You are on page 1of 14

SPE 56597

Incorporating Crossflow and Spurt-Loss Effects in Filtration Modeling Within


a Fully 3D Fracture-Growth Simulator
James M. McGowen, SPE, Halliburton Energy Services, Inc., Robert D. Barree, SPE, Marathon Oil Company,
and Michael W. Conway, SPE, Stim-Lab, Inc.
Copyright 1999, Society of Petroleum Engineers, Inc.
This paper was prepared for presentation at the 1999 SPE Annual Technical Conference and
Exhibition held in Houston, Texas, Oct. 3-6.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Permission to copy is restricted to an abstract of not more than 300
words. Illustrations may not be copied. The abstract should contain conspicuous acknowledgment of where and by whom the paper is presented. Write Librarian, SPE, P.O. Box 833836,
Richardson, TX 75083-3836, U.S.A, fax 01-972-952-9435.

Abstract
A practical numerical fracturing-fluid filtration model has been
successfully incorporated into a fully three-dimensional (3D)
fracture-growth simulator. This filtration model incorporates
four principal filtration mechanisms: (1) compressibility and
flow of the reservoir fluid, (2) invasion of non-Newtonian
filtrate, (3) filter-cake deposition, and (4) quasi-equilibrium of
filter-cake thickness with the fluid shear stress. When this
filtration model is used with a numerical 3D fracture-growth
simulator, the impact of these filtration mechanisms on fracturetreatment design can be evaluated. Because the 3D fracturegrowth simulator allows the filtration properties to vary spatially, the simulator can accurately evaluate the impact of changing filtration parameters such as filtration pressure, shear rate,
fluid rheological properties, and fluid composition.
Before evaluation in the fracture simulator, the results of
this new filtration model are first verified against laboratorygenerated filtration data, and then compared to the classical
filtration theory. The fully 3D fracture-growth simulator is then
used to verify the impact of the fluid-leakoff model. The comparisons show that this filtration model can have a pronounced
effect on treatment design when spurt loss is significant, and in
treatments in which the quasi-equilibrium filter-cake formation
(caused by fluid flow over the filter cake) causes dramatic
deviations from classical filtration models. The simulations
highlight concerns about the accuracy of present minifracturing
analysis and treatment-design techniques.

References at the end of the paper.

Introduction
One of the most critical components of fracturing-fluid mechanics is the role of fracturing-fluid invasion and filtration into
porous media. Although many researchers have measured this
process, current industry modeling practices are still rudimentary compared to other elements of fracturing technology. The
summaries compiled by Howard and Fast1 and Penny and
Conway2 will represent classical filtration theory in this paper.
Almost all currently available fracturing simulators are based on
the classical theory of fluid invasion and filtration. McGowen
and Vitthal3-6 rekindled awareness of the major mechanisms
controlling filtration of common fracturing fluids and the limitations of current techniques. Their findings, however, have not
been used in any fracturing simulators because of the absence of
an acceptable methodology. This paper reviews the classical
filtration theory and proposes a methodology that more accurately simulates the filtration process of fracturing fluids. The
new filtration model, filtration with linear-invasion and crossflow
(FLIC), is described, and FLIC simulations are compared to
laboratory measurements as a means of confirming general
behavior of the fluid-leakoff model. The model is then compared with the classical filtration theory, allowing us to determine the classical theorys magnitude of potential error. Finally,
a fully 3D fracturing simulator is used with the FLIC fluidleakoff model so that simulations can be generated to establish
FLICs impact under normal fracture-growth behavior. A short
discussion is then presented on the implications of using the
FLIC model on minifrac analysis and treatment design.
Classical Filtration Theory
Fracturing-fluid leakoff is normally described within the industry with three fluid-loss coefficients. These coefficients represent the fluid leakoff rate through the filter cake, through the
invasion zone, and through the noninvaded zone. A short review
of this classical theory is presented below; the reader can review
Howard and Fast7 and Penny and Conway2 for more detailed
discussion.
Noninvaded Region. To derive the resistance to filtration into
the noninvaded zone, the equations of continuity and Darcys
law are used together. The differential equation used when the

INCORPORATING CROSSFLOW AND SPURT-LOSS EFFECTS IN FILTRATION


MODELING WITHIN A FULLY 3D FRACTURE-GROWTH SIMULATOR

same fluid as the reservoir fluid invades an infinite reservoir,


assuming a slightly compressible fluid, can be approximated as

2 p c p
=
................................................... (1)
k t
y 2
Howard and Fast7 derived the solution for the case of
constant pressure to obtain the following expressions for the
filtration rate and volume.

SPE 56597

Substitution of the expressions for Cc and Cv (Eqs. 2 and 4)


and defining new variables Cc* and Cv* allows the determination of Cvc such that

C vc =

2C v* C c*

where

C v* + C v* 2 + 4C c* 2

C c* =

k c ct

St

k v S t .... (6)
2 v

C v* =

and

V = 2C c t
Cc =

vs =

and

k c ct

Cc
t

where
......... (2)

Sc

Invaded Region. The derivation of the leakoff velocity for the


invaded zone is a simple extension of Darcys law, with the
assumptions of Newtonian filtrate properties, piston-like displacement with 100% saturation of the fluid filtrate, and incompressible filtrate and formation. Starting with Darcys law,

vs =

k v p v
v L

where

L=

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

and integrating (since vs = dV/dt) with the assumption of constant pressure, yields the solution for the leakoff volume and
thus the leakoff velocity such that

V = 2C v t
Cv =

and

kv S v
2 v

vs =
where

Cv
t

where
Sv = Sn

... (4)

Combining Noninvaded and Invaded Regions. Since both


effects occur simultaneously, the fluid-loss coefficients of the
noninvaded and invaded regions are generally combined for an
overall reservoir-leakoff coefficient for predicting filtration
without filter cakes. The combination-coefficient solution assumes constant regional pressures and no moving boundary
between the two regions. For this overall reservoir-rate coefficient to be obtained, the pressures of the individual regions are
summed according to the following:

p t = p c + p v

where

p v = p n ......... (5)

Filter-Cake Region. The fluid-loss rate controlled by the deposition of a filter cake on the formation surface is the wallbuilding coefficient. This coefficient must be measured within
the laboratory and is determined from the slope of the filtrate
volume vs. the square root of filtrate time. Aside from the filtercake coefficient, the spurt-volume (filtrate volume required to
generate the filter cake) and spurt-time are obtained. The expressions for this laboratory modeling are presented as Eq. 7,
with the early spurt data captured with the spurt-loss value and
the late time data assumed to be dominated by the filter-cake
properties.

