SPE/ISRM 78157
Continuous Scratch Testing on Core Allows Effective Calibration of Log-Derived
Mechanical Properties for Use in Sanding Prediction Evaluation
Roberto Suárez-Rivera, SPE, TerraTek Inc., Jørn Stenebråten, TerraTek Inc.,
Fabrice Dagrain, Faculté Polytechnique de Mons
Copyright 2002, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the SPE/ISRM Rock Mechanics Conference held
in Irving, Texas, 20-23 October 2002.
This paper was selected for presentation by an SPE/ISRM Program Committee following
review of information contained in an abstract submitted by the author(s). Contents of the
paper, as presented, have not been reviewed by the Society of Petroleum Engineers or
International Society of Rock Mechanics and are subject to correction by the author(s). The
material, as presented, does not necessarily reflect any position of the Society of Petroleum
Engineers, International Society of Rock Mechanics, its officers, or members. Papers
presented at SPE/ISRM meetings are subject to publication review by Editorial Committees of
the Society of 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 prohibited. Permission to reproduce in print is restricted to an abstract of not more
than 300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
Abstract
This paper discusses the use of continuous scratch testing for
evaluation of rock heterogeneity and effective correction and
calibration of mechanical properties predictions using wire
line logging. Understanding the vertical variability of rock
properties (strength and moduli) along the reservoir thickness
is of primordial importance for sanding prediction analysis.
Logging measurements (GR, Sonic, Density and Porosity) are
traditionally used to identify troublesome intervals, to select
sections for laboratory testing and for extrapolating laboratory
data to untested reservoir sections. However, depending on
the scale of the reservoir heterogeneity, the log measurements,
by providing integrated values over their window of vertical
resolution, can considerably underestimate the strong reservoir
sections and overestimate the weak reservoir sections. This
leads to misrepresentation of critical weak layers with high
potential for sanding and potentially to large errors in the
completion strategy with costly long-term consequences.
Corrections are thus necessary and are traditionally based on
simple unconfined compression (UCS) tests conducted
sparsely or at high density, varying from operator to operator.
This paper presents a method for log-correction based on
continuous core scratching in the laboratory and shows that
the methodology is effective in evaluating changes in
properties with high accuracy and a resolution of at least 1 cm.
Introduction
Rock mechanical properties are fundamental to analysis of
wellbore stability, sanding potential, mud weight
requirements, casing collapse, compaction potential and many
other problems related with the short- and long term
mechanical stability of the well and the reservoir. These
predictions, when correct, may help saving millions of dollars
in drilling and completions costs, and may allow preventing
long term and costly consequences. A particular problem in
petroleum geosciences is that rock mechanics data is hard to
come by and limited to small sections of the reservoir. The
lack of data and, when available, the lack of significant
volume representation of these data, has frustrated many
investigators and has led them to pursue alternative avenues
(e.g., correlations, statistical estimates) to obtain rock
mechanical properties. In petroleum-related rock mechanics,
computational capabilities for numerical simulations of
wellbore stability, sanding potential, reservoir compaction,
casing failure during production and other pressing
mechanical problems have evolved into sophisticated tools
whose demands for field data (e.g., in-situ stress, rock
properties, coupled hydraulic-thermal-mechanical properties,
and the lateral and vertical variability of these properties
across a reservoir) far exceeds the industry’s capacity to
provide it. Thus engineers struggle to develop alternative
sources of data via correlations, inferences from seismic
prospecting, inferences from log measurements, or by
developing knowledge-based criteria to guess reasonable
values. However, lacking good quality data, rock mechanics
predictions are weak and most importantly, the industry
confidence for implementing them to field applications is low.
Thus rock mechanics analysis, limited by lack of data, is a
