Anda di halaman 1dari 9

Journal of Computational Physics 230 (2011) 8304–8312

Contents lists available at SciVerse ScienceDirect

Journal of Computational Physics


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

Are upwind techniques in multi-phase flow models necessary?


C.-H. Park a,b,⇑, N. Böttcher c, W. Wang a, O. Kolditz a,d
a
Department of Environmental Informatics, Helmholtz Centre for Environmental Research, UFZ, Permoserstr. 15, 04318 DE, Germany
b
Department of Geothermal Resources, Korea Institute of Geoscience and Mineral Resources (KIGAM), Gwahang-no 92, Yuseong-gu,
Daejeon 305-350, Republic of Korea
c
Institute for Groundwater Management, TU Dresden, Helmholtzstr. 10, Dresden 01069, DE, Germany
d
Applied Environmental System Analysis, TU Dresden, Helmholtzstr. 10, Dresden 01069, DE, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Two alternatives of primary variables are compared for two-phase flow in heterogeneous
Received 1 November 2010 media by solving fully established benchmarks. The first combination utilizes pressure of
Received in revised form 21 June 2011 the wetting fluid and saturation of the non-wetting fluid as primary variables, while the
Accepted 28 July 2011
second employs capillary pressure of the wetting fluid and pressure of the non-wetting
Available online 9 August 2011
fluid. While the standard Galerkin finite element method (SGFEM) is known to fail in the
physical reproduction of two-phase flow in heterogeneous media (unless employing a fully
Keywords:
upwind correction), the second scheme with capillary pressure as a primary variable with-
Two-phase flow
Galerkin finite element methods
out applying an upwind technique produces correct physical fluid behaviour in heteroge-
Upwind schemes neous media, as observed from experiments.
Ó 2011 Elsevier Inc. All rights reserved.

1. Introduction

Regardless of the numerical modelling approach to simulate multi-phase flow, selection of primary variables must be
made to solve the underlying governing equations. Then, under the choice made, the governing equations are discretized
by local mass conservative methods such as finite difference methods and finite volume methods or by global mass conser-
vative methods such as finite element methods. In general, the importance of using local mass conservative numerical meth-
ods are emphasized in the area of fluid such as petroleum reservoir engineering, geothermal studies, and groundwater
literature [4,3,10,20].
A comprehensive review on the selection of primary variables for modelling multi-phase flow can be found in [25] focus-
ing on the local mass conservative methods. The majority of previous research has utilized pressure and saturation as pri-
mary variables with an upwind technique [1,2,6,8,9,11,14,19]. Helmig and Huber [8] have also reported that, without a fully
upwind technique, multi-phase models fail to accurately reproduce flow behaviour in heterogeneous media, particularly
with the Brooks–Corey model used for relative permeability and capillary pressure. Interestingly, all the models mentioned
above are local mass conservative and have utilized one phase pressure along with a single saturation of either phase as
primary variables.
Differently from the typical choice, the importance of capillary pressure in heterogeneous media with respect to the
choice of primary variables was the motivation of our work [15,16,21], particularly when the continuity of capillary pressure
at the interface between the two different media is important. In this paper, we conduct a comparative study of primary var-
iable selection in modelling two-phase flow in non-deformable porous media, and compare the results of two simple,

⇑ Corresponding author at: Department of Geothermal Resources, Korea Institute of Geoscience and Mineral Resources (KIGAM), Gwahang-no 92,
Yuseong-gu, Daejeon 305-350, Republic of Korea.
E-mail address: chanhee.park@kigam.re.kr (C.-H. Park).

0021-9991/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcp.2011.07.030
C.-H. Park et al. / Journal of Computational Physics 230 (2011) 8304–8312 8305

reproducible and well-known benchmarks and one experiment [7,14,17]. We also explain why the upwind technique is re-
quired with association of the pressure and saturation choice in the standard Galerkin finite element method (SGFEM). All
simulations conducted in this paper utilize SGFEM with no upwind weighting technique. The open source scientific software
OpenGeoSys for the simulation of coupled thermo-hydro-mechanical-chemical processes in porous media [12,18,24] is used
for the study.

