Sei sulla pagina 1di 9

Chemhd&&eering Science. Vol. 41, No. 2, pp. 273-281, 1986 uax-2so!wt6 53.00+0.

00
Printed in Great Britain. Pergamoa Pmss Ltd.

A FILTRATION MODEL FOR THE FLOW OF DILUTE,


STABLE EMULSIONS IN POROUS MEDIA-II. PARAMETER
EVALUATION AND ESTIMATION

H. SOO, M. C. WILLIAMS and C. J. RADKEt


Chemical Engineering Department, University of California, Berkeley, CA 94720, U.S.A.

(Received 25 June 1984)

Abstract-Part I of this work outlined a new theory, based on deep-bed tiltration concepts, to describe the
flow of dilute, stable emulsions in underground porous media. Here, in Part II, we quantitatively test the
proposed theory against experimental data and we indicate how the liltration model parameters can be
estimated from first principles.
Comparison is made between the theory and data on transient permeability and efauent concentration for
dilute, oil-in-water emulsions of mean drop-size diameters ranging from 1 to IO- and volume
concentrations of 0.5-2.5 /, flowing in quartx sandpacks of 0.57-2.0 & permeability. The pH of the
continuous aqueous phase is kept constant at 10. Filtration theory successfully represents the data, permitting
unambiguous evaluation of the theoretical parameters
Procedures are described for o priori calculation of the filtration parameters from knowledge of the drop
sixe and the pore-size and grain-sixe distributions of the porous medium. God agreement is achieved between
the experimentally determined parameters and their estimated values. Thus, the proposed filtration model
provides a reliable tool for predicting emulsion flow behaviour in porous media

lNTRODUCTION sion flow models (McAuliffe, 1973; Devereux, 1974a)


Available experimental evidence indicates that stable revealed the critical import of testing model predic-
emulsions percolate through underground porous tions against effluent drop concentrations in addition
media by capture of the disperse phase with sub- to validating flow resistance_
sequent permeability reduction to the continuous In this second part we bring theory and experiment
phase (McAuliffe, 1973; Devereux, 197423; Alvarado together. The filter coefficient and the flow-diversion
and Marsden, 1979; Soo and Radke, 1984). While parameter, which both control the processes of drop
flowing in the porous medium, droplets clog pores capture in the medium, are evaluated from emulsion-
having sixes smaller than their own by lodging between concentration histories. Conversely, the flow-restric-
sand grains (straining capture). Also, they are trapped tion parameter, which characterizes how effective
in recirculation eddies, wedge in the crevices, or retained drops are in obstructing flow, is obtained
sometimes attach to the pore walls due to van der from permeability data. Quantitative comparison is
Waals, electrical, gravitational and hydrodynamic made between the theory and the experimental results
forces (interception capture). As drop retention builds, for dilute, oil-in-water emulsions of mean drop di-
the permeability of the medium is reduced and 3ow ameters ranging from 1 to 10 pm and volume con-
diverts to the larger pores where capture probability is centrations of 0.5-2.5 okflowing in quartz sandpacks of
lowered because of the increasing pore sixes. 0.57-2-O pm* permeability. The proposed filtration
In Part I of this series we presented a filtration-based model successfully represents the experimental data,
flow model which correctly accounts for the actual both for permeability-reduction and dropcon-
physical phenomena occurring in the porous medium. centration histories.
In that theory, transient flow behaviour is character- Since the three filtration parameters have well-
ized by three parameters: a filter coefficient, a dow- defined meanings, we develop simpli6ed procedures
diversion parameter and a flow-restriction parameter. for their estimation from physical properties of the
The filter coefficient controls the sharpness of the system, such as the medium porosity, the pore-size and
emulsion front, the flow-diversion parameter dictates grain-size distributions, and the drop size. The
the steady-state retention as well as the flow- periodic-constricted-tube model is adopted to describe
redistribution phenomenon, and the flow-restriction the porous medium (Payatakes et al., 1973).
parameter describes the effectiveness of retained drops Accordingly, the filter coefRcient is estimated based on
in reducing permeability. Qualitative comparison of the likely capture probability of drops in a unit bed
the filtration theory to the continuum-viscous element (Payatakes et al., 1974). The flow-
(Alvarado and Marsden, 1979) and retardation emul- redistribution parameter is estimated from fractional
monolayer coverages of drops on grain surfaces and
%%zsent address: Union Carbide Corporation, South from the deduced volume fraction of blocked pores.
Charleston,WV 25303, U.S.A. Finally, the flow-restriction parameter is estimated
*To whom correspondence should be addressed. from hydrodynamic calculations of the increased
273
274 H. Soo er al.

pressure drop due to retained drops in a constricted evaluated from experimental effluent-concentration
tube. histories, cL = c(T, 1). In contrast, fi influences flow
In the next section we deal with the evaluation of resistance (see Fig. 5 of Part I), and is, therefore, best
the filtration parameters from experimental data. This evaluated from experimental pressure-drop data.
is followed by a discussion of parameter estimation Details of how the filtration parameters are actually
procedures. fit from the experimental results may bc found in
Appendix A.
PARAMETER EVALUATION