service that does not fulfill its potential for minimizing well
construction costs or preventing, long term, production-related
problems.
In rock mechanics applications, logging measurements (GR,
Sonic, Density and Porosity) are traditionally used to identify
troublesome intervals, select sections for laboratory testing
and for extrapolating laboratory data to untested reservoir
sections. However, depending on the scale of the reservoir
heterogeneity, the log measurements, by providing integrated
values over their window of vertical resolution, may
considerably underestimate the strong reservoir sections and
overestimate the weak reservoir sections. Consequently, the
potential troublesome sections of the reservoir may be
underestimated or misrepresented altogether. This problem
2 R. SUÁREZ-RIVERA, J. STENEBRÅTEN, AND F. DAGRAIN
may be corrected by increasing the vertical resolution of the
logs and by obtaining data from direct measurements on cores
(e.g., UCS tests) for calibrating the logs. Traditionally, log
calibrations are obtained by shifting the log by a constant
value to match laboratory data. However, when sufficiently
large number of direct measurements is available, log shifting
alone is not sufficient for honoring the experimental data.
This may be done using optimization routines or neural
network algorithms. For example Morita[1] suggests
conducting foot by foot rock strength measurements via point
load, brinell hardness or UCS tests. In this paper, we present
methodology for calibration of logs based on continuous
scratch test measurements and show that the methodology is
effective in evaluating changes in properties with high
accuracy and a resolution of at least 1 cm.
Equipment and Testing Procedures
Experiments were conducted using the TerraTek TSI™
scratch test system (Figure 1). The system consists of a
moving cart with a sample holder and a loading fixture
capable of “scratching” the rock sample. A load cell is
mounted on the loading frame and measures the horizontal
force (in the cutting direction) and the vertical force (normal
to the cutting surface) in the range from 10 N to 4000 N, and
with an accuracy of 1 N. Computer control allows variable
“scratching” velocity, automatic data acquisition, and real-
time data analysis. The cutting velocity can be set in a range
from approximately 0.1 mm/second to 5 mm/second and the
depths of cut in a range from 0.1 mm to 1 mm. The cutter was
a replaceable polycrystalline diamond compact (PDC) cutter
of 10 mm width, with a sharp geometry. Measurements were
conducted on slabbed core sections with a pre-slab diameter of
4 inches and a length of 3 feet. The rock surface was
scratched at a constant depth of cut of 0.5 mm. Normal and
tangential forces on the cutter were measured; and automatic
data processing provided estimates of rock strength along the
cut.
Eighty-five feet of primarily reservoir sandstone from two
depth ranges (approximately 360 ft apart) were scratched for
mechanical property evaluation. The core was cut in 3 ft long
sections and slabbed prior to scratch testing. Because of
inherent irregularities on the slabbed surface of the cores,
multiple scratch passes were initially conducted to obtain flat
scratching surfaces. Subsequent to these preliminary
scratches, two scratch measurements were conducted and
averaged for property evaluation. Because of limitations on
the travel length of the system and requirements for alignment
and parallelism between the sample surface and the scratch
tool, continous scratches were restricted to short segments of
approximately 10 cm to 20 cm length.
Visual inspection of the core prior to scratching indicated
strong vertical heterogeneity in a scale ranging from few
inches to few feet. The core heterogeneity was characterized
SPE/ISRM 78157
by abrupt changes in rock type, bed dip, rock quality, and by
the presence of fractured zones. Scratch measurements
showed strong correlation with these visually observed core
features. This information was important for calculating the
percentage of mechanically weak, thin beads, across the core,
and thus for mechanical modeling and better selection of
alternative completions systems for sanding control or
sand exclusion.
Theory and Definitions
Figure 2 shows the forces acting on the cutter. These have
been described and studied by Detournay et. al and others[2-8].
In summary, the principal forces for scratch testing analysis
are the cutting force (Fc) and the frictional force (Ff).
Detournay at al.,[2,3] further decomposed these forces in their