2. Governing equations

Governing equations for isothermal two-phase flow consist of the continuity equation, the momentum equation (Darcy’s
law), the relationship between capillary pressure and phase saturation, and the equation that summates the saturation of
phases to be unity. Depending on the choice of primary variables the closed form of the continuity equation can be expanded
differently. The continuity equation of each phase (a = w, nw. The indices w and nw specify wetting or non-wetting phase
respectively.) can be defined by
@ð/Sa qa Þ
¼ O  ðqa v a Þ þ Xa ; ð1Þ
@t
where / is porosity, Sa is saturation, qa is density, Xa is a source or sink term, and va is the Darcy velocity for phase a defined
by
kr a
va ¼  k  ðOPa  qa gÞ; ð2Þ
la
where kra is relative permeability, la is viscosity, Pa is pressure for phase a, k is intrinsic permeability tensor, and g is the
gravitational vector.
Inherently for saturation, the sum of partial saturations in pore space is
X
Sa ¼ 1: ð3Þ
a

Assuming that relative preference (i.e., wettability) of the fluid to the media exists and is not negligible, the capillary pres-
sures relation for a two-phase system is defined over a representative elementary volume (REV) by
Pc ¼ P nw  P w ; ð4Þ
where Pc is capillary pressure, Pnw is pressure for the non-wetting phase fluid, and Pw is for the wetting phase fluid.

3. Primary variables

The overwhelming choice for primary variables for two-phase flow is to have one phase pressure and a single saturation
from either phase [8,14,23]. For simplicity, the expanded continuity equations with the Darcy velocity embedded are, in this
paper, provided based on the assumption that both phase fluids are incompressible. Extending from single- to two-phase
merely requires the inclusion of a second continuity equation for the second fluid,
 
@Sw @Snw krw
/ ¼ / ¼O k  ðOPw  qw gÞ þ Xw ; ð5Þ
@t @t l
 w 
@Snw @Sw krnw
/ ¼ / ¼O k  ðOðPc þ Pw Þ  qnw gÞ þ Xnw : ð6Þ
@t @t lnw
The advantage for this scheme comes from easy handling of capillary pressure, which occurs within the spatial, rather
than temporal, derivative. While Eq. (5) is an elliptic equation for pressure, Eq. (6) is a first order equation for saturation
as a primary variable, thus shows hyperbolic nature. Note that to deal with heterogeneity most of numerical models based
on this scheme use the upwind weighting technique to prevent numerical difficulties of the combination of Eqs. (5) and (6).
Capillary pressure can be obtained secondarily from the primary variable. Helmig and Huber [8] have reported that even
using full upwinding under SGFEM, it is impossible to reproduce the discontinuity of saturation at the interface between dif-
ferent media.
An alternative choice for primary variables is to utilize capillary pressure as unknown of the wetting phase mass balance
Eq. (1) and phase pressure of the non-wetting phase [15,16,21] for Eq. (1). Thus, capillary pressure appears in the time deriv-
ative term both in the wetting and non-wetting continuity equations
 
dSw @Pc krw
/ ¼O k  ðOðPnw  Pc Þ  qw gÞ þ Xw ; ð7Þ
dPc @t l
w 
dSnw @Pc krnw
/ ¼O k  ðOPnw  qnw gÞ þ Xnw : ð8Þ
dP c @t lnw
8306 C.-H. Park et al. / Journal of Computational Physics 230 (2011) 8304–8312

An advantage of this choice is that capillary pressure is approximately continuous in points or cells of the given numerical
methods. This is particularly important in heterogeneous media and when continuity of capillary pressure across elemental
boundaries becomes critical, and is further discussed below. In contrast to the first choice of the primary variables, both Eqs.
(7) and (8) are elliptic equations for both of primary variables.
Since both choices of primary variables are discretized under SGFEM, neither of the choice is local mass conservative. This
is because SGFEM multiplies the mass balance equation by the weighting function of choice and further minimize or force
the resulting expression to be zero in the process of SGFEM, while FDM or FVM uses either approximations to the terms or
the integral form of the mass conservation equation as the starting point. However, SGFEM guarantees mass conservation
globally over the solution domain. So does FDM or FVM.

