Permeability of Idealized Fractured Rock

R. W. PARSONS MARA THON OIL CO.

JUNIOR MEMBER AIME LITTLETON, COLO.

the plane of interest at right angles with the

ABSTRACT
fractures. This concept represents the horizontal
The over-all apparent single-phase permeability flow in a reservoir system with the fractures
of fracture-rock systems was studied using two oriented vertically. Only single-phase, laminar
different two-dimensional models. In a strict flow is considered. The regular fracture-rock system
sense the results are applicable only to these permeability is calculable from a simple analytical
models, yet may give some insight into real expression. The heterogeneous system requires
reservoir behavior. a computer program for calculating overall pennea-
The regular fracture-matrix model can have any bility .
number of regular fracture sets. All fractures with-
in each set have the same orientation, width and FLOW IN A SINGLE FRACTIJRE
spacing. The gross behavior is analytically
Fluid flow in a single fracture is assumed
equivalent to that of an anisotropic permeable
equal to laminar flow between two infinite parallel
medium.
smooth plates. This is expressed as: 1
In the heterogeneous fracture systems, individual
fracture conductivities vary within a two-dinensional
network pattern. A computer program calculates w2 dp
(1)
the overall permeability after the conductivities v
12 fl dL
are randomly placed in the network. The calculated
permeability depends upon the distribution of these
conductivities and the particular fracture pattern dp
into which they are placed. q _ . . . . (2)
Although detailed information on actual reservoir dL
fracture systems is not available, this study
shows how some of the possible variables may Turbulent flow in fractured reservoirs is quite
affect observed gross reservoir characteristics. unlikely. As with flow in circular pipes, a Reynolds
number can be defined to characterize the flow
INTRODUCTION regime (2Wvp/ fl). Values of this dimensionless
group at the" onset of the transition to turbulent
Characteristics of reservoir rocks are determined flow vary from 1,800 to 4,000, depending on the
by direct (core analysis) or indirect (logging and reference. 1 - 3 Surface roughness on the fracture
production tests) examination. Both methods, faces lowers this number. 2-4
singly and together, are widely used in reservoir The following example shows the order of magni-
evaluations. For fractured reservoirs a problem tude of flows necessary to get into the turbulent
exists in relating characteristics of cores to the in region. Assume a rough fracture with a Reynolds
situ reservoir properties. Knowing fracture charac- number of 200 as the cutoff point for laminar flow.
teristics can be of value in determining production The corresponding volumetric flow rate per foot of
mechanisms. Because of the limitations of observa- fracture height perpendicular to the flow direction
tion, the intimate details of a fractured rock system IS independent of fracture width and will be:
will never be known. One approach in analyzing
fractures is to study the behavior of some con-
ceptually simple models.
q bbl/day
The two models analyzed contain features which h of fracture
may be realistic, yet are mathematically tractable.
Flow is considered in two dimensions only, with

Original manuscript received in Society of Petroleum

Engineers office Aug. 2, 1965. Revised manuscript of SPE
16.6 .
p (g~:c) (3)

1289 received March 10, 1966. Paper was presented at SPE

Annual Fall Meeting held in Denver, Colo., Oct. 3-6, 1965. lReferences given at end of paper.

126 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 REGULAR FRACTURE-MATRIX SYSTEM across a fracrure-matrix surface does not alter
the flow behavior of either system. A marked
The regular fracrure-matrix system is taken to
similarity between Eqs. 4 and 5 is noted. In
be a homogeneous permeable rock laced with a particular, the flow per unit cross-sectional area
number of sets of vertical fracrures. All fracrures
perpendicular to the over-all pressure gradient is
in a given set have the same orientation, width directly proportional to the pressure gradient. This
and spacing. There is no restriction to flow of shows that the pressure fields in the two systems
fluid between the fracrure and the permeable matrix
are identical, and in this situation the superposition
blocks.
of the two systems will cause the flow in each to
In developing the permeability expression, we
beha ve as if the other were not presen t.
will consider a slab of rock removed from a reservoir
The following equa tion define s an over-all
so the slab faces are vertical and perpendicular to
permeability for the system in the direction of the
the pressure gradient. Flow is considered in a
pressure gradient by analogy to Darcy's law:
cross-section of this slab for three separate
cases.
1. Homogeneous permeable medium (Fig. IA):
the flow per unit area perpendicular to the pressure
kfr . . . . . . (6)
gradient for this case is given by Darcy's equation: =

Substiruting Eqs. 4 and 5 into 6 gIves:

. . . . . . . (4)