fundamental components for analysis of the physics of the
scratching process. In the present study, the energy needed to
scratch a unit volume of rock is defined as the standard
intrinsic specific energy ε10, and is calculated from the ratio of
the cutting force and the cross sectional area of the cut
(A=wd). This definition is valid for tests performed with a 10
mm width sharp cutter[8]. Equation 1[8] shows
this relationship.
ε10 = Fh , (1)
wd
where Fh is the mean of the horizontal force over the
tested zone.
The standard intrinsic specific energy is assumed to be a
characteristic of the rock and thus represents a rock property
that is directly correlated to its unconfined compression
strength (UCS). Figure 3 shows an example of plots generated
during scratch testing. In this, the intrinsic specific energy
(vertical axis) is plotted versus the scratch distance (horizontal
axis). The standard intrinsic specific energy is used directly as
an indicator of the rock heterogeneity by plotting it as a
function of scratch length or by correlating it with the rock
unconfined strength. Figure 4 shows the correlation between
the measured specific energy (horizontal axis) and unconfined
compressive strengths (vertical axis) [8]. Within the range of
UCS values from approximately 10 to 120 MPa (1450 to
17,400 psi), the intrinsic energy versus unconfined strength is
reasonably well defined and linear.
Multiple (up to ten) to single scratch measurements are
possible for data evaluation. Because of its simplicity and fast
turn around, and average of two scratch measurements were
used in this project. The selection of the depth of cut is of
great importance when using a limited number of scratches, to
avoid tip effects (if the cut is too shallow) and to avoid
chipping (if the cut is too deep)[5,6]. Measurements with
limited number of scratches are known to underestimate or
 CONTINUOUS SCRATCH TESTING ON CORE ALLOWS EFFECTIVE CALIBRATION OF LOG-DERIVED
SPE/ISRM 78157 MECHANICAL PROPERTIES FOR USE IN SANDING PREDICTION EVALUATION
overestimate the specific energy depending on whether the
scratch tool tip force is non negligible (for shallow cuts) or
when the lateral forces (side effects) are non negligible[8]. By
carefully selecting the size of the cutter and the appropriate
scratch depth these unwanted effects were minimized.
Subsequent to scratch testing, vertically oriented samples,
representative of weak, intermediate and strong core sections,
were obtained for laboratory unconfined compression (UCS)
testing. Unconfined compression testing was conducted on
samples of 1-inch diameter and 2-inches length and with flat
and parallel surfaces, to within a tolerance of ±0.001 inch, for
comparison with scratch test predictions. Prior to testing, the
individual test samples were CT-scanned for quality control,
consisting on two cross-sectional images and one axial scan.
UCS tests were conducted on vertical samples and with their
native condition of saturation.
Results and Discussion
Photographs, high-resolution FMI images, log-based
mechanical property predictions, laboratory UCS
measurements and continuous scratch tests of two cored
sandstone reservoir intervals of 55ft and 30ft length were
obtained, and compared. Laboratory UCS measurements were
limited to 12 samples and were acquired from locations
representing strong, intermediate and weak reservoir rock.
Unfortunately, measured UCS values were not always
acquired on locations where the scratch test was conducted.
This was so because approximately half of the UCS samples
were obtained prior to scratch testing. In addition, some
sections of the core could not be scratched because of lack of
material (e.g., horizontal plugs for routine and specialized core
analysis had been taken at regular intervals, prior to scratch
testing) or because the material was too weak or too damaged
for scratch testing.
Figures 5 and 6 compare laboratory UCS measurements
(triangles) with scratch-derived and log-derived UCS values.
The latter were obtained using log-interpreted matrix volumes,
porosity permeability and raw compressional and shear
velocities. Figure 5 shows measurements on the upper core
section. Figure 6 shows measurements on the lower core
section (approximately 360 ft deeper). Data is presented as a
function of core distance from the top of the cored (TOC)
section. Laboratory based UCS values ranged from 118 psi to
10,300 psi. Core sections weaker than 118 psi were present
but could not be sampled (i.e., were either severely fractured
or damaged). Scratch-derived UCS values ranged from 145
