Sei sulla pagina 1di 14

1

Mark D. Zoback, David D. Pollard

HydraulicFracturePropagation
and the Interpretation of Pressure-Time
2ecordsfor In-Situ Stress Determinations

HYDRAULIC FRACTURE PROPAœATION A•;D THE I?!TERPRETATION

OF PRESSURE-TIME RECORDS FOR Ir•-SITU STRESS DETERMINATIONS

Mark D. Zoback
David D. Pollard
U.S. Geological Survey
345 Middlefield Road
Menlo Park, California 94025

IHTRODUCTIOH the rock surrounding the borehole behaves elastically;


it is impermeable: accurate tensile strenoths can be
Followino the stress measurement carried out as determined; the theoretically anticipated fracture is
part of the Rangely earthquake control experiment generated rather than a pre-existing one; and the
(Raleigh etal, 1972; and Faimson, 1973), numerous breakdown pressure represents the pressure at which
investigators began makino in-situ stress measure- fracture initiation occurs.
ments using the hydraulic fracturing technique.
Haimson(1977) and McGarrand Gay (1977) review much
of this work. Because the reliability of these
measurements depends upon understanding the fracture • Pb
initiation and extension process, we consider P
several aspects of hydraulic fracture prooagation
relevant to use of the technique for the determina-
tion of in-situ stress.
s
s
Pumping Pressure
Hubbert and Willis (1957) showedthat the pres- u
sure to hydraulically fracture a wellbore depends on r
•ISIP
*he tensile strength of the rock and tectonic e
stresses. A numberof authors (Scheidegger, 1962;
Kehle, 1964; and Haimson and Fairhurst, 1971) sub-
sequently pointed out that the pressure-time records
of a hydraulic fracturing operation could be used to
compute the tectonic stresses if the tensile strength
of the rock were known. Hubbert and Willis showed Time
that a vertical fracture should form in a vertical
wellbore at the azimuth of the maximumcompressive
horizontal principal stress, SHmax(it was presumed Figure 1. Idealized Pressure-Time Curve.
that one principal stress, Sv, is due only to the The BreakdownPressure, Pb, PumpingPres-
sure, and Instantaneous Shut-In Pressure,
weight of the overburden and is vertical). Briefly
sumamrizine the manner in which in-situ stresses are ISIP, are Indicated.
determined from pressure-time data (such as the
idealized data shown in Figure 1), the fracture Haimsonand Fairhurst (1970) investigated pore
should form whenthe borehole pressure, Pb, reaches fluid permeation into the rock surrounding the bore-
hole prior to breakdown. Whensuch permeation occurs
Pb= T + •Hmin
• - 'Hmax
• - Po (1) a factor can be incorporated into Eqn. (1) to compen-
in an impermeablerock with pore pressure, Po, where sate for the phenomena(see Haimsonand Fairhurst,
SHmin is the maenitude of least orincipal horizontal 1970). In practice, however, the factor is usuall•
stress (compresSion
is taken as positive throughout close to unity and Eqn. (1) is used in the form shown
this work) and T is the tensile strength of the rock.
Pb is presumably indicated by an abrupt drop in bore- Whenpre-existino fractures are present Abou-
hole pressure and is thus termed the breakdown pres- Sayed et al (1977) suggest that it is possible to
sure (see Fieure 1). Because the fracture should estimate SHmax
whenthe length of suchfractures are
propaga;e perpendicular to the direction of least known. However, because there is considerable uncer-
compressionwhenSHmin Sv, SFiminis presumedto be tainty in knowing the length of pre-existing fractures
eeual to the instantaneous shut-in pressure (ISIP, they acknowledgethe limited applicability of such a
see Figure 1). Interpretation of shut-in pressures technique. Rather than use pre-existing fractures,
in cases when-•-im•n
•. - _
Sv is •iscussed at leneth it has been shown that appropriate steps may be taken
below. to overcometheir presence. Haimson(1974) found that
a plaster of paris borehole liner enabled a correctly
EQuation (I) can be solced to yield SHmax oriented fracture to be generated in a highly foliate
cause Pb and SHminare measured from the pressure- shale, and Zoback et al (1977b) found that if viscous
time curvet Po can De measured(or estimated); and T drilling mudwas used as the fracturing fluid a cor-
can be •easured for core samples. The reliability of rectly oriented fracture could be generated in pre-
fractured rock.
this deCermination depends on several key assumptions:
Mark D. Zoback, David D. Pollard

