You are on page 1of 11

2958

A Rapid Method of Predicting Width and


Extent of Hydraulically Induced Fractures
J. Geertsma, SPE-AIME, Koninklijke/Shell Exploratie en Produktie Laboratorium
F. de Klerk, Koninkli jke/Shell Exploratie en Produktie Laboratorium

Introduction
During the hydraulic fracturing treatment of an oil or ing assumptions have had to be made:
gas well the liquid pressure in the borehole is in- 1. The formation is homogeneous and isotropic as
creased until tensile stress in the surrounding rock regards those of its properties that influence the frac-
exceeds tensile strength. Once a tensile fracture is ture-propagation process.
initiated, it is penetrated by liquid from the borehole 2. The deformations of the formation during frac-
and fracture propagation under continuous hydraulic ture propagation can be derived from linear elastic
action takes place. The fracturing liquid carries a stress-strain relations.
propping agent to ensure a highly permeable flow 3. The fracturing fluid behaves like a purely viscous
channel after pressure release. liquid; i.e., any peculiar flow behavior due to the addi-
Field results range from failure to obtain increased tion of gelling agents or other additives is neglected.
production to outstanding success. In all cases, how- Moreover, the effect of the propping agent distribu-
ever, itunfortunately remains uncertain whether the tion on the distribution of fluid viscosity in the frac-
values chosen for the operational parameters, such as ture is not taken into account.
injection rate, pumping time and fluid viscosity, were 4. Fluid flow in the fracture is everywhere laminar.
in fact the ideal ones. Though experience provides a 5. Simple geometric fracture-extension patterns are
lead, a more satisfactory way to predict results would assumed either radially symmetrical propagation
seem to be to subject the fracture propagation process from a point source @lg. 1A) or rectilinear propaga-
to a theoretical analysis that (1) makes the maximum tion originating from a line source (Fig. 1B). In the
use of the relevant physical information and (2) so first case the periphery of the fracture is circular, in
simpliies the resulting calculations that the field en- the second case it is rectangular.
gineer gets practical data that he can handle comfort- 6. A rectilinear propagation mode can be acconl-
ably. We are attempting here to do this in comection plished only by injection over a large perforated in-
with the prediction of fracture width and areal extent terval, thus forming a lime source. Such a rectilinear
before pressure release. What remains of the fracture fracture must therefore be located in the vertical
afterwards depends on the distribution of the prop- plane. A circular propagation mode might be ex-
ping agent between the fracture walls, and that is a pected from injection through a narrow band of per-
separate story. forations. This forms a point source. Because gravity
effects are excluded from our considerations, the
Idealization of the Problem fracture propagation plane may in this case assume
To keep the problem tractable, a number of simplify- any angle with respect to the wellbore as far as the

I
With the design charts presented here, and nothing more eiaborate than c slide ride,
it is possible to predict the dimensions oj either a linearly or a radially propagating,
hydraulically induced jracture around a wellbore.

DECEMBER, 1%9 1571


-.

application of the theory is concerned. infinitely large fluid velocities at the injection point.
For a linearly expanding fracture, we may equally
Mathematical Formulation of the well tdte f.,0 = o without introducing ) ficul~.
Fracture-Extension Process The theory of elasticity provides general relations be-
Under conditions of laminar flow, liquid in a narrow tween the fracture width w and an arbitrary normal-
space between two parallel surfaces experiences a stress distribution p over the fracture walls. For a
resistance due to its viscosity, To balance this resist- linearly expandkg fracture, England and Green, as-
ance the liquid pressure in the fracture decreases with suming plane strain conditions, found that
increasing distance from the injection point. Viscous-
flow theory (Poisseuille) relates the fluid-pressure w= 4(1 V)L f,df,
gdid ~0
AI_
IIIC
- l~iCLUIC
r--------- ....A+l.
WIUU1.
ma- ~mI:-n*l.,
AU. UU=.J
awwaA-
w~Y-.u-
(r
ing fracture in impermeable rock under the condition G !{ ~-~
of constant injection rates, thk leads to: iz
.
pmp= J ~;... f,. P (fI) d fl
+s~l j.
12p QL (1) ~J ~f, f,
h ~3 )
f1.w . . . . . . . . . . . (3)
and for a radially expanding fracture it leads to: For a radially expanding fracture, Sneddon2 obtained

into
In these formulas f. = x/L and j~ = r/R represent
fractions of the fracture extent, L and R, respectively,
fEw = RJR, if Rw = wellbore radius or
equivalent height of perforated
interval . . . . . . . . . . (4)
fLw = R,JL In these formulas G and v represent the elastic de-
P = local fluid pressure, PW= fluid pressure formation constants of the rock (the shear modulus
at wellbore and Poissons ratio), fl and f? are fmctions of the ex-
Q = injection rate, assumed to be constant tent of the fracture, and S is a constant external load-
1 = fracturing-fluid viscosity ing that represents the tectonic stress normal to the
h = height of a linearly expanding fracture, fracture plane.
and The two equations (Eqs. 1 and 3 or Eqs. 2 and 4),
w = local fracture width.
together with the appropriate boundary conditions,
In the case of a radially expanding fracture, the determine the shape of the fracture. The hydraulic
wellbore radius must be introduced in order to avoid fracturing process in a brittle material such as a res-

_. --
. _.. -.
---- .

