Anda di halaman 1dari 9

ARMA 10-221

Analysis of sand production processes at the pore scale using the discrete element method and
lattice Boltzman procedures.
Raquel Quadros Velloso
Dept. of Civil Engineering, Catholic University, Rio de Janeiro, Brazil
Euripedes A. Vargas Jr.
Associate Professor, Dept. of Civil Engineering, Catholic University, Rio de Janeiro, Brazil
Clemente J. Gonçalves
Senior Engineer, Research Division (CENPES), Petrobras, Rio de Janeiro, Brazil.
Armando Prestes
Research Engineer, Dept. of Civil Engineering, Catholic University, Rio de Janeiro, Brazil.
Copyright 2010 ARMA, American Rock Mechanics Association
This paper was prepared for presentation at the 44th US Rock Mechanics Symposium and 5th U.S.-Canada Rock Mechanics Symposium, held in
Salt Lake City, UT June 27–30, 2010.
This paper was selected for presentation at the symposium by an ARMA Technical Program Committee based on a technical and critical review of
the paper by a minimum of two technical reviewers. The material, as presented, does not necessarily reflect any position of ARMA, its officers, or
members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of ARMA
is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The
abstract must contain conspicuous acknowledgement of where and by whom the paper was presented.
ABSTRACT: It is believed that coupled fluid-mechanical-erosional phenomena associated to sand production derived from
microscale based mechanisms as the disaggregation of particles from the rock matrix and their transport through the pores are best
described at the particle and pore level. The present work presents results obtained from a numerical simulation study of these
phenomena at the pore scale level. This study was carried out by coupling the discrete element method for the analysis of the
mechanical behavior of the rock and lattice-Boltzman method for the simulation of fluid flow at the pore level. Computer codes
were written to simulate both processes using the procedures described. One objective of the work was an improvement in the
knowledge of micromechanical processes leading to solids production. It is believed that such understanding will improve the
constitutive relationships necessary for sand prediction and used in continuum based methods of prediction, more amenable for
applications in real case situations. Rock properties and texture, boundary conditions (stresses and drawdown) in the
disaggregation and particle transport were considered. Preliminary results of the DEM-two phase LBM coupling are also
presented. Advantages and limitations of this micromechanical approach are also discussed.

