You are on page 1of 13

SPE 63172

Impact of Rock Plasticity on Hydraulic Fracture Propagation and Closure


D.B. van Dam, SPE, Delft U. of Technology, P. Papanastasiou, Schlumberger Cambridge Research and C.J. de Pater,
SPE, Delft U. of Technology

Copyright 2000, Society of Petroleum Engineers Inc.


plastic (shear) deformation around the fracture tip, as depicted
This paper was prepared for presentation at the 2000 SPE Annual Technical Conference and in Figure 1.
Exhibition held in Dallas, Texas, 1–4 October 2000.
This paper was selected for presentation by an SPE Program Committee following review of It has been proposed that such plastic behaviour may induce a
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 high tip pressure1,2, that could lead to high wellbore pressures
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
and a fracture that is much wider and shorter compared with
SPE meetings are subject to publication review by Editorial Committees of the Society of elastic behaviour. In fact there is evidence that propagation
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is pressure in soft formations is much larger than predicted by
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 elastic models. Apart from a high tip pressure, the apparently
words; illustrations may not be copied. The abstract must contain conspicuous acknowledg-
ment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, high net pressures can also be explained by underestimation of
Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
the rock modulus. Net propagation pressure is almost propor-
tional to rock modulus and the modulus can (especially in soft
Abstract rock) be underestimated by a factor of two to three. Also, un-
We performed scaled laboratory experiments of hydraulic frac- derestimation of fracture closure pressure could lead to signifi-
ture propagation and closure in soft artificial rock and outcrop cant errors, since the absolute level of net pressure is low in
rock samples. We also performed numerical simulations of the soft rocks.
fracture behaviour in plastic rocks, using independently meas- In routine jobs, the net pressure is used for estimating frac-
ured rock properties. The simulations aided us in interpreting ture penetration. Significant errors in the relation between net
the measurements and extrapolating the results to field scale. pressure and fracture width could result in poor coverage of
Plasticity induces a larger width for a given net pressure, com- the interval or insufficient fracture length. Also, fracturing is
pared with elastic rock. However, the pressure to propagate nowadays used for waste disposal in shallow, unconsolidated
fractures is only marginally increased and in the case of the formations. It is obvious that a correct prediction of fracture
laboratory tests was actually lower than expected from elastic geometry is vital for safely designing such a process. Apart
behaviour. The most dramatic effect of plasticity is that closure from fracture geometry, it is important to establish the stress
is much lower than the confining stress due to strong stress development after fracturing soft formations. Sandface stabili-
redistribution along the fracture. sation in unconsolidated formations is an important by-product
Introduction of fracturing. Insight in the induced stress changes can aid in
Rock is characterised by a very low tensile strength compared optimising such treatments.
to its compressive strength. Often, tensile fracturing will there- This paper presents an overview of a study of rock plastic-
fore prevail if rock is loaded to rupture. This is also expected ity in hydraulic fracturing (for details we refer to refs. 3 to 13).
for fluid driven fractures, where the high fluid pressures will We performed physical model tests to determine the relevant
lead to extensile loading of the rock ahead of the fracture tip. phenomena and to validate numerical models of fracture
Straight ahead of the crack, the stresses will go to zero, but at propagation. Small-scale lab tests cannot be directly applied to
some offset from the fracture plane, the stress decreases
strongly in the direction perpendicular to the fracture plane,
whereas the stress along the fracture plane will be much
higher. This loading induces shearing of the rock, which gives
rise to deviations from brittle tensile failure. Shear failure can
be described on a global scale by plasticity theory, although on
a small scale the plastic deformation consists either of so-
called axial cleavage fractures along the maximum stress or
small shear fractures. In view of the large shear stress, fracture
propagation in weak rocks is expected to induce a significant
Figure 1: Stresses and process zone at a hydraulic fracture tip.
2 D.B. van Dam, P. Papanastasiou and C.J. de Pater SPE 63172

Figure 2: Schematic view of the experimental set-up for hydraulic fracturing experiments: overview of total set-up (left) and overview of
ultrasonic measurements (right).
field applications: one needs to use a numerical model that is and calculate the fracture radius and the size of the non-
checked against experiments that represent the essential phys- penetrated zone. In experiments on plaster with non-penetrated
ics of the process. Validation of the numerical code gives con- zone sizes of 1 to 2 cm, the non-penetrated zone size agreed
fidence in the field application of the numerical model. Of well with the size of the dry zone measured after opening of
course, it is always possible that field behaviour is governed by the block. When the fracture stops growing, the diffraction
phenomena such as formation heterogeneity (e.g. layering and disappears. Then, the ultrasonic shear-wave transmissions gave
small scale faults) that are neglected in the current model. us the radius over which the fracture was open. These meas-
We will first review the most important experimental ob- urements have an accuracy of approximately ± 1 cm.
servations. Then we will describe the results of simulations We used artificial rock samples made of cement, plaster and
with the two (independent) numerical models. The two models diatomite as natural rock, see table 1. The strength of the plas-
are based on the same physical assumptions, but we use the ter blocks depends on the water content. We did experiments
first model for comparison with the experiments and the sec- with strong and weak plaster. The diatomite blocks were ob-
ond model is used for extrapolation of the results to field scale. tained from a mine in Lompoc, CA, U.S.A. In diatomite, the
Finally, we will discuss the practical implications of the re- hydraulic fracture plane is oriented approximately parallel with
sults. the bedding plane.
Experimental set-up and method Scaling: The experiments are scaled in terms of energy
Fig. 2a shows a schematic view of the experimental set-up rates associated with fracture opening, fluid flow, and rock
used for the hydraulic fracturing experiments. Cubic blocks of fracturing3. In order to have sufficient time for doing meas-
3
0.30 m size are loaded in a true triaxial machine to simulate in- urements, the time scale for the propagation phase is about 10
situ stress states; no pore pressure can be applied. We used 0.1 s. The fracture radius is approximately 0.1 m. The combination
mm thin Teflon sheets greased with Vaseline to reduce friction of these practical conditions with the scaling laws requires a
between the block and the loading platens. Six LVDT’s meas- relatively high fluid viscosity and low fracture toughness in the
ure the block deformation. During fracture propagation, a high experiments.
pressure pump injects fluid into the wellbore. The wellbore Pressure and width in hydraulic fracturing
pressure is measured at a dead string. An LVDT, mounted with experiments
clamps in the wellbore, measures the fracture width with a It is most convenient to compare experiment and simulation as
measuring error of approximately 10 %. The block extension a function of radius. Fracture width, bw, and pressure, pnet,w,
during propagation is a measure for the fracture volume. The are made dimensionless using the wellbore radius Rw, the ef-
fracturing fluid we use is silicon oil which behaves approxi- fective flowrate ieff, the channel flow viscosity µ (being 12m),
mately Newtonian at the shear rates of interest. 2
Fig. 2b shows a schematic view of the ultrasonic measure- and the crack opening modulus E (being E/4(1-n )). By using
ments. Direct transmissions through the fracture yield the the actual value ieff of the flowrate, the dimensionless wellbore
width profile of the fracture14. The fracture tip acts as a dif- pressure and width can be compared with the simulation when
fractor. During fracture propagation, two diffractions are we assume that the propagating fracture adjusts its width and
measured, which we interpret as coming from the fluid front pressure instantaneously to the variations in the effective
and fracture tip position. From the traveltime of these diffrac- flowrate.
tions and the material velocities we can construct the ray path,
SPE 63172 IMPACT OF ROCK PLASTICITY ON HYDRAULIC FRACTURE PROPAGATION AND CLOSURE 3

