- Citado por SciELO
versão impressa ISSN 2238-3603
versão On-line ISSN 1807-0302
Comput. Appl. Math. v.23 n.2-3 Petrópolis maio/dez. 2004
Size exclusion during particle suspension transport in porous media: stochastic and averaged equations
A. SantosI; P. BedrikovetskyII
IUniversidade Católica do Rio de Janeiro (PUC-Rio/DEC/GTEP) Rua Marques de São Vicente 225, 22453-900, Rio de Janeiro, RJ, Brazil E-mails: email@example.com
IIUniversidade Estadual do Norte Fluminense (UENF/LENEP) Rod. Amaral Peixoto, km 163 - Av. Brenand s/nº 27925-310, Macaé , RJ, Brazil E-mails: firstname.lastname@example.org
A pore scale population balance model is formulated for deep bed filtration of stable particulate suspensions in porous media. The particle capture from the suspension by the rock occurs by the size exclusion mechanism. The equations for particle and pore size distributions have been derived from the stochastic Master equation. The model proposed is a generalization of stochastic Sharma-Yortsos equations it accounts for particle flux reduction due to restriction for large particles to move via small pores. Analytical solution for low particle concentration is obtained for any particle and pore size distributions. The averaged macro scale equations, derived from the stochastic pore scale model, significantly differ from the traditional deep bed filtration model.
Mathematical subject classification: 76S05, 60H15.
Key words: deep bed filtration, particle size distribution, pore size distribution, size exclusion mechanism, stochastic model, averaging.
Injection of particulate suspensions in porous media has long been recognised to cause formation damage, which occurs due to particle capture by the porous media. This phenomenon is named deep bed filtration and its understanding is essential for industrial and environmental technologies such as water flooding of oil reservoirs, water filtration, wastewater treatments, etc.
Different physical mechanisms (electrical forces, size exclusion, diffusion, bridging, etc.) are responsible for the capture of suspended particles during fluid injection. In the current work, the size exclusion mechanism (where large particles are captured by thin pores) is discussed.
The core-scale deep bed filtration model operates with averaged concentrations for suspended and deposited particles [8, 10]. The model is widely used in industrial filtration [8, 10, 14], in poor quality water injection into oil reservoirs [17, 24] and in environmental studies [7, 12]. Several analytical solutions for direct prediction problems [8, 14, 24] and for inverse data treatment problems [3, 4, 24] were obtained. The model exhibits a good agreement with laboratory data and can be used for prediction purposes, such as forecast of well injectivity decline based on laboratory coreflood data.
Nevertheless, the model does not distinguish between different mechanisms of formation damage. Also, the model operates with phenomenological filtration coefficient and does not include real parameters such as pore and particle sizes, ionic strength and electrical charges. Therefore, it cannot be used for diagnostic predictive purposes based on real physical parameters, such as forecast of well injectivity decline based on pore and particle sizes data.
For the size exclusion mechanism, the larger the particles are and the smaller the pores are, the more intensive the capture is, resulting in major formation damage [1, 18, 20]. Simultaneous zoom of the porous media and flowing particle ensemble would not affect the capture rate. Therefore, the filtration coefficient in the size exclusion mechanism should be a monotonically increasing function of the particle-pore size ratio. Nevertheless, several attempts to correlate the formation damage with sizes of particles and pores were unsuccessful [4, 16]. This failure means that either size exclusion mechanisms did not dominate in the laboratory experiments studied, or that the phenomenological model for average concentrations does not describe adequately the size exclusion filtration. One way around the problem is micro-scale modelling of each capture mechanism.
The capture rate due to the size exclusion mechanism is determined by the concentration of particles that are larger than the pores. Therefore, it is essential to consider the particle and pore size distributions in modelling the size exclusion filtration. The stochastic approach using size distributions has been used to model the deep bed filtration in pore networks [12, 19, 21].
A phenomenological deep bed filtration model accounting for pore and particle size distributions is also present in the literature. The population balance equations are derived in [20-22] from continuity equations for each particle size. The closed system of equations is presented. The model captures change of pore and particle size distributions due to different retention mechanisms. Several analytical solutions have been obtained and interpreted physically. Nevertheless, the model assumes that particles of any size move with carrier water velocity, while in reality particles move only through large pores resulting in reduction of particle flow.