Another observation in the study of Zoback et al. Consider a two-dimensional fracture of variable
(1977b) was that fracture initiation could occur sub- length •. In case 1, the fracture is subjected to a
stantially before breakdown when highly viscous dril- mathematical point load, or in reality a uniform pres-
ling mud was present. The authors hypothesized that sure, P, acting over a small segment, 2r, of its
•his phenomenawas caused by large pressure drops length (r << 4). In case 2 the pressure acts uniform-
occurring in the fracture resulting in initially ly over the fracture's entire length. The stress in-
stable fracture propagation. The presence of dril- tensity factor at the fracture tip, K, and the frac-
ling mud and even a mud "cake" in wells is a common ture opening displacement of the fracture wall, D, are
occurrence, and it was speculJted that manypublished given by
estimates of SHmaxmay be erroneously low since the
breakdown pressure, Pb, may have exceeded the frac- (case 1) K = 2Pr/•-•-•,
ture ini•-tion pressure, Pc. Understanding this D: 2Pr(1-v)[1-(2x/•)2]'2/• (2)
phen•me • is clearly quite important and is con-
sider at length below. (case 2) K = PC •=,
D = P (1-v)[1-(2x/•)212/2p (3)
Finally, the ultimate usefullness of Eqn. (1)
must be considered in light of the difficulty of where p is the shear modulus and v is Poisson's ratio
determining accurate values of tensile strength. Un- of the material. These formulae are readily derived
less a well were extensively cored, a substantial from the more general expressions presented in
number of samples would not be available for testing Appendix A. The change in displacement along the
and accurate determination of the tensile strength fracture's length illustrates that an elliptically
would not be possible. Bredehoeftet al (1976) and shaped crack arises in both cases. If we consider
Zobacket al (1977a) have suggested that Eqn. (1) case 1 as a model representing borehole pressuriza-
can be used with T : 0 and "breakdown" pressures from tion without pressure acting within the fracture (due
pumpingcycles after fracture initiation. Investi- to pumping at a high flow rate or using a very viscous
gating the validity of such an approach is also con- fluid), then the stress intensity is seen to mono-
sidered below. tonically decrease as the fracture extends. This
represents stable fracture propagation because growth
A MODEL FOR FRACTURE EXTENSION would stop when the stress intensity factor fell below
the value of the fracture toughness of the material.
In attempting to intuitively understand the Such stable propagation necessitates an increasing
fracture initiation and extension process, it is borehole pressure with time to propagate the fracture
necessary to consider the coupled problem of the and would thus not produce breakdownbehavior. For
elastic deformation of a fracture and viscous fluid case 1, when P and r are constant, the wall displace-
flow into it. The necessity of considering this ment at the center of the fracture is unchanged as it
coupled problem is illustrated by the extreme cases propagates. In case 2, with constant pressure acting
shown in Figure 2. over the entire fracture length, we see that fracture

2r
CASE I CASE 2

I--
Z
(./3
Z

I--

Figure 2. Stress Intensity Factor and Fracture OpeningDisplacement


SchematicallyShown
for
a Two-Dimensional Fracture with Constant Internal Pressure, P, Applied Over a Short Central
Portion where 2r << • (case 1) or Over the Entire Fracture Length (case 2).

15
Mark D. Zoback, David D. Pollard

propagation is extremely unstable since the stress length, •, (where r < • < h) into which fluid is
ntensity rapidly increases as the fracture propagates. being pumpedat a particular rate and borehole pres-
Further, since the wall displacements of the fracture sure. We assumethat the fracture propagates per-
also increase as the fracture propagates, fluid flow pendicular to the least principal compressive stress,
into the fracture is enhancedby the fracture's in- S3, and we ignore stress componentsin the plane of
creased width. Clearly, a propagating fracture can- the fracture. The fracture is subjected to an inter-
not be represented precisely by either of these ex- nal pressure distribution generated by fluid flow
treme models. Fluid pressure may act in the fracture into the fracture and to a constant pressure over the
to somedegree, but not necessarily such that fracture portion of the fracture corresponsing to the wellbore.
propagation is unstable at all times. Shear stresses on the fracture face due to the fluid
flow are ignored.
We consider a very idealized and simplified ver-
sion of the coupled problem of deformation of the Three important points about the model should be
fracture and fluid flow into it. We utilize a two- made clear. First, while we prescribe a volumetric
dimensional plane strain fracture model in an infinite flow rate into the fracture, we assume that increases
continuumof linearly elastic, homogeneous, and iso- in the fracture's volume(due to propagation) pre-
tropic material. We also consider steady, constant vents pressure buildups due to mass accumulation.
property flow of a Newtonian viscous fluid. While Second, we analyze fracture extension by considering
such an ideal model cannot describe precisely the the process quasi-statically. That is, the model acts
complexprocess we are attempting to understand, it as if we have stopped a fracture in time. We have a
can be of significant use as a tool for gaining in- fracture with a prescribed length, and the pressure
sight into the process. Assuminga model with more in the borehole has reached the prescribed value.
complexgeometry(e.g., a three-dimensionalpenny- The analysis then determines the fracture wldth, the
shaped fracture) would introduce Quantitative dif- pressure profile in the fracture, and the stress in-
ferences in the results presented here, but the basic tensity factor at the fracture's tip. By considering
physics of the problem should not be altered. In how the stress intensity factor changes with fracture
terms of an hydraulic fracture propagating outward length and comparing the stress intensity with a
from a vertical well of radius r, the model (Figure value for the critical stress intensity or fracture
3) approximatesa fracture of height h, and overall toughness, we can tell whether this fracture is likely

