Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Introduction
This section reviews the basic physics of hydraulic fracture propagation, growth, and final fracture
geometry. In addition to leading toward a general understanding of fracture growth, reviewing the basic
physics emphasizes which variables are critical to treatment design. As with all real world processes, the
behavior of hydraulic fracture is a complex function of many variables, however, for the most processes,
a minor number of these variables dominate the process and control; the results. These are critical
parameters or variables, and good information must be available for these critical variables in order to
understand and design the process (in this case to design a hydraulic fracture treatment). For fracturing,
the critical parameters are height, (or the in-situ stress which controls the height); Youngs modulus, ;
fluid loss coefficient, ;and (sometimes) fracture toughness, , e.g.; HECK. Other variables which are
generally less important in controlling fracture geometry include pump rate, , and fluid viscosity,.
Finally, it should be noted that fundamental fracture geometry models discussed in this section from
theoretical basis for Fracturing Pressure Analysis.
Continuity Equation
Determining (or designing) a created hydraulic fracture length can be subdivided into two brad
categories- fracture geometry and fluid loss. On a volume basis this is written as
= + (1)
Which, rearranged, states that the end product of fracture volume is equal to the volume pumped minus
the volume which leaks-off to the formation during pumping. This is the volume balance or material
balance equation, or continuity equation, and the components can be broken down as seen in Eq. (2)
= =
(2)
3 + 2
(3)
. . , (4)
end product,
Combining these gives an approximate relation for one of the major design variables, fracture length.
where is pump rate in ft3/min (5.614 BBL/min) is total pump time in minutes. C is the fluid loss
coefficient in units of feet/min (with typical values of .0005 to .1). Hp
is the permeable or leak-off height of fracture while is the total gross height of fracture in feet, Sp is
spurt loss in ft3/ft2(.00134 gal/100ft2), with typical values of 0 to 20 gal/100ft2, and w is the average
fracture width (averaged over fracture length and over fracture height ) in feet.
This simple relation is at the heart of all fracture design and is, in itself, a powerful tool. By assuming a
nominal fracture width (say inch = .02 feet 0.6 cm, this simple equation can give surprisingly good
estimates of how altering input data will affect fracture length.
Eq. (5) also illustrates the principal variables in fracture design: 1), width, which is discussed below;2)
C Hp leak off capacity or fluid loss coefficient times leak-off time;3) H, gross or total fracture height.
If H=Hp, as is often the case, then generally height becomes the dominant variable governing fracture
length as seen in Fig. (1).The effect of leak off or fluid loss coefficient, C on fracture length is illustrated
in Fig (2). Note that fracture length versus height and length versus C , in the figures are based on two
dimensional fracture geometry models, which are discussed in more detail below.
Fracture Area
In the previous section fracture geometry, or fracture area, was identified with three terms , and .
For the particular case of a fixed height fracture (and ignoring spurt loss for the moment), the area would
be equal.
= (6)
and the fluid loss area would equal
= = = (7)
The first fracture area (or fracture growth) equation was from Carter [1] and this implicitly inclined
= 1.
2 2
= 2 () + 1,
4
(8)
=2
Harrington [2], et al, in 1973 developed a general relation for fluid volume loss (dale to "matrix" or
"solid loss coefficient" loss) as
= 8 (9)
and Nolte [3] showed that under various bonds the coefficient ( 8 in this case) of this equation could
only vary between 8/3 and 8. Thus, for = 1, and no spurt loss,
3 (10)
re-including spurt loss revises this equation to
3 + 2 (11)
and using the material balance equation gives
= (12)
3 + 2 +
Fracture Geometry
The above discussion arid the area equations leads to an extremely important concept in fracturing, the
fracture fluid efficiency. This is simply the fracture volume divided by the total volume pumped. Using
the equations above this can be written as
=
==
3 + 2 +
(13)
=
which is a powerful number for treatment scheduling (e.g., pad volume determination and proppant
addition scheduled. It is also interesting to note how fluid efficiency varies with time, and how "spurt
loss" effects florid efficiency. Also, note that fluid efficiency is referred to as a single number (since
efficiency varies with time during pumping), the reference is to the efficiency just at shutdown.
Reviewing the equations above, it is clear that for spurt loss equal to "0", then for small values of time the
fluid efficiency will always be quite high (e.g., for a time of "0", efficiency will be 100%) and that
efficiency will be monotonically decrease with time. However, for a "non-0", value of spurt, the early
time efficiency will be lower (possibly much lower), and during early stage of fracture growth, fluid
efficiency increases as fracture width increases. As time increases the fluid loss coefficient, C, tern will
become more dominant, and again (even with spurt loss) fluid efficiency will begin to decrease. Just as
with many aspects of fracturing, behavior can vary from place-to-place depending on the exact formation
variables!