1.0 3.0
ce01 ce02
ce02 ce03
ce03 2.5 ce04
0.8 ce06
ce04
ce06 ce07
ce07 2.0 ce08
b w* ce09
0.6 ce08
pnet,w* ce09 (-) ce10
ce10 1.5 sim ulation
(-) simulation
0.4
1.0

0.2
0.5

0 0
0 2 4 6 8 10 0 2 4 6 8 10
R* (-) R* (-)
* *
Figure 3: Dimensionless net pressure pnet,w and fracture width bw at the wellbore in cement, as a function of fracture radius, normalised
on the wellbore radius. In test ce09, the flow rate was increased by a factor of 30 during the test, resulting in the sudden change in pres-
sure and width.

1.2 3.0 sp01


sp01 sp02
sp02 sp03
1.0 sp03 2.5 sp04
sp04 sp05
sp05 sp06
0.8 sp06
pnet,w* sp07 bw* 2.0 sp07
sp08
(-) sp08 (-) sp09
0.6 sp09 1.5 sp10
sp10
simulation
simulation
0.4 1.0

0.2 0.5

0 0
0 2 4 6 8 10 0 2 4 6 8 10
R* (-) R* (-)
Figure 4: Dimensionless net pressure and width at the wellbore as a function of normalised radius in strong plaster.

1.2 3.0
wp01 wp01
wp02 wp02
1.0 wp03 2.5 wp03
wp04 wp04
wp05 wp05
0.8 wp06 2.0 wp06
pnet,w* wp07 b w* wp07
(-) simulation (-) sim ulation
0.6 1.5

0.4 1.0

0.2 0.5

0 0
0 2 4 6 8 10 0 2 4 6 8 10
R* (-) R* (-)
Fig. 5: Dimensionless net pressure and width at the wellbore as a function of normalised radius in weak plaster.
4 D.B. van Dam, P. Papanastasiou and C.J. de Pater SPE 63172

The non-dimensionalisation is based on the expressions that For diatomite, we could not obtain the full curves of pressure
describe fracture propagation as given in Ref. 15, which yields and width versus radius, since the acoustic monitoring did not
for pnet,w and bw: work due to strong attenuation of the acoustic waves. How-
1 ever, the pressure and measured width agreed with the elastic
1
 µi E 3  4 simulation at the end of the tests when we used the radius of
 µi R  4
∝  and b ∝  eff  ...................(2)
eff the fracture seen after opening the blocks.
pnet ,w
 R3  w
 E 
 
We normalise the fracture radius on the wellbore radius: R*
=R/Rw. 0.12 0.06
Figure 3 shows the pressure and width for the cement blocks. R a)
R fluid
In these figures we also plotted the predicted pressure and ω
width for an elastic simulation. The pressure was somewhat
0.08 0.04
lower than the prediction, but agrees within experimental error. R ω
In order to investigate directly the importance of toughness in
(m) shut-in (m)
the cement experiments, we did an experiment (numbered ce08
in table 2) in which we prefractured the block with water. The 0.04 0.02
radius of the water fracture was 8.8 cm, while the maximum
attained width was 18 µm. The fracture surface was smooth,
and the fracture closed almost completely after shut-in, which 0 0
was indicated by the acoustic transmission measurements. We 0 1000 2000 3000

reopened this fracture with viscous silicone oil. The net pres- t-t i (s)
sure and fracture width at the wellbore during reopening b)
agreed with the fractures created in intact cement. Also, the shut−in
break-down pressure is unaffected by the absence of tough- 1