S$

Figure 3. Two-DimensionalModel used in this Study. Model is Infinite in Vertical Extent


althoughFluid FlowSolutionsRequireFinite Heightto be Specified. A Far-Field Stress, S3,
Acts •lormalto the Fracture Planeand a Variable Internal Pressure,Resultingfrom Fluid Flow
into the Fracture also is Present. The Inset Illustrates the Mannerin Uhicha Varying Pres-
sure Distribution is Approximatedby a Series of ConstantPressuresActinc in Adjacent
Intervals. •

16
Mark D. Zoback, David D. Pollard

to propagate. We recognize that considering a frac- tures, they assume constant pressure acts throughout
ture of changing length which is subjected to the the fracture. Geertsma and de Kleerk (1969) also
same borehole pressure may not seem.reasonable. Wow- considered the coupled elastic deformation-fluid flow
ever, the largest effect a change in borehole pressure problem with a two-dimensional model, but they as-
can have is a proportionate change in the stress in- sumed a constant pressure acted in the fracture to
tensity factor and displacement (Eqn. 2, 3), and arrive at simple formulas for fracture length and
since we are primarily concerned with what is happen- width. Secor and Pollard (1975) and Pollard (1976)
ing as the borehole pressure is building up, assuming used a two-dimensional fracture model to determine
a constant pressure in the borehole merely limits the the form and stability of hydraulic fractures, but in
solution to be valid for borehole pressures close to their work fluid flow into the fractures was ignored
the chosen value. Finally, since a simple slit is and fluid pressure within the fractures was arbitrarily
used to r' resent the fracture, we are, in effect, prescribed as either uniform or linearly varying.
ignorin• .ne borehole geometry and the stress concen- Abe et al (1976) considered the dynamics of linearly
trat• caused by the hole. However, because we in- propagating, three-dimensional, penny-shaped cracks
corporate the stress normal to the fracture, S3, into and assumed that the fractures were propagating at
the model by considering the net pressure in all frac- constant velocity with a "statistically equivalent"
ture segmentsoutside the borehole to be Pi-S3, this fluid pressure acting over almost the entire fracture
is not a serious flaw. For a crack extending from a area. Hsu (1975) evaluates hydraulic fracture pro-
borehole the stress normal to the fracture would be pagation with a two-dimensional model without incor-
at most 2S3 (since the stress concentration is 3S3-S1 porating fluid flow, and chooses to use the stress
and S1 •S3) at the edgeof the hole and this stress intensity formulae of Newman(1971) for two cracks
decays rapidly to S3 in about one borehole radius. growing from a borehole. While such a model seems
desirable, the stress intensity values for fracture
The detailed manner in which the model works is lengths at several wellbore diameters are nearly the
illustrated in the inset of Figure 3. We approximate sameas those used in this work (for the limiting
the pressure distribution which results from the case of pressure in the borehole only). Moreover,
fluid flow into the fracture as a series of different because the stress intensity factors determined by
uniform pressures applied along adjacent intervals of Newmanwere arrived at numerically, it would be quite
the fracture. Using the stress function method of difficult to evaluate the stress intensity and frac-
Muskhelishvelli (1975), the effect of the constant ture width profiles resulting from non-uniform pres-
pressure acting within each fracture interval on the sure distributions.
stress intensity factor at the fracture tip and the
fracture width (for the entire fracture) is computed. Model Application
The appropriate stress functions are presented in
Appendix A. By sun•ningthe effects of the individual In the laboratory experiments of Zoback et al
intervals, the effect of the pressure distributed (1977b), cubical specimenswere uniaxially loaded and
within the fracture is taken into account. The pres- a central borehole (oriented perpendicular to the
sure within each fracture interval is determined after applied stress) was pressurized at a variety of rates
first computing a pressure drop for each interval with with either water or high-viscosity (~100 cp) bento-
a laminar, steady-state, one-dimensional flow law nitic drilling mud. In choosing the model parameters,
(see Appendix B). We can use such flow laws because the appropriate flow rate and borehole radius were
the Reynolds number if small and because the diffusion used(0.0204cm3/secand0.6 mmrespectively), the
time to reach a steady-state pressure profile in the fracture "height" was taken to be the length of the
fractures is quite small. pressurized region of the borehole (5 cm), and the
constant borehole pressure used in the analysis was
The pressure distribution in the fracture and the taken to be the fracture initiation pressure of a
fracture width are highly dependent on one another. particular experiment in which the acoustic emission
To arrive at the pressure distribution in the fracture, data indicated that fracturing initiated at 124 bars,
the fracture's width, and the stress intensity factor although breakdown did not occur until 380 bars.
at the fracture's tip, we iteratively solve the fluid Since the fracture formed parallel to the applied
flow and elasticity solutions. The basic solution stress, zero compressive stress normal to the fracture
procedure involves assuming an initial pressure dis- plane was incorporated into the model. The elastic
tribution and then computing the width along the frac- moduli used were those of the rock in the study, Ruhr
ture resulting from this pressure. The width distri- sandstone, which has a Youn•l'S modulus, E, of 4.2 X
bution is used to compute a new pressure profile 10-5 bars and a Poissonsratio, v, of 0.15.
(while the pressure along the portion of the fracture
correspondingto the borehole is kept constant), From the analysis of the stress intensity at the
which is then used to compute a new width profile, and fracture tip as a function of fluid viscosity and
so on. Convergence with this procedure was usually fracture length (Figure 4) it is clear that for a
•eached in five to ten iterations. given fluid viscosity, a critical fracture length
exists. At lengths shorter than the critical length,
A number of authors have modelled the process of stress inteisity decreases with increased length
hydraulic fracture extension, although without the aim (along the 1000 cp curve), whereas at greater lengths
of rationalizing real pressure-time histories. Per- the stress intensity rapidly increases with fracture
kins and Kern (1961) and Nordgren (1972) considered length. Sincethe latter c•ndition wouldresult in
the coupled problem of fluid flow and elastic rock unstable fracture propagation, it is apparent that
deformation with a two-dimensional fracture model, but for low-viscosity fluid such as water (1 cp), unstable
their elastic solution assumesa uniform pressure fracture propagation is indicated for all lengths.
along the fracture (of infinite length) whereas the For the higher viscosity fluids, however, the initial
fluid flow solution demandsa pressure drop. Like- fracture propagation is stable because of the greater
wise, in predicting the length (or radius) of frac- pressure drop in the fractures. The analysis showed
17
Mark D. Zoback, David D. Pollard

