You are on page 1of 11

A New General Model of Fluid Loss

in Hydraulic Fracturing
A. Settari,

SPE, Simtech Consulting Services Ltd.

Abstract
This paper gives a new fonnulation of fluid loss in
hydraulic fracturing that is much more general than the
classical theory while retaining its simplicity. The model
allows many parameters to vary during filtration and
can, therefore, simulate nonlinear effects.
The model has been validated against laboratory data
for Newtonian fluids and crosslinked gels. The results
show that the finite length of the core, viscosity
screenout, and shear sensitivity are important parameters
that. can be represented by the model. The standard
analysis gives values of leakoff coefficients that will give
incorrect, considerably higher leakoff when applied to
field conditions.
Introduction
The estimate of fluid loss is an important part of a
hydraulic fracturing treatment design. Although the control of fluid loss has improved with the use of modem
fracturing fluids, the size of the generated fracture areas
increases with the size of a job. Consequently, fluid loss
can be important even in low-penneability reservoirs for
large treatments.
For design calculations, fluid loss has been treated in
the past by use of the simplified theory proposed by
Howard and Fast, 1 which expresses the rate of filtration
perpendicular to a fracture wall as a simple function of
leakoff coefficients.
The advantage of this approach, besides its simplicity,
is that it can be directly (if not always correctly) related
to experimental data on fluid filtration obtained in a
laboratory. Apart from the correction of the derivation of
the combined leakoff coefficient, 2,3 very little has been
done to improve the classical theory.
With the recent development of a simulation approach
to fracturing design,4,5 it has been recognized that fluid
loss can be computed directly by solving the basic
multiphase flow equations in porous media. Such an approach is more general and does not have many of the
assumptions that limit the chtssical theory. 6 However,
the computational cost is much higher and the data required to describe the process are difficult to measure.
This paper presents a generalization of the classical approach that includes the effect of several parameters that
are variable in the field. The mathematical fonnulation
includes the model of filter-cake behavior developed by
the author 6 and the results of the work of Biot et al. , 7
which improves the calculation of flow in the reservoir.
The model is then fonnulated numerically, which allows
Copyright 1985 Society of Petroleum Engineers

AUGUST 1985

us to introduce the effects of variable pressure, fluid


viscosity, and different fluids contacting the wall in the
filtration process, in accordance with real conditions during the treatment. Comparison with the experimental
data of McDaniel et al. 8 shows that the model is capable
of exhibiting nonlinear behavior matching the laboratory
data, which cannot be explained in tenns of the previous
simple theory. An important feature of the model is incorporation of the length of the core, which produces
nonlinear behavior and can cause large errors in
calculating the true value of the leakoff coefficient when
the simple fonnulas are used.
The new model retaihs the simplicity of the classical
leakoff theory, although it is more comprehensive and
potentially more accurate than the simulation-type
leakoff calculations, because it is fonnulated in tenns of
measurable variables.
Leak-off Models vs. Simulation
The flow of fracturing fluid into the reservoir can be
described, at least in principle, by the equations of
multiphase flow in porous media. It would thus seem
natural that an improved treatment of fluid loss would
use numerical simulation of flow in the reservoir with the
properties and pressure at the wall (behind the filter
cake) as the boundary conditions. This approach, which
we have taken in our current work,6 is indeed more
general. It is not restricted by the assumption of onedimensional (10) flow, and it includes the effects of
relative penneability and capillary pressure and handles
changing conditions at the fracture face.
However, the simulation approach also has problems.
First, the process of fracture fluid filtration is more complicated than the reservoir multiphase flow. The properties of the invading fluid are greatly different from the
reservoir fluid and are changing with time because of
breakers, temperature changes, and mixing. The fluid
can be miscible with one of the resident fluids. The proper fonnulation would require solution of three-phase
flow (one phase being the fracture fluid) with relative
penneability, capillary pressure, and viscosities changing with time. Even though such a fonnulation and solution is possible, the multiphase data are almost impossible to obtain because of the nonlinearity and instability
of the gels. Consequently, one must make simplifying
assumptions (e. g., the filtrate assumes the properties of
the reservoir water).
On the numerical level, an extremely fine grid would
be required owing to usually very small penetration of
the fracture fluid. To avoid this, we have found it
491

ILR' The fluid loss additives and/or high-molecularweight polymers may form a layer of filter cake on the
fracture wall, as shown in Fig. 1. If the fluid pressure is
Pf and the original reservoir pressure P R , the total
pressure drop, tl.p=Pf-PR, is a result of the flow
resistance of the cake and pressure drop in the porous
medium.
The classical theory treats the problem as three
separate regions.

