: Fig. 1 summarizes the different approaches. There are two assembly [18]. Air is therefore not modelled as an active compo-
major types of modelling approaches: the one-domain approach or nent in the assembly. The position of the moisture source in this
the two-domain approach. In the first one, a single set system of lumped approach must be determined according to practical expe-
transport equations is used in the whole domain including both rience (Fig. 1c).
porous media and air channels, with location dependent coeffi- On site investigations in France [19] proved that leaking air
cients. For example, air transfer is computed with a one-domain mostly flows through porous materials and thin air channels due
approach in Fig. 1d. and e. In the second-one, different transport to material imperfections and construction tolerances. In addition,
equations are used in each domain (e.g. Darcy law in porous media air inlet and outlet are not necessarily close to each other, which
and Navier–Stokes equation in air channel), so that an interface makes air leakage paths through the building envelope multidi-
condition must be written to connect both computational domains. mensional and difficult to map. Very few existing HAM models
In this regard, the one-domain approach enable easier numerical enable such complex air leakage geometries to be dealt with.
implementation, as it only requires one solver to solve the equa- Regarding air gaps in contact with air permeable porous mate-
tion over both domains (there is no interface coupling). More rial, one of the most detailed modelling approaches is proposed by
details about these two approaches can be found in [15]. [20] and has been recently implemented in DELPHIN by [21]. A
Air transfer computational domains are presented in the first one-domain approach is used to compute air velocity field in both
row and those of heat and moisture transfers in the second row. air permeable porous material and air channels, with Darcy law
In each domain, indicated by a thick line, transfer processes are and the averaged Poiseuille law, respectively. Natural convection
ruled by the same partial derivative equations (PDEs). Inside cer- is captured using Boussinesq approximation. Heat and moisture
tain domains, dotted lines separate areas of different properties. transfers in porous material and air channels are computed sepa-
In case of air channels in contact with airtight porous materials, rately, using two-domain approach. The interaction between the
[16] compares a one-domain approach (Fig. 1a.) and a line source two domains is taken into account using constant heat and vapour
approach (Fig. 1b.). The first one consists in solving the velocity surface film coefficients along interfaces (Fig. 1d).
field in the air channel alone with CFD (Computational Fluid For 2D tortuous air gaps, the assumption of simplified boundary
Dynamics) codes, and then solving simultaneously hygrothermal conditions at air gaps interfaces implied by a two-domain
transfers in both air gap and porous domains. No surface film coef- approach might be questioned as effective values of surface film
ficient is thus required to couple fluid and airtight porous material. coefficients vary significantly between rectilinear and bend sec-
The line source approach (which belongs to two-domain tions. It is possible to dispense with these boundary conditions
approaches) reduces the computational effort, which is profitable by describing HAM transfers continuously in air gaps and porous
for long-term simulations. The basic principle is to describe the media with one single system of equations. Such a one-domain
air gap in 1D and calculate a velocity with a macroscopic law such approach (Fig. 1e.) is proposed in the present paper. This newly-
as the power law. Then the convective transfer of moisture and developed HAM model is designated as HAM-Lea (‘‘Lea” standing
energy can be modelled along the air channel axis with 1D balance for ‘‘Leakage”).
equations and surface film coefficients. An overall good agreement In a first part of the paper, the general governing equations for
is obtained compared to the comprehensive one-domain approach coupled HAM transfers in both multilayered porous media and air
Airtight
Navier Stokes Poiseuille law Poiseuille law
1D constant velocity
Airtight Airtight
Darcy law Darcy law
natural
Coupling
convection
HEAT & MOISTURE
HM HM HM HAM HAM
hconv β moisture
source
1D HAM hconv β
HAM HAM HAM
HM HM HM HAM
a. Two-domain approach b. Line source approach: c. Lumped approach d. One-domain approach e. Proposed approach:
for air flow a two-domain approach for with moisture source for air flow One-domain approach
One-domain approach for air flow and HM transfer Two-domain approach for for air flow and HM transfer
HM transfer HM transfer
channels are presented, and suitable assumptions are made, lead- This equation can be simplified, depending on air velocities and
ing to the HAM-Lea system of equations. Secondly, a numerical val- the level of detail needed for the simulation. The pore Reynolds
idation of this system is presented, based on published numerical number gives information on the flow in the porous medium.
benchmarks. Finally, HAM-Lea is applied to a complex 2D geome-
Lc qair u
try including air channels and porous media, subjected to transient Rep ¼ ð6Þ
boundary conditions. This last section illustrates the applicability
lair
of HAM-Lea to real problems from building physics. The characteristic length Lc ½m for the flow is generally
approximated by the square root of the permeability:
2. Governing equations pffiffiffiffiffiffiffiffiffi
Lc ¼ kmat ð7Þ
HAM transfers are ruled by conservation laws written as PDE. For building physics applications, the flow regime in the porous
For this purpose, the continuum medium approximation is used: medium remains laminar. However, the drag term differs depend-
all properties are averaged on representative elementary volumes ing on the range of Rep . For Rep > 10 the drag term is quadratic
(REVs). with velocity and thus both Darcy and Forchheimer terms must
The general form of a conservation law states that the rate of be considered. In contrast, for Rep of order of unity, the drag is
increase of a quantity A in an REV equals the net inflow of this linear and Darcy term is the only one required.