that along the 1 cp curve almost constant pressure fluid pressure) to unstable growth (increasing stress
existed everywhere within the fracture (as in case 2), intensity with length from nearly uniform pressure
and along the 1000 cp curve pressure acted only in distribution).
that portion of the fracture corresponding to the
borehole (as in case 1). DETERMIr,•TION
OFSHmax

The analysis presented above seems to confirm the


hypothesis that stable hydraulic fracture growth can
precede breakdown when high viscosity fracturing
fluids are used. We thus conclude that the presence
of viscous mud has, at least in some cases, led to
erroneously low determinations of SHmax. Haimsonand
Stahl (1970) refer to an experiment in NewYork in
which breakdown pressures in two out of three wells
clearly were anomalously high because, they suggested,
of the presence of drilling mud. To computeSHmax
with Eqn. (1) in cases such as this, sophisticated
down-hole equipment (such as acoustic emission moni-
tors) wouldhave to be used in order to determinePc-
Considerable benefit could be gained from sub-
stantiating the suggestionthat SHmax
can be deter-
mined from secondary breakdown pressures. If such were
0 i 2 $ 4 5 6
the case both the problemsof Pb exceeding Pc and per-
FRACTURE HALF LENGTH, cat meability lowering Pb would be vitiated and core would
not be needed to determine tensile strength.

Figure 6 shows a pressure-time record in which


Figure 4. Stress Intensity Factor as a five pumping cycles were performed in an attempt to
Function of Fracture Half-Length for determine the minimumpressure at which the fracture
Various Fluid Viscosities. Parameters would open (after Zoback et al, 1977a). The pressure
Chosen are for Laboratory Scale Experi- drop preceding breakdown was caused by momentarily
ments (see text). For 10 through lO0 shutting off the pump. Breakdownoccurred at 25.2 MPa
cp,the Stress Intensities Decrease Ini- and the first cycle ISIP was '11.4 MPa. On cycle 2 the
tially along the Curve Labelled 1000 cp, well was shut-in before breakdown actually occurred
then Increase Abruptly. The Position (thus indicating that breakdownwould have been greater
and Slope of the 10, 30, and 100 cp Curves than 13 MPa). Cycles 2 to 5 showa progressive de-
are Approximate Since only Selected Points crease in the breakdown pressure which apparently
along the Curves were Analyzed. levels off at 9.3 MPa. The shut-in characteristics of
this record are discussed below.
In considering the details of the effect seen in
Figure 4, it is helpful to think in terms of how the Breakdown represents the condition when an open
width of the fracture depends on fracture length since fracture accepts fluid faster than the pumpsupplies
the pressure drop at any poing in the fracture is it. Thus, when breakdown occurs the fracture must
inversely proportional to the cube of the width be openat the boreholeandfrom (1) Pb>_3SHmin
-
(AppendixB). Figure 5 presents wall displacementat SHmax- Po- From the view of determining SHmax,
the center of the fracture as a function of fracture
length. The limiting cases I and 2 are presented as SHmax>3SHmin - Pb - Po (4)
well as the data corresponding to Figure 4. The
displacements for 1 cp fluid coincide as expected with so that at the very least, a lower bound estimate can
case 2 (uniform pressure) and the width proportionate- be gotten from Pb. Operational procedures like that
ly increases with length, thereby enhancing fluid flow illustrated in Figure 6 are intended to determine a
into the fracture. Conversely, with the lOO0 cp minimumPb, Pbmin, so that the lower boundestimate of
fluid, since the pressure acted only in the borehole, SHmaxdeterminedwith (4) is close to the actual value.
the increase in crack displacement with length remain
quite small, demonstrating behavior similar to case 1 Two cases are important to consider, when Pbmin<
rather than case 2. It is important to note, how- S3 andwhenPbmin•_ S3. In the formercase, consider•
ever, that for a fracture with a finite pressurized able fluid cannot be pumpedinto the fracture since
interval, the displacements are considerably larger for pressures less than S3 the fracture will be closed
than for the point load case, and that even for crack away from the borehole. A key operational step in
length to borehole diameter ratios approaching lO, such a case is to pumpat low flow rates so that the
the wall displacement still increases with increased moderate fluid volume escaping into the fracture
length. For the intermediate viscosity (10, 20, 100 causes a noticeable change in the rate of borehole
cp) fluids, the critical fracture lengths apparent in pressurization. In Figure 6, note that Pb on cycles
Figure 4 (that at which unstable fracture propagation 4, 5, and 6 is less than Sp•ni
n and after the fracture
begins) are seen clearly to represent the length at opens at the borehole the pumping pressure builds up
which sufficient width resulted to allow flow into to about SHmin, as expected. The pumping rate in
the fracture. Once this process began, the increased these experiments was about 8 liters/min. Had the
width resulting from the distributed pressure caused pumpingrate been considerably greater (as is usually
the abrupt transition from stable fracture growth the case in hdyrofrac operations) a substantially
(decreasing stress intensity with length from confined lower estimate of SHmax may have resulted since a
18
MarkD. ZobacK,DavidD. Pollara o