Models that try to predict sand production rates and


1. INTRODUCTION produced volumes are generally continuum based,
The different processes involved in sand production are incorporating fluid-mechanical coupling and erosion
generally associated to deformation and failure of the phenomena [1, 2, 3, 4]. For the solution of these
rock, rock-fluid interaction, and solids transport. Sand problems it is necessary that a constitutive relation for
production can be described by two mechanisms [1]: the eroded mass of solids is established. Several
constitutive relations of this type have been proposed for
1. Mechanical instabilities and localized failure of the
that purpose in order to reproduce laboratory test results.
rock in the vicinity of the well due to stress
Such relations are inspired in the filtration theories for
concentrations;
fines through a solid matrix. Besides the constitutive
2. Hydromechanical instabilities due to erosion which relation for erosion (relating the erosion process to pore
manifests in the disaggregation and particle transport flow), relationships between mechanical processes
caused by seepage forces. (deformation and failure) with erosion have also to be
established. All the mechanisms that couple mechanical,
These two mechanisms are coupled as stress
flow and erosion happen at the microscopic scale as
concentrations lead to localized failure which in turn
disaggregation of a particle from the rock mass and its
leads to amounts of displaced particles transported by
transport through the pore structure is best described at
the fluid. The displacement of particles leads to an
the particle and pore scale. A better understanding of
increase of rock porosity causing a readjustment of the
such basic mechanisms may help substantially in the
interparticle forces leading to a further increase in the
improvement of the constitutive relations necessary for
failed zone of the rock mass.
sand production prediction.
Numerical modeling at the particle and pore scale is a A particular aspect that should be very relevant in the
tool that may be useful in the study of the fundamental process of sand production is related to capillary forces.
mechanisms as described for sand production. One way However, little consideration has been devoted to its
to carry out this modeling is by coupling a method that quantification in the microscale. Gili and Alonso [33]
treats particle movement and their interaction with fluid coupled DEM with a pore network in 2D for the study of
flow at pore level. For the simulation of movement and unsaturated soils where capillary forces are calculated
particle interaction, the DEM (Discrete Element Method) through menisci established between two spheres.
[5] has been widely used. Ibañez [34] used a similar approach in order to model
the behavior of unsaturated residual soils. Grof et al.
As for coupling of DEM with flow, several approaches
[16] introduced an approximate solution for the capillary
can be followed. Perhaps the simplest way is to use a
forces on a coupled simulation using the DEM with a
Darcy based continuum model for flow superimposed to
numerical solution (finite volume) of Navier-Stokes
a discrete particle code. Some implementations of this
equations in 3D.
type have been used for sand production studies and
hydraulic fracturing [6, 7, 8, 9, 10]. An alternative One problem associated to the application of
approach involves the local averaged solution of Navier- conventional LBM to the solution of Navier-Stokes
Stokes equations developed by Anderson and Jackson equations in the case of incompressible fluids is the
[11]. In this method, flow variables are solved in a cell effect of compressibility itself [35]. This effect becomes
containing various solid particles and fluid-particles particularly acute when imposing pressure gradients at
interaction are considered through semi-empirical the boundaries. In the present work, the formulation
relationships. Such methods have been applied to proposed by He and Luo [35], one that reduces the effect
fluidization studies [12, 13]. of compressibility in conventional LBM was used. This
work presents initially results obtained with a computer
The simulation of flow at the pore scale in the real
implementation of He and Luo’s formulation for LBM
geometry demands the solution of Navier-Stokes and
coupled to DEM for one phase flow conditions. The
mass conservation equations in this domain. However,
work also presents preliminary simulations using
the traditional computational fluid dynamics methods
coupled DEM-LBM for two phase flow conditions. In
can be very inefficient when dealing with complex
both cases sand production conditions are focused.
geometries as the ones encountered in porous media.
When flow is coupled with particle movement, these
solution procedures become even less efficient as it is
2. THE METHODS
generally necessary that meshes have to be regenerated
along the simulation stages. Hu [14] and Maury [15] 2.1. Solid particle motion
present simulation of the coupling between particle Motion of particles and their interaction are modeled in
movements and fluid flow using finite elements in 2D, the present work by the Discrete Element Method
with a low fraction of solids and without particle (DEM). DEM was introduced by Cundall [36] for the
interaction. Grof et al. [16] coupled DEM to a flow analysis of rock mechanics problems and later
simulator considering stationary flow conditions. specialized to granular media by Cundall and Strack [5].
An alternative method for the solution of the flow In this method, Newton´s second law is used to describe
equations is the denominated lattice-Boltzmann method the motion of an individual particle. The equation that
(LBM) which has been successfully used in flow governs the translation motion of an individual particle i
simulations of complex geometries and demonstrated to is:
be more efficient than the more traditional methods dv i nci
under such conditions [17, 18]. LBM has often been mi = F f ,i + ∑ Fc,ij + Fg ,i (1)
dt j =1
used in one and two phase flow simulations in porous
media as soil and rocks [19, 20, 21, 22, 23, 24, 25, 26, where mi e vi are respectively mass and velocity of
27, 28, 29]. In these simulations the main objective was particle i, nci is the number of contacts of particle i.
the determination of permeability and saturation- Forces acting on particle are fluid transferred force Ff,i,
capillary pressure relationships of the porous medium. gravitational force Fg,i, and contact forces between
Ladd [30] proposed the application of LBM in the particles i and j Fc,ij. The equation that governs the
analysis of suspended particles in a fluid by coupling rotational motion of particle i is:
LBM with the movement of the solid particles without
interaction amongst them. Cook and Noble [31] coupled dϖ i nci