quantity r qA into this REV, plus a source term s: The Brinkman term is a viscous drag, similar to the laplacian
@A term in the Navier–Stokes equation. It is required when the
¼ r qA þ s ð1Þ non-slip condition near an impermeable wall interface needs to
@t
be captured. An effective viscosity l ~ ½Pa s is introduced, which is
where qA is the flux density of the A quantity. Depending on the bal- close to air dynamic viscosity for high porosity materials. As we
ance equation, quantity A represents dry air mass, dry air momen- do not need a fine description of boundary layer at wall interfaces,
tum, water mass, or energy, respectively. In the momentum Brinkman term will not be considered in our model.
conservation equation, the right term (net inflow of momentum) Given the low velocities, the inertial term is usually small com-
corresponds to external forces applied to the REV. pared to the drag terms [22]. Moreover, recent research work
In the following sections, equations for each quantity will be proved that it is justified to neglect transient effect of air transport,
detailed differentiating transport in porous media and in air because of much smaller time scale compared to heat and moisture
channels. transport [21]. This assumption is retained for the newly-
developed model. For steady state flow, the momentum equation
2.1. Air reduces to the general form of the phenomenological Darcy law:
kmat
2.1.1. Air flow in porous media u¼ ðrP qair ðTÞg Þ ð8Þ
In the porous medium, the general equation for dry air mass lair
conservation, also called continuity equation, reads: where g ½m2 =s is the gravitational acceleration vector. The vertical
@q axis (y coordinate) is oriented upwards. The body force, even if
e air ¼ r ðqair uÞ ð2Þ small for air in comparison of liquid, is included as it might have
@t
a significant contribution on the overall air flow. The body force
where e ½— is the open porosity of the porous medium, qair ½kg=m3 can be considered as constant, or temperature dependent when nat-
the dry air density and u ½m=s the Darcy velocity vector, which cor- ural convection needs to be captured. Natural convection is caused
responds to an averaged velocity over a REV. One neglects ‘‘source by air density gradient driven by temperature differences. Consider-
term”. The Dupuis–Forchheimer relationship provides the link ing a temperature dependent air density in all conservation equa-
between the Darcy velocity u and the intrinsic velocity v: tions dramatically increases the system complexity. The
u ¼ ev ð3Þ Boussinesq approximation allows capturing natural convection
restraining density variation to a buoyancy force in the momentum
In building physics, air velocities remain sufficiently low, conse- equation as shown in (8).
quently the assumption of incompressible flow is widely accepted. Within this approximation, air density temperature depen-
The continuity equation simplifies to: dence can be linearized using a Taylor series near an equilibrium
ru¼0 ð4Þ temperature T 0 , assuming small variation around this value.
The general form of momentum conservation for laminar air qair ðTÞ ¼ qair ð1 bðT T 0 ÞÞ ð9Þ
flow in porous media is obtained by analogy with the Navier–
with b ½K1 the coefficient of thermal expansion of air. An alterna-
Stokes equation applied to an air volume, with additional drag
tive choice is to use the ideal gas law, using the molar mass of air
terms due to the resistance of the rigid porous matrix against the
Mair ½kg=mol, an equilibrium pressure P0 , the universal gas constant
flow. Interested readers may refer to handbooks about heat and
R ½J=ðmol KÞ, and T the absolute temperature [K]:
mass transfer in fluids and porous media [22,23] for further details.
P0 M air
qair @u qair qair ðTÞ ¼ ð10Þ
þ ðu rÞðuÞ ¼ rfflP}
|fflffl{zffl þqair g RT
e @t
|fflfflffl{zfflfflffl} e2 ffl{zfflfflfflfflfflfflfflfflfflfflffl}
|fflfflfflfflfflfflfflfflfflffl |fflfflffl{zfflfflffl}
pressure force
local acceleration advective acceleration
body force Resulting pressure differences across the building envelope are
1 1=2 1 the driving force of air transport. In the present work, driving
l air kmat u cF kmat qair juju þ l ~ Du ð5Þ pressure was assumed equal to dry air pressure. Indeed, water
|fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflffleffl{zfflfflfflfflffl}
Darcy drag term Forchheimer drag term vapour contribution to total pressure is in general less than 2%.
Brinkman term
Resulting pressure differences are due to wind, stack effect and
where P ½Pa is the total air pressure, lair ½Pa s air dynamic viscosity, mechanical ventilation. Typical values are within [0–10 Pa]
l~ ½Pa s an effective viscosity, kmat ½m2 the material intrinsic perme- according to [7]. For an order of magnitude, a 160 mm thick
ability, and cF a dimensionless form-drag constant. layer of highly permeable porous material (glass fibre batt,
456 C. Belleudy et al. / International Journal of Heat and Mass Transfer 95 (2016) 453–465
qmat ¼ 16 kg=m3 ; kmat ¼ 3:85 109 m2 [24]) subjected to 10 Pa from air inlet or bends. This level of accuracy is deemed to be sat-
pressure difference, gives an air velocity of 13 mm/s and a pore Rey- isfactory regarding the characteristic lengths of leaking building
nolds number of 0.055. This proves the validity of Darcy law for assemblies (see Section 4).
