Sie sind auf Seite 1von 10

International Journal of Heat and Mass Transfer 45 (2002) 509518

www.elsevier.com/locate/ijhmt

A new mathematical method to solve highly coupled


equations of heat and mass transfer in porous media
N. Mendes a,*, P.C. Philippi b, R. Lamberts b
a
Department of Mechanical Engineering, Pontical Catholic University of Parana, Rua Imaculada Conceica~o,
1155 Prado Velho, Curitiba, PR, 80215-901, Brazil
b
Department of Mechanical Engineering, Federal University of Santa Catarina, Florian
opolis-SC, 88040-900, Brazil
Received 7 January 2000; received in revised form 9 May 2001

Abstract
Heat and mass conservation equations in porous media are coupled and, in general, solved, iteratively, by
using the values of temperature and moisture content from previous iteration to calculate source terms. This is
the traditional mathematical method and numerical stability is only ensured for small time steps, depending on
the source term's magnitudes. This is specially important, when material properties have strong variations with
moisture content. This paper presents an unconditionally stable numerical method, conceived accordingly to a
new methodology, which considers: (i) linearization of the term giving the vapor exchanged at the boundaries in
terms of temperature and moisture content and (ii) introduction of a new generic algorithm to solve, simulta-
neously, the governing equations, for each time step. Numerical stability of these two methods are compared and
it is shown that, in addition to avoid numerical unstability for arbitrary time steps and material properties,
convergence is always quickly reached, in the presently proposed calculation method. 2001 Elsevier Science
Ltd. All rights reserved.

1. Introduction transport of moisture in the porous structure of the


walls. However, walls are normally submitted to both
In several elds such as drying processes, agriculture thermal and moisture gradients so that an accurate heat
and building energy analysis, it is important to know the transfer determination requires a simultaneous calcula-
dynamics of temperature and moisture content tion of both sensible and latent eects.
distributions and how they are related to each other Besides its eect on heat transfer, moisture has
to evaluate heat ux and mass ow through porous other implications, especially in hot/humid climates.
media. It is well known that moisture can cause damage to
In spite of the main focus of this paper is on the the building structure and can promote the growth of
presentation of a new mathematical method and show mold and mildew, aecting the health of building
how it is applied in hygrothermal studies of porous occupants.
building materials, the heat and mass transfer formula- Several investigators have developed models for
tion presented can also be used in many other elds moisture transport in porous materials. Cunningham [1]
where the mass and energy balances are the governing developed a mathematical model for hygroscopic ma-
equations. terials in at structures that uses an electrical analogy
In building energy analysis, calculated heat conduc- with resistances for the vapor ow and an exponential
tion through walls usually neglects the storage and approximation function with constant mass transport
coecients. Kerestecioglu and Gu [2] investigated the
phenomenon using evaporationcondensation theory in
*
Corresponding author. Tel.: +55-41-330-1322; fax: +55-41- the pendular state (unsaturated liquid ow stage). The
330-1349. application of this theory is limited to low moisture
E-mail address: nmendes@ccet.pucpr.br (N. Mendes). content. Burch and Thomas [3] developed a computer

0017-9310/02/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 1 7 - 9 3 1 0 ( 0 1 ) 0 0 1 7 2 - 7
510 N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518

