Professional Documents
Culture Documents
1. Introduction
Flows of bi-phase fluids in porous media occur in a
large number of industrial applications and in natural
systems. Among them, we may indicate boiling and
pressurized water reactors, phase separators used in oil
production, geo-meteorological phenomena, including
sedimentation and soil erosion, blood flow [1–4].
2015 The Author(s) Published by the Royal Society. All rights reserved.
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
2. Model
We consider a two-dimensional channel in the plane XY, limited by two parallel lines placed
at y = ±h/2, where the fluid flow takes places under a pressure gradient applied along the
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
(a)
3
L
1 2 3 4 h
Figure 1. (a) First and (b) third generations of the Apollonian packing inside the channel of height h and extension L. The
numerical integration considers non-slip boundary conditions in the interface between any solid object (all discs in the packing
and channel walls) and any fluid. (Online version in colour.)
X-direction. In its interior, a set of discs, grouped according to an Apollonian packing, models
a fractured porous region, composed of isolated obstacles. This packing was introduced by
Apollonius of Perga in the third century BC as the solution to the problem of determining the
minimum number of non-overlapping circles necessary to cover a region in the plane [19]. In our
model, the packing is constructed according to the following rule. Initially, at the first packing
generation G = 1, four tangent discs of the same size, with radius R = h/4, are placed together
within the channel, in such a way that their centres are located on the vertices of a square of
side h/2. The second generation G = 2 is obtained by adding three more discs: one in the centre
of the G = 1 packing, which touches the four G = 1 discs, and two discs in the space limited by
two discs and one of the walls, which are considered to be circles with infinite radii placed at
y = ±∞. Higher order generations G > 2 are obtained by adding up new tangent circles in all
spaces limited by any three discs in the previous generation. The number of discs in the packing
at generation G is indicated by NG . We recall that, given any set of three mutually tangent circles
with known radii and centre locations, it is possible to determine the radii and the coordinates of
the centres of a fourth circle that is tangent to the three previous ones [19].
The touching discs in the described configuration refrain fluid movement under the pressure
gradient for any value of G. To allow for fluid displacement, it becomes necessary to open
channels between the discs, which is achieved by uniformly reducing their radii Ri by a factor
s < 1. In this work, we consider s = 0.7, 0.8 and 0.9, and G = 1, 2 and 3. In this way, the packing
porosity, which takes account of the space between the discs within a square of area h2 , depends
both on s and on G, i.e. φ = φ(s, G). For the purpose of finding the numerical solution within the
CFD procedure, the otherwise infinite channel is considered to have a finite length L.
The model set-up is illustrated in figure 1, for G = 1 (a) and G = 3 (b), and s = 0.9. For the
purpose of keeping track of the presence of the defending (oil) and invading (water) fluids in
different regions of the channel, it is divided into four zones Zi i = 1 4, indicated by different
colours (grey tones). With this division, it is possible to use a facility in the CFD application to
choose the area in which we want to calculate the volume of each of the phases. At the initial
integration time (t = t0 = 0), the invading fluid (water) occupies zone Z1 . The void space of zone
Z2 , as well as zones Z3 and Z4 are fully filled with oil.