2. Impermeable medium with a regula. fracture

set (Fig. IB): Every fracture has the same orientation . . . . (7)
and width, and there is a constant spacing between
fractures. The average flow per unit area perpen-
dicular to the pressure gradient comes from the
equa tion for flow between two parallel plates: When other fracrure sets are added, an additional
assumption is that the effect of deviations from
WA 3 cos 2 ex dp the normal flow patterns occurring at fracrure
. . (5) juctions is negligible. The generalized regular
121L A dL fracture-matrix permeability equation is:

The cos 2 a term comes from the component of

dp/ dL along the fracture direction and from the
number of fracrures per unit area of the face. kfr = kr + a cos 2 ex
3. Composite of Cases 1 and 2 (Fig. IC): this
is a regularly fracrured, homogeneous permeable + b cos 2 f3 + . . . . . (8)
matrix. It is assumed that fracture widths Ware
small in comparison with spacing A so the permeable where (maximum permeability of individual fracrure
medium cross-sectional area is not changed from sets):
Case 1. It is also assumed that the fluid flowing

dp/dL

b W 3 /12B . . . . . . . . . . (9)
B

TT
TT a 5.4-4-6 x 10 10 WA 3/ A if dimensions are in

1
inches and permea-

1 7 3
bility in millidarcys.

8.4-4- x 10 WA /A if dimensions are in

A. B. C. millimeters and per-
PERMEABLE REGULAR COMPOSITE meability in milli-
MATRIX FRACTURES OF darcys.
ONLY ONLY A and B
FIG. 1 - SLABS FROM IDEAL RESERVOIRS.
. (10)

JUNE, 1966 127

This over-all permeability is a function of the
direction of measurement (direction of pressure h WI 3
gradient) through the angles a, {3, . . . For a qfi = ~p i . . . . . . (12)
given direction, the quantity of fluid flowing 12 J1 L i
through each system will be proportional to each
term of Eq. 8. . (flow)(conductance)(potential)
This is equivalent to the permeability of a
homogeneous, anisotropic medium. The proof of The distributions of the conductance values
this is presented in Appendix A along with a are given in Appendix B along with various statistical
scheme for determining maximum and minimum properties of these distributions. To compare
permeability axes and their orientations from results, all the distributions have been adjusted
the properties of the fracture-matrix sy stem. This to median values of 1.0. Calculations with the
will allow direct application of published results first four distributions in the square network are
concerning flow in anisotropic media to this original wi th this work. The remaining distributions
particular ideal model of a fractured reservoir. and results are from the work of others. 5 6
A later section of this paper reviews some of the For a given calculation discrete conductance
literature on anisotropic media. values from one distribution were placed in a
For the three-dimensional case, the third given network pattern in some random order.
major permeability axis is vertical, the value Boundary conditions were imposed to give quasi-
being: linear flow. For a given random placement of
conductances, separate calcula tions were made
with the over-all pressure drop being parallel to
kf r k r + a + b .... . . . . . . . (11)
the two main directions in the square fracture
pattern (x and y directions) and being oriented
where a, band c have the same significance as 45 to the x and y directions. In setting up the 45
Eq. 9. A comparison shows that the vertical boundaries, half the original fractures are excluded
permeability will always be greater than the from the pattern. To ensure the inclusion of all
horizontal one. values, the original random conductances are
placed only in one-quarter of the fracture pattern.
HETEROGENEOUS FRACTURE SYSTEM The other three quarters are made duplicates of
the first.
The heterogeneous fracture model has varying If an anisotropy was desired, all the fracture
hydraulic conductance fractures placed in a given conductances oriented in one direction were
pattern in an impermeable medium. Values from a doubled. For the anisotropic, 45 case the no-flow
given distribution of conductances are randomly boundaries were modified. Each point on the no-
placed in the pattern (Fig. 2) and the over-all flow boundary was replaced by a constant potential
permeability computed. The problem is equivalent representing either a linear gradient between the
to determining the over-all conductance (resistance) normal constant potential boundaries or a gradient
of a heterogeneous network of resistors. established by averaging the potentials across
The relationship between the conductance all planes perpendicular to the flow direction. This
values used in the network and the physical way, a truer quasi-linear flow was obtained, and
properties of the individual fractures can be obtained the calculated over-all permeability was equivalent
by applying Eq. 1 to the fracture between two node to that measttred in the direction of the pressure
points. drop.
Table 1 shows the different combinations of the
variables included in this study along with the
calculated permeability. Where several different
random sequences were used for placing con-
ductances with other variables unchanged, an
average and standard deviation of the calculated
over-all permeabilities are given.
Potentials, or pressures, at all node points in
the network pattern were determined on the digital
computer using an alternating direction relaxation
technique. The total flow in the fracture system
was obtained by summation of the flows over a plane
perpendicular to the over-all pressure gradient.
This total flow was put into Darcy's law to cal-
SQUARE TRIPLE HEXAGONAL culate an over-all apparent permeability in the
TWO DIMENTIONAL TWO DIMENTIONAL direction of the pressure gradient.