In the current paper, a population balance model for filtration with size exclusion mechanism is derived from the stochastic Master equation . The particle flow reduction due to restriction for large particle flow through small pores is considered. The derived non-linear system allows for linearization for low suspended particle concentration. An analytical solution for injection of low particle concentration suspension is presented. Averaging of the developed model allows proposing corrections to the widely used classical deep bed filtration theory.
In section 1, the classical deep bed filtration model is presented. Its stochastic generalization accounting for pore and particle size distributions is derived in Section 2. The initial-boundary value problem for suspension injection is of Goursat type , yielding exact formulae for particle and pore populations at the inlet cross-section without solving the initial-boundary value problem (Section 3). Section 4 contains the linearized system for the injection of low suspended particle concentration. Exact analytical solutions and averaged equations for deep bed filtration are also derived in section 4. Finally, a section explaining the nomenclature used is included at the end of this paper.
For high injected concentration, the non-linear system derived allows finding asymptotic solutions, which are subject of a forthcoming publication . The analysis of the averaged non-linear system for overall concentrations with analytical solutions for direct and inverse problems will appear in .
1 Classical deep bed filtration theory
Let us discuss the classical deep bed filtration model in order to compare it with the averaged deep bed filtration equations derived further in the paper.
The deep bed filtration model consists of equations for particle balance, for kinetics of the particle capture and of the modified Darcy's law [8, 10]:
where dimensionless length and time are given by the following formulae:
In the above, c(X,T) is the suspended particle concentration (number of suspended particles per unit pore volume), s(X,T) is the deposited particle concentration (number of captured particles per unit rock volume), f is the porosity, L is the core length. The filtration coefficient l(s) expresses the particle capture rate. The formation damage function k(s) shows how permeability decreases due to particle retention.
The velocity U is independent of the linear coordinate X due to the aqueous suspension incompressibility.
The first two equations represent the kinematics of particle transport and capture; the third equation is a dynamical model that predicts pressure gradient increase due to permeability decline resulting from particle deposition. The third Eq. (1.1) decouples from the first and second Eqs. (1.1): the system of the two kinematics equations can be solved independently.
The initial and boundary conditions for particulate suspension injection into a porous media initially filled by particle-free fluid are:
For constant filtration coefficient, the solution of the initial-boundary value problem (1.1), (1.3) for T > X (see [8, 14, 24]) is:
Ahead of the concentration front (T < X) both concentrations are zero.
The suspended concentration distribution reaches the steady state (first formula (1.4)) at the moment T = 1. Therefore, for constant filtration coefficient, the outlet concentration is equal to zero before the particle breakthrough T < 1, and it is constant afterwards. After T = 1, the deposition rate becomes constant. The deposit accumulates, and the deposited concentration grows proportionally to time (second formula (1.4)).
The explicit formula for suspended concentration (1.4) allows calculating the constant filtration coefficient from the constant outlet concentration [17, 24]. For general l(s), the breakthrough curve c(1,T) determines the filtration coefficient, and the solution of the inverse problem is unique and stable .
The solution (1.4) allows for the following physical interpretation of the filtration coefficient - the average dimensionless particle penetration depth is equal to 1/l:
It is worth mentioning that particles move with the carrier water velocity according to the continuity equation (1.1). The analytical solution for one-dimensional deep bed filtration (1.4) contains a suspended concentration shock that moves with the carrier water velocity along the trajectory X = T .
2 Stochastic equations for deep bed filtration
In this chapter we derive a population balance model for the transport of suspension with particle radius distribution through rock with pore radius distribution.
Figs. 1a,b present the geometry of the porous space adopted. The porous medium is represented by a bundle of parallel capillaries periodically intersected by mixing cameras. The radius of capillaries (rp) is distributed. Mixing cameras alternate with capillaries (Fig. 1b). Size exclusion happens at the outlet of each camera: a particle enters a capillary if and only if the particle radius (rs) does not exceed the pore radius.