program, MOIST, using the nite-dierence method to moisture equation is solved in a two-step procedure.
estimate the heat and mass transfer through composite First, vapor pressure is calculated using suction pressure
walls under non-isothermal conditions. El Diasty et al. data of the previous step. Then, the new vapor pressure
[4] used an analytical approach that assumed isothermal data are used as a known term, to calculate the new
conditions and constant transport coecients. Liesen [5] suction values. No iteration inside a time step is applied.
used evaporationcondensation theory and a response In MATCH, the amount of vapor condensing or evap-
factor method to develop and implement a model of orating in the control volume is taken, explicitly, from
heat and mass transfer in the building thermal simula- the former time step and, according to what was stated
tion program integrated building loads analysis and by Pedersen [9], in some cases it may be relevant to let
system thermodynamics (IBLAST). Hygrothermal liquid moisture transport term out of the calculation. If
property variations were neglected and liquid transfer this is not followed, the non-linearity of the problem
was not considered. In this way, IBLAST is restricted to would not guarantee the overall moisture balance to be
a very low moisture content, although requiring short kept.
calculation time. Yik et al. [6] developed a fast model This statement [9] shows that the algorithm used in
integrated with air-conditioning system component MATCH may have unstability problems even if the
models that employs evaporationcondensation theory source term (the amount of vapor condensing or evap-
with dierential permeability. In fact, above-described orating in the control volume) is calculated from the
models originated many simulation programs to predict former time step, which means that the source term is
the heat and moisture transfer through porous building written as a constant for each time step so that no it-
walls, but using calculation methods that are condition- eration is needed. In this way, unstability problems
ally stable. might appear because the solver does not, really, solve,
As the present paper is focused on the numerical simultaneously, the highly coupled governing equations.
method rather than on heat and mass transfer model MOIST [10] normally uses 1-h time step and provides
itself, a complete discussion about mathematical models reasonable results. However, the use of a such large time
was avoided. Actually, all those models have nearly the step for low density insulation materials may make
same origin (heat and mass balance equations, Philip MOIST simulations unstable and physically unrealistic.
and De Vries model and the laws of Fourier, Fick and This shows that MOIST algorithm is not always stable
Darcy). The main dierence among them is related to and small time steps may be required to avoid numerical
the particular assumptions used. In this way, it is be- convergence problems.
lieved that the models are, generally, reliable and the In general, as it can be noticed by IEA Annex 24
main aspect that will be treated, here, is the calculation description [7], by Pedersen [9] and by Burch and Chi
method itself. [10], any of the codes, mentioned above, does not have
In IEA Annex 24 [7], 5 models are commented in an algorithm with such a robust solver to calculate
detail: 1D-HAM, WUFIZ, MATCH, HYGRAN24 and simultaneously the temperature and moisture content
LATENITE. Below, they are briey described in terms distributions. Instead, they have stability criteria and
of their solvers. time step control devices to reduce numerical conver-
1D-HAM models air ow (as a known ux), heat gence problems. In the present paper a new calculation
conduction and vapor diusion. Time step depends on a method is presented, in order to enable yearly simulation
stability criterion for the explicit forward dierences of heat and moisture transfer in building porous walls,
scheme. The transport coecients are assumed to be possible to be performed with reduced processing times,
constant. WUFIZ matrix equations are solved per time considering arbitrary material properties and boundary
step by using an alternating-direction implicit method conditions. The method was conceived to preserve nu-
(ADI), with iteration between heat and moisture until a merical stability as a result of two new considerations.
desired accuracy is reached. It uses an implicit forward The rst considers the vapor exchanged between the wall
control volume method. ADI method is a commonly surfaces and the air, as a linear function of temperature
used line-by-line method, which was introduced by and moisture content, rather than vapor concentration.
Peaceman and Rachford in 1955 [8]. HYGRAN24 The second introduces a new generic algorithm to solve,
models air ow, heat conduction and vapor and liquid simultaneously, the governing equations.
diusion. For each time step, air ow is, rst, calculated. This new tridiagonal algorithm uses tensors instead
Then, heat and moisture matrix equations are solved of scalar numbers so that it reaches numerical conver-
consecutively, in an iterative loop, until the desired ac- gence very quickly by solving, at the same iteration level,
curacy is reached. It uses CranckNicholson scheme. both equations together, independently of their coupling
LATENITE equations can be solved in either iterative terms.
or block form. A delta form is employed for the dis- It is shown that the presently proposed method leads
continuous set of strongly coupled equations. MATCH to reliable results, free of numerical-stability related
heat matrix equation is solved for each time step. Then problems.
N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518 511