LBM to the DEM for the study of bed erosion in 2D. Ii


dt
= T f ,i + ∑ Tij (2)
j =1
Boutt et al. [32] applied the coupled system LBDEM to
the study of hydraulic fracturing problems. where ωi is the angular velocity of particle i, Ii is the
moment of inertia of particle i, Tij is the torque generated
by the contact forces between particles i and j and Tf,i is b
the torque generated by the fluid force on particle i. p(x, t ) = ∑ f i (x, t ) (6)
i =1
Motion equations (1) and (2) are integrated using a
central difference scheme. Details of expressions for 1 b
accelerations, velocities and displacements can be found u(x, t ) = ∑ ei fi (x, t )
p0 i =1
(7)
in Cundall and Strack [5].
Contact forces between particles are calculated through a Viscosity and fluid density are given by:
force-displacement law. The contact behavior depends
c2  τ 1 
on the contact constitutive behavior. In the present work ν= ∆t  −  (8)
a linear elastic constitutive behavior was adopted for the 3  ∆t 2 
contact. Contacts have bonds that are allowed to fail in p ( x, t )
shear and tension. ρ (x, t ) = (9)
cs2
2.2. One phase flow
One phase flow occurring in the pores of a porous where cs is the sound velocity in the fluid and for D2Q9
medium can be described by Navier-Stokes equations. and D3Q19 lattices is given by:
LBM is a technique for the solution of these equations.
In the present work the formulation proposed by He and c2
cs = (10)
Luo [35] is used. This formulation reduces considerably 3
the errors due to compressibility in relation to the
conventional formulation of LBM, allowing application
of larger pressure gradients. 2.3. Fluidmechanical coupling
In the LBM, the problem domain is discretized by a Coupling of flow processes simulated with LBM and
lattice where an evolution equation is given by: particle movement was initially proposed by Ladd [30]
with the objective of the simulating the particular case of
f i (x + ei ∆t , t + ∆t ) = suspended particles. Contact between particles was not
∆t (3) modeled in that work. Cook and Noble [31]
f i (x, t ) − ( fi (x, t ) − fieq (x, t )) implemented the coupling between DEM and LBM for
τ
the solution of 2D fluidmechanical problems using the
where τ is the collision relaxation time, ∆t is time immersed moving boundary scheme as proposed by
interval, fi is the local pressure distribution function, ei Noble and Torczynsky [37].
represents the lattice velocity and i = 1, .., b, b being the
number of discrete velocities of the lattice. In the present Coupling between fluid flow and particle movement
work, D2Q9 (two dimensional, 9 velocities) and D3Q19 involves two stages:
(three dimensional, 19 velocities) lattices are used. The 1. Definition of the boundary condition imposed to the
equilibrium distribution function is given by [35]: fluid by the solid particles in motion.
fieq (x, t ) = 2. Calculation and transfer of the drag forces from the
fluid flow to the solid particles.
  (e ⋅ u ) 9 (ei ⋅ u )2 3 u 2   (4)
wi  p + p0 3 i 2 + −  In order to simulate the fluidmechanical interactions
  c 2 c4 2 c 2   between solid particles and fluid flow, LBM must be
modified in order to incorporate the boundary condition
where p0 is the initial (constant) pressure and of the solids in motion. The boundary condition at the
p = p0 + δp, δp being the pressure fluctuation, u is the solid surface is one of no-slip, that is, the solid is
fluid velocity, wi are coefficients that depend on the impermeable and the fluid adjacent to the solid surface
lattice and c is magnitude of the lattice velocity given moves with the same velocity of the solid. The immersed
by: moving boundary condition, proposed by Noble and
∆x Torczynsky [37], the one adopted in the present work,
c= (5) modifies LB equation (Eq. (3)) in order to impose the
∆t
no-slip condition in the sites of the lattice covered by the
∆x being the lattice spacing. After imposing initial and particles. The modified LB equation is then given by:
boundary conditions, LBM equations are solved and the
macroscopic variables of flow can be obtained through fi (x + ei ∆t , t + ∆t ) = fi (x, t ) −
the following expressions: ∆t (11)
(1 − B(x, t ))( fi (x, t ) − fieq ( x, t )) + B(x, t )Ωis (x, t )
τ
where B is a weighing function given by:
ε (x, t )(τ / ∆t − 0.5) b
B ( x, t ) = (12) N (x, t ) = ∑ N i (x, t ) (18)
(1 − ε (x, t )) + (τ / ∆t − 0.5) i =1