common building materials in standard conditions. Furthermore, While integrating the parabolic velocity profile (14) over the y-
as the model is not dedicated to severe weather conditions with direction, the mean velocity profile can be expressed in the air gap
temperature gradients over 40 C, we can safely neglect natural con- and an equivalent permeability of the air channel appears:
vection in porous materials [8]. As a result, Darcy law including a
keq e2
constant body force (8) will be used to calculate the velocity field u¼ rP with keq ¼ ð17Þ
lair 12
in porous materials. The expression of Darcy law will be condensed
introducing the driving pressure P ¼ P qair g y. Here, g is the However, the use of Poiseuille law might be questioned while
absolute value of the gravitational acceleration and the y-axis is ori- modelling air channels that are in contact with permeable porous
ented upwards: material, as the no-slip condition is no longer verified at the inter-
face. The expression of a slip flow boundary condition between the
kmat
u¼ rP ð11Þ fluid and the air permeable porous medium has been introduced
lair by [27]:
Disregarding natural convection simplifies the problem as air duf abj
convection is decoupled from hygrothermal field. This enable the ¼ pffiffiffiffiffiffiffiffiffi ðusurf umat Þ ð18Þ
dy kmat
velocity field to be solved prior to the hygrothermal field, which
enhances simulation performance, while keeping good precision where abj is a dimensionless quantity depending on the structure of
of results. the permeable porous material, uf ; umat and usurf the y-velocity com-
ponents in the fluid, the material and at the interface, respectively.
2.1.2. Air flow in air channels As a consequence, the predicted flow rate in the air gap and hence
The continuity Eq. (4) describes air mass conservation in thin air the equivalent permeability keq are greater compared to those cal-
channels. culated with Poiseuille law. While resolving (13) with u ¼ 0 and
The momentum conservation applied to a fluid particle in (18) as boundary conditions at both sides of the channel, we obtain
motion is the Navier–Stokes equation. The variation rate of the following velocity expression:
momentum is equal to the volumic forces applied to this particle,
keq 3ðr þ 2abj Þ e
which are hydrostatic pressure, a body force, and a viscous drag u ¼ ð1 þ /Þ rP with / ¼ and r ¼ pffiffiffiffiffiffiffiffiffi
due to the fluid viscosity:
l rð1 þ abj rÞ kmat
ð19Þ
@u
qair þ ðu rÞðuÞ ¼ rP þ qair g þ lair Du ð12Þ If abj is assumed of order of unity, as suggested by [27], a 2 mm
@t
wide air channel in contact with a highly permeable porous mate-
Eqs. (12) and (5) are very similar. In air channels the porosity is rial (glass fibre batt, qmat ¼ 16 kg=m3 ; kmat ¼ 3:85 109 m2 [24])
equal to one and drag terms simplify to lair Du. As previously men- gives / 0:1, which corresponds to an air flow increase in 10%
tioned, the transient terms can be safely omitted in (12). As previ- compared those calculated with (13). This is an extreme case: for
ously, we consider a constant body force, and we do not account
less permeable insulation material (kmat 1010 m2 ), this flowrate
for natural convection. This leads to Poiseuille law [25], valid for
increase stays below 3%. As we are firstly interested in the magni-
2D fully-developed laminar flow in cylindrical pipes or between
tude order of this flowrate, Poiseuille law will be used in the model
parallel plates. Considering Cartesian coordinates with the x-axis
to calculate the equivalent permeability of the air channel.
parallel to the flow direction, and the driving pressure P , Poiseuille
law may be expressed as:
2.1.3. Boundary conditions
@2u 1 @P Boundary conditions are straightforward: air pressure or air
¼ ð13Þ velocity may be applied on air inlets, and a reference pressure at
@y2 lair @x
the air outlet. In most cases, resulting pressure differences between
In case of a flow between infinite parallel plates, (13) can be indoor and outdoor can be measured on field. Sometimes, air inlet
integrated as: velocity is easier to access, in particular for experimental setups
where air flow is supplied by a sample pump at a controlled rate
1 @P 2
uðyÞ ¼ ðy eyÞ ð14Þ through an airtightness defect [28]. Whether pressure or velocity
2lair @x
is imposed at air inlet, it is of importance to set a reference pres-
The air channel related Reynolds number is written using the sure elsewhere in the domain to ensure that the problem is well-
channel thickness e ½m as a characteristic length: posed. A slip boundary condition is written on airtight boundaries:
capillary suction flux densities, respectively, expressed in where @wðuÞ=@ u is directly the slope of the sorption isotherm.