1----
Fig. lA-Cross.setion of radially propagating fracture Fig. lB-Schematic view of linearly propagating f racture
A+h Isw-ninar
... .,, ., . .. . f!~w.
. ....-. fli!i~ with Iaminar fluid flow.

1572 JOURNAL OF PETROLEUM TECHNOLOGY


ervoir rock requires a particular boundary condition for an average value of Poissons ratio, v = 0.25; and
at the tip of the fracture. This condition, for the first the shape of the fracture, except in a narrow wedge-
time suggested by Zheltov and Khristianovitch,3 and Iike zone near the tip, is more or less elliptical:
later clarilied by Barenblatt,4 states tha~ in the case w2=w~(1fL2) . . . . . . . (8)
of a fr~~~~~~in ~,~bi!e eql~~ibnlurn.propagating in a
brittle solid, the distribution of normal pressure ex- Eq. 7 is valid for
erted by the fracturing fluid on the fracture walls must
be such that the faces of the fracture close smoothly ~QG3
<<1, say <0.05 . . (9)
at the edges. The condition of smooth closing implies i S L h
that This means that by combining condition (9) with
dw dw SL W.
Eq. 7, the theory is valid at least for WW< ~ or ~
()
d f f~=,
=0 and
() f~ r~.,
d
= O, respectively.

<~ . For instance if G = 10 kgf/sq cm, at a depth


Barenblatt proved that this ensures that the normal-
stress component at k
.:- w: I.- g.n..~ .- ;. Gmitia -..sn14
Up W UIG U CL- Ldl w 10 AU- wheres = 200 kg/sq cm, Wu must be smaller than
equal to the tensile strength of the rock. The tensile 20 mm for L = 10 m.
strength can be assumed to be of neg@ible influence
Under these conditions, it develops that the fluid-
for large-sized fractures in the practical range of over-
injection pressure with respect to the tectonic stress
burden pressures (see Perkins and Krech). Substitu-
perpendicular to the fracture walls, S, is
tion of the above boundary condition in Eqs. 3 and 4
leads to
Pw=s+q% . . . . . . (lo)

(5) Because according to Eq. 7 WWincreases in propor-


tion to ~~, it is found that PWdecreases with increas-
ing fracture length and approaches S for large values
for the linear configuration, and of L. Such pressure behavior is in agreement with
reported field observations. A check for the validity
of the assumption of laminar flow is that the Reynolds
number, ~R, equals Q p/h P less than 1,000, where P
is the liquid density.
for the circular. For a radially propagating fracture, the maximum
(Note that if the tensile strength had been taken width at the wellbore, again for v = 0.25, is approxi-
into account, Eqs. 5 and 6 would have been found mately
to be
(11)
1

K
r
o
P (fL) d fL
~1 fL2
=+s+
~2L
and the shape is parabolic except for a narrow zone
near the tip
W2 =
and Ww(ljz). . . . . . . (12)
1
~QG3
fR P (fR) d fR K Eq. 12 is valid for << 1, again say
d s R
f ~1 fR2 s+~n
jRIO SR
<0.05, or w. < .
respectively, G
The fluid pressure at the entrance of the fracture
~Ea (r= RJ decreases with increasing fracture radius R
in which K = = Barenblatts cohesion
d lv according to
modulus. In this- expression E = Youngs modulus
and a the specific surface energy. Our theory thus 5 Gww
4 K, pm=S~&nfrw-
assumes that 2L > ~ ~ and that 2R >$, re-
In terms of the Reynolds number, laminar flow con-
spectively.) ditions are now fullilled provided NRe equals Q p/
~ ~ ~ ~ !e~~ than 1,000. The fracturing fluid will
Equations for Fracture Wldtb and Shape usually behave in a Iaminar fashion, except in a cer-
In Appendices A and B approximate solutions for the tain area near the wellbore. As long as this area is
sets of equations @qs. 1, 3, 5 and Eqs. 2, 4, 6) are Iimited to a few well radii, it will hardly invalidate
derived. For a linearly propagating fracture the maxi- the theory given.
mum width at the origin amounts approximately to
Effect of Formation Permeability
4 ~QL2 on Fracture Dimensions
Ww =2.1 T,..... (7)
i
Communication between fracture volume and the