Experimental Comparison to experiment


Only a brief account of the experimental procedures Figure 1 compares typical experimental flow data to
is presented here. Considerable detail may be found the filtration theory. The transient reduced overall
elsewhere (Soo, 1983; Soo and Radke, 1984). Dilute, permeability K/K0 and the reduced effluent drop-
oil-in-water emulsions of mean drop sizes (D, > volume concentration cL/ci are plotted against the
ranging from 1 to IO pm and of known drop-size cumulative injected volume of oil, TC~_Solid symbols
distributions were injected into quartz sandpacks of correspond to transient permeability reductions and
known grain-size distribution, pore-size distribution open symbols correspond to effluent emulsion concen-
and initial permeability KO. Drop-volume concentra- trations. Dashed lines show the best fit of the theory to
tions ci of 0.5, 0.6, 1.2 and 2.5 y0 were used, and the the effluent concentrations according to eqs (1) and (2),
superficial velocity u was fixed at 0.07 mm/s. The as described in Appendix A. Likewise, the solid lines
emulsion droplets were adjusted to neutral buoyancy are best-fit transient permeabilities according to eq. (3).
with carbon tetrachloride. A dilute aqueous sodium Figure 1 demonstrates that the proposed filtration
oleate-oleic acid mixture at a pH of 10 stabilized the theory successfully portrays both the transient per-
emulsion. Under this alkaline condition there is signifi- meabilities and the drop breakthrough concentrations,
cant double-layer repulsion between the oil drops and even with the assumption that model parameters a and
between the drops and the quartz sand grains. B are retention-independent. Additional confirmatory
Sensitive transducers (Validyne) monitored the pres- flow data are reported in Figs 6 through 12 of Soo and
sure drop. To permit flow-model discrimination, the Radke (1984), where in those figures the solid and
core effluent was sampled periodically and analysed dashed lines are best fit according to the filtration
(Zeiss Particle Counter) for drop concentration and theory. Table 1 summarizes the filtration model par-
size distribution. ameters obtained in this work.
In Fig. 1, K. = 1.15 pm2 and the inlet mean drop
Filtration theory sizes vary from 2.1 to 6.1 pm. As the drop sizes increase,
According to the filtration flow theory proposed in the effluent concentration curves reveal that drop
Part I, the transient drop-volume retention and drop- retention increases so that 1/a rises (i.e. the steady-state
volume concentration profiles c (t, JC)and c (t, x) obey retention is an increasing function of drop size).
the following expressions in a linear medium of length Permeability at any 7ci decreases as the mean drop size
L: increases from 2.1 to 4.5 pm. Yet, initially, the rate of
1 -exp (aA,,c,T) permeability reduction is smaller when the drops
c(T,Z)
(1) increase in size from 4.5 to 6.1 pm. This is @cause fl is a
____%I = a [l -exp (A,%) -exp (ahs,ciT)]
decreasing function of drop size, due to its unit volume
definition, as verified in Table 1. At equal volumes of
c(T, 3)
-=- Q cc 2)
ci uiG7 <OdhF" 1
where f = x/L is a reduced axial distance, and T =
z --f is a shifted time variable with 7 = ut/EOL. The
subscript i refers to the bed inlet and E,,is the clean-bed
porosity. As, = &,L is the reduced filter coefficient,
made dimensionless by the column length L, while a is Ci=OOOS _
the flow-redistribution parameter.
Emulsion flow behaviour is reflected in the transient
permeability reduction, K/K,,. According to Part I,
this ratio depends on the local retention as

1 TCi* PORE VOLUME OIL

-= [ 1 - #?a/~~] - 1 dn, (3)


Fig. 1. Experimental permeability-reduction (solid symbols)
and breakthrough-concentration (open symbols) histories for
where fi is the flow-restriction parameter. varying drop sizes (2.1, 3.1, 4.5 and 6.1 Nrn) in the 1.15 pm2
permeability core. The oil viscosity is 1.5 mPa s. Solid (per-
From eqs (1) and (2), the filtration parameters A,, meability reduction) and dashed lines (breakthrough-
and a are seen to control droplet chromatography (see concentration) are best _.fit according to _.
the filtration theory
. . -_. _
also Figs 3 and 4 of Part I). Therefore, they are best with the correspondmg parameters hsted m .laOle 1.
Filtration model for the flow of dilute, stable emulsionsin porous media-11 275

Table 1. Filtration parameters

2.5 0.57 0.038 15.0 4.7 0.570


3.3 0.57 0.045 14.5 4.4 0.563
3.6 0.57 0.047 14.5 1.6 0.682
5.3 0.57 0.250 3.6 - 0.900
2.1 1.15 0.019 13.9 0.69 0.264
3.1 1.15 0.032 14.5 1.7 0.464
3.7 1.15 0.027 12.6 1.8 0.340
3.8 1.15 0.032 7.9 2.4 0.253
4.3 1.15 0.051 14.0 3.5 0.714
4.4 1.15 11.7 1.5 0.702
4.5 1.15 :z 10.0 1.9 0.530
4.5 1.15 0.074 10.2 1.9 0.755
5.8 1.15 0.095 5.9 1.7 0.561
6.1 1.15 0.170 4.8 1.4 0.816
6.4 1.15 0.200 3.6 2.8 0.720

retained oil, smaller drops are more finely dispersed


locally in the pore assemblages, and are therefore more
effective in reducing permeability. Hence, before drop-
let breakthrough where retention volumes are equal,
the 4.5 m drop emulsion has a lower transient c;
permeability than the 6.1 pm drop emulsion. A 0.006
0 0.012
Eventually, at steady state, large drop-size emulsions 0 0.025
reduce permeability more than small drop-size emul-
sions. That is, in a given porous medium, p/a is an
increasing function of drop size (i.e. at steady state
K-/K,, = 1 -/?/a). Table 1 also confirms this observa- ol; 0 I I I
0
tion, with some scatter. 0.05
c, (T-1)
0.10 0.15