Fracture Width
If a Slit inside an elastic rock formation is opened by internal pressure, then that silt will open up into
an elliptical cross section crack (Fig. (3)) with a maximum width given by [4]
2 ( )
= =2 , (14)
where = . With having typical values (for rocks) of .15 to .25, is essentially equal to for all
12
practical purposes.
This fundamental elasticity solution was applied to hydraulic fracturing, but it was applied in two
different ways. This resulted in two competing fracture width models, and there is still debate over
which is correct. The question of which model to use is very important since they can predict quite
different fluid, pad, and proppant volumes needed to achieve a required stimulation as seen in Table (2).
NOTE: This example is for a low efficiency (fracture volume/injected volume) case. For low efficiency,
length is dominated by fluid loss and is thus almost width model independent. For higher efficiency, the
resulting lengths of different models would differ much more significantly.
The two width models are commonly called the Perkins & Kern [5] PKN or Geertsma-deKlerk [6] GdK
models and are illustrated in Fig. (4) (along with a third "width model" for radial or penny shaped
fractures for horizontal fractures or vertical fractures with no height confinements). Some basic
mathematical development for a Geertsma-deKlerk, GdK fracture model was proposed by a Russian,
Khristianovic [7]; thus often this fracture geometry is referred to as the KGD model (Khristianovic-
Geertsma-deKlerk).
Though there is debate, general evidence is that the PKN model is more applicable. In a 1976 article [8]
Geertsma concluded For practical application these basic differences suggest that the Perkins and Kern
PKN model is most appropriate for length/height ratios much in excess of unity, while the GdK model is
most appropriate for small values of L/H. "This analytical conclusion was later supported by fracture
width measurements made with a down hole television camera [9], and by mine back experiments done
by Sandia Laboratories [10]. This is also seen in a comparison of calculated fracture widths in Fig. (5).
The correctness of ignoring the fracture tip considerations for PKN fracture geometry (e.g. L > H) in most
situations arises from the fact that net pressure inside a hydraulic fracture (and thus fracture width) is a
result of interaction of fluid flow in the fracture, and the elasticity of the rock. In effect, viscous properties
of the fluid flowing down the entire length of a fracture are generally more dominant than rock
mechanics considerations near the fracture tip. For a PKN model, this relation between fluid flow/fracture
width is discussed below. The relation between fluid flow and fracture width is summarized for all the
remodels in Table (4). [NOTE - Ignoring detailed fracture mechanics processes near the fracture tip is all
right for many cases, particular for fractures with some height confinement and for fractures in normal
rock. However, for other cases, fracture toughness and fracture tip effects can dominate the process. This
is particularly true for radial fractures and fractures in very soft (e.g. low modulus) rocks, and this is
discussed in more detail later.
For a Newtonian viscous fluid, in laminar flow down narrow silt, pressure drop is given by
( ) 12
= = (15)
3
For the PKN model, the elliptical cross section is assumed to be a vertical cross section of the fracture
(e.g., the fracture is closed at to top and bottom) so that
2
= = = (16)
4 4
The fluid flow and width equations can be parameterized as
12
3 = (17)
and
2
= (18)
And these combined to give
1
4 (19)
1
2 +2
(19)
( 1)
where is the consistency index, and is the non Newtonian flow index for a power law fluid. For a
Newtonian fluid = 1, = and this reduces to Eq.
This can be recombined with the width equation to predict the net pressure
3
4 (20)
= ( ) [ ]
showing that as fracture length extends, the net treating pressure increases.
Eq. (20) for the PKN model and the similar relations in Table (4-1,4-2) for other models are very
illustrative of the "sensitivity" of fracture width to other parameters (viscosity and pump rate). For a
1
parameter sensitivity only to the 1/4 power, a 50% error results in only a 10% change (1.54 = 1.1) in the
calculated value so we see the calculations for fracture width are relatively insensitive to reasonable
uncertainties in these variables. This limited dependence on fluid viscosity is very important since
rheology of typical fracturing fluids (particularly gelled oil and cross-linked gelled water) are difficult to
characterize and measure, and exact rheology is sensitive to field mixing conditions. Thus there is
generally ALWAYS some uncertainty with respect to fracturing fluid viscosity.
Fracture Tip Effects, Fracture Toughness
While it was not emphasized, developing the relations between net pressure, , and height, pump rater
etc., in Table 4-1, implicitly assumed net pressure was "zero" right at the fracture tip. This is clearly not
the case since some non-zero pressure is required to break the rock and keep the fracture propagating. In a
Appendix
PKN Model:
qi x f
w = 0.19[ ]1 / 4 (in field units, Newtonian fluid)
E'
w
wmax =
0.628
Non-Newtonian Fluid:
1 n '
2n'+1 n ' 0.9775 5.61 n ' 1 /( 2 n '+2 ) qi K ' x f h f 1 /( 2 n '+2 )
n'
128
wmax = 12[( )(n'+1)( ) ( )( ) ] ( )
3 n' 144 60 E
qi x f
2
wmax= w = 0.34[ ]1 / 4 ( )
E' h f 4