DECEMBER, 1%9 1573


pore space of a permeable formation results in fluid 18) into the material-balance equation @q. 15) pro-
loss and, among other things, in a pore-pressure build- vides the equations
up in the immediate surroundings of the fracture.
Q _ 3T w,, + ~Sp dL
Thus the formation expands locally, reducing the
fracture width. It has been demonstrated, however,
that in most practical cases this expansion effect on

h ( 4 ) dt

fracture width is negligibly small.


Aithough, for a given exteiit (L Cii R), the ei?ect d
fluid loss on fracture width is thus small, fluid loss
definitely reduces the fracture extent for a given for the linear two-sided fracture, and
pumping time. In other words, knowledge of injec-
tion rate and pumping time is not enough to permit
the prediction of the extent of a fracture. Instead it Q=(~ww+2~sP)~
may be investigated by means of the material balance
that relates the amount of fluid injected to the fracture dR d,
volume, the dtierence between the two being due to +27C dtg~ (20)
the fluid loss into the formation. .f
o
For the linear configuration, the fracture volume
at any time z amounts to V = 7/4 h L WWfor a one- for the radial case.
sided fracture, and thus If in the last two equations the fracture width w
were independent of fracture extent (L or R) the
V=;h Lw,, , . . . . . . (13) solution of the problem would be a well known one.
In the present case this solution can still be used be-
cause . th.a
In .,,W mrnrhl~t
y,w.. . w(~A .--, the
., , . . //#t) --- yap:afions ill w as
if the fracture extends on both sides of the wellbore.
For the radial configuration, a function of time are much smaller than those of
dA/dt.
8: The constant w-value is chosen in such a way
=15 WR 14) that the true solution is approached as closely as
possible. Therefore WWin Eqs. 19 and 20 is replaced
provided R,,,/R << 1. This follows from Eqs. 8 and
by @WW,,where wW,is the fracture width near the
12, respectively, and involves the reasonable assump-
wellbore at the time the pumps stop, and /3 is a factor
tion that the effect of fluid loss on the pressure dis-
that causes the approximate solution of Eq. 19, or
tribution in the fracture, and thus on fracture shape,
Eq. 20, to become numerically equal to the exact one
can be neglected.
for the case where C = O, Sp = O (no fluid loss, no
The material balance reads
spurt loss). The exact solution follows from the mate-
dV rial-balance equation dV/dt = Q. Direct integration
Q QmSP$$, . . . (15) leads to V = Q L With the aid of Eq. 13, or Eq. 14,
dt
this gives us for the extent of linear and radial frac-
where Q? represents the rate of fdtration loss into the tures:
f. _-_.,_ do
[OrrhdUUri>
o .L.- ..- .4 1-
UK spbl L 1uS3, ~ii d ~ the tcXa~eXpCH!

surface area (increasing with time) of the fracture. L,= ~ and R, ==, respectively.
A equals approximately 4Lh for the linear case (two- ~ we

sided fracture) and 2xRS for the radial propagation


If, however, we again take C = O, Sp = O and inte-
mode. The rate of filtration loss per unit of exposed
grate Eq. 19 or Eq. 20, we find that, treating WWas a
surface of the fracture is indicated by u, and thus,
constant:

Qt= /4ud A= !u$+d, . . (16) L.= ~ and R,=%, respectively.


/
% %
In order to bring these results into numerical agree-
For most practical situations, ment, a correction factor /3 = 2/3 for the linear case,
--A o 0 jn s-- .5- *..A;.1 O- a rn,~ct ha ktrndl]r.erl {f
c tUIU ~ 0/ 7 LUl U G Iau,cu WC& .,. -.. u,- .u...v...--.V -
u WW, is chosen for wti in the approximate calculation.
@(17)
After the introduction of these &values, applica-
where C represents the so-called fracturing fluid co- tion of Laplace transform and the convolution theo-
..= .:--+ Clllu
G1llUIGUL ---l -1A+
-
t ~ ~qre$~~t~ ~~~ t~rn.~.of ~x.
rem. to Eqs. 19 and 20 leads to
posure of the fluid to the permeable surface area A. 2
&L .9
Substitution of Eq. 17 into Eq. 16 leads to f?aL2dc CIL,
dt h( zWme + 8SP)
in which
. (18)