FILTRATE

RESERVOIR

1 -I-

Filter cake. In the absence of the porous media, the flow


velocity through the cake, u w , and the filtration volume
per unit area V w, are given by a square-root relatio~
ship, V w This relation is found experimentally 10
filtration tests with thin wafers. It can be derived from
the assumptions that (1) the deposition of the cake is proportional to the volume passed through the surface; (2)
the permeability of the layer is independent of its
thickness; and (3) flow through the cake obeys Darcy's
law. Then, according to Fig. 2a,

Jr.

dV w

kw

I1pw

=-=_._-=
w
dt
ILf U w

k wexl1pw

........... (1)

IL f V w

from which we obtain the expression 8

\
\

\
\

----I'filt
~---------------I'

x
Fig. 1-Representation of regions in the analytical leakoff
theory.

necessary to track the invaded zone on a subgrid scal~ in


a piston-like displacement, which resembles the classIcal
leakoff theory .
Although direct coupling with the simulation of reservoir flow is necessary for high-Ieakoff applications (fractured water injectors, steam fractures), its accuracy for
typical fracturing application depends on adjusting the
data that cannot be measured to agree with laboratory experiments. In addition, the. mechanism of filter-cake
buildup is always expressed empirically, and there is little difference between the simple theory .and simulation
if the leakoff is controlled by the filter cake (coefficient
C w )' All this gives a good argument for revisiting the ID
leakoff theory, which is the subject of this paper.
Classical Leakoff Theory
Before the new model formulation is presented, it is
useful to summarize the standard theory because the
development of the equations is sketchy in the literature
and the underlying assumptions are not always
appreciated.
Consider filtration of a fracturing fluid with viscosity
ILf into an infinite porous medium ~ith ~rmea?ilit~ k
and porosity cb, occupied by a reservOIr flUid of VISCOSIty
492

and
V w=2C w.Jt+ V spt ' . . . . . . . . . . . . . . . . . . . . . . . (2b)

where ex is a constant of deposition (ex units of volume


passed will deposit unit volume of cake). Spurt loss,
Vspt. is the integration constant and is equi~alent to the
assumption that volume V spt does not deposit cake.
Filtrate Zone. The flow in porous media is divided into
two zones: the part invaded by the fracture fluid and the
part with the original reservoir fluid. To calculate the
flow through the invaded zone, one assumes (Fig. 2b)
(1) constant pressure difference, tl.pv =Pw -P c; (2)
piston-like Darcy displacement with 100% saturation of
fluid filtrate of constant viscosity, ILfilt; and (3) incompressible fluid and rock. These conditions are approached if there is no filter cake and the resident fluid
resistance is small (gas reservoir or evacuated core) such
that the backpressure is constant. Then the flow through
the region is instantaneously steady state, and from
k

tl.pv

dV v

U v=-'-=-

Milt

..................... (3)

dt

and the equation for the fluid front movement


U v =cbVv =cbluvdt, ....................... (4)
SOCIETY OF PETROLEUM ENGINEERS JOURNAL

~Lw= Vw/cx

one obtains

r.

uv=C)'Vt,wlthC v =

~kcf>APv
---,

............ (5a)

2t-tfilt

, .

Pf

and
Vv =2C v .Ji. ............................. (5b)

Reservoir zone. To calculate the flow in the uninvaded


zone, one assumes (Fig. 2c) (1) constant pressure, Pc>
and therefore constant Ap c = Pc - P R; (2) compressible
flow with constant total compressibility, C T; and that (3)
position of the front, x e, does not change with time.
Then one can consider the position of the front to be the
face of the porous medium and apply the well-known
analytical solution to calculate the flow rate at the
face. 9,10

......J

il
......J

and

I
I
I

.'

Pw

I ' ..
I
, '

~
::-,..-----Pc

Ve =2Ce .Ji. .............................. (7)


x

It should be noted that to apply Eq. 5 or 7 one must

assume infinite medium. Also, the last condition for C e


cannot be physically satisfied and will be discussed later
in relation to the work of Biot et al. 7
For calculations, one knows only Pi and P R and is interested in u w' Because of the incompressibility assumption, we have

Uw=Uv=Ue(Xe)=U ....................... (8a)

-~=---==~~---PR

and
x

These conditions can be used to express the fluid loss in


terms of the combined leakoff coefficient

Fig. 2-a. Flow through filter cake. b. Flow through the zone
invaded by the filtrate. c. Flow in the reservoir zone.

U=C/.Ji, V=2C.Ji. ........................ (9)

In the case without filter cake, Williams 2 derived the


correct form,
2C e C v
C ve = - - - - - -

Cv+.JC~+4C~ , .................. (lOa)

(Note that Eq. lOa is quoted with error in Ref. 3). A


similar formula can be derived for the general coefficient
C wve , if the spurt volume is ignored (Eq. 6.27 of Ref. 3).
McDaniel et al. 8 give the corresponding formula for
C wv when C e is not important (i.e., laboratory experiments) and the equation for conversion from C wv to
C wve

where Ce and Cv are the nominal coefficients, in which


overall Ap is substituted for Ape and Apv. Eq. lOa is an
improvement over the previous harmonic mean,
C ve =

CeC v
A

Cv+C e
AUGUST 1985

(1 Ob)

However, all these equations implicitly include all


the previous assumptions, which greatly restrict their
accuracy. Because they are all of the form V - C .Jt.
they are not able to explain nonlinear behavior in .Ji, frequently found in laboratory tests. In such cases determination of C w from laboratory data is not possible and
the values quoted are quite arbitrary.
493

