Rock acoustics and rock mechanics:

Their link in petroleum engineering

Downloaded 02/20/17 to 137.195.101.233. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

By ERLING FJÆR
IKU Petroleum Research

and RUNE M. HOLT

IKU Petroleum Research/NTH Norwegian Institute of Technology
Trondheim, Norway

S eismic data represent the most important source of infor- to be largest for low stresses; for higher stresses the velocities
mation about underground formations that have not been become increasingly independent of the stress.
accessed by excavations or drilling. When target formations The effect of a stress increase in only one direction is quite
can be reached by drilling, more accurate velocity data are different (Figure 3). First, the velocities that are either prop-
available via logging. The primary use of the seismic infor- agating or are being polarized in the same direction as the
mation is still to map structure. However, there is a growing increasing stress are also increasing; the other velocities
interest in using seismic data to derive intrinsic information remain constant. At higher stresses, the velocities of wave
about rock properties such as porosity, clay content, etc. propagating or being polarized normal to the increasing stress
Porosity estimation has traditionally also been the main ap- are decreasing significantly. The result is an increasing acous-
plication of sonic logs. In fact, an established school of tic anisotropy, reflecting the stress anisotropy. The degree of
thought holds that velocities in sedimentary rocks are primar- this stress-induced anisotropy differs from one rock to an-
ily a measure of porosity-the lower the velocity, the higher other. Also, the degree depends on the confining pressure.
the porosity. The effect is smaller for higher confining pressures.
In this context, it is important to notice that velocities may
be strongly dependent on the stress state as well as the stress C racks. It turns out that stress-induced changes in velocities
history of the rock. This is of particular importance when we (Figures 2 and 3), as well as stress history effects (Figure 1),
wish to compare velocities measured under different stress can be explained micromechanically only in terms of cracks.
conditions-in situ stresses, near wellbore stresses, no This seemingly simple-minded explanation of a rather com-
stresses at all-and we often ignore or are unable to account plex behavior does require a somewhat wider definition of
for the effects of the stress history of the rocks. An example the term cracks; but, it is both attractive and useful that the
of the potential impacts of the stress history is shown in Figure stress and stress history effects of sound velocities in many
1. The diagram shows the P-wave velocity versus porosity rocks can be intuitively explained by tracking the stress-in-
(measured under identical conditions) for a set of samples that duced opening and closure of cracks.
have the same lithological characteristics except for stress Any soft inclusion in the material, like a spherical pore or
history. The normal trend, with the velocity decreasing with a flat crack, will reduce both stiffness and, to a lesser degree,
increasing porosity, is inverted, indicating the strong impact density so that the net result is a reduction in the sound
of stress history. These results will be explained in terms of velocities. This explains the normal trend (as shown in Figure
micromechanical processes later in this paper. 1) where sound velocities decrease with increasing porosity.
Theoretically, a sound velocity V in a material is given by A flat crack has a very small volume compared to its exten-
a stiffness C and the density D as Given that density sion, hence the reduction in density due to a crack is marginal
can be obtained by other means, the velocity can be regarded compared to the reduction in stiffness, and the reduction in
as a measure of the stiffness. Experience shows there is often velocities comparatively strong. However, the effect of a
a significant difference between a “dynamic” stiffness ob- crack on the stiffness depends strongly on the orientation of
tained from velocity measurements and the corresponding the crack: the stiffness measured normal to a crack is strongly
stiffness obtained by static measurements. This difference reduced because of the crack but the stiffness parallel to the
depends both on the rock type and the stress state. crack is only marginally affected. Thus, a nonuniform distri-
bution of the orientation of cracks will result in acoustic
I mpact of stress on sound velocities: observations. Sound anisotropy. An example of a remarkably strong crack-in-
velocities in granular rocks generally increase with increasing duced anisotropy was presented by Holt et al. (Canadian
external pressure (Figure 2). This is well known and can be Journal of Exploration Geophysics, 1993) regarding the de-
intuitively explained as being due to a broadening and corre- teriorated soapstone from the Nidaros Cathedral in
sponding stiffening of the grain contacts. The effect is seen Trondheim. The shear wave anisotropy was found to be