2. Mathematical model where hT1 Tx0 represents the heat exchanged with
the outside air, described by the surface conductance h,
The governing partial dierential equations to model aqr is the absorbed short-wave radiation and hm qv;1
heat and mass transfer through porous media are given qv;x0 , the phase change energy term. The solar
by Eqs. (1) and (2). They were derived from conserva- absorptivity is dened as a and the mass convec-
tion of mass and energy ow in a 1-D elemental volume tion coecient as hm which is related to h by Lewis'
of porous material. The energy conservation equation is relation.
written as The mass balance at the outside surface x 0 is
  described as
oT o oT o  
q0 cm T ; h kT ; h LT jv ; 1
ot ox ox ox o oh oT
Dh T ; h DT T ; h
ox ox ox x0
while the mass conservation equation as
  hm
oh o j q qx0 : 6
; 2 ql v;1
ot ox ql
The same above equations apply to the inside surface
where q is the solid matrix density, cm , mean specic x L, with the omission of short-wave related terms.
heat, T, temperature, t, time, k, thermal conductivity, L,
latent heat of vaporization hLV , h, volume basis
moisture content, jv , vapor ow, j, total ow and ql the 3. The traditional heat and mass transfer calculation
water density. method
Note that Eq. (1) diers from Fourier's equation for
transient heat ow by a source term due to phase change The governing equations are discretized by using the
within the pores. According to Philip and DeVries [11], control-volume formulation method. The interpolation
vapor ow density is given by method used in this 1-D problem is the central-dierence
jv oT oh scheme (CDS).
DTv T ; h Dhv T ; h 3 Due to coupling between moisture content h and
ql ox ox
temperature T, Eqs. (1)(4) are, as usually, discretized in
and for total mass ow density the form given below. For internal points, the mass
j oT oh conservation equation is written as
DT T ; h Dh T ; h 4  
ql ox ox Dx Dhe Dhw Dhe Dhw Dx
hp hE hW hop
Dt dxe dxw dxe dxw Dt
with DT DTl DTv and Dh Dhl Dhv , where DTl is 
the liquid phase transport coecient associated to a DTe o
T TPo
temperature gradient, DTv , vapor phase transport dxe E

coecient associated to a temperature gradient, Dhl , DTw o o
liquid phase transport coecient associated to a T TW 7
dxw P
moisture content gradient, Dhv , vapor phase transport
coecient associated to a moisture content gradient, and the energy conservation equation is written as,
DT , mass transport coecient associated to a tem-  
Dx ke Lql DTve kw Lql DTvw
perature gradient (m2 /s C) and Dh , mass transport q0 cm TP
Dt dxe dxe dxw dxw
coecient associated to a moisture content gradient    
(m2 /s). ke Lql DTve kw Lql DTvw
TE TW
dxe dxe dxw dxw
 o o
2.1. Boundary conditions Dx o Dhve hE hP
q0 cm T Lql
Dt P dxe
The associated conservation equations at the outside 
Dhvw hoP hoW
and inside wall surface are as follows. For the outside ; 8
dxw
surface x 0, it was considered that the wall is ex-
posed to short-wave radiation, convection heat and where superscript ``o'' indicates that the variable as-
mass transfer, and phase change. Thus, the energy bal- sumes its previous value. Subscripts p, e(E) and w(W)
ance becomes are related, as usually, to node p and its neighbors at
  east and west side, respectively.
oT
kT ; h LT jv x0 In the same way, Eqs. (5) and (6) are put in discrete
ox x0
form to obtain the following expressions for the half-
hT1 Tx0 aqr LT hm qv;1 qv;x0 ; 5 volume node at the boundaries S 0; L:
512 N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518
 
Dx Dhe Dhe Dx o opposite side of the wall is higher and can happen
hS hnS h S
2Dt dxe dxe 2Dt continuously for high time steps, preventing the solution
 o  to converge. Therefore, for the traditional method, nu-
T nS T o S
DTe merical stability is only ensured for very small time
dxe
steps.
hmt
q qS 9 To solve the stability problem and to increase the
ql v;1 time step, vapor concentration dierence was linearized
and in terms of temperature and moisture content dierences
  to strengthen the main diagonal. Linearization is dis-
Dx ke DTve cussed in the following section.
q0 cm Lql h T S
2Dt dxe dxe
 