The General Fluid Loss Model


The new model is developed by considering the flow
resistance of the three regions in Fig. 1 as a function of
several additional variables, and then solving for the
common velocity and unknown pressure drops in each
region by use of Eq. 8. Although it is not necessary, we
have formulated the equations in terms of the leakoff
coefficients for easier understanding and relating to data.
Flow Through Cake. The development of the equations
for the filter cake follows our earlier work. 6 The
resistance of the cake is expressed by the transmissibility, T w, according to
UW

=Twt:..pw' ............................ (11)

If C w is the leakoff coefficient measured in a static test


with a pressure differential t:..p~ and fluid of viscosity
ILl, then transmissibility in test, T~, is

where

if w is a "modified volume"

V w = 0 if V w < V spl ;

and

Thus, T w is infinite until the filtration volume reaches


V SPI ' Eq. 15 can be numerically integrated for any arbitrary variation of t:..pw and f-tf to give the filtration V(t)
(neglecting the resistance of the porous media) if the
physical mechanisms, Eqs. 13 and 14, are specified. The
requirement of T w independent of t:..p suggests a function
fk=t:..p;1 t:..pw for t:..pw >D.pmin

and
T~ =2C w 2 I(t:..p~ V w)' ..................... (12)

The following mechanisms will modify T w for conditions different from the test.
1. Cake erosion. This effect is well documented 2.11,12
and is measured by a minimum (steady-state) velocity,
Us, which corresponds to an equilibrium cake thickness.
The value of Us obviously depends on the shear rate of
the flow along the fractllre face, but systematic data are
lacking in the literature. The data of McDaniel et al. 8
can be interpreted as supporting this effect, as will be
discussed later. Equilibrium is reached when the volume
of filtrate reaches Vs =2C wlu s and the transmissibility
for V w > Vs is constant, which is obtained by using Vs
instead of V w in Eq. 12 ..
2. Cake compaction. In the early measurements it has
been observed that filtration is approximately independent of t:..p w until D.p w is decreased to 100 to 300 psia
[690 to 2068 kPa]. This can be explained by considering
kw in Eq. 1 a function of t:..pw such that kw -lIt:..pw' In
general, we can write

3. Viscosity effect. As Eq. 1 indicates,

Uw

is a function

of ILf' and therefore

where the viscosity in the fracture ILf may vary widely


from the laboratory value ILl because of different shear,
temperature degradation, etc.
The above effects are introduced in Eq. 12 to give

and

From the assumption of Darcy flow,


fIL =f-tJlf-tf

These functions are currently used in the model, for lack


of better data. Note that the filtration of the polymers
should be a function of the permeability and pore structure of the porous media as indicated by the theory of
polymers 13 and laboratory evidence. 14 Therefore, the
C w data should be obtained with porous material
representative of field conditions.
Recent laboratory measurements 15 suggest that the
filter cake is compressible at low pressure differentials
and becomes incompressible at high D.p, contrary to the
older data.
Eqs. 15 a~d 16 can be written in terms of a variable
coefficient, C w , which by Eq. 11 is
Cw=Twt:..pJt.

Flow Through Invaded Zone. The flow in the filtrate


zone is much more complicated than suggested by Eq. 5.
Because of the usually small penetration distance, mixing and diffusion can play a role even though the
displacement usually has a favorable mobility ratio. The
dispersion zones in porous media 16 are known to be on
the order of magnitude of t:..L v The polymers will be
screened off on entrance, and the filtrate viscosity, IL fill ,
will generally be lower than /Lf. When the fracture fluid
is not miscible with the reservoir fluids, relative
permeability and capillary pressure can also playa role.
Water-based fluids will also mix with the connate reservoir water. Finally, 'the displacement will be under varying pressure because of changes in p f and t:..p w.
These effects can be introduced in the flow equation,
Eq. 3. The effects of relative permeability are approximated by using the effective properties ke and e in
place of k and . The effects of dispersion and mixing
can be accounted for by using a "mixed" viscosity
IL M filt

494

= f(f-tfilt, f-tmix'

t:..L mix ), ............... (17)

SOCIETY OF PETROLEUM ENGJNEERS JOURNAL

10

...a:

e:J

...

...J

w
w

- - PRESENT MODEL
- - - STANDARD THEORY

II

a:
0

II

.---...,

I
II

P - CASE A
v

L____ _

4
MIXING
VOLUME

..

I.

Fig. 3-Representation of flow in the filtrate zone for the


general model.

where Mmix is a mixing-zone length (Fig. 3). In addition, we will consider the practical case when the wall is
contacted by several batches of fluids with distinct
viscosities /.t 1, /.t2 ... , as shown in Fig. 3, and arbitrary
variation of tlp v .
As derived in Appendix A, the filtration velocity ~an
then be expressed in terms of a variable coefficient, C v:
uv=Cvl.Jt=C v

(~)I.Jt,
C avg

............. (18)

where C v is the instantaneous coefficient

Cv(t)=~ kecPetlPv(t) ,

.................... (19)

2/.te(t)

C avg is an average C v , and /.te is an effective viscosity. The ke and cPe are adjusted for mobility and saturation of the displacing fluid. The coefficient C avg is
simply

C avg =VvI2.Jt,
where

.......................... (20)

Vv is again a modified filtration volume

II
tim . min

II

10

12

1l2

Fig. 4-Effect of step change in pressure (A) and of finite


length of core (8) on leakoff.