APRIL 1994 THE LEADING EDGE 255

 50-100 percent, and for a specific direction in the rock the
fastest shear wave was even faster than the compressional
wave!
Although the effects of an open crack on the acoustic
properties of a material have often been described theoreti-
cally for more than three decades, the theories went untested
due to the difficulties in providing physical models with well
defmed crack geometry. The situation has changed recently
Downloaded 02/20/17 to 137.195.101.233. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

and reports on experimental tests are now appearing. A tech-

nique developed by Jaswant Rathore et al. (see “Experimental
versus theoretical acoustic anisotropy in controlled crack
synthetics”, SEG Expanded Abstracts, 199 1) allows manu-
facture of artificial sandstones containing cracks that are
exactly defined in size, shape, numbers, and orientation. The
technique has been used to demonstrate that flow of fluid
between cracks and pores may have significant impact on the
velocity of compressional waves.
Figure 1. P-wave velocity versus porosity for a set of dry
synthetic sandstones. The velocities are measured parallel S tress effect on sound velocities: explanation. In terms of
(closed symbols) and perpendicular (open symbols) to the the concept of cracks, the effects of stresses on sound veloc-
minor principal stress during unloading. ities can be explained as follows:
Compressive stresses will tend to close the cracks oriented
normal to the direction of stress. When the cracks are closed,
their impact on sound velocities is eliminated. Thus, a hydro-
static stress will produce an increase in sound velocities due
to closure of cracks.
If the definition of cracks is expanded to include poorly
cemented grain contacts, this concept generally explains why
sound velocities increase with increasing confining stress.
The possibilities for closing new cracks by an additional
increase in the confining stress depend on the number of
cracks still open. Since this number decreases with increasing
stress, it follows that the rate of velocity increase will decrease
with increasing stress (Figure 2).
Closed cracks oriented at a skew angle relative to the
increasing stress in a uniaxial stress situation may at some
point start to slide (Figure 4). This motion of the crack faces
will normally be accompanied by the opening of so-called
“wing cracks” that are oriented normal to the minimum
Figure 2. Typical development of sound velocities as a principal stress, i.e., parallel to the increasing stress in a
function of external hydrostatic pressure. uniaxial stress situation. This process is not initiated before
the rest friction of the crack faces has been overcome; hence,
the effect will mostly be seen at higher stresses. For propa-
gating elastic waves, only the open wing cracks will be
visible; consequently, the effect is to reduce the velocities of
the waves that are propagating or are being polarized normal
to the increasing stress.

S tress history. As mentioned earlier, the compressional and

shear velocities depend not only on the stress state but also
on the stress history of the rock. This is clearly illustrated in
Figure 1. We shall now show how the data in Figure 1 can be
explained in terms of crack development in the samples. The
samples mapped in Figure 1 are artificial sandstones, made
of sand grains coated by a thin film of epoxy. Before the
epoxy hardens, the sand-epoxy mixture is placed in a casting
form and put under pressure. After the epoxy has hardened,
the sample (now a porous, epoxy-cemented sandstone) is
Figure 3. Typical development of sound velocities as func- unloaded and the sound velocities measured. The only differ-
tions of the strain parallel to an external uniaxial com- ence in the manufacturing of the various samples is the stress
pressive stress, up to the peak stress. The symbols on the level that the samples are subject to during the hardening
curves show direction of propagation and polarization period of the epoxy-the forming stress of the samples.
relative to the applied stress. Naturally, samples subject to the largest forming stress will

256 THE LEADING EDGE APRIL 1994

 of their initial values during unloading. The vertical P-wave
velocity was reduced more strongly and in advance of the
horizontal P-wave, reflecting the anisotropic stress path. Re-
loading back to the forming stress resulted in only a partial
recovery of the velocities. The observed 10-20 percent ve-
locity reduction between measurements before and after the
unloading/reloading cycle represents a permanent core dam-
age. These results suggest that core damage may be identified,
and possibly quantified, by comparing velocity measure-
Downloaded 02/20/17 to 137.195.101.233. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

ments on the cores to velocities obtained by in situ log

measurements.