A stochastic model for deep bed filtration should capture the feature that the larger the particle is, the higher its capture rate is. A particle can be captured only by a smaller pore, rp < rs (the so called size exclusion mechanism of particle retention). Pore and particle sizes vary up to 2-3 orders of magnitude, and there is a significant overlap between pore and particle size distributions causing size exclusion capture [9, 18, 19]. Therefore, the distributions of pore and particle sizes should be taken into account in deep bed filtration models.
Let us consider the concentration C(rs,x,t)drs of suspended particles with radii in the interval [rs,rs+drs]:
where c(x,t) is the total suspended particle concentration (total number of suspended particles per unit volume) and fs(rs,x,t) is the normalised suspended particle size distribution:
Integration of (2.1) considering (2.2) allows calculating the total suspended particle concentration:
The concentration S(rs ,x,t) drs of captured particles with radii in the interval [rs ,rs +drs] and the total captured particle concentration s(x,t) are related in the same way:
and fs(rs ,x,t) is the normalised captured particle size distribution:
The vacant pore concentration with radii in the interval [rp,rp+drp] is H(rp,x,t)drp. It is related to the total vacant pore concentration and pore size distribution as follows:
where fp (rp ,x,t) is the normalised pore size distribution function:
Integration of (2.7) accounting for (2.8) allows calculation of the total vacant pore concentration:
Let us consider the transport of an rs-particle population C(rs ,x,t) through a porous medium with a pore size distribution H(rp,x,t). In a time interval t, the x-coordinates of individual particles will increase by an amount D due to the carrier water flow. The total rs-particle population variation is equal to the sum of all ''transitions'' that accumulate rs-particles at position x, minus the sum of all ''transitions'' that remove rs-particles from the position x (Fig. 2):
where f is the porosity and Y(rs,x,t ;D,t) denotes the probability density that a suspended rs-particle moves from x to x+D during time t without being captured.
The first integral in (2.10) corresponds to all displacements that bring suspended particles to the position x and, hence, it represents a net gain to the function [ fC(rs,x,t)+S(rs ,x,t) ]. Similarly, the second integral (2.10) corresponds to all transitions that remove suspended particles from x and, hence, it represents a net loss to the function [ fC(rs ,x,t)+S(rs ,x,t) ].
The Eq. (2.10) is called the Master equation . The integration on the right hand side of (2.10) is performed from zero to infinity because it is assumed that particles cannot perform negative displacements.
Let us consider the probability b(D,t) dD that the carrier fluid experiences a shift between D and D+dD during a time interval t:
The velocity U is an average fluid displacement per unit time:
Let us assume that, due to the carrier fluid flow, particles perform advective ''jumps'' along the distance x without dispersion. In this case, the probability density b(D,t) for a fluid particle to perform a displacement D during the interval t is:
where d is Dirac's delta function (Fig. 3).
On the other hand, for size exclusion mechanism, only the fraction of the fluid flow through large pores will effectively transport suspended particles. So, the probability Y that an rs-particle flows through pores larger than rs is:
where q(rp ,x,t)drp is the flow through pores with radii in the interval [rp,rp+drp] and the ratio on right hand side of (2.14) is the flow fraction through large pores (Fig. 1a). So, flow of large particles through small pores is not allowed.
Expanding the right hand side of Eq. (2.10) in Taylor series in D results in:
Substituting (2.15) into (2.10) and considering (2.14) and (2.13) we obtain:
The fuild velocity U is given by (2.12). Tending t to zero in (2.16) and neglecting porosity variations due to particle tetention yields:
Eq. (2.18) is the population balance equation for rs-particles.