The substitution of the fracture volume equations


(Eqs. 13 or 14) and the filtration loss equation @q. and

1574 JOURNAL OF PETROLEUM TECHNOLOGY


~dR Q 15 These results are thus independent of SP, arid are
_= tR2 ed~ C2R ,
dt 477 ( 4 w,., -t 15 sp ) practical formulas permitting an estimate of the extent
of large-size fractures. Eqs. 7 and 11 subsequently
in which provide the corresponding fracture width.
15c~z
aR = 4WW,+ 15SP
Fracture Design Charts
For a <8 we must solve the sets of equations (Eqs.
Integration of these relationships shows that 7 and 21 and Eqs. 11 and 22) in ftdl. For this purpose
i+i.
. . u pnnwpni~rltto
. , ---. -. in~~duce di,mensiodew ~OUpS. ~
32 =Qh
~,(T Wme + W) the linear case,

2aL

(
1 + ea.z dc UL , . (21)
G )

This changes Eq. 21 into

30$
2a.
Cz (4 Wwe + 15 s,)
~L=+i7H~++!+=q
(77
1 + fi?a.zetic
CC~
)
. . (22)

For the product e erfc a, various approximate ex-


pressions are known for various ranges of values of
a. For a > 2, the asymptotic expansion . . . . . . . . . . (21a)
1 In the radial case the introduction of Ku, Ks and
eU2erfc a e 1+: (l)
aqz [ m=] R, c
= KR changes Eq. 22 into
1 .3.. , (2m 1) ~

is suitable. This also shows that for


(2a2)m 1
a>>l, ea2erfca+ --&<<l,
z~ ->>. 40-l!_
l.~olvra>> l,Eq. 21 RXbXX~
whereas --=
v~
to . . . . . . . . (22a)

Lq~ Qv~ III Fig.2, KL is plotted as a function of K,, for O < K.


(23)
27= < 2.80, and various values of Ks. Ks = co means
no spurt loss. This last curve has an asymptote KL =
and Eq. 22 SiI@ifif3S to
% T, corresponding to the asymptotic solution, Eq.
1 4~ 23. In Fig. 3, where the same procedure is followed,
(24) Kn is now plotted against Ku for O < K. < 1.40.
+7 4 F
Here the curve for Ks = cm has an asymptote KR =
l/r2, correspondkg to the asymptotic solution, Eq.
24.
~ .x Furthermore the fracture width equations (Eqs. 7
L Qw and 11) can also be written as relations between the
m,--
-------- ~~------- ..._. _ . . dimensionless K-values as follows
p ...,...,. ;
1
for the linear case, . (7a)
u 2 = ~
and
1
for the radial case, . (1 la)
8 R = G
where

K,. = 8.724 ~~) (~) , . (25)

and

Fig. 2Fracture design chart No. 1, linear case. PE=2(*iJ(~)2 - 2)


DECEMBER, 1%9 1575
The asymptotic solutions for C = O,SP = O which,
... ,
as alreaoy nas men shown, lead @ !4 ~ h L wW=Qf
in the linear case, and to 8r/15 WWR2 = Q t in the
radkd configuration are also indicated in Figs. 2
and 3. They are straight-line relationships between
Ku and K., and K and K., respectively:

:KlIrEF7
Analytkml
Computer
results,

results
this
of Baron t d
poper 1

8?? _.
KU=~KL,
15 R

6+-1=--F \
Asymptote for WW [in,. WP=r)
These figures are in fact convenient graphical solu-

\\
12

\
tions of the general problem. 1.

Discussion
Fracture-width equations of the type repmented by
Eqs. 7 and 11 except for significant differences in 400 I I I
lu?
I
%e

the proportionality constants can already be found


in the literature, originally in a paper by Perkins and
Kern. The reasons for the numerical diilerences are
discussed in detail in Appendices A and B. Results of
material-balance studies leading to expressions simiiar
in form to Eqs. 21 and 22, (again except for the nu-
mencsj date from &G . -wu~h
-..1. -: nod
U1~a. rneat-~
a, . p..xu in nn
- ---
appendix to a paper by Howard and Fast.8 Never-
theless, these partial results have never been thought o! I I I I I I I I I ;0
,..4 2 5 ,..3 2 5 ,0-2 2 5 ,~- 1

of as leading to sets of equations with two unknowns, C cm/Gec

which thus contain the complete solution of the prob-


Fig. 4-Fracture dimensions as a function of fluid-loss
lem of fracture dimensions. coefficient (linearly propagating fracture).
We must male one exception to this point. Re-
cently, Baron et al.ll of the Institut Fran@s du P&role
followed a reasoning basically similar to ours. The ve

Fm
m
main difference between their approach and ours is
that the French authors have not used an analytical