S tatic and dynamic moduli. We said before that there are

theoretically direct relations between velocities and elastic
stiffness. These relations can be useful. For instance, when
predicting reservoir compaction or the stability of production
cavities as results of hydrocarbon production, the static elastic
moduli-and the rock strength-are the main design param-
eters. Although core measurements can provide these param-
eters, there is a need for more data since there are rarely
enough cores to provide continuous coverage. Well logs, on
the other hand, normally provide continuous coverage of
velocity data. Thus it is desirable to be able to estimate elastic
Figure 4. Cracks due to uniaxial compressive stress. stiffness, and perhaps also strength, from the velocity data.
But, as mentioned earlier, there is a significant difference
be most compacted and have the lowest porosity. Initially, the between the “dynamic” stiffness obtained from velocity mea-
sound velocity will also be largest for these samples. During surements and the corresponding stiffness obtained by static
the unloading process, some of the grains that were pressed measurements. For saturated materials, a part of this differ-
together by the forming stress will be ripped apart, despite the ence may be related to the fluid compressibility, which will
epoxy cement. Broken grain contacts tend to occur in clus- contribute to rock stiffness for a dynamic (undrained) mea-
ters-or “cracks” if we like. These cracks will have a strong surement but not for an undrained static measurement. How-
effect on the velocities, but only a marginal effect on the ever, the fluid contribution is generally not enough to explain
porosity, since the volumes of the cracks are relatively small. the difference-which may also be significant even for dry
Samples that were subject to the largest forming stress also materials. The difference is particularly large for very weak
will have the longest unloading paths; therefore these samples sandstones at low stress levels where we have observed
will be most affected by this cracking process. Figure 1 thus differences of nearly an order of magnitude. Also, close to the
shows that the cracks that develop due to the unloading peak stress at rock failure, the stiffness of soft rocks tends to
process have an effect on the sound velocities that is stronger vanish while the corresponding dynamic stiffness seems to
than the impact of porosity. be rather unaffected.
Empirical correlations between sound velocities and static
E xample: core damage and uplift. This type of experiment stiffness (and rock strength) were established in the 1960s.
is not just an exotic laboratory exercise-it is performed Extensive mapping of these properties resulted in widely used
because it resembles processes in petroleum-related rock correlations for sedimentary rocks like sandstones, lime-
mechanics. For instance, the situation described above resem- stones, and shale. In petroleum-related rock mechanics, such
bles, to some degree, uplift. In such areas, one should thus correlations may be developed at the scale of a specific field
expect low velocities and strong anisotropy to be prominent or basin. However, in order to take the dependency of stress
features. and stress history into account, a more fundamental theory is
The process a rock sample undergoes when it is cored is needed. For this purpose, the crack concept appears to be
another example. When the cores are to be used for measure- highly relevant. Closed cracks that slide under a static shear
ments of rock mechanical parameters like elastic moduli or load will not be triggered under the low stress amplitudes of
strength, we need to question whether what we measure in acoustic waves. Hence, the stiffness is larger for the dynamic
the laboratory is representative of in situ rock or to what than for the static measurements.
extent the real behavior is masked by coring induced damage.
During coring, the external stress field (and the pore pressure) S ummary and outlook. Sound velocities are extremely
is released in a few seconds, and with the vertical stress sensitive to cracks. Velocity anisotropy reflects crack orien-
removed before the horizontal stress. The process has been tation distributions. Cracks originate and develop due to
simulated in the laboratory for weak artificial sandstones. The changes in the stress field. This means that acoustic measur-
samples were cemented under stress, then unloaded to simu- ables contain information about current stress field and tec-
late the coring process, and then reloaded back to the forming tonic history. It also means that, in order to obtain reliable
stress in order to look for possible damage caused by the estimates of porosity, static moduli, or other physical param-
coring process. The results, reported by Holt et al. (Interna- eters, we must correct for the stress and stress history effects.
tional Journal of Rock Mechanics and Mineral Sciences, The application of acoustics as a tool to investigate rock
1993) show that the velocities were reduced to less than half mechanical behavior is currently in an active phase of devel-

APRIL 1994 THE LEADING EDGE 257

 opment. Further development requires new knowledge, new
acquisition and processing routines, and probably also new
tools. One of the keys for future success is probably to
combine the interdisciplinary efforts of three sciences that are
closely related but often work in separate worlds--geophys-
ics, petrophysics, and rock

Erling Fjaer received his diploma in

Downloaded 02/20/17 to 137.195.101.233. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

physics (2977) and his Dr Ing degree in

solid state physics (1983) from the Uni-
versity of Trondheim. For the last nine
years he has been working at IKV Pe-
troleum Research, currently as a man-
ager for the Formation Physics Section.

Rune M. Halt received his diploma in

physics (I 975) and his Dr Ing degree in
solid state physics (1980) from the Uni-
versity of Trondheim (NTH). Since then
he has been working with the SINTEF
group in Trondheim, most of the time at
IKU Petroleum Research. Since Janu-
ary 1993, he has been a professor in
petroleum technology and applied geo-
physics at NTH and also a part-time
senior scientist at IKU.

258 THE LEADING EDGE APRIL 1994