ness. This again indicates that the influence of fracture tough- normalized energy 0.8
ness is negligible in cement under our conditions of stress,
flow rate, and viscosity.
0.6 9.6 8.1 6.6 5.1 3.6 radius (cm)
Figure 4 and 5 show the pressure and width for strong and
weak plaster, respectively. The net pressure is generally
0.4
somewhat lower than in the simulation, which is the same kind
of deviation as in cement. The best agreement is obtained for 0.2
experiment sp02, with a relatively small externally applied
stress deviator. The largest difference between experiment and 0
simulation is for experiments sp09 and sp10, where a very 1800 2000 2200 2400 2600
t (s)
large stress difference was applied. This suggests a systematic
deviation between the measurements and the simulation, which 0.12
however is within the experimental error. c) sp05
For soft plaster, the pressure was much lower and the width sp06
sp09
was much higher than the elastic simulation. The stresses and
0.08
pump rates varied in these experiments, see Table 2. We could
R
not find a clear influence of the stresses on the measured pres-
sure and width. Also, there was no systematic influence of the (m)
injection rate. Triaxial tests show that creep may be important 0.04
in weak plaster3. We did three experiments (wp04, wp05, and
wp06) to determine the significance of creep in this material.
In these experiments, the fracture efficiency and product of
0
viscosity and flow rate were equal, which would lead to the 0 1000 2000
same net pressure in the elastic case. The time scales of propa- t-tsi (s)
4 3 2
gation were 10 s (wp04), 10 s (wp05), and 10 s (wp06).
From these experiments, it appears that creep leads to a Figure 6: Behaviour of fracture radius after shut-in. (a) Meas-
smaller wellbore pressure and larger wellbore width, although ured radius and non-penetrated zone (NPZ) size in cement, (b)
the effect is within the normally observed variation (see figure transmitted shear energy through the fracture during closure
5). at different radial distances from the wellbore in plaster blocks
and (c) radius during closure inferred from shear wave trans-
missions for three experiments on plaster.
SPE 63172 IMPACT OF ROCK PLASTICITY ON HYDRAULIC FRACTURE PROPAGATION AND CLOSURE 5

Fracture Closure
In all experiments we measured radius changes after shut-in. In Closure at the borehole can be detected directly from the width
cement the radius grows, which goes together with a decreas- measurement. Only in relatively strong plaster, the closure
ing size of the non-penetrated zone. In plaster we measure re- point is clearly visible on a plot of pressure versus G-function
cession after shut-in, which we can approximate by a radius (see Fig. 7). In soft plaster and diatomite, we measured a
declining linearly with time after shut-in, see Figure 6. It shows smoothly decaying pressure without a clear break at closure.
that the fracture surfaces make mechanical contact at the tip This could be attributed to a gradual decay of fracture compli-
first, and then subsequently towards the wellbore. ance in combination with a relatively impermeable residual
We found that the pressure decline after shut-in was strongly fracture. In cement, the leak-off rate was very low, which made
influenced by the diminishing leak-off area. When we incorpo- it hard to detect closure.
rate the measured radius in plaster in the model and assume The fluid pressure at which the fracture closes in our experi-
that the closed part of the fracture is impermeable, we indeed ments on relatively strong plaster is between 10 and 25 % too
found a fair agreement of the modelled fracture volume with high (see for example Fig. 7a). Possible causes for this could
the measured volume5. be fracture roughness near the wellbore and poroelastic effects.
20 300 The test on diatomite showed normal closure behaviour: the
(a) pw width was almost zero at a pressure equal to the minimum
bw stress. In experiments on soft plaster the fluid pressure at
15 200 which the fracture closes is significantly lower than the confin-
p b ing stress working perpendicular to the fracture plane. Fig. 7c
(MPa) closure at the wellbore (µm) shows an example of this.
↓ Detailed analysis of the pressure decline shows that closure at
10 100
the wellbore does not take place at the point were the deriva-
closure stress
tive starts to deviate from the approximately linear slope in the
beginning. Instead, it appears that the fracture typically closes
5 0
0 0.5 1.0 1.5
near the point where the derivative reaches a local minimum5.
Numerical Modelling of Laboratory Tests
G (tD) (-)
Elasto-plastic Simulations
2.2 600 We use two different numerical codes to simulate the propaga-
(b) pw
tion and closure of fractures in plastic rock: a custom built
bw
Finite Element code (developed at SCR) that fully couples
1.8 400 rock deformation and fluid flow and a commercial finite dif-
p b ference code (FLAC) with custom routines that couple the rock
(MPa) (µm) deformation with fluid flow using self-similar propagation.
1.4 200 The constitutive behaviour of the solid is based on elasto-
Borehole

closure stress

1.0 0
0 4.0 8.0 12.0 16.0
0.15
G (tD) (-)

16 400
(c) pw
0.10
bw
Fluid pressure

12 300
p b
(MPa) (µm)
8 closure stress 200
0.05

4 100

0
0 0
0 0.5 1.0 1.5 2.0 2.5 Fluid pressure Fixed y-coordinates
G (tD) (-) 0 0.05 0.10 0.15

Figure 7: Wellbore pressure as a function of the dimensionless


loss function in the high leak-off limit for (a) plaster, (b) diato- Fracture radius
mite, and (c) soft plaster. Figure 8: Boundary conditions and grid.
6 D.B. van Dam, P. Papanastasiou and C.J. de Pater SPE 63172