V = Vo + 2C w t

and

vs =

Cw
t

................... (7)

A derivation of the physical mechanism is provided below


that describes this relationship. The essential assumption necessary for this derivation is that 100% of the solute after the spurtloss phase is retained within the filter cake. The assumption is
normally valid for static filtration, but may not be valid if
crossflow exists. This assumption allows the filter-cake thickness (Lw) to be defined in terms of the leakoff volume and the
ratio of the solute concentration (M) in the bulk solution and
filter cake such that

Lw = V

Mb
..................................................... (8)
Mw

Substitution into Darcys law and integrating yields

V = 2C w t + V o
with

Cw =

and

kwM w S w
2 wM b

vs =

Cw
t
..... (9)

SPE 56597

If all the other variables are independent of pressure, then


the Cw coefficient should be related to P. The filter cake in this
case is termed incompressible since the filter-cake permeability
does not alter with pressure. Since the filtrate viscosity and the
bulk-solute concentration should be independent of pressure,
the pressure dependence of the cake permeability and cakesolute concentration determines whether a particular filter cake
is compressible. A significant amount of literature exists for
incompressible filter cakes, which are common if the solute is
composed of rigid, uniform particles. Compressible filter-cake
theory is discussed later in this paper.
Overall Leakoff Coefficient. Since all three filtration regions
exist after filter-cake formation, the wall-building coefficient is
normally combined with the Cc and Cv coefficients to form an
overall fluid-loss coefficient (Ct). To combine these coefficients, all previous literature assumes incompressible filtercake theory (Cw P) and constant overall pressure differential. With these assumptions, the pressures of the three regions
are summed so that

p t = p c + p v + p w

where

p v = p n (10)

Substituting expressions for Cc, Cv, and Cw and defining a new


variable from Cw* allows the determination of Ct so that

Ct =

2Cc Cv Cw

Cv C w + C w Cv + 4Cc (Cv + Cw )
*2

Filtration with Linear-Invasion and Crossflow (FLIC)


The classical filtration theory described in the preceding section
has many limitations. The most severe limitations are the theorys
inability to model the following: (1) variable filtration pressure,
(2) non-Newtonian filtrates, (3) a filter-cake model that does not
retain 100% of the solute, and (4) a compressible filter cake.
These limitations are addressed in the formulation of the new
FLIC model. The FLIC model is divided into six main tasks: (1)
modeling the noninvaded region, (2) modeling the filtrate invasion, (3) modeling the filter cake, (4) formulating FLIC for
infinite media, (5) applying FLIC with changing filtration
conditions, and (6) determining assumptions and limitations. In
the FLIC model, expressions are developed for the pressure
drop in each of the filtration regions in terms of the fluid
properties, rock properties, and leakoff velocity and volume.
Because time is removed as a dependent variable, the FLIC
formulation can adequately address variable pressure situations.
Modeling the Noninvaded Region. The solution of the classical filtration theory (Eq. 2) can be rearranged to remove the time
variable and can be expressed as
2
v sV = 2C c ....................................................... (13)

This solution can then be rearranged and solved for pc, where

pc =

.......(11)

k M S
with Cw* = w w t
2 wM b

With the combined coefficients, the fluid-loss volume and


leakoff velocity are estimated using expressions

V = 2C vc t
C
v s = vc
t

for

and

vs =

........(12)

Ct
t

for

V > Vo

with

c =

2 k c ct
c ........ (14)

v v sV
k v ................................................... (15)

This equation can be rearranged similarly to Eq. 14 so that

V Vo

p n =

V = 2C t t

v sV
c

Modeling the Filtrate Invasion. The process for modeling the


invaded region depends on the fluid-rheology model used. For
the simple case of a Newtonian filtrate, the pressure drop from
the filtrate invasion is expressed from Darcys law as

p n =

and

J.M. MCGOWEN, R.D. BARREE, M.W. CONWAY

v sV
n

with n =

k v
v ...................... (16)

Significant laboratory data, however, suggests that, before


filter-cake formation, filtrates exhibit properties similar to those
of the bulk fluid. To model this invasion of non-Newtonian
fluid, a model for the porous media is needed to allow prediction
of the average shear rate within the porous media and the
apparent fluid viscosity. Fluid flow through porous media is best
explained by modeling porous media as a bundle of straight
capillaries as proposed by Kozeny8 and later expanded by
Carman.9 This model is derived from the basic equations for

INCORPORATING CROSSFLOW AND SPURT-LOSS EFFECTS IN FILTRATION


MODELING WITHIN A FULLY 3D FRACTURE-GROWTH SIMULATOR

shear rate and shear stress for Newtonian flow in capillaries,


leading to

4v s
r

and =

rp
2L

with

r=

8k
...........(17)

This bundled-tube model ignores both tortuosity within the


porous media and potential non-Newtonian fluid effects such as
slippage or depleted-layer effects. Authors who have proposed
significant solutions include Bird, Stewart, and Lightfoot,10
Christopher and Middleman,11 Teeuw and Hesselink,12 and
Chaveuteau.13 These approaches differ in the location and application of the shear rate and shear stress tortuosity factors, and in
application of either a pulsating-flow solution or depleted-layer
solution. Chaveuteaus technique was chosen for use within the
FLIC formulation. This method adjusts the shear rate by a
tortuosity correction factor and uses a depleted-layer model for
flow through a cylindrical pore. The equation for calculating the
shear rate from this approach can be expressed as

1 4v s
8k ......................................................... (18)

Chaveuteaus approach was selected because the tortuosity


factor for the shear-rate correction is confirmed experimentally
by matching the porous media low-shear transition point between the low-shear Newtonian plateau and the power-law
region to that of the bulk-fluid transition point. This confirmation of the shear-rate calculation is not present in the other
solution techniques. Chauveteau observed 1 values for glass or
sand packs ranging from 1.05 to 2.5, while his lone measurement on sandstone is reported as 4.5. Fletcher et al.15 measured
1 values on 17 cores and observed values ranging from 3 to 10.
As expected, a general trend can be observed with the lower 1
values seen on the higher core permeabilities.
The depleted-layer model proposed by Chaveuteau for
predicting the apparent viscosity within the porous media,
formulated as a bulk-fluid correction factor, is listed below with
dl equal to the viscosity in the depleted layer and rdl as the
thickness of the depleted layer.