ke DTve Dx o
Lql T nS q0 cm T S 4. Mathematical considerations for the vapor exchanged
dxe dxe 2Dt
 o  between the wall surfaces and air
h nS ho S
Lql Dhve hT1 :
dxe Vapor ow between a porous surface and air is
a1 dS; Lqr Lhm qv;1 qv S: 10 normally written as a function of vapor concentration
dierence Dqv which requires vapor ow to be
where nS is the internal point neighbor to S and da; b evaluated with previous values of temperature and
is the Kronecker delta (da; b 1 for a b and 0 moisture content. In this way, it is, in general, included
otherwise). in the source term of the system of equations, giving
Eqs. (7)(10) have the form additional unstablity source terms.
AP UP AE UE AW UW AoP UW D; 11 It is described below a new procedure to calculate
vapor ow, independently of previous values of tem-
where AP AE AW AoP , D AoP UoP F and U is a perature and moisture content by writing Dqv as a linear
dependent variable (T or h). combination of temperature and moisture concentration
The source term D contains besides the transient dierence.
term (Aop Uop ) all the terms qv S qv h; T which are The rst assumption is that water vapor behaves like
non-linear functions of the dependent variables. As the a perfect gas so that the concentration dierence can be
solution method is iterative, the values calculated for T written as
and h from the previous iteration are used for calculat-
ing the source terms in the above equations. In this P1 M P sM
qv;1 qv s ; 12
traditional method, mass and energy conservation RT1 RT s
equations are thus uncoupled, as coupling terms are where R is the universal gas constant, and M, the
always evaluated at a previous iteration. molecular mass.
Replacing the vapor pressure by the product Ps / of
3.1. Stability conditions saturated pressure Ps and relative humidity /,

Linearized Eqs. (7)(10) can be seen as two un- Ps;1 /1 M Ps s/sM


qv;1 qv s : 13
coupled tridiagonal system of equations one for T and RT1 RT s
the other for h, where the main diagonal is given by the
Eq. (13) can also be written as
coecients that accompany UP Eq. (11).
Decoupling of heat and mass conservation equations qv;1 qv s
results in high magnitude source terms, which are added  
M Ps;1 /1 Ps s/1 Ps s/1 Ps s/s
to the non-linear surface terms, associated to qv h; T ;
R T1 T s T s T s
dependence. Linear systems of equations are numeri-
cally solvable when main diagonal terms are dominant. 14
It is, thus, suitable that the source term in Eq. (11) be or,
dismembered in way to decrease its magnitude and in-
crease the main diagonal terms Ap . In the present case, qv;1 qv s
 
positiveness of vapor concentration dierence qv;1 M Ps;1 Ps s M Ps s
/1 / /s; 15
qv S, at a given step, means a given amount of vapor R T1 T s R T s 1
condensing on the surface and may raise, substantially,
the surface moisture content and temperature, produc- At this point, the vapor concentration dierence is
ing vapor ow inversion. This unphysical, ow inversion written as a linear combination of relative humidity
occurs because the mass ow resistance towards the dierence and the dierence between the quotient of
N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518 513

saturation pressure by temperature. This quotient can be However, the inuence of the previous iteration is still
calculated by a rst-degree function with small errors for expressive, since there are temperature terms in the mass
a certain range of temperatures. For increasing accu- conservation equation and vice versa. This fact inspired
racy, this is performed, introducing a residual function, a modication in the solution method, which was im-
RT , related to temperature as proved to calculate, simultaneously, the temperature
Ps and moisture content proles. This is shown in the fol-
RT AT B; 16 lowing section.
T
where A and B are the straight-line coecients from
the approximation Ps =T AT B. The ratio Ps =T is 5. A generic algorithm to solve highly coupled governing
calculated by the saturation pressure expression from equations
ASHRAE [12].
Therefore, the dierence between pressure and tem- The development of a generic algorithm, to solve
perature quotients of Eq. (17) may be written as heat and mass transfer coupled equations, emerged from
the need of obtaining all the dependent variable proles
Ps;1 Ps s simultaneously at a given time step avoiding numerical
AT1 T s RT1 RT s: 17
T1 T s divergence caused by the evaluation of coupled terms
from previous iteration values. In this way, the method
Similarly, it is possible to dene a residual function re-
becomes stable and closer to the nature of the physical
lated to moisture content dierence Rhprev s, then
phenomenon of combined heat and mass transfer in
the term associated to the relative humidity of Eq. (15),
porous media.
is redened as
Discretization of conservation equations in the
 prev
M Ps s M Ps s physical domain gives the following algebraic equations:
/1 /s
R T s R T s Ai  xi Bi  xi1 Ci  xi 1 Di ; 21
 
o/
 h1 hs Rhprev S ; where x is a vector containing the dependent variables.
oh In the present case
 