psi to 23,780 psi and matched the laboratory UCS values well.
Log-derived UCS values ranged from 2320 psi to 12,905 psi
and consistently overestimated the strength of the weakest
rock sections and underestimated strength of the strongest
rock sections.
3
The scale of the core vertical heterogeneity was on average
approximately ½ ft. The vertical resolution of the log data
was unknown but was evaluated to be approximately 3 to 4 ft,
by calculating moving averages of the scratch test data to
match the log data (Figure 7). Because of the low vertical
resolution of the log in relation to the scale of the rock
heterogeneity, the log-derived UCS values represent moving-
averaged values and do not resolve the rapid transitions in
rock strength from high to low values. This results in
predictions that underestimate the strength of the stronger
sections of the core and overestimate the strength of the
weakest sections of the core. For evaluations of wellbore
stability and sanding potential, the underestimation of the
stronger core sections is of limited importance; however, the
overestimation of the weaker core sections has important
implications because it misrepresents the most critical and
most problematic rock along the length of the well.
Figures 8 through 11 show comparisons of scratch derived
UCS values, log-derived UCS values and laboratory UCS
measurements for selected 3 ft sections of core. Viewed in
this smaller scale, the data shows good agreement between the
laboratory measurements and the scratch predictions but large
discrepancies between the laboratory measurements and the
log predictions. The log predictions consistently overestimate
the weak intervals.
Log calibrations are often conducted by applying a constant
shift to the log data to match laboratory UCS values. This
method was not adequate because shifts of opposite directions
were required to match the low strength and the high strength
rock. Instead, log calibration was obtained readily by
comparison with continuous scratch measurements using an
optimization algorithm that minimized the error between the
scratch-derived and log-derived UCS values. For convenience,
and to avoid predicting negative UCS vales, a cut off value of
100 psi was selected. Once the correction function was
obtained based on the upper core section, the same parameters
were used to match the lower section of the core. Figures 12
and 13 show the calibrated log measurements. The corrected
log data show good agreement with the scratch test data and
the measured UCS. Comparing the corrected log data with a
record of wellbore breakouts (Figure 14), it is evident that the
weaker zones are identified reasonably well. The weak zones
are zones with highest probability for wellbore stability
problems (during drilling) and sanding (during production).
The corrected log allows us to evaluate their relative
pervasiveness along the length of the reservoir and thus
estimate the relative magnitude of the potential sanding
problem during production. Because the calibrated log
compares well with laboratory data, the degree of confidence
in rock mechanics predictions is high. This implies that costly
decisions, with long-term consequences to the reservoir, can
be taken with higher degree of confidence.
4 R. SUÁREZ-RIVERA, J. STENEBRÅTEN, AND F. DAGRAIN
In addition to identifying the scale and magnitude of
heterogeneity in the core and being important for correcting
the log measurements, the potential of continuous scratch
measurements to identify the adequate number of core
sections for rock mechanics testing is invaluable. At present,
decisions of the number of intervals to be tested are based on
some form of log data (e.g., GR, sonic log), changes in
physical properties (ambient porosity and density), measured
at discrete distances along the core, or geological analysis.
Continuous scratch measurements provide a direct
measurement of mechanical continuity of the various
lithological units in a cored section and thus allow positive
selection of the units with largest contrasting mechanical
properties for testing.
Continuous scratch measurements also offer the potential for
elastic property calibration[7] and for rapid evaluation of the
effect of drilling fluids on overburden shales[9]. The effect of
rock-fluid interactions can be easily detected by measuring
their effect on the scratching resistance of a rock surface.
Surface measurements appear to provide higher resolution to
rock-fluid interactions than bulk measurements (e.g., wave