CH
2

app
b

dl
b


1 1

dl
b

r
1 dl
8k

Eq. 19) of increasing values with increasing pore size. Since


depleted-layer effects would not occur with Newtonian fluids
(i.e. dl /b equal to 1.0), the 2 value would equal 1 for these
fluids.
Applying the Chaveuteau Approach to the Carreau Rheology Model. The rheology model selected for use with the
FLIC formulation is the Carreau rheology model14 expressed as

app = +

Measured values of 2 range from 0.50 to 0.80 for most


sandstone or sand packs, with the obvious trend (as predicted by

o

1 +
l

1 n
2

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

The Carreau models advantage over the more popular


power-law rheology model is that it predicts both the low- and
high-shear Newtonian plateaus, while also modeling power-law
behavior between the confines of o and . In general, the
Carreau model provides a reasonable approximation for predicting the rheological behavior of most commonly used fracturing fluids. Applying the bulk-fluid correction factor to the
Carreau model and substituting the equation for the porousmedia shear rate yields

app = +

2o

1 + 4 v s 1

8 k
l

1 n
2

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

Substituting Darcys law and rearranging for the pressure drop


yields

Scar

....(19)

SPE 56597

car

v
= V s
car

v
= vs s
kv

with

o 2

4v
s 1
1 +
l 8kv

1 n
2

+ ... (22)

This solution will reduce to the Newtonian solution (Eq. 16) for
n=1.

SPE 56597

Modeling the Filter Cake. Filter-cake deposition under shear


can deviate from the classical square root of time relationship.
This crossflow of the fracturing fluid over the filter cake cannot
be predicted accurately with the classical filter-cake model. For
most viscous fluids, the filter cake under dynamic conditions
tends toward a thickness controlled by the quasi-equilibrium
between the shear stress and hydrodynamic forces and the
filtration pressure. This quasi-equilibrium filter cake will cause
the filtrate rate to become linear with time, resulting in leakoff
data that can be significantly higher than data collected under
static conditions. Many authors have attempted to overcome the
non-square-root filtration behavior by adapting Eq. 7 to include
a third or fourth variable that would model crossflow effects.
The most successful empirical filter-cake model and the one that
will be used within FLIC is the equation proposed initially by
Roodhart,16 which adds a third variable, denoted as Cd, to
account for the quasi-equilibrium filtration behavior. This equation can be expressed as

V = Vo + 2Cw t + Cd t

and

vs =

Cw
+ Cd .......(23)
t

This empirical filter-cake equation has a natural transition


between the early-time filter-cake deposition, in which the
leakoff velocity is inversely proportional to the square root of
time, and the quasi-equilibrium phase, in which the leakoff
velocity approaches a constant.
Modeling Compressibility. Classical incompressible filtercake theory, as described by Eq. 9, would indicate that the filtercake coefficient, Cw, would be proportional to pw. Measurement of the filter-cake coefficient under different differential
pressures, however, has shown that this pressure-dependency
normally varies between 0 and 0.2 psi for common fracturing
fluids.17 It should not be surprising that fracturing filter cakes
composed of flexible, widely varying molecular-weight polymers would have compressible behavior. To account for the
cake compressibility, we can define a compressibility factor
from the relationship of Cw obtained from static laboratory
testing at different pressures so that

C w p ....................................................... (24)
From the derivation of Cw (Eq. 9), it is possible to relate the
product of the cake permeability and cake density to the compressibility factor, so that kwMwpw2-1. Since the Cd coefficient
can be approximated by the Darcy flow of the filtrate through a
constant thickness filter cake, it can be shown that

C d p w

J.M. MCGOWEN, R.D. BARREE, M.W. CONWAY

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

Empirical Filtercake Model. As discussed previously, Eq. 23


is the recommended empirical equation for predicting the mecha-

nisms of filter-cake deposition under significant fluid shear


stress. This equation can be rewritten as

C w + C w + C d C w (V Vo )
2

vs =

V Vo

+ C d ..... (26)

From Eqs. 24 and 25, the two filter-cake coefficients Cw and Cd


can be expressed as functions of Pw as

Pw
C w = C ref
Pw
*
w

and

Pw
C d = C ref
Pw
*
d

.....(27)

These expressions can be substituted into Eq. 26, which can be


rearranged to yield
1

v 2
3w = 3wref
d
where

C w* + C *w + C d* C w* (V Vo )

......(28)

(V Vo )

+ Cd

FLIC Formulation for Infinite Media. Before the filter cake


is developed, the filtration can be modeled by combining the
noninvaded region model and the viscous invasion model. In
general, the principle is to combine the pressure drops within the
individual regions so that

St = Sc + S v
where S v = S n or S car

.............................. (29)

The noninvaded region is modeled using Eq. 14, while the


viscous invasion model is determined from Eqs. 16 or 22.
Combining these equations leads to the general solution, presented as

3t =

Vvs
c

where

v
+ s V
v
= n or car

................................ (30)

Once a filter cake forms, two additional effects must be


modeled. The first effect is to add the filter-cake pressure drop
to Eq. 29. The second effect is caused by the invasion of the
postfilter-cake filtrate into the matrix already invaded by the
spurt-loss non-Newtonian filtrate. The way this postfilter-cake
filtrate displaces, mixes, or fingers through the previous filtrate

INCORPORATING CROSSFLOW AND SPURT-LOSS EFFECTS IN FILTRATION


MODELING WITHIN A FULLY 3D FRACTURE-GROWTH SIMULATOR

will lead to differing solutions for the pv term. Because of the


need to track many fluid changes, we selected a simple mixing
solution. When a mixing solution is used, the modification of
Eq. 29 can be expressed as