4. Numerical solution scheme

SGFEM is applied for the numerical solution of the above partial differential equations (PDEs) resulted in a system of alge-
braic equations. The first choice will be referred to as the PwSnw scheme and the other choice as the PcPnw. After applying
SGFEM the corresponding ordinary differential equations for the two schemes are straightforward, in a global-implicit form,
  ( )  ( )  
CPw Pw CPw Snw d bw
P KPw Pw KPw Snw bw
P ^r Pw
þ ¼
CSnw Pw CSnw Snw dt b S nw KSnw Pw KSnw Snw b
S nw ^r Snw ð9Þ
|fflfflfflfflfflfflfflfflfflfflfflfflfflffl
ffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflffl{zfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflffl{zfflfflfflffl}
C x K r

for the PwSnw scheme and


  ( )  ( )  
CP c P c CPc Pnw d bc
P KPc Pc KPc Pnw bc
P ^r Pc
þ ¼
CPnw Pc CPnw Pnw dt b nw
P KPnw Pc KPnw Pnw b nw
P ^r Pnw ð10Þ
|fflfflfflfflfflfflfflfflfflfflfflfflfflffl
ffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflffl{zfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflffl{zfflfflfflffl}
C x K r

for the PcPnw scheme. The expressions of sub-matrices and sub-vectors of Eqs. (9) and (10) are given in detail in (A). For time
discretization the generalized implicit finite difference scheme
h i h i
n
Cnþ1 nþ1
¼ CMt  ð1  hÞKnþ1 xn þ r
Mt
þ hKnþ1 x ð11Þ

is applied with convergence of the nonlinear equations obtained iteratively by the Picard method. Since the global matrix
formed by this discretization is unsymmetrical, we have used ILU preconditioners with iterative solvers. In the case of con-
vergence failure for iterative linear solvers, the direct solver PARDISO [5,22] has been successfully used to solve the two-
phase flow problem. A direct solver is often necessary for solving non-linear problems with global-implicit schemes.

5. Benchmarks

As briefly stated in the introduction, we investigate the necessity of upwind schemes in numerical models for multi-phase
flow to represent the deterministic nature of two-phase flow in heterogeneous media through benchmarks. In the section of
the benchmarks, we have selected carefully test examples prone to the determination of these flow characteristics.

5.1. McWhorter–Sunada problem

McWhorter and Sunada [17] derived the analytical solution of two-phase flow that includes the term for capillarity in the
governing equation. The analytical solution has been used frequently to verify a variety of numerical models for two-phase
flow. In this paper, problems are configured to compare the two schemes (PcPnw and PwSnw) against the analytical solutions.
The benchmark is to test the fluid behaviour at the interface of two different porous media (medium L and R). Initially, each
medium has a different ratio of two immiscible fluids to generate flow due to pressure gradient generated by capillarity across
the interface of the two media under various different wetting conditions. Medium L expands infinitely to the left, while med-
ium R does to the right. Two fluids of phases w and nw reside in medium L and R. Initially, the wetting phase saturation in
medium L differs from medium R ðSLw – SRw Þ. Thus, a flux of two fluids is generated to equilibrate saturation in the left and right
domain without external forces under different medium conditions (Fig. 1). Fluid properties of the two synthetic fluids but
with same physical properties are specified in Table 1 and hydraulic properties of three different sands for the Brooks–Corey
model are summarized in Table 2. When the two media have the same properties in Fig. 2 (homogeneous), both schemes are
comparable in the prediction of saturation and demonstrate good agreement with the analytical solution.
In contrast, a striking difference right at the interface of the media appears between the two schemes, when the medium
properties of mediums L and R are different. From Figs. 3 and 4, it can clearly be seen that there exists one or two saturation
values at the interface. According to the analytical solution, depending on either side of medium approaching, the saturation
value must be different at the interface. However, the PwSnw scheme calculates one saturation value at the interface
regardless of the different medium properties.
C.-H. Park et al. / Journal of Computational Physics 230 (2011) 8304–8312 8307

Fig. 1. Initial saturations in media L and R.