Let us discuss particle capture and pore plugging kinetics. The probability P that a particle with radius rs meets a pore with radius rp is proportional to the product of the number of particles with radii rs by the fraction of flow through pores with radii rp [8, 20]:
The concentration of particles with radii rs captured by pores with radii rp is S(rs ,rp ,x,t). The number of particles with radii rs captured by pores rp per unit time is called the particle-capture rate. This rate is proportional to the probability P:
The proportionality coefficient is denoted by l'(rs,rp). It equals zero for rs > rp when capture does not occur. Consequently,
Integration of (2.20) in rp from zero to infinity, accounting for (2.21), yields the expression for the total number of captured particles with radii rs per unit time:
Substitution of (2.20) in (2.22) results in:
It is assumed that the capture of one particle plugs only one pore. Therefore, the change of the total number of pores with radii rp is equal to the total amount of particles captured in the pores with size rp:
The integral in (2.24) represents the total number of particles captured in pores with radii rp. The integration on the right hand side of (2.24) is performed from rp to infinity because
|' (rs, rp) = 0 for rs < rp.
Differentiation of (2.24) over t results in:
Substituting (2.20) into (2.25), we obtain:
Eqs. (2.18), (2.23) and (2.26) form a closed system for three unknowns C(rs,x,t), S(rs,x,t) and H(rp,x,t). Following  we assume that the flow through a pore with radii rp is proportional to , as in Hagen-Poiseuille flow . This assumption allows to specify the flux expression in (2.17) and the closed system (2.18), (2.23), (2.26), takes the following form:
Here it is assumed that the particulate suspension is incompressible, i.e. U = U(t), so it was taken out of the brackets in the first equation (2.27). The system (2.27) takes the following form in dimensionless coordinates (1.2):
where l = l' L.
The boundary condition at the core/reservoir inlet (X = 0) corresponds to the injection of particle suspension with a given size distribution. The initial conditions correspond to the absence of particles in the porous medium and to a given pore size distribution of the medium before the injection:
It is important to emphasize that the population balance equation (first Eq. (2.28)), compared with the model proposed in , contains the delay factor a in the advective term:
The water carrying particles with a given size flows through the pores that are larger than the particles. So, the particles are carried by a fraction of the overall flux U (Fig. 1a). Consequently, the delay factor is smaller or equals one, i.e. the particle front lags behind the injected water front.
3 Dynamics of particle and pore populations at the inlet cross section
The second and third equations of the system (2.28) do not contain X-derivative, so the corresponding species concentrations should not be posed at the inlet boundary X=0 (this is a Goursat problem, ). This means that the concentration of the immobile species should not be specified at the inlet. Nevertheless, these values can be calculated using the boundary conditions for the mobile species (2.29) and the kinetic equations for the immobile species (second and third equation of the system (2.28)).
Let us find the particle-capture and pore-plugging kinetics at the inlet cross section X = 0. We substitute the boundary condition (2.29) into the second and third equations of the system (2.28). This substitution yields a system of two ordinary integro-differential equations for captured particle and vacant pore concentrations at the inlet cross-section:
where: H(0)(rp ,T) = H(rp ,X = 0,T), S(0)(rs ,T) = S(rs ,X = 0,T).
The second equation (3.1) is independent of the first equation, and it can be solved separately. Afterwards, the first equation allows calculating the deposition kinetics.
There are no deposited particles and plugged pores at the beginning of deep bed filtration. This fact provides the initial conditions for the system of ordinary integro-differential equations (3.1):
The solution of the system of ordinary integro-differential equations (3.1) determines the particle-capture and pore-plugging kinetics at the inlet cross section.
4 Low suspended particle concentration
In many deep bed filtration processes of practical interest, such as flooding of petroleum reservoirs by seawater or water filtration in aquifers, the injected suspended particle concentration is low:
This assumption allows to linearize the basic system of governing equations (2.28) and to obtain exact analytical solutions for deep bed filtration of low concentration suspensions. It also permits derivation of averaged equations for the transport of low concentration suspensions in natural rocks.
4.1 Linearization of governing equations
Substituting the second equation (2.28) into the first one, we reduce the system of three equations (2.28) to two equations:
Let us consider a suspension with particle radius distribution constant in time injected into a rock with homogeneous pore radius distribution. So, the boundary and initial distributions in (2.29) become:
For low concentration suspensions (4.1), the problem (4.2), (4.3) contains a small parameter
where: cinj = Cinj( rs) drs and h0 = H0(rp ) drp .