18 h
xi i 22
Ti
where the Rhprev
0 is the residue calculated by the sorp-
tion isotherm, evaluated at the previous iteration. and, dierently from the traditional TDMA, coecients
In this way, the above equation gives A, B and C are the second-order tensors. Vector xi can
  be expressed as a function of xi1
M Ps s
/1 /s
R T s xi Pi  xi1 qi ; 23
 prev  prev
M Ps s o/ where Pi is, now, a second-order tensor.
h1 hs
R T s ohs Replacing Eq. (23), evaluated at point i 1 in (21),
 prev the following equation is obtained:
M Ps s
Rhprev S: 19 Ai Ci  Pi 1 xi Bi  xi1 Ci  qi Di : 24
R T s 1

Finally, replacing Eqs. (17) and (19) in the vapor con- Writing Eq. (24) in an explicit way for xi
centration dierence expression i
xi Ai Ci  Pi 1 1  Bi  xi1
qv;1 qv s M1 T1 T s M2 h1 hs M3 ; Ai Ci  Pi 1 1 Ci  qi 1 di : 25
20
Thus, a comparison between Eqs. (25) and (23) gives the
following new recursive expressions:
where i
 prev  prev Pi Ai Ci  Pi 1 1  Bi 26
M M Ps s o/
M1 A /; M2 ;
R R T s ohs and
 prev
M Ps s qi Ai Ci  Pi 1 1 Ci  qi Di : 27
M3 Rhprev s /1 RT1 1
R T s
 The use of this new algorithm is illustrated in the sec-
RT prev s : tions below, by developing the new mathematical
method.
514 N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518

6. The new heat and mass transfer calculation method for simulating building walls that are largely used in hot
and humid countries.
The new method was derived from the traditional The basic dry-basis material properties are given in
one by using the new algorithm and the linearization of Table 1. In this table open porosity is the ratio of the
vapor concentration dierence, to solve simultaneously volume of open pores (i.e., pores with openings that
the governing equations on a very stable way. Thus, for have a path to both wall surfaces) to the total volume.
this new method, the coecients for internal points are: The available material data gathered from Perrin [15]
2 Dx D allow all the transport coecients to be modeled as a
dxhee Ddxhww
4
Dt
  function of moisture content. Figs. 13 show the prop-
Ai
Lql Ddxhvee Ddxhvw erty data for lime mortar as a function of moisture
w

DTe
3 content. Considering the temperature range of interest in
dxe
DdxTww building applications, temperature dependence of lime
  5;
ke
LDdxTve dxkww LDdxTvw q0 cm Dx mortar properties was, here, neglected when compared
dxe Dt
2 D
e w
3 to their dependence with moisture content.
D
dxe
he
dx
Te
In Fig. 1, vapor and total (liquid plus vapor) trans-
Bi 4 e  5; 28 port coecients for lime mortar are shown.
Lql Ddxhvee ke
dxe
LDdxTve
e
2 Dhw DTw
3
dxw dxw
Ci 4   5;
Table 1
Lql Ddxhvw kw
dxw
LDdxTvw
w w Dry-basis material properties for lime mortar
" #
Dx o
h
Dt P q0 kg=m3 2050.0
Di
q0 cm Dx To k (W/m K) 1.96
Dt P
cm (J/kg K) 950.0
and for the boundary volume S, Open porosity 0.18
" Dx Dhe
2Dt
dx M2 hqml
AS Lq D e
l hve
dxe
LM2 hm
DTe
#
dxe
M1 hqml
;
Dx
q0 cm 2Dt dxkee Lqdx
l DTve
h LM1 hm
2 D D
e
3
he Te
dxe dx
BS 4 Lq D e  5;
29
l
dxe
hve ke
dxe
Lqdx
l DTve
e
 
0 0
CS ;
0 0
2 hm Dx o
3
ql
M1 T1 M2 h1 M3 2Dt h S
6 7
DS 4 hT1 a1 dS; Lqr Lhm M1 T1 5:
Dx o
M2 h1 M3 q0 cm 2Dt T S
Fig. 1. Mass transport coecients for lime mortar.
This new method is much less dependent on the T and h
elds of the previous iteration. Nevertheless, it is still
necessary to make iterations because transport coef-
cients and residues are calculated by using the previous
T and h elds. However, this does not have any signif-
icant inuence on the solution of the system of equations
by the new mathematical method.

7. Results

7.1. Material properties

In order to show the numerical performance of the


proposed method, material property data are provided Fig. 2. Thermal conductivity for lime mortar.
N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518 515

Fig. 3. Sorption isotherm curve for lime mortar.