If all the variables are constant, Eq. 18 will reduce to


the standard expression. However, with changing conditions, integration of Eq. 5a with instantaneous C v will
give erroneous answers. For example, the effect of a step
change in tlp v and of the length of the core are shown as
Cases A and B in Fig. 4. The core and fluid data used
here are the same as for Test 3 discussed in the next section. Note particularly the large error that results if we
neglect the finite length of the core.

Flow in the Reservoir. The correct solution of this flow


region is more complicated because it is transient. The
two major assumptions made to obtain Eq. 6 are the stationary boundary and constant pressure Pc. Recently,
Biot et at. 7 obtained the solution for the correct formulation with moving boundary between the regions. Under
the standard assumptions about other variables as in the
preceding section, they derived an expression for C ve in
the case of constant tlp,

C ve

=Cv~ ~

1+~

, ......................... (21)

and

v v = V core

if V v > V core'

. . . . . . . . . . . . . . . . . . . (21 )

and V core is the filtration volume per unit area at which


the front Xc reaches the end of the core,

The effective viscosity is a simple average of viscosities


/.ti weighted by the volume filtered for each fluid, as
derived in Appendix A.
AUGUST 1985

where

is a function of

and

They claim that Eq. 21 gives much higher leakoff than


the old formula (Eq. lOb).
495

0.'

f,- EQUATION

where k rf is the average relative permeability to the


filtrate in the invaded zone. The effects of capillary
pressure are minimized in most cases by the large
pressure differences created during fracturing.
In the absence of data on the filtrate, the values of S her
and krf can be estimated from wateiflood data for waterbased gels.
Other limitations result from the assumption of ID
flow and infinite extent of the reservoir in the direction
perpendicular to the fracture face. These limitations will
be important only for very high leakoff situations
(waterflood fractures) and are not important in fracturing
treatments .

bObl

F2 - EQUATION ('001

F,-BIOT ET AL,EQN(,91

0' 1L-_-'--..L-..L..J...LL..I.J.:'.,.....---''--.L......JL..J.....J....I...LLL_--'---I--L...L.Ji..J...L.I.J
.01

0.1

10

Implementation and Numerical Solution


Fig. 5-Comparison of three methods of calculation of Cve.

While this is true, our analysis in Appendix B shows


that most of the difference seen is between the Eqs. lOa
and lOb. Biot's formula (Eq. 21), in fact, agrees very
well with Eq. lOa except in cases when leakoff is
dominated by Ce and cTt:..Pve is large (Fig. 5). As shown
in Appendix B, Biot's solution can be used to correct Eq.
lOa for the moving boundary if C Tt:..P ve < 1 by replacing
Ce by

C =C .
e
e
-

7r

3(I-c Tt:..Pve)

J'h

. ............. (22)

Use of Eg. 2? gives the same leakoff as Eq. 21 in the


lim!! for Ce/C v ->0. In addition, we have found that use
of C e in Eq. lOa approximates very accurately the more
complicated formula (Eq. 21) in the range of practical
data.
The effect of variable pressure is not as easy to correct.
However, this is important only if Ce controls the filtration (C e ~ Cv' C w), which is rarely the case in practice.
In any case, our calculations presented in the following
section, as well as laboratory experience, suggest that t:..p
will stabilize under constant conditions.

Limitations of the Model


The model just described is limited by the assumptions
remaining in its formulation. One of the more important
of these is the assumption of piston-like displacement in
the invaded zone, which replaces the more realistic
model of two-phase Darcy flow with capillary and
relative permeability description. To assess the accuracy
of this approximation, it would be necessary to obtain
detailed saturation profiles of injected filtrate. If these
were known, the present model could be adjusted by using effective porosity to filtrate as

where S we is the connate water saturation and S her is the


average residual hydrocarbon saturation left in the invaded zone. Similarly, the effective permeability will be

496

At any time, the filtration velocity u and pressure drops


across the regions must satisfy Eqs. 8a and b. Because of
the nonlinearities now introduced in the coefficients T w
and C v , analytical expression is no longer possible. The
solution procedure is as follows.
1. At time tn, we know the pressure drops t:..p!, t:..p~,
and t:..p e' total filtration volumes V! and V~, fluid
pressure Pf' and viscosity /J-f for the next time increment
t:..t.
2. Calculate T!+l and C~+l, which satisfy the
equations

and
Un
w+1 =Tn+l
w....Apn+l
w

n+1 / rt,
=u ve
n+l =C ve"
t ..... . (23b)

where T!+l is given by Eq. 16 and


C~+l

2C~+lC~+1
= _________
_

Cv+v'C~+4C~
-

............ (24)
-

In Eq. 24, C v is given by Eq. 18 and C e by Eq. 22,


with the overall pressure difference t:..p~+l instead of
t:..p v and t:..p e. Because the terms in Eq. 23b depend on
the unknown boundary pressure, Pw, and volumes,
V n + 1 (through the constraints on V wand V v ), the solution must be obtained iteratively.
3. Integrate the loss volume,

V~

t n+1

+ 1 = V vn
.

udt.. ................... (25)