This model has already been used to investigate the flow of single-phase fluid under different
boundary conditions. Although the particular choice of disc sizes and geometrical placement
can hardly be found in natural systems, the use of the Apollonian packing to model porous
geometry has some advantages, which result from this specific packing geometry. Indeed, among
many other properties, it is observed that the disc sizes decay exponentially as their number
in the packing grow, the distribution of discs with respect to their number of neighbours in
the packing decays as a power law, the network obtained connecting the centres of touching
discs, which reproduces the structure of contact forces, is topologically invariant. When it is
used in any numerical modelling, the packing is limited to a finite number of sites but, due to
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
its deterministic nature, it is easy to improve the quality of the results, just by increasing the
4
value of G (and number of discs) in a controlled way. As we mentioned in the Introduction,
the geometrical properties of the packing provide an interesting way to analyse the properties of
3. Methods
The equations of mass and momentum conservation are integrated with the help of ANSYS-
FLUENT CFD package, under the assumption of incompressible and isothermal regime [22]. The
equation of conservation of mass, or the continuity equation, is written as
∂ρ
+ ∇ · (ρv) = Sm . (3.1)
∂t
This equation is valid for both compressible fluids as for incompressible fluids, and the term Sm
is the flow mass added to the continuous phase from a dispersed phase. The incompressibility
assumption requires that ρ does not depend on time. Thus, if Sm = 0, the continuity equation
reduces to
∇ · v = 0. (3.2)
where μ is the molecular viscosity, I is the unit tensor and the second term on the r.h.s. describes
the effect of volume expansion. Thus, ignoring external forces, and using the non-slip condition
and the continuity equation in equations (3.3) and (3.4), the momentum conservation equation is
reduced to the Navier–Stokes equation
∂v
ρ + ρv. ∇v = −∇P + μ∇ 2 v. (3.5)
∂t
The numerical integration takes into account the usual non-slip boundary condition (v = 0)
between all solid–fluid interfaces. The CFD framework requires the definition of a mesh, which
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
results from a division of the domain into discrete, non-overlapping cells [9]. The differential
5
equations expressing mass and momentum conservation are transformed into algebraic equations
for the values of the pertinent fields on the faces of the cells. The resulting values for each of
4. Results
The results we obtained were limited to the low Reynolds number regime (Re < 10), where
hv̄ρ
Re = . (4.1)
μ
For this range of Re values, the properties of the flow also depend on the capillary number nc ,
defined by
μv̄
nc = . (4.2)
σ
In equations (4.1) and (4.2), h indicates the channel height, v̄ the constant inlet speed of the
invading fluid, ρ and μ represent the invading fluid density and viscosity, respectively, while σ
is the surface tension between the two fluids [5,23]. In this investigation, the capillary number,
which expresses the relationship between the viscous force and surface tension between the
two fluids, takes values in the interval 3.0 × 10−6 < nc < 3.5 × 10−5 . So it covers not only a
range conditions where capillary forces dominate viscous forces nc ∼ = 10−5 , but also the opposite
situation in which viscous forces dominate the flow.
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
The effect of the packing on the drainage of oil from the channel by the less viscous fluid
depends on the values of the following parameters: contact angle, inlet speed, channel size,
fluid viscosities and surface tension [23]. For each of the quoted values of reduction factor s and
generation G, we carried on numerical integration of the system for five different values of inlet
speed v̄ indicated in table 1. During this procedure, the values of all other parameters were kept
fixed.
For the invading fluid, we used the properties of water at 20◦ C, which corresponds to
ρinv = 998.2 kg m−3 and μinv = 0.001003 kg ms−1 , while the defending fluid is characterized by
ρdef = 922 kg m−3 and μdef = 0.05 kg ms−1 . This corresponds to oil with typical density (in respect
to that of water) expressed by American Petroleum Institute (API) index 22. Further parameter
values are σ = 0.0312 N m−1 , contact angle between fluids θ = π/2, and time integration step
τ = 2.5 × 10−4 s.
The choice of the inlet speed range was based on the analysis of the flow of the same two-phase
system in a channel without any obstacle. We selected values of v̄ such that, in the packing-free
condition, all oil could be expelled by the invading water. In this way, we can be sure that any oil
retention was caused only by the presence of the packing. The corresponding value of Re were
low enough to be in the laminar (table 1). There is an intermediate range of values of the inlet
velocity in which the regime is still laminar but the invading fluid is not able to completely expel
the oil from the channel. This regime is characterized by a particular fingering pattern, in which
a single finger is opened connecting inlet to outlet, through which water flows. In the stationary
regime, a certain amount of oil remains adhered to the channel walls [23].