One unique characteristic of the filtration model is


that both effluent concentrations and transient per- Fig. 2. Experimental breakthrough-concentration histories
meabilities should be independent of the disperse oil- for varying injection concentrations (0.6, 1.2 and 2.5 vol %) in
phase viscosity. Figure 11 of Soo and Radke (1984) the 2 pm permeability core. The oil viscosity is 1.5 mPa s
while the dashed line is best fit from the 0.6 and 1.2 vol /,
shows that upon changing the oil viscosity from 1.5 to concentration data according to the filtration theory
23 mPa s, while keeping all other variables constant, (l/a = 0.015, fi = 15 and AsI = 1.6cm-I).
no change in transient permeability or emulsion-
concentration history is observed. This provides con-
vincing evidence that the filtration model correctly I
I I I
1
reflects the phenomena controlling dilute, stable emul-
sion flow in unconsolidated porous media.
A second distinguishing feature of the filtration
model in eqs (l)-(3) is that graphs of cL/ci and K/K, vs.
the variable ci (7 - 1) should be independent of inlet
concentration. Figures 2 and 3 give results for injecting
a 4 m mean drop-size emulsion of ci = 0.006, 0.012
and 0.025 volume concentrations into a 2 pm2 per-
meability sandpack. Both cL/ci (Fig. 2) and K/K,
(Fig. 3) obey the filtration theory for the two most
dilute suspensions. Deviations toward higher retention ci (T- I)
and larger permeability reduction are found with the ci
= 0.025 emulsion. Such deviations may be attributed Fig. 3. Experimental permeability-reduction histories for
to larger hydrodynamic resistances (e.g. ~JF, > 1) or varying injection concentrations (0.6,1.2 and 2.5 vol %) in the
2 pm* permeability core. The oil viscosity is 1.5 mPa s while
to a nonlinear dependence of drop capture kinetics on the solid line is best fit from the 0.6 and 1.2 vol y0
concentration in eq. (4) of Part I. concentration data according to the filtration theory (l/a
The delayed effluentconcentration history of the = 0.015, /I = 15 and AsI = 1.6cm-).
0.025concentration emulsion in Fig. 2 suggests non-
linear kinetics as the cause of the discrepancy. There solid particulates (Herzig et al., 1970). The present
are citations in many deep-bed filtration studies that ci study indicates linearity prevails for liquid droplets up
= 0.001 is the upper limit of linear capture kinetics for until about ci = 0.01. Drop capture kinetics are in-
276 H. so0 et al.
fluenced by interactions among the drops. In par- As indicated later, stabilized liquid drops are not
ticular, high drop concentrations can lead to several captured on top of previously retained droplets.
drops being captured simultaneously in a pore, which Hence, the constricted-tube transition diameter should
at low drop concentration does not plug. The dashed lie, from simple geometric arguments, between 2 and 3
line in Fig. 2 and the solid line in Fig. 3 represent the drop diameters. From our experimental results for a
best-fit calculations of the two lower concentration and /I, we will show D, to be close to 2 <D,>.
experiments according to eqs (lF(3). Thus, the fil- The straining porosity, ss, may now be calculated,
tration model successfully interprets the experimental since the pore-entry diameter distributions of the
results up to about ct z 0.01. sandpacks are known from centrifugal water-drainage
We conclude that the transient filtration model experiments (Soo, 1983). These experiments show that
accounts for all the experimental findings for emulsion the volume frequency of pore throats (i.e. E,,/E~) is
flow cited in Part I. Furthermore, it quantitatively fits distributed according to log-normal statistics. Hence,
pressure drop-flow rate data and e5uent drop concen- es follows directly:
trations with a single set of parameters.
exp ( - 5/2) dth (9)
PARAMETER ESTIMATION
Prediction offiltration parameters where In ii and 6 are, respectively, the mean and the
The phenomenological filtration parameters A,,, a, standard deviation of the Gaussian distribution in
and /l all have well-defined physical meanings. From In D, which serves as the generating function for the
the theoretical analysis of Part I the expressions for log-normal distribution (Thomasian, 1967). Once ii
these parameters are [see eqs (lo), (11) and (13) of and 6 .are obtained from experiments and D, is
Part I]:
specified, eq. (9) dictates the value of ES.Further, eq. (9)
states that es. and consequently A,,, are functions of
[In (0,/p)]/& the dimensionless transition throat dia-
meter based on a log-normal pore-sire distribution. It
remains to determine < 1, > and < 1, >.
As discussed previously in Part I, drop capture
efficiencies, q,,, in the straining pores are close to unity.

and
XdD,
J, (5)
Therefore, < 1, > may be approximated by the inverse
of the thickness of the average unit bed element,