Apart from the necessity of the solution of Eqs. 23, the


model retains the simplicity of the classical leakoff
calculations. On the other hand, admitting this complication allows us to vary, during the calculation, a number
of important parameters in a multistage job: fracturing
pressure, Pf; fluid apparent viscosity, /J-f; wall-building
properties of each fluid; and viscosity reduction by a
cake (/J-filt//J-f). In addition, the constraint ofEq. 21 takes
into account the effect of the length of the core.
In our numerical implementation, Eqs. 23 are solved
by a Newton-Raphson technique. The integration of Eq.
25 is done by a trapezoidal rule except for the first
SOCIETY OF PETROLEUM ENGINEERS JOURNAL

TABLE 1-DATA FOR THE SIMULATION OF


LABORATORY TESTS

TABLE 2-PARAMETERS USED FOR


MATCH OF DYNAMIC DATA

Core
Test

1
2
3

(cp)

</>

k
(md)

361
262

0.2
0.2
0.2

0.10
0.28
0.20

III

Fluid
Glycerol
Glycerol
Crosslinked gel

Match by

Match by f~

Us

L
(cm)

ilp
(psia)

Shear rate
(sec -1)

(ft/min)

(ft/min)

6.0
1.0
2.0

1,000
550
300

0
41
82
123

0.0
0.00066
0.00087
0.0011

1.0
1.0
1.0
1.0

0.0
0.00032
0.00032
0.00032

1.0
1.93
2.47
3.15

timestep, where the singularity is integrated out, which


yields VI =2C I ill. The iterative scheme is very dependable, and the model has since been successfully integrated into a general-purpose hydraulic fracturing
simulator.
Testing of the Model
In the first instance, the model was tested against the
results obtained previously by numerical simulation, 6
with the results being very similar. Because of the simple
nature of the examples, these tests did not involve any of
the nonlinearities built into the model.
For comparison with actual laboratory data, we have
used the data of McDaniel el al. 8 They measured the
filtration of linear fluid as well as crosslinked polymer
under various shear rates and used cores of different
lengths. We have simulated two tests with Newtonian
fluid (glycerol) and one test with crosslinked natural
polymer. The pertinent data are summarized in Table 1;
the experimental results are those of Figs. 2 and 3 of
Ref. 8.
Tests 1 and 2. Because glycerol is a Newtonian fluid
and does not form a cake, one would expect the results to
follow the classical theory. This is observed for Test 1 at
ilp=I,OOO psia [6.89 MPa], which is matched by the
model with C w =0 as shown in Fig. 6. The fluid front is
still within the core at the end of the test and the behavior
is linear in .Jt.
The second test was conducted with a V2-in. [1-cm]
core, which gives a PV, V core' of about 2.3 mL. The
leakoff should therefore be linear with I for V> V core,

Us

Us

and this trend can be seen in the later part of the curve.
The core length alone cannot explain the large initial rate
of leakoff, which could be produced in the model only
by assuming that initially the effective viscosity is much
lower.
Once this is recognized, a reasonable agreement can
be achieved by several assumptions. Fig. 6 shows three
different simulations: (1) initial batch of a fluid with
P-f= 10 cp [0.01 Pa' s] followed by the test fluid; (2)
larger batch of P-f=26 cp [0.026 Pa' s]; and (3) mixing
zone with P-mix = 1 cp [0.001 Pa' s], because the sample
was saturated initially by water, and V mix =2Vcore '
The last of these gives slightly lower filtration, but the
trend ofthe curve follows the data very well. The results
clearly indicate that some kind of mixing or dilution
mechanism is present, which can be included in the
leakoff model in a very simple fashion.
Test 3. The filtration is controlled in this case by the
filter cake and less so by the core, and there is a marked
dependence on shear rate.
To begin with, the data for zero shear are linear
in .Jt,
which agrees with C w theory with Us =0 (no cake erosion). To match the test, it was necessary to assume that
the filtrate viscosity was completely screened out and
equal to water viscosity. We have learned subsequently* that the same conclusion has been reached experimentally. The match gave a value of C w = 0.002 ftf
..fmin, [0.00001 mf../s], which was then kept for all
runs with shear.
"A. Deyserkar, private communication. Dresser Titan, (Houston), 1982.

,>

/2
7.5

60

TEST

DATA
50

DATA

----

SIMULATION,
VAAIAB lE Us

ME 5,0

1 - - - --

BATCH OF 10 cp

2-----

BATCH OF 28cp

3- - -

MIXING

40

ZONE

"3o
>

20

1
SHEAR. sec-

10

10
12
time,mtn'l2

14

16

18

Fig. 6-Filtration of Newtonian fluid-comparison of data


and simulation for Tests 1 and 2.
AUGUST 1985

10
ttme. mln

1/2

12

14

16

18

Fig.7-Filtration of crosslinked gel-comparison of data


and simulation by variable us'
497

TABLE 3-DATA FOR THE EXAMPLE OF A


THREE-STAGE FILTRATION

/'23
I.

TEST 3

P,-PR

Cw

(psia)

III

Illilt

Stage

(cp)

(cp)

(ft/min '(2)

Pad
Fluid I
Fluid II

600
800
900

1.0
20.0
100.0

1.0
10.5
1.05

0.01
0.001

!J

- - DATA
50

-- -

SIMULATION.