½kg=ðs m2 Þ. The storage term @wðuÞ=@t, can be derived from the Moreover [31,32] recommend using relative humidity as driving
sorption curve mentioned below. potential for physical and numerical reasons. Of course, the choice
Moisture transfer consist of three main phenomena: vapour dif- of driving potentials does not impact the physical results, as under-
fusion, vapour advection by air flow, and capillary suction. The lined by [31–33]. For all these reasons, relative humidity is chosen
moisture diffusion flux, ruled by Fick law, is pointing in the oppo- as driven potential for all moisture transfers.
site direction of the vapour pressure gradient: The relative humidity is defined as the ratio of the partial
vapour pressure to the saturation vapour pressure P sat ½Pa:
g diff ¼ dmat ðuÞrpv ðT; uÞ ð22Þ
pv ½Pa is the partial pressure of water vapour and dmat ½s the vapour pv ðT; uÞ ¼ uPsat ðTÞ ð30Þ
permeability of the material. In order to make this property more Among the different existing correlations to calculate P sat ðTÞ,
meaningful, the ratio between air vapour permeability d0 and dmat , the following one from [34] has been implemented in HAM-Lea:
called the vapour resistance factor or l factor, is often used:
4042:9
d0 Psat ðTÞ ¼ u exp 23:5771 ð31Þ
l factorðuÞ ¼ ð23Þ T 37:58
dmat ðuÞ
The sorption curve gives the equilibrium moisture content in a
dmat and l factor depend on relative humidity u. Air is considered
material in contact with surrounding moist air. Within typical tem-
as an ideal gas mixture containing dry air and water vapour. Water
perature range considered in building physics, temperature impact
vapour carried by air flow is referred as advected moisture flow:
on the sorption curve can generally be omitted. It is thus measured
g adv ¼ qv ap ðT; uÞu ð24Þ at T ref ¼ 298:15 K and called the sorption isotherm. In addition, for
the sake of simplification, no hysteresis phenomena will be consid-
qv ap ½kg=m3 is the water vapour content of air, also called humidity ered between absorption and desorption.
by volume. According to ideal gas law: By replacing flux expressions (22), (24) and (28) in (21) and
Mw rewriting the equation with relative humidity as moisture state
qv ap ðT; uÞ ¼ p ðT; uÞ ð25Þ
RT v variable, the moisture conservation equation can be formulated
using a general coefficient form PDE:
Liquid water transport occurs firstly in smaller pores subjected
to capillary condensation, which generates a suction pressure @u
gradient. This liquid flow, namely capillary suction flow, can be da þ r ðcru au þ cÞ þ bru þ au ¼ f ð32Þ
@t
expressed with a Darcy law, as in (11):
The different coefficients are given below:
g liq ¼ K l ðuÞrPsuc ð26Þ
@wðuÞ
where K l ½s is called the liquid water permeability of the material. da ¼ @u
The body force is disregarded in the nabla operator, as its influence K l ðuÞqw RT ref
c ¼ dmat ðuÞPsat ðTÞ þ uMw
is generally negligible compared to those of the suction pressure [29].
Suction pressure is the pressure difference between liquid and or c ¼ dmat ðuÞPsat ðTÞ þ Dw ðuÞ @wð
@u
uÞ
vapour phases. Kelvin law states the equilibrium between both ðTÞ
a ¼ dmat ðuÞ dPsat
dT
rT ð33Þ
phases on the pore scale. Similarly to [30], P suc will be considered
as a function of relative humidity only. T ref ¼ 298:15 K is com- c ¼ 0
Mw P sat ðTÞ
monly chosen as reference temperature. The validity of this b ¼ u
RT
assumption will be assessed thanks to numerical benchmarks pre- a ¼ Mw rTu 1 dPsat ðTÞ
PsatT 2ðTÞ
R T dT
sented in the next section.
f ¼ 0
qw R
Psuc ðuÞ ¼ T ref ln u ð27Þ
Mw
where qw ½kg=m3 is the density of liquid water and Mw ½kg=mol its 2.2.2. Moisture transport in air channels
molar mass. Capillary suction flow can also be expressed with mois- Moisture transport in air channels is mainly due to advection by
ture content or relative humidity as potential: air movements and to vapour diffusion in air. Moreover, as water
content of air is generally several orders of magnitude lower than
@wðuÞ water content in materials, we can therefore assume a negligible
g liq ¼ Dw ðuÞrwðuÞ ¼ Dw ðuÞ ru ð28Þ
@u dependency of water vapour content of air to the temperature,
i.e. @ qv ap =@T ¼ 0.
where Dw ðuÞ ½m2 =s is the moisture diffusivity. It can be experimen- The general coefficient form PDE is identical to (32), and the dif-
tally measured and corresponding values for common building ferent coefficients read:
materials are available in literature.
The choice of adequate potential for moisture balance equation da ¼ Mw
Psat ðTÞ
RT
is still discussed in the literature. The continuity across material
c ¼ d0 Psat ðTÞ
interfaces is one of the important elements. In this regard, water ðTÞ
vapour pressure and relative humidity are both continuous func- a ¼ d0 dPsat
dT
rT
tions at the interface between two materials. The writing of the c ¼ 0 ð34Þ
M w P sat ðTÞ
chosen general PDE formulation (32) appeared to be easier with b ¼ u
relative humidity as potential. For example, the storage term
RT
1 dPsat ðTÞ
expressed with relative humidity (29), writes: a ¼ M w rTu
R T dT
PsatT 2ðTÞ
@wðuÞ @wðuÞ @ u f ¼ 0
¼ ð29Þ
@t @ u @t
where d0 ½s is the vapour permeability of air.