-r
approximate solution, but have solved the problem
Anolytmi solutlon, this poper
numerically. Some of their computer results are com- - -computer r.sdts d Baron t at 1
pared with our analytical ones in Figs. 4 and 5. These
figures show how the fluid-loss coefficient influences ~
fracture width and extent for some chosen values of -J t -80 mm
I
the shear modulus of the formation, the fracturing
fluid viscosity, pumping time and injection rate, cor-
responding to those used by Baron et aL Because the
French authors dld not take into account any spurt-
Ioss effectj all comparisons are made for KS = co.
Asym ptote for Wwe (this POper)
3
I

1
~R=A!l
Qw 1

+11~
1

1: :1 \
o~o x:-+ --I--L--------L-- --- L--L -:--------?
\\
? \\ 2

\\
\ \\
\
l\I. 1

\\ )

J&
.325 ,0-2 2 5

,-1
0

C cm &
K.. -
W.?
Fig. 5-Fracture dimensions as a function of fluid-loss
Fig. 3-Fracture design chart No. 2, radial case. coefficient (radially propagating fracture).

1576 JOURNAL OF PETROLEUM TECHNOLOGY


These figures show that from a practical point of view TABLE 1PREDICTED FRAtitiRE titiEf$S:=S,
FIELD CASE
the agreement is usually satisfactory. The most sig-
nificant differences are found for the asymptotic solu-
tions, i.e., for C ~ O. As under these conditions our
approach is straightforward, involving the least num-
ber of assumptions, the asymptotic values for C ~ O
obtained by Baron et al. are questionable.

Applications
In Figs. 6 and 7 we show how the fracture width at
the wellbore increases with pumping time for practical
values of G, p and C. From these last two figures it is
clear that even after a very short injection time a frac- O =35 bbl/min t =33 min
ture frequently becomes wide enough to permit the
passage of propping agents of rather large size. Again, h =70 ft G= 5 X104 kg/cml
m
in these examples the spurt-loss is not talcen into ac-
count. The need to use a highly viscous fracturing would have been adequate, the new composition was
fluid for a particular field case can be determined by not. Under the altered conditions, a fracture length
constructing such plots, using the appropriate data of of only 28.5 f~ and a fracture width at the wellbore
G, Q and C. of 2.6 mm were predicted after a pumping time of
The practical significance of the design graphs, 33 minutes. So a sand-out was to be expected. The
Figs. 2 and 3, can further be illustrated by a field actual job confirmed this prediction. After the frac-
-s 11
exaiiipie. A se~i~~ UL $..CIVLUIV
.+... t.. ~hf.
UW.UW..W
tc ~~~ qlu~~fj~
turing fluid formulation had been improved the job
in a rather new operating area, where the choice of was earned out successfully.
chemicals for adequate fluid-loss control was limited.
The service company in charge of the fracturing op- Nomenclature
eration recommended a fracturing fluid formulation a = specific surface energy
that would provide a fluid-loss coefficient of C = 5.2 A = total fracture surface area
X 10-3 ft/min~ and a viscosity of 20 cp. The chemi- C = fracturing fluid coefficient
cals required to achieve this low fluid loss were not E = Youngs modulus
available, however, and the formulation of the frac- ~,, fz, f~, and jR = fractions of fracture extent
turing fluid had to be changed on location, Laboratory G = shear modulus of formation
tests on actual core materials revealed that the fluid- h = fracture height (linear
10SScharacteristics of this new fluid were unfavorable: configuration)
C = 90 X 10-3 ft/min~ and P = 10 cp. With the aid .x = &i-en
...- ~innleqs
.-... LWOUD
~ ..=

of Fig. 2 (because vertical fractures were expected) it L= fracture lengtl-(linear


was possible to evaluate the consequences of these configuration)
changes at once. Table 1 has been compiled from Fig. Reynolds number
2 with the help of only a slide rule and the conversion fluid pressure
factors mentioned in Appendix C. It was assumed injection rate
that there was no spurt loss. From this table it is seen fracture radhzs (radial
that while the originally recommended fracturing fluid configuration)

Wwe
mm mm
16 5
I I
I
IL
L
12 Q=5&

10 ~
3
( # T ~=#!L*
8 // ~

, 2 {
6 a
//{ Q=I*

L 4/-- G- 109kgf/c.ti
1
~w 25 CP
I
I C _ aol cm/ K&

I
o
o 1 2 3 L 5 6 1 8 mm
t t
F@ ~ln~rease in fracture width with time Fig. 7increase in fracture width with time
(linearly propagating fracture). (ratiiaihy --- .+1.. S-+,,rml
prupagcauit~ ,8-...-. -,.

