Вы находитесь на странице: 1из 17

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/216756662

Advances in Theory of Fluid Motion in Porous Media

Article  in  Industrial & Engineering Chemistry · December 1969


DOI: 10.1021/ie50720a004

CITATIONS READS

445 691

1 author:

Stephen Whitaker
University of California, Davis
230 PUBLICATIONS   13,902 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Diffusion of Charged Species in Liquids View project

Interpore Interview with Prof. Stephen Whitaker View project

All content following this page was uploaded by Stephen Whitaker on 06 October 2017.

The user has requested enhancement of the downloaded file.


Vol. 61, December 1969

I~EC
Copyright 1969 b ' Pages 14-28
S . y the Am .
Reprinted from oClety and reprinted b enc.an. Chemical
the copyright owner y permission of
General theory of creep flow is proposed which indicates permeability

tensor can be represented in terms of only four scalar constants

Advances in Theory of

Figure 1. Spatially periodic porous media

from the early work of Hubbert (75) who has given a


T hean problem of flow in porous media is of interest to
enormous range of engineers and scientists- lucid account of the famous work of Darcy (7) and has
from the chemical engineer who quite likely is concerned presented, on an intuitive basis, a logical route to a
with multicomponent reacting systems, to the civil theoretical treatment of this problem; from the exhaus-
engineer who is often faced with a two-phase (air-water) tive study by Brenner (5) on creep flow in a spatially
flow process, to the seismologist who must deal with a periodic porous media; from the work of Slattery (30,
deformable porous media in a transient state. Even with 37) on the flow of viscoelastic fluids in porous media; and
this vast array of important problems awaiting our from the author's own attempt at a theoretical treatment
attention, this paper will consider only the simplest of of this problem (38).
all possible cases; the steady, incompressible flow of a There are perhaps three fairly distinct, although cer-
Newtonian fluid under conditions such that inertial tainly not mutually exclusive, approaches to the problem
effects can be neglected. A careful survey of this prob- of flow in porous media. Since a porous medium is
.lem will hopefully provide a sound basis for the study of generally not an ordered structure, the idea of developin~
some of the more important aspects of flow in porous statistical models is appealing, and a number of attempts
media. have been made in this area [7, (3, Chap. 6), (28,
The ideas presented in this paper are drawn primarily Sec. 6.6), (29)]. Another approach is the method of

14 INDUSTRIAL AND ENGINEERING CHEMISTRY


FLOW THROUGH POROUS MEDIA SYMPOSIUM

Fluid Motion in Porous Media


STEPHEN WHITAKER

geometric modeling in which one postulates a geometry


which hopefully bears some resemblance to the porous
media, yet is sufficiently simple to allow the governing
differential equations to be solved. Only two geometric
models will be referred to in this work; the spatially
periodic model of Brenner (Figure 1), and a skewed
capillary model (Figure 2).
In Brenner's model the porous media are made up of an
infinite array of unit cells (two shown in Figure 1) having
boundaries of arbitrary shape and containing one or more
solid particles of arbitary shape. So that the flow in a
unit cell will be amenable to analysis, it is required that
_ the boundary of the unit cell be a fluid surface. Thus
• one must accept the existence of rigid supports of negligi-
ble hydrodynamic resistance which hold the matrix of I, .1
unit cells in place. Figure 2. Skewed capillary model of porous media

VOL 6 1 N O. 1 2 0 EC EM B E R 1 9 6 9 15
The skewed capillary model shown in Figure 2 is a volume V which contains a volume of fluid V f. The
minor variation of the well-known straight capillary volume porosity is given by
model (28, page 114), but it cannot be obtained from
that model by a simple transformation of coordinates (1-4)
because the boundary conditions, which may be imposed
on the flow, also change under a coordinate transforma-
tion. A third approach, which lies somewhere between or more explicitly in terms of the void volume distri- -
bution function
statistical models and geometric models, is the develop-
ment of correct averaged forms of the governing differ-
ential equations. These equations should be valid for any
cp = _1
V
r a(r)dV
Jv (1-5)
geometry, thus the results obtained for specific geometric
models must satisfy the averaged equations. In addition, Throughout this work we will require that the averaging
any statistical model should be in accord with the averaged volume be constant in time and space; however, the
equations. All three methods lead to unspecified param- magnitude of the fluid volume V f may vary with position
eters which must be determined experimentally, and subject to certain restrictions to be discussed later.
the primary objective of theoretical work in this general Before attacking the problem at hand-i.e., obtaining
area is to aid in the interpretation of experimental data. the averaged forms of the governing equations-we must
In attacking the problem of incompressible flow in establish some ideas about area and volume averages
porous media, one is confronted with the fact that the and what is meant by a "meaningful average," and we
final result is pretty well established-i.e., Darcy's must develop an averaging theorem to relate averages
law gives an accurate description of the flow. Because of derivatives to the derivatives of averages.
of this, it is easy to proceed along a variety of ap-
proaches, some of which might well be erroneous or 2. Averages
wholly intuitive, to the correct final result. We will try to In treating problems of flow in porous media, we
avoid this pitfall in the present study and establish assume that some microscopic characteristic length, d,
as carefully as possible a logical, correct route to the exists which is representative of the distance over which
final result. significant variations in the point velocity, V, take place.
Similarly, we designate the macroscopic characteristic
length as L, and assume it is representative of the dis-
1. The Problem
tance over which significant variations in the volume
We are concerned here with the incompressible, creep averaged velocity, (v), take place. In general, one
flow of a constant viscosity Newtonian fluid in a rigid associates d with the poorly defined but intuitively A
porous media. The equations governing this process appealing mean pore diameter, and L with some . .
are: macroscopic dimension representative of the process
1. The continuity equation under consideration. In Darcy's original work one
would consider d to be on the order of magnitude of the
v .V = ° (1-1) "diameter" of the sand grains used in the filter, while L
would be associated with the diameter of the filter bed.
The velocity, V, and position vector, r, are measured Letting if; be some point quantity (tensor of any
relative to an inertial frame imbedded in the porous order) associated with the fluid (velocity, density), we
media; thus the porous media may move with a con- define the volume average of if; as
stant velocity, U, relative to a frame that is stationary
with respect to the fixed stars.
2. The equations of motion
(if;) = ~
V
r
Jv!
if;dV (2-1)

°= - Vp + pg + JJ.V2v (1-2) One can now ask whether the average (if;) is a function
that is suitable to use in analyzing flow in porous media.
The restriction of constant viscosity is imposed by the If we were to plot (if;) versus V we might obtain a function
form of Equation 1-2. similar to that in Figure 3. Such a curve would be
3. The void volume distribution function obtained if the point associated with V was in the solid
region; thus V f would be zero for small values of V.
( ) = {1, if r lies in the fluid region (1-3)
As V becomes larger, portions of the fluid are contained
a r 0, if r lies in the solid region within V and the average increases from zero going
through some fluctuations representative of the varia-
If the function a(r) were known it would be possible, in tions in the point value of if;. For values of V larger
principle, to solve the governing equations and thus than V* the microscopic variations in if; are essentially
determine completely the pressure and velocity fields. smoothed out, but the value of (if;) need not become con-
However, a(r) is never known and we are obviously stant. It should be clear that (if;) is a continuous func- _
forced into a different type of analysis-i.e., a derivation tion for any value of V, but for values of V larger than _
of the volume averaged equations. In this approach V* the volume average (if;) becomes "smooth," and
we associate with every point in space an averaging thus amenable to the type of analysis we have in mind.

