Particle Technology and Fluidization

lncompressive Cake Filtration: Mechanism,

Parameters, and Modeling
Faruk Civan
School of Petroleum and Geological Engineering, The University of Oklahoma, Norman, OK 73019

Improved models developed verify interpretation and prediction of incompressible fil-

ter cake thickness, filtrate volume, and rate data for linear and radial filtration cases at
static and dynamic conditions. Based on the methods for determining model parameters
from experimental filtration data, three sets of experimental data were analyzed, which
demonstrate the diagnostic and predictive capabilities of the model, These models pro-
vide insight into the mechanism of incompressible cake filtration and offer practical
means of interpreting experimental data, estimating the model parameters and simulat-
ing the linear and radial filtration processes.

Introduction
Many industrial filtration processes facilitate plate and However, the model is also valid for the reverse operation.
drum filters for slurry filtration. Instruments used for various Consider incompressible particles, carrier fluid, and filter
purposes may utilize porous plate, tube, and hollow disk type cake, no small particle invasion and deposition within the fil-
filters. Linear and radial filtration models are needed for ter cake, and a particle free filtrate. The inner and outer radii
prediction, simulation, and optimization of the various filtra- (cm) of the hollow cylindrical filter are denoted, respectively,
tion processes, and analysis and interpretation of experimen- by rw and re. The radius of the slurry side cake surface is
tal data. Simple models, such as those presented in this study, denoted by rc. The thickness and width (cm) of the cake are
are preferred in many applications, because of their conve- denoted by S and h, respectively. The cross-flow (or tangen-
nience and the reduced computational effort. tial) velocity of the slurry flowing over the inner filter surface
The applicability of the majority of the previously reported (cm/s> is denoted by v and the filtration velocity of the parti-
simple analytical models, such as by Collins (19611, Hermia cle free carrier fluid normal to the inner filter surface (cm/s)
(19821, and de Nevers (19921, are usually limited to linear by u. The filtration conditions are referred to as the static
and constant rate filtration. However, models for constant when v = 0 and the dynamic when v > 0.
pressure filtration are also required for certain applications. The volumetric balance of the cake forming particles is
When the filter cake and the filter thicknesses are small rela- given by (Civan, 1994, 1996)
tive to the radius of the filter surface exposed to slurry, linear
filtration models can sufficiently approximate radial filtra-
tion. Otherwise, radial filtration processes can be better de-
scribed in radial geometry. in which the volume fraction of the cake forming particles eS
In the following, improved simple filtration models for ra- in terms of the cake porosity +c is given by
dial and linear incompressible cake filtration processes, with
additional features such as considering variable rate, effect of
filter thickness, and static and dynamic filtration conditions,
are presented and verified by means of the simplified analyti-
t is time(s), p, is the particle density (g/cm3), and R,, de-
cal models and the reported experimental data.
notes the net rate of particle deposition over the progressing
cake surface (gm/s/cm2).
Formulation The present problem deals with a forced filtration during
For the derivation of the radial filtration model, consider the flow of slurry over the filter surface at a sufficiently rapid
that the slurry is applied to the inner surface and the filtrate rate, such as in the case of the filtration of the drilling mud
leaves from the outer surface of a hollow cylindrical filter. slurry in wells. In this case, deposition of particles is directly

AIChE Journal November 1998 Vol. 44,No. 11 2379

related to the filtration mass flux of particles of the slurry roughness, and particle stickiness (Civan, 1996) on the parti-
approaching the filter, colloidal detachment is negligible be- cle detachment. Therefore, the actual value of the critical
cause there is no alteration of the physico-chemical condi- shear stress may be several fold different than that predicted
tions in ordinary filtration, and erosion of particles from the by Eq. 5 using the particle size and separation distance data.
cake surface by the hydrodynamic shear force is dominant by Hence, Ravi et al. (1992) recommend experimental determi-
several orders of magnitude. Particles can only be detached if nation of the critical shcar stress.
the prevailing fluid shear exceeds the critical, minimum shear The filter cake thickness S (cm) is the difference between
stress necessary to dislocate the particles from the cake sur- the radius, r, of the filter surface exposed to the slurry, and
face. It takes a certain minimum shear stress to dislocate par- the filter cake radius re
ticles from their places on the cake surface, especially if the
particles are sticky. Peng and Peden (1992) omitted this. It is
reasonable to assume that the erosion rate is proportional
with the excess of the shear stress above the minimum critical Thus, substituting Eqs. 2 and 3 into Eq. I yields (Civan. 1994,
shear stress necessary for detachment of the cake particles 1996)
from the filter cake surface. Therefore, the net mass rate of
deposition of particles per unit area of the cake surface is dr,. k,,u, c, - k,,(.r - r,,.)
given by the difference between the deposition and the ero- (7)
sion rates as (Civan, 1994, 1996)

in which set k , = 0 when 7 < rC1.