Table 1
Fluid properties for the wetting and non-wetting fluids.

Property Unit Wetting fluid Non-wetting fluid


3
Density kg/m 1000 1000
Viscosity Pa  s 103 103

Table 2
Hydraulic properties of Sands A, B, and C for the Brook-Corey model.

Property Parameter Sand A Sand B Sand C


Entry pressure pd (Pa) 5000 5000 3700
Distribution index k() 2 2 1.8
Residual saturation Swr() 0.0 0.0 0.0
Permeability k(m2) 1013 1013 1014
Porosity /() 0.40 0.30 0.30

1
wetting-phase saturation Sw[-]

A A
homogeneous medium
0.8

0.6

0.4

analytical solution
0.2 the PcPnw scheme
the PwSnw scheme

0
-0.4 -0.2 0 0.2 0.4
x[m]

Fig. 2. Saturation profiles of wetting phase after t = 10,000 s in a homogeneous medium. Initial saturations of wetting phase are SLi ¼ 0:2 and SRi ¼ 0:8.

Mathematically speaking, the interface of two media is not a place where calculation of the derivative of dPc/dSw is pos-
sible due to two different relationships of capillary pressure and saturation across the interface. From the typical relationship
of capillary pressure and saturation, e.g. Brooks–Corey and van Genuchten functions in Fig. 5, the saturation is discontinuous
at the interface unless two relationships are exactly same (i.e., homogeneous media). Therefore, depending on the approach-
ing direction to the interface, the interface is not differentiable to make the derivative of dPc/dSw due to a break by two dif-
ferent saturation values in heterogeneous media. However, both schemes have dPc/dSw either explicitly or implicitly in the
governing equation.
The PwSnw scheme has this term implicitly in the spatial derivative term of Eq. (6), while the PcPnw scheme has explicitly in
the time derivative term of (7) and (8). From the example of Figs. 3 and 4, the interface is space-related at 10,000 s. Without
telling the approaching direction spatially, a typical averaging for the value of dPc/dSw would lead to non-physical flow
8308 C.-H. Park et al. / Journal of Computational Physics 230 (2011) 8304–8312

wetting-phase saturation Sw[-]


C A
nC ≠ nA
0.8
KC ≠ KA

0.6 pd,C ≠ pd,A

λC ≠ λA
0.4

analytical solution
0.2 the PcPnw scheme
the PwSnw scheme

0
-0.4 -0.2 0 0.2 0.4
x [m]

Fig. 3. Saturation profiles of wetting phase after t = 10,000 s at the interface between two materials C and A, which have different porosities and
permeabilities as well as different retention curves.

1
wetting-phase saturation SW[-]

nC = nB C B
0.8 KC = KB

pd,C ≠ pd,B
0.6
λC ≠ λB

0.4

analytical solution
0.2 the PcPnw scheme
the PwSnw scheme

0
-0.4 -0.2 0 0.2 0.4
x [m]

Fig. 4. Saturation profiles of wetting phase after t = 10,000 s at the interface between two materials C and B, which have different porosities and
permeabilities identical retention properties.

Fig. 5. Relationships of capillary pressure and the wetting fluid saturation: Brooks–Corey and van Genuchten models.

behaviour. This is where the fully upwind scheme does the job in the PwSnw scheme. Unfortunately, no standard FEM-based
model has successful implementation of this upwind scheme in simulating two-phase flow in heterogeneous media, yet.
C.-H. Park et al. / Journal of Computational Physics 230 (2011) 8304–8312 8309

Fig. 6. Configuration of heterogeneous media in parallel-plate cell.

