Porosity-Dependent Mechanisms of Failure Around
Boreholes in Brea Sandstone
Bezalel C. Haimson and Insun Song [Geological Engineering Program, University of
Wisconsin]
We have conducted laboratory-drilling experiments in Berea sandstone blocks of two
different porosities subjected to truly triaxial far-field stress conditions representing
depths reaching five kilometers. Depending on the simulated in situ stress conditions
applied to the blocks, drilling led to borehole failure (breakouts). We then used a
petrographic microscope to study thin sections of the drilled boreholes and inferred from
them the mechanism of failure that resulted in the formation of breakouts. The observed
behavior throws new light on the mechanical behavior of poorly consolidated sandstones and
on a possible source of sand production.
In the well-consolidated Berea sandstone (17% porosity) failure took the form of a
V-shaped breakout occurring along the s h springline, similar to
observations in granite and other crystalline rocks. This shape results from successive
failure of rock flakes formed by dilatant extensile microcracks subparallel to s H
direction. The observed microcracks preceding breakout failure are mainly intergranular
(they propagate with greater ease in the much weaker rock matrix), with intermittent
intragranular grain-splitting cracks.
An entirely different failure mechanism was observed in the more poorly consolidated
Berea sandstone (22.5% porosity). In this rock borehole failure occurred at the same
locations as above (along the s h springline), but surprisingly the final shape
of the breakout is best described as a narrow linear fracture or slot perpendicular to s H
direction. At first, this behavior appears to be counterintuitive. Our interpretation of
the mechanism forming this type of breakout is that it is related to the reduced cementing
matrix material between the grains as inferred from the higher porosity. This also implies
weak bonding between the quartzitic grains. Under sufficient far-field stresses, failure
that first occurs in the much weaker rock matrix at the points of high stress
concentration around the hole, partially or totally debonds the quartz grains. The
debonded and exposed grains, assisted perhaps by the circulating drilling fluid, spall off
in non-dilatant fashion, creating a higher stress concentration behind them. This in turn
facilitates further disintegration of the cementing matrix, more non-dilatant grain
debonding, and additional grains being removed along the s h spring line. The
rugged free surface of the fracture-shaped breakout shows whole quartz grains left intact
during this failure process, supporting the assertion that failure is limited to that of
the matrix material. The final shape of breakouts in this sandstone and the inferred
fracturing mechanism are very similar to the non-dilatational behavior observed in
poorly-sintered glass bead bricks by Bessinger et al. (1997). This newly observed behavior
could be a source of sand production, and may have major significance in studying
instability problems in oil wells intersecting poorly consolidated sandstones.
Micromechnaics of Poorly Consolidated Media
Larry Myer, Kurt Nihei [Earth Sciences Division, Lawrence Berkeley National Laboratory]
There are significant differences between the macroscopic mechanical properties
(including static and dynamic deformation and strength) of well consolidated, indurated
clastic rock and cohesionless granular material, and these differences arise from
different micromechanical processes. Poorly consolidated rocks have intermediate
properties. This work is focused on understanding the transition in micromechanical
processes between cohensionless and well consolidated rock. The approach is to perform a
variety of mechanical measurements on poorly consolidated samples which vary, first of
all, in the amount of cohesion between grains. Qualitative and quantitative relationships
are sought between micromechanical parameters and macroscopic properties. Correlations are
also sought between macroscopic properties, for example, between seismic wave velocities
and attenuation and other properties such as fracture toughness or compressive strength.
Previous work has shown that under differential compressive stress, extensile crack
growth accompanied by frictional dissipation fundamentally affects the macroscopic
stress-strain strength properties of well consolidated granular rock. The extensile cracks
form at the grain scale and are oriented parallel to the maximum compressive stress
direction. Around a wellbore, where the major compressive stress is a hoop stress,
extensile cracks form parallel to the surface of the wellbore, leading to typically
observed wellbore failure features.
In cohesionless material, grain sliding, rotation and rearrangement are dominate
micromechanical processes under a large range of differential stress. In poorly
consolidated rock it might be supposed that there is a transition between behavior
dominated by grain sliding and rearrangements, and extensile crack growth. Experimental
results indicate that the character of borehole breakouts may be significantly different
if grains kinematically move instead of fracturing.
Seismic wave propagation has been studied in saturated granular media under changing
compressive stress, where the compressive stress can be considered as providing an
effective cohesion between grains. As uniaxial stress was increased a monotonic increase
in P-wave velocity was observed. However, quite different trends were observed in the
amplitude and frequency content of the pulses. At low stress the behavior was best
explained by a model characterizing the sample as a suspension of solid particles in a
liquid. As stress is increased there is a transition to a condition in which energy is
primarily propagating through the granular skeleton with the stiffness of the grain
contacts having a first order effect on the amplitude and frequency content, as well as
the velocity of the P-wave.