The cake formation begins at the filter surface exposed to
where u , is the carrier fluid filtration velocity normal to the the slurry, and, therefore, the initial cake radius is given by
slurry side cake surface (cm/s). c p is the slurry particle con-
centration expressed as the mass of particles per unit volume
of the carrier fluid in the slurry (g/cm3 carrier fluid). For
dynamic filter cake formation, the non-Newtonian slurry fluid The volumetric flux of the particle frce carrier fluid. that is,
wall shear-stress (dyne/cm2) T is given by the Rabinowitsch- the filtrate, into the cake surface is given by
Mooney equation (Metzner and Reed, 1955)
4
u, = - (9)
2?rr,.lz

where k' and n' are the consistency constant (dyne/cm2/s"') where q is the particle free carrier fluid filtrate flow rate
and flow index (dimensionless), respectively. For Newtonian (cm3/s).
slurry fluids, k' = p where p is the slurry viscosity (cp) and Applying Eq. 9, Eq. 7 can be written as
n' = 1. k , and k, are the particle deposition (dimensionless)
and erosion rate constants (s/cm), and T~~ is the minimum
critical fluid shear stress necessary to dislocate the particles (10)
from the cake surface (dyne/cm2). Therefore, set k , = 0 for
T < T ~ For~ . static filter cake formation v = 0, and, therefore,
T = 0. Hence, the second term on the right of Eq. 3 drops where
out.
The deposition and erosion rate constants depend on the (11)
properties of the particles and carrier fluid, and the condi-
tions of the slurry, such as particle concentration, flow rate,
and pressure. and
Ravi et al. (1992) have determined that the following equa-
tion proposed by Potanin and Uriev (1991) predicts the criti-
cal shear stress with the same order of magnitude accuracy of
their experimental measurements

T~~ = H/(24di2) The volumetric flux of the particle free filtrate is given by
Darcy's law
where H = 3.0X lo-" erg is the Hamaker coefficient (erg), rl
is the average particle diameter, and 1 is the separation dis- (13)
tance (cm) between the particle surfaces in the filter cake.
However, the values calculated from Eq. 5 can only be used
as a first-order accurate estimate because Eq. 5 has been de- Note that the volumetric ratcs of fluids flowing through the
rived from an ideal theory. The actual value can be different, filter cake and filter media in series are the same, irrespec-
because the ideal theory does not take into account the effect tive of whether it is a constant rate or a constant pressure
of the other factors, such as aging (Ravi et al., 1992), surface filtration. In constant pressure filtration, the rate varies but

2380 November 1998 Vol. 44, No. 11 AIChE Journal

the rates of fluid flowing through the filter cake and the filter which can be numerically evaluated by the Gaussian quadra-
media are equal at all times. In constant rate filtration, the ture rule.
slurry precsure varies when the filter effluent side pressure is The radial flow equations derived above can be readily
kept constant. Thus, even when q varies temporally, its value converted to a linear one-dimensional case of laboratory core
remains spatially the same over the radial distance at any tests by means of the following transformation
given time. Thercfore, Eq. 13 can be integrated over the ra-
dial distance rL 5 r I re before and during the cake buildup x = r2 (23)
from the slurry to the effluent sides to obtain the following
pressure over halance expressions, respectively Thus, the equations corresponding lo Eqs. 14,15, 17-19,9-12
and 20-22 are obtained, respectively, as
(14)

where p , and p, are the pressures of the slurry and the filter where a is the cross-sectional area of the filter (cm').
outlet (atm), respectively. qo and q are the flow rates before Equating Eqs. 24 and 25 and rearranging, the filter cake
and during the filter cake buildup (cm3/s). K , and K f are thickness is given by
the filter cake and filter permeabilities (darcy).
Thus, equating Eqs. 14 and 15 and rearranging leads to 6 = C/q -D (26)