DECEMBER, 1%9 1s77


.

radial distance simplified approach that leads to the practical for-


tectonic stress normal to fracture mula, Eq. 7. To this end, we assume a plausible pres-
plane sure distribution in the fracture and calculate from
spurt loss Eq. 3 the fracture shape, and from Eq. 1 the pressure
pumping time distribution in such a fracture. This will show whether
time of exposure to fluid loss or not the assumption is acceptable.
rate of fluid loss per unit surface Barenblatts condition that closure must be smoot!!
area implies infinite flow resistance at the very tip of the
fracture volume fracture, so that pressure here must be zero. Since
fracture width with smooth closure the increase in fracture width is
longitudhtal distance more than proportional to the dktartce from the tip,
penetration depth of fracturing the pressure gradient decreases by at least the third
fluid in formation power of the distance (see Eq. 1). It therefore rapidly
fluid viscosity becomes very small, and it is plausible to approximate
Poissons ratio of formation the pressure distribution in the fracture by the dis-
fluid density continuous one*
tlit

. ..1..s--
/kkiKlwlcss~nwFhm
4.

We wish to thank A. C. van der Vlis for providing where we suppose, a priori, fLo+ 1. Barenblatts con-
the data on the field application.-mentioned
. ..- in the last dition equation (Eq. 5) gives, with such a distribution,
section, and the management ot Shell Kesearcit N.-v.,
The Hague, The Netherlands, for permission to pub-
~Lo= sin~~ . . . . . . . (A-1)
lish this paper.
The fracture shape resulting from this pressure dis-
References
tribution follows from Eq. 3
1. England, A. H. and Green, A. E.: Some Two-Dimen-
sional Punch and Crack Problems in Classical Elastici-
ty, Proc., Cambridge Phil. Sot. ( 1963) 59, 489.
2. Sneddon, I. N.: The Distribution of Stress in the Neigh-
borhood of a Crack in an Elastic Solid, Proc., Royal
Society of London (1946) A 187, 229.
3. Zheitov, Yu. P. and Khristianovitch, S. A.: The Hy-
draulic . Fracturing . .-=
of an ,-8.Oil-Producing Formation,
Izvest. AKad. NauKxi3K, udd ~eicii%iik (~%5 ) No.
3, 41.
4. Barenblatt, G. I.: The Mathematical Theory of Equi-
librium Cracks in Brittle Fracture, Advances in Applied
Mechanics ( 1962) 7, 56.
5. Perkins, T. K. and Krech, W. W.: The Energy Balance This shows that the maximum fracture width at the
Concept of Hydraulic Fracturing, .SOC. Pet. Eng. J.
(March, 1968) 1-12. wellbore amounts to
6. Geertsma, J.: Problems of Rock Mechanics in Petro-
leum Production Engineering, Proc., First Cong. Intl.
Sot. of Mech., Lisbon (1966) I, 585.
7. Sadowsky, M. A.: Thermal Shock on a Circular Surface
of Exposure of an Elastic Half Space, J. Appl. Mech. For fLo+ 1 this reduces to
(1955) 22, 177.
8. Howard, G. C. and Fast, C. R.: Optimum Fluid Char-
acterishcs for Fracture Extension, Drill. and Prod. Prac.,
API (1957) 261.
9. Handbook of Mathematical Functions, M. Abramowitz whereas a good approximation of the equ~lbrium
and 1. A. Stegun, eds., Natl. Bureau of Standards, New condition (Eq. A-1) is
York (1964).
.,, ,-. ,.
IU.rerKiiss,T. K. aitd Kern, ~. It.: W;A*L
,.11. .
af ....-----
uv~=~~li~
Fractures, J. Pet. Tech. (Sept., 1961) 937-949.
11. Baron, G,. et al.: Fracturaiion nytirauiique; bases iiik-
oriques, etudes de iaboratoire, essais sur champ, Proc.,
Seventh World Pet. Cong. (1967)3, 371. Combining the last two approximations leads to
i2 Khristianovitch, S. A. and Zheltov,4 Yu. P.: Forrnatjon
of Vertical Fractures by Means of Highly Viscous Fluid,
Proc., Fourth World Pet. Cong. ( 1955) II, 579.

For a given pressure in the fracture in excess of the


APPENDIX A tectonic stress, the fracture width at the origin is thus
Fracture Width Determination for a tO a first iIpprOXhNitiOtI independent Of fLo; k, it
Linear Mode of Propagation does not depend on the extent of the region of zero
The behavior of a linearly propagating fracture has
This approach to solve the problem, proposed for the first time
been considered in some detail by Khristianovitch and in Ref. 12 is frequently misinterpreted as asauming a void near
Zheltov. A conformal mapping technique was used the tip of the fracture. However, only the pressure is off, et Ieest
in our theory. The fracturing extends right to the tip of the fracture
for finding the displacement field. We give here a with a local steep pressure gradient.