propagation, UCS). Work currently underway[9], suggests that
fundamental effects of shale-fluid interaction (osmosis, ion-
diffusion, the effect of shale dehydration) can be monitored
easily and continuously via scratch testing.
Conclusions
Continuous scratch measurements were conducted on
approximately 85 ft of reservoir sandstone for evaluation of
core heterogeneity and for comparison with mechanical
properties predicted from logs. Results of scratch testing can
be used directly, as an index of the rock resistance to
scratching, or via correlations with unconfined compression
tests, as an indicator of the variability of rock strength along
the length of the core. Rock scratch measurements are limited
to rocks with unconfined strengths larger than 100 psi.
Furthermore, the core should be well preserved, undamaged
and, when of low strength, provided with a supporting sleeve
to prevent damage during handling. Scratch tests are not
possible on rocks requiring special conditions of handling and
preservation (e.g., poorly consolidated sandstones requiring
handling under frozen conditions).
Continuous scratch measurements provide a quantitative
evaluation of the vertical scale and magnitude of the rock
heterogeneity. These results are helpful for (i) identifying the
vertical heterogeneity of the core, (ii) for selecting a
representative number of core sections for laboratory rock
mechanics analysis (iii) for specifying the minimum vertical
resolution for display and evaluation of the logs and (iv) if
logs with the desired vertical resolution are not available, to
anticipate misrepresentations by the logs due to the averaging
process.
SPE/ISRM 78157
Continuous scratch measurements may be conducted on-site
or in the laboratory with minimum requirements for sample
preparation. Results are available immediately (real-time) and
can be used readily for designing the laboratory-testing
program, for calibration of logs and for preliminary
evaluations wellbore stability and sanding potential.
Continuous scratch measurements, after calibration and
conversion to unconfined compression strength values, in
combination with a comprehensive rock mechanics
characterization of an appropriate number of core sections,
may offer the best alternative for in depth mechanical
characterization of the entire core interval. This implies that
laboratory analysis may be extended from specific core
depths, limited to few inches, to the entire length of the core
(tens to hundred feet), thus expanding by several fold the
current volumetric representation of laboratory mechanical
data. This expansion would also apply to rock mechanics
analysis and evaluations of mechanical stability (e.g., sanding
potential).
Continuous scratch measurements, in combination to well
calibrated logs, allows the extension of high quality core data
to the entire logged interval of the well. This extension,
currently conducted via correlations of logs with a limited
number of UCS data points, will considerably enhance the
quality of log-inferred mechanical data and consequently
increase the degree of confidence on rock mechanics
predictions.
In an industry starving for field data, any equipment, system
or methodology that expands the industry’s current capacity
for obtaining data should be welcomed. Frustration by data
starvation has lead many in the industry to promote artificial
data and to move away from laboratory data altogether. There
are direct benefits to be gained by increasing the industry’s
capacity to collect data. For mechanical analysis, the scratch
test measurement provides an opportunity for increasing the
volumetric representation of laboratory data.
Acknowledgements
The authors thank BHP Billiton Petroleum for allowing us
publishing these data. Special gratitude is extended to Bill
Begnaud and KaiSoon Tan for their close and willing
cooperation with the authors in undertaking this investigation.
References
1. N. Morita. Field and Laboratory Verification of Sand-
Production Prediction Models. SPE Drilling and
Completion, December 1994.
2. E. Detournay and P. Defourny. A phenomenological
model for the drilling action of drag bits. Int. J. Rock
Mech. Min. Sci. and Geomech. Abstr, 29(1) :13-23,1992
3. E. Detournay, A. Drescher, P. Defourny, and D.
Fourmaintraux. Assessment of rock strength properties
from cutting tests: Preliminary experimental evidence. In
Chalk and Shales Colloquium, Brussels, 1995.
 CONTINUOUS SCRATCH TESTING ON CORE ALLOWS EFFECTIVE CALIBRATION OF LOG-DERIVED
SPE/ISRM 78157 MECHANICAL PROPERTIES FOR USE IN SANDING PREDICTION EVALUATION 5