FIG. 2 - HETEROGENEOUS FRACTURE MODELS.

128 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 CEq. 12).
L q{1 ) Perhaps the overall permeability might correlate
kp ' ( hT L\p ... . (13)
with some statistical property of the conductance
overall values distribution, as has been found previously for
This is independent of viscosity p., height h cubic networks? Fig. 3 shows a plot of these
and pressure drop L\p since these values al"so variables for the isotropic situations. Many of
enter into the determination of the over-all flow the data points are from single calculations and
q from the summation of individual fracture flows may be misleading becau se of permeability variations

TABLE 1 -
PROPERTIES OF HETEROGENEOUS FRACTURE SYSTEMS
Direction Number of Standard
Conductance Network Network of Random Over-All Deviation
Distribution Element Size Pressure Directional Condo Permeabil ity of Kp
Number (Appendix B) Geometry (X) (Y) Gradient Properties Placements K;(ave.) (a) Remarks

Square

2
83
Same
20

20
20

20
X None

None
10

10
0.772

0.801
0.033

0.036
3 Same 20 20 None 10 0.801 0.028 Combined data for 1, 2 and
3 gives K;
= 0.791
a = 0.03~
4 Same 20 20 X ex doubled 10 1.240 0.060
5 Same 20 20 Y ex doubled 10 0.832 0.032
6 Same 20 20 45 ex doubled 10 1.084 0.032
7 II Same 20 20 X None 10 0.945 0.016
8 II Same 20 20 Y None 10 0.958 0.015
9 II Same 20 20 45 None 10 0.954 0.006 Combined data for 7, 8 and
9 gives Kp' = 0.952
a=0.014
10 II Same 20 20 X ex doubled 10 1.578 0.027
11 II Same 20 20 Y ex doubled 10 0.966 0.015
12 II Same 20 20 45 ex doubled 10 1.303 0.013
13 III Same 20 20 X None 10 0.713 0.137
14 III Same 20 20 Y None 20 0.688 0.136
15 III Same 20 20 45 None 20 0.844 0.151 Combined data for 13, 14
and 15 gives K;= 0.755
a = 0.150
16 III Same 20 20 ex doubled 10 1.203 0.242
17 III Same 20 20 ex doubled 10 0.766 0.171
18 III Same 20 20 ex doubled 23 1.207 0.145
19 IV Same 20 20 None 10 0.908 0.082
20 IV Same 20 20 None 10 0.909 0.043
21 IV Same 20 20 None 10 0.919 0.033 Combined data for 19, 20
and 21 gives K'; = 0.912
a = 0.057
22 IV Same 20 20 X ex doubled 10 1.635 0.076
23 IV Same 20 20 Y ex doubled 10 0.966 0.040
24 IV Same 20 20 45 ex doubled 10 1.340 0.047
25* V Same 11 80 X None 0.854
26* VI X None 0.976
27* VII Same 11 80 X None 14.06
28* V Square but 11 80 X None 0.244
with 1/3 of
I ink s random-
ly removed
29* VI Same 11 80 X None 0.108
Triple

30* V H)J5(' II 79 X None 1.536

31* VI Same 11 79 X None 2.521

32** VIII Same 11 11 X None 0.758
33* V Triple hex- 11 79 X None 0.644
agonal but
with 1/3 of
Iinks random-
ly removed
34* VI Same 11 79 X None 0.641
* Data is from unpubl ished work of D. H. Thurnau. 6
"*Data is from Fatt. S