16 INDUSTRIAL AND ENGINEERING CHEMISTRY


The linear term in Equation 2-4 has dropped out since Xi

is measured from the centroid of V, thus

Iv XidV = 0 (2-6)

If (I/;) is a linear function of Xi> Equation 2-2 is immedi-


ately satisfied; however, we wish to consider a some-
what more general case.
An order of magnitude estimate of the first integral
in Equation 2-5 can be expressed as,

_1_
V Jv
r xixjdV = 0 ([2) (2-7)

and the order of magnitude estimate for the second


Figure 3. Dependence of average on averaging volume derivative of (I/;) is given by
2
0 (1/;») = 0 «I/;)/V)
( ox/ox} (2-8)
0
If we let l be a characteristic length for the averaging
volume, we can impose our first restriction (throughout Neglecting the higher order terms in Equation 2-5,
this paper restrictions will be denoted by R.1, R.2, we express the average of the average as
and assumptions by A.1, A.2, and so forth).
«I/;» = (I/;) + O[(I/;)(l/L)2] (2-9)
R.1
Clearly our analysis must be bound by the restriction
In addition to this restriction we will require that the
R.2
average of the average be equal to the average, a concept
perhaps more dearly expressed by in order that Equation 2-2 be satisfied.
(2-2) AREA AVERAGES
This is an important restriction, for it must be true Subject to restriction R.2, it will be of interest to com-
if the average is to be a useful quantity with which to pare area averages with volume averages to determine
work. A similar situation occurs in the time averaging any possible differences between the area and volume
of turbulent fields (39, pp 187, 209). In considering averaged forms of the basic equations. To make a com-
(1/;), we note that it is defined everywhere in space, not parison, it is necessary to choose an averaging volume V
just in the space occupied by the fluid. Thus the proper which can bound a plane averaging area Au .k) such
average of this function is given by that the area is always constant. As an example, we

«I/;» = _1
V
r (l/;)dV
Jv
(2-3)
choose V to be a cube of volume [3 and AU,k) to be an
area of magnitude [2 parallel to the (j, k) plane. We let
the center of the cube be the origin of our coordinate
For convenience only, we choose the point with which system, the sides of the cube being orthogonal to the
we associate (I/;) to be the centroid of V and expand (I/;) coordinate axes. The area average of I/; at the origin
in a Taylor series (32, p 228) choosing the centroid as is given by
the origin of our coordinate system. (Here Xi is the
position vector in index notation. Repeated indices 1I0 U ,k) = _1_
AU,k)
r I/;dA
J AI
(2-10)
are summed from 1 to 3.)
Here A, represents the fluid area contained within
A(j,k),the dependence of A, on the particular coordinate
surface under consideration being understood. The
volume average can now be expressed as
(2-4)
(1/;)0 = -
1
l
i XC ') =

XC') = -
+ 1/2 U
1/2
lI ,k)dx(i) (2-11 )
Here the subscript zero indicates that the function is
evaluated at Xi = O. Forming the average of (I/;) where (i), (j), and (k) are distinct. If we expand
indicated by Equation 2-3 yields if;U,k) in a Taylor series in X (ih we obtain

(2-5) (2-12)

VOL. 6 1 N O. 1 2 0 E C EM B E R 1 9 6 9 17
where the summation convention does not apply to with gradients of point functions, and we need to explore
indices in parentheses. Substituting Equation 2-12 the problem of averaging gradients of functions for we
into Equation 2-11 and evaluating the integrals yield are interested in obtaining gradients of averages rather
than averages of gradients.
(2-13) A general relationship between gradients of averages
and averages of gradients for functions defined in both
the solid and the liquid phase and suffering a jump dis-
Estimating the order of magnitude of the second deriva-
continuity at the phase interface has been presented by
tive as
the author (36); however, the result (37) is not available
CA/(J,k») in the readily accessible literature, and furthermore the
( -OX2(t)- 0 = O(if;U,k)/V) (2-14) development is a rather heavy-handed one. A similar
result has been obtained by Slattery (30) for functions
we can express Equation 2-13 as defined in the fluid phase by a rather ingenious applica-
tion of the general transport theorem [(35, p 347),
(2-15) (39, p 88)]. Since the original presentation given by
Slattery is rather terse, and since the development has
Thus if our analysis is to be restricted by I « L, it follows value in the analysis of more complex systems than the
that volume and area averages are essentially equivalent. one treated here, we will present a detailed version of
It is of some interest to apply this result to the void the theoretical development.
volume distribution function, a, which leads us to The general transport theorem can be written as

(a) ~ c':u ,k)


Since the volume average of a is the volume porosity, <P,
(2-16)
-d
dt
i Vet)
ifidV = i Vet)
- dV
oif;
ot
+ f
A(t)
if;w·ndA (3-1)

while the area average is the plane porosity, we find where Vet) is a volume bounded by the surface A(t)
that our restriction I «L requires the volume porosity moving at a velocity w which may be different from the
be essentially equal to the three plane porosities velocity of the fluid, v. If w is taken to be equal to V
the volume Vet) is the region occupied by a body and
(2-17)
Equation 3-1 is referred to as the Reynolds' transport
This relation between the porosities applies in general to theorem (2, p 84). To apply the general transport
anisotropic porous media; however, it is not necessarily theorem to the problem of averaging gradients, we
valid for a highly structured porous media in which the consider a point in the porous media (in either the fluid
order of magnitude estimate given by Equation 2-14 or solid) located on an arbitrary, continuous curve, the
might not apply to aU ,k). Note that if porous media are arc length measured along this curve being s. With
homogeneous or uniform, Equation 2-17 takes the form each point on the curve, we associate an averaging
volume V(s) bounded by a surface A(s). Of this volume,
<P = a(1,2) = a(2.3) = a(3,1) (2-18)
a portion is occupied by the fluid; we designate this
regardless of whether or not the medium is isotropic. portion by VIes) and its bounding surface by A!(s).
The local volume porosity is then given by
3. SlaHery's Averaging Theorem
<p(s) = V!(s) (3-2)
Up to this point we have been discussing the averaging V(s)
of some point function if;; however, in developing the
averaged form of the basic equations we will be dealing We consider now how the integral over VIes) of some
quantity if; changes as a function of s. Although the
transport theorem is generally viewed as a rule for taking
AUTHOR Stephen Whitaker is Associate Professor in the the time derivative of an integral having limits which are
Department of Chemical Engineering, University of California, functions of time, the analysis can be carried over directly
Davis, CalzJ. This work was supported by the National to the problem of obtaining the directional derivative
Science Foundation, Grant GK-477. The author also acknowl- (16, p 107). Thus, from Equation 3-1, we may proceed
edges the courtesy of the hand-written notes made available to directly to
him by Professor Howard Brenner, Carnegie-Mellon Univer-
sity, and the helpful discussions with Professor J. C. Slattery,
Northwestern University. This paper was presented as part of
-di
ds· VI(S)
if;dV = i VI(S)
- dV
(Oif;)
OS
+ f AI(s)
if; (dT)
-
ds
. ndA

(3-3)
the Symposium on Flow Through Porous Media, The Carnegie
Institution, Washington, D.C., June 9-17, 7969. where n is the outwardly directed unit normal for AAs).