Finally, to provide some preliminary observations of fracture in poorly consolidated
media, three point bending tests were performed on sintered glass bead samples. Values of
Young's modulus, E, were obtained from acoustic velocity measurements. Both Klc and E for
the lightly sintered glass bead material were very low compared to rock data. The data
support intuition that values of Klc for materials will go to zero as E approaches zero.
Visual observation during the tests suggest that the process of failure in these very
weak samples was similar to that in stronger rock. Results imply that the principles of
fracture mechanics can be applied to crack propagation in granular media as long as there
is non zero bond strength between grains.
3D Pore Scale Geometry of Sandstones, Basalts and Fractures
Brent Lindquist [State Univ. of New York at Stony Brook, Dept. of Applied Mathematics
and Statistics]
We have developed a computational package, 3DMA, to use as a tool to provide analysis
of the geometry of the pore and grain phases of digitized three dimensional images.
Current analyses for each phase include determination of volume fraction, specific surface
area, 2-point spatial covariance, disconnected volume distribution, the medial axis (1D
skeleton), path length and tortuosity distribution, coordination number, throat size and
nodal pore volume distributions. We have applied this analysis to a variety of porous
media images including rocks (sandstones, basalts, carbonates), glass micro-bead packs,
cellulose fiber networks, biological images (neurons) and two-fluid mixtures.
Medial axis analysis is a central component of the package. It provides a one
dimensional skeleton enabling efficient searching and geometrical characterization as well
as a graph for the application of graph theoretic network tools.
I will present the results of our analyses for Fontainebleau sandstone, vesiculated
basalt, and fractures in Harcourt granite and Carrara marble.
Micromechanical Studies of Fracture
Dr. Anthony J. C. Ladd [Department of Chemical Engineering, University of Florida]
Preliminary work aimed at developing a micromechanical model for the deformation and
failure of porous rocks will be described. A detailed description of the microstructure
and mineral distribution in specimens of Berea sandstone has been obtained by
high-resolution x-ray tomography. An efficient finite-element model has been developed
that uses these x-ray tomographs for structural input; stress and strain fields in samples
of more than 10 million elements can be determined using this model. Fracture can be
modeled by the progressive failure of elements under high tensile loads; local tensile
stresses develop in microstructural models, even if the macroscopic loads are compressive.
An example of predicted and measured fracture patterns will be given.
Strain Localization in Compacting Rock
William A. Olsson [Sandia National Laboratories]
Mollema and Antonellini (1996) recognized a new type of geologic structure, which they
called compaction bands, in porous sandstones from the Navajo formation. These bands were
tabular zones of crushed material that have undergone no shear, just pure compressional
deformation and a consequent reduction in porosity. Haimson and Song (1997) found features
that would seem to be identical to compaction bands in a laboratory study of borehole
breakouts. Lajtai (1974) also reported what appear to be compaction bands in association
with shear cracks in laboratory models. Because of the potential deleterious effects on
the permeability of porous reservoir rocks, I used theory and experiment to investigate
the possibility of producing compaction bands in Castlegate sandstone that had about 25%
porosity.
I applied the strain localization theory of Rudnicki and Rice (1975) to the problem of
inhomogeneous compaction and found that the theory predicts that bands of pure compaction
can occur for certain combinations of Poisson's ratio and dilatancy factor. Some features
that may indeed be compaction bands were formed experimentally. The agreement between
theory and experiment, at first sight, seems somewhat inconclusive because compaction
bands and thick shear bands sometimes occur in the same specimen. But when subtle
relations between constitutive properties of Castlegate sandstone and the predictions of
the strain localization theory are considered, the interrelation between theory, field,
and experiment can be explained.
Haimson, B. C., and I. Song, Laboratory study of borehole breakouts in two Berea
sandstones reveals dramatically different failure mechanisms, (abstract), EOS Trans. AGU,
78, F710, 1997.
Lajtai, E. Z., Brittle fracture in compression, Int. J. Fracture 10, 525--536, 1974.
Mollema, P. N., and M. A. Antonellini, Compaction bands: a structural analog for
anti-mode I cracks in aeolian sandstone, Tectonophysics, 267, 209--228, 1996.
Rudnicki, J. W., and J. R. Rice, Conditions for the localization of deformation in
pressure-sensitive dilatant materials, J. Mech. Phys. Solids 23, 371--394, 1975.