JUNE. 1966 129

connected with different random conductance Also plotted are points representing the average
placements. From this plot and from examination of the calculated kx and kyo From Fig. 4, the
of the data in Table 1, only some very generalized following trends are noted. These are tentative
trends appear. because of the limited data involved; for example,
1. Con sider only the properties of the con- only the square fracture pattern.
ductivity distributions (Appendix B). As the , 1. The anisotropic permeability in the y direction
standard deviation increases from "Zero (as the increases somewhat over the isotropic case. For
distributions become more disperse), the various quasi-linear flow in the y direction, there is some
means depart further from 1.0, the arithmetic fluid flowing in individual x direction fractures.
mean increasing and the geometric and harmonic Increasing these x conductances would, therefore,
means decreasing. A particular distribution, tend to increase the total net flow, thus increasing
however, is not uniquely defined by these means. the y permeability.
2. The larger the standard deviation of the 2. The anisotropic permeability in the x direction
conductance distribution, the larger is the spread is less than double the isotropic permeability. A
(standard deviation) of permeabilities calculated qualitative argument similar to that presented in
for different random orderings. The spreads were 1 above can be applied here also. It is just as
approximately normal. valid, but easier to visualize the effect on the
3. For conductance distributions which are not x permeability if the y conductances are halved.
too disperse, say with standard deviations of 1.0 The total flow is decreased, reducing the x permea-
or less, the geometric mean value will be a fair bili ty.
approximation to the permeability of square network 3. As an approximation for the 45 permeability,
patterns. If there are trends for other patterns or the homogeneous formula, kp (45') = (k x + ky)/2
for more disperse distributions, there are not (Eq. 14), can be used with kx and ky being the
enough data here to show them. actual measured (calculated) values. In other
If a homogeneous, isotropic medium has Its words, the anisotropic heterogeneous fracture
permeability in the x direction doubled, then the system is roughly analogous to an anisotropic
permeability of this anisotropic medium can be homogeneous system. The one case where the
found from Eq. 14. For the x, y and 45 directions, approximation breaks down is for a high standard
this relation ship is shown as the solid lines in deviation of the conductance distribution (Distribu-
Fig. 4. For the heterogeneous fracture system, tion III, of Table 2). Whether this was due to the
values of the calculated permeability wi th no limited network size or not was not checked.
directional properties (called the isotropic case)
are plotted vs the anisotropic permeabilities after MA TRIX FLOW IN
the x direction conductances have all been doubled. HETEROGENEOUS FRACTURE SYSTEM

4.0 ,. ,. A brief study was made of the effect of allowing

the rock matrix in the heterogeneous fracture
system to have a finite permeability. A single
3.5 matrix block was represented by a two-dimensional,
o Arithmetic "ton homogeneous square. Fluid flow into or out of
A GeometriC Meon
D Harmonic Mean
a particular matrix block was assumed insignificant
3.0
Ii " Median in comparison with the flow in the fracture system.
;; o A D V Square fracture pattern
+4'9 t 1/3 In other word;;, the pressure distribution along the
J Squa.. wotll
*-69-V' Triple hexaoonol
of hn.........d

j JlTpSJ'rr Triple Mlavonal With 1/3 of links removed matrix block boundaries was determined by the
2.5
.!O! fracture flow only, with constant, but in general
~
E not equal, pressure gradients along each side.
Ii
:z:
2.0
The single-phase fluid flow characteristics

"
;
E
0

~ l.O~-+--+--+--+----;,f----+--f---:;,--+--+-~

!-
1.5
"

.;:
c ~ . !T'iP
0
loE2
All fracturll thl ICImI
Square pottern
U
a::
o
,8~-+---+---+----;~-~~~-+- __~-+-~
1.0 0::
~ 1--1-----1-
~ ~?
,6
o

.5
4
It>

\t>
(/)

;!I!J.
.'

.0 Q - 0 - ..... ;I'-6>-9- >- ,2 LO 1.2 1.4 1.6 L8 2.0
o .5 1.0 1.5 2.0 2.5

O.erall permeability k;' kp (ANISOTROPIC)

FIG. 3 - HETEROGENEOUS MODEL PERMEABILITY FIG. 4 - RELATION BETWEEN ISOlROPIC AND

VS PROPERTIES OF FRACTURE CONDUCTANCE ANISOTROPIC PERMEABILITIES IN HETERO-
DISTRIBUTION. GENEOUS FRACTURE SYSTEM.

130 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 TABLE 2 - PROPERTIES OF DISTRIBUTIONS OF CONDUCTANCES
Standard
Frequency Function Arithmetic Harmonic Deviation
Distribution f(c) Median Mean Geometric Mean Mean (a)

I. Rectangular 0.5 1.0 1.0 (1.0)" 0.736 (0.737)" 0 (0.275)" 0.577 (0.577)"
II. Normal
0.3yZii
1 8 C-Q.3YJ
exp - -
1
1.0 (1.0)" (0.949)" (0.886)" 0.300 (0.298)"

III. Rectangular in
widths (conduc- 1/6 L'2/3 1.0 2.0 (2.0)" 0.399 (0.400)" (0)" 2.27 (2.27)"
tances l/3 )
IV. Normal in widths 1.0 (1.267)" (0.855)" (0.362)" (0.036)"
(C 1/3)
V. Right triangular -0.1716C + 0.5858 1.0 1.138 0.762 0 0.804
VI. Right triangular in -0.0572C-1/3 + 0.1953C- 2/ 3 1.0 3.98 0.442 0 6.38
(cl/3)
VII. Right triangular in -0.0343C- 3/5 + 0.1172C- 4/5 1.0 22.07 0.256 0 52.6
(C 1/ 5)
VIII. Fatt (permeabi lity) 1.0 (2.887)" (0.6913) (0.0273)" (6.00)"
Val ues are calcul ated from the di screte di stributions; others are calcul ated from the continuous frequency functions.