18 INDUSTRIAL AND ENGINEERING CHEMISTRY


to the interface, thus

(~). n = 0, on the solid-fluid interface (3-6)

We may express the area A,(s) as


(3-7)

where Ae(s) represents the area of the entrances and


exits on the surface A(s), and Aj(s) represents the area
of the solid-fluid interface contained within V(s). In
view of Equations 3-6 and 3-7, we may write Equation
3-4 as

-d
ds
i Vr(s)
1/IdV = i A.(s)
1/1(dr)
- 'ndA
ds
(3-8)

Letting ToeS) be the position vector locating the reference


point on the arbitrary curve and pes) the position vector
locating points on A,(s) relative to the reference point
Figure 4. Physical significance of derivative of volume integral with
respect to s (which is at the center of a sphere for the case illustrated
in Figure 4), we can write

The formal mathematical development of Equation


res) = Toe s) + pes) (3-9)

3-3 is obtained by replacing t in the derivation of the The directional derivative (39, p 232) can be expressed
transport theorem with the arc length, s. This simply in terms of ToeS) as -
means that we assume a continuous and invertible
mapping exists ~ = (dro) . V (3-10)
ds ds
r = r[~, s]
Substituting Equations 3-9 and 3-10 into Equation 3-8
~ = r(s = 0)
yields
which maps V,(s = 0) into V,(s). Here r and ~ are
comparable to spatial and material position vectors dr o). V
( ds
r
J Vr(s)
1/IdV = r
J A,(s)
1/1 (dro) . ndA
ds
+
(39, p 76). We will concern ourselves with quantities

i (dP)
which are explicit functions of time and the spatial
coordinates only, thus 01/l/os = 0 giving us 1/1 - ·ndA (3-11 )
A,(s) ds

-d
ds
~
Vr(s)
1/Ids = f Ar(s)
1/1 (dr)
- • ndA
ds
(3-4) Since dro/ ds is not a function of the limits of integration
for a fixed value of s, we may remove dro/ ds from the
integral sign to obtain
To clarify the dependence of 1/1 upon s, we note that 1/1

Jr Jr
is, in general, a function of the spatial coordinates and dro
time, 1/1 = 1/1 (Xj, t), and the spatial dependence may be (V 1/IdV - 1/IndA)
ds Vr(s) A.(s)
represented in terms of the arc-length s along an arbi-
trary curve to give, 1/1 = 1/I[Xj(s), t]. It should be clear
that the total derivative of 1/1 with respect to s is generally
fA.(S) 1/Ie~) . ndA (3-12)
nonzero and given by
Provided the volume V(s) is translated without rotation
along the arbitary curve C, any differential variation in P
(dVt)
ds
(01/1 ) (dXj)
OXj ds
(3-5)
is a tangent vector to the surface of the volume (21,
p 168)-i.e.,
while the partial derivative is zero since 1/1 does not
A depend explicitly upon s. The physical significance of dP) _ {a unit tangent vector} (3-13)
, . the derivative of the volume integral with respect to s ( ds - to the surface Ae(s)
is shown in Figure 4. It is clear from that figure that
dr/ ds over the solid-fluid interface is a tangent vector Thus dp/ ds and n are orthogonal and the right-hand

VOL. 6 1 N O. 1 2 0 EC EM B E R 1 9 6 9 19
side of Equation 3-12 is zero. Since dro/ ds is an arbitary porous media is moving at a constant velocity u, we
vector, we obtain must remember that V is to be interpreted as the relative
velocity which is also zero over Ai. With this in mind
v r
Jv!
if;dV = r
JA,
if;ndA (3-14) we write Equation 4-2 as

Here the dependence of V f and Ae upon s has been


V . (v) = 0 (4-3)
omitted; however, we must remember that the fluid where it is understood that (v) is the average fluid
volume V, contained within the averaging volume V velocity relative to the constant porous media velocity,
very definitely depends upon the spatial coordinates. u.
The same may also be said of A •. Before attacking the equations of motion we define a
The above result, due to Slattery (30), can be used to piezometric pressure, P, as
provide a useful expression for the average of a deriva-
tive in terms of the derivative of the average. Making P = p - po + PI{! (4-4)
use of the divergence theorem, where po is some reference pressure and I{! is the gravita-
tional potential function which must satisfy the condi-
r Vif;dV = JAi
Jv,
r if;ndA + JA.
r if;ndA (3-15) tion

we may write Equation 3-14 as


g = -VI{! (4-5)
Equation 4-5 specifies I{! to within an arbitrary constant,
r Vif;dV =
Jv!
V r if;dV + JAi
Jv!
r if;ndA (3-16) and we will choose this constant so that I{! = 0 when p =
po. In terms of Equation 4-4, the equations of motion
Remembering that the volume average is defined as can be written as

(if;) = ~ r if;dV
o= -VP + p,V2v (4-6)
V Jv Again, it should be kept in mind that v is the velocity
relative to u. Taking the divergence of Equation 4-6
we may divide Equation 3-16 by V (remember that V
and making use of the continuity equation show that P
is a constant) in order to obtain
must satisfy the Laplace equation

A more general form of this result, relating the average


of the derivative to the derivative of the average, has been
presented elsewhere by the author (37). We must
(3-17)
If we now operate on Equation 4-6 with the Laplacian,
the velocity must satisfy the biharmonic equation
(4-7)

(4-8)
,
remember that there was one important restriction In addition, the velocity must satisfy the "no slip"
made in the derivation of Equation 3-17 namely condition at the solid-fluid interface.

The averaging volume V is constant, and its orienta- v = 0 on At of V (4-9)


tion relative to some inertial frame remams un- If we were to specify v over the entrances and exits of any
changed R.3 averaging volume V-i.e.,
Having discussed averages and averages of derivatives, v = f(x) on A. of V (4-10)
we are now in a position to proceed with our analysis
of flow in porous media. a unique solution for v would result (77, Chap. 2) in the
region V. It follows that Equations 4-8, 4-9, and 4-10
4. Darcy's Law are sufficient to determine (v) for the particular region in
The governing equations that we wish to analyze question. Thus, these three equations lead to
are given by Equations 1-1 and 1-2. The volume (v) = g(ro) (4-11)
averaged form of the continuity equation may be written
as where ro is the point with which the averaging volume is
associated.
2.- r V.
V Jv!
vdV = 0 (4-1) Now we are going to assume that this problem can be,
in effect, turned around and stated in the following
and by our averaging theorem, Equation 3-17, we way:
obtain Given the governing differential equation (Eq.
4-8), the no-slip condition (Equation 4-9), and
+ _1 r =
V . (v)
V JAi
V· ndA 0 (4-2) the average velocity (Equation 4-11), the veloc-
ity field is uniquely determined A.l t
if the porous media is fixed in space v = 0 over At One can prove this for a number of flows. For example,
and the second term in Equation 4-2 is zero. If the the one-dimensional flow in a capillary tube is uniquely