St = S c + ( Sv )mix + S w
with

Sv )mix = mixing solution of

S n or S car

...(31)

Substituting Eq. 14, Eqs. 16 or 22, and Eq. 28 yields the


general FLIC filter-cake solution

Vvs v s

Pt =
+
c v mix

n mix

1
2

v
V + s Pref ... (32)
d

Applying FLIC with Changing Filtration Conditions. While


the formulation presented thus far can handle variable filtration
pressures, the model requires additional rules for modeling
changing shear rates, temperatures, or fluid properties. Additionally, methods for populating the reservoir layers for the
empirical constants Vo, Cw, and Cd are required. For simplified
input requirements, these empirical constants are defined as
fluid properties at the conditions listed in Table 1, Page 10. With
these standard conditions defined, the empirical constants for
any local condition can be predicted. While a detailed discussion about adjusting these constants for local conditions is
beyond the scope of this paper, the following general
functionalities are proposed:

Vo = f(Voref1 , k, S S ref , Db ,T,T ref )


C *w = f(Cwref1 ,C wref2 , k,T,T ref )
C d* = f(Cdref1 , k,  Db ,T,T ref ) ............. (33)
*
v = f( b ,C w ,t)
Two Cw coefficients (Cwref1 and Cwref2) are required to allow
local variation of the Cw value vs. permeability. Crosslinked
gels, for example, exhibit very little change in the Cw value vs.
permeability. Linear gels, however, only form filter-cakes on
low permeability. Because Cw values can be entered at 1 md and
1000 md in the FLIC model, it is possible to correctly predict the
Cw and filtrate properties.
Eq. 33 implies that all fluid, temperature, shear-rate, and
pressure changes lead to immediate alteration in the local
constants. The only exception is that the spurt-loss value for a
local condition is not altered once a filter cake is established.
(i.e. Vo is held constant after filter-cake formation). This exception is overcome by allowing the filtrate properties to vary with
the Cw coefficient, allowing a transition from a fluid in which

SPE 56597

filtration is controlled by a filter cake to a fluid in which filtration


is controlled by viscous invasion.
Assumptions and Limitations. The assumptions and limitations in the proposed FLIC formulation for infinite media should
be emphasized. The major assumptions imposed were as follows:
isotropic porous media containing ideal constant compressibility fluid
infinite reservoir boundaries (i.e., vertical fracture)
incompressible filtrate and formation
no moving boundary between the filtrate and the noninvaded
zone
piston-like displacement of reservoir fluid with 100% saturation of filtrate
filtrate before filter-cake formation characterized by the
bulk-fluid properties
filtrate and bulk-fluid properties characterized by the Carreau
rheology model
filter-cake filtration accurately modeled using Cw and Cd
coefficients
changes in Cw and Cd coefficients occurring instantaneously
with changes in the filtration conditions
constant filter-cake compressibility
ideal mixing of filtrates
Validating FLIC with Laboratory Data
Before FLIC data are compared with laboratory-generated data,
several modifications of FLIC are required so that filtration in
finite media can be predicted. For laboratory situations, compressibility of the noninvaded region is normally neglected,
while the pressure drop of the initial saturation fluid must be
modeled. An additional constraint for this situation is also
applied, since the filtration occurs in a finite core length (L).
These factors must be taken into account when the pressure
drops are added and when the filtrates are mixed within the
porous media. A version of FLIC modified for the finite medium
was used to compare against laboratory data collected using
procedures defined by McGowen and Vitthal.4 Three laboratory
tests with borate-crosslinked HPG fluids are compared to the
FLIC prediction in Figs. 1 through 4 (Page 12) . Input data for
the FLIC predictions are included as Table 2 (Page 11) . A linear
relationship of Vo vs. permeability is assumed within the simulations, while the Cw coefficients are assumed to transition
slowly between References 1 and 2. The Cd coefficients were
assumed to transition slowly between the Reference 1 condition
and zero at 1000 md. The Cd coefficients are also assumed to be
a function of the shear rate, such that the Cd value declines
steadily with decreasing shear rate until the value reaches 0 at a
0 shear rate.
The first example is a test conducted with constant filtration
conditions of 500 psi and 50 sec-1 on cores with permeability of

SPE 56597

J.M. MCGOWEN, R.D. BARREE, M.W. CONWAY

0.46 md (Case 1) and 78 md (Case 2). A comparison of the


measured filtration volumes vs. the FLIC predictions is included
as Figs. 1 and 2. Good agreement is observed with the same input
parameters for the two core permeabilities. This agreement is
essential to validate the relationship of permeability on the
empirical constants Vo, Cw, and Cd. The second test was conducted at 1,000 psi on a core with a permeability of 0.34 md
(Case 3). The shear rate during this test was lowered in stages
from 100 sec-1 initially to 75 sec1, 50 sec-1, and finally 25 sec-1.
Again, there was good agreement between the FLIC prediction
and the laboratory measurements observed in Fig. 3 (Page 12).
A comparison of a FLIC solution with the shear rate constant at
100 sec-1 is also included to show the magnitude of the shear-rate
correction. The agreement between the FLIC solution and
laboratory data shows that the filter-cake coefficients can be
instantaneously corrected for the local shear rate without significant error. The last comparison (Case 4) was conducted with
the filtration pressure altering stepwise during the filtration on
a 0.54-md core. Good agreement is observed in Fig. 4 (Page 12),
indicating accurate prediction of the filter-cake compressibility
effects with a = 0.10. A FLIC prediction with the = 0.50
(incompressible cake behavior as assumed in the classical
filtration theory) is shown for comparison. As shown, = 0.50
incorrectly predicts the effect of changing filtration pressure.
FLIC vs. Classical Filtration Theory
The FLIC formulation for infinite media is compared to the
classical filtration theory in Figs. 5 through 10 (Pages 12 and
13). The properties from Case 1 were used in these simulations
unless specified differently in Table 3 (Page 11). The same set
of corrections for the empirical constants used in the laboratory
comparisons were used in these simulations. Each simulation
compares the predicted leakoff volume for the FLIC solution
and the classical filtration theory, using the same set of input
parameters Vo and Cw.
The first comparison (Case 5) simulates the fluid loss on a
0.1-md oil-saturated formation. Because of the limited reservoir-fluid compressibility, little difference between the classical
and FLIC solutions is observed in Fig. 5 (Page 12) . The FLIC
solution was independent of shear rate in this case, indicating the
lack of pressure differential across the filter cake. The next
comparison (Case 6) is the same simulation except that the
formation gas-saturated. The two models in Fig. 6 (Page 12)
show a more pronounced difference since the pressure drop
within the compressible region is significantly lower. The FLIC
solution is dependent on the shear rate and predicts higher
leakoff rates because of cake compressibility and crossflow.
The lack of difference between the 50 and 100 sec-1 FLIC
predictions is caused by significant pressure drop in the invaded
region. Similar results are observed in Cases 7 and 8, which
simulate leakoff on 10 and 100-md oil-saturated formations
(Figs. 7 and 8, respectively, Page 13). The cases show the
relative importance of the cake-compressibility effects by the
difference between the classical and 0 sec -1 solutions, while the