Let us find the first order expansion of the solution for the problem (4.2), (4.3) over the small parameter e:
Substituting the expansion (4.5) into (4.2) and (4.3), we obtain the zero order approximation system:
subject to initial and boundary conditions
The solution C0(rs,X,T) = 0 satisfies the first equation (4.6) and boundary conditions (4.7). From the second equation (4.6) and (4.7) it follows that
Therefore, if only the zero order approximation is considered, the solution shows that there is no particle transport or pore plugging.
In first order approximation, the initial-boundary value problem (4.2), (4.3) becomes:
Let us consider the case of constant l. From the first equation (4.9) it follows that the flux reduction factor a is independent of X and T. In this case, the first two equations (2.28), considering the first order approximation (4.5), become linear:
The carrier water velocity in dimensionless variables (1.2) is equal to one. The velocity of characteristic lines in the first equation (4.11) (i.e., the velocity of particles with size rs) equals a(rs) < 1. The capture rate in the model (2.28) is proportional to the flow through small pores, so in (4.11) the term [1-a(rs)] does appear in the expression for the capture kinetics. Particles of different sizes undergo deep bed filtration independently of each other.
The effect of particle velocity reduction compared with carrier water velocity was observed in laboratory tests .
4.2 Analytical solution
The analytical solution for the linear system (4.11) subject to initial and boundary conditions (2.29) is obtained using the method of characteristics :
The concentration front for particles with size rs moves with velocity a(rs) (see (4.13) and (4.15)). The decrement of the suspension concentration is also a function of particle size.
Let us consider the existence of minimum and maximum pore sizes. The flux reduction term a(rs) for large particles (rs > rpmax) equals zero. From Eqs. (4.13)-(4.15) it follows that:
So, large particles (rs > rpmax) do not penetrate into the porous medium. All large injected particles are captured at the inlet cross-section (X = 0). The deposited concentration at the inlet cross section is obtained from the first equation (3.1):
For small particles (rs < rpmin), the flux reduction term a(rs) is equal to one, Eq. (4.12). For a(rs) = 1, Eqs. (4.13)-(4.15) become:
Equations (4.18)-(4.20) show that small particles flow through porous media without being captured.
Fig. 4 exhibits the concentration waves for small, intermediate and large particles. Notice that the particle concentration for each particle size C(rs,X,T) is zero ahead of the propagation fronts: Xf(rs,T) = a(rs)T (see (4.15)). The concentration wave for small particles propagates through the reservoir without change (curve 1), while the concentration of large particles in the reservoir equals zero (curve 4). The smaller the particle is, the faster its concentration front propagates and the slower its capture is (curve 2 is located ahead and above curve 3).
The dynamics of suspended particle and vacancy distributions is shown in Fig. 5. The injected particle size distribution and the initial pore size distribution are presented in Fig. 5a. Fig. 5b shows the suspended particle and vacancy distributions for T > 0. In first order approximation, the flux reduction factor is not affected by deep bed filtration of low concentration suspension. In the interval [rpmin,rpmax] the particle concentration decreases from Cinj(rs = rpmin) to zero for rs = rpmax.
The larger the particles are, the larger the decrement (1-a(rs))/a(rs) of the exponent in (4.13) is. So smaller particles have higher relative concentration (C(rs,X,T)/Cinj(rs)) and their concentration profile moves faster. Fig. 6 shows different particle size concentration history at the core outlet (X = 1). The smaller the particle is, the earlier it arrives to the outlet and the higher its relative concentration is.
The simplified assumption that the filtration coefficient is constant was used in the current and previous sections in order to illustrate the effects of the flux reduction factor a on the velocity of the concentration front and on the concentration decrement. In reality, analytical solutions like (4.13)-(4.15) can be obtained for any l = l(rs , rp). When the filtration coefficient is independent of pore radius, l = l(rs), the obtained system (4.11) is not changed. Therefore, the solution (4.13)-(4.15) is still valid. In the general case l = l(rs ,rp), the term l(1-a) in (4.11), (4.13) and (4.14) must be substituted by
i.e., in general, the effects of capture and flux reduction cannot be separated.