VARIABLE

~f

40
~E
u

w
,. 30

!lp0= 1.000 psia. f!.P'in= 100 pSia, Us =0.0005 Itlmin,


f!.L mix = 0.0, and cross-sectional area = 1 sq It.

20

10

The behavior with shear is physically much more complicated. It was found by numerical experiments that the
observed dependence on shear rate could be matched in
two ways.
1. One can assume that the steady-state filtration
velocity, Us, depends on shear rate. The results of the
match are shown in Fig. 7, and the values of Us are in
Table 2. The assumption is reasonable, but the results
cannot be considered conclusive because of the strong
shear sensitivity of the gel. To isolate the mechanism of
the cake buildup, data similar to these should be
measured with linear wall-building fluids, which is experimentally difficult.
2. One can assume that the coefficient C w itself is a
function of shear rate in the fracture. This can be
justified in several ways. One possible argument is that
the material of the cake will have different structure and,
therefore, different kw, if the fluid was sheared at different rates. Another possibility is the effect of the shear
on the flow through the cake, where at least at the entrance the fluid will be nonlinear. We have simulated the
effect by a constant nominal C w equal to that of the static
test, and by varying 11/ according to the shear. Fig. 8
shows the results of a match. Note that the values of apparent viscosity that can be calculated from the theory of
non-Newtonian flow in porous media 17 would be on the
same order as I1ffor the match in Table 2.
While the results in Figs. 7 and 8 have similar
character, the pressures behind the cake, Pw, shown in
Fig. 9, are more revealing. The stabilized pressures at

10
12
tim.:min'l'l

18

14

18

Fig. 8-Filtration of crosslinked gel-comparison of data


and simulation by variable Cwo

the end are similar for both hypotheses, but the early
pressure history is completely different. The measurement of pressures in addition to filtrate volume will be
important and should be included in future laboratory
work.

Model Application
The cases shown in the previous section show the ability
of the model to represent nonlinear filtration. Also, they
bring out the peril in characterizing the process in terms
of a single number, as is the practice in the service industry. For the case of Test 2, fitting a straight line
through data gives apparent C v almost twice the actual
value, as a result of the end effect. Use of this coefficient
for field calculations would then give larger leakoff,
while the present model will automatically correct for the
difference. Similarly, the values of the leakoff coefficient quoted in Table 2 of Ref. 8, which have been obtained by fitting the ends of the curves, will give considerably higher fluid loss compared with the data if applied for the entire time period (for example, 100 mL vs.
65 mL for the highest shear rate).

TEST 3
150

100

- - VARIATION OF

I'f

- -

u.

VARIATION OF

----,

- - - - - - - - - - - - - - 1~3

FLUID J

2000
41

50

--'" ~
~---- 0

-.

~
W
ti ... ,M'fl

12

SHEAR,Me"'

'. ..

Fig. 9-Pressures behind cake for the two methods of matching the f,ltration of crosslinked gel.

498

I
I
I

~
~.

-------------- u

, _____________

I
I--1

~
FLUID.

.02

I
I

:I

i
z

II'\~."'''-:::.:::::::-:=.::=.-:-,,------

"'Ci:I

Fig.10-Example of calculated fluid loss for a multifluid


treatment.
SOCIETY OF PETROLEUM ENGINEERS JOURNAL

Use of the model allows us to deduce the basic values


of C wand IL filt, which can then be used to calculate
filtration at conditions different from the test, provided
that the variation resulting from other parameters is
calibrated by laboratory data. The amount of dynamic
laboratory tests necessary for a given fluid could potentially be greatly reduced if the dependency on shear rate,
temperature, etc., is established with the help of matching with the model.
Once the filtration is characterized in terms of the
basic properties from laboratory tests, the same model
must be used to compute loss in the field. As an example
of the generality of conditions possible, Fig. 10 shows a
hypothetical case of a fracture wall contacted by three
fluids. The data, shown in Table 3, resemble a pad
followed by two fluid stages. The model predicts
simultaneously the filtration volume, velocity, and
pressures between the regions, which all vary during the
job.
Conclusions
A new mathematical formulation of fluid loss has been
developed, implemented numerically, and tested against
laboratory data. The main results of the theoretical
analysis are as follows.
1. The new model removes many of the assumptions
restricting the applicability of the classicalleakoff theory
while it retains its simplicity. The formulation allows
time variation of pressure, fluid viscosity, and filtration
characteristics during the process and models the effects
of finite core length, viscosity screenout and mixing, and
cake erosion.
2. The model can produce nonlinear behavior, which
cannot be obtained from the classical theory, and thus
can be used to analyze the laboratory data.
The main results of the comparison with the data of
McDaniel et al. 8 are as follows.
1. The finite length of the core causes large deviation
from linear behavior in.Jt. Standard analysis will predict too much leakoff in the field while the present model
yields values unaffected by the length of the core.
2. Leakoff sensitivity to shear rate can be obtained in
the model by variation of cake erosion or of the effective
coefficient. More laboratory research is required to
determine the mechanism.
3. The difference between the fracture fluid and filtrate
viscosity resulting from screenout is important and can
be obtained by matching laboratory data.
The model is suitable for integration in fracture design
programs and it is computationally extremely efficient.
Acknowledgment
The author thanks A. Deyserkar and R. McDaniel of
Dresser Titan for their discussions and input during the
course of this study.
Nomenclature
Note: All volumes are per unit cross-sectional area except in results of laboratory experiments.
CT