Localization in Dilating Rock
David J. Holcomb [Sandia National Laboratories]
An experimental and modeling approach has been pursued in testing the applicability of
a bifurcation theory of shear-localization developed by Rudnicki and Rice (RR) to the
process of shear deformation in dilating rocks. The modeling approach uses the results of
triaxial testing to determine the constitutive parameters required by the RR formulation.
A model was defined with sufficient flexibility to capture the major features of the
triaxial tests, hardening, softening, pressure-dependence and shear-strain-driven
dilatancy. Using a simplex fitting technique, the model parameters were determined that
best fit all of the data. This approach avoids over-emphasizing any one data set, as in
common in the use of uniaxial test results to determine fitting parameters. Fitting the
stress-strain data and then differentiating the fitted analytical expressions, eliminated
much of the numerical noise that differentiating the experimental results would have
introduced. Having derived expressions for the plastic tangent modulus, slope of the yield
surface and dilatancy parameter as a function of mean stress and plastic shear strain, it
was then possible to calculate the results for any loading path for comparison with
experiments. In addition to stress and strain, the criterion for localization could be
calculated for a general stress path. Results for Tennessee marble under plane strain were
in good agreement with the calculated results.
Determining that localization has occurred is in general difficult using measures of
strain that average over the entire sample. Using acoustic emissions it is possible to
observe the localization process as the location of acoustic emissions evolves from a more
or less random cloud of events to a well-defined planar feature. Results for Gosford
sandstone under triaxial and plane strain show that localization occurs in the softening
regime (post-peak) for triaxial test conditions and in the hardening regime (pre-peak) for
plane strain, confirming a key prediction of the RR localization theory.
Frequency-Dependent
Acoustic Anisotropy from Fractures - Modeling
Kurt T. Nihei, Seiji Nakagawa, Michael Schoenberg, and Larry R.
Myer[Lawrence Berkeley National Laboratory]
Visiting scientist at
Lawrence Berkeley National Laboratory; on leave from Schlumberger-Doll Research,
Ridgefield, CT
Fractures in rock are localized planes of compliance. When fractures
are aligned in near parallel sets, as often occurs in nature, this ordered compliance can
give rise to anisotropy in the elastic properties. The conventional approach for
determining the elastic anisotropy of such a medium is based on a static (i.e., zero
frequency) approximation. The static description predicts directionally-dependent wave
velocities and polarizations of the quasi-compressional and the two quasi-shear waves. The
static approach, however, does not predict frequency-dependent wave phenomena such as
attenuation, dispersion, and fracture guided waves that have been observed in laboratory
and field measurements in fractured rock. One of the objectives of this research is to
quantify the effects of wave frequency, fracture stiffness, and fracture spacing on the
velocities and amplitudes of waves propagating in fractured rock.
To investigate the frequency-dependent characteristics of seismic waves
arising from a set of multiple, parallel fractures, we have formulated an analytic
expression for elastic wave propagation in a medium composed of an infinite series of
parallel fractures. The solution uses Floquet theory to impose periodic boundary
conditions on the stress and displacement fields. Fractures are modeled as
displacement-discontinuity boundaries. The equation was solved numerically to obtain
slowness and wave surfaces for a range of frequencies and fracture stiffnesses. Analysis
of this fully dynamic effective medium theory for fractured rock reveals that deviations
from the static effective medium velocity can occur, even when the wavelength is much
larger than the fracture spacing, provided that the fracture stiffnesses are low.
The Floquet analysis also predicts pass-stop band transmission and
reflection for waves propagating at oblique incidence to the fractures. This behavior was
examined using a one-dimensional propagator code and a broadband source. When the entire
waveform including the coda (i.e., the multiply reflected waves) is used to compute the
spectrum, half-wavelength, inter-fracture resonances produce a transmission coefficient
with alternating regions of complete transmission and of complete reflection. However,
when the spectrum is computed from just the first coherent pulse, the transmission
coefficient shows a strong low-pass filtering that results from reflection losses. These
results suggest that the presence of fractures can be inferred from the low-pass filtering
of the transmitted first coherent arrival and, possibly, from the full waveform
transmission coefficient, provided that the source generates wavelengths that are on the
order of the fracture spacing.
Frequency Dependent Acoustic Anisotropy from Fractures:
Experimental
Laura J. Pyrak-Nolte [Department of Physics, Department of Earth and Atmospheric
Sciences, Purdue University]
Many
reservoirs include multiple, near-planar discontinuities such as joints and fractures that
occur on multiple length scales. The presence of a set or sets of fractures produces an
anisotropy of material properties. Traditionally, the wave propagation through a fractured
rock has been modeled using effective medium theories that use a static approximation to
develop analytic expressions for the elastic moduli of a population of microcracks or
planar fractures. The effective medium approach assumes that a fracture will reduce the
modulus of the rock, which in turn reduces the seismic velocity. The effect of the
fracture is distributed throughout the bulk and the discreteness of the fracture is lost.