where

Equation 16 can be written as

r,/r,=exp(-C/q+ D) (17)

where
The filter cake volumetric balance leads to

for which y,, is determined by Eq. 14 and d 6/dt = A4 - B (30)

where

Thus, substituting Eq. 16 into Eq. 13 and rearranging yields

the filtration flow rate equation as

dq/dt = ( - 1jC ) q 2 [Aq exp (C/q - D ) - B]exp(C/q - D )

in which the shear-stress for linear filtration is given by
(20) (Metzner and Reed, 1955)

subject to the initial condition given by

T = k'(8v)"' (33)
q=40, t = O (21) Invoking Eq. 26 into Eq. 30 leads to
where q, is the injection rate given by Eq. 14 before the filter
cake buildup.
The wall shear-stress is calculated by Eq. 4 €or the varying
cake radius r, = r < ( t ) .The filter cake thickness is calculated subject to
by means of Eqs. 6 and 17. Equations 20 and 21 can be solved
numerically using an appropriate method such as the
q=q,, t=O (35)
Runge-Kutta method. However, for thin cakes, it is reason-
able to assume that the wall-shear stress is approximately The analytical solution of Eqs. 34 and 35 is
constant because rc = rw. Then, Eq. 20 can b e integrated as

C A
t=-- - In
B B

AIChE Journal November 1998 Vol. 44, No. 11 2381

The cumulative filtrate volume is calculated by t A 1
_ = _
Q+- (45)
Q C (I,

Q = /'qdt (37)
0 Equation 45 is similar to Eq. 12 of de Nevers (1992) and Eq.
21 of Hermia (1982).
Eliminating q between Eqs. 26 and 36 yields another expres- As can be seen by these exercises, the present improved
sion as model can be simplified to reproduce the models given by
Collins (19611, Hermia (1982), and de Nevers (1992) when

'{
B
1"' B
B
A
1 B - C
C
S0+D
the inherent simplifying assumptions of these models are in-
troduced into the present model

Determinationof Model Parameters and Diagnostic

Charts
in which, usually, 6, = 0 at t = 0, that is, no initial filter cake. The majority of the reported filtration studies have not
Equation 38 is different from Eq. 7-94 of Collins (19611, made attempts at measuring a full set of measurable parame-
because Collins did not consider the filter cake erosion. ters. The filtration models presented in this study may pro-
Therefore, Collins' equation applies for static filtration. To vide some guidance for the types of parameters which are
obtain Collins' result, k , = 0 or B = 0 must be substituted in needed for simulation studies.
Eq. 30. Thus, eliminating q between Eqs. 26 and 30, and As listed in Table 1, the present filtration models require
then integrating, yields the following equation for the filter the values of 20 parameters for simulation. Only five of these
cake thickness parameters may not be directly or conveniently measurable
with the conventional techniques. These are the permeability
K , and porosity CPC of the filter cake, the deposition and ero-
sion rate constants k, and k,, and the critical shear-stress T,,
which results in Eq. 7-94 of Collins (1961) by invoking Eqs. for the particles. However, given the experimental measure-
24, 27, 28, and 31 and expressing the mass of suspended par- ments of the filtrate volume Q (cm3>or rate q and the filter
ticles per unit volume of the carrier fluid in terms of the vol- cake thickness 6 as functions of the filtration time t, some of
ume fraction up of the particles in the slurry according to these parameters can be determined by means of the diag-
nostic charts constructed as described in the following. These
are presented separately for the linear and radial filtration
processes.
Collins (1961) did not derive the expressions for the filtrate
flow rate and the cumulative filtrate volume. These expres- Linear filtration
sions can be readily obtained by integrating Eq. 34 for B = 0 A plot of Eq. 30 for d6/dt vs. q yields a straight line. Sub-
and applying Eq. 37, respectively, as stituting the slope (A) and intercept ( - B) of this line into
Eqs. 31 and 32 yields, respectively, the following expressions
for the particle deposition and erosion rate constants

and

Equation 41 expresses that the filtrate rate declines by time In dynamic filtration, the filter cake thickness attains a cer-
due to static filter cake buildup. Donaldson and Chernoglazov tain limit value 8, when the particle deposition and erosion
(1987) represented such rate declines by an empirical decay rates equate. Simultaneously, the filtration rate also reaches
function a limiting value, determined by Eq. 30 as

in which p is an empirically determined coefficient.