We measure the effectiveness of the oil drainage by the quantities B2 and B2,3 . Let Oi , i = 2, 3,
denote the amount of oil retained in zones Z2 and Z3 . The amount of oil initially placed in the
G
zone Z2 corresponds to A2 = h2 − N i=1 π ri (s), where ri (s) indicates the radius of each disc in the
2
O2 O2 + O3
B2 = and B2,3 = , (4.3)
A2 A2
so that B2 informs how much oil was expelled from Z2 by the invading fluid, and B2,3 measures
the total amount of oil that could not be recovered at the end of the process.
(a)
7
(c)
Figure 2. Typical trapped oil pattern in the first-generation packing for (a) v̄ = 0.001 m s−1 , (b) v̄ = 0.00075 m s−1 and (c)
v̄ = 0.0005 m s−1 . Patterns in (b,c) do not display inversion symmetry. (Online version in colour.)
flow through the central channel, horizontally placed between the large G = 1 discs, has been
observed. The central channel is the preferential path taken by the invading fluid.
Water reaches Z3 in three different patterns: (i) exclusively through the central channel;
(ii) through the central channel and one of the two channels formed by a large circle and the
wall; and (iii) through all three channels. In this last case, the flow through Z3 and Z4 is much like
that observed in the packing-free channel at the same value of v̄. The invading fluid completely
expels oil from these regions. The amount of retained oil (if any) is trapped within the channels
between the discs and, eventually, a small drop attached to the right part of the discs in the
downstream direction.
In case (i), the invading fluid flowing through the central channel pressures oil towards the
walls. For the chosen values of v̄, the invading fluid causes the progressive decrease of the width
of the oil layers adhered to the walls, until the rupture of the resulting thin films occurs. This
happens at a certain distances (not necessarily the same for the two walls) from the point marking
the beginning of Z3 . At the moment this occurs, the large surface tension causes the oil left behind
to flow back in the −x-direction. An amount of oil remains attached to the large circle, part in zone
Z2 , part in Z3 . Similarly, the region occupied by the downstream oil is also reshaped towards
the usual interface pattern of the packing-free invasion process. Finally, the dynamics in case
(ii) follows a combination of the processes observed in (i) and (iii).
The different scenarios described above depend on the number and size of discs through
changes in the total area of the walls of the channel, which impact on the influence of the viscous
forces. It is possible to identify a trend towards the enhancement in the amount of oil expelled
with decreasing speed, which results from changes between capillary forces and viscous forces.
The changes in oil retention profile are also related to accumulation in the downstream region
after the packing, as can be seen in figures 2 and 3.
The qualitative analysis of the flow by the direct comparison of resulting patterns indicates
that the presence of new circles causes the water to penetrate the porous region, as can be seen by
the comparison between figures 2 and 3.
These illustrations indicate that, in a laminar flow with low Re, despite the larger energy
dissipation, owing to the increase in the fluid–wall interface caused the increase of the number
of discs, geometric aspects of the porous region can facilitate the invasion of water into thinner
channels. However, this is not a guarantee of an enhancement of oil removal, because oil can
be trapped in other regions. These can be better understood by combining the observation of
trapping patterns with quantitative measures expressed by the parameters B2 and B2,3 .
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
(a)
8
(c)
Figure 3. Typical trapped oil patterns for G = 2, s = 0.8 and (a) v̄ = 0.0005 m s−1 , (b) v̄ = 0.00075 m s−1 and (c) v̄ =
0.001 m s−1 . All patterns lack inversion symmetry with respect to the centre of the channel. (Online version in colour.)
(a) (b)
60 60
50 50
40 40
30 30
20 20
60 60
50 50
B2,3
B2
40 40
30 30
20 20
60 60
50 50
40 40
30 30
20 20
0 0.02 0.04 0.06 0.08 0.10 0 0.02 0.04 0.06 0.08 0.10
v–(cm s–1) v–(cm s–1)
Figure 4. Volume fractions (a) B2 and (b) B2,3 as function of v̄. Squares, circles and triangles correspond, respectively, to G = 1, 2
and 3. From top to bottom, s increases from 0.7 to 0.8 and 0.9. (Online version in colour.)