and ε is the solid fraction at the site, defined by fraction


of the volume at the lattice site occupied by the solid
(ii) Generation of surface tension
particle. In Eq. (11), Ωis is the additional collision term
that modifies pressure distribution functions at the lattice Initially, color gradient is determined:
nodes covered by solid particles in a way that the no-slip
ei
condition is imposed. This term is given by: g=∑
c
∑ [R j (x + e i ∆t , t ) − B j (x + e i ∆t , t )] (18)
i j
Ωis (x, t ) = f −i (x, t ) − fi (x, t ) +
(13) And with this value a perturbation is added to the
fieq ( ρ , u s ) − f −eqi (x, t ) populations which creates a surface tension:
where us is particle velocity at position x, in time t and –i  (e ⋅ g )2 1 
represents opposite direction to direction i. N i" (x, t ) = N i' (x, t ) + wi A g  i 2 −  (19)
 c2 g 2
In LBM, forces that are transferred to the solid particles  
can be determined directly from the summation of
A is a parameter chosen to set the magnitude of the
momentum transfer from the fluid sites to the solid
surface tension.
surface. In the immersed moving boundary scheme, the
fluid force in a particle is given by: (iii) Color redistribution in order to minimize the
diffusion from one color to the other
∆x3
Ff = − ∑ Bn ∑ Ωisei (14) At this stage, the method proposed by Latva-Kokko and
cs2 ∆t n i Rothman [40] was adopted:
n being the sites covered by the particle. Torque R ( x, t )
regarding the center of mass of the particle, xp, is given Ri" (x, t ) = N i" (x, t ) + ∆ i (20a)
R ( x, t ) + B ( x, t )
by:
B ( x, t )
∆x 3  s  Bi" ( x, t ) = N i" (x, t ) − ∆ i (20b)
Tf = − ∑ ( x − x p ) × 
 Bn ∑ Ω i ei  (15) R ( x, t ) + B ( x, t )
cs2 ∆t n  i 
R ( x, t ) B ( x, t )
∆i = β 2
N 0 ieq cos ϕ i (20c)
( R ( x, t ) + B ( x, t ) )
2.4. Two Phase Flow
The formulation for LBM used in the present work for being R and B the total number of red and blue particles
the case of two phase fluid flow follows the work of at one specific site, β is the parameter that provides the
Gunstensen and Rothman [38] and Rothman and Zaleski tendency of the two fluids to separate, φi is the angle
[39]. In that formulation, at each lattice site, two types of between the color gradient g and direction ei, and
fluid particles can coexist, the red and blue particles. N 0 ieq is given by Eq. (17) assuming that velocity u is
One defines Ri (x,t) as the red population at site x, and zero.
time t, velocity in direction i. Bi(x,t) can be defined
likewise for the blue fluid in a way that (iv) Propagation of populations to adjacent sites
Ni(x,t) = Ri(x,t) + Bi(x,t). The main idea is to separate
Ri (x + e i ∆t , t + ∆t ) = Ri" ( x, t )
the evolution of the population in four stages: (21)
(i) Collision Bi (x + e i ∆t , t + ∆t ) = Bi" ( x, t )
Wettability of a fluid in the solid phase is controlled by
N i' (x, t ) = N i (x, t ) −
∆t
τ
[N (x, t ) − N
i
eq
i ( x, t ) ] (16) only one parameter p ≥ 0, one that measures the red
color fraction at the solid sites. At the beginning of the
being: simulation, color at the solid sites is attributed with the
objective of calculating color gradients. If the red fluid is
N ieq (x, t ) = totally wetting then p=1, if it is partially wetting then
 1< p <0, and if both fluids have the same wettability then
(e i ⋅ u ) 9 (e i ⋅ u )2 3 u 2  (17)
p=0.
wi N (x, t ) 1 + 3 2 + − 
 c 2 c4 2 c 2 