for this model were analyzed using a digital was by far the most common situation found in the
computer. The rock was represented by a homo- previous heterogeneous fracture calculations. For
geneous conductance network. A 20 x 20 network this case, along a given face, the flow is either
was adequate. The boundary conditions were set all into or all out of the matrix block. In Figs. 5C
up using various combinations of pressures at the and 5D, the maximum and minimum pressures
four corners of the square, then considering the occur on adjacent corners. For this situation,
pressure distribution along a particular face of flow can be into the matrix block in one section
the rock to be linear between the two corners. and out of the matrix block in another section of
although the detailed results of these calculations the same face. This would occur at the upper and
are not especially useful for the quantitative lower faces of Fig. 5C. It would occur on all four
evaluation of flow in real systems, some of the faces of Fig. 5D.
qualitative observations can give an insight into Actual flow rates through potential patterns
what may happen. Four typical examples of the such as these will be determined by the length of
resulting potential maps are shown in Fig. 5. the stream lines and the potential gradients along
In general, the equipotentials (and their orthogonal these lines. Thus, there will be a spectrum of flow
stream lines, which are not shown) are curves rates through the various portions of a matrix
which are not fit by simple equations. block.
In Figs. 5A and 5B, the maximum and minimum The actual volumetric flows into or out of a
pressures are located on opposite corners. This given face are shown in Figs. 6 and 7 as a

.8
0 0

.7

.5
0::
0
a A .5 ;;::
0
-'
lL
.4 ~
f--
w
.3 z
-'

f--
0
.2 f--

.2 ,3.4.6 .9
Net Flow Crossing one Boundary(q//kh6p) 0 B

C AD,
FIG. 5 - EXAMPLES OF ISOPOTENTIALS IN MATRIX FIG. 6 - FLOW CROSSING MATRIX BLOCKS
BLOCKS. BOUNDARIES.

JUNE, 1966 131

 function of the pressures at the corners of the
square. The flows given are in dimensionless
units equal to qll/kht!p. Values for real flow rates
can be found from this expression. This corresponds to measuring the permeability
For the regular fracture-matrix system, the rock across an infinite slab of rock. If the permeability
permeability could be added to the fracture permea- .is measured in the direction of the velocity vector
bility to give the composite value (Eq. 8). For a or the stream line, then the resulting equation is:
heterogeneous fracture-matrix system, this cannot
be done. Cursory calculations show that the over-
all permeability for this system will be the fracture + . . . . . . (15)
permeability plus some fraction of the rock permea-
bility. Apparently this fraction will always be less
than one and dependent upon the distribution of This is equal to the permeability measured on a
fracture conductance and the fracture pattern. very long, thin piece of material with the flow
forced by the boundary to be in the direction that
REVIEW OF FLOW k is measured.
IN ANISOTROPIC MEDIA Perhaps the most convenient analytical tool
available for dealing with anisotropic media
Because of the similarity in the flows in fractured involves a coordinate transformation. The result
and ani sotropic media, a brief mention of some of is an isotropic medium with the original boundaries
the literature references on this subj ect will be being stretched or compressed along the directions
useful. There are several books which discuss of the principal permeability axis. This transforma-
fluid flow in anisotropic media, mainly steady- tion has been used to solve several problems.
state situations, to vatying degrees. 8, 12 The Marcus 13 has analytically solved for the apparent
fluid flow is complicated by the fact that the permeability of a rectangular, two-dimensional
pressure gradient and velocity vector are not anisotropic sample. This is a function of the
oriented in the same direction, except along the orientation and magnitude of the principal permea-
major and minor permeability axes. In an unbounded bility axes and the length to width ratio of the
anisotropic medium there are two limiting expressions sample.
for the permeability .1 2 ,13 If the permeability is Arnold et al,14 have outlined a procedure for
measured in direction of the pressure gradient, the determining anisotropy from steady-state reservoir
result is: tests. Their technique relies on the matching of

1.0

Cf? 9
m

Note: Straight lines represent net flow. "