The filtration equations presented by Hermia (1982) and At this condition, Eq. 26 yields the limiting value of the filter
de Nevers (1992) can be derived by simplifying the improved cake thickness as
model developed in this work. For constant rate ( q ) filtra-
tion, Eq. 37 yields

Q = qt (44) Consequently, substituting Eqs. 27, 28, 31, and 32 for A, B,

C, and D into Eqs. 48 and 49 leads to the following relation-
Thus, invoking Eq. 44 into Eq. 42 results in the conventional ships for the cake permeability and the ratio of the erosion
filtration equation as and deposition rate constants, respectively, as

2382 November 1998 Vol. 44, No. 11 AIChE Journal

 Table 1. Data for Laboratory Filtration Applications
Radial Flow Linear Flow Linear Flow
Fisk et al. Jiao and Sharma Willis et al.
Parameters (1991) (1994) (1983)
Suspension type Seawater-based partially Fresh water ben- Lucite in water
hydrolized polyacrylamide tonite suspension suspension
drilling mud
Filter permeability, K f (darcy) 6* * 0.104' -
Cake permeability, K , (darcy) 1.35 X 10W6 2.1 x 1 0 - ~ -
Cake porosity, I $ ~ 0.4@ 0.4@ 0.388%
Filter length, Lf (cm) - 20.34' -
Filter diameter, D (cm) - 2.54' -
Slurry injection side filter radius, rw (cm) 2.5* * - -
Filtrate outlet side filter radius, re (cm) 3.8** - -
Filter width, h (cm) l.P* - -
Filtrate density, pw (g/cm3) 1.0 1.0 0.997'
Particle density, pp (g/cm3) 2.5* 2.5* 1.18*
Particle mass per carrier fluid volume, c p (g/cm') 0.56** 0.04' 0.055'
Deposition rate constant, k , 1.1 4.3 -
Erosion rate constant, k, (s/cm) 3 X lo-' 7.4~10-7 -
Critical shear-stress, T ~ ,(dyne/cm') 0.5 5.0 -
Filtrate (water) viscosity, p (cp) 1.0 1.0 0.969%
Consistency constant, k' (dyne/cm*/s"') 8.0' 8.0' -
Flow index, n' 0.319' 0.319' -
Slurry tangential velocity, u (cm/s) 125.** 8.61' -
Slurry application pressure, p c (atm) 34.* * 6.89' 1.7'
Filter outlet side back pressure, p , (atm) 1 1 1*

*Data assumed.
**Data from Fisk et al. (1991).
'Data from Jiao and Sharma (1994)
'Data from Willis et al. (1983).

and k,, if the cake porosity +c and the critical shear stress 7cr
are known. += can be measured. rcr can be estimated by Eq.
5, but the ideal theory may not yield a correct value as ex-
plained previously by Ravi et al. (1992) and in this article.
Equation 34 can be rearranged in a linear form as Therefore, Ravi et al. (1992) suggested that 7,, should be
measured directly.

Radial filtration
Given the filter cake thickness 6, the progressing surface
cake radius r, can be calculated by Eq. 6. Then a straight
Thus, the intercept ( B / C ) and slope ( - A / C ) of the line plot of ln(rc/rw) vs. (l/q) data according to Eq. 17 yields
straight-line plot of Eq. 52 can be used with Eqs. 27, 28, 31, the values of C and D as the slope and intercept of this line,
and 32 to obtain the following expressions respectively. A straightline plot of [ - d/dt(rJr,,,)] vs.
[q/frc/rw)]data according to Eq. 10 yields the values of A
and B as the slope and intercept of this line, respectively. At
(53) static filtration conditions, v = 0 and T = 0 according to Eq.
4. Therefore, B = 0 according to Eq. 12. Consequently, sub-
stituting B = 0 and Eq. 18, Eq. 20 can be expressed in the
following linear form