3. RESULTS AND DISCUSSION


3.1. Verification of the fluidmechanical coupling with constant velocity in a fluid, considering low
Two examples are initially shown in order to Reynolds numbers, Stokes' law provides the drag force:
demonstrate the ability of the incompressible Fd = 6πρυR(-up) (22)
formulation of LBM in order to correctly determine the
force transferred from the fluid to the solid particle. A Where R is the sphere radius and up is the sphere
2D situation simulates a still particle while a 3D velocity. Simulation data, in non dimensional values are:
situation simulates a moving particle. up = 0.005; R = 5∆x; υ = 1/6; ρ = 1.0; ∆x = 1.0; ∆t = 1.0.
The first example consists of a fixed disk located For these values Reynolds number is 0.15 and the drag
between two parallel walls where one phase flow is force is 0.0785. The lattice dimension is 60x60x60. An
imposed as shown in Figure 1. This verification is evolution of the drag force is presented in Figure 3
particularly relevant as it is representative of drag forces where a satisfactory agreement is obtained between
transmitted to a particle in high solids concentrations (in LBM and the analytical solution, an error of
this case the walls are located near to the particle). The approximately 5 % was found for this lattice dimension.
parameters used in the simulations are: Reynolds
number = 0.0002; ∆x = 5.0E-5m; ∆t = 1.0E-4s; υ =
2.0E-6m2/s; ρ = 1000kg/m3; disk radius, a = 10∆x;
domain length, L = 20a. The purpose of the exercise was
to calculate the force transmitted by the fluid to the
particle.

Fig. 3. Comparison between the analytical solution (blue line)


and the numerical solution (red line) for the drag force. The
Fig. 1. Geometry and boundary conditions of 2D example for figure on the side shows the velocities and flow lines obtained
the verification of the drag force determination. in the numerical simulation.

The results obtained with LBM were compared with the 3.2. Simulation of sand production
results obtained by Richou et al. [41] who simulated the In order to evaluate the implemented procedures for the
same flow problem using finite differences. Figure 2 fluid mechanical coupling applied to deformation, failure
shows a comparison between the two procedures and a and erosion of granular media, 2D situations were
good agreement was obtained. analyzed. Initially, 2D simulations were carried out in
order to evaluate the stress-strain-failure of synthetic
granular media.
Considering the limitations of LBM regarding pressure
differential that can be applied when using LBM, the
synthetic granular media used in the simulations have a
higher deformability and smaller strength than the real
materials. Parameters of the synthetic materials are
presented in table 1 and the stress-strain curves of the
biaxial tests are presented in figure 4. It is possible to
observe that the materials present a behavior compatible
Fig. 2.Comparison of the results obtained with incompressible with the one of geologic materials.
LBM and results obtained by Richou et al. [41].

In order to evaluate the motion of solid particles in the Table 1. Mechanical parameters of the materials.
lattice and the evaluation of forces transferred from the
Mat# Rmin Rmax µ KN KS σcontact τcontact
fluid to the particles, a sphere immersed into an initially
[mm] [mm] [-] N/m N/m [Pa] [Pa]
still fluid is considered. A constant velocity is imposed
for the particle and a periodic flow condition is imposed 1 0.20 0.25 0.25 1.0e6 1.0e6 25.0 200.0
in the flow direction. For the case of a sphere moving 2 0.20 0.25 0.50 1.0e6 1.0e6 50.0 200.0
3 0.15 0.25 0.25 1.0e6 1.0e6 25.0 200.0
pressure field and flow velocities for the case
M3∆P2S100 is presented in Figure 9.
Subsequently, simulation of sand production is carried
out. In the simulation stages carried out, the granular
medium is submitted to a confining stress equal in the
vertical and horizontal directions and a hole is
introduced at the right wall with concomitant application
of a pressure difference between the three remaining
walls and the hole (Figure 5). These conditions
(confining stresses and pressure differences) are
maintained constant during the simulation, the wall at
the left hand side (where the hole is) is maintained fixed.
In order to allow fluid connectivity in this 2D scheme,
the same approach used by Boutt et al. [32] was used,
one which considers a hydraulics radius of the solid Fig. 5. Boundary conditions for the sand production
simulations.
particles as a percentage of the mechanical radius. In this
case the hydraulic radius was taken as 80% of the
mechanical radius.

Fig. 6. Sand production curves for materials M1 and M2.

Fig. 4. Stress-strain behavior simulated with DEM for the


materials used in the sand production simulations.