1E agsrl, dD,. According to Payatakes et al. (1973), 1


50
is related to the physical properties of the unconsoli-
The symbols appearing in eqs (4)-(6) are defined in
dated porous medium by
Part I. In this study the surface properties of the drops
and the sand grains are Iixed. Therefore, eqs (4)-(6) 1
-yl= &,) <D:>]l (12
permit calculation of I,,, a and /l given fundamental 2
<s>
information about the drop size and pore-size distri-
bution along with some estimates of the porous- where D, is the grain diameter and ( Di>
medium structure (& D,, up,_,), of the interception = wfgD;dD,, where fg is the volume grain-size
capture probability (A,,), and of the retention hydro- s
dynamic resistance (/$,). In the following sections we distkbution. We find ( D3 > and hence < As > from sieve
develop procedures for predicting the filtration para- analysis of the sand grails (Soo, 1983).
meters. The experimentally determined values listed in Next, <A, > in the interception pores, where only
Table 1 are then compared to the a priori predicted surface capture occurs, can be estimated from trajec-
ones. tory calculations of a particle moving through a
constricted tube, or around a cylindrical or spherical
Filter coeficient collector (Tien and Payatakes, 1979). The calculation
Overall drop capture efficiency, or the filter coef- for a constricted tube, including all possible physical
ficient, may be estimated from eq. (4): forces, follows from the work of Payatakes et al.
(1974). For a mean drop-size to pore-size ratio between
ils, = < 11,> (as/so) + <A, > (1 -es/so), (7) 0.01 and 0.1 (i.e. in the interception pores), the capture
where < L, > and <AI > are the porosity-averaged filter efficiency with no net colloidal interaction forces
coefficients in the pores where straining occurs (D, ranges from 1O- 3 to 1O- [see Fig. 2 of Payatakes et al.
c DJ and where interception occurs (D, > D,), (1974)]. Thus, within the unit-bed-element model we
respectively. First, we define the porosity of the take (I, > to be equal or less than about 0.01 /L This
straining pores, es, in terms of the transition pore- completes the estimation of parameters needed to
throat diameter between straining and interception calculate A,, from eq. (7).
pores, D,, so that Figure 4 demonstrates the relation between
et E,,dD,.
(8) (&J)- and the dimensionless transition diameter
Es = [In (D,/p)]/S for two values of <I, >. Here D, is taken
I0
Filtrationmodel for the flow of dilute, stable emulsionsin porous media-11 277

distribution, the small drops percolate through the


porous medium more quickly than the large ones. The
spread of the experimentally measured breakthrough
curve, which determines IsI, therefore is wider than
that of a monodisperse system with the same mean
drop size. A wider effluent drop-concentration spread
is interpreted by the theory as a smaller or under-
estimated A,,. This size-separation phenomenon is
- KoomJ=) o most pronounced for the larger drops in Fig. 4, since
,*I- 0 1.15 the emulsions with a large average drop size possess a
l 0.57
wider distribution (Soo and Radke, 1984). Fortunately,
when 1,, is large, it has little influence on the emulsion
Dt - 2<Dcl> flow behaviour (see Fig. 3 of Part I).
lo I

-3.0 -2.0
I t
-10
[In (Dt/&i )I /8 Flow-diversion parameter
An estimate of CL,the flow-diversion parameter, is
Fig. 4. The inverse of filter coefficient, expressed as &I,
available from eq. (5). At steady state, the retention in
correlatedagainst the dimensionless logarithmic transition
pore-throat diameter [In (D,/ji)]/6. D, is taken as 2 <Dd > in
each classification of pore space is at a fixed value so
this figure. that cp - ss for D, -z D, and up - ur,,, for D, > D,.
Under this circumstance, eqs (4) and (5) reduce to:
as 2 <Dd>. The physical meaning of (&I)- is the a m= s0/gm = so/(es+ <o,,,>) 9 (11)
number of unit bed elements (i.e. the number of grain
where < ur,m> is the average maximum retention in the
diameters) a drop travels before it is captured. Open
interception pore space and ss is available from eq. (8).
circles in the figure represent experimental results of a
In this study drops are not captured on previously
1.15 m2 permeability sandpack and solid circles rep-
retained drops. Hence, the maximum interception
resent those of a 0.57 pm2 permeability sandpack.
retention is that of complete monolayer cover-
Selected error bars denote the expected variation of the
experimental values. The log-normal distribution par- age or oI.m = [2x(1 -so) (so --ES) <D,>ll(fi <D,>)
ameters, reported in terms of R and 6, are obtained (Payatakes et al., 1973).
from centrifuge water-drainage measurements (Soo, In reality, however, complete surface monolayer
1983). For the 1.15 pm2 sandpack, R is measured as coverage is highly unlikely, especially for large drops
25.5 pm and 6 is 0.54 (i.e. <D,> = 29.5 pm+). For the and for repulsive double-layer interactions. Therefore,
0.57 m2 sandpack, ji is 15.3 pm and 6 is 0.49 (i.e. a fractional coverage, 8, is introduced:
<Dp > = 17.3 m). The unit-bed-element length, 1, is 2n-50(1 -so) (1 - E~/EO)<D, > e