Comparing Eqs. 51 and 53 yields an alternative expression

for determination of the limit filtrate rate as

Thus, a straightline plot of In[ - q - 3 dq/dt] vs. ( l / q ) yields

Equation 55 can be used to check the value of qm obtained by the values of ( 2 C ) and [In(A/C)-2CKf/q,] as the slope and
Eq. 48. Equation 50 can be used to determine the filter cake intercept of this line, respectively. This allows for determina-
permeability K,. Equations 46 and 51 or 53 and 54 can be tion of the A and C coefficients only. The determination of a
used to calculate the particle deposition and erosion rates k , full set of A , €3, C, and D from Eqs. 10 and 20 requires both

AIChE Journal November 1998 Vol. 44, No. 11 2383

the filtrate flow rate (or volume) and the cake thickness vs. 02 I I I
the filtration time data. Once these coefficients are deter-
mined, then their values can be used in Eqs. 11, 12, 18, and
19 to determine the values of the deposition and erosion rate
constants k , and k,. The discussion of the linear filtration
about the determination of T ~ , .by Eq. 5 is valid also in the
radial filtration case.
At dynamic equilibrium, the filter cake thickness and the
filtrate flow rate attain certain limiting values S, and qx.
Then, substituting Eq. 6 into Eqs. 10 and 17 yields the follow-
ing relationships, respectively
0
j I
I
1

Aq, = B(1- S&,) (57) 10 20 30 40 50 60

Reciprocal Filtrate Flow Rate, l/q, rninlml
1 - SJr, = exp(- C/q, +D) (58)
Figure 2. Correlation of Jiao and Sharma (1994) pre-
dicted filter cake thickness data according to
The filter cake permeability is determined by Eq. 19 as Eq. 26.

the filtrate volumes or ratcs. Therefore, to the author's

The equations and the linear plotting schemes developed knowledge. there is no one complete set of suitable data
in this section allow for determination of the parameters of available in the literature that can be used for full-scale ex-
the filtration models, mentioned at the beginning of this sec- perimental verification of the present formulations. Here, the
tion, from experimental filtrate flow rate (or volume) and/or Willis et al. (1983) and Jiao and Sharma (1994) data for lin-
filter cake thickness data. The remaining parameters should ear filtration and the Fisk ct al. (1991) data for radial filtra-
be either directly measured or estimated. In the following tion are used, because these data provide more information
applications, the best estimates of the missing data have been than the other reported studies. The data arc presented in
determined by adjusting their values to fit the experimental Table 1 in consistent Darcy units, which are more convenient
data. This is an exercise similar to several other studies, in- for flow through porous media.
cluding the ones by Liu and Civan (1996) and Tien et al. Jiao and Sharma (1994) carried out linear filtration experi-
(1997). They have resorted to a model-assisted estimation of ments using concentrated bentonite suspensions. They only
the parameters, because there is no direct method of mea- measured the filtrate volume and predicted the filter cake
surement for some of these parameters. thickness using a simple algebraic model. These data are given
in their Figures 3 and 10, respectively. In Figures 1 to 3, their
Applications data are plotted according to the linear plotting schcmes pre-
sented in the previous section for determination of parame-
The numerical solutions of the present models require the ters. As can be seen from these figures, the coefficients of
information on the characteristics of the slurries, particu- Eqs. 52, 26. and 30 obtained by the least-squares regression
lates, carrier fluids, filters and filter cakes, the actual condi- method and the corresponding coefficients of regression are
tions of the tests conducted, and the measurements of all the given, respectively, by
system parameters and variables. The reported studies of the
slurry filtration have measured only a few parameters and

0 002

02 I I I I /

00015
E
x
;;
:z-
0
om1
%%
c v
P
00005
>

1 . 1 I I , 0
0 0 02 0 04 0 06 0 08 01
04 I I I
0 015 0 02 0 025 0 03 0 035 0 04 Filtrate Flow Rate, q, mllmin
Filtrate Flow Rate. q, mUmin
Figure 3. Correlation of Jiao and Sharma (1994) pre-
Figure 1. Correlation of Jiao and Sharma (1994) experi- dicted filter cake thickness data according to
mental data according to Eq. 52. Eq. 30.