plastic theory. We used a Mohr-Coulomb model for the plastic metry considerations. Around the tip a fine grid is needed. As
potential function. The cohesion and friction angle varied as a there is no remeshing possibility, a fine grid was required
ps
function of the accumulative plastic strain ε . around the whole fracture. The used grid size is fine enough to
We use the Poisseuille equation and the continuity equation to follow the strong gradients near the fracture tip. We expect
calculate the pressure profile inside the fracture. We assume that our grid is fine enough, based on the validation of the
impermeable fracture walls, because leak-off is only of secon- model using the model of Barr17. The grid point co-ordinates
dary importance for the pressure profile, when we compare in the plane of the fracture were initially fixed in the direction
fracture pressure at a given radius. Assuming self-similar perpendicular to the fracture plane. Fracturing of these grid
propagation facilitates convergence of the numerical integra- points took place when the tensile stress exceeded the tensile
tion scheme. This assumption means that we neglect the time strength, and was applied by removing the displacement con-
derivative of the width in the mass conservation equation. straint. After fracturing, the gridpoint positions were fixed
We implemented a cohesive-zone model16 for the rupture again when the initial position was reached (which happens
process at the fracture tip. This is the simplest approach be- after unloading of the pressure working on the fracture sur-
yond Linear Elastic Fracture Mechanics, that is obviously in- face).
applicable. In the cohesive zone, a tensile stress σcoh is applied Numerical results
on the fracture faces as a function of the fracture width b. The We determined the friction angle, cohesion, and cap pressure
relation between tensile stress and width is obtained from uni- as a function of the hardening parameters used in FLAC. Using
axial tensile tests. In the simulation presented, we used a linear these data, we simulated our triaxial extension tests on soft
relation between opening b and stress σcoh after the maximum plaster in order to obtain appropriate model parameters by
stress T0. We assumed that the cohesive zone is permeable for fitting the simulations to the measured stress-strain curves. In
fluid flow. The cohesive zone is thought to consist of liga- the simulations, we used i=2.5 cc/s, µ=100 Pa⋅s, E=5.6 GPa, n
ments that connect the opposite fracture faces, which are ex- =0.2, T0=1.0 MPa, and bc=62 µm. In most simulations (except
pected not to hinder fluid flow significantly. The tensile cohe-
when indicated otherwise), we used σmin=2.5 MPa and σmax=4
sive stress is added to the fluid pressure loading in the cohe-
MPa as externally applied stresses working respectively per-
sive zone.
pendicular to and parallel with the fracture plane.
Grid and Boundary conditions First, we made a comparison with a fully coupled simulation
Figure 8 shows the boundary conditions and axisymmetric using the program of Barr17. We found that the width and
grid. We modelled half of the fracture width because of sym- pressure profiles of both programs agree well. This supports
12
elastic-plastic
the plausibility of the self-similar assumption, and indicates
elastic that the grid is fine enough and that the solution has indeed
8 converged far enough.
σy
Figure 9 shows the pressure and width profile in the elasto-
(MPa) fluid front plastic calculations for the final calculation step. The profiles
4 position
closure stress
are compared with the elastic case, for equal fracture length.

-4
0 0.02 0.04 0.06 0.08 0.10
r (m)
120
elastic-plastic
elastic

80
b/2
(µm) fluid front
position
40

0
0 0.02 0.04 0.06 0.08 0.10
r (m) Figure 10: Measured and simulated size of the dry tip as a function
Figure 9: Stress in the plane of the fracture and width profile of the maximum size (corresponding to a constant pressure in the
for elastic and elasto-plastic simulation. The fracture tip in fracture). Simulations and tests had different stresses and injection
both cases is at the same position. rates, yielding a different maximum fluid lag.
SPE 63172 IMPACT OF ROCK PLASTICITY ON HYDRAULIC FRACTURE PROPAGATION AND CLOSURE 7

The initial width is still visible in the final width profile of the tions, see Figure 10. Simulations based on LEFM tend to un-
elasto-plastic calculations, and causes a small irregularity in derestimate the observed size of the fluid lag. The same result
the pressure profile. These figures show that in the elasto- is obtained with a cohesive zone model with a very large ten-
plastic case, the fluid comes closer to the fracture tip, so that sile strength. Plasticity gives a larger length of the fluid lag,
the fluid lag length is smaller. Also when we take the change in which agrees with the experiments. For small and moderate
wellbore pressure into account, the size of the fluid lag be- values of the externally applied stress deviator, we found a fair
comes relatively smaller. This shows that plasticity ’screens’ agreement between the size of the fluid lag in our FLAC simu-
the fluid pressure loading, so that stronger loading (closer to lations and the experiments. In the experimental results, we
the tip) is present during propagation. Identical results were could not discern a influence of the varying degree of plastic-
obtained by Papanastasiou et al.8. ity. When the block was loaded close to failure by a very high
Figure 9 illustrates that plasticity leads to a larger wellbore maximum stress, the dry tip after splitting of the block disap-
width and smaller net pressure at the wellbore. This qualita- peared. This agreed with the observed trend in the simulations,
tively agrees with the experimental observations. When we that used acohesive zone with plasticity.
compare the elasto-plastic simulations with the elastic case, we Influence of plasticity on closure mechanism
can identify two processes that influence the wellbore pressure. Figure 11 shows the stress and opening during closure of the
First, in the body of the fracture the pressure gradient is fracture. The closure process is modelled with a constant pres-
smaller, because of the larger width. Second, the width at the sure in the open part of the fracture, which is equal to the con-
position of the fluid front is somewhat smaller for the elasto- fining stress. The figures show that the elasto-plastic fractures
plastic fractures, which increases the pressure gradient in this stay open over a certain radius, which is smaller than the origi-
region. Both effects influence the wellbore pressure. In our nal radius. In the closed part of the fracture, a high stress per-
experiments, the influence of the smaller gradient in the body pendicular to the fracture plane develops.
of the fracture is dominant. However, there are conditions in Elongation in the direction perpendicular to the fracture plane
which the influence of the tip gradient dominates the effect of occurs, and contraction in the direction parallel with the frac-
the gradient in the body of the fracture, so that a high net pres- ture plane. It is this elongation that acts like a wedge during
sure results while the fluid does not reach the fracture tip. closure, which causes the high stress that tends to keep the
We compared the measured size of the dry tip in plaster with fracture open. The plastic contraction in the direction parallel
the value obtained by fully coupled numerical elastic simula- with the fracture plane also hinders the fracture surfaces to
12
return to its original position during closure. This also tends to
elastic-plastic
elastic keep the fracture open, but causes lower stresses at the closed
8 fracture surfaces.
σy A Finite Element Fully Coupled Elastoplastic
(MPa) Hydraulic Fracturing Model – Field Application.
4 In the following we present computational results of another
closure stress
independent numerical study which verify the experimental
0
findings and extrapolates them to field scale. The study was
based on fracture propagation and closure of a plane strain
(KGD) elasto-plastic fracture. Rock deformation is modelled
-4 by Mohr-Coulomb plasticity for a cohesive-frictional dilatant
0 0.02 0.04 0.06 0.08 0.10
material. A cohesive model based on the softening behaviour
r (m)
120 of rocks is employed as the propagation criterion16. Fluid flow
elastic-plastic in the fracture is modelled by lubrication theory and the com-
elastic
plete mass conservation equation.
Depending on formation properties, insitu stresses and pump-
80
ing parameters, the fluid may not necessarily reach the fracture
b/2
tip thus allowing for the possibility of a dry zone (fluid lag)
(µm) near the fracture tip. A special continuation method based on
40 the volume of injected fluid in the fracture is used for direct
coupling of the fluid-flow with rock deformation and for driv-
ing the solution during propagation. A mesh similar to the one
shown in Fig. 8 was used and a meshing/remeshing scheme
0
0 0.02 0.04 0.06 0.08 0.10 was employed in order to carry out longer propagations with
r (m) fine mesh near the fracture tip. Detailed information on the
Figure 11: Stress in the plane of the fracture and width profile numerical implementation are provided in Ref. 10.
for elastic and elasto-plastic simulation, after fracture closure. The parameters upon which the numerical computations were
In the open part of the fracture, the applied fluid pressure
equals the confining stress.
based are given in Table 3. The only extra material parameters
8 D.B. van Dam, P. Papanastasiou and C.J. de Pater SPE 63172