1578 JOURNAL OF PETROLEUM TECHNOLOGY


pressure. However, the shape near the tip does depend Expressing w,. and v- ifI terms of (P s),
we find with fLo + 1,
strongly on f~~. If ~ were applied over the full frac-
tUre length (fLo = 1), integration of Eq. 3 would re-
sult in an elliptically shaped fracture. With fLo< 1, ,-s= y-~(+)+,
Eq. A-2 represents a fracmm of the shape ,ui(ith
shown in
nC the
Fig. li3, but with the ---- SiilllGlll~A~,,IuIJ.r~--. --
-..:m,,m W~CiI ieaiis to
elliptically shaped fracture.
It remains to be seen whether the pressure distribu- 4 84(lv) IJQL
. ~ .
w~ =
tion in a fracture of the shape given by Eq. A-2 cor- d T G h P,,
...-A=
lvapllw. tfi
. the
---- assumed
-..----.distribution, and what value From these it foiiows that it is of M& practical imp-
should be given to ~. From Eq. 1 it follows that ortance for the determination of WWwhether we take
~ = PWor a slightly smaller value (for examPlej ~
equal to the average value of 7 along the fracture).
Taking ~ = pw, and a suitable value for Poissons
ratio, v = 0.25, Eq. 7 follows. Furthermore,

;=~+
V%(A)3 ;
where W = w/ww and follows from Eq. A-2. The
result of numerical evaluation of the integrals is shown or, with p,. = P,
in Fig. 8. It will be seen that the curves approach the
assumed distribution as f~.oapproaches ufity. The PUJ=~+2(~::)L
pressure at the origin
jLo

12p QL d fL
pw =
h W,.3 r W
.0 16p QG3 ;
is examined in Fig. 9 for a number of values of f~.. f+l-v~T ()
Again, when f,,. ~ 1, a good approximation appears
to be With ~ = p., j- S and tie condition fLO >0.95 Y
condition (9) results. It follows that the region of zero
fLo

d fL 7 pressure i.e., L(l fLo) is independent of the


4\/1
extent of the fracture L.
f
w fLo
o If fLo < O.%, a condition that is ufllkely to be
met for most practical conditions, the more compli-
Thus cated calculations by Khristianovitch and Zheltovlz
21p QL are required. This implies either very viscous frac-
h ~-1 fLo2
w * WW3 turing fluids or tectonic stress conditions of a very

/0
+s
,____ ,.
0.5- --1 ~-
I
.:1 ~ I
o ~o;cu,atd Wants ~ .
I
~~fLj, ,,l:c ~ - - :
08

~, Ws=l 55

0.6 -
: ----TTx

11 \ \ 1
P
o, .+
,/
--- ; ---1- :------- --:-- y
0.4 I
\ I
I ;!
I
! I I J
1
i!i o 088 0w.
I 100 096 092
0.2 k
~~;j!!~

I I Fig. 9-Determination of an algebraic expression for


I I
00 a2 0.4 a6 m 1.0
fL

Fig. -Pressure distribution in a linearly


propagating crack.
1579
DECEMBER, 1959
low level. The latter implies fracturing at shallow At the wellbore fR = fRW << fRO. perform~g the
depth. Ub..u.,
--e-.cnm ..-.-& -- m&ing use of Eq. B-1,
-...---., a~~
intsurntinns

It is of interest to note that Perkins and Kern we find the fracture width at the fracture origin to be
obtained expressions for the fracture width of hy-
draulically induced fractures for some other geome- ~,o=4(lv)R ~l_E)a+
tries, including the one to be discussed in Appendix TG [
B. The results show qualitative agreement with the
theory outlined here, Their determination of the flow
resistance in the fracture, however, is not based on
&e(S+aln2)

In the first integral appearing in Eq. B-2, fRw2 maY


I . . . . (B-3)

the Barenblatt theory. In view of this, the numerical


values for fracture width according to these authors be neglected with respect to f,, and this also applies
are too low. to the reciprocal integrand. Furthermore,

APPENDIX B
Fracture Width Determination for the
fR.
Radial Mode of Propagation
During radial fracture propagation, the situation near
the tip is essentially the same as that in the case of
linear propagation a smoothly closing region of in.