4. J.I. Adachi. Determination of rock strength parameters

from cutting tests. Paper submitted to the 2nd North
American Rock Mechanics Symposium (NARMS 96), ψ
α
Montreal, 1996. θ
5. T. Richard, E. Detournay, A. Drescher, P. Nicodème, and θ
D. Fourmaintraux. The scratch test as a means to measure
strength of sedimentary rocks. Paper submitted to the
SPE/ISRM Eurock’98, Trondheim, 1998. (SPE 47196). φ
6. T. Lhomme Frictional contact at a rock-tool interface: an
experimental study. Master’s thesis. University of
Minnesota, 1999.
7. G. Schei, E. Fjaer, E. Detournay, C.J. Kenter, G.F. Fuh, F. Figure 2: Forces acting on a cutter (after Dagrain[8]).
Zausa. The scratch test: an attractive technique for
determining strength and elastic properties of
sedimentary rocks. Paper submitted to the 2000 SPE
Annual Technical Conference and exhibition, Dallas,
2000. (SPE 63255).
8. F. Dagrain. Influence of the cutter geometry in rock
cutting with PDC cutter: an experimental approach.
Master’s thesis. University of Minnesota, 2001.
9. J. Stenebraten, R Suarez-Rivera and F. Dagrain.
Evaluation of Shale Drilling Fluid Compatibility via
Scratch Testing. OilRocks, 2002. Paper No. 78161.

SI Metric Conversion Factor

feet × 3.048* E-01 =m
in. × 2.54* E+00 = cm
lbf × 4.448 222 E+00 =N
psi × 6.894 757 E+00 = kPa Figure 3: Intrinsic specific energy [vertical axis] versus sample
distance from the beginning of the scratch [horizontal axis].
*Conversion factor is exact

160
UCS = 1.0038 ⋅ ε10
R 2 = 0.864
140

120

100
Rc (MPa)

80

60

40

20

0
0 20 40 60 80 100 120 140 160
ε 10 (MPa)

Figure 4: Correlation between specific energy and unconfined rock

strength (After Dagrain, 2001).

Figure 1: The Scratch Index System.

6 R. SUÁREZ-RIVERA, J. STENEBRÅTEN, AND F. DAGRAIN
Figure 5: Unconfined compression versus distance from the top of
the core (TOC) for the upper core section. Values are
provided in MPa (left) and psi (right). Scratch test
measurements (solid lines) and Log predictions (circles) are
compared to laboratory measurements (triangles).
Figure 6: Unconfined compression versus distance from the top of
the core (TOC) for the lower core section (approximately
360 ft deeprt than the uper core section). Values are
provided in MPa (left) and psi (right). Scratch test
measurements (solid lines) and Log predictions (circles) are
compared to laboratory measurements (triangles).
SPE/ISRM 78157
Figure 7: Unconfined compression versus distance from the top of
the core (TOC) for the upper core section. The vertical
resolution of the mechanical log was evaluated to be 3 to 4
ft, by comparison with moving averages of the scratch test
data.
Figure 8: Unconfined compression versus distance from the top of
the core (TOC) for the upper core section. .
Figure 9: Unconfined compression versus distance from the top of
the core (TOC) for the upper core section.
 CONTINUOUS SCRATCH TESTING ON CORE ALLOWS EFFECTIVE CALIBRATION OF LOG-DERIVED
SPE/ISRM 78157 MECHANICAL PROPERTIES FOR USE IN SANDING PREDICTION EVALUATION 7

Figure 10: Unconfined compression versus versus distance from

the top of the core (TOC) for the upper core section.
Figure 13: Unconfined compression versus versus distance from the
top of the core (TOC) for the lower core section. The log
measurements were corrected to represent more truthfully
the UCS values evaluated through laboratory testing and
scratch testing measurements.

Figure 11: Unconfined compression versus versus distance from the

top of the core (TOC) for the upper core section.

Figure 14: Corrected log versus distance from the top of the core
(TOC) for the upper core section. Borehole breakout
occurrences coincide well with regions of low strength.

Figure 12: Unconfined compression versus distance from the top of

the core (TOC) for the upper core section. The log
measurements were corrected to represent more truthfully
the UCS values evaluated through laboratory testing and
scratch testing measurements.