−3 5
x 10
4
plasticity
1 3

net−pressure (MPa)
2 elasticity
0.5
width (m)

1
0 0

−1
−0.5
−2
−1 −3

0 0.5 1 1.5 2 2.5


0 0.5 1 1.5 2 distance from wellbore (m)
distance from wellbore (m)
Figure 14: Net-pressure profiles on fractures
Figure 12: Fracture profiles for an elasto-plastic (solid lines) and
an elastic (dashed-lines) fracture. opening of the elastic fracture results in greater pressure drop
required for propagating an elastoplastic fracture are the plas- near the fracture tip and significant fluid-lag. Higher energy is
tic constants. With the chosen material parameters and in-situ required for propagating the elastoplastic fracture. Most of this
stresses the rock is initially elastic but very close to a yielding energy is coming from the smaller fluid lag region. This result
state. again contradicts the dilation hypothesis.
Figure 12 shows the profiles of propagating elastic (solid- Plastic yielding near the tip of a propagating fracture provides
lines) and elasto-plastic (dashed-lines) fractures for the same an effective shielding, resulting in an increase in the effective
fracture lengths The cusping of the crack tips with zero slope is rock fracture toughness11. This is shown in Fig. 15 where we
a result of the cohesive model which was incorporated as the plotted the increase in the effective fracture toughness as a
propagation criterion. As mentioned earlier, the model of the function of the fracture growth. The effective fracture tough-
elastoplastic fracture requires an initial fracture length which ness was determined using the calculated, path independent, J-
was set to 0.5 m; the influence of this can be seen in the width integral18. The value of the effective fracture toughness is di-
profiles in Fig. 12. If we compare the fracture openings in the rectly related to the size of the plastic zones.
region where the fractures were propagated (i.e., distance from The elastoplastic fracture was propagated further to reach a
wellbore between 0.5 and 2.2 m) we see that the width profile length of 8 m before examining the closure pattern (Fig. 17).
of the elastoplastic fracture is much wider than the width pro- Figure 16 shows the width profile of a receding elastoplastic
file of the elastic fracture. fracture. It was assumed that there was no fluid-flow in the
Figure 14 shows the comparison of the net-pressure profiles in fracture during closure, therefore the pressure acting along the
the fracture for the same fracture length. The very narrow fracture was constant. The elastoplastic fracture makes contact
initially near the tip and then the closure moves towards the
−3
x 10 wellbore13. This closure pattern agrees with the experimental
3

10
2
apparent fracture toughness (MPa m^0.5)

1 8
width (m)

0 6

−1 4

−2 2
fracture toughness

−3 0
0 2 4 6 8 0 0.05 0.1 0.15 0.2
distance from wellbore (m) fracture extension (m)
Figure 13: Propagation of an elasto-plastic fracture. Figure 15: Apparent fracture toughness vs fracture extension
SPE 63172 IMPACT OF ROCK PLASTICITY ON HYDRAULIC FRACTURE PROPAGATION AND CLOSURE 9

−3
x 10
3 6
5
2

net−pressure (MPa)
4
1
3
width (m)

propagation
0 2
1
−1
0

−2 −1 closure
−2
−3 0 2 4 6 8
0 2 4 6 8 fracture length (m)
distance from wellbore (m)
Figure 17: Net-pressure at wellbore vs fracture length
Figure 16: Closure of elasto-plastic fracture
results reported earlier on soft rock samples (Fig. 7). Discussion
Figure 17 shows the corresponding net-pressure at the well- Rock plasticity leads generally to a relatively small increase of
bore vs fracture length during propagation and closure after the the fracture pressure. In the conditions of our lab tests we even
fracture has reached a) 5 m and b) 8 m length. These results found that the propagation pressure was lower than expected
show that the net-pressure drops to zero while the fracture is from elastic models. However, plasticity has an important in-
still wide-open along a large proportion of the original length. fluence on the geometry of the fracture: the width becomes
The decrease of the fracture surface during closure will result much larger compared with elastic rock behaviour. Numerical
in the underestimation of the insitu leak-off coefficient at the modelling explains the deviations of the geometry of plastic
late stage of the pressure decline analysis. More accurate leak- fractures compared with elastic rock. In particular, we show
off predictions can be obtained in the early stage of pressure that the observations at laboratory scale are also valid at large
decline analysis, immediately after shut-off of the pumping scale.
stage. The results of Fig. 17 suggest that the fracture will close An important conclusion from our numerical work is that the
completely at negative net-pressures (fluid pressure less than rock behaviour at the fracture tip cannot be described by linear
the far-field stress). Application of classical analysis, which elastic fracture mechanics. A better description of the tip be-
assumes that the fracture closes completely when the fluid- haviour is obtained with a cohesive zone over which the rock
pressure drops to the value of the far-field stress, would lead to looses its bearing capacity. In the lab tests we observed a fluid
the underestimation of the minimum insitu stress. Furthermore, lag size that was much bigger than the prediction from elastic
special considerations are necessary to identify the time instant models. Field observations of the fluid lag are hard to perform,
the pressure in the fracture is equal to the far-field confining but the available data23 seem to indicate a much larger fluid
stress. This has been recognized by Nolte19 who introduced lag than predicted by elastic models. This appears to agree
diagnostic methods to detect deviations from ideal behaviour
in practice. 0
Figure 18 shows the stress profiles on the propagation plane
after the last closure step. The irregularity in the stress
distribution in the area near the wellbore is due to the effect of S_hmin
principal stresses (MPa)