20 I N 0 U S T R I A LAN 0 ENG I NEE R I N G C HEM 1ST R Y


determined by the equations of motion, the no-slip length measured along this curve and substituting
condition at the walls of the capillary, and the average Equation 4-12 for v allow us to write
velocity. Flow past a sphere is uniquely determined by
dP
the equations of motion, the no-slip condition, and the - = p.:A. • (V"lM) . (v) (4-16)
ds
average velocity, provided the velocity is uniform at
infinity. Brenner (5) has been able to prove assumption Here we have used the relation :A. . V = d/ ds and used
Ai, provided the flow is spatially periodic. It seems the same order of magnitude arguments that were im-
doubtful that a general proof of assumption Ai will posed between Equation 4-13 and Equation 4-14.
ever be given, for in each case where we know it holds, (This step and several subsequent ones are all rigorous
there is always one additional requirement placed on the if (v) is a constant vector or is a linear function of the
flow-i.e., the flow is one-dimensional, or uniform at spatial coordinates. For the general case, we constantly
infinity, or spatially periodic. Still, assumption Ai invoke the argument that the spatial variations in (v)
has some intuitive appeal, and, one way or another, it are small compared to those of M.) Integrating from
appears in all previous theoretical treatments of this s = 0 where P = Po to any point on the arbitrary curve
problem. yields
Since v and (v) are continuous and defined every-
r~=8(T) }
where in the region under consideration, we can map per) = Po + p. { J~=o [:A.' (V2M) ]d'l) . (v) ( 4-17)
(v) into v-i.e.,
v = M . (v) (4-12) The term in braces depends only on the geometry
through Equation 4-14 and on the position vector r
where M is the transformation which maps the average through the upper limit of integration. Somewhat
velocity into the point velocity. (Tensors are indicated more simply we can state that P is determined to within
by capital letters in boldface type, while vectors are an arbitrary constant by the expression
represented by lower case letters in boldface type.) By
the assumption Ai, the mapping is unique although not P = -p.m· (v) (4-18)
necessarily linear. If we substitute the above expression
where m depends on the position vector r in addition to
for v into Equation 4-8 we obtain
the detailed structure of the porous media. (The minus
(4-13 ) sign is included in Equation 4-18 so that the perme-
ability for isotropic porous media will be positive.)
From previous order of magnitude arguments, it is clear
Returning now to Equation 3-14, we can substitute
that V 4(v) is negligible and the transformation matrix
P for if! on the left-hand side and - p'm • (v) for if! on the
is determined by
right-hand side and divide by sides by V to obtain
(4-14a)
We must keep in mind here that Equation 4-14a follows V [ ~ fv,PdvJ = - ~i, p'm . (v)ndA (4-19)
rigorously from Equation 4-13 if (v) is a constant vector
or depends on the spatial coordinates to the third order This can be expressed as
or less. For the more general case, we need only re- V(P) = _p.K-1 . (v) (4-20)
member that significant variations in (v) take place over
a distance L and significant variations in M take place where K -1 is given by
over a distance d where d < < < L.

M = OonA; (4-14b)
K-1 = ~
V
r nmdA
JA,
(M) = U at ro ( 4-14c) (4-21 )
where Equation 4-14c is a logical extension of assump-
tion A1. From Equations 4-14 we conclude that M We are now confronted with the question as to whether
is independent of (v), thus the point velocity is a linear K -1 has an inverse. We know that if P is constant, the
vector function of the average velocity. From Halmos pressure variation is hydrostatic, the point velocity is
(13, p 62) we know that a linear transformation is zero, and (v) = O. The same line of reasoning does not
invertible if, and only if, M . (v) = 0 implies that (v) = necessarily follow if (P) is constant; however, it is
0; however, v = 0 everywhere in the solid regardless intuitively appealing, and we will make the assumption
of the value of (v). Thus we cannot say that M has an
V(P) = 0 implies (v) = 0 A2

,
inverse.
Returning now to Equation 4-6, we form the scalar The theorem of Halmos (13, p 62) then indicates that
product with :A. K -1 has an inverse which we designate by K so that
(4-15) Eq ua tion 4-19 takes the form

where :A. is a tangent vector to some arbitrary curve lying 1


(v) = -- [K V(P)] (4-22)
wholly within the fluid region. Letting s be the arc p.

VOL. 61 NO. 12 DECEMBER 1969 21


Up to this point we have dealt with the equations of In general then, Equation 4-29 takes the form
motion in a form which kept the analysis as neat as
1
possible. Now we would like to express the result in a (v) = - -K {<I>[V(p - po), - pg] +
form which is more directly applicable to the interpreta- J.I.
tion of experimental data. Substitution of Equation [(P - po),- pr· g]V<I>} (4-33) '-"
4-4 yields
We have already restricted the analysis to porous media
having only slowly changing physical properties, thus
(v) = (4-23)
the practical form of Equation 4-32 is

Here we have removed the density from the average <l>K


(v) = - - [V(p - po), - pg] (4-34)
since we are considering only incompressible flows. J.I.
To within an arbitrary constant, we can express the
Note that the average form used for the velocity has been
gravitational potential function as
chosen with the idea that volumetric flow rates can be
(4-24) readily calculated by the relation
'" = -r· g
and the average becomes Q= Is (v) . ndS (4-35)

("') = - _1
V
r
Jv!
(r. g)dV (4-25) while the pressure and body force terms are given in a
form most suitable for interpretation by the experi-
mentalist. (Here the surface area S must be large
We note that g is constant which reduces to
compared to [2 so that the integrals accurately represent
(4-26) the volumetric flow rate.) Massarani (20) has experi-
mentally verified Equation 4-34 for one-dimensional
where r is the position vector locating the centroid of flow in a nonhomogeneous porous media.
the fluid volume, V,. Because of the restrictions we have We can carry out analysis just a bit further by examin-
placed on the variation of the porosity, r can be con- ing Equation 4-21 and noting that an order of magnitude
sidered essentially identical to ro, the position vector estimate gives
locating the centroid of the averaging volume, V.
An experimental measurement of the fluid pressure
in a porous medium would probably determine the thus
quantity (p - po), which is defined K-l = O(d-2)

(p - po), =
1
v r (p
, Jv! - po)dV (4-27) and the inverse of K
by
-1 has an order of magnitude given

Thus the average pressure III Equation 4-23 can be


written as
We can define a dimensionless permeability tensor It
(p - po) = <I>(p - po), by the equation
(4-28)

Substitution of Equations 4-26 and 4-28 into Equation


4-23 yields and express Equation 4-34 as

1 <I>d2) _
(v) = - - K . {V[<I>«P - Po), - pr·g)J} (4-29) (v) = - ( ----;; K· [V(p - po), - pg] (4-36)
J.I.
At this point we can say little about K. except that it
The general interpretation of the quantity Vr is
depends on the structure of the porous media. In
Vr = U (4-30) general, the determination of the scalar components ofK.
represents an enormous experimental task, although
where U is the unit tensor. In taking the gradient of geometric models can be helpful in shedding some light
averaged functions, we think of these functions as on this problem. In particular, one can argue from the
being defined at points in space, and if we associate the capillary model of Kozeny (4, p 196) that It should be
centroid of the averaging volume it appears consistent proportional to <1>2/(1 - <1»2, thus the difficulty in corre-
to write lating data is reduced. The spatially periodic model of
Brenner (5) indicates that K. should be symmetric;
Vro = U (4-31)
thus the results of those investigators who have tacitly
In as much as r is essentially identical to To we can ex- assumed this to be self-evident become more credulous, 6.
press the body force term in Equation 4-29 as and the experimentalist is faced with six undetermined ...
components instead of nine. The experimental deter-
V[-<I>pi· g] = -<I>pg - pi . gV<I> (4-32) mination of the scalar components of K. indeed repre-