2.8 ÷

2.6

2.4

2.2

2.0

cAsœ2 •

1.6

1.4

1.2

1.0

iOO0 cp
0.8
(FINITE LOAD

0.6

0.4, CASEI (POINT


LOAD)

0.2
I
/
0 i 1
0 1.0 2.0 :5.0 4.0 5.0 6.0 7.0

FRACTURE HALF LENGTH, CM

Figure 5. Displacement at Center of Crack as a Function of Fracture Half-Length for Computa-


tions Presented in Figure 4 and Limiting Cases 1 and 2.

greaterboreholepressure
wouldhavebeengenerated
in These observations then, suggest a three-step
order to open the fracture to accomodate greater flow. procedure for using secondary breakdown pressures;
(1) pumpat low flow rates, (2) repeatedly pressurize
In cases whenPbmin> SHmin,it is still impor- the borehole to progressively extend the fracture and
tant to pumpat restricted flow rates, but the neces- attempt to detect a decrease in Pb to Pbmin, and
sity of performing many pressurizing cycles, or sub- (3) pumpat still lower flow rates to verify that
stantially propagating the fracture, takes added sig- Pbminhad, in fact, beenobservedin step (2). These
nificance. Figure 5 illustrates that in the presence steps should allow Pbminto be observedand success-
of viscous fluids the center wall displacement is con- ful use of the secondary breakdown pressure method
siderably restricted for small fracture lengths. The for determinationof SHmax.
30 and lO0 cp curves, however, illustrate that beyond
a certain length the wall displacement increases to The observation that after a fracture has been
its maximumvalue. Roughly speaking, the secondary formed the borehole pressure can continue to rise
breakdown pressure will be inversely proportional to leads to explanation of a variety of observed pheno-
crack width. In other words, for relatively small mena. For example, the observation that in the
fractures the secondary breakdown pressure will presence of existing fractures pressurization with
decrease with fracture length. We nowsee that Pb viscous mud allowed generation of a properly oriented
decreased as the volume of fluid into the fracture hydraulic fracture is also explainable in terms of
increases in Figure 6 and fracture length increased. the pressure drop in the fractures. Flow of viscous
Had not the final pumpingcycles been done and what fluid into the existing fracture resulted in a suf-
appears to be the minimum fracture opening pressure ficiently large pressure drop that initiation of the
observed, a lower estimate of SHmaxwould have been properly oriented fracture occurred (this is illus-
computed. trated in Figure 6B of Zoback et al, 1977b). Growth

1 9
Mark D. ZObacK, uavia U. Pol lara

BREAKDOWN
250

200

--
ISIP
LEAST
PRINCIPAL
-,,, MINIMUM
SECON
HORIZONTAL

/
STRESS BREAKDOWN PRESSURE-

-P PSI ..... P SI "' P SI


I--i H--I I t I I---I II I I H---I
I I I I I I I I I I I I I I

0 I0 20 50 40 50 60 70
TI ME (MINUTES)
Figure 6. Hydrofrac Pressure-TimeRecordsRecordedat a Site near SanArdo, California (after Zoback
et al, 1976a). Arrows Indicate the BreakdownPressure, Instantaneous Shut-In Pressure (ISIP), and
Secondary BreakdownPressures. P Indicates Pumping, SI Indicates the Well was Shut-In.