For injection of low suspended particle concentration, the system (3.1) allows calculating immobile concentrations at the inlet cross-section:
The lower limit of the integral in (4.22) is rpmin because S(rs,X,T) is zero for rs < rpmin.
4.3 Penetration depth
Let us calculate the average dimensionless penetration depth for particles with different sizes á X(rs ,T) ñ :
where: a = a(rs).
The particle concentration C(rs,X,T) is zero ahead of the propagation front Xf(rs,T) = aT, so the integration in (4.23) is performed from zero to aT. Substituting (4.13) and (4.15) into (4.23), we calculate the average dimensionless penetration depth dynamics for particles with intermediate sizes:
The maximum dimensionless penetration depth distribution for an ensemble with a fixed particle size á X(rs ) ñ max is obtained by letting T tend to infinity in (4.24):
Substituting (4.24) into (4.25) we obtain the expression for maximum dimensionless penetration depth:
Fig. 7 shows how the maximum penetration depth depends of the particle radii. Three different probability distribution functions (PDF) were used for pore sizes: normal, lognormal and uniform. These three PDFs are bi-parametrical, so we assumed the same mean and standard deviation values ( á rp ñ = 11 mm, s = 0.577mm) for the three examples.
Particles with radii larger than rpmax do not penetrate into porous medium, so á X(rs > rpmax ) ñ max = 0. Particles with radii smaller than rpmin flow without being captured, so á X(rs < rp min ) ñ max tends to infinity when the time T tends to infinity. Finally, the smaller the particles are, the deeper they penetrate into the porous medium (Fig. 6).
4.4 Averaged concentration model
Let us derive the model for the average concentration and compare it with the classical model for deep bed filtration (1.1).
For small particles (rs < rpmin), the flux reduction factor equals one (a = 1). Integration of both equations (4.11) in rs from zero to rpmin, yields the averaged equations for the population of small particles:
where c1 is the total concentration of small particles:
The solution for the system (4.27) subject to initial and boundary conditions (2.29) is:
Equations (4.29) and (4.30) show that small particles move with the carrier water velocity without entrapment.
The equations for the total concentrations of intermediate size particles are obtained by integration of the first and second Eqs. (4.11) in rs from rpmin to rpmax:
where c2 is the total concentration of intermediate size particles:
In Eq. (4.31), the average flux reduction factor áañ for intermediate size particles is:
So, when compared with the classical deep bed filtration model (1.1), the advective term in the population balance equation (4.31) for intermediate size particles contains an average flux reduction term á a ñ . The second equation (4.31) shows that the filtration coefficient for the average concentration model equals the filtration coefficient for the classical model (1.1) times (1 - á a ñ). The term (1 - á a ñ) corresponds to the fraction of water flux through thin pores where the capture takes place.
For large particles, a(rs) is equal to zero. Integration of system (4.11) in rs from rpmax to infinity allows obtaining the averaged model for large particles:
The solution for the system (4.34) is:
Equation (4.36) shows that there is no transport of large particles; all the suspended large particles are captured by the porous medium at the inlet cross section.
The solutions for systems (4.27), (4.31) and (4.34) show that only intermediate size particles perform deep bed filtration. This means that in the classical theory for deep bed filtration the concentration "c" used in system (1.1) must be equal to c2. Fig. 8 illustrates the portions of overall particle concentration (dashed areas) that should be neglected in the classical theory for deep bed filtration (1.1).
Furthermore, the flux reduction factor (4.33) should be taken into account in the classical theory (1.1).
5 Summary and conclusions
A population balance model for deep bed filtration with size exclusion mechanism was derived from the Master equation. The model accounts for particle flux reduction due to motion of particles with a given size in larger pores only.
For low concentration suspensions, the change of flux reduction factor due to pore plugging can be neglected, allowing linearization of the governing equations. The linearized deep bed filtration problem has exact analytical solution for any particle and pore size distributions. The analysis of the solution leads to the following conclusions:
1. The absence of rs-particle flow in pores smaller than rs results in particle flux reduction when compared with the carrier water flux. Consequently, a flux reduction factor a appears in the advective term of the population balance equation.