Table 3
Hydraulic properties of sands for the Brooks–Corey model.

Properties pd (Pa) k() Swr () k (m2) / ()


1 369.73 3.86 0.078 5.04  1010 0.4
2 434.45 3.51 0.069 2.05  1010 0.39
3 1323.95 2.49 0.098 5.26  1011 0.39
4 3246.15 3.30 0.189 8.19  1012 0.41

On the contrary, the PcPnw scheme handles the proper direction of this phenomenon correctly, as demonstrated in Figs. 3
and 4. As summarized in Table 2, the variation of the medium properties are porosity, permeability, entry pressure and k in
Brooks–Corey’s model. This behaviour is obvious, particularly when the entry pressure of the media at the interface is dif-
ferent. Both numerical results are compared to the analytical solution [7].

5.2. Kueper–Frind problem

In addition to the former benchmarks, both schemes are further tested for model validation with an experiment chosen to
prove experimental observation against numerical solutions. Kueper and Frind [14] developed a model to simulate their
experiment for DNAPL penetration [13]. In their experiment, a 60-cm-high  80-cm-wide  0.6-cm-thick parallel-plate
glass-lined cell was packed with four types of sands and saturated fully with water. The configuration of the assembled sand
lenses and the two sets of the boundary conditions for the PwSnw and PcPnw schemes are illustrated in Fig. 6. As for the rela-
tions between relative permeability and saturation as well as capillary pressure and saturation, the authors have used the
Brooks–Corey model. The properties of sands for the Brooks–Corey model are measured experimentally and summarized
in Table 3. The numerical solutions obtained from the PwSnw scheme and PcPnw schemes for the benchmark of [14] are com-
pared in Fig. 7.
Both schemes produce DNAPL plume propagation physically until the plume reaches the heterogeneous interface be-
tween the topmost permeable layer and the lower permeability zone beneath. A striking difference occurs at this interface.
While the PwSnw scheme simulates the plume to infiltrate into the less permeable medium, the PcPnw scheme allows the
plume to bypass this zone generating pressure build-up at the interface. A similar experiment and simulation comparison
against experimental observation are also conducted by Helmig and Huber [8]. The authors have reported unphysical fluid
behaviour captured by the PwSnw scheme: a phenomenon that can be avoided with a fully upwind technique in the box
model [8].
8310 C.-H. Park et al. / Journal of Computational Physics 230 (2011) 8304–8312

Fig. 7. Comparison of the results obtained from the both schemes. The right-hand column shows a good agreement with observed distribution of DNAPL of
the experiment.

6. Conclusions

In this paper, we have compared the PwSnw and PcPnw primary variable schemes through several benchmarks designed to
show the deterministic nature of two-phase flow in heterogeneous media. The schemes show a striking difference in the
numerical results when entry pressure is different in heterogeneous porous media. To prevent non-physical flow behaviour
at material interfaces, those approaches that use the PwSnw scheme must be corrected with upwind techniques or an appro-
priate selection of interpolation functions to maintain continuity of capillary pressure across the heterogeneous interface to
achieve accuracy. On the other hand, the PcPnw scheme does not require any upwinding at all.
We have demonstrated the importance of primary variable selection in modelling two-phase flow. This is due to the dif-
ferent handling of the derivative term that represents the relation of capillarity and saturation. Handling the term temporally
does have an advantage in estimating the two different values for saturation at the same point of material interfaces in het-
erogeneous porous media.
Although local mass conservative discretization methods are preferred in the previous research [3,25], the benchmark
results obtained using global mass conservative discretization methods in this paper clearly indicate that continuity of cap-
illary pressure across the heterogeneous interface is more important in achieving accurate solution of two-phase flow in
C.-H. Park et al. / Journal of Computational Physics 230 (2011) 8304–8312 8311

heterogeneous media. The use of the PcPnw scheme as primary variables maintains this continuity naturally. Since there is no
additional code required for the upwind technique, the PcPnw scheme even can lead to improved efficiency. This is important
if two-phase flow is being coupled to other processes such as consolidation and heat transport. To the authors’ knowledge
[8], this paper presents the first numerical solutions for the benchmarks obtained purely from the standard Galerkin finite
element method. Before this work, none of the attempts using the standard Galerkin finite element method have reported
success in achieving accurate solutions in heterogeneous media.