22 IN 0 U 5 T R I A LAN 0 ENG I NEE R I N G C HEM 1ST R Y


sents a formidable task and in the next section we would where At are the scalar components of the unit vector
like to address this problem. pointing in the direction of the capillary tubes-i.e.,
Al = a, A2 = /3, and A3 = 'Y. Here we have found that
5. The Permeability Tensor the permeability tensor is symmetric, an interesting result
a In discussing the problem of the experimental deter- especially in light of Brenner's analysis of spatially
• mination of the scalar components of K or K, we will periodic porous media. In that structure, the anisot-
consider two geometric models: Brenner's spatially ropy results from the geometry and orientation of a
periodic model shown in Figure 1 and the skewed solid particle within a unit cell; thus the orientation
capillary model shown in Figure 2. For the former, occurs on a length scale comparable to t. The situation
Brenner has proved that K is symmetric even if the is quite different in the skewed capillary model where the
porous media is anisotropic. The skewed capillary orientation occurs over a length scale on the order of L.
model shown in Figure 2 has apparently never been Since these two rather different types of porous media
analyzed, presumably because it appears to be nothing have symmetric permeability tensors, it is appealing to
more than a minor variation of the straight capillary assume that K is always symmetric, and a number of
tube model. This is quite true-still there is something investigators have presented arguments in support of
to be learned from the analysis. this point of view [11,25, (27, p 625) J.
The skewed capillary model consists of a slab of thick- Our next step is to try to put forth some general ideas
ness L, infinite in the y- and z-directions, and embedded about the analysis of anisotropic porous media and to
with capillary tubes of diameter d. We require that the offer a proof, suggested by Slattery (37), that the perme-
tube diameter be much smaller than the slab thickness, ability tensor is always symmetric.
d «< L, so that our requirements for meaningful aver- Following the work of Ericksen (8, 9), we will assume
ages are satisfied. The flow occurring in the capillaries at the outset that there is a unique orientation vector w,
will be one-dimensional, laminar, and will satisfy assump- representing a preferred direction, which describes the
tions A.l and A.2. Thus the average pressure gradient anisotropic nature of the porous media:
will be related by Darcy's law, Equation 4-22, to the
average velocity.
K = K(w) (5-2)
In analyzing the flow that takes place in the capil- Subject to the constraints on the porous media listed in
laries, we imagine that either the pressure or the pressure Sec. 2, the vector, w, may depend on the spatial co-
gradient can be maintained at some constant value over ordinates. For isotropic porous media, w will clearly
y-z planes at x = 0 and x = L. Under these circum- be zero, and for the skewed capillary model, w is ob-
a stances we can consider three distinct flow processes viously parallel to the capillaries. If the particles in
. , given by Brenner's spatially periodic porous media are ellipsoids,
w would be parallel to the major axis, but what can we
o(P) o(P) __ o(P)
I. -Cl , oy oz o say about the orientation vector for the more general
ox porous media? Very little, it seems, at this point;
(V)(l) uo(l)(ia + j/3 + ky) however, we will overlook for the present the fact that the
three scalar components of w may have to be determined
UO(l) Cl (arP/32j.1.) experimentally and proceed.
o(P) = o(P) = 0 o(P) = _ C2 If a porous media is symmetric about one plane, then a
II. change of coordinates of the type
ox oz 'oy
(V)(2) = uo(2)(ia + j/3 + k'Y) (5-3)
should leave the permeability tensor unchanged, thus
Uo (2) = C2 (/3rP /32j.1.)
Kij = Ktj
o(P) = o(P) = 0 o(P) = -C
III. 3 for the transformation given by Equation 5-3. This
ox oy , oz
requires that
(V) (3)
K12 = K21 = K31 = Kl3 = 0 (5-4)
and the array of coefficients becomes
Here a, /3, and 'Y represent the direction cosines for the
orientation of the capillary tube, and Uo represents the Ktj = (~11 ~22 ~23) (5-5)
magnitude of the average velocity determined by the o K32 K33
Hagen-Poiseuille law. Substituting these expressions
for the pressure gradients and average velocities into If the media are symmetric about two orthogonal planes,
6 Equation 4-22 allows us to determine the scalar compo- a coordinate transformation of the form,
~ents
. ofK given by (5-6)
Ktj = G~) AtAj (5-1)
in addition to the transformation given by Equation 5-3,
leaves the permeability tensor unchanged. In addition

VOL. 6 1 NO. 1 2 0 E C E M B E R 1 96 9 23
to the restrictions given by Equation 5-4, we now require To put our analysis of the dependence of K upon (0)
that on a sound basis, we first note that the vector (0) can be
Kn = Ka2 = 0 (5-7) represented by the skew-symmetric tensor 0 by the
relation
and the array of coefficients becomes
1
Kll Wi = 2 eijSljk or Q jk = ejkiWi (5-11)
Kij = ~ (5-8)
( The functional dependence of the scalar components of
K can therefore be expressed as
Clearly porous media symmetric about two orthogonal
planes must be symmetric about a third orthogonal (5-12)
plane so that only the diagonal elements are nonzero
We now assume that Kij can be expressed as a poly-
in the principal axis coordinate system established by
nomial in Qnp, and apply the Caley-Hamilton theorem
the three orthogonal planes of symmetry. Such ma-
(74, p 64) to obtain a closed-form expression.
terials are called orthotropic (72, p 159) and K is com-
pletely specified by the principal values K ll , K 22 , and Kij = Ail + AijmnIlQmn + AijmnIllQmpQpn (5-13)
Kaa. One can easily show that if K is symmetric in one
Here the A-tensors are polynomial functions of the single
coordinate system it is symmetric in all coordinate scalar invariant of Qmn-i.e., the magnitude of w. Sub-
systems; thus the permeability tensor for orthotropic stituting Equation 5-11 into Equation 5-13 yields
materials is symmetric. The skewed capillary model
represents an or tho tropic material-one plane of sym-
metry being perpendicular to the direction of the capil-
laries and the other two planes being orthogonal to the (5-14)
first. With a little algebraic effort we can show that
If, in addition to the transformations given by Equa-
tions 5-3 and 5-6, the permeability tensor is unchanged (5-15)
by a transformation of the type Since the A-tensors are unspecified we can redefine and
Xl = Xl cos 8 + X2 sin 8 regroup these terms to obtain

X2 = -Xl sin 8 + X2 cos 8 (5-9) (5-16)

Xa = Xa where once again the B-tensors are polynomial functions


of the magnitude of w. In going from Equations 5-14
the porous media are said to be transversely isotropic
and 5-15 we have replaced (0) with wJ.. where J.. is the unit
(12, p 159) and we find that
orientation vector. No generality is lost in this step
K22 = Kaa since the B-tensors are already functions of w. If J..
is indistinguishable from - J.., as it will be for any ortho-
At this point we have learned nothing about how one
tropic material, the second term must vanish to yield
might expect the orientation vector, (0), to depend on
the nature of the porous media. For orthotropic, (5-17)
transversely isotropic, and isotropic media, one might
for orthotropic porous media. If the porous media is
guess the form of (0) to be
isotropic, J.. = 0 and BiF must be an isotropic second-
order tensor and we obtain
(0) = e(l)
KllV3
[ _ I 1] + e(2)
[K22
_ ~
V3 - 1] +
V KijKij V KijKij (5-18)