2. The capture rate is proportional to the flux through small pores. So, the factor (1-a) appears in the expression for the capture kinetics.
3. Populations of particles with different sizes filtrate independently. The velocity of each population is a(rs) times lower than the carrier water velocity. The capture rate for each population is also a function of the particle radius.
4. The penetration depth for the intermediate size particles equals the flux reduction factor a divided by the filtration coefficient times (1-a). The smaller the particles are, the deeper they penetrate.
5. The averaged model includes three systems of equations: one for small particles, another for large particles and the third for intermediate size particles.
6. Deep bed filtration occurs only when there is an overlap between pore and particle size distributions. Particles larger than the maximum pore radii do not enter the porous media and particles smaller than the minimum pore radii are transported without being captured. Consequently, only intermediate size particle concentration should be taken into account in the classical deep bed filtration theory (1.1) for size exclusion mechanism.
7. The averaged model for intermediate size particles differs from the traditional deep bed filtration model by the flux reduction factor a that appears in the advective term of population balance equation. This difference is due to particle motion through large pores only. The two models also differ from each other by the filtration coefficient expression, which in the averaged model is proportional to (1-a). This is due to particle capture by small pores only.
This work was supported by FINEP (under CTPETRO grants 64.99.0465.00, 21.01.0386.00 and 65.00.0327.00), ANP and FENORTE.
Authors are mostly grateful to Petrobras colleagues A. G. Siqueira, A. L. de Souza and F. Shecaira Shecaira for fruitful and highly motivated discussions during several years. Many thanks are due to Prof. Dan Marchesin (Institute of Pure and Applied Math) and Prof. A. Shapiro (Technical University of Denmark) for useful suggestions and improvement of the text. Detailed discussions with Prof. Yannis Yortsos (University of South California) are also highly acknowledged.
Especial gratitude is due to Themis Carageorgos and Poliana Deolindo (North Fluminense State University, Brazil) for permanent support and encouragement.
total suspended particle concentration, no. L-3
concentration distribution of suspended particles with radius rs, no. L-4
probability distribution function, L-1
total vacant pore concentration, no. L-3
concentration distribution of vacant pores with radius rp, no. L-4
particle flux, L-3.T-1
core length, L
probability density that a particle with radius rs meets a pore with radius rp, L-2
pore radius, L
particle radius, L
standard deviation, L
concentration distribution of retained particles with radius rs, no. L-4
dimensional and dimensionless times
fluid velocity, L.T-1
dimensional and dimensionless linear co-ordinates
á X ñ
dimensionless average penetration depth
flux reduction factor
Dirac's delta function
dimensionless filtration coefficient
total captured particle concentration, no. L-3
 F. Al-Abduwani, van den Broek and Currie, P.K., Visual Observation of Produced Water Re-injection Under Laboratory Conditions, SPE 68977 presented at the SPE European Formation Damage Conference, The Hague, 21-22 May, (2001). [ Links ]
 B. Amix, R. Bass and A. Whiting, Applied Reservoir Engineering, McGraw Hill Book Co, NY, (1964). [ Links ]
 P. Bedrikovetsky, D. Marchesin, G. Hime, A. Alvarez, A.O. Marchesin, A.G. Siqueira, A.L.S. Souza, F.S. Shecaira and J.R. Rodrigues, Porous media deposition damage from injection of water with particles, Proc., VIII ECMOR (European Conference on Mathematics in Oil Recovery), Austria, Leoben, Sept. 3-6, Sept., (2002). [ Links ]