Acknowledgements

We acknowledge the funding by the German Federal Ministry of Education and Research (BMBF) within the framework of
the CO2-MoPa and the CLEAN joint projects as parts of the Special Program GEOTECHNOLOGIEN and by Basic Research Pro-
gram at Korea Institute of Geoscience and Mineral Resources. We thank the anonymous reviewers for their fruitful reviews
based on which the manuscript was substantially improved.

Appendix A. Matrix and Vector of FEM

The matrices and vectors in Eqs. (9) and (10), which arises from the finite element discretization of the weak forms of the
governing equations, are given hereafter.
Z
CP w P w ¼ /N T N dX; ðA:1Þ
X
CPw Snw ¼ 0; ðA:2Þ
CSnw Pw ¼ 0; ðA:3Þ
Z
CSnw Snw ¼  /NT N dX; ðA:4Þ
Z X
krw
KPw Pw ¼ ðrNÞT k rN dX; ðA:5Þ
X lw
KPw Snw ¼ 0; ðA:6Þ
KSnw Pw ¼ 0; ðA:7Þ
Z
krnw
KSnw Snw ¼ ðrNÞT k rN dX; ðA:8Þ
X lnw
Z Z Z
krw
^rPw ¼ Xw N dX  ðrNÞT  k  qw gdX  qw N dC; ðA:9Þ
lw C
ZX ZX
T krw
^rSnw ¼ Xnw NdX  ðrNÞ  k  qnw g dX
X X lw
Z Z
krnw
þ ðrNÞT   rPw dX  qnw N dC: ðA:10Þ
X lnw C

where N is the shape function vector of N ¼ ðN 1 ; N 2 ; . . . ; N n Þ; n 2 N, qw and qnw denote the Neumann boundary values, and
subscript T means transpose.
Z
dSw T
CP c P c ¼ / N N dX; ðA:11Þ
X dPc
CPc Pnw ¼ 0; ðA:12Þ
CPnw Pc ¼ 0; ðA:13Þ
Z
dSw T
CPnw Pnw ¼  / N N dX; ðA:14Þ
X dPc
Z
krw
KPc Pc ¼ ðrNÞT k rN dX; ðA:15Þ
X lw
Z
krw
KPc Pnw ¼  ðrNÞT krN dX; ðA:16Þ
X lw
KPnw Pc ¼ 0; ðA:17Þ
Z
krnw
KPnw Pnw ¼ ðrNÞT k rN dX; ðA:18Þ
X lnw
Z Z Z
krw
^rPc ¼ Xw N dX  ðrNÞT  k  qw gdX  qw N dC; ðA:19Þ
lw C
ZX X
Z Z
T krw
^rPnw ¼ Xnw N dX  ðrNÞ  k  qnw gdX  qnw N dC: ðA:20Þ
X X lw C
8312 C.-H. Park et al. / Journal of Computational Physics 230 (2011) 8304–8312