the initial fracture length. As a result of the fracture closing


−5
first near the tip, the closure stress σhmin is higher than the in-
situ stress in this area (5-8 m) and lower near the wellbore.
This implies that the closure stress on the proppant will be
higher near the fracture tip and lower in the area near the
S_hmax
wellbore (σhmin in Fig. 18). In addition, after fracture closure, −10
due to permanent deformation, the stress state does not return
to the original state. The risk of formation failure is
significantly reduced after fracturing and closure not only due
to increase of minimum insitu stress as a result of the enlarged S_vert
−15
propped fracture but also due to decrease in the other two 0 2 4 6 8 10
distance from wellbore (m)
insitu stresses13.
Figure 18: Stresses on fracture plane after closure (Note that
compressive stress is negative in this figure).
10 D.B. van Dam, P. Papanastasiou and C.J. de Pater SPE 63172

with the laboratory observations. By modelling the cohesive than the first load static modulus8. Also, the large discrepan-
zone and rock plasticity, we obtain a fair match between the cies between predicted net pressure and observed net pressure
observed length of the fluid lag and the predicted length by the in soft rock could be explained by incorrect determination of
numerical model. We also found that rock dilation leads to an the closure stress.
increased fracture width at the tip. This is opposite to the so- Our experimental data show that interpreting fracture geometry
called dilation hypothesis that has been proposed to explain based on pressure measurement alone is insufficient. In the lab
high net pressures2. It was assumed that dilation of the rock tests we have benefited from active acoustic monitoring that
would suppress the width in the fluid lag region, thereby en- gave us independent measurements of the fracture geometry.
hancing the size of the fluid lag. The longer fluid lag would This is also becoming feasible in field applications in view of
lead a much higher propagation resistance. Instead of dilation, recent developments in low frequency sonic logs that can look
the most important effect of plasticity is in the shielding of the away from the wellbore. Furthermore, fracture mapping with
fluid load from the fracture tip. This results in a strong redis- passive microseismics21 and tiltmeters22 can provide important
tribution of the stresses over the fracture length and since a data to confirm fracture models in real rock. For soft forma-
higher pressure is needed to rupture the rock at the tip, the tions, tiltmeters may be most useful since these formations do
width in the fluid lag region becomes actually larger. The not induce much seismic activity.
stress redistribution in plastic fractures yields a much larger Our study was limited to lab experiments and numerical simu-
width for a given net pressure. In this respect, the idea that lation. The findings still need to be confirmed with field obser-
fractures are much wider (and shorter) in reality than predicted vations, although the available data generally agree with our
by elastic models is corroborated by our study. The ‘forgiving’ results. More measurements of fracture length are needed in
nature of hydraulic fracturing might even result in a correct soft formations. Also, measurements of the fluid lag in field
design when one picks the closure from a pressure decline, applications would be needed to confirm the modelling results.
uses the high net pressure with a modulus that is wrong to cal- Conclusions
culate a fracture width and fracture length. However, for opti- • Soft plaster blocks showed a wellbore width that was sig-
mising treatments it will be better to use the correct modelling nificantly larger and a pressure that was lower than the
assumptions, rather than relying on serendipitous cancelling elastic simulation.
out of many errors. • Compared to the elastic case, plasticity results in larger
When should one expect plasticity to play a role in fracturing? width everywhere in the fracture even though dilatancy
As the experimental results on different rock types show, it is tends to close the fracture near the tip. This effect influ-
not so much the strength of the rock, but the brittleness that ences the wellbore pressure.
determines the level of plasticity. Diatomite is very weak, but • Plastic deformation around the fracture tip influences the
in tensile tests, it ruptured in a brittle way. Soft plaster was relation between wellbore pressure and fracture width. For
very ductile in tensile and extension tests and showed signify- the conditions in our experiments on soft plaster, plastic
cant deviations from elastic behaviour in hydraulic fracture deformation around the tip tends to keep the fracture open
tests. This seems a good indicator of the importance of plastic- at the wellbore at a lower pressure than in the elastic case.
ity: a large plastic deformation in a tensile or extension test. • Numerical simulations with a cohesive zone model and
Conventional models of hydraulic fracture closure make as- plastic behaviour quantitatively predict the size of the dry
sumptions such as constant fracture area and constant stiffness. tip in plaster fairly well for small and moderate values of
We found that the fracture recessed during closure, but the the applied stress deviator.
most dramatic consequence of plasticity is a significant de- • Closure pressure in plastic rock may be significantly lower
crease in closure pressure. It appears that the opening of the than the far field stress on the fracture plane. This applies
fracture changes the stress distribution such that the fracture both for laboratory size fractures and large scale fractures.
remains open at the wellbore. The fracture closes only when • Stress redistribution after closure of the fracture is impor-
the pressure falls much below the minimum stress. tant in plastic rocks. The shear stress near the wellbore is
The observed deviations in closure pressure from the nominal significantly lower than expected from elastic behaviour.
confining stress have a big impact on estimating the net frac- Acknowledgments
ture pressure in practice. Also, the change in closure pressure We thank Gerard Mathu and Jan Etienne for the technical as-
is important for designing frac&pack treatments, where sand- sistance with the experiments. The research at TU Delft was
face stabilisation is an important main objective. sponsored by the companies in the Delfrac consortium and the
Practical Implications Dutch Technology Foundation (STW). Part of this research
Controlled lab experiments show that hydraulic fracture behav- was carried out in Schlumberger Cambridge Research.
iour can be explained in a consistent manner, provided one References
takes account of the physics of the rupture process at the tip. 1. Shlyapoberski, J., G.K. Wong and W.W. Walhaug (1988), "Over-
Our results indicate that in many field applications so-called pressure calibrated design of hydraulic fracture stimulations",
high net pressures can be explained by underestimating the SPE Paper 18194.
rock modulus (which may be closer to the dynamic modulus
SPE 63172 IMPACT OF ROCK PLASTICITY ON HYDRAULIC FRACTURE PROPAGATION AND CLOSURE 11