2384 November 1998 Vol. 44, No. 11 AIChE Journal

 A/C = 8.297 min/cm6, B/C = 0.1136 cmp3, R2 = 0.8713
(60)

C = 0.0034 cm4/min, D = 0.0076 cm, R2 = 0.949 (61)

A = 0.0229 cm-’. B = 0.0003 cm/min, R 2 = 0.9873 (62)

The coefficients o f regressions very close to 1.0 indicate that

the present equations closely represent the data. The coeffi-
cient of regression R 2 = 0.8713 indicated by Figure 1 and Eq.
60 is lower than those indicated by Figures 2 and 3, and Eqs.
61 and 62, inferring the possibility of larger measurement er-
rors involved in the filtrate volume data. Another source of 0 100 200 300 400 500 600
errors may bc duc to the three-point finite difference numer- Filtration Time, 1, rnin

ical differentiation of the filtrate volume data to obtain the Figure 4. Predicted vs. measured filtrate volumes for
filtrate flow rate data used to construct Figure 1. The data linear filtration of fresh water bentonite sus-
necessary for Figure 1 were obtained by a series of numerical pension.
procedures, first to calculate q = dQ/dt from the filtrate vol-
ume Q data, and then (l/q) and [d/dt(l/q)].
The initial filtrate volume rate is obtained as q, = 0.096 data over the period of the filtration process. Then. the +c
mL/min by a three-point forward differentiation of the mea- and T ~ values
, obtained this way were used in Eqs. 46 and 51
sured, initial filtrate volume data. This data is expected to to calculate the k , and k , values.
involve a larger error because of the possibility of relatively Using the slurry tangential velocity of v = 8.61 cm/s, the
larger errors involved in the early filtrate volume data. The typical particle diameter of d = 2.5 X lo-‘ cm and the parti-
noisy data had to be smoothed prior to numerical differentia- ’
cle separation distance of 1 = 2. X 10 cm in Eq. 5, the criti-
tion, which may have introduced further errors. Because of cal shear stress for particle detachment is estimated to be
the propagation of the significantly larger measurement er- T,, = 1.25 x lo3 dyne/cm2, whereas, the prevailing shear stress

rors involved in the early filtrate volume data, the first two of calculated by Eq. 33 is T = 16 dyne/cm2. Under these condi-
the [ d / d t ( l / y ) ] values degenerated and deviated significantly tions, theoretically, the cake erosion should not occur, be-
from the expected straightline trend. Therefore, these two cause T << T ~ Therefore,
~ . the value of the coefficient B
data points formed the outliers for linear regression and had should be zero. In contrast, as indicated by Eq. 62, the pre-
to be discarded. sent analysis of the data has led to a small, but nonzero value
Substituting the values given in Eq. 60 into Eq. 55 yields of B = 3 . x 1 0 p 4 cm/min. Recall that we used this value in
the limiting filtrate flow rate as qx = 0.014 mL/min. On the Eq. 48 to calculate the limiting flow rate of q, = 0.013
other hand, substituting the values given in Eq. 62 into Eq. mL/min. This value was shown to be very close to the qa=
48 yields 9%= 0.013 mL/min. These two values obtained from 0.014 mL/min value calculated by Eq. 55 and the approxi-
the filtrate flow rate and cake thickness data, respectively, mate value of q- = 0.017 mL/min obtained by extrapolating
are very close to each other. The limiting filtrate volume rate the filtrate flow rate data beyond the range of the experimen-
q, estimated by an extrapolation of the derivatives of the fil- tal data. Thus, it is reasonable to assume that B = 3.X l o p 4
trate volume data beyond the range of the experimental data cm/min is a meaningful value and not just a numerical result
is qx = 0.017 mL/min and close to the values obtained by the of the least-squares regression of Eq. 30 to data, because the
present regression method. This is an indication of the valid- coefficient of regression R2 = 0.9873 is very close to one.
ity of the present filtration model. Hence, it can be inferred that T > T , and the cake erosion
Using q x = 0.014 mL/min in Eq. 26 yields the limiting fil- occurred in the actual experimental conditions of Jiao and
ter cake thickness as 6, = 0.24 cm. The predicted cake thick- Sharma (1994). In view of this discussion, it becomes appar-
ness data presented in Figure 10 of Jiao and Sharma (1994) ent that the theoretical value obtained by Eq. 5 is not realis-
indicates a value of approximately 0.17 cm. Therefore, their tic.
prediction of ths limiting filter cake thickness appears to be The Jiao and Sharma (1994) data and the missing parame-
an underestimation compared to the 0.24 cm value obtained ter values which have been approximated by fitting the exper-
by the present analysis. imental data are given in Table 1. The results presented in
The above obtained values can now be used to determine Figure 4 indicate that the model represents the measured fil-
the values o f the model parameters as follows. The filter cake trate volumes over the complete range of 600 min of filtra-
permeability can be calculated by Eq. 50. Equations 46, 47, tion time as closely a5 the quality of their experimental data
51, 53 and 54 form a set of alternative equations to deter- permits. However, they did not measure the cake thickness,
mine the deposition and erosion rate constants k , and k,. but predicted it using a simple algebraic model. As shown in
Here, Eqs. 46 and 51 were selected for this purpose. How- Figure 5, the cake thicknesses predicted by Jiao and Sharma
ever, Jiao and Sharma (1994) d o not offer any data on the (1994) and the present study are close to each other.
cake porosity +<, and the critical shear stress T,, necessary Willis et al. (1983) conducted linear filtration experiments
for detachment of the particles from the progressing cake using a suspension of l u c k in water. As shown in Table 1,
surface. Thercforc, the 4c and T,, parameters had to be esti- they reported only a few parameter values. They only provide
mated and used with Eqs. 46 and 51 to match the filtration some measured filtrate flow rate and cake thickness data in