m
---~.8
o
Deviation from straight lines is gross flow
m

_.-+-- 6

--- ~

7
.4

A .5
.2
4

.3 - - - -

05~---4~--+---~--~~~O---*--~--~--~~~--~~~~~~~~-7.;--------O~D--~I~
Net Flow

A B
FIG. 7 - FLOW CROSSING MATRIX BLOCKS BOUNDARIES.
132 SOCIETY OF PETROLEUM ENGINEERS JOURNAL
pressures ovserved at one producing and three A, B ... fracture spacing, perpendicular
shut-in observation wells with theoretical steady- distance between fractures for
state equipotentials. The steady-state equipo- sets A, B, ...
tentials from a point source are a series of ellipses
A' B' or C: ... constants for a given regularly
of the same eccentricity.
The effect of anisotropy on areal sweep efficiency fractured homogeneous rock
and flow capacity has been examined by Mortada (defined by Eqs. 22 through
and Nabor 15 and Landrum and Crawford. 16 Both 24, Appendix B)
of these studies use the steady-state case with C conductance of a single fracture
fluid mobilities equal ahead of and behind the (Eq. 12)
front. h vertical height (perpendicular to
Assuming the pressure equilization between pressure gradient)
fracture and matrix at a given point is instantaneous, permeability of fracture - rock
than some unsteady-state results from anisotropic system (in direction of pressure
media can be applied directly to a fracture-matrix gradient)
system. The usual pressure draw down equations permeability of anisotropic
are found to apply, except that distances are medium (in direction of pressure
changed in compliance with the coordinate trans- gradient)
formation and the permeability is replaced by the
apparent permeability of hetero-
geometric mean of the maximum and minimum
geneous fracture system (in
anisotropic permeabilities. 17,18 Thus, routine
direction of pressure gradient)
analysis of simple pressure drawdown tests yields
a permeability value of ykxky. If other pressure k, rock permeability of homogeneous
observation points are available, values of k:x;,ky rock matrix
and orientation can be selected to give the best permeability of anisotropic
match of field and calculated data. medium (in direction of stream
line)
CONCLUSIONS == maximum and minimum permea-
bilities in anisotropic medium
The conclusions presented here are applicable
in a strict sense only to the particular models L length
used. Although there are field data which indicate P pressure
that a natural fractured system will behave like v superficial velocity in porous
an anisotropic medium,18 there is no complete medium or average velocity
way of physically comparing the natural fractured in fracture
systems with the particular models proposed. q == volumetric flow rate
Keeping the limitations in mind, the conclusions
q! == volumetric flow rate (in fractures)
of this study are:
1. The gross single-phase fluid flow behavior q, volumetric flow rate (in rock
matrix)
in a naturally fractured porous rock is equivalent
to that of an anisotropic permeable medium. As T horizontal width (perpendicular
with unfractured reservoirs, the possibility exists to pressure gradient)
of larger-scale heterogeneities overshadowing or W fracture width
modifying the flow effects due to anisotropy.19 x, y coordinate axes corresponding to
2. The regular fracture-matrix flow model is directions of maximum and
exactly equivalent to an anisotropic medium. The minimum permeabilities (also
characteristics of this medium are calculable from used as subscripts for prop-
the rock and fracture properties. erties measured In these
3. For the heterogeneous. fracture system, the directions)
overall permeability is a function of both the a, {3, .. == angle between fractures, A, B, ...
distribution of the fracture conductances and the and over-all pressure gradient
fracture pattern. For distributions which are not
0, f, angle between fracture set A
too disperse (grouped closely around some median
and sets B, C, ...
value) and for the square fracture pattern, the
geometric mean of the distribution is a fair e angle between maximum permea-
approximation to the over-all permeability. For bility direction and direction
the anisotropic cases, the permeabilities can be of pressure gradient
approximated by an equivalent anisotropic homo- fJ. == viscosity
geneous medium. This again appears to hold only p == density
for distributions not too disperse. == angle between maximum permea-
bility direction and velocity
NOMENCLATURE vector
a, b ... == fracture set permeabilities