2. Johnson, E., and M.P. Cleary, 1991, "Implications of Recent Labo- 1990 SPE Annual Technical Conference and Exhibition, New
ratory Experimental results for Hydraulic Fractures", SPE Orleans, LA, Sept. 23-26.
21846, paper presented at the Rocky Mountains Regional Meet- 20. Hoek, P.J. van den, J.T.M. van den Berg, J. Shlyapobersky, 1993,
ing and Low-Permeability Reservoirs Symposium, Denver, "Theoretical and Experimental Investigation of Rock Dilatancy
Colorado, April 15-17. Near the Tip of a Propagating Hydraulic Fracture", Int. J. Rock.
3. Dam, D.B. van: "The Influence of inelastic rock deformation on Mech. Min. Sci. & Geomech. Abstr. 30, No. 7, pp. 1261-1264.
hydraulic fracture geometry", Ph.D. Thesis, Delft University of 21. Warpinski, N.R., Branagan, P.T., Peterson, R.E., Fix, J.E., Uhl,
Technology (1999). J.E., Engler, B.P., and Wilmer, R., Microseismic and deforma-
4. Dam, D.B. van, C.J. de Pater, and R. Romijn: “Experimental Study tion imaging of hydraulic fracture growth and geometry in the c
of the Impact of Plastic Rock Deformation on Hydraulic Frac- sand interval, gri/doe m-site project., In Proceedings of the SPE
ture Geometry”, Int. J. Rock Mech. & Min. Sci. (1997) 34, No. 1997 Annual and Technical Conference and Exhibition. Society
3-4, Paper No. 318, 439. of Petroleum Engineers,, 1997., SPE 38573.
5. Dam, D.B. van, C.J. de Pater, and R. Romijn: "Analysis of Hy- 22. Warpinsky, N.R., P.T. Branagan, B.P. Engelder, R. Wilmer, and
draulic Fracture Closure in Laboratory Experiments", SPE S.L. Wolhart, (1997), "Evaluation of a Downhole tiltmeter Ar-
47380, paper presented at the 1998 SPE/ISRM Eurock Sympo- ray for Monitoring hydraulic Fractures", Int. J. Rock. Mech.
sium, July 8-10, Trondheim, Norway. Min. Sci. & Geomech. Abstr., Vol. 34: 3-4, paper no. 329.
6. Dam, D.B. van, C.J. de Pater, and R. Romijn: "Reopening of Hy- 23. Warpinski, N.R., (1985), “Measurement of Width and Pressure in
draulic Fractures in Laboratory Experiments", paper presented a Propagating Hydraulic Fracture”, SPE Paper 11648, SPEJ, p.
th
at the 9 ISRM congress on rock mechanics, August 1999, 46.
Paris, France. Nomenclature
7. D.B. van Dam and C.J. de Pater, (1999), “Roughness of Hydraulic Units: SI (m= metre, s= second, Pa=Pascal), Dimensions:
Fractures: The Importance of In-Situ Stress and Tip Processes”, m=mass, L=length, t=time
SPE Paper Proc. 1999 SPE Annual Technical Conference and Variable Description Units Dimensions
Exhibition held in Houston, Texas, 3–6 October 1999.
8. Papanastasiou, P., M. Thiercelin (1993), “Influence of inelastic b : width of the fracture [m] (L)
rock behaviour in hydraulic fracturing”, Int. J. Min. Sci. & E : Young’s modulus, E = E/4(1-n2) [Pa] (m/Lt2)
Geom. Abstr., Vol 30, No 7, pp 1241-1248. ϕ : porosity
9. Papanastasiou, P. (1997). A coupled elastoplastic hydraulic frac- 3 3
i : injection flowrate L /t m /s
turing model. International Journal of Rock Mechanics and 2 2
k : permeability, L , m
Mining Science, 34: 3-4, paper No. 240. 1/2 2 1/2
KIc : fracture toughness, m/L t Pa⋅m
10. Papanastasiou, P. (1999a). An efficient algorithm for propagating
fluid-driven fractures. Computational Mechanics, 24, 258-267. m : viscosity, m =12m [Pa s] (m/Lt)
11. Papanastasiou, P. (1999b). The effective fracture toughness ap- n : Poisson ratio [-] (-)
proach in hydraulic fracturing. International Journal of Frac- pn : net fracture pressure [Pa] (m/Lt2)
ture, 96, 127-147. R : fracture radius L m
12. Papanastasiou, P. (1999c) Formation stability after hydraulic S’vert : Vertical effective stress [Pa] (m/Lt2)
fracturing. International Journal for Numerical and Analytical
S’hmax : Minimum effective horizontal stress [Pa] (m/Lt2)
Methods in Geomechanics, 23, 1927-1944.
13. Papanastasiou, P. (2000). Hydraulic fracture closure in a pres- S’hmin : Maximum effective horizontal stress [Pa] (m/Lt2)
2
sure-sensitive elastoplastic medium. International Journal of σc : stress perpendicular to fracture surface Pa, m/Lt
Fracture, (in print). σhmax : largest stress parallel with frac surface Pa , m/Lt2
14. Groenenboom, J. 1998. Acoustic Monitoring of Hydraulic Frac- Su : Initial UCS [Pa] (m/Lt2)
ture Growth. PhD thesis, Delft Univ. of Technology, The Neth- t : time [s] (t)
erlands. T0 : tensile strength [Pa] (m/Lt2)
15. Crockett, A.R, N.M. Okusu, and M.P. Cleary, 1986, "A Complete
Integrated Model for Design and Real-Time Analysis of Hy-
w : length of the non-penetrated zone [m] (L)
th
draulic Fracturing Operation", SPE 15069, Proc. 56 California y : Dilation angle [deg] [-]
Regional Meeting of Petroleum Engineers, Oakland, CA, April SI Metric Conversion Factors
*
2-4. cp × 1.0 E − 03 = Pa ⋅ s
*
16. Hillerborg, A., M. Modeer, and P.-E. Petersson, 1976, "Analysis ft × 3.048 E − 01 = m
of crack formation and crack growth in concrete by means of 2 * 2
ft × 9.290 304 E − 02 = m
fracture mechanics and finite elements", Cement and Concrete 3 3
Research, Vol. 6, pp. 773-782.
ft × 2.831 685 E − 02 = m
*
17. Barr, D.T. 1991. Leading-Edge Analysis for Correct Simulation in. × 2.54 E + 00 = cm
of Interface Separation and Hydraulic Fracturing. PhD thesis, lbf × 4.448 222 E + 00 = N
Massachusetts Institute of Technology, USA. md × 9.869 233 E − 04 = µm
18. Rice, J.R. (1968), “A path independent integral and the ap-
*
psi × 6.894 757 E + 00 = kPa
proximate analysis of strain concentration by notches and Conversion factor is exact.
cracks”, J. Appl. Mech., Trans. ASME, 35, p379-386.
19. Nolte, K.G., 1990, “Fracturing Pressure Analysis: Deviations
From Ideal Assumptions”, paper SPE 20704 presented at the
12 D.B. van Dam, P. Papanastasiou and C.J. de Pater SPE 63172