AIChE Journal November 1998 Vol. 44, No. 11 2385

 2 . I I I I

06

E
1.5
3-
E
05
rd
ui u 04
0-
f 1
+ <5 0 3
E
m
x
-
E
-
m
02
0.5 li

01

0 s 1 I I I I 0
0 1W 200 300 400 500 600 0 10 20 30 40 50 60
FiltrationTime, 1. min FiltrationTime, t, min

Figure 5. Predicted cake thicknesses for linear filtration Figure 7. Predicted vs. measured filtrate volumes for
of fresh water bentonite suspension. radial filtration of a sea-water based partially
hydrolyzed polyacrylamide drilling mud.

their Table 2. However, the filtration time data is missing.

Therefore, a full-scale simulation of their filtration process as means of Eqs. 10 and 20 requires both the filtration volume
a function of time could not be carried out. Only the linear and filter cake thickness measurements. Therefore, the Fisk
plotting of the measured data according to Eq. 26 could be et al. (1991) radial filtration data has more missing parameter
accomplished. As indicated by Figure 6, the best linear fit of values which had to be approximated as given in Table 1.
Eq. 26 with the least-squares method has been obtained with Figure 7 shows that the model predicts the measured dy-
a coefficient of regression of R 2 = 0.9921 very close to 1.0. namic and static filtrate volumes with reasonable accuracy in
This reconfirms the validity of the present filtration model. view of the uncertainties involved in the estimated values of
Fisk et al. (1991) conducted radial filtration experiments the missing data. Fisk et al. (1991) did not report any results
using a seawater-based partially hydrolyzed polyacrylamide on the filter cake thickness and, therefore, a comparison of
mud. Their Figure 4 provides the measured dynamic and the cake thicknesses could not be made in the radial filtra-
static filtrate volumes vs. filtration time data. Judging by their tion case.
Figure 4, their static filtration data contains only three dis-
tinct measured values. This data is insufficient to extract a
meaningful information on the values of the A and C coeffi- Conclusion
cients by regression of Eq. 56, because the calculation of The improved models developed in this study offer practi-
In[ - q - 3 dq/dt] requires a two step, sequential numerical cal means of interpreting experimental data, estimating the
differentiation, first to obtain the filtrate flow rate q = dQ/dt model parameters, and simulating the linear and radial, in-
by differentiating the filtrate volume Q, and then differenti- compressive cake filtration processes at static and dynamic
ating q to obtain dq/dt. O n the other hand, their dynamic filtration conditions. The simplified forms of the new models
filtration data is limited to the filtrate volume. As explained conform with the well-recognized simplified models reported
in the previous section on the determination of parameters, in the literature. The new models are capable of capturing
the determination of all coefficients of A , B , C, and D by the responses of typical laboratory filtration tests while pro-
viding insight into the governing mechanisms.