K a3 v'3 ] for isotropic porous media.


e(3) [ - 1 (5-10) At this point we proceed with a development sug-
v'KijK ij
gested by Slattery (31) and require that Darcy's
to within a constant multiplier. Here K ll , K 22 , Ka3 law satisfy the principle of material frame-indifference
represent the principal values. Here also e(l), e(2h and [(22), (34, P 41)]. This follows from the fact that
e(3) are the orthogonal base vectors for the principal Darcy's law is nothing more than a balance of forces,
directions. For isotropic porous media, Kll = K22 = and forces are regarded as being frame-indifferent (33,
K 33 , and (0) equals zero as required. For transversely p 27). Thus the equation
isotropic media, the angle between (0) and e(3) (in our
example) would depend on the ratio of Kll to K 22 . ~ [B i].1 o(P)
;".
+ BijkI n
"k
. o(P)
;".
+
For the skewed capillary model, (0) would be parallel }Jo VXj VXj

to the capillary tubes as required. The general problem


seems to indicate that (0) must be measured experi-
mentally; however, it would seem that good intuition
\. O(P)]
B ijmn Ill\."m"n Ox j (5-19) e.,
and some ingenious modeling might lead to some general must be frame-indifferent. The most general change of
rules regarding the nature of (0). frame is given by (34, p 41)

24 I N D U 5 T R I A LAN DEN GIN E E R I N G C HEM 1ST R Y


x* c(t) + Q(t) . [x - Xo] (5-20) This result is consistent with that given by Slattery (31)
and indicates that the permeability tensor is symmetric.
t*=t-a (5-21)
If Bm = 0, we obtain the proper form of Darcy's law
where c(t) is a time-dependent point, Q(t) is a time- for the skewed capillary model, and the symmetry
dependent orthogonal tensor, Xo is a fixed point that is requirement is in accord with the spatially periodic
mapped into c(t). The time-dependent orthogonal model. From the experimentalist's point of view, the
transformation, Q(t), represents a rotation and, possibly, problem has been reduced to measuring four scalars
a reflection from a right-(left)-handed coordinate system (B(1), B(2h A1, A2) although the problem could be simpli-
to a left-(right)-handed coordinate system. A vector fied if a method of making a good guess about the nature
b is said to be frame-indifferent if it satisfies the condition of ~ could be developed. Our guess, given by Equation
5-10 is not entirely out of line with the result obtained
b* = Q(t) . b (5-22)
here. Comparing Equation 5-28 and Equation 4-36,
In considering Equations 5-19 we express the functional we find
dependence of <v) upon ~ and V<P) as (5-29)
<v) = F(~, V<P» (5-23) Taking the trace of this expression allows us to solve for
and require that (33, p 34) B(2)

<v)* = F(~*, V<P)*) (5-24) (5-30)

This expression means that the material properties of the For orthotropic materials, we can put Equation 5-29
porous media-i.e., the relationships among velocity, in the principal axis coordinate system and solve for ~
orientation, and pressure gradient are indifferent to the
choice of observer. In writing Equation 5-24 we have
~ -_ e (1) ~ _
Kll/
_
B(l)
/ _
-) 1 +
(KijKij B(l) - 3
used the fact that the relative velocity is frame-indifferent.
We also regard forces-i.e., V<P)-as being frame- e J (K22/B(1» - 1 + e J (R3a1B(l» - 1
indifferent (33, p 27), and we assume that the orientation (2)" _ _ / _ )
(KijKij Bm - 3
(3) " ( _ _ _ )
KijKij/B(1) - 3
vector ~ is frame-indifferent. Thus ~ must be thought
of as the gradient of some scalar property of the media. We must keep in mind that two assumptions were
Equation 5-24 requires that the B-tensors satisfy the made in the course of this development:
The anisotropic nature of a porous media can be

e
conditions
uniquely described by a single orientation vector, II)
Bi/* = Bi/ A.3
Bjj"Il* = Bjj"Il (5-25) The principle of material frame-indifference applies
B ijmnIIl* -- B ijmnIII to Darcy's law A.4

which leads to the restriction that the B-tensors must be 6. Transient Creep Flow
isotropic (not that the porous media are isotropic) (34, For transient creep flow our governing equations take
p 23), and we conclude the form
Bd = BIOij (5-26a) ov
Bij"Il = 0 (5-26b) p ot -VP + ~V1-v (6-1)

BjjmnIII = B(l)IIlOjjOmn + B(2)IIl(OijOmn + OimOjn) + v . v = 0 (6-2)


B(3)1II(OjjOmn - OinOjm) (5-26c) We can follow the analysis of Sections 4 and 5, provided
We note here that the third-order tensor ejj" is isotropic assumption A.1 can still be applied; a reasonable proce-
only to right-(left)-handed transformations and must be dure only if the flow is quasi-steady. This means that
considered a hemitropic tensor-valued function, if the left-hand side of Equation 6-1 is negligible and the
reflections are to be allowed (34, p 24). time dependence of <v) and <P) enters the problem only
Subject to Equations 5-26, we write Equation 5-19 as through the boundary conditions. We can examine the
problem of transient flow in a tube (4, p 129) to gain
1
<Vi) = -- [B(l)Oij
~
+ B(2)AjAj] O<P)
~
UXj
(5-27) some insight as to when the quasi-steady assumption
might be valid. If a tube filled with fluid is subjected
to a sudden change in the pressure drop, essentially
Putting this result in the form expressed by Equation
steady flow occurs for times greater than to where
4-36, we obtain our final form of Darcy's law for aniso-
tropic porous media (6-3)

e <v) = _(if>:) (BmU + B(2)~~) . [V<p - po) f - pg] Here d is the tube diameter and v is the kinematic
viscosity. For the purpose of estimating microscopic
(5-28) transient times in porous media, a practical lower