(a) (b)
60 60 9
50 50
B2,3
B2
40 40
30 30
20 20
60 60
50 50
40 40
30 30
20 20
60 60
50 50
40 40
30 30
20 20
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65
f(s, g) f(s, g)
Figure 5. Volume fractions (a) B2 and (b) B2,3 as function of φ(s, G). Squares, circles and triangles correspond, respectively,
to G = 1, 2 and 3. For fixed G, φ decreases for increasing values of s = 0.7, 0.8 and 0.9. From top to bottom, v̄ assumes the
following values (in cm s−1 ): 0.01, 0.025, 0.05, 0.075 and 0.1. (Online version in colour.)
and walls remain blocked. The only exception observed for the smallest value of v̄ when s = 0.7
results from opening one channel between one of the left discs and the wall. On the other hand,
for the same values of s and v̄, the dependence of B2,3 is subject to larger variations. This results
from the different amounts oil that remain attached to the right of the discs, most of which occupy
regions in zone Z3 .
This picture changes for the higher generations, mainly because of the presence of the disc
with the second largest radius in the central part of the packing. This disc alters the flow pattern
in different ways for different values of s and v̄. The dependence of B2 as function of v̄ is
characterized by large variations for all values of s, some of which with opposite tendencies.
Fluctuations in the behaviour of B2,3 are enhanced, since it results from oil stacked within the
small channels as well as in the region behind the large circles zone Z3 .
Despite the large fluctuations in the values of B2 and B2,3 with respect to v̄, an overall general
tendency is the increase in the fraction of retained oil as the value of s increases. Although this
aspect becomes more evident for B2,3 , it can also be detected in the case of B2 . This corresponds
to the reported behaviour in other previous studies and the overall known resistance to remove a
fluid from dense packing and media with small porosity. The larger differences between B2 and
B2,3 when the value of s increases is explained by the fact that the amount of oil O3 is less sensitive
to variations in the value of s than O2 and A2 .
It is important to note that the two different measures can guide expected results for other
packing arrangements that can be built in similar way. If the whole channel is filled with discs,
one should expect the value of B2 as a good approximation. However, for inhomogeneous
packing, where regions with obstacles and obstacle-free alternate, B2,3 should provide a better
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
approximation for the fraction of retained oil. This is useful when one compares the dependency
10
of these measures as function of v̄ as well as for the analysis of the dependency with respect to
the porosity, that is carried out in the following subsection.
5. Conclusion
This work discussed the numerical investigation of a two-phase flow. The main focus was to
obtain a quantitative relationship between the amount of oil retained within a two-dimensional
channel with a packing of discs of different sizes in its interior when the system is invaded by
a less viscous fluid. Although the flow differs from that occurring in a Hele–Shaw set-up by
neglecting the viscosity effect between fluids and limiting plates, viscous fingering phenomena
are also present [24]. The packing is constructed by a geometric recurrence rule, and the flow
proceeds at low speed to warrant a laminar flow. The regime of low velocity was also reduced to
warrant that, at all invading velocities used in the investigation, no oil would be retained inside
the channel in the packing-free condition.
The total amount of retained oil depends on the invading fluid velocity as well as on several
geometrical features, including the number, position and size of the discs. No simple relation
could be obtained between this amount and any single specific geometric feature of the packing.
Global tendencies, however, can be detected. They are consistent with empirical findings that
extraction is difficult when porosity is small and the invading velocity increases.
The quantitative results were expressed in terms of the fraction of oil retained in the region
where the packing was contained, as well as in terms of all remaining oil. This makes it possible
to extrapolate the results to situations in which a larger number of obstacles is uniformly placed
within the channel, or when they follow a sequence of regions with and without obstacles.
Different effects follow the increase of the number of discs inside the channel. On one side, it
creates barriers within the porous region and increases the area of contact of fluid with the walls.
At the same time, changes in the flow pattern may enhance the efficiency of the invading fluid in
expelling the resident one.