calculated from eq. (10). For the 1.15 pm2 sandpack it <QI,m> = fQ,,@ =
is 120 pm (i.e. a mean grain diameter of 107 pm), and ,b<D,> .
for the 0.57 pm2 sandpack it is 83 pm (i.e. a mean grain (12)
diameter of 75 pm). Equations (9), (11) and (12) permit calculation of am,
At small values of the logarithmic-reduced trans- which we take as a good approximation to a based on
ition diameter, or equivalently for small mean drop the results of Fig. 2 of Part I.
sixes, the drops travel many tens of unit bed elements Figure 5 displays the functionality of l/a with the
before they are captured. As the mean drop size dimensionless logarithmic-transition pore-throat di-
increases, the drops travel a fewer number of unit bed ameter [ln (D&i)]/& Circles give the experimental
elements before they stick. data for the two different permeability cores. This
Solid lines in Fig. 4 represent a priori predictions figure demonstrates that as drop size increases, l/a
based on eqs (7~(10) for <A,> equal to lo-/1 and increases. As drop size increases,. more drops are
lo- j/l. At large values of the logarithmic-reduced captured in the porous medium. Consequently, flow
transition diameter, the prediction shows that strain- redistribution is more manifest, and steady-state reten-
ing capture dominates. Most drops are retained tion is higher.
within one unit-bed-element distance, and therefore Solid lines in the plot represent the a priori calcu-
(&tW approaches unity. At small values of lations of a for various values of interception coverage
[ln (D,/R)]/S, (1,,Z)- approaches (<A,> I)- . Signili- 8 and for D, fixed at 2 <D, >. It is not possible to obtain
cant deviations are seen for the larger drops a unique estimate of 8 and D, using only data in Fig. 5.
independent of < 1, >. This may be explained qualitat- Raising D, simply lowers 0 because a larger D, de-
ively by the drop-size separation effect reported by Soo signates more retention in the straining mode and
and Radke (1984). In emulsions with a drop-size therefore less retention in the interception mode. Thus,
the dashed line in Fig. 5 for D, = 2.2 < Dd > and 0 = 0.1
tFrom the definitionof the log-normal distributionfunc- lies above the solid line for D, = 2.0 < D, > and 0 = 0.1.
tion, the mean-volume Pore-throat diameter is given by: To obtain 8 and D,, we utilize both the data for a in
(D, > = ji [exp (S2/2)] (Thomasian, 1967; Soo, i983). Fig. 5 and the data for fi in Fig. 6 (to be presented).
278 H. Soo et nl.

tion occurs near pore constrictions and then permits


straining in pores whose throat diameters are less than
twice the drop size (i.e. D, - 2 <D, >). We denote this
mode of capture as multi-drop straining.
There is a distinction between multi-drop straining
and the bridging of solid particulates during deep-bed
filtration (Sakthivadivel, 1966; Payatakes and Tien,
1976; Payatakes, 1977; Payatakes and Gradon, 1980).
When a solid suspension flows through a porous
medium, particles are retained not only by the porous
material but also by the previously retained particles.
Since a retained particle protrudes from the pore wall,
it has a greater chance to capture oncoming particles.
Therefore as the process proceeds, particles are cap-
-3.0 -2.0 -1.0 tured largely by the previously retained particles and
ClnK+/p)l/s
form dendrites. These dendrites grow with little inter-
ruption until they connect with others. A dendrite may
Fig. 5. The inverse of the flow-diversion parameter, 01,
correlated against the dimensionless logarithmic transition
build sufficiently to encounter another dendrite on the
pore-throat diameter [ln(D,/@)]/& B is the fractional grain opposing side of a pore. Once such a bridge forms, it
surface coverage of drops captured in the interception mode. may eventually block the pore completely. Due to this
Two estimates of Dt are reported (solid and dashed lines). bridging phenomenon, solid particles clog pores of
sizes much larger than the size of the particles
(Sakthivadivel, 1966; Payatakes and Tien, 1976;
20 I I I I Payatakes, 1977; Payatakes and Gradon, 1980).
2 0, e = 04 1 For stable emulsion flow in porous media, however,
in which significant repulsion exists among the drops
and grain surfaces, interception capture apparently
occurs only on the porous medium surfaces. Dendrites
do not form and therefore bridging is not possible.
Thus, in emulsion flow, a steady state exists with all
flow in the larger non-straining pores.
For those solid suspensions which can form
dendrites, a steady state is unlikely except at high
flow rates in high-permeability media where bridging
is prevented by hydrodynamic forces. Soo and Radke
(1986) studied the flow of 2.2pm monodisperse poly-
styrene latexes [Dow (No. lAl2)j at ci = 0.005 in the
1.15 pm2 sandpack. The surface charge density of the
stabilized latexes was equivalent to that of the emul-
sion liquid droplets. Few solid particles eluted and no
9%2=
0 I L I steady state was achieved. Microscopic examination of
-3.0 -1.0 the clogged sandpack revealed the presence of de-
b n Cf$%],s
ndrites (Soo, 1983; Soo and Radke, 1986).
Fig. 6. The flow-restriction parameter j3 correlated against In spite of no liquid-droplet bridging, straining of
the dimensionless logarithmic transition pore-throat dia- emulsions is not restricted to pores of throat diameters
meter, [ln (Dt /F)] /6. Solid and long dashed lines correspond
smaller. than the droplet diameters. When drops are
to <fit > = 18 while the short dashed line corresponds to
<#?t> = 20. e is the fractional coverage of grain surfaces intercepted near a throat, they reduce the open pore
occupied by drops retained due to interception. diameter. If the resulting throat diameter is smaller
than the diameter of an approaching droplet, the
oncoming drop is then strained out and completely
Matching both sets of data permits rough estimates blocks the pore. By this multi-drop straining mechan-
of D, and 8. A transition pore diameter of around ism, pores with throat diameters as large as three drop
B, = 2 (D, > is found and a surface coverage at steady diameters could possibly plug. However, Fig. 5 reveals
state of about 8 = 0.1 emerges. that the actual blocking pore-throat size is close to two
The dotted-dashed line labelled by straining only drop diameters.
in Fig. 5 reports a calculation assuming the transition The amount of interception capture discovered
diameter is equal to one drop diameter, and assuming from Fig. 5 is not large (i.e. 6 equals about 10% at
no interception capture. Equivalently, drops can be steady state). This low amount of surface coverage is
strained only in pores whose throat diameters are less reasonable in view of the strong repulsive forces
than their own. This hypothesis is clearly in error and between the drops and the sand surfaces. A large
indicates that interception occurs. Further, intercep- portion of the interception capture is probably in dead
Filtrationmodel for the flow of dilute, stable emulsionsin porous media-11 279