The simulation of sand production are referred by the


acronym Mm∆PpSs where m refers to the material (table
1), p is the pressure gradient imposed: p=1 for ∆P/L =
0.8kPa/m, and p = 2 for ∆P/L = 1.6kPa/m, s is the
confining pressure in Pa. Figure 6 shows the solids
production along time. One observes that material M1
(less resistant) produced solids only when submitted to Fig. 7. Sand production curves for materials M1 and M3.
the highest confining stress (100 Pa) while in the other
situations a stable arch was formed.
A higher confining stress may produce a higher
deviatoric state of stress in the vicinity of the hole,
leading to the creation of a failed zone prone to be
eroded. Figure 7 shows the production curves obtained
with materials M1 and M3, the only difference between
them is the fact that material M3 has smaller grain sizes
and it is more non-uniform than M1. The production rate
of material M3 was the highest encountered for the two
simulated cases. Figure 8 shows the domain at the final
stages of simulation in the two cases: a stable arch Fig. 8. Final stages of simulations M3∆P2SC100 (left hand
(M2∆P2S080), solids produced (M3∆P2S100). The side) and M1∆P2SC080 (right hand side). In the back of
image, the pressure field is plotted. (in Pa).
Fig. 9. Final stages of simulation M3∆P2SC100. On the left Fig. 11. Comparison between the results obtained with two
hand side the velocity field [m/s] is plotted and on the right phase flow LBM simulations and the solution presented in Gili
hand side the pressure field [Pa] is plotted. Some solid and Alonso [33].
particles have been erased for a better visualization of the
fields.