cases also show the importance of crossflow. The crossflow


effect on the 100-md case is lower because of the reduced
influence of crossflow on higher permeability. The results on a
1000-md oil-saturated formation (Case 9, Fig. 9, Page 13) show
that the crossflow effects are no longer important; however,
there is a difference between the FLIC and classical solutions in
the spurt region. This difference in the spurt region is caused by
the use of the bulk fluid properties in the FLIC solution, while the
classical solution assumes a Newtonian filtrate. The last comparison (Case 10, Fig. 10, Page 13) illustrates the importance of the
filter-cake compressibility. The leakoff volume for the filtration
on a 10-md oil-saturated formation is predicted at filtration
pressures of 2,000 and 8,000 psi, comparing the results from the
FLIC (with = 0.10) and the classical theory (which assumes
= 0.50). At 2,000 psi, the slight impact of the increased pressure
is observed between the FLIC solution and the previous FLIC
solution at 1,000 psi and 50 sec-1(Fig. 7, Page 13). The classical
solution, however, is more affected because the cake lacks compressibility. This effect is also observed at 8,000 psi, with the
FLIC solution only significantly affected by the increase in spurt
volume. The classical solution, however, is drastically increased
by the increase in pressure. This behavior is easily explainable
since the Cw from the classical theory increases by 280% (square
root of 8) between 1,000 psi and 8,000 psi, while the Cw of the
FLIC solution (with = 0.1) only increases by 23% and the Cd
increases by 52%.
Incorporating FLIC within a Fully 3D Fracture
Simulator
The FLIC formulation for infinite media was implemented
within a fully 3D fracture simulator, previously described by
Barree16 and systematically enhanced through many years of
use. The 3D simulator allows spatial variation of the formation
properties, fluid properties, filtrate properties, and filter-cake
properties. The ability to spatially vary these parameters is
essential for accurate evaluation of the FLIC formulation and
the impact of usage. To best show the effect of incorporating the
FLIC formulation, several simulations were generated that
compared the classical filtration theory (with as the Newtonian
viscosity) and the FLIC formulation. Each simulation compares
the predicted fracture pressure and fluid efficiency for a minifrac
treatment for a well-confined fracture. Reservoir properties for
the reservoir layers for each simulation are described in Table
4 (Page 11), with the fracture always initiating out of the middle
of the pay interval. Fluid properties are the same as in Case 1
unless specified differently in Table 4. The same set of corrections was used for the empirical constants as in the laboratory
comparisons. With the conditions specified, a minifrac injection
of 15,000 gal at 15 bbl/min is simulated, followed by the
pressure decline until fracture closure. A plot of the predicted
fracture pressures and fluid efficiencies for these minifrac
simulations are presented as Figs. 11 through 14 (Pages 13 and
14).

INCORPORATING CROSSFLOW AND SPURT-LOSS EFFECTS IN FILTRATION


MODELING WITHIN A FULLY 3D FRACTURE-GROWTH SIMULATOR

SPE 56597

As would be expected, the simulations using the 3D fracture


model behave similarly to the FLIC solutions presented previously. The first simulation (Case 11) is for a 0.1-md oil well. Only
slight differences are observed in Fig. 11 (Page 13) between the
classical and FLIC solutions because of the role of reservoir
compressibility. Case 12, however, shows a large difference
between the classical and FLIC solutions when the reservoir fluid
is changed to gas (Fig. 12, Page 13). This behavior is similar to the
results of a 10-md oil reservoir (Case 13) shown in Fig. 13 (Page
14). This case shows that the efficiencies for the FLIC and
classical solutions cross at roughly 2 minutes. This crossing is
caused by the use of a higher viscosity in the FLIC solution during
the spurt invasion followed by the increased leakoff velocity from
the crossflow across the filter cake. Less difference in the two
methods is observed when the permeability is increased to 100 md
(Case 14) as shown in Fig. 14 (Page 14). The lack of difference
is caused by a tradeoff of reduced early-time leakoff from the
non-Newtonian fluid invasion as compared to the classical solution, and slightly higher leakoff caused by crossflow effects.
Because the crossflow effects are less on higher permeability,
only slightly higher leakoff volumes are observed later in time
with the FLIC solution. It should be recognized that all of these
simulations were conducted at filtration pressures below 2,200
psi; thus, little effect of cake compressibility is observed.

borate-fluid pads were followed by HEC proppant-laden fluids.


