The Confining Effect of End Roughness on Unconfined Compressive
Strength
Z. Szczepanik, D. Milne & C. Hawkes
Department of Civil and Geological Engineering, University of Saskatchewan, Canada
Proceedings of the 1st Canada-US Rock Mechanics Symposium, Vancouver, Canada,
2007, pp. 191-198.
ABSTRACT: The influence of sample end effects on the unconfined compressive strength of rock core is well
recognised. ASTM standards exist to ensure that minimum standards of sample smoothness are maintained to
minimise the influence of friction between the samples ends and loading platens. Sample end preparation is
also done to avoid stress concentrations on irregularities on the end surfaces. This paper describes tests that
have been conducted on relatively uniform grey granite from northern Manitoba, Canada to investigate the influence of sample end effects. End conditions were varied by polishing the sample ends and by using loading
platens with varying degrees of roughness. In one series of tests, lead foil was placed between the sample ends
and the loading platens to further decrease frictional effects. In all tests, except the lead foil tests, procedures
and sample preparations were conducted within the ASTM standards for unconfined compressive strength
(UCS) testing.
The test results presented show that sample “hourglassing”, as measured using circumferential strain gauges
located near sample ends and at sample mid-points, resulted in strengths as low as 50% of the standard UCS
values. Rougher sample ends and platens produced sample “barrelling” with strengths the same, or slightly
higher than results from standard tests. These results suggest that standard UCS tests are conducted with a
significant degree of effective sample confinement generated by sample end friction.
1 INTRODUCTION
Research into factors influencing unconfined compressive strength (UCS) tests have been conducted at
the University of Saskatchewan for over 10 years.
Early testing was conducted to look at crack initiation and propagation in granitic samples (Eberhardt,
1998). Subsequent creep testing was conducted to
determine if the strength of granitic samples was reduced under long term loads in excess of the theoretical strength, causing unstable crack growth
(Sczcepanik et al., 2003). Based on the results of
these tests, research is concentrating on the influence
of sample end conditions on sample strength
(Sczcepanik et al., 2005). Test results have shown a
relationship between the ratio of circumferential
strain at the sample mid point and sample ends versus the sample UCS, and that varying sample to
platen contact friction can change the resulting sample UCS by up to about 100%. This paper presents
the results of continued tests in this area. Modifications have been made to test procedures to try and
vary sample end friction as much as possible.
2 SAMPLES AND PLATENS USED FOR THE
TESTING PROGRAM
2.1 Sample Descriptions
Samples of a medium-grained grey granite from
northern Manitoba have been tested. The granitic
rock was divided into two groups based on the
P-wave velocity measured for each sample. The
slower velocity samples ranged from 3161 metres
per second (m/s) to 4373 m/s and the faster velocities were between 4496 m/s to 5134 m/s. Two
groups of sample sizes were tested as well. The
smaller samples were 35 mm in diameter and the
larger samples had diameters of 61 mm. All tests
had a length to diameter ratio between 2.0 and 2.5,
which is within ASTM (1987) specifications.
2.2 Sample Instrumentation
All samples were strain gauged. Circumferential
strain gauges were mounted 10 mm from each sample end and at the sample midpoint.
 For the 35 mm diameter samples, both 14mm and
90 mm long strain gauges were used. For the
samples instrumented with 14mm long gauges, 2
gauges were installed at each location and 2 axial
strain gauges were also used. For the tests conducted with 90 mm long gauges, only 1 circumferential gauge was used at each location.
 For the 61 mm diameter tests, 90 mm long strain
gauges were used with circumferential gauges 10
mm from each sample end and the sample midpoint. Two axial strain gauges were also used.
Two gauges were used at each location to provide
redundancy.
2.3 Surface Roughness Measurements
Both sample end conditions and platen conditions
were varied and a method of quantifying the surface
conditions of the sample ends and platens was
adopted. A portable surface roughness tester (profilometer) was used to measure sample end roughness to the nearest 0.01 μm. The roughness tester
measured roughness along a 12.5 mm profile length.
Average surface roughness, Ra, was recorded. Ra is
calculated by first determining an average straight
profile to represent the surface trace. The areas
above and below the profile are calculated and added
together. This total area is divided by the straight
line profile length to determine the average profile
roughness parameter Ra. To account for any directional anisotropy in the roughness of the sample
ends, roughness was measured at 60° increments on
the sample surface to produce a “roughness rosette”.
An average roughness value for each sample end
was obtained by averaging the roughness rosette
values. Figure 1 shows how the Ra value is calculated.
effects between the sample ends and platens and did
not conform to ASTM standards.
Three platen conditions were also used for testing. In all cases the steel platens had a hardness in
excess of 58 (Hardness Rockwell C) HRC, as specified in the ASTM standards (1987). No effort was
made to match the elastic properties of the rock and
platens. Instead, the contact friction between the
sample and platens was varied. Polished and striated
platens were used, all of which conformed to ASTM
standards. The polished platen was prepared on a
thin section polishing wheel whereas the striated
platen (Figure 2) was prepared on a fine grinding
wheel.
A third platen type was used that consisted of
concentric grooves with a roughness in excess of
ASTM standards. This platen is shown in Figure 3.
The variation in platen types was used to vary
sample to platen contact friction. The three types of
sample end finish and three platen types that were
used are summarized in Table 1, along with corresponding ranges in average roughness. The sample
tested using lead foil are also listed in this table. The
thickness of these foils were 0.015 mm, 0.03 mm, or
in three cases, 1 mm.