Preliminary results for more complex configurations, with two or more Apollonian plugs
placed inside the channel, confirm that the devised framework can lead to similar results for
more realistic patterns (see §2), which will be discussed in a forthcoming contribution. The
computational efforts to investigate many distinct configurations are much larger, as we have
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
to consider all difficulties that this study made evident for the simulation of two-phase flows
11
within CFD. A good mesh discretization becomes crucial due to the necessity to account for the
fractions of two fluids in the same cell. The control of Courant number, which is not relevant to
References
1. Scheidegger AE. 1974 The physics of flow through porous media, 3rd edn. Toronto, Canada:
University Toronto Press.
2. Bear J. 2013 Dynamics of fluids in porous media. New York, NY: Dover Civil and Mechanical
Engineering, Dover Publications.
3. Sahimi M. 2011 Flow and transport in porous media and fractured rock: from classical methods to
modern approaches, 2nd edn. Weinheim, Germany: Wiley-VCH.
4. Dullien FAL, Brenner H. 2012 Porous media: fluid transport and pore structure. Amsterdam, The
Netherlands: Elsevier Science.
5. Tiab D, Donaldson EC. 2012 Petrophysics: theory and practice of measuring reservoir rock and fluid
transport properties, 2nd edn. Burlington, Gulf Professional Publication.
6. Vandenbossche J, Yox-theimer D, Vidic R, Brantley S, Abad J. 2013 Impact of
shale gas development on regional water quality. Science 340, 1235009. (doi:10.1126/
science.1235009)
7. Male F, Patzek T, Marder M. 2013 Gas production in the Barnett Shale obeys a simple scaling
theory. Proc. Natl Acad. Sci. USA 110, 19 731–19 736. (doi:10.1073/pnas.1313380110)
8. Versteeg HK, Malalasekera W. 2007 An introduction to computational fluid dynamics: the finite
volume method. Harlow, UK: Longman Scientific and Technical.
9. Tu J, Yeoh GH, Liu C. 2012 Computational fluid dynamics: a practical approach. London, UK:
Butterworth-Heinemann.
10. Rothman DH, Zaleski. S. 1994 Lattice-gas models of phase separation: interfaces, phase
transitions, and multiphase flow. Rev. Mod. Phys. 66, 1417–1479. (doi:10.1103/RevModPhys.
66.1417)
11. Wilkinson D, Willemsen JF. 1983 Invasion percolation: a new form of percolation theory.
J. Phys. A 16, 3365–3376. (doi:10.1088/0305-4470/16/14/028)
12. Ferreol B, Rothman DH. 1995 Lattice-Boltzmann simulations of flow through fontainebleau
sandstone. In Multiphase flow in porous media (ed. PM Adler), pp. 3–20. Berlin, Germany:
Springer.
13. Adler P. 1985 Transport processes in fractals. I-Conductivity and permeability of a Leibniz
packing in the lubrication limit. Int. J. Multiphase Flow 11, 91–108. (doi:10.1016/0301-
9322(85)90007-2)
14. Adler PM, Thorvert JF. 1998 Real porous media: local geometry and macroscopic properties.
Appl. Mech. Rev. 51, 537–585. (doi:10.1115/1.3099022)
15. Doye JPK, Massen CP. 2005 Self-similar disk packings as model spatial scale-free networks.
Phys. Rev. E 71, 016128. (doi:10.1103/PhysRevE.71.016128)
16. JS Andrade, Herrmann HJ, Andrade RFS, da Silva LR. 2005 Apollonian networks:
simultaneously scale-free, small world, euclidean, space filling, and with matching graphs.
Phys. Rev. Lett. 94, 018702. (doi:10.1103/PhysRevLett.94.018702)
Downloaded from http://rsta.royalsocietypublishing.org/ on February 28, 2018
17. Oliveira RS, Andrade JS, Andrade. RFS. 2010 Fluid flow through Apollonian packings. Phys.
12
Rev. E 81, 047302. (doi:10.1103/PhysRevE.81.047302)
18. Oliveira RS, Andrade JS, Andrade. RFS. 2015 Fluid flow through packings of rotating