TABLE 1- MATERIAL PROPERTIES AND EXPERIMENTAL CONDITIONS


Material Eunl k ϕ KIc S0 µf
(GPa) (mD) (-) (MPa√m) (Pa⋅s) (-)
-3
Cement 18 10 0.2 0.5 - -
1
strong plaster 9 5⋅10 0.42 0.3 6.2 0.47
1
weak plaster 5 5⋅10 0.42 0.3 2.6 0.47
Diatomite 0.2 0.2 0.70 0.03* 0.45 0.47
* Estimated value.

TABLE 2:OVERVIEW OF THE EXPERIMENTAL CONDITIONS IN ALL EXPERIMENTS.


Material name σh,1 σh,2 σc Qpump µ
(MPa) (MPa) (MPa) (cc/min) ( Pa⋅s)
cement ce01 33 33 23 0.088 500
cement ce02 33 33 23 0.088 500
cement ce03 33 33 23 0.088 500
cement ce04 12 12 8 0.088 500
cement ce06 33 33 23 0.088 500
cement ce07 12 12 8 0.088 500
cement** ce08 33 33 23 0.088 500
cement ce09 33 33 23 0.022, 0.704 500
cement ce10 33 33 23 0.352 500
Strong plaster sp01 12 12 8 3 500
strong plaster sp02 20 20 16 3 500
strong plaster sp03 28 28 16 3 500
strong plaster sp04 20 20 8 3 500
strong plaster sp05 12 12 8 3 500
strong plaster sp06 20 20 8 3 500
strong plaster sp07 24 24 8 3 500
strong plaster sp08 24 17 10 3 500
strong plaster sp09 26 26 8 3 500
strong plaster sp10 20 20 4 3 500
weak plaster* wp01 8 8 4 3 500
weak plaster wp02 12 12 8 3 500
weak plaster* wp03 20 20 8 3 500
weak plaster wp04 4 4 2.5 2.5 100
weak plaster wp05 4 4 2.5 4.17 60
weak plaster wp06 4 4 2.5 20 12.5
weak plaster wp07 7.5 7.5 2.5 20 12.5
weak plaster wp09 7.5 7.5 6 20 12.5
diatomite di01 1.28 1.28 0.85 0.5 500
diatomite di02 1.28 1.28 0.85 0.5 12.5
diatomite di03 2.125 2.125 0.85 2.5 100
diatomite di04 2.45 2.45 1.30 2.5 100
diatomite di05 2 1.5 1 4.17, 0.104 60
* intermediate water content.
** after previously creating a fracture with a small width using water.
SPE 63172 IMPACT OF ROCK PLASTICITY ON HYDRAULIC FRACTURE PROPAGATION AND CLOSURE 13

TABLE 3. INPUT PARAMETERS FOR FINITE ELEMENT COMPUTA-


TIONS.
Pumping parameters
Fluid viscocity µ=0.1 Pa sec=100cp
Flow rate i = 0.0005 m3/sec
Effective Insitu stresses
Vertical S’vert = 14.49 MPa
Minimum horizontal S’hmax = 9 MPa
Maximum horizontal S’hmin = 3.7 MPa
Elastic constants
Young’s modulus Eunloading = 9Eloading MPa
Poisson ratio ν = 0.3
Plastic parameters
Friction angle φ = 28o
Dilation angle ψ= 28o
Loading modulus Eloading = 1.785 GPa
Initial UCS Su = 4 MPa
Fracturing parameters
Fracture toughness KIc = 1 MPa m1/2
Tensile strength T0 = 0.5 MPa

You might also like