Fig. 4. Convergence error as a function of Bim , for
Fom 7e 4.
It is possible to see from Fig. 1 that the coecient  
responsible for the ow of liquid due to a temperature hs hprev
s
d abs :
gradient (DTl ) is very small compared to the one for hprev
s
vapor since the dierences between the curves of DT and
DTv are very small. Fig. 4 illustrates the inuence of the Biot and Fourier
Fig. 2 presents the thermal conductivity curves for numbers on the convergence error function.
both materials. In these curves, the vapor diusion and It is noted, for the traditional method, as the Bim
phase change eects were not considered, which means increases the error function goes up very quickly show-
that they show the value for pure thermal conductivity ing a high sensitivity to this dimensionless number,
for heat transfer conduction in Fourier`s law. which can be even higher than that caused by the mass
Fig. 3 presents the sorption isotherm curve for lime Fourier number.
mortar. This curve is the average between adsorption The mass conservation equation at the boundaries
and desorption curves. can be written as
 
1
7.2. Simulation results: comparative analysis 1 hs
2Fom
hAnt s qv;1 qv s
Comparative studies between the traditional and the hs 1 Bim
2Fom ql
new methods are presented. First, the inuence of di- DT prev prev
mensionless numbers on the stability of the traditional T s T 1:
Dh
method is analyzed. Then, comparisons are made in
terms of vapor ow, temperature, moisture content, heat The system of equations is stable when the left-hand
ux and time averaged heat ux through a porous term is much higher than the source terms (from the 2nd
building wall for both methods. At the end of this sec- to the 4th right-hand terms). The last right-hand term is
tion, the inuence of high time-step on the performance normally lower than the 2nd and 3rd ones. As the sec-
of the new method is described. ond right-hand term is comparable to the rst parcel at
the right-hand side of the equation above, it is believed
7.2.1. Dimensionless numbers inuence that a good stability criterion, dependent on the
A dimensionless number that could numerically ex- boundary condition, but independent on the Fourier
press when numerical instability arises is the Biot num- number, could be proposed as
ber for moisture diusion (Bim ). This number can be qv;1 qv s
written as Bim  1;
ql
hm L
Bim ; which can be satised for materials with high Dh or for
Dh
very low permeation surfaces, under any boundary
which means, for high Bim numbers, that wall higric condition value. However, as it can be seen in Fig. 4, the
resistance is much higher than the corresponding surface higher the Fom number the lower the Bim number to
resistance. avoid numerical divergence.
Another important dimensionless number is the mass
Fourier number, Fom Dh Dt=Dx2 , expressing a mesh- 7.2.2. Comparison between the traditional and the new
size parameter that could be related to a convergence methods
error function d, dened as the relative dierence be- In order to illustrate the vapor ow inversion prob-
tween the wall surface moisture content values lem and the advantages of the new method, a symmetric
516 N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518