Cc
Cv

= total reservoir compressibility


= reservoir flow coefficient (C n )
= filtrate displacement coefficient

AUGUST 1985

C vc

= combined leakoff coefficient for the porous

media
C w = wall-building coefficient (C m )
F 1,2,3 = functions defined in Appendix B
k = reservoir permeability
kw = filter-cake permeability
!:.L core = core length
!:.Lv = depth of the invaded zone
Pc = pressure at the filtrate front
Pf = pressure in the fracture
P R = initial (undisturbed) reservoir pressure
P w = pressure at the face (behind cake)
t:..p = pressure drop
P = dimensionless compressibility, C p1.p
R = ratio CclC v
t = time
T w = filter cake transmissibility
u = Darcy flow velocity
V = filtrated volume per unit area
V c = fictitious volume per unit area of reservoir
fluid flowing through the cross section Xc
V mix = "mixing" volume
Vs = spurt volume per unit area
V v = volume of filtrate in the invaded zone per
unit area
V w = volume passed through cake per unit area
ex = cake deposition constant
IL f = fracture fluid (apparent) viscosity
ILfilt = filtrate viscosity after passing through cake
ILmix = "mixing" viscosity
IL R = reservoir fluid viscosity
~ = parameter defined by Eq. B-1
4> = reservoir porosity
Subscripts
e = effective
f = fracture
filt = filtrate
hcr = hydrocarbon residual
i = batch of filtrate fluid with distinct properties
N = total number of fluid batches
R = reservoir
spt = spurt
v = invaded zone (filtrate)
w = filter cake (wall)
Superscripts
M = mixed
n = time step number
- = modified value to account for additional
variables
= value of leakoff coefficient related to the
overall pressure drop, Pf-P R
o = reference or laboratory value
References

(C I )

1. Howard, G.C. and Fast, C.R.: "Optimum Fluid Characteristics


for Fracture Extension," Drill. and Prod. Prac., API, Dallas
(1957).
499

2. Williams, B.B.: "Fluid Loss From Hydraulically Induced Fractures," J. Pet. Tech. (July 1970) 882-88.
3. Williams, B.B., Gidley, J.L., and Schechter, R.S.: Acidizing
Fundamentals, Monograph Series, SPE, Richardson, TX (1979)
6.
4. Settari, A.: "Simulation of Hydraulic Fracturing Processes," Soc.
Pet. Eng. J. (Dec. 1980) 487-500.
5. Settari, A. and Cleary, M.P.: "Three-Dimensional Simulation of
Hydraulic Fracturing," J. Pet. Tech. (July 1984) 1177-90.
6. Settari, A. and Price, H.S.: "Simulation of Hydraulic Fracturing
in Low-Permeability Reservoirs," Soc. Pet. Eng. J. (April 1984)
141-52.
7. Biot, M.A., Masse, L., and Medlin, W.L.: "A Two-Dimensional
Theory of Fracture Propagation," paper SPE 11067 presented at
the 1982 SPE Annual Technical Conference and Exhibition, New
Orleans, Sept. 26-29; to be published in Jan. 1986 Production
Eng.
8. McDaniel, R.R. et al.: "An Improved Method for Measuring
Fluid Loss at Simulated Fracture Conditions," Soc. Pet. Eng. J.
(Aug. 1985).
9. Collins, R.E.: Flow of Fluids Through Porous Materials, Van
Nostrand Reinhold, New York City, 1961.
10. Carslaw, H.S. and Jaeger, J .C.: Conduction of Heat in Solids, second edition, Clarendon Press, Oxford, 1959.
11. Hall, C.D. and Dollarhide, F.E.: "Effects of Fracturing-Fluid
Velocity on Fluid-Loss Agent Performance," J. Pet. Tech. (May
1964) 555-60; Trans., AIME, 231.
12. Hall, C.D. and Dollarhide, F.E.: "Performance of Fracturing
Fluid Loss Agents Under Dynamic Conditions," J. Pet. Tech.
(July 1968) 763-68; Trans., AIME, 243.
13. Stright, D.H. Jr.: "The Use of Polymers for Enhanced Oil
Recovery, A Review," Applications Report AR-2, Petroleum
Recovery Institute, Calgary, 1976.
14. Gulbis, J.: "Dynamic Fluid-Loss Study of Fracturing Fluids,"
paper 82-33-18, 33rd Annual Technical Meeting of CIM,
Calgary, 1982.
15. Roodhart, L.P.: "Fracturing Fluid, Fluid-Loss Measurement
Under Dynamic Conditions," paper SPE 11900 presented at the
1983 SPE European Petroleum Conference, Aberdeen, Sept. 6-9;
to be published in Oct. 1985 Soc. Pet. Eng. J ..
16. Perkins, T.K. Jr. and Johnston, O.c.: "A Preview of Diffusion
and Dispersion in Porous Media," Soc. Pet. Eng. J. (March 1963)
70-84; Trans., AIME, 228.
17. Ikoku, C.U. and Ramey, H.J. Jr.: "Transient Flow of NonNewtonian Power-Law Fluids in Porous Media," Soc. Pet. Eng.
J. (June 1979) 164-74.