pore-space caverns or in recirculation eddies. reveals the increasing importance of straining capture.
Interception capture is important and cannot be Solid lines in Fig. 6 correspond to eq. (13) at several
discounted, even in situations where straining capture values of the fractional surface coverage 0 for a fixed
is significant. D, = 2<D,> and with <fir> = 18. As in Fig. 5 the
A point worth noting is that for some of the large choices of D, and 8 are not unique. Here, however,
drop-size emulsions, in which [In ( DC/p)]/6 is greater increasing D, shifts 8 to lower values (compare the long
than - 1.5, drop breakthrough is not observed even dashed line at 0 = 0.1) because more retention in the
after 100 pore volumes of emulsion injection (i.e. straining mode causes less flow restriction (i.e. < & >
steady state is not achieved). This indicates that the c < fl, >). The consequence of this opposing depen-
maximum size of pores in the contiguous void space dence of Q and fi on B is that it is possible to estimate
through which steady state must be achieved is ap- both D, and 0 from Figs S and 6.
proximately equal to the transition throat diameter for Clearly, the data for the 0.57 pm2 sandpack do not
which [ln (D&i)]/& is - 1.5 or D, = ji exp( - l.SS). align with those of the 1.15 pm2 sandpack. It is
possible, by adjusting < /?, >, to effect a better fit of the
Flow-restriction parameter OS7 pm data. This is shown by the short dashed line
Equation (6) allows determination
of the flow- in Fig. 6 which corresponds to <fir > = 20. However,
restriction parameter /3 again at steady state: altering < & > does not change our conclusion that D,
and 0 are near 2 < D, > and 0.1, respectively.
PO3= (<Bs>++ <BI> <a*,,>)/(%+ <~LIll>). From the above discussion we conclude that the
(13)
filtration model not only explains quantitatively the
where < & > and < & > are the average flow-restriction experimental results, but also that the phenomeno-
parameters in the straining and interception Pore logical parameters A,, , a and /3 can be interpreted
spaces, respectively. As mentioned earlier in Part I, physically. Several of the prediction parameters, in-
<&.> in the straining pore space is around unity. cluding 4, <A, >. <Q~,~ > and <fl, >, cannot be as-
To estimate the flow-restriction parameter in the signed uniquely, although they cannot be changed by
interception region of the porous medium, <& >, we large amounts from plausibly predicted values.
must calculate the pressure-drop increase across a The results of Figs 4-6 have some very important
constricted tube due to the retained drops. The ramifications. First, surface capture exists and can
pressure-drop increase depends strongly on the pos- cause straining in pores of throat diameters larger than
ition of the drop in the constricted tube. Payatakes et the drop diameter. The largest straining pore diameter
al. (1973) calculate fl, for two types of symmetric is approximately two drop diameters. Second, the
retention patterns. For a smooth monolayer coating grain surfaces are not completely covered with drops
upstream of the pore throat, #I, is about 3, and for a (i.e. 0 = 0.1). This is probably due to the strong
smooth monolayer coating throughout the pore it is repulsive drop-sand colloidal interactions. This obser-
10. Unfortunately, with the drop sizes dealt with here, vation is sensitive to the surface chemistry of the
smooth symmetric coatings are not anticipated. An system and must not be taken as general. Finally, these
alternate approach is to consider drops attached to the figures explain the increase in retention, or in l/a, with
surface of the constricted tube. The fluid mechanics for increasing drop size, and the concomitant decrease in p
this case are not available. However, Happel and with increasing drop size, as found experimentally.
Brenner (1965) present results for a spherical particle
in a cylindrical tube. Their results may be transformed, CONCLUSIONS

as outlined in Appendix B, into a value for fir of 13. We The theoretical filtration-flow model is quantitat-
expect & to be larger than this value in a constricted ively compared to experimental effluent-drop concen-
tube, especially if the drop lodges near the pore throat. tration data and transient permeability data for dilute,
Hence, we arbitrarily take </3, > as between 1S and 20. stable oil-in-water emulsions of mean drop sizes
Equations (9), (12) and (13) then determine fl, again ranging from 1 to 10 pm and with inlet volume
with the assumption that fi = pou. concentrations of 0.5, 0.6, 1.2 and 2.5 0/0 flowing in
Figure 6 gives the flow-restriction parameter /? as a unconsolidated quartz sandpacks of about 1 pm*
function of the dimensionless logarithmic-transition permeability at constant ionic strength and pH of the
diameter. The experimental results demonstrate that fl continuous aqueous phase. For emulsion concentra-
decreases with increasing drop size, or that small-size tions up to 1.O y0 and for oil-drop viscosities between
drops are more effective in reducing the permeability 1.5 and 23 mPa s, the filtration model successfully
than large-size drops, at the same volume retention. represents the experimental data, permitting direct
Similar to the l/a: plot, this figure indicates that evaluation of the phenomenological flow and retention
interception capture does exist. If only straining cap- parameters: rZ,, a and fi.
ture were to occur, /3 = < & > shouId be very close to Theoretical expressions for the filtration parameters
unity (Herzig et al., 1970). However, for small drop-size are developed which allow their a priori prediction
emulsions, /? is close to 15 showing that interception given information about the drop size, the pore-size
capture is dominant. As the drop size of the system and grain-size distributions and given some knowledge
increases, especially in the case that [ln (D&)1/S is of porous medium structure, of interception capture
larger than -2, the value of /3 drops to about 3. This probabilities and of hydrodynamic resistances of cap-

CES41:2--E
280 H. so0 er al.