VOL. 6 1 NO. 1 2 DEC E M B E R 1 96 9 2S


I
bound on is 10-2 cm2/sec, and an upper bound on d Since the boundary conditions are independent of time,
might be on the order of 10-1 cm. This gives a micro- we conclude that M cannot depend on time, but only
scopic transient time on the order of 1 sec, and if the on the structure of the porous media. This immediately
transient time for the macroscopic process is much leads us to the conclusion that the time dependence of
larger (say on the order of minutes), we should be able the point velocity is completely determined by the time
to treat the flow as quasi-steady and the analysis presented dependence of the average velocity-i.e.,
in Sections 4 and 5 will be valid. This means that
Equation 5-28 and Equation 4-3 are to be solved subject
v =M . g(ro, t) (6-14)
to boundary conditions of the type This, of course, is the quasi-steady flow condition, and it is
(v) = g(r, t) over S (6-4) not reasonable at all for the conditions we are now
considering. This unhappy condition is brought about
where S is the surface of some macroscopic region. by the assumption A,l', for in making that assumption
If the left-hand side of Equation 6-1 is not negligible, the time dependence of Equation 6-7 is replaced by
we reformulate the problem so the governing equation for the time dependence of Equation 6-8. We might
v is consider the simplification in going from Equation
6-10 to Equation 6-11 as a possible error in our analysis;
(6-5) however, this simplification is quite reasonable, and is
strictly satisfied if (v) is a linear function of the spatial
and the boundary conditions are coordinates. As an example, consider the case of one-
dimensional, transient, incompressible flow in homo-
v = 0 on At of V (6-6)
geneous porous media. For such a flow, the average
v = /(r, t) on Ae of V (6-7) velocity (v) is always independent of the spatial coordi-
nates, being a function of time alone, and Equation
The solution of Equation 6-5, subject to these boundary
6-11 logically follows from Equation 6-10.
conditions, is unique (17, Chap. 4), thus Equations 6-5,
If we were to throw caution to the winds, we might
6-6, and 6-7 are sufficient to determine (v) for the region
discard assumption A,l' but retain Equation 6-9, allow-
in question, and we write
ing M to be a function of time. Under these circum-
(v) = g(ro, t) (6-8) stances we can modify the development given in Section
4 as follows. Beginning at Equation 4-15 we would
Following the development in Section 4, we assume
write
Given the governing differential equation
(Equation 6-5), the no slip condition (Equa- A . VP = P.A . ~v - PA . (~:) (6-15)
tion 6-6), and the average velocity (Equation
6-8), the velocity field is uniquely determined A,l' Substitution of Equation 6-9, keeping in mind that M
is a function of time leads to
This assumption has less appeal than assumption A,1,
primarily because we do not have a set of special cases
for which we know it holds true. Nevertheless, once
the assumption is made we can express V in terms of (v)
~: = P.A . (v2M) . (v) - {A . e:)]. (v) -

by the unique mapping, M p(A . M) . o(v) (6-16)


ot
v = M . (v) (6-9)
As usual, we assume negligible spatial variation of (v)
Substitution of Equation 6-9 into the governing differen- and o(v)/ot to obtain
tial equation yields
f~=8(T) [
P(r) = Po + p. J~=o
[(~ IIV2)~M ] { A . (V2M) -
- . (v) +

M . [(~ - II~ )V2


(V) ] = 0 (6-10)

Following the arguments given in Section 4, we may


conclude that V 2(v) is negligible and M must satisfy the
equation Here we see that, to within an arbitrary constant, P is
given by
(~-II~)~M = 0 (6-11)
P = -p.m· (v) - pb . o~:) (6-18)
subject to the boundary conditions of
M = 0 on Ai (6-12)
Here we must remember that m is a function of and t, _ I
and that b is a function of t. Rearranging and sub-
(M) = U at ro (6-13) stituting into Equation 3-14 yield

26 I N 0 U 5 T R I A LAN 0 ENG IN E E R I N G C HEM 1ST R Y


V[~ f
V J VI
(p - b . p ()(V»)
()t
dV] = -~ f p.m·
V J A.
(v)ndA
deviations of the local point velocity from the local
average velocity
(6-19) v = (v) + iJ (7-S)
This can be expressed as Substituting this expression into Equation 7-4 and noting
()(v)
V(P) - V [ (b) . p Tt 1 = _p.K-l . (v) (6-20)
«v» = (v) and therefore (fJ) = 0, we find

pe~~) + (v) . V(V») = -V(P) + p.(~v) +


If the spatial variation of (b) is large compared to the
spatial variation of ()(v)/()t, we could express Equation
6-20 as ~ f PndA - pV . (iJfJ) (7-6)
V JAi
()(v) Assuming that (v) = 0 if, and only if, v is everywhere
V(P) - B . p - = _p.K-l . (v) (6-21)
()t equal to zero, a mapping of (v) into v exists which cannot
The existence of an inverse for K -1 is not so easy to argue be linear nor is it likely to be unique, and we can write
in favor of as it was in the case of steady flow, neverthe- v = M . (v) (7-7)
less we will assume it exists and write Equation 6-21 as
Expressing the last two terms in Equation 7-6 as a force
(v) = - !p. [K . V(P) - (Kt) . p ()(v)]
()t
(6-22) I, we can write

where K . B has been replaced by K '. It must be pe~~) + (v) . V(V») =


remembered that both K and K' are time dependent.
Several workers (19, 23, 24, 34) have suggested forms
-V(P) + p.(~M) . (v) +I (7-8)
similar to this, but have assumed that K' = K which If (v) = 0 implies I = 0 we can map (v) into I by the
requires that B be equivalent to the unit tensor. It expression
should be remembered that two assumptions were re- I = A . (v) (7-9)
quired to arrive at Equation 6-22.
and rewrite Equation 7-8 in the form
A unique time-dependent mapping M exists
)(V) (v) . V(v) ) = - V(P)
( Tt + +
which maps (v) into V A.S p R . (v) (7-10)
The time-dependent tensor K -1 has an inverse A.6
Further theoretical work and experimental studies are
Here we have written R = p.(~M) +
A where R is a
total resistance tensor. It will depend upon the struc-
certainly needed to confirm or reject the intuitive de- ture of the porous media, the local velocity, and perhaps
velopment outlined here. If assumptions A.S and A.6 velocity gradients, in addition to the viscosity and density
prove to be satisfactory, the principle of material frame- of the fluid. In general we would expect the compo-
indifference can again be invoked to put K and K' nents of R to be proportional to the viscosity at low
in the simplified form suggested by Equation S-28. Reynolds numbers and proportional to the density and
velocity at high Reynolds. This generalized type of
7. Inertial Effects
approach to the analysis of high Reynolds number flows
In considering inertial effects, we restrict our analysis does not appear to have been considered previously,
to incompressible flows so that the governing equations probably because such flows are usually one-dimensional
are and one need only determine a friction factor or drag

p(~: + v . vv) = -VP + p.~v (7-1)


coefficient to characterize the macroscopic flow field.

Conclusions
V· V = 0 (7-2) The development presented in this paper has as the
The volume averaged continuity equation quickly primary objective a careful listing of the assumptions
becomes and restrictions that must be imposed if the point equa-
tions describing steady, incompressible, creep flow in a
V • (v) = 0 (7-3) rigid porous media can be integrated over an averaging
while the volume averaged Navier-Stokes equations may volume to produce Darcy's law. The functions to be
be written as averaged must behave in a manner such that the follow-
ing restriction is satisfied
e~~) + (v . VV») = d«I«L

e - V(P) + ~V JA;
f PndA + p.(V v) 2
(7-4)
and the averaging must be performed in such a way that
The averaging volume is constant and main-
tains a fixed orientation relative to an inertial
We can define a new velocity Ii which represents the frame