The classical theorys inability to handle non-Newtonian filtrates has led to difficulty in analyzing minifrac data and
uncertainty in simulating the main treatment. A common error
resulting from incorrect modeling of the non-Newtonian filtrate
effect is to overpredict leakoff from a minifrac treatment and
design excessive pad for the main treatment. A good discussion
of this difficulty has been presented by Dusterhoft et al.19
The crossflow effects, from use of Eq. 23, cause the leakoff
volume at a constant shear rate to start at a proportionality of 0.5
and transition toward a linear proportionality with time. While
the effect will be smaller with steadily declining shear rates,
incorporating crossflow effects will still lead to significant
variation from the assumption used in conventional minifrac
analysis. The result of crossflow will cause conventional minifrac
analysis to underpredict the amount of leakoff that will occur on
larger main treatment volumes and the occurrence of earlierthan-expected tip-screenout behavior. An interesting application of the effect of crossflow is the successful use of linear-gel
pads in low-permeability gas wells to reduce leakoff. The linear
gels do not exhibit crossflow effects because of their reduced
viscosity, and will have lower leakoff volumes than crosslinked
gels at low permeability. A good review of this practice is
presented by Fritcher et al.20

Importance of FLIC on Minifrac Analysis and


Treatment Design
The impact of incorporating non-Newtonian filtrate invasion,
crossflow effects, and filter-cake compressibility in filtration
modeling leads to an obvious concern of how they would impact
conventional minifrac analysis and treatment design. While an
extensive discussion is beyond the scope of the paper, it should
be understood that all conventional analysis and predictions are
based on the classical filtration theory or, more specifically, on
forms of the Carter equation from Howard and Fast.1 This
assumption maintains that the leakoff volume is always proportional to the square root of the leakoff time. Two elements of the
FLIC formulation violate this assumption: the non-Newtonian
filtrate during the spurt phase and the crossflow effects on the
filter cake. Additionally, filter-cake compressibility prevents
the removal of the reservoir components from the overall
solution method. While the impact of the cake compressibility
does not necessarily alter conventional minifrac analysis (except for methods of incorporating laboratory data), the other
effects require more discussion.
The non-Newtonian filtrate causes the time proportionality
of the leakoff volume to become dependent on the fluid n so
that, with constant pressure, the leakoff volume will be a
function of n/(n+1). Thus, if n equals 0.5, the time proportionality for constant pressure filtration would be dependent on t1/3.
The effect of the non-Newtonian filtrate will be even more
critical if the fluids are altered during the minifrac or main
treatment, such as in cases presented by Petit et al.,18 in which

Conclusions
A methodology for predicting fracturing-fluid leakoff, filtration with linear-invasion with crossflow (FLIC), is presented. This formulation models the noninvaded region,
non-Newtonian filtrate invasion, and a compressible filtercake by using an empirical equation that incorporates the
effect of fluid crossflow.
The new FLIC formulation, modified for finite media,
successfully predicted laboratory filtration experiments
including cases with changing shear rates and pressures.
Simulations comparing FLIC to classical filtration theory
show the importance of correctly modeling for the effects of
non-Newtonian filtrate invasion, crossflow, and filter-cake
compressibility.
The FLIC formulation was successfully incorporated into a
fully 3D fracturing simulator. This incorporation confirms
the presence of serious limitations with the classical filtration theory and highlights concerns about the accuracy of
present minifracturing analysis and treatment-design techniques.
While inaccuracies from the use of the classical filtration
theory were small when the noninvaded region was important, significant errors were observed in all other conditions. In low- to moderate-permeability formations,
crossflow and cake-compressibility effects cause deviations from classical-theory predictions. In moderate- to
high-permeability formations, non-Newtonian filtrate in-

SPE 56597

J.M. MCGOWEN, R.D. BARREE, M.W. CONWAY

vasion during the spurt phase causes deviation from predictions made with the classical filtration theory.
Acknowledgments
The authors thank the management of Halliburton Energy Services, Inc., Marathon Oil Company, and Stim-Lab, Inc. for
permission to publish this paper. Special thanks are extended to
Diederik van Batenburg and Harold Walters, both of Halliburton
Energy Services, Inc. for valuable discussions and assistance on
the formulation of the FLIC model.

vc =
w =
y =
Superscripts
* =
ref =
ref1 =
ref2 =

combination of c and v regions


wall-building region
yield
evaluated at Pt (classical) or Pwref (FLIC)
measured at reference condition
measured at reference condition
measured at reference condition

References
Nomenclature
c =
k =
n =
r =
t =
v =
y =
C =
D =
E =
K =
L =
P =
M =
V =
=

=
=
=
=
=
=

1.

compressibility (Pa-1)
permeability (m2)
flow-behavior index (dimensionless)
average capillary radius (m)
filtration time (sec)
velocity (m sec-1)
distance in the direction of the filtration (m)
fluid leakoff coefficient (m sec-1/2)
degradation ratio compared to t = 0
Youngs Modulus (psi)
consistency index (Pa secn)
length (m)
pressure (Pa)
solute concentration (kg m-3)
filtration volume per area (m3 m-2)
intermediate variable (units dependent on
subscript)
porosity (dimensionless)
shear rate (sec-1)
compressibility constant (dimensionless)
apparent viscosity (Pa sec)
shear stress (Pa)
Poissons Ratio (dimensionless)

rock factor (dimensionless)


change (dimensionless)

Subscripts
a =
app =
b =
c =
d =
dl =
i =
mix =
n =
new =
o =
old =
s =
t =
v =

average capillary
apparent
bulk polymer solution
noninvaded region
dynamic equilibrium region
depleted layer
interior of pore
mixture
Newtonian
new
initial
old
superficial
total
viscous invasion region

2.

3.

4.

5.

6.

7.

8.
9.
10.
11.
12.

13.

14.
15.