of the properly oriented fracture occurred because the DETERMINATION


OF SHmin
applied stress acting normal to its plane was less.
Another interesting observation is that breakdown has
been observed in wells subjected to either constant or
Shut-in pressures are usually reliable, and es-
cyclic borehole pressurization. This can be explained timates of the least principal stress are generally
in terms of fluid diffusion away from the wellbore
quite good (see the reviews of Haimson, 1977, and
which increases the interstitial pore pressure and McGarrand Gay, 1977). The pressure measuredimme-
results in breakdown at a pressure lower than that at diately after shut-in is used for determination of
which it would have occurred without penetration. the least principal stress because there is typically
Another explanation, however, follows from the above a pressure decay during shut-in. This decay can be
caused by operational problems like pressure leaks in
analysis. If a hydraulic fracture had been initiated
and was in a condition of stable propagation, break- the tubing or packers, or by processes such as con-
down would occur when a sufficient length of the
tinued fracture propagation during shut-in, or fluid
fracture became pressurized. In other words, unstable permeation into the formation. However, complicated
shut-in behavior occurs when the least principal
propagation would occur at some critical length, stress is vertical and fractures are formed between
analogous to the conditions illustrated in Figures 4
straddle packers. In this case the stress concentra-
and 5. Finally, when fracture proppants such as sand tion leading to generation of horizontal fractures
are deposited in a wellbore rather than a fracture
(Kehle, 1964) is not present and a vertical fracture
during massive hydraulic fracturing operations, will form at the wellbore (see Haimsonand Fairhurst,
screening out is said to have occurred. Conditions
which could lead to such phenomenaare recognizable
1970). As this fracture propagates it will "roll
over" into a horizontal plane and propagate perpen-
in terms of the above analysis. Figure 5 clearly
dicular to the least compressive stress. The pressure-
illustrated the marked effect of viscosity on fracture
time record presented in Figure 6 illustrates such a
wall displacement. If overly viscous fluid were
process.
being injected at too fast a rate for a fracture of
given length, insufficient width could result and not
When a well is shut-in, the rate of immediate
permit passage of the proppant.
pressure drop is the rate at which fluids bleed out
into the fracture. The characteristic fast then slow
It is important to realize that while the effect
shut-in decay apparent on cycles 2-5 of Figure 6 is
of high viscosity fluid is singled out in this analy-
sis and discussion, very high flow rates should lead what would be expected if flow were initially through
an open vertical fracture than through this fracture
to similar phenomena(Equns. B (1,2)), and the magni- after it has "closed" and then into a horizontal one.
tude of the least compressive stress will be quite
Thus, the pumping pressures on cycles 2-4 are about
important in determining the characteristics of
fracture propagation. It seems imperative, therefore, the sameas the ISIP (and the magnitudeof SHmin)and
that in economic applications of hydraulic fracturing the long-term shut-in pressures of cycles 3-5 approach
the lithostat. Records such as this have now been
fluid viscosity and flow rate be carefully matched
recorded many times in wells near the San Andreas
against the formation modulus and the magnitude of
the least principal stress. fault where, at least at The depths <300 m, SHmin>Sv.