References

[1] L.M. Abriola, G.F. Pinder, A multiphase approach to the modeling of porous-media contamination by organic-compounds. 1. Equation development,
Water Resources Research 21 (1985) 11–18.
[2] K. Aziz, A. Settari, Petroleum reservoir simulation, 1979.
[3] M.A. Celia, E.T. Bouloutas, R.L. Zarba, A general mass-conservative numerical-solution for the unsaturated flow equation, Water Resources Research 26
(1990) 1483–1496.
[4] K. Coats, Implicit compositional simulation of single-porosity and dual-porosity reservoirs, in: SPE18427, 1989, pp. 239–275.
[5] Development of Software Infrastructure for Large Scale Scientific Simulation Team, C., 2007. Lis mannual.
[6] P.A. Forsyth, Simulation of nonaqueous phase groundwater contamination, Advances in Water Resources 11 (1988) 74–83.
[7] R. Fuc̆ík, Numerical analysis of multiphase porous media flow in groundwater contamination problems, Ph.D. Thesis, 2006.
[8] R. Helmig, R. Huber, Comparison of galerkin-type discretization techniques for two-phase flow in heterogeneous porous media, Advances in Water
Resources 21 (1998) 697–711.
[9] D.P. Hochmuth, D.K. Sunada, Groundwater model of 2-phase immiscible flow in coarse material, Ground Water 23 (1985) 617–626.
[10] P.S. Huyakorn, S. Panday, Y.S. Wu, A three-dimensional multiphase flow model for assessing napl contamination in porous and fractured media. 1.
Formulation, Journal of Contaminant Hydrology 16 (1994) 109–130.
[11] J.J. Kaluarachchi, J.C. Parker, An efficient finite-element method for modeling multiphase flow (vol. 25, p. 43, 1989), Water Resources Research 30
(1994) 2485–2486.
[12] O. Kolditz, J.O. Delfs, C. Burger, M. Beinhorn, C.H. Park, Numerical analysis of coupled hydrosystems based on an object-oriented compartment
approach, Journal of Hydroinformatics 10 (2008) 227–244.
[13] B.H. Kueper, W. Abbott, G. Farquhar, Experimental observations of multiphase flow in heterogeneous porous media, Journal of Contaminant Hydrology
5 (1989) 83–95.
[14] B.H. Kueper, E.O. Frind, Two-phase flow in heterogeneous porous-media. 1. Model development, Water Resources Research 27 (1991) 1049–1057.
[15] R.W. Lewis, P.J. Roberts, B.A. Schrefler, Finite-element modeling of 2-phase heat and fluid-flow in deforming porous-media, Transport in Porous Media
4 (1989) 319–334.
[16] R.W. Lewis, B.A. Schrefler, N.A. Rahman, A finite element analysis of multiphase immiscible flow in deforming porous media for subsurface systems,
Communications in Numerical Methods in Engineering 14 (1998) 135–149.
[17] D.B. McWhorter, D.K. Sunada, Exact integral solutions for 2-phase flow, Water Resources Research 26 (1990) 399–413.
[18] C.H. Park, C. Beyer, S. Bauer, O. Kolditz, Using global node-based velocity in random walk particle tracking in variably saturated porous media:
application to contaminant leaching from road constructions, Environmental Geology 55 (2008) 1755–1766.
[19] G.F. Pinder, L.M. Abriola, On the simulation of nonaqueous phase organic-compounds in the subsurface, Water Resources Research 22 (1986) S109–
S119.
[20] K. Pruess, Tough user’s guide. Lawrence Berkeley National Laboratory, LBL-20700, 1987.
[21] L. Sanavia, F. Pesavento, B.A. Schrefler, Finite element analysis of non-isothermal multiphase geomaterials with application to strain localization
simulation, Computational Mechanics 37 (2006) 331–348.
[22] O. Schenk, K. Gartner, Solving unsymmetric sparse systems of linear equations with pardiso, Future Generation Computer Systems 20 (2004) 475–487.
[23] C. Thorenz, G. Kosakowski, O. Kolditz, B. Berkowitz, An experimental and numerical investigation of saltwater movement in partially saturated
systems, Water Resources Research 38 (2002), doi:10.1029/2001WR000364.
[24] W. Wang, G. Kosakowski, O. Kolditz, A parallel finite element scheme for thermo-hydro-mechanical (thm) coupled problems in porous media,
Computers & Geosciences 35 (2009) 1631–1641.
[25] Y.S. Wu, P.A. Forsyth, On the selection of primary variables in numerical formulation for modeling multiphase flow in porous media, Journal of
Contaminant Hydrology 48 (2001) 277–304.

Anda mungkin juga menyukai