tured particles. The estimation procedures adequately u local oil retention, oil-drop volume/bed volume
reflect the measured experimental parameters in both or local retention of drops in pores of throat
the interception and straining dominated capture diameter D,, m - l
regimes, and they justify why &t increases and why cx r emulsion pore volumes injected, ut/~eL
and fi both decrease with increasing drop size, as
observed experimentally. Additionally, _ the estimation Subscripts
exercise reveals that for repulsive drop-sand grain d droplet ~
interactions little surface capture of drops occurs. Even e emulsion
small interception surface coverages of around 10 %, grain
however, lead to straining in pores of throat diameters : interception
larger than the drop diameters. The largest straining i inlet
pore diameter is found to be around two drop L exit
diameters. m maximum
The filtration theory provides a new tool for under- 0 initial
standing the flow behaviour of dilute, stable emulsions pore
in underground porous media. : straining
SI straining and interception
Acknowledgements-This research was supported by the U.S. t transition
Department of Energy under Grant W-7405-ENG-48 to the W water phase
Lawrence Berkeley Laboratory. H.S. acknowledges financial
co steady state
assistance from the Chevron Oil Field Research Company.
The authors would like to thank Mr. R. F. Hettler of-Dow <> denotes volume average
Chemical Company for supplying the polystyrene latexes.
REFERENCES
Alvarado, D. A. and Marsden, S. S., 1979, Flow of oil-in-water
NOTATION emulsions through tubea and porous media. Sot. Petrol.
twice the particle centre distance to a wall, m Engng J. 19, 369-377.
Devereux, 0. F., 1974a, Emulsion flow in porous solids--I. A
volume concentration of oil drops in emulsion,
flow model. Chem. Engng 1. 7, 121-128.
volume of drops/flowing volume Devereux, 0. F., l974b, Emulsion flow in porous solids-II.
drop diameter, m Experiments with crude oil-in-water emulsion in porous
grain diameter, m sandstone. Chem. Engng J. 7, 129-136.
pore-throat diameter, m Happel, I. and Brenner, H., 1965, Zmv Reynolds Number
Hydrodynamics with Special Application to Particulate
transition pore-throat diameter, m Media. Prentice-Hall, Englewood Cliffs, NJ.
volume fraction of sand grains with diameter Herxig, J. P., Leclerc, D. M. and LeGoff, P., 1970, Flow of
D,, m-i suspensions through porous media-applications to deep
overall permeability, m2 bed filtration. Ind. them. Engng 62, 8-35.
McAuliffe, C. D., 1973, Oil-in-water emulsions and their flow
core length, m
properties in porous media. J. Petrol. Technol. 25.727-733.
unit-bed-element thickness, m Payatakes, A. C., 1977, Model of transient aerosol particle
length of a pore, m deposition in fibrous medii with dendrltic pattern.
pressure, N/m2 A.1.Ch.E. J. 23, 192-202.
time, s Payatakes. A. C. and Gradon, L., 1980, Dendritic deposition
of aerosol particles in fibrous media by inertial impaction
shifted time variable, z - 5 and interception. Chem. Engng Sci. 35, 1083-1096.
velocity, m/s Payatakes. A. C. and Tien, C.. 1976, Particle deposition in
distance, m fibrous media with dendrite-like pattern: a preliminary
reduced axial distance, x/L model. J. Aerosol Sci. 7, 85-100.
Payatakes, A. C., Tien, C. and Tutian, R. M., 1973, A new
model for granular porous medii. A.I.Ch.E. J. 19, 58-66.
Greek letters Pavatakes. A. C.. Tien. C. and Turian. R. M., 1974, Traiectory
flow-diversion parameter &lculation of particle depositions in deep bed fiRration.
how-restriction parameter A.1.Ch.E. J. 20, -905.
flow-restriction parameter of retained drops in Sakthivadivel. R.. 1966. Theorv and mechanism of filtration
of non-colloidal fines through a porous medium. Tech.
pores of throat diameter D, Rep. HEL 155. Hydraulic Engng Lab., University of
variance of Gaussian distribution in In D, California. Berkeley.
bed porosity, void volume/bed volume Soo, H., 1983, Flow of dilute, stable emulsions in porous
porosity contribution of pores of throat dia- media. Ph.D. Thesis, University of California, Berkeley.
Soo, H. and Radke, C. J., 1984, The flow mechanism of dilute,
meter D,, m-l
stable emulsions in porous media. Jmf. Engng Cheer.
drop capture efficiency in pores of diameter D, Fundam. 23, 342-347.
fraction of monolayer coverage Soo, H. and Radke, C. J., 1986, Flow of dilute, stable liquid
filter coefficient, m-i and solid dispersions in underground porous media.
reduced dimensionless filter coefficient, rZL A.I.Ch.E. J. (in press).
Thomasian, A. J., 1967, The Structure of Probrrbifity Theory
viscosity, mPa s with Applications. McGraw-Hill, New York.
ln F is the mean of the Gaussian distribution Tien, C. and Payatakes, A. C.. 1979, Advances in deep bed
in In D,, m filtration. A.I.Ch.E. J. 25, 737-759.
Filtration model for the flow of dilute, stable emulsions in porous media-II 281