3.3. Capillary force simulation


This section focuses on a particular aspect relevant to
solids production which is the capillary force generated
by the presence of various fluid phases inside the pores.
The results try to demonstrate the ability of LBM in
simulating correctly the capillary force that appears in a
meniscus between solid particles. The example consists Fig. 12. Meniscus evolution with particle motion.
in two fixed solid spheres surrounded by a non-wetting
fluid and having an amount of a wetting fluid at their
contact. Simulation using LBM proceeds until a
permanent and stable condition is reached (figure 10). 4. CONCLUSIONS
The capillary forces determined with two phase LBM
were compared with an analytical solution presented by The present work presented a methodology for the
Gili and Alonso [33] for the capillary force generated by fluidmechanical coupling between DEM and LBM in
the meniscus between the two spheres (figure 11). One order to simulate sand production problems in the
notices that the analytical solution is valid for wetting microscale. Considering one phase flow problems,
angles up to approximately 50º. Figure 12 shows simulations were carried out varying the following
meniscus evolution when a constant force is applied parameters: strength, boundary stresses, pressure
upwards on the upper sphere. This force has a magnitude gradients and grain size. The incompressible LBM
larger than the capillary force generated at the meniscus formulation used reproduced well the forces transmitted
shown. to a solid particle but, although able to apply pressure
differentials up to 50 times higher than values applied
using the conventional formulation, these pressure
differentials are not high enough to simulate real
problems.
Two phase flow simulations were carried out and the
results obtained suggest that LBM is able to correctly
simulate the action of the capillary forces on solid
Fig. 10. LBM simulation results for equilibrium configurations particles. An advantage of two phase flow simulations in
of menisci between two spheres.
relation to using analytical solutions is that it is not
limited to particles of regular geometries or low wetting
angles. Preliminary results concerning DEM-LBM
coupling for two phase problems were presented.
The main advantages of LBM are its ability to simulate
different mechanisms in the microscale as the
determination of drag forces and capillary forces, both
relevant to the simulation of sand production problems.
This way, LBM coupled to DEM appears to be an
adequate tool for the understanding and verification of 8. Preece, D.S, R.P. Jensen, E.D. Perkins, J.R. Williams.
stress-strain x failure x flow x erosion constitutive 1999. Sand production modeling using superquadric
relations. discrete elements and coupling of fluid flow and
particle motion. Rock Mechanics for Industry.
LBM is not, however, without problems and what could
9. Bruno, M.S., A. Dorfmann, K. Lao. 2001. Coupled
be considered as its limitations at present are:
particle and fluid flow modeling of fracture and slurry
1. Relatively low pressure differentials can be imposed injection in weakly consolidated granular media. Rock
in the simulations. This limitation precludes the adoption Mechanics in the National Interest.
of strength parameters for the materials compatible to 10. Liming, L., E. Papamichos, P. Cerasi. 2006.
real geologic materials. Investigation of sand production mechanisms using
DEM with fluid flow. In Proc. Int. Symp. ISRM, Eurock
2. High computational costs. This however is a common
2006, Liège, Belgium, 9-12 May 2006.
feature of microscale simulations. One should observe,
however, that numerical implementation of DEM-LBM 11. Anderson, T.B., R. Jackson, R.. 1967. A fluid
coupling is prone to parallelization. Also, LBM can be mechanical description of fluidized beds: equations of
considered as a rigorous method, especially when motion. Ind. Eng. Chem. Fund. 6: 527-539.
dealing with two-phase flow, free from simplifications 12. Xu, B.H., A.B. Yu, A.B. 1997. Numerical simulation of
and it is understandable that it may be slower that other the gas-solid flow in a fluidized bed by combining
methods that adopt simplifications. discrete particle method with computational fluid
dynamics. Chem. Eng. Sci. 52(16): 2785-2809.
Despite the shortcomings stated above, one should note
that the use of a coupled DEM-LBM as proposed in the 13. Di Renzo, A., F.P. Di Maio. 2007. Homogeneous and
bubbling fluidization regimes in DEM-CFD
present work for sand production simulations is not to
simulations:Hydrodynamics stability of gas and liquid
solve real scale problems but to evaluate and even fluidized beds. Chem. Eng. Sci. 62: 116-130.
suggest macroscopic constitutive relations to be used in
more conventional simulators. 14. Hu, H.H. 1996. Direct simulation of flows of solid-
liquid mixtures. Int. J. Multiphase Flow 22(2): 335-352.
15. Maury, B. 1999. Direct simulations of 2D fluid-particle
REFERENCES flows in biperiodic domains. J. Comp. Physics 156:
325-351.
1. Vardoulakis, I., M. Stavropoulou, P. Papanastasiou.
1996. Hydro-mechanical aspects of the sand production 16. Grof, Z., J. Cook, C.J. Lawrence, F. Štěpánek. 2009.
problem. Transport in Porous Media 22: 225-244. The interactiom between small clusters of cohesive
particles and laminar flow: Coupled DEM/CFD