problem was rstly studied, when a lime mortar wall,


initially with a moisture content of 3% in volume and
temperature 20C, was put in contact with air at a
temperature of 28C and 80%, at both wall surfaces.
Solar radiation was ignored. Considering its equilibrium
value, average wall moisture content increases mono-
tonically, with time, toward its equilibrium value.
Fig. 5 shows, at the very beginning of the simulation,
in the rst time step, the vapor ow at the surfaces of a
100-mm lime mortar wall for each iteration. Three it-
erations were needed, in the new method, to reach
convergence. However, for the traditional method,
convergence was not reached since the Fom and Bim Fig. 6. Vapor ow at the surfaces of a 100-mm lime mortar wall
numbers were too high to avoid numerical divergence. for 30-s time step.
This non-physical behavior can be explained by consid-
ering the numerical algorithm related to mass conser-
vation equation, in traditional a method, that produces a In Fig. 6, a good agreement for vapor ow calcula-
very high quantity of moisture that comes to the wall tion by both methods is also indicated when a 30-s time
surface. As the wall higric resistance is much higher than step is used. Nonetheless, by looking at the second it-
the air lm resistance, moisture content at the surface eration, it is remarkable that the new method still shows
increases substantially. In the sequence, energy balance a better numerical performance.
equations are solved and the wall may lose energy to the As material properties vary signicantly with moist-
air, especially if the source term was calculated with the ure content and residual functions are calculated with
new values of moisture content. If not, for the next it- previous iteration values of temperature and moisture
eration, the vapor concentration dierence at the wall content, it is necessary to make at least a single iteration
surface might be higher than that of the air, inverting when the new method is used, however convergence is
vapor ow and preventing a stable solution. rapidly achieved.
Fig. 6 shows the same example illustrated in Fig. 5, Fig. 7 displays the temperature at the wall surface.
for a 30-s time step. In this case, both models behave It is noticed, for a high time step, high uctuations with
similarly showing a dierence of 0.126% for the 1st it- the traditional method even for temperature, due to the
eration. After the third iteration, this dierence re- latent term in the energy balance at the boundaries. It is
mained constant at the value of 0.032%, which is very perceived again that it would not be necessary to make
satisfactory. more than three iterations by using the new method.
Therefore, the traditional method can only be used In Fig. 8, it is shown how the wall temperature dis-
for small time steps or small Fom numbers. The Fourier tribution behaves for each iteration, at the very begin-
mass number decreased 125 times from the one used for ning of the simulation, with the new method. It is
Fig. 5 to that for Fig. 6, allowing convergence for the remarkable how fast convergence is attained, by using
same Bim number. Actually, this is what many con- this method.
ditionally stable method programs do to avoid diver- Fig. 9 presents the same temperature proles, ob-
gence problems or even physically unrealistic solutions. tained by the traditional method. It can be observed that

Fig. 5. Vapor ow at the surfaces of a 100-mm lime mortar wall Fig. 7. Temperature at the surfaces of a 100-mm lime mortar
for 1-h time step. wall for 1-h time step.
N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518 517

In conclusion, for high moisture diusion Biot


numbers and high time steps or high Fom numbers, the
use of the traditional heat and moisture transfer calcu-
lation method will, likely, experience serious conver-
gence problems.
Although, as was shown above, the new method al-
lows to adopt any time step without numerical stability
concerns, accuracy problems related to high time-step
persist. In order to evaluate the errors associated to the
time step magnitude, simulation results with the new
method are presented for porous walls of air-con-
Fig. 8. Temperature proles obtained by using the new ditioned buildings, in terms of heat ux. A 100-mm thick
method. wall was simulated, without any additional mass ow
resistance such as painting or vapor barriers, rep-
resenting a critical situation for heat and moisture
transfer.
The weather le used in the simulations was the
Singapore Typical Meteorological Year (TMY). Data
for solar radiation and internal psychrometric con-
ditions were gathered from Mendes [13]. Internal tem-
perature and relative humidity were obtained by using
the DOE-2 building simulation program [14].
It was found for the temperature proles on a very
critical day of 14 January at 1:00 am (not shown) a
largest dierence of 0.7C between simulations with time
steps of 1 and 24 h.
Fig. 9. Temperature proles obtained by using the traditional Fig. 11 shows the dierences on yearly integrated
method. heat ux (sensible and latent) for the climate of Singa-
pore, relative to 1-s time-step simulations. It is observed,
in Fig. 11, that the use of 1-day time step gives a 10.9%
the temperature prole goes up and down at each dierence when compared to the 1-s time-step simula-
iteration and does not stabilize. The even (or odd) iter- tion, which is very reasonable for estimating the initial
ations have about the same temperature proles conditions by a previous simulation. In fact, when a
between themselves due to the vapor ow ``come-and- building porous envelope was simulated in Singapore
go'' problem. with 24-h time step for the rst year and 1-h time step
Fig. 10 shows the evolution in time of temperature at for the second year, a dierence of 0.05% was found in
the wall surfaces for both models with a 30 s time step. the calculation of integrated yearly heat ux, in com-
At each 50 iterations, a time step is incremented. In this parison with simulations with 1 h time step for both
case, it is noted that surface temperature evolution in years. It is, also, noticed that the use of time steps higher
time, for both models, is almost identical. than 1 h would be suitable in situations where hourly