VOL. 6 1 N O. 1 2 0 EC EM B E R 1 9 6 9 27
Here d, I, and L represent characteristic lengths for: a(i. k ) plane porosity for the (j, k) plane
p fluid density, g/cc
(1) the structure of the porous media, (2) the averaging <p gravitational potential function, cm2/sec 2
volume, and (3) the macroscopic process. With these <I> volume porosity (= (a»
restrictions placed on the averaging process, Darcy's '" any tensor valued function associated with the fluid
j" unit tangent vector to an arbitrary curve, unit orientation
law can be derived from the point equations on the vector
basis of two assumptions " kinematic viscosity, cm2 /sec
Il coefficient of viscosity, dyne sec/cm2
v is a unique function of (v) w = orientation vector
V(P) = 0 implies (v) = 0 o skew-symmetric orientation tensor
Ii = reference position vector, cm
On the basis of two additional assumptions
The anisotropic nature of a porous media can REFERENCES
be uniquely represented by a single orienta- (1) Aranow, R. H., "Statistical approach to flow through porous media," Phys.
tion vector, (,) Fluids, 9,1721 (1966).
(2) Aris, Rutherford, "Vectors, Tensors, and the Basic Equations of Fluids Me-
Darcy's law must satisfy the principle of ma- chanics," 1962, Prentice-Hall, Englewood Clifts, N. J.
terial indifference (3) Beran, M. J., H Statistical Continuum Theories," 1968, Interscience, New York,
N.Y.
one can prove that the permeability tensor is symmetric (4) Bird, R. B., Stewart, W. E., and Lightfoot, E. N., "Transport Phenomena,.'
1960, Wiley & Sons, Inc.,NewYork,N. Y.
and takes the form (5) Brenner, Howard, "Elements of transport processes in porous media,1I to be
published as a monograph by Springer-Verlag.
Kij = B(l)~ij + B(2)W/WJ (6) Brenner, Howard, "The Stokes resistance of an arbitrary particle," Chern. Engr.
Sci., 18, 1 (1963).
(7) Darcy, Henry, "Les Fontaines Publiques de la Ville de Dijon," 1856, Victor
The analysis of transient flow indicates that the relation- Dalmont, Paris.
ship between velocity, pressure gradient, and local (8) Ericksen, J. L.," Anisotropic fluids," Archive Rat. Mech. Anal., 4, 231 (1959-60).
(9) Ericksen, J. L., "General solutions in the hydrostatic theory of liquid crystals,"
acceleration is an extremely complex one, and further Trans. Soc. Rheol., 11, 5 (1967).
(10) Ericksen, J. L., "Tensor fields,1I "Handbuch cler Physik," 1960, Vol. III,
theoretical study is obviously needed. Pt. 1, SpringerwVerlag, Berlin.
(11) Ferrandon, J., Le Genie Civil, 125, 24 (1948).
Nomenclature (12) Green, A. E., and Zerna, W., "Theoretical Elasticity," 1963, Oxford Press.
(13) Halmos, P. R., "FinitewDimensional Vector Spaces," 2nd ed., 1958, D. van
Roman Letters Nostrand Co., Princeton, N. J.
(14) Hildebrand, F. B., "Methods of Applied Mathematics," 1952, Prentice-Hall,
surface of the averaging volume, cm~ Englewood Cliffs, N. J.
surface ofthe fluid volume contained within the averaging (15) Hubbert, M. K., "Darcy's law and the field equations of the flow of under ..
volume, cm~ ground fluids," A.I.M.E. Petrol. Trans., 207, 222 (1956).
(16) Kaplan, Wilfred, "Advanced Calculus," 1952, AddisonwWesley, Reading,
area of entrances and exits of A" cm~ Mass.
area of solid-fluid interface of AI, cm~ (17) Ladyzhenskraya, O. A., "The Mathematical Theory of Viscous Incompressible
= an averaging area parallel to the (j, k) plane, cm~ Flow," 1963, Gordon and Breach, New York, N. Y.
fluid area contained within A(j.kh cm~ (18) Landau, L. D., and Lifshitz, E. M., "Fluid Mechanics," 1959, Addison ..
characteristic length for the structure of the porous
media, cm
Wesley, Reading, Mass.
(19) Lapwood, E. R., PTOC. Camb. Phil. Soc., 44, 508 (1948).
(20) Massarani, GiuHo, "Criticism on the validity of the law of Darcy in hetero- •
a
g gravity vector, cm/sec~ geneous media," Compt. Rend., 265, 19, 587a (1967).
K permeability tensor, cm2 (21) McConnell, A. J., "Applications of Tensor Analysis," 1957, Dover Pub.,
New York, N. Y.
i. dimensionless permeability tensor (22) Noll, W.," On the continuity of the solid and fluid states," J. Rat. Meeh. Anal.,
K-I inverse of the permeability tensor, cm-2 4,3(1955).
(23) Orovy Ann, T. and Sil'vyann, E., Rev. Mecan. Appl., 5, 215 (1960).
k isotropic permeability, cm~
(24) Paria, G., Appl. Mech. Rev., 16,421 (1963).
Kij permeability tensor in index notation, cm2 (25) Polubarinova-Kochin P. Ya., "Theory of Ground Water Movement," 1960,
L characteristic length for the macroscopic process, cm Princeton University Press, Princeton, N. J.
[ characteristic length for the averaging volume, cm (26) Poreh, M. and C. Elata, "Analytical Derivation of Darcy's Law," Israel J.
Tech., 4, 214 (1966).
M the tensor mapping (v) into v
(27) Scheidegger, A. E., "Handbuch der Physik," 1960, Vol 3, Part I, Springer-
m a vector relating the average pressure to the average Verlag, Berlin.
velocity, cm-I (28) Scheidegger, A. E., "The Physics of Flow Through Porous Media," 1960,
n outwordly directed unit normal for V and VI Macmillan,NewYork,N. Y.
p pressure, dyne/cm2 (29) Scheidegger, A. E., "Statistical theory of flow through porous media," Trans.
Soc. Rheol., 9, 313 (1965).
po reference pressure, dyne/cm2 (30) Slattery, J. C.," Flow of viscoelastic fluids through porous media," A.I.Ch.E. J.,
p position vector relative to the centroid of the averaging 13,1066 (1967).
volume, cm (31) Slattery, J. C., "Single-phase flow through porous media," in press, A.I.Ch.E.
J.
P piezometric pressure, dyne/cm2 (32) Taylor, A. E., "Advanced Calculus," 1955, Ginn and Co., New York, N. Y.
T position vector, cm (33) Truesdell, C., "The Elements of Continuum Mechanics," 1966, Springer...
To position vector locating the centroid of the averaging Verlag, New York, N. Y.
volume, cm (34) Truesdell, C., and Noll, W., "The nonwlinear field theories of mechanics,'-
"Handbuch der Physik," 1965, Vol. 3, Pt. 3, Springer-Verlag, Berlin.
position vector locating the centroid of the fluid volume (35) Truesdell, C., and Toupin, R., "The classical field theories," ibid., 1960, Vol.
contained within the averaging volume, cm 3, Pt. 1.
R resistance tensor, dyne-sec/cm4 (36) Whitaker, S., "Diffusion and dispersion in porous media," A.I.Ch.E. J., 13,
420 (1967).
s arc length, cm
(37) Whitaker, S., Document 9234 deposited with the American Documentation
S surface area of a macroscopic region, cm~ Institute, Photociuplication Service, Library of Congress, Washington 25, D. C.
U unit tensor (38) Whitaker, S., "The equations of motion in porous media," Chern. Engr. Sci.,
21, 291 (1966).
u constant porous media velocity, cm/sec.
(39) Whitaker, S., "Introduction to Fluid Mechanics," 1968, Prentice-Hall, Engle-
V averaging volume, cc wood Cliffs, N. J.
VI fluid volume contained within the averaging volume, cc
V velocity vector, cm/sec
(v) = volume averaged velocity vector, cm/sec
Xi position vector for rectangular Cartesian coordinates in
index notation, cm
Greek Letters
a(X) = void volume distribution function
(a) = volume porosity

28 INDUSTRIAL AND ENGINEERING CHEMISTRY

View publication stats

Вам также может понравиться