JUNE, 1966 133

 ljJ == angle between maximum permea-
APPENDIX A
bility axis and fracture set A
(+ counterclockwise)
/l,.p over-all pressure gradient (== dp/ EQUIVALENCE OF REGULAR
dL) FRACTURE-MA TRIX AND ANISOTROPIC
SYSTEMS
ACKNOWLEDGMENTS In a homogeneous, isotropic medium, permea-
The computer program for determining the potential bility is a proportionality constant between the
distribution in conductance network was furnished pressure gradient and velocity vector, the two
by D. H. Thurnau. always being parallel. For a homogeneous, ani-
sotropic medium (one whose properties vary wi th
REFERENCES
direction, but not with distance in a given direction)
this is not the case. There exists instead a set
1. Huitt, J. L.: "Fluid Flow in Simulated Fractures", of perpendicular axes, along which velocity and
A/ehE Jour. (1956) Vol. 2, 259. pressure gradient are parallel. These are the
2. Baker, W. J.: "Flow in Fissured Formations", Proc., maximum and minimum permeability axes. In all
Fourth World Petroleum Congress, Section II (1955)
379.
other directions, velocity and pressure gradient
3. Townsend, A. A.: "Turbulence", in Handbook of
will not be parallel. Instead, by analogy to Darcy's
Fluid Dynamics, Victor L. Streeter, Editor, McGraw- law and by experimental evidence the following
Hill Book Co., New York, N.Y. (1961) 10. equations hold:
4. Parrish, D. R.: "Fluid Flow in Rough Fractures",
SPE Production Research Symposium, The U. of
Oklahoma (April 29-30, 1963).
5. Fa tt, I.: "The Network Model of Porous Media, III. (A-I)
Dynamic Properties of Networks with Tube Radius
Distribution", Trans., AIME (1956) Vol. 207, 164.
6. Thurnau, D. H.: private communication.
7. Warren, J. E. and Price, H. S.: "Flow in Hetero-
geneous Porous Media", Soc. Pet. Eng. Jour.
The x and y axes correspond to the maximum and
(1961) 153. minimum permeability axes. The general relation-
8. Collins, R. E.: Flow of Fluids Through Porous ship between the pressure gradient, velocity vector
Materials, Reinhold Publishing Corp., New York, and the x and y axes can be seen from Fig. 8.
N. Y. (1961). The permeability in the direction of the pressure
9. Maasland, M.: "Soil Anisotropy and Land Drainage", gradient is that calculated considering flow across
in Drainage of Agricultural Lands, edited by James
N. Luthin, American Soc. of Agronomy (1957).
a plane perpendicular to the pressure gradient.
From Darcy's law:
10. Muskat, M.: Physical Principles of Oil Production,
McGraw-Hill Book Co., New York, N.Y. (1949).
11. Polubarinova - Kochina, P. Ya.: Theory of Ground- f.1 (component of -: i n\l p direct i on)
water Movement. translated from Russian by J. M.
Roger de Wiest, Princeton U. Press (1962). \lp
12. Scheidegger, A. E.: ThA physics of Flow through (A-2)
Porous Media, The MacMillan Co., New York, N.Y.
(1957).
From Fig. ~8, the following relations hold:
13. Marcus, H.: "The Permeability of a Sample of an
Anisotropic Porous Medium", Jour. Geophys. Res.
(Dec., 1962) Vol. 67, 5215.
14. Arnold, M. D., Gonzales, H. T. and Crawford, P.
B.: "Estimation of Reservoir Anisotropy from
y
Production Data", Jour. Pet. Tech. (June, 1962)
909.
15. Mortada, M. and Nabor, G. W.: "An Approximate
Method for Determining Areal Sweep Efficiency and
Flow Capacity in Formations with Anisotropic
Permeability", Soc. Pet. Eng. Jour. (1961) 277.
16. LandrUlT', B. L. and Crawford, P. B.: "Effect of
Directional Permeability on Sweep Efficiency and
Production Capacity", Trans., AIME (1960) Vol.
219, 67.
17. Jones, P.: "Reservoir Limit Test on Gas Wells",
Jour. Pet. Tech. (June 1962) 613.
18. Elkin, L. F. and Skov, A. M.: "Determination of
Fracture Orientation from Pressure Interference",
Trans .. AIME (1960) Vol. 219, 301.
~------~--~-----------------x
19. Greenkorn, R. A., Johnson, C. R. and Shallenberger,
L. K.: "Directional Permeability of Heterogeneous
Anisotropic Porous Media", Soc. Pet. Eng. Jour. FIG. 8 -- PRESSURE GRADIENT AND FLOW
(1964) 124. DIRECTION IN ANISOTROPIC MED IA.

134 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 -+
component of v in\7p direction = v cos (e - )
sin = Vy/v

ap/ax = \7p cos e ap/ay = \7p si n e

+ C Sin
2 (] + 2 sin a cos a
(A-3)

Combining Eqs. A-I, A-2 and A-3 gives:

[b sin 0 cos 0 + C sin ( cos (]

(A-4) (A-6)

The bracketed expressions In Eq. A-6 are

Eqs. Sand A-4 express the permeability in the constants dependent upon the physical charac-
direction of the pressure gradient, one for the teristics of the fracture-rock system. A unique
fracture-matrix sy stem and the other for an aniso-
set of variables, kx' ky and l/J, is sought to make
tropic system. The following derivation shows that corresponding bracketed expressions of Eqs. A-S
these equations are uniquely identical: i.e., for and A-6 equal. Equating these expressions gIves:
gross flow considerations, a regular fracture-matrix
system can be replaced by an anisotropic medium.
Fig. 9 defines the various angles involved in the
derivation.
When the relationship = a - l/J is substituted e
into Eq. A-4 the trigonometric expansion performed + b COS 2 0 + c COS 2 ( _ A' . . (A-7)
and the resulting expression expanded and re-
arranged, the following equation results:
. 2
kx sin l/J + kY cos 2 l/J =
kr

+ b sin 2 0 + C sin2 ( _ B' . . (A-S)

+ 2 sin a cos a [(kx - ky) sin l/J cos l/J] (kx - ky) sin l/J cos l/J = b sin 0 cos 0