Hence, a seismic reduction in velocity is observed, but the location of the cause of the
reduction is lost. An alternative method for analyzing the seismic reponse of a fracture
that retains the discreteness of the effect of the fracture is the displacement
discontinuity or non-welded contact theory. Several investigators have modeled a fracture
as a non-welded contact which is assumed to have negligible thickness compared to the
seismic wavelength. The non-welded contact is represented as an interface across which
stresses are continuous but displacements are not. This pureley elastic theory gives rise
to frequency dependent transmission and reflection coefficients as well as frequency
dependent velocities. Field and laboratory measurements have shown that a fracture behaves
as a low-pass filter that attenuates the wave by removing the high-frequency components of
the signal and produces a frequency-dependent time delay. The displacement discontinuity
boundary conditions have also been used to derive plane wave transmission and reflection
coefficients, group time delays, interface waves, and guided love waves.
In this study, experimentally
measured compressional wave anisotropy caused by the presence of a single fracture and
multiple parallel fractures is compared to theoretical predictions based on the
displacement discontinuity theory. The comparison between the theoretically predicted
transmission coefficients and the experimentally determined values enabled us to
investigate the robustness of the displacement discontinuity theory for obliquely incident
compressional waves on a single fracture and on multiple parallel fractures. The
analytical solution for plane-wave propagation across a displacement discontinuity
predicts: (1) that at high frequencies the group time delay is not as sensitive to changes
in stiffness as transmitted amplitudes; (2) that the transmission coefficient decreases in
magnitude with increasing frequency; (3) that the transmission coefficient decreases for
glancing angles of incidence to the fracture; (4) that a discontinuity in the value of the
transmission coefficient should occur at the orientation when the trajectory of a wave
crosses additional fractures; and (5) that single values of normal and tangential fracture
specific stiffness can be used to fit all angles of incidence for all frequencies. These
predictions are examined in light of the experimental data.
While the displacement
discontinuity theory is able to fit the measured compressional wave transmission
coefficients well for most angles of incidence, discrepancies between theory and
experiment still exist. The deviations for glancing angles of incidence and for angles of
incidence near the transition from multiple fractures to a single fracture are not
completely understood. The plane-wave solution used in this analysis does not account for
the partitioning of energy into interface waves, i.e., waves that propagate along the
fracture and which depend on fracture specific stiffness. A full numerical method needs to
be applied to compare oblique incidence wave propagation across a fracture to accurately
assess energy partitioning by a fracture or sets of fracture. In addition, further effort
is needed to explore the frequency dependent fracture stiffness. If the frequency
dependence gives an indication of the distribution of asperities or contact regions in a
fracture, this information could be used to predict the hydraulic properties of fractures.
Materials Modeling - The Challenge to Explore the Relation
Between Microstructures and Mechanical Properties in Crystalline Solids.
Richard G. Hoagland [Washington State University, Pullman, WA]
This talk touches on methodologies devised to help understand/predict the mechanical
behavior of solids from the viewpoint of a metallurgist / materials scientist. The various
approaches in this area that presently exist can be categorized into essentially five
groups, each group distinguished from the others by a characteristic length scale range,
namely: electronic structure (to about 0.15 nm), atomistic simulations (to about 15 nm),
micro-scale modeling (to 15,000 nm), mesoscale modeling (to 5 x 10^6 nm), and continuum
plasticity (to ?). The focus of each of these groups will be described briefly and it will
be noted that, generally, within each category there exists sophisticated, highly
predictive capabilities. The principal challenge remains, however, to link these
categories so that an intelligent and intelligible transfer of information is able to take
place among the groups. Currently, in the bulk of the metals fabrication industry, there
is no need for this linkage. For example, empirical constitutive relations developed for
structural steels together with 3D finite element models are usually adequate for
predicting (with satisfactory accuracy) the outcomes of most complex forming operations.
Similar applications involving more complex materials (e.g., complicated by anisotropy)
are also commonplace. However, new materials continue to be developed and recently, some
of these materials have strength levels that are so high that new forms of plasticity and
failure mechanisms have appeared. In addition, exotic applications, e.g. involving very
small sizes, operation at very high temperature or other extreme environments, are
creating a new set of demands on manufacturing processes. In such circumstances, the
development of an understanding of behavior and predictive capability is best served
(economically) by linking activity over the entire hierarchy of length scales. We will
give some examples of results of the atomistic and micro-scale modeling and briefly
describe the kind of information that needs to move up the length scale chain from
smallest to largest.