2O
Mark D. Zoback, David D. Pollard

The manner in which a fracture "rolls over" is Muskhelishvili, N. I., Somebasic problems of the
obviously quite complicated. Recent studies by mathematical theory of elasticity, 334-361,
Pollard (1978) suggest that a single fracture plane Noodhoff, Leyden, Netherl ands, 1975.
may break up into a set of en echelon fractures to Newman,J. C., An improved method of collacation for
accomplish this. The shut-in behavior of such a the stress analysis of cracked plates with
system would be difficult to understand. The ISIP various shaped boundaries, NASATN D-6373, 1971.
values of the 1st and 2rid pressurization cycles shown Nordgren, R. P., Propagation of a vertical hydraulic
in Figure 6 are approximately the same (+ 2 bars) fracture, Soc. Petrol. Eng. Jour., 306-314,
and comparewell with the pumpingpressuFesof cycles August, 1972.
2-4. This value is clearly that of SHmin. Once the Paris, P. C., and G. C. Sih, Stress analysis of cracks,
fracture has propagated, however, the ISIP does not in Fracture ToughnessTesting and Its Applica-
reflect S•min. This, then, again indicates the bene- tions, pp. 30-81, STP No. 381, ASTM, Philadelphia,
fits of inducting multiple pressurization cycles
(so tP • SHmincan be determinedbefore appreciable Perkins, T. K., and L. R. Kern, Widths of Hydraulic
fracture propagation) and pumpingat low rates (so Fractures, J. Pet. Tech., 937-949, September, 1961.
that the pumping pressure is approximately a measure Pollard, D. D., On the form and stability of open
of SHmin}. It also emphasizesthe importanceof es- hydraulic fractures in the earth's crust, Geophys.
tablishing a careful test programto insure that ' Res. Letters, 3, 513-516, 1976.
operational disfunctions do not obscure the nature of Pollard, D. D., The form of hydraulic fractures as
shut-in pressures. deduced from magmatic fractures, 19th Symposium
on Rock Mechanics {this volume), 1978.
REFERENCES Raleigh, C. B., J. H. Healy, and J. D. Bredehoeft,
Faulting and crustal stress at Rangely, Colorado,
Abe, H., T. )lura, and L. M. Keer, Growth rate of a in Flow and Fracture of Rocks, ed. H. C. Heard,
penny-shaped crack in hydraulic fracturing of I. Y. Borg, N. C. Carter, and C. B. Raleigh,
rocks, J. Geophys.Res., 81 {29), 5335-5343, Amer. Geophys. Union, Washington, D.C., 1972.
1976. Rummel, F., and R. Jung, Hydraulic fracturing stress
Bredehoeft, J. D., R. G. Wolff, W. S. Keys, and measurements near the Hohenzollern-gaben struc-
EugeneShuter, Hydraulic fracturing to determine ture, SWGermany,Pure and App1. Geoph., v. ll3,
the regional in-situ field, Picenace Basin, no. 1/2 321-330, 1975.
Colorado, Geol. Soc. Amer. Bull., 87, 250-258, Secor, D. T., and D. D. Pollard, On the stability of
1976. open hydraulic fractures in the earth's crust,
Geertsma, J., and F. deKleerk, A rapid method of pre- Geophys.Res. Letters, 2, 510-513, 1975.
dicting width and extent of hydraulically in- Scheidegger,A. E., stresses in the earth's crust as
duced fractures, Soc. Petrol. Eng. Journal, determined from hydraulic fracturing data,
p. 1571, Dec., 1969. Geologie und Bauwesen,27, H,2, 45-53, 1962.
Haimson, B.C., and C. Fairhurst, In-situ stress Sokolnikoff, I. S., Mathematical Theory of Elasticity,
determination at great depth by means of McGraw Hill, New York, 476 pp., 1956.
hydraulic fracturing, in Rock Mech. Theory and Zoback, M.D., J. Healy, and J. Roller, Preliminary
Practice, Proc. of llth Symp. of Rock Mech., stress measurements in central California using
Univ. of Ca., Berkeley, CA., June 1969, ed. by the hydraulic fracturing technique, Pure and
W. H. Somerton, Soc. Mining Eng. AIME, New York, Applied Geophysics,in press, 1977a.
1970. Zoback, M.D., F. Run•nel. R. Jung, H. J. Alheid, and
Haimson, B.C., and E. Stahl, Hydraulic fracturing C. B. Raleigh, Rate controlled hydraulic frac-
and the extraction of minerals through wells, in turing experiments in intact and pre-fractured ß
3rd Symp. on Salt, Northern Ohio Geol. Soc., rock, Int'l. Jour. RockMech., MininO Sci., in
Cleveland, Ohio, p. 421-432, 1972. press, 1977b.
Haimson, B.C., Earthquake related stresses at Rangely,
Colorado, in New Horizons in Rock Mechanics, APPENDIX A
Proc. 14th Symp. on Rock Mechanics, Eds. Hardy
and Stefanko, ASCE, New York, 689-708, 1973. The complex stress function method involves find-
Haimson, B.C., Determination of in-situ stress around ing two analytic functions, •(z} and9_(z), which taken
undergroundexcavations by meansof hydraulic in the formU -- Re[•_(.z)+ 9_(•z)]solve the fundamental
fracturing, Final Tech. Rpt. to ARPA,Contract biharmonic equation, v•U : O, of two-dimensional elas-
H-220080, Monitored by U.S. Bur. Mines, 1974. ticity theory and also give the appropriate boundary
Haimson, B.C., Crustal stress in the continental conditions (see e.g. Muskhelishvelli, 1975; Sokolni-
U.S. as derived from hydraulic fracturing koff, 1956). For the fracture problem it is convenient
measurements,in The Earth's Crust, Geophys. to map the region outside a slit of length • = 4R in
Mono. 20, ed. J. G. Hickock, Amer. Geophys. the z : x + iy plane.to the region outside a unit
Union, Wash., D.C., p. 576-592, 1977. circle in the c + pe•e plane. The slit lies on the
Hsu, Y. C., Critical borehole pressure for a vertical real axis, centered at the origin. The transformation
hydraulic crack in the presenceof two principal equations are
total stresses, Los Alamos Scientific Lab.
R•, No. LA-6115-MS,1975. z : u(C) : R(C+ C-1) (A-l)
Hubbert, M. K., and D. G. Willis, Mechanicsof hy- • : 1/2{(z/R)+ [(z/R)2 - 4]1/2} (A-2)
draulic fracturing, Pet. Trans. AIME, v. 210,
pp. 153-168, 1957. We consider a uniform pressure, P, acting on the slit
Kehle, R. 0., The determination of tectonic stresses wall from x1 to x2 and on the opposite wall from x4 to
through analysis of hydraulic well fracturing, x3. Onthe unit circle, wherep : l, these points are
J. Geophys.Res., 69 {2}, 259-273, 1964. c1 to c4- The analytic functions which solve this
Lachenbruch,A. H., Depth and spacing of tension problem are
cracks, J. Geophys.Res., 66, 4273-4292, 1961.
21
Mark D. Zoback, David D. Pollard -'

+(c) : (PR/2•i) {2c-11n(c2/•1)