458 C. Belleudy et al. / International Journal of Heat and Mass Transfer 95 (2016) 453–465
2.2.3. Boundary conditions qmat ½kg=m3 is the density of dry material, cpair the heat capacity
At air inlets, the most straightforward condition is to impose at constant pressure of dry air.
relative humidity. We denote uinlet the inlet velocity, uamb and
usurf the interior ambient and surface relative humidity, dH ¼ ½qmat cmat þ wðuÞcw dT ð40Þ
respectively:
dH ¼ qair cpair dT ð41Þ
usurf ¼ uamb ð35Þ
The enthalpy variation rate of a REV is driven by three flux den-
Alternatively, the inward moisture flux can be imposed. It is sities, namely heat conduction qcond , heat convection by dry air
useful for instance when a moisture flux due to driving rain must qconv , latent and sensible heat carried by moisture qmoist , expressed
be included. Written with the formalism of (32), with bamb ½s=m in ½W=m2 .
as the surface film coefficient for vapour transfer, it reads:
h i @H
¼ r qcond r qconv r qmoist ð42Þ
n ðcru au þ cÞ ¼ bamb pv ðT amb ; uamb Þ pv ðT surf ; usurf ÞÞ @t
h i
þ uinlet qv ap ðT amb ; uamb Þ qv ap ðT surf ; usurf Þ According to Fourier law, the conductive flux reads:
where T amb is the ambient temperature, T surf the surface tempera- where kmat ðuÞ½W=ðm KÞ is the thermal conductivity of the medium
and depends on moisture. qconv is the heat flux density due to dry air
ture. The velocity related term is often omitted by researchers
convection.
[12,13,21], because it is small compared to the others in case of
air tight materials in combination with low pressure differences.
qconv ¼ qair cpair Tu ð44Þ
At air outlets, air exits at the boundary relative humidity, which
gives the following boundary condition: The total moisture flow g moist ¼ g diff þ g adv þ g liq causes both
h i sensible and latent heat fluxes, which are sometimes included in
n ðcru au þ cÞ ¼ bamb pv ðT amb ; uamb Þ pv ðT surf ; usurf ÞÞ a source term, as shown in (1). The flux density due to latent and
ð37Þ sensible heat carried by moisture reads:
In the general case, a two-temperature approach must be qmoist ¼ qlatent ¼ Lv ðg adv þ g diff Þ ð46Þ
adopted when dealing with REV containing both solid and gas
phases. In this case, methods exist to determine the interfacial con- Replacing the different fluxes in (42) gives:
vective heat transfer coefficient in porous media [35]. In building @T
0 0
physics area, given low leakage rates and high porosities of air per- da þ r ðc0 rT a0 T þ c0 Þ þ b0 rT þ a0 T ¼ f ð47Þ
@t
meable materials, thermal equilibrium between air and solid mate-
rial is attained within a small distance compared to the wall
0 uÞ
dimensions, as demonstrated by [36]. Therefore, it is justified to da ¼ qmat cmat þ wð
q cw
mat
2.3.3. Boundary conditions no longer coupled, which enables the moisture equation to be
Similarly to moisture boundary conditions, temperature (50) or solved analytically. The initial relative humidity of the material is
inward heat flux (51) can be imposed at air inlets: u ¼ 0:95, corresponding to an initial moisture content of
w ¼ 84:8 kg=m3 . At t ¼ 0, relative humidity falls to u ¼ 0:45 on
T surf ¼ T amb ð50Þ
the exterior side and u ¼ 0:65 on the interior side, corresponding
or to a moisture content of 19:5 kg=m3 and 30:5 kg=m3 respectively.
Heat surface film coefficient of 25 W=ðm2 KÞ and moisture surface
n ðc0 rT a0 T þ c0 Þ ¼ hamb ðT amb T surf Þ
h i film coefficient of 1:0 103 s=m are applied on both exterior and
þ Lv bamb pv ðT amb ; uamb Þ pv ðT surf ; usurf Þ
h i interior sides. Water content profiles in the wall simulated with
þ qair cpair uinlet ðT amb T surf Þ þ Lv uinlet qv ap ðT amb ; uamb Þ qv ap ðT surf ; usurf Þ ð51Þ the model are compared with analytical solutions provided by
the benchmark at 100 h, 300 h and 1000 h, as presented in Fig. 2.
As for moisture boundary conditions, the velocity related term On this figure, exterior and interior side of the monolithic structure
is often disregarded by researchers. At air outlets, the boundary is on the left side and on the right side of the x-axis, respectively.
condition reduces to: Excellent agreement is found between analytical and simulated
profiles.