APPENDIX A
Derivation of the General Coefficient C v
Let us first consider displacement with constant ilp v ,
where batches of several fluids are introduced with
viscosities /LI, /L2" /LN, as shown in Fig. 3. If we
denote by VI, V 2 ... V N the filtered volumes per unit
area of each fluid, and assume that the batches do not
mix, we can calculate the distances LlL I ... LlL N, and
pressure drops are given by Darcy's law,

where U v is the common velocity. We now want to find


an effective viscosity that will give the overall pressure
drop, ilpv, over a distance LlLv:

cf>k

.............. (A-2)

This can be expressed more conveniently by a recurrence


relationship,

where V N- I and (/Le)N-I are the volume and the effective viscosity at the end of the previous batch. Thus, only
the cumulative values need to be saved. It is easy to See
how Eq. A-3 can be generalized for a continuous change
in viscosity.
To express U v in the form of Eq. 9, we write Eq. A-2
as

=c v

/Le=
500

(~/LY)lVv .

.................... (A-3)

C avg

............. (A-5)

Note that this equation is valid for any ilp v and therefore
is general, provided C v is evaluated with the current
ilpv' However, if the length of the porous media is
finite, Eq. A-2 is valid only for LlLv < LlLcore' or
V v < V core' and for larger time it must be replaced by
uvLlLcore/Le
ilp = - - - v
k

U v Vcore/Le

cf>k

Both cases can be treated by Eq. A-5 with a modified


volume

and
Vv=Vcore if V v > Vcore'

. . . . . . . . . . . . . . . . . . (A-6)

Finally, we will describe a simple method of adjusting


the viscosity /Lfilt for mixing. We define a "mixing
length," LlLmix' and the "mixing viscosity" of the fluid
ahead of the front, /L mix' Then the mixed viscosity of the
displacing fluid, /Li\\t, will change continuously from
/L mix when V c = 0 to the full viscosity /L filt at
Vv = V mix = LlLmixlcf>. We used a simple linear mixing:

The viscosity /Lmix may be reservoir water viscosity or a


viscosity of a previous stage when mixing of stages is
considered. Different mixing rules (such as the onequarter-power rule) may be used for miscible fluids. The
concept of the mixing can also approximate the effects of
multiphase flow (i.e., changing mobility).

APPENDIX B
Comparison of Formulas to Calculate C vc
We wish to compare Eqs. lOa, lOb, and 21, in which

~=
N

(~)/Jt=Cv/Jt.

1/2

[CTilP vc -1 + [(CTilp vc -1)2 +

8CTil~VC/Lfilt ]

V2]. ..................... (B-1)

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

For this purpose, it is convenient to write them as


C vc = Cv . F, and compare the functions F, which will be
functions of two parameters:

and

quires either high reservoir-fluid viscosity (which is


unlikely for high-compressibility fluid) or a low
fracturing-fluid viscosity.
Assuming that P < 1, we can derive the limiting value
of C ve according to Eq. 19 for R-+O and compare it with
the limit ofE~s. lOa and b, which is C vc =C c . For small
R, ~;"(7r/3)R l(l-P) and F3 ;,,~, and therefore in the
limit
, - - ; -

Then
F I =RI(l +R) (Eq. lOb),
F 2 =2RI(I+v'1+4R 2 ) (Eq. lOa),

~=1/2 [P-I+v'(P-l)2 + (47r/3) R2 ],

and
F3 =v'~/(l +~) (Biot et al. Eq. 21).
The plot of these three functions is shown in Fig. 5. It
is apparent that Eq. lOa gives higher values of leakoff
than Eq. lOb, with the maximum difference of-23% at
CclC v -1. Biot's formula depends on the compressibility parameter P. In the limit for P=O, the curve F3 is
practically identical with F2 and Biot's method gives the
same result as Eq. lOa. Only for fluids with large compressibility and for Cc < Cv are there large differences.
This is, however, a physically unlikely situation,
because to have a large compressibility and small R re-

AUGUST 1985

Ccv(R=O)=~ ~ C c =C c . ............ (B-2)

3(I-P)

This is an interesting result because it shows clearly the


effect of including the moving boundary between the
regions. Eq. B-2 shows that the error is very small if the
compressibility is small, and grows with increasing P.
Because Eq. B-2 gives us a correction for R-+O, and
Biot's formula has the same limit for large R, one would
expect that use of C c in Eq. lOa would give a good approximation to Eq. 19. This is indeed true in the entire range of Fig. 5 as long as P< 1, as can be verifie\~
by calculating "corrected" F 2 by substituting
[7r/3(I-P)] '12 R for R in the formula.

SI Metric Conversion Factors


cu in. X 1.638 706
E+Ol
psi x 6.894 757
E+OO
SPEJ
Original manuscript received in the Society of Petroleum Engineers office Feb. 28,
1983. Paper accepted for publication July 19, 1984. Revised manuscript received
Sept. 17, 1984. Paper (SPE 11625) first presented at the 1983 SPE Low Permeability
Symposium held in Denver March 14-16.

501

You might also like