(A-S)
+csin( cos ( -C' (A-9)

If Eq. S were considered for this three-fracture

set example and if the substitutions f3 = a - 0 and Solving explicity for kx and ky gives the following
y = a - ( are made followed by expansion and expr-essions:
rearrangement, the following is found:

All angles are in

the half-plane A' + B' + (A' - B,)2 . (A-lO)
90 from fracture kX =
set A.
Angles are positive
in the counter-
clockwise direction
\
A' + B' - (A' - B' )2 .
\
, ky =
(A-II)
\
\
\
\
\ Although l/J can also be found explicity, it is
\
\ more convenient to consider the following expression:
~~~-,--~ __L -______
\
\
\
C' sin l/J cos l/J . . . . . (A-I2)
\ ,
\
A' - B' 2
cos l/J - s i n2 l/J
FIG. 9 - RELA TIONSHIP OF ANGLES IN ANISOTROPIC
AND REGULAR FRACTURED MEDIA:. l/J is not a single valued function of C'/A ' - B:

JUNE, 1966 135

 5.0 :II

f'b
4.8
-- 4 I.~

A
4 .4 1.0
4.2
.~
4o
"'0 0
38 1.0 2.0 .~ 1.0 I.~ 2.0
Conductivity.
.9 36
3 ,4
.8
/ 3.2 3.0 ill :nz::
/ 30
,J-"
'1..1.: .7
/ 2.8 2.0 .1

-- / 2.6 .4
If J 2.4
1.0
/ J 22
.2

/ I 2.0 0
4.0 8.0 0
4 6
/ / I.
/ I.
r-7 f-

.~ l~~
I.4 .10
/ / I.
D
+o / / o
I. .4

2
/ .O~
o I 2
- / /" .2
;/
/
J /
o I 2 3 4 0
20 40

0/
o 10 15 30 35 40 4

~L
.10

FIG. 10 - ANGLE FOR ANISOTROPY ORIENTATION.

Examination of this relationship and Eq. A-9 (C'

and sin tjJ cos tjJ must have same algebraic sign)
"~
.OO~ 0 ~ 10
.O~
o I 2 3

leads to Table 3 for determining the true tjJ value. o 200 400 0

tjJ'is given in Fig. 10.

FIG. 11 - FRACTURE CONDUCTANCE (PERMEA-
The derivation of the equations for kx' ky and BILITY) DISTRIBUTIONS.
tjJ leads to two sets of solutions. In the second
set, kx and ky (Eqs. A-10 and A-H) are reversed TABLE 3_0RIENTATION OF MAXIMUM PERMEABILITY
and the angle tjJ (measured from kxaxis) is displaced if C" is
by 90 so that the results are identical.
QndA"-B~is +
In summary, then, to use this system the follow~
ing steps are performed: then 'I' is in the range _90 0 to _45 0 _45 0 to 0 0 to 45 45 to 90

1. Select one fracture set as the base set A. and the true 'P = - qt'
2. Calculate a, b, c, .,. (Eq. 10).
3. Calculate A:
B' and C' (Eqs. A-7 through relative frequency of occurrence of the vanous
A-9) remembering angles 0, f, . , . are measured conductances. For example, in Fig. 11-1 all
from base fracture set (corresponding to a) to other conductances from 0 to 2 occur with equal frequency,
fracture sets so that the angles are between -90 while icr Fig. ll-III the lower the conductance
and +90 (+ is in coun terclockwi se direction). value the more frequently it will occur. For numerical
4. Calculate kx and k (Eqs. A-10 and A-H). calculations, these curves have to be divided into
5. Calculate C I(A ' - 13 ') and find tjJ' on graph increments to determine the number of conductance
(Fig. 10). elements corresponding to each discrete con-
6. Knowing the algebraic signs of C' and A'- ductance value.
B: look up in Table 3 to find what the true tjJ is. To have a basis for comparing the results, each
of the distributions has been adjusted to a median
APPENDIX B conductivity value of 1. In addition, the curves have
been normalized so that the total area is 1. In
DISTRIBUTION OF HETERCX;ENEOUS other words, the areas above and below the con-
FRACTURES ductivity value of 1 are both equal to Yz.
Some properties of these distributions are pre-
No data are available on the di stribution of the sented in Table 2. Distributions V, VI and VII were
permeability (or other properties) of naturally used by D. H. Thurnau 6 in some unpublished work.
occurring fractures. The eight artificial distribu- Distribution VIII is from 1. Fatt's paperS with the
tions considered in this report are shown in Fig. data smoothed to be presentable in the same form
11. The frequency function is a measure of the as the other distributions. >f. >f. >f.

136 SOCIETY OF PETROLEUM ENGINEERS JOURNAL