low pressure followed by a region of almost constant


pressure. Close to the welIbore, however, the small V(1 f,) (1 fRo)
Wi[
cross-sectional area of the fracture no longer permits
a negligible pressure gradient. It will therefore be as- \/l jR + VI fRo
.
(fRo fR) in
sumed that the pressure distribution is logarithmic, as ti]jRo fR]
it would be if the fracture width; were constant up The same procedure applies to the reciprocal inte-
to a certain radius jn~; beyond jRothe pressure is con- grand. These approximations of the integrals and use
sidered to be zero. Thus, 01 Eq. B-3 lead to

and fR = r/R. Barenblatts condition, Eq. 6, provides


with fRw << 1,
(l&)p,O=Salnf~W , . . . . (B-1)
where E = ~ 1 fRo2<< 1. The shape of the frac-
ture corresponding to the assumed pressure dktnbu-
tion follows from Eq. 4. The resuit, at jR < jh, is

. . . . . . . . . . (B-4)
With the exception of the region close to the tip, i.e.,
when fR < fR., the 10f@hIUiC h?rms can be further
simplified to

fRo

:
+a
w fR

1.
f12

flz
fR2

f,twz d f
This provides

f
f,2
2fR
fRo2
df, + . . . . . . . . . . (B-5)
h )
fm The second term in this expression for w is a correc-
tion to the elliptical shape predicted by the first term.
+~~lfR
{[
h
I - ;;:fR.2 (k)]+:} This second term disappears for a ~ O (that is, if the
pressure drop due to flow resistance is neglected) and
. . . . . . . . . . (B-2) thus p is constant. The exact level of the correction
The same expression holds fOr jR > fRo, provided that term is not so important. The value of a depends on
fRo is replaced by fR in the lower limits of the integrals. ~, which will not be greater than WW.In what follows

1580 JOURNAL OF PETROIJ3UM TECHNOLOGY


.

P
we willtake;= 2/3 is ~ equal to the aver-
wW, that m
1
age value of the ellipsoid having the semiaxes w. and 1
I
RfR., and prevmg when P is cons~t. In fact> ~k- I

ing~= WWhas led to only insignificant diilerences. 0.75 !


!
With w given by Eq. B-4, the pressure distribution is I
I I
found by integrating Eq. 2. The value of f~~ is then 0.5 :
found from the condition that p = O at fR~. The im- I I
1
plicit nature of the relations makes it necessary to use !
0.25 ;
a numerical trial-and-error method for a range of (
I
values of S and RW. In appropriate dimensionless
0.05 -
form, 0 & az r3L Ki6

- ----- d.,,
Fig. io-fiuki PM=U,= ~i +Pifyltinq
-.-.
in a radially
propagating fracture for various values of fgo
and (f,. = 0.015andv== 0.25J.
fjw = 0.005 to 0.095.
First a value of jEo is assumed. Then WWfollows At the smallest value used for the S-parameter, the
from Eq. B-3, with the factor a expressed in terms of distance R Ro increased by 10 percent and another
w*. Next p~ is determined from Eq. B-1 and p by 5 percent at the greater values of fRW.With the em-
integration of Eq. B-2. This gives the values of p at pirical relation of wW,the value of the injection pres-
f. = f,., which should be made zero by adjusting j~o. sure is found from Eq. B-1 and reads
I=-.VI mA..-.
D,&Q~m$~ -~a~o, ~ = 0.25 was agfi chosen. From
the computer results the following empirical relation
was derived:
The fluid pressure distribution p/pw for various values
of f~ois shown in Fig. 10.
WW=2 4 ~ d (standard deviation, 5 percent).

The shape of the fracture Nrned out to be parabolic AmmDIx c


rather than elliptical, giving Eq. 12. Conversion Factors
At values of over 10 for the S-parameter, (1 fE~)
For both the dimensionless plots and the formulas, all
was found to be independent of jRICI ad in inverse
proportion to R. Thus, the distance R R. is inde- quantities must be expressed in a consistent system of
utits, for instance in dynes, centimeters and seconds.
pendent of the size of the fracture, a situation similar
It should be remembered that
to that in the linear case. We obtain specifically
1 bbl/rnin = 2,640 cclsec
1 cp = 10- dyne sec/sq cm
1 psi = 6,89 X 10 dyne/sq cm
Original manuscript received in Society of Petroleum Engineem 1 ft/min~ = 3.9 cmlsec~
office Dec. 12, 1968. Revised msnuacript of paper SPE 2458 re-
ceived Sept. 8: 1969. @ Copyright 1969 Amarican Institute of Min- lmin=60sec
ing, Metallumtcsl, and petroleum Enaineem Inc. ift= 30.5 cm
This paper will
will cover 1969.
be printed in Transactions volume 246, which
1 in. = 2.54 cm tJPT

DECEMBER, 1%9 1s81

You might also like