Howard, G.C., and Fast, C.R.: Hydraulic Fracturing, Monograph


Series, SPE, Richardson, TX (1970) 34-41.
Penny, G.S. and Conway, M.W.: Chapter 8: Fluid Leak-off, in
Recent Advances in Hydraulic Fracturing, Gidley, J.L, Holditch,
S.A., Nierode, D.E., and Veatch, R.W. Jr. (eds), SPE Monograph
Volume 12, Richardson, TX, 1989.
McGowen, J.M., Vitthal, S., Parker, M.A., Rahimi, A., and
Martch, W.E. Jr: Fluid Selection from High Permeability Formations, paper SPE 26559 presented at the 1993 SPE Annual
Technical Conference, Houston, TX, Oct. 3-6.
McGowen, J.M. and Vitthal, S.: Fracturing Fluid Leak-off Under
Dynamic Conditions: Part 1: Development of a Realistic Laboratory Testing Procedure, paper SPE 36492 presented at the 1996
Annual Technical Conference, Denver, CO, Oct. 6-9.
Vitthal, S. and McGowen, J.M.: Fracturing Fluid Leak-off Under
Dynamic Conditions: Part 2: Effect of Shear Rate, Permeability,
and Pressure, paper SPE 36493 presented at the 1996 Annual
Technical Conference, Denver, CO, Oct. 6-9.
McGowen, J.M. and Vitthal, S.: Evaluation of Particulate and
Hydrocarbon Fracturing Fluid-Loss Additives Under Dynamic
Conditions, paper SPE 37488 presented at the 1997 Production
Operations Symposium, Oklahoma City, OK, Mar. 10-12.
Howard, G.C., and Fast, C.R.: Optimum Fluid Characteristics
for Fracture Extension, Drill. and Prod. Prac., API (1957) 261270 (Appendix by E.D. Carter).
Kozeny, J.: Uber kapillare Leitung des Wassers im Beden,
Berichte Wien Akad., 186-2A (1927) 271.
Carman, P.C.: Fluid Flow Through Granular Beds, Trans. Inst.
Chem. Eng. 16 (1937) 150-165.
Bird, R.B., Stewart, W.E., and Lightfoot, E.N.: Transport Phenomena, John Wiley and Sons, New York (1960) 197-207.
Christopher, R.H., and Middleman, S.: Power-Law Flow Through
a Packed Tube, I&EC Fundamentals (November 1965) 422-426.
Teeuw, D., and Hesselink, F.T.: Power-Law Flow and Hydrodynamic Behavior of Biopolymer Solutions in Porous Media,
presented at the 1980 Fifth International Symposium on Oilfield
and Geothermal Chemistry, Stanford, CA, May 28-30.
Chauveteau, G.: Rod Like Polymer Solutions Flow Through Fine
Pores: Influence of Pore Size on Rheological Behavior, Journal
of Rheology, Volume 26 No. 2 (1982) 111-142.
Carreau, P.J.: Rheological Equations from Molecular Network
Theories, Trans. Soc. Rheol. (1972) 16, 99-127.
Fletcher, A.J.P., Flew, S.R.G., Lamb, S.P., Lund, T., Bjrnestad,
E.., Stauland, A., and Gjvikli, N.B.: Measurements of Polysaccharide Polymer Properties, in Porous Media, paper SPE 21018
presented at the 1991 International Symposium on Oilfield Chemistry, Anaheim, CA, Feb. 20-22.

10

INCORPORATING CROSSFLOW AND SPURT-LOSS EFFECTS IN FILTRATION


MODELING WITHIN A FULLY 3D FRACTURE-GROWTH SIMULATOR

16. Roodhart, L.P.: Fracturing Fluids: Fluid-Loss Measurements


Under Dynamic Conditions, SPEJ (Oct. 1985) 629-636.
17. Barree, Robert D.: A Practical Numerical Simulator for ThreeDimensional Fracture Propagation in Heterogeneous Media,
paper SPE 12273 presented at the 1983 Reservoir Simulation
Symposium, San Francisco, CA, Nov. 15-18.
18. Petit, G., Leschi, P., and Dusterhoft, R.: Frac and Pack Stimulation: Application and Field Experience from Hylia Gabon, West
Africa, paper SPE 30115 presented at the 1995 European Formation Damage Symposium, The Hague, The Netherlands, May
15-16.
19. Dusterhoft, R.D., Vitthal, S., McMechan, D.E., and Walters,
H.G.: Improved Minifrac Analysis Technique in High Permeability Formations, paper SPE 30103 presented at the 1995
Formation Damage Symposium, The Hague, The Netherlands,
May 15-16.
20. Fritcher, E., Byrd, A., and Stegent, N.: Optimized Frac Pad and Gel
Improve Well Productivity, Oil & Gas J. (Mar. 20, 1995) 90-94.

SI Metric Conversion Factors


m
3.28084E + 00
m
2.45424E + 01
m s-1/2
2.54133E + 01
m s-1
1.96850E + 02
kg
2.20462E + 00
kg m-3
8.34541E + 00
pa
1.45038E - 04
pa-1
6.89476E + 03
pa s
1.00000E + 03
pa s
2.08850E - 02
m2
1.01325E + 15
pa secn
2.08850E - 02

Table 1Defining Empirical Constants


Variable

Defined At

ref1

1 md, p , 50 sec , Tref, no fluid degradation

Cwref1
Cwref2
Cdref1

1 md, pref, 50 sec-1, Tref, no fluid degradation

Vo

ref

-1

1000 md, pref, 50 sec-1, Tref, no fluid degradation


1 md, pref, 50 sec-1, Tref, no fluid degradation

SPE 56597

=
=
=
=
=
=
=
=
=
=
=
=

ft
gal ft-2
ft min-1/2
ft min-1
lb
lb mgal-1
psi
psi-1
cp
lbf sec ft-2
md
lbf secn ft-2

SPE 56597

11

J.M. MCGOWEN, R.D. BARREE, M.W. CONWAY

Table 2Fluid Parameters for Laboratory Comparisons


Property

Case 1

Case 2

Case 3

Case 4

0.32

0.32

0.25

0.32

K (lbf secn/ft2)
o (cp)

0.5

0.5

0.9

0.5

6,000

6,000

12,000

6,000

(cp)

i (cp)

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0.004

0.004

0.0023

0.004

0.003

0.003

0.0032

0.0035

n (cp)
Vo

ref1

(gal/ft )

Cwref1 (ft/min-1)
Cwref2 (ft/min-1)
Cdref1 (ft/min)

0.003

0.003

0.0032

0.0035

0.0006

0.0006

0.0012

0.00052

0.1

0.1

0.1

0.1

p (psi)
k (md)
ki/kv

1,000

1,000

1,000

1,000

0.46

78

0.34

0.54

0.16

0.23

0.15

0.16

p (psi)

500

500

1,000

Varies

(sec )

50

50

Varies

50

ref

-1

Table 3Fluid Parameters for Laboratory vs. Classical Comparisons


Property
Case 5
ki (md)
0.1

0.1

10

100

1000

10

0.20

0.20

0.20

0.20

0.20

0.20

1,000

1,000

1,000

1,000

1,000

Varies

3.00E-05
1

3.00E-04
0.02

3.00E-05
1

3.00E-05
1

3.00E-05
1

3.00E-05
1

p (psi)
-1

ct (psi )
c (cp)

Case 6

Case 7

Case 8

Case 9

Case 10

Table 4Reservoir Properties for 3D-Model Simulations


Property

Case 11

Case 12

Case 13

Case 14

Pay Height (ft)