0.2
Acknowledgment
The author gratefully acknowledges the support of the School of
0.15
Petroleum and Geological Engineering at the University of Okla-
homa.
d
ui
a
4
.-
0.1
Notation
e
Y a =cross-sectional area of the filter, cm’
0
0.05 A , B , C , D =parameters defined by Eqs. 11, 12, 18 and 19 for the
radial model and by Eqs.21, 32, 27 and 28 for the linear
model, respectively, cm--, cm/s, cm4/s, cm
h =filter width, cm
0 d =particle diameter, cm
0.5 1 1.5 2 2.5 L -filter length, cm
Reciprocal Filtrate Flow Rate, l l q , 104s/m3 ni =flow
-
index
r =radial distance, cm
Figure 6. Correlation of Willis et al. (1983) measured fil- p =viscosity of fluids, cp
ter cake thickness data according to Eq. 26. x =linear distance, cm

2386 November 1998 Vol. 44, No. 11 AIChE Journal

Subscripts Jiao, D., and M. M. Sharma, “Mechanism of Cake Buildup in Cross-
flow Filtration of Colloidal Suspensions,” J. Colloid and Inter. Sci.,
c =cake or slurry side cake surface 162, 454 (1994).
d =deposition Liu, X., and F. Civan, “Formation Damage and Filter Cake Buildup
e =erosion or effluent side filter surface in Laboratory Core Tests: Modeling and Model-Assisted Analysis,”
p =particle SPE Formation Evaluation J., 11, 26 (1996).
w =slurry side filter surface Metzner, A. B., and J. C. Reed, “Flow of Non-Newtonian
Fluids-Correlation of the Laminar, Transition, and Turbulent
Flow Regions,” AZChE J., 1, 434 (1955).
Literature Cited Peng, S. J., and J. M. Peden, “Prediction of Filtration Under Dy-
Civan, F., “A Multi-Phase Mud Filtrate Invasion and Well Bore Fil- namic Conditions,” SPE Int. Symp. on Formation Damage Control,
ter Cake Formation Model,” Proc. SPE Int. Petrol. Con$ and Exhib. Paper SPE 23824, Lafayette, LA 503 (Feb. 26-27, 1992).
of Mexico, Veracruz, Mexico, SPE 28709 399 (Oct. 10-13, 1994). Potanin, A. A,, and N. B. Uriev, “Micro-rheological Models of Ag-
Civan, F., “A Multi-Purpose Formation Damage Model,” Proc. SPE gregated Suspensions in Shear Flow,” J. Coll. Int. Sci., 142, 385
Formation Damage Control Symp.,Lafayette, LA, Paper SPE 31101, (1991).
p. 311 (Feb. 14-15, 1996). Ravi, K. M., R. M. Beirute, and R. L. Covington, “Erodability of
Collins, E. R.. Flow of Fluids Through Porous Materials, Penn Well Partially Dehydrated Gelled Drilling Fluid and Filter Cake,” Proc.
Publishing, Tulsa, OK (1961). SPE Technical Conf and Exhib., SPE 24571, Washington, DC, p.
de Nevers, N., “Product in the Way Processes,” Chem. Eng. Educa- 219 (Oct. 4-7, 1992).
tion, p. 146 (Summer, 1992). Tien, C., R. Bai, and B. V. Ramarao, “Analysis of Cake Growth in
Donaldson, E. C., and V. Chernoglazov, “Drilling Mud Fluid Inva- Cake Filtration: Effect of Fine Particle Retention,” AIChE J., 43,
sion Model,”J. Petrol. Sci. Eng., 1, 3 (1987). 33 (1997).
Fisk, J. V., S. S. Shaffer, and S. Helmy, “The Use of Filtration The- Willis, M. S., R. M. Collins, and W. G. Bridges, “Complete Analysis
ory in Developing a Mechanism for Filter-Cake Deposition by of Non-Parabolic Filtration Behavior,” Chem. Eng. Res. Des., 61, 96
Drilling Fluids in Laminar Flow,” SPE Drilling Eng., 6, 196 (1991). (1983).
Hermia, J., “Constant Pressure Blocking Filtration Laws-Applica-
tion to Power-Law Non-Newtonian Fluids,” Trans. IChemE, 60, 183
(1982). Manuscript receiwd Dec. 15, 1997, and revision receiwd Sept. 3, 1998.

AIChE Journal November 1998 Vol. 44, No. 11 2387