n ðc0 rT a0 T þ c0 Þ ¼ hamb ðT amb T surf Þ
h i
þ Lv bamb pv ðT amb ; uamb Þ pv ðT surf ; usurf Þ 3.2. Insulated roof
ð52Þ
This benchmark pushes the model validation a step further as a
This boundary condition also holds for non adiabatic moisture two-layer wall is subjected to transient thermal and moisture con-
permeable interfaces. ditions. This case originally aims to assess the model ability to pre-
dict interstitial condensation between materials. A load bearing
2.4. Numerical tool wall is insulated on the interior side and a vapour barrier is placed
on the exterior side. The whole structure is perfectly airtight. Mate-
The developed numerical model is implemented in the rials have different moisture properties: the load bearing material
commercial simulation software COMSOL Multiphysics [37] which is hygroscopic and capillary active whereas the insulation material
is manipulated via its user-friendly GUI (Graphic User Interface). is less hygroscopic and not capillary active. The simulation is per-
The user can either built a geometry directly or import a formed over four years. One year transient boundary conditions are
CAD-file. Constant material properties and values can be entered repeated every year. The total moisture content in the load bearing
as parameters, and variable coefficients can be defined either as material computed by HAM-Lea is plotted in the fifth year, and
analytical functions or as a list of discrete values generating inter- compared with results obtained by different universities and
polated functions. Conservations laws, written as PDEs (Eqs. (11), research institutes, as presented on Fig. 3. The simulated results
(17), (32), (47)), are affected to each domain (various materials, are contained within the envelope formed by the other plots,
air channels), and the values of the corresponding coefficients are which proves good performance of the model.
entered. The mesh can be generated directly from the GUI, and it
is possible to refine mesh in regions where steep gradients are 3.3. Lightweight wall
expected. As indicated previously, disregarding natural convection
enables the stationary air flow balance to be solved prior to the In this third and last studied benchmark case, air transfer is
transient moisture and energy balance. The resulting velocity field finally coupled to hygrothermal transfers. In addition to vapour
is used while solving coupled heat and moisture equations. Heat diffusion and liquid water transport, moisture can be advected
and moisture equation are solved simultaneously with COMSOL’s by air flow. The internal side of a 200 mm thick wall is air and
built-in time-dependent solver MUMPS. It is based on the finite vapour permeable, whereas the external side is air permeable
element method and an explicit scheme with variable time step- but vapour tight. During the first 20 days (i.e. 480 h) an exfiltration
ping. It is possible to define a maximum timestep, coinciding with is simulated: air flows from the interior to the exterior side. After-
the one of transient boundary conditions. When convergence issue wards, from the 20th to the 100th day, air direction is reversed to
arise, it can be useful to adjust some of the solver settings such as simulate an infiltration. Fig. 4 pictures the moisture content at
damping factor, relative tolerance, maximum number of iterations. x ¼ 0:1 m against time. It can be seen that the material stores
moistures when air exfiltrates because this moisture cannot exit
on the exterior side. On the contrary, a drying of the wall occurs
3. Benchmarking of the model
in the infiltration phase. An excellent agreement is obtained
between the benchmark and the model.
In the previous sections, coupled HAM equations implemented
in HAM-Lea model were presented. Now, in order to gain confi- 90
dence in model results, a validation of HAM-Lea using published 80 100h
benchmarks is achieved. They have been developed in the frame- 70
work of the European HAMSTAD project, which aimed to standard- 300h
w (kg/m3)
60
ize HAM calculation methods [30]. Three benchmark cases have
50
been chosen, and a step by step methodology was followed for this 1000h
40
validation, adding physical processes and coupling terms one after
30
another. A more comprehensive description of used material Analytical
20 HAM-Lea
properties can be found in [30].
10
0 0.05 0.1 0.15 0.2
3.1. Homogeneous wall
Thickness of material (m)
In this benchmark, a monolayer wall is maintained at a constant Fig. 2. Comparison between analytical solution and HAM-Lea outputs. The interior
temperature T 0 ¼ 20 C. Thus, moisture and energy equations are side is located.
460 C. Belleudy et al. / International Journal of Heat and Mass Transfer 95 (2016) 453–465
11
4.2. Boundary conditions and solver settings
10
To simulate potentially problematic conditions, a high moisture
0 2000 4000 6000 8000
scenario from WUFI weather database is used. It includes both tem-
Time (h)
perature and relative humidity variations over a year (Fig. 7). No
Fig. 3. Total moisture content in the load bearing material during the fifth year. external moisture flux due to driving rain is considered. The relative
humidity is set at 0.8 as initial value for the whole building assem-
bly. As long term behaviour is of interest here, mean annual varia-
120 tions of temperature and relative humidity are considered. They are
all benchmark results described by analytical functions, and presented in Figs. 7 and 8.
100 The initial temperature is set at ½T int ðt ¼ 0Þ þ T ext ðt ¼ 0Þ=2 ¼ 10 C
HAM-Lea
80 for the whole building assembly. As described in Section 2, pre-
w (kg/m3)
scribed boundary conditions are adopted at air inlets for both mois-
60
ture and heat equations. Heat and moisture surface film coefficients
40 are applied on interfaces in contact with ambient air. Typical mois-
ture film coefficients suggested by [34] are used:
20
exfiltration infiltration
0 bint ¼ 18:5 109 s=m bext ¼ 140 109 s=m ð53Þ
0 400 800 1200 1600 2000 2400
Heat surface film coefficients are drawn from EN Standards
Time (h)
[34]:
Fig. 4. Moisture content at x = 0.1 m during 100 days.