2. Papamichos, E., I. Vardoulakis, J. Tronvoll, A.
approach. J. Petr. Sci. Eng. 66: 24-32.
Skærstein. 2001. Volumetric sand production model
and experiment. Int. J. Num. An. Meth. Geomech. 25: 17. Bernsdorf, J., F. Durst, M. Schäfer. 1999. Comparison
789-808. of cellular automata and finite volume techniques for
simulation of incompressible flows in complex
3. Wang, J. 2003. Mathematical and Numerical Modeling
geometries. Int. J. Numer. Meth. Fluids 29: 251-264.
of Sand Production as a Coupled Geomechanics-
Hydrodynamics Problem. Ph.D. Thesis, University of 18. Geller, S., M. Krafczyk, J. Tölke, S. Turek, J. Hron.
Calgary. 2006. Benchmark computations based on lattice-
Boltzmann, finite element and finite volume methods
4. Detournay, C., C. Tan, B. Wu. 2006. Modeling the
for laminar flows. Computers & Fluids 35: 888-897.
mechanism and rate of sand production using FLAC. In
Proc. of the 4th Int. FLAC Symp. on Numerical 19. Gunstensen, A.K., D. Rothman. 1993. Lattice-
Modeling in Geomechanics, Madrid. Paper No. 08-10. Boltzmann studies of immiscible two-phase flow
R. Hart and P. Varona, Eds. trough porous media. J. Geoph. Res. 98(B4): 6431-
6441.
5. Cundall, P.A., O.D.L. Strack. 1979. A discrete
numerical model for granular assemblies. Géothecnique 20. Ferreol, B., D.H. Rothman. 1995. Lattice-Boltzmann
29(1): 47-65. simulations of flow through Fontainebleau sandstone.
Transport in Porous Media 20: 3-20.
6. O'Connor, R.M., J.R. Torczynski, D.S. Preece, J.T.
Klosek, J.R. Williams. 1997. Discrete element 21. Martys, N., H. Chen, H. 1996. Simulation of
modeling of sand production. Int. J. Rock Mech. & Min. multicomponent fluids in complex three-dimensional
Sci. 34: paper No. 231. geometries by the lattice Boltzmann method. Phys. Rev.
E 53: 743-750.
7. Dorfmann, A., L. Rothenburg, M.S. Bruno. 1997.
Micromechanical modeling of sand production and 22. Olson, J.F., D. Rothman. 1997. Two-fluid flow in
arching effects around a cavity. Int. J. Rock Mech. & sedimentary rock: simulation, transport and complexity.
Min. Sci. 34: paper No. 068. J. Fluid Mech. 341: 343-370.
23. Hazlett, R.D., S.Y. Chen, W.E. Soll. 1998. Wettability 37. Noble, D.R., J.R. Torczynsky. 1998. A lattice-
and rate effects on immiscible displacement: Lattice Boltzmann method for partially saturated computational
Boltzmann simulation in microtomographic images of cells. Int. J. Modern Phys. C 9(8): 1189-1201.
reservoir rocks. J. Petr. Sci. Eng. 20: 167-175.
38. Gunstensen, A.K., D.H. Rothman. 1991. Lattice
24. Keehm, Y. 2003. Computational Rock Physics: Boltzmann model for immiscicle fluids. Phys. Rev. A
Transport Properties in Porous Media and 43(8): 4320-4327.
Applications. Ph.D. Thesis, Stanford University.
39. Rothman, D.H., S. Zaleski. 2004. Lattice-Gas cellular
25. Kutay, M.E., A.H. Aydilek, E. Masad. 2006. Automata. 1st paperback ed. Cambridge. Cambrige
Laboratory validation of lattice Boltzmann method for University Press.
modeling pore-scale flow in granular materials.
40. Latva-Kokko, M., D.H. Rothman. 2005. Diffusion
Computers and Geotechnics 33: 381-395.
properties of gradient-based lattice Boltzmann models
26. Ramstad, T., P.-E. Øren, S. Bakke. 2009. Simulation of of immiscible fluids. Phys. Rev. E 71: 056702.
two-phase flow in reservoir rocks using a lattice
41. Richou, A.B., A. Ambari, J.K. Nacin. 2004. Drag force
Boltzmann method. In Prec. SPE Annual Tech. Conf.
on a circular cylinder midway between two parallel
and Exhib. New Orleans, Lousiana, USA, 4-7 October
plates at very low Reynolds numbers - Part 1: Poiseuille
2009. SPE124617.
flow (numerical). Chem. Eng. Sci. 59: 3215-3222.
27. Pan, C., M. Hilpert, C.T. Miller. 2004. Lattice-
Boltzmann simulation of two-phase flow in porous
media. Water Resour. Res. 40: W01501.
28. Schaap, M.G., B.S.B. Christensen, M.L. Porter, D.
Wildenschild. 2006. Linking experimental capillary
pressure-saturation data with lattice Boltzmann
simulations. In Computational Methods in Water
Resources - XVI International Conference,
Copenhagen, Dinamarca.
29. Porter, M.L., M.G. Schaap, D. Wildenschild. 2009.
Lattice-Boltzmann simulations of the capillary
pressure-saturation-interfacial area relationship for
porous media. Adv. Water Res. 32: 1632-1640.
30. Ladd, A.J.C. 1994. Numerical simulations of particle
suspensions via a discretized Boltzmann equation: Part
I. Theoretical foundation. J. Fluid Mech. 271: 285-309.
31. Cook B.K., D.R. Noble. 2004. A direct simulation
method for particle-fluid systems. Eng. Comp. 21: 151-
168.
32. Boutt, D.F., B.K. Cook, B.J.O.L. McPherson, J.R.
Williams. 2007. Direct simulation of fluid-solid
mechanics in porous media using the discrete element
and lattice-Boltzmann methods. J. Geoph. Res. 112:
B10209.
33. Gili, J.A., Alonso, E.E. 2002. Microstructural
deformation mechanisms of unsaturated granular soils.
Int. J. Numer. Ana. Meth. Geomech. 26:433-468.
34. Ibañez, J.P. 2008. Discrete micromechanical modeling
of residual soils. 394p. PhD. Thesis. PUC/Rio. Rio de
Janeiro, Brazil (in Portuguese).
35. He, X., L.-S. Luo. 1997. Lattice Boltzmann Model for
Incompressible Navier-Stokes Equation. J. Stat.
Physics 88(3/4): 927-944.
36. Cundall, P.A. 1971. A computer model for simulating
progressive, large scale movements in blocky rock
systems. In Proc. Symp. Int. Soc. Rock Mech., Nancy 2,
No.8.

Anda mungkin juga menyukai