APPENDIX A. PARAMETER DETERMINATION At early injection times (T > 1) both bi and cL are very much
This appendix indicates how the filtration parameters less than ss, so that eq. (19) of Part I becomes:
AsI. a and fl are determined from the emulsion flow data. 1
KO
-gp++ AG In (ai/ by)
Since the reduced filter coefficient A,, and the flow-diversion L48)
parameter .x reflect the processe s of the drop retention in the K(r) (l-_/x) (l-a/B) (I+xci) .
sandpack, they should be evaluated from emulsion Substituting cJcL for ui/uL from eq. (2), and using
breakthrough data. The flow-restriction parameter fi. which ln(cL/Ci) nAsr(l -acir) from eq. (A7), we obtain
characterizes how effective the retained drops are in reducing
permeability, is found from pressure-drop data.
First, we relate the inlet and outlet concentration datain the -zKO B
1+ -Ci(T+l)r fA9)
form of the ratio cL/ci. Since eq. (2) shows that cL/ci = uL/ui, K(r) 1-8/e
the next step is to evaluate uL and cri.From eq. (I), with T(L) where we have also neglected higher-order terms in xei, which
=s-landT(O)=r,wehave is assumed to be small. This expression can be fitted to data in
a plot of K,/K vs. ci7 to obtain p/(1 - /?/a) from the slope at
1 - exp [aAs,ci (r - l)]
-=
CL=
(Al) short times. This, combined with a determined by other
&Cl 1 - exp As, - exp [ahsrc; (T - l)] methods, yields 8. The problem here is that a cannot be
and determined from eq. (A4) as before, if breakthrough (cL/ci
uia
- exp [aAstcir] - 1 (A2)
= 0.5) does not occur. In such a case, one hnds a by exercising
,- exp [aAsrcir] . judgement in reconciling estimates based on a projected K,
jthus using eq. (A6) and fits of the short-time data in terms of
Division of these two expressions yields:

(1 -exp [aAsrci(r - l)])exp (aAsIcir)


~-
UL CL - (A3)
ui ci (1 -exp As,-exp [aAsrci(z - I)]) [l -exp (aAsicir)] .

Since drop/sixes and pore sixes in this study are about the K (r)]. Fortunately, the latter is not sensitive to the value of
same order, a large value of the filter coefficient is anticipated. the very large As), so computed values of a and fl are not
Under this circumstance exp Asr * 1 and capture occurs in a influenced significantly by it.
sharp front.
Next, we focus on the breakthrough region and, in
particular, consider the point where cL/ci = 0.5. As will be
justified below, this condition requires that APPENDIX B. DROPLET FLOW RESTRICTION IN A
CYLINDRICAL PORE
aci(r-I)= 1, (A4) This appendix calculates the average flow-restriction par-
independent of the value of As,. In such a case, terms in eq. ameter in the interception pores, <& >, from a mode1 of a
(A3) that contain products such as Ax, aci (r - 1) become of spherical drop resting on the inside surface of a cylindrical
order AsI and thus all exponentials in that equation exceed pore. Happel and Brenner (1965) give the increased pressure
unity greatly. (Clearly, long times are implied here, so that drop AP* due to the presence of a single, stationary solid
r - 1 and r have similarly large magnitudes.) Under these sphere in a cylindrical tube (not necessarily at the wall) under
conditions, eq. (A3) simplifies to: creeping Row [see also Payatakes (1977)]:

exp CaAstci (r - 240, u


CL
-=
1)) AP=- l-(Dp--;)) (I+;&, (Bl)
(A5)
=i exp A,, + exp [aAsrci (r - l)] D;
Note that this function assumes the value 0.5 precisely when where b is the twice the distance from thecentre ofthe particle
eq. (A4) holds, thus providing the justification promised to the tube wall. Equation (Bl) applies to a liquid drop as long
above. The point cL/ci = 0.5 in the efIIuent-concentration as the droplet surface is laden with enough surfactant to
history is used to specify the flow-diversion parameter a, support the no-slip condition. For a drop attached to the
according to eq. (A4). Once a is known, Ast may be obtained wall, b = Dd. Now, when Darcys law along with the
by a best fit of the entire emulsion effluent volume concen- Hagen-Poiseuille result for the pressure drop in an open tube
tration history via eq. (A5). of length L is used to predict AP,, and thus define K,, then
The flow-restriction parameter p is found from the steady- combination with eq. (Bl) (i.e. AP = AP, + AP*) yields an
state permeability: expression for the permeability reduction:
K,fK, = 1 -$/a. (A6) K
-43(3)(2-3. (B2)
With a known, fi follows directly from the measured value of KO
K,/Ko. This parameter estimation procedure does not
involve the entire unsteady permeability behaviour. Only the We recognize that U/E,, is proportional to the ratio of the drop
steady-state pressure in combination with the elIIuent emul- volume to the tube volume. With this observation, com-
sion volume concentration is required. parison of eq. (14) of Part I and eq. (B2) gives the desired
An exceptional case arises when the drop sixes are so large result:
compared to the pore sixes that emulsion breakthrough does
not occur within an experimentally conveniet time frame. In &=:(2-$)ls 14(1-s)l. (B3)
this situation, the value of 01is small while the value of A,,
remains large. Furthermore, in this case we consider rather For the ease of D, 4 D,, j?, = 14. If we take the mean drop-
short times in the sense of r z 1. Thus, we have the dual sixe to pore-size ratio to be 0.1, then & = 13. Equation (B3)
conditions: exp Ast 9 1 and exp [a Asrcis] g 1. Hence, eq. indicates that smaller drops are more effective in reducing
(A3) can be written as: permeability on a unit volume retention basis, in that &
decreases with increasing drop size. This is consistent with the
CL exp (aAs ci + )/exp As,. (A7) experimental findings of Table 1 and Fig. 6.
ci -

Potrebbero piacerti anche