Fig. 10. Evolution in time of temperature at the surfaces of a Fig. 11. Dierences of yearly integrated heat ux calculation
lime mortar wall for a 30-s time step. relative to 1-s time step simulations.
518 N. Mendes et al. / International Journal of Heat and Mass Transfer 45 (2002) 509518

information is not needed or available and/or when [3] D.M. Burch, W.C. Thomas, An Analysis of Moisture
computer running time plays an important role. Accumulation in Wood Frame Wall Subjected to Winter
Therefore, it is easy to see that the use of 1-h time Climate, NISTIR 4674, National Institute of Standards
step simulations with the new method (unfeasible using and Technology, Gaithersburg, 1991.
[4] R. El Diasty, P. Fazio, I. Budaiwi, Dynamic modeling of
the traditional one) allows to reach high accuracy,
moisture absorption and desorption in buildings, Build.
although with a computer running time reduction Environ. 28 (1993) 2132.
of nearly 1:3000. [5] R.J. Liesen, Development of a response factor approach
for modeling the energy eects of combined heat and mass
transfer with vapor adsorption in building elements, Ph.D.
8. Conclusions Thesis, Mechanical Engineering Department, University of
Illinois, Chicago, IL, 1994.
Presently proposed method has shown a considerable [6] F.W.H. Yik, C.P. Underwood, W.K. Chow, Simultaneous
progress when compared to traditional calculation modeling of heat and moisture transfer and air-condition-
methods, when a physical problem is described by ing systems in buildings, in: Proceedings of the 4th IBPSA
International Building Performance Simulation Associ-
strongly coupled equations.
ation Conference, Madison, WI, 1995.
The new method was applied to the problem of heat [7] IEA (International Energy Agency) Annex 24 Final
and moisture transfer through porous walls, by solving Report, Heat, Air, and Moisture Transfer in Insulated
the governing equations, simultaneously and by con- Envelope Parts, vol. 1, Task 1: Modelling, Belgium, 1996.
sidering the vapor exchanged between the wall and the [8] S. Patankar, Numerical Heat Transfer and Fluid Flow,
air as a linear function of temperature and moisture Hemisphere, New York, 1980.
content dierences. [9] C.R. Pedersen, A transient model for analyzing the
The use of this new method avoids numerical oscil- hygrothermal behavior of building constructions, building
lations, since it keeps the discrete equations strongly and simulation'91, in: Proceedings of the 3rd IBPSA
coupled between themselves, preventing the occurrence International Building Performance Simulation Associ-
ation Conference, France, 1991.
of physically unrealistic behavior when time step is in-
[10] D.M. Burch, J. Chi, MOIST: A PC Program for Predicting
creased and producing an unconditionally-stable nu- Heat and Moisture Transfer in Building Envelopes,
merical method, which is very suitable to be used in Release 3.0, NIST Special Publication 917, National
building yearly energy simulation programs. Institute of Standards and technology, Gaithersburg, 1997.
[11] J.R. Philip, D.A. DeVries, Moisture movement in porous
materials under temperature gradients, Trans. Amer.
Acknowledgements Geophys. Union 38 (2) (1957) 222232.
[12] ASHRAE, ASHRAE Handbook Fundamentals, Amer-
ican Society of Heating, Refrigerating and Air-Condition-
The authors thank CNPq Conselho Nacional de ing Engineers, Atlanta, 1993.
Desenvolvimento Cientico e Tecnol
ogico of the [13] N. Mendes, Models for prediction of heat and moisture
Secretary for Science and Technology of Brazil for transfer through porous building elements, Ph.D. Thesis,
support of this work. 225, Federal University of Santa Catarina (UFSC),
Florian opolis, SC, Brazil, 1997 (in Portuguese).
[14] B.E. Birdsall, W.F. Buhl, K.L. Ellington, A.E. Erdem,
References F.C. Winkelmann, DOE-2 Basics version 2.1E, Energy
and Environment Division, Lawrence Berkeley National
[1] M.J. Cunningham, The moisture performance of framed Laboratory, University of California Berkeley, CA, 1994.
structures: a mathematical model, Build. Environ. 23 [15] B. Perrin, Etude des transferts couples de chaleur et de
(1988) 123135. masse dans des materiaux poreux consolides non satures
[2] A. Kerestecioglu, L. Gu, Incorporation of the eective utilises en genie civil, These Docteur d'Etat, Universite
penetration depth theory into TRNSYS, Draft Report, Paul Sabatier de Toulouse, Toulouse, France, 1985.
Florida Solar Energy Center, Cape Canaveral, FL, 1989.

Das könnte Ihnen auch gefallen