 P. Bedrikovetsky, P. Tran, W.M.G.T. Van den Broek, D. Marchesin, E. Rezende, A. Siqueira, A.L. Souza and F. Shecaira, Damage Characterization of Deep Bed Filtration from Pressure Measurements, SPE Journal P&F, 3 (2003), 119-128. [ Links ]
 P. Bedrikovetsky, A. Santos, A. Siqueira, A.L. Souza and F. Shecaira, A Stochastic Model for Deep Bed Filtration and Well Impairment, SPE 82230 presented at the SPE European Formation Damage Conference to be held in The Hague, The Netherlands, 13-14 May (Submitted to SPE Journal in 2003). [ Links ]
 P. Bedrikovetsky and A. Santos, A Stochastic Model for Particulate Suspension Flow in Porous Media, submitted to Transport in Porous Media, (2003). [ Links ]
 G. Elimelech, J. Gregory, X. Jia and R.A. Williams, Particle Deposition and Aggregation, Butterworth-Heinemann, USA. (1995). [ Links ]
 J.P. Herzig, D.M. Leclerc P.L. and Goff, Flow of Suspensions through Porous Media - Application to Deep Filtration, Industrial and Engineering Chemistry, 62 (5) (1970), May, 8-35. [ Links ]
 A.O. Imdakm and M. Sahimi, Transport of large particles in flow through porous media, Physical Review A, 36 (1987), 5304-5309. [ Links ]
 T. Iwasaki, Some notes on sand filtration, J. Am. Water Works Ass., (1937) 1591-1602. [ Links ]
 N.G. Van Kampen, Stochastic Processes in Physics and Chemistry, North-Holland, Amsterdam-Oxford, (1984). [ Links ]
 K. Khilar and S. Fogler, Migration of Fines in Porous Media, Kluwer Academic Publishers, Dordrecht/London/Boston, (1998). [ Links ]
 L.D. Landau and E.M. Lifshitz, Fluid Mechanics (Course of Theoretical Physics, v.6), 2nd edition, Pergamon Press, Oxford, (1987). [ Links ]
 J.D. Logan, Transport Modelling in Hydrogeochemical Systems, Springer-Verlag, NY-Berlin, (2000). [ Links ]
 N. Massei, M. Lacroix, H.Q. Wang and J.P. Dupont, Transport of particulate material and dissolved tracer in a highly permeable porous medium: comparison of the transfer parameters, Journal of Contaminant Hydrology 57 (2002), 21-39. [ Links ]
 E. Van Oort, J.F.G. Van Velzen and K. Leerlooijer, Impairment by Suspended Solids Invasion: Testing and Prediction, SPE 23822, (1993). [ Links ]
 S. Pang and M.M. Sharma, A Model for Predicting Injectivity Decline in Water Injection Wells, SPE 28489, 275-284. Presented at Society of Petroleum Engineers Annual Technical Conference and Exhibition, New Orleans, LA, September, (1994). [ Links ]
 A.S. Payatakes, R. Rajagopalan, and C. Tien, Application of Porous Medium Models tothe Study of Deep Bed Filtration, The Canadian Journal of Chemical Engineering, 52(1974), 727. [ Links ]
 A.G. Siqueira, E.J. Bonet and F.S. Shecaira, A 3D Network Model of Rock Permeability Impairment Due to Suspended Particles in Injected Water, SPE 82232 presented at the SPE European Formation Damage Conference held in The Hague, The Netherlands, 13-14 May, (2003). [ Links ]
 M.M. Sharma and Y.C. Yortsos, Transport of Particulate Suspensions in Porous Media: Model Formulation, AIChe J., 33 (10) (1987a), 1636. [ Links ]
 M.M. Sharma and Y.C. Yortsos, A network model for deep bed filtration processes, AIChE J., 33 (10) (1987b), 1644-1653. [ Links ]
 M.M. Sharma and Y.C. Yortsos, Fines migration in porous media, AIChE J., 33 (10)(1987c), 1654-1662. [ Links ]
 A.N. Tikhonov and A.A. Samarskii, Equations of Mathematical Physics, Dover, New York, (1990). [ Links ]
 K.E. Wennberg and M.M. Sharma, Determination of the filtration coefficient and the transition time for water injection, Proceedings of the SPE European Formation Damage Conference, SPE 38181, The Hague, The Netherlands, June 2-3, (1997). [ Links ]
Received: 31/III/03. Accepted: 24/XI/04.
This article has received corrections in agreement with the ERRATUM published in Volume 24 Number 3.