hint ¼ 7:7 W=ðm2 KÞ hext ¼ 25 W=ðm2 KÞ ð54Þ
Pressure differences are set between air inlets and outlets. Two
Following this numerical validation using three of the pressure differences will be tested, resulting in two flow rates. As
HAMSTAD benchmarks, HAM-Lea was successfully compared with previously mentioned, pressure differences across the envelope
experimental measurements in an additional paper [38]. This case are in general lower than 10 Pa, and smaller values are more likely
represents the impact of air flow through loose fill thermal to be maintained over long periods of time. That is why 0:1 Pa and
insulation. 1 Pa were chosen as pressure differences for the simulations. If we
calculate the velocity field using Darcy’s law and Poiseuille’s law as
4. HAM transfers through porous media and air channels: a case described in Section (2.1), we do not obtain the same flow rate in
study infiltration and exfiltration for a given pressure difference
jPinside P outside j. This is due to the gravity force, which creates an
4.1. Configuration hydrostatic pressure corresponding to the weight of an air column.
As the orifice of the air gap in the exterior side is lower than those
After validating the model equations, it is proposed to investi- in the interior side, the weight of the air column acts from the
gate coupled HAM transfers in a complex configuration including top down, and generates an air flow from the interior side to
porous media – air permeable or not – and thin air channels. In this the exterior side. This contribution is permanent, even if
regard, a typical envelope leakage encountered in wooden-frame jPinside P outside j ¼ 0. In order to fairly compare the impact of air
buildings is chosen as case study [39]. 2D vertical section gives flow on HAM transfer in infiltration and exfiltration, it is of impor-
the composition (Fig. 5) and the dimensions of the studied config- tance to have the same flow rate for both cases, for a given pressure
uration. Thermal insulation is wood fibreboard except near the difference. This is the reason why, even if the gravity creates a sig-
concrete slab where XPS is used. As a consequence of poor work- nificant contribution in the overall air flow, we will not take it into
manship, flexible sealing may be overlooked and the vapour barrier account it in the following analysis. In exfiltration, when gravity is
not sealed properly, which may lead to potential air leakage path included, the predicted flow rate is twice as high as the one when
between the wood bottom wall plate and the foundation wall. A gravity is disregarded. Therefore, even if the considered flow rate
potential 2 mm air gaps created by tolerances between materials are biased compared to the real ones, they remain in the same
links both the interior and the exterior sides. The vapour barrier order of magnitude, which maintains the relevancy of this analysis.
has been removed to reproduce the effect of strong discontinuities. Omitting gravity finally leads to total flowrates of 0:04 m3 =h and
Another approximation is done regarding the thermal insulation 0:4 m3 =h, for pressure differences of 0.1 Pa and 1 Pa, respectively.
layer located between vertical studs. As it is not directly subjected For each flow rate, both infiltration and exfiltration scenario are
to air pressure gradient, it has been considered as airtight to limit tested. The simulation is performed over 4 years. During the first
the air computation domain. These assumptions allow us to have year, there is no air flow: only heat and moisture diffusion occur.
clearly defined air inlets, outlets, as well as an air computational At the end of the first year, pressure differences are applied, gener-
domain for the simulation (Fig. 6). It is assumed that concrete, ating air flow through the wall assembly.
wood and XPS insulation layers are airtight as well. As natural con- The hygrothermal field at the end of the first year is used as
vection is not taken into account, the velocity can be solved prior to initial conditions for 3-year exfiltration and infiltration simula-
energy and moisture equations. As XPS insulation is neither hygro- tions. An overview of the simulation scheme is provided in Fig. 9.
scopic nor capillary active, its water content was set equal to air Boundary conditions are summarized in Fig. 10.
C. Belleudy et al. / International Journal of Heat and Mass Transfer 95 (2016) 453–465 461
Interior 1.0
φext 0.0190
conditions
0.8
ρvap (kg/m3)
0.6 0.0140
φint
φ (-)
Exterior
conditions 0.4 ρvap-int
0.0090
0.2 ρvap-ext
0.0 0.0040
0 2000 4000 6000 8000
Time (h)
Fig. 6. Computed fluid domain and air path for an infiltration scenario.
25
Tint
20
15
T (°C)
10
Text
5
0
0 2000 4000 6000 8000 Fig. 9. Pressure boundary conditions for all five scenarii.
Time (h)
4.3. Results and discussion
Fig. 7. Inside and outside temperatures over a year.
To assess the impact of air flow on the hygrothermal field in the
building component, the considered indicator is the averaged
The geometry has a total of 250,000 meshes built with COMSOL moisture content of the wood bottom plate (Fig. 5). This rectangu-
meshing built-in interface. We used an unstructured meshing, lar area appeared to be particularly impacted by the air flow, hence
which is refined in narrow regions and in regions where high gra- the relevancy of this choice. The averaged moisture content of the
dients are expected, for instance in the vicinity of the air channel wood bottom plate can be calculated as follows:
and on interfaces with ambient air. The HAM model requires on RR
wðx; y; tÞdxdy
average 2 h to run an annual simulation, using an Intel Xeon E5- wmoy ðtÞ ¼ S
ð55Þ
1650 CPU v2 at 3.5 GHz and 128 GB RAM workstation. Only 5 GB S
RAM are required to run the simulation. where S is the surface of the section.