+ (c+c-1-c1-c1-1)
ln[(cl-c)/(cl-l-c)]
- (•+•-1-C2-•2-1)ln[(c2-c)/(c2-1-c)]}(A-3)
•(C) = (PR/2•i) {[4C/(C2-1]ln(c2/C1)
- (c1-C1-1-c2+c2-1) [(1+c2)/(c2-1)]
- (el+el-I) ln[(cl-C)/(c2-1-c)]
+ (C2+C2-1)ln[(c2-c)/(c2-1-C)]} (A-4)
(Muskhelishvelli, 1975, p. 354-358).
Here•(z) : d•/dz : •[m(C)] : u(c). From+(c) and
m(c), the normal displacement of the slit wall, D, is
calculated using
D = (1/2p) Im [(3-4v) +(c)-m(c)+-•[•l/m-"-C•')
-
-•--('•-] (A-5)
(Muskhelishvelli, 1975, p. 192-194)
where p : shear modulus; v : Poisson's ratio; prime
denotes differentiation with respect to C; a bar de-
notes the complex conjugate. The stress intensity
factor, K, for the ends of the slit is calculated
using
K: (2/R)1/2 •'(c), c : + 1 (A-6)
(Paris and Sih, 1961, p. 37). Uponsubstitution of
(A-3) we find
K(•) : [P(2R)l/2/(2•i)][-21n(c2/c1)
• (el-el-1-c2+•2
-1) (A-7)
where the plus and minus signs refer to the slit ends
at x : +2R and x : -2R.

The special case 1 referred to in the text is


found by letting C1 = +l; C2 = -1. The special case
2 is foundby consideringthe limit Ix2 - xll + 0
while specifying that the force F = P(mx2- xll)
remainfinite. In the,c - plane Ix2 - xll :
RIC2+ C2-1 - C1- •1-•1ß Tofind the caseof a
center loaded slit we let c2 + +i; Cl + +i.
APPENDIX B

The pressure within each fracture interval is


determined after computing a pressure drop for each
interval, aPi, with a steady state flow law and as-
sumingthat the width of each fracture interval, wi,
is equal to the width of the interval's midpoint.
Whenwe consider the fracturing process under field
conditions, we use the parallel plate type of flow
law (after Happel and Brenner (1965))

Ap
i : 6•4
HW_•__•p.
aX (B-l)
where Q is the volumetric flow rate into the fracture,
p is the fluid viscosity, ax is the size of the in-
terval, and H is the fracture height.

Whenthe fracturing process under laboratory con-


ditions is considered (and the ratio of the fracture
height to width is very much smaller) we assumethe
fracture has elliptical shape in vertical cross-
section. We then use a flow law like that derived
by Perkins and Kern (1961),

Ap
i - 10.2
HWiQ]•
• Jx (B-2
Comparisonof Eqns. (B-2) and (B-3) showsthem to be
identical except for the multiplicative constant.
(U!U,J/IDõ)

i!

$T•/
--4m i

--; _z.J .

$T-• I/•LL BOR•

ø8 8 •o
•042 m

(.•1)

Maximum Flow

W.ler LIQvi.9 • •jL..-$1)i..er Idoving

!I \ .•_•-•m-•...•o,.

g • o

Zero
FIow•-•
Geometry of Spinner Survey
6540 6550 6560 6570 6580

ß COMPOSITE OF 8 SPINNER RUNS


WNILE PUMPING
ß SiNGlE BACKGROUNDRUN WITH PUI• :• OFF

rp$

ß m
mmmmm ß

me
ß
me
ß
ß ß

• 14' -..-.-.i

OEPTH (m)

(!sd) 3•INSS3•IdI-:•

o
o

(•1•1) 3BI'ISS3Eid 1-33


10.91KPo/$
(9•9
I•i/min}
I•'lø/Po
(.'.'.'.9.?
xIO'e/psi)

o
o

TIME (rain)

EXPERIMENT 50 b (I)
PRIMACORDEXPT IN EE-I 8/28/75
(!sd) 3•JI35531•11-33
i
I ] I I [ I I
I •
N3dO-
ß •

{3.

Id.I --

\
• • X

0.574liter/s(9.1gpm)
- •00

(0diN) 3•JI9SS3t•11-33
0 I ] I i I i i l ,0
o 50
TIME (rain)
EXPERIMENT 51 (3(I)
INDUCED POTENTIAL 8/28/75

EE- I PRESSURE ( M Pa )
o c• o
o

cm

-m :]>__.
.FLOW
RATE

0.5•4
liter/•500
õ0
TIME (rain)
I (.n I

EE - I PRESSURE( psi)
(.md)

GT-2 9450 FEET


I I I I
AUG 6, 1975

'•25- FLOW
RATE
O.41
liter/s(6.5cjpm)

• iOi- / _
TIME (rain)

ti/'r) • m (S,lm:l)
3Hns•3Hd 3o•ns

t 1 I I I I

Gt - 2 9550 FEET
AUG 6, 1975

•25 - FLOW RATE(•41 liter/s (6.5 gpm)

--

--

o
o I 2 3 4 5 6
TIME (rain)

Potrebbero piacerti anche