100

100

100

100

k (md)
kc/kv

0.1

0.1

10

100

ct (psi )
c (cp)

3.00E-05

3.00E-04

5.00E-05

5.00E-05

0.02

BH Pressure (psi)

3,000

3,000

3,000

3,000

E (psi)

3.00E-06

3.00E-06

3.00E-06

1.00E-06

Pay Zone Stress (psi)

4,250

4,250

4,250

4,250

0.25

0.25

0.25

0.25

-1

INCORPORATING CROSSFLOW AND SPURT-LOSS EFFECTS IN FILTRATION


MODELING WITHIN A FULLY 3D FRACTURE-GROWTH SIMULATOR

12

0.040

0.030

0.035
0.030

Lab Data

0.025

FLIC

0.025

V (m)

V (m)

0.040

FLIC

0.035

0.020
0.015

Lab Data

0.020
0.015

0.010

0.010

0.005

0.005

0.000
0.00

2.00

4.00

6.00

8.00

10.00

0.000
0.00

12.00

4.00

2.00

Time (min)

75 sec-1 50 sec-1

12.00

0.05

25 sec-1

2,000
psi

0.060

FLIC

0.040

Lab Data

0.030

1,500
psi

1,000
psi

FLIC

0.04

FLIC
at 100 sec-1

0.050

V (m)

10.00

8.00

Fig. 2FLIC vs laboratory data for Case 2 (78 md).

V (m)

0.070

100 sec-1

6.00

Time (min)

Fig. 1FLIC vs. laboratory data for Case 1 (0.46 md).

0.080

SPE 56597

FLIC
w i t h k = 0.5

0.03
Lab Data
0.02

0.020

0.01

0.010
0.000
0.00

2.00

4.00

6.00

8.00

10.00

0.00

12.00

Time (min)

10

12

14

Fig. 4FLIC vs. laboratory data for Case 4 (0.54 md).

0.006

0.030

0.005

FLIC at 0 to100 sec

-1

0.004

FLIC at 100 sec-1

0.025

FLIC at 50 sec-1

Classical
0.020

V (m)

V (m)

Time (min)

Fig. 3FLIC vs. laboratory data for Case 3 (0.34 md).

0.003

0.010

0.001

0.005

2.00

4.00

6.00

8.00

10.00

12.00

14.00

Time (min)
Fig. 5FLIC vs. classical theory for Case 5 (0.1-md oil reservoir).

FLIC at 0 sec-1

0.015

0.002

0.000
0.00

0.000
0.00

Classical

2.00

4.00

6.00

8.00

10.00

12.00

14.00

Time (min)
Fig. 6FLIC vs. classical theory for Case 6 (0.1-md gas reservoir).

SPE 56597

0.040
FLIC at 50 sec

0.030

FLIC at 0 sec-1

0.015

Classical

0.020

0.005

0.005
0.000

10

12

Classical

0.015
0.010

FLIC at 0 sec-1

0.025

0.010

FLIC at 50 sec-1

0.030

0.025
0.020

FLIC at 100 sec-1

0.035

-1

V (m)
14

Time (min)
Fig. 7FLIC vs. classical theory for Case 7 (10-md oil reservoir).

12

14

0.06

0.10

0.05

Classical

0.08

Classical at 8,000 psi

FLIC at 50 sec-1

FLIC at 8,000 psi

0.04

0.06

V (m)

V (m)

10

Fig. 8FLIC vs. classical theory for Case 8 (100-md oil reservoir).

0.12

0.04

FLIC at 2,000 psi

0.03
0.02

0.02
0.00

Time (min)

Classical at 2,000 psi

0.01
0

10

12

14

0.00

Time (min)

10

12

14

Time (min)

Fig. 9FLIC vs. classical theory for Case 9 (1000-md oil reservoir).

Fig. 10FLIC vs. classical theory for Case 10 (10-md oil reservoir).
1.0

5,300

Fracture Pressure (psi)

5,200

0.8

Classic Eff

5,100

0.6

FLIC Eff

5,000

0.4

4,900

0.2

4,800

0.0

4,700

-0.2

Classic

4,600

-0.4

FLIC

4,500

-0.6
-0.8

4,400
4,300

10

20

30

Time (min)

40

-1.0
50

Fig. 11FLIC vs. classical theory for Case 11 (0.1-md oil reservoir).

Efficiency (%)

V (m)

0.040

FLIC at 100 sec-1

0.035

0.000

13

J.M. MCGOWEN, R.D. BARREE, M.W. CONWAY

5,300

1.0

5,200

0.8

5,100

0.6
Classic Eff

5,000
4,900

0.4
0.2

FLIC Eff

4,800

0.0

4,700

-0.2

4,600

-0.6

FLIC

4,400
4,300

-0.4

Classic

4,500

10

15

20

25

Time (min)

Efficiency (%)

Fracture Pressure (psi)

INCORPORATING CROSSFLOW AND SPURT-LOSS EFFECTS IN FILTRATION


MODELING WITHIN A FULLY 3D FRACTURE-GROWTH SIMULATOR

-0.8
30

35

40

45

-1.0

5,300

1.0

5,200

0.8

5,100

0.6

5,000

4,800

0.2
0.0

FLIC Eff

4,700
4,600

-0.2
-0.4

Classic

4,500

-0.6

FLIC

4,400
4,300

0.4

Classic Eff

4,900

10

Efficiency (%)

Fracture Pressure (psi)

Fig. 12FLIC vs. classical theory for Case 12 (0.1-md gas reservoir).

-0.8

20

-1.0
40

30

Time (min)

Fig. 13FLIC vs. classical theory for Case 13 (10-md oil reservoir).

1.0

4,800

0.8
4,700

0.6

FLIC Eff

0.4

Classic Eff

4,600

0.2
0.0
-0.2

4,500

-0.4
4,400

4,300

Classic
FLIC
0

10

15

20

25

Time (min)

30

35

40

-0.6
-0.8
-1.0
45

Fig. 14FLIC vs. classical theory for Case 14 (100-md oil reservoir).

Efficiency (%)

Fracture Pressure (psi)

14

SPE 56597

You might also like