462 C. Belleudy et al. / International Journal of Heat and Mass Transfer 95 (2016) 453–465
Plots in Figs. 11 and 12) show this averaged moisture content 130
for infiltration and exfiltration scenarii respectively, with the two exfiltration 0.1Pa
120
pressure differences. exfiltration 1Pa
The four-year hygrothermal simulation with no air flow shows 110
w (kg/m3)
no airflow
that the assembly reaches periodic hygric equilibrium after one 100
year. Moisture content is fluctuating with an amplitude of 90
5 kg=m3 around its mean value is of 80 kg=m3 . Beginning at the
80
end of the first year, the HAM simulation of the infiltration scenario
shows drying of the wood bottom plate. In this case, two years are 70
no airflow
needed to reach hygric equilibrium at around 72 kg=m3 for the 60
flowrate corresponding to P inside Poutside ¼ 0:1 Pa. For 0 1 2 3 4
Pinside P outside ¼ 1 Pa, the drying process is slightly more pro- Time (years)
nounced with an averaged moisture content of 70 kg=m3 . For both
Fig. 12. Averaged moisture content of wood bottom plate for exfiltration.
flow rates, moisture content amplitudes are very close to those
observed without air flow.
The exfiltration scenario (Fig. 12) shows opposite tendencies.
The exfiltrated air flow causes a significant increase in moisture content in the wood bottom plate. For Pinside Poutside ¼ 0:1 Pa, the
equilibrium moisture content increases by 20 kg=m3 compared to
the one without air flow. For P inside P outside ¼ 1 Pa, the predicted
moisture content reaches 100 kg=m3 as mean value, with an ampli-
130
infiltration 0.1Pa tude of nearly 50 kg=m3 . When interior moist air exfiltrates
120 through the air channel, it comes into contact with increasingly
infiltration 1Pa
110 cold building components, which increases its relative humidity
no airflow
w (kg/m3)
References
[27] G.S. Beavers, D.D. Joseph, Boundary conditions at a natural permeable wall, J. [34] Hugo Hens, Building Physics – Heat, Air and Moisture: Fundamentals and
Fluid Mech. 30 (1967) 197–207. Engineering Methods with Examples and Exercises, Ernst & Sohn, Berlin,
[28] Clément Belleudy, Ahmad Kayello, Monika Woloszyn, Hua Ge, Paul Fazio, Marx Germany, 2012.
Chhay, Daniel Quenard, A heat-airflow model for simulating the effects of air [35] Fujio Kuwahara, Mitsuhiro Shirota, Akira Nakayama, A numerical study of
leakage on the temperature field in porous insulation, in: 10th Nordic interfacial convective heat transfer coefficient in two-energy equation model
Symposium on Building Physics, Lund, Sweden, 2014, pp. 79–86. for convection in porous media, Int. J. Heat Mass Transfer 44 (6) (March 2001)
[29] Hartwig M. Künzel, Simultaneous Heat and Moisture Transport in Building 1153–1159.
Component (Ph.D. thesis), Fraunhofer IBP, 1995. [36] Christopher R. Buchanan, Max H. Sherman, A mathematical model for
[30] Carl-Eric Hagentoft, A.S. Kalagasidis, B. Adl-Zarrabi, S. Roels, J. Carmeliet, H. infiltration heat recovery, Report LBNL-44294, Lawrence Berkeley National
Hens, J. Grunewald, M. Funk, R. Becker, D. Shamir, O. Adan, H. Brocken, K. Laboratory, May 2000.
Kumaran, R. Djebbar, Assessment method of numerical prediction models for [37] COMSOL, COMSOL Multiphysics User’s Guide- version 5.0, October 2014.
combined heat, air and moisture transfer in building components: [38] Clément Belleudy, Ahmad Kayello, Monika Woloszyn, Hua Ge, Experimental
benchmarks for one-dimensional cases, J. Build. Phys. 27 (4) (2004) 327–352. and numerical investigations of the effects of air leakage on temperature and
[31] Graham H. Galbraith, R.C. McLean, J. Guo, The selection of appropriate flow moisture fields in porous insulation, Build. Environ. 94 (December 2015) 457–
potentials for moisture transport models, in: Proceedings of the 6th IBPSA 466.
Conference, Prague, Czech Republic September, 1997, pp. 8–10. [39] CETE de Lyon, Carnets Prebat Mininfil, Mémento étanchéité – Construction
[32] Hans Janssen, Simulation efficiency and accuracy of different moisture transfer Ossature Bois – Isolation Thermique Intégrée. Report, CETE de Lyon, November
potentials, J. Build. Perform. Simul. 7 (5) (September 2014) 379–389. 2010
[33] Max Funk, Karim Ghazi Wakili, Driving potentials of heat and mass transport [40] Anton TenWolde, William B. Rose, Moisture control strategies for the building
in porous building materials: a comparison between general linear, envelope, J. Building Phys. 19 (3) (1996) 206–214.
thermodynamic and micromechanical derivation schemes, Transp. Porous [41] Arnold Janssens, Hugo Hens, Interstitial condensation due to air leakage: a
Media 72 (3) (April 2008) 273–294. sensitivity analysis, J. Building Phys. 27 (1) (July 2003) 15–19.