Profile length
Figure 2. Striated platen showing scratch marks. Scale bar
shows 1 cm
j
a b
Roughness
average (Ra)
c
d
e f
g
h
Centreline
i
Ra 
k
l
m
n o p
ya  yb  yc    yn
n
Figure 1: Average roughness, Ra. (after Hebert, 2004)
2.4 Platen and Sample End Preparation
Three sample groups were prepared with rough,
standard and polished end conditions. All of these
sample end conditions were within ASTM standards
(ASTM, 1987). An additional suite of samples was
tested with lead foil placed between the platens and
the rock surface. This was done to reduce frictional
Figure 3. Rough platen showing concentric grooves. Scale bar
shows 1 cm
Table 1. Platen types and sample end finishes used in laboratory UCS tests
Platen type / Sample end finish
 Polished platens
 Smooth / striated platens
 Concentric grooved platens
 Rough sample end condition
 Standard / smooth sample end
condition
 Polished sample end condition
Average Roughness
0.17 μm<Ra<0.21 μm
0.8 μm<Ra<1.0 μm
4.0 μm <Ra<4.6 μm
3.8 μm <Ra< 4.3 μm
2.4 μm <Ra< 3.0 μm
0.6 μm <Ra< 1.2 μm
3 TEST RESULTS
ASTM testing procedures were followed for sample
testing unless otherwise stated. The sample loading
rate was kept constant at 8.1 MPa per minute, which
failed the samples in the ASTM recommended time
period of 5 to 15 minutes for all but the strongest test
results. Load and strain values were continuously
recorded.
After each test, the sample strains at 50% of the
sample UCS were assessed. The ratio of the circumferential strain at the sample mid-point to the average circumferential strain at the sample ends was
calculated. The smallest value for this ratio represents the sample end that experienced the greatest
strain, which is interpreted as the end with the minimum surface to platen confining friction.
Table 2 summarizes the ratios of minimum midpoint strain to average end-point strain measured at
50% of the sample UCS. It is interesting to note that
the polished samples tested on the striated platens
show the highest circumferential strain ratio, suggesting that the contacts between polished samples
and striated platens experienced the highest friction.
Rough samples tested on the grooved platens, however, show approximately 1-to-1 mid-point to endpoint circumferential strain ratios, indicating significantly less friction. It appears that the relatively
rounded groove surfaces of the concentric grooved
platens produce less rock-to-platen friction under
loading than the polished or striated platens, even
though these have much lower roughness (Ra) values.
Additional tests were also conducted on the large
samples with the lower compressional wave velocity. Brazilian tests were conducted, according to
ASTM standards (ASTM, 1995), to obtain an estimate of the tensile strength of the rock. 8 samples
were tested and gave an average tensile strength of
12.1 MPa, with a standard deviation of 1.0 MPa.
4 TEST INTERPRETATION
The frictional bond between the loading platens and
sample ends is not well understood, however, the
application of circumferential strain gauges at sam-
ple ends and mid-point allows the frictional effect to
be indirectly quantified. Figure 4 shows measured
sample UCS versus the sample circumferential strain
ratio. Four groups of test results are shown. Tests
conducted with lead foil on the sample ends have
been circled. The granite samples with the lower
compressional wave velocities were expected to be
weaker and in general a slightly lower strength is obtained for these samples at a given circumferential
strain ratio (Figure 4).
Tests with a mid-point to end-point circumferential strain ratio greater than 1.0 are deforming to a
barrel shape, showing significant end friction. In this
circumferential strain ratio range, relatively high
sample strengths between 200 and 250 MPa were
consistently measured. At circumferential strain ratios less than 1.0 the samples developed an hourglass
shape, signifying reduced end friction. Sample
strengths in this circumferential strain ratio range
show a distinct trend of reduced strength with increased hourglassing. Strengths range from just over
200 MPa to under 100MPa. Of the 16 samples that
show some degree of hourglassing, 8 were tested
with lead foil, 7 were tested with polished ends and
polished platens, and one was tested with polished
platens and standard smooth ends. It is interesting to
note that the smooth striated platens produced some
of the highest sample strengths and the largest degree of sample barrelling. The striated platen was
finished on a fine grained aluminum oxide polishing
wheel that left very shallow scratches on the platen
surface. The scratched surface produced a low Ra
roughness value of 0.8 to 1.0 μm, however, the
scratched or striated surface produced an apparent
high friction between the samples and platen.
The failure mechanism, as well as strength, between hourglassing and barrelling samples appears
to be different (Figures 5 and 6). Figure 7 shows the
location of the circumferential strain gauges. Figure
5 shows a typical failed sample that deformed to a
barrelled shape before failure. The sample ends were
relatively intact after failure and a shear type failure
developed in the sample. Figure 6 shows a typical
failed sample that deformed to an hourglass shape
before failure. Axial fracturing developed in the
sample and the UCS was significantly lower than for
the samples that deformed to a barrel shape prior to
failure. The three test results for the lower velocity
samples with the greatest degree of hourglassing before failure had circumferential strain ratios between
0.48 and 0.50 (Table 2) and an average UCS of 98
MPa (Figure 6). It is interesting to note that the average unconfined compressive strength for these three
samples is 8.1 times the tensile strength. This is remarkably close to the theoretical 8:1 ratio between
UCS and tensile strength predicted with Griffith’s
crack theory (Griffith, 1924).
Table 2. Minimum ratio of sample mid-point to end-point circumferential strain
Circumferential
strain ratio for
rough sample
ends
Sample / platen description
61mm diameter samples:
Concentric grooved platens
Circumferential
strain ratio for
smooth / standard ends
Circmferential
strain ratio for
polished sample ends
1.113
1.029
Smooth striated platens
1.449
1.847
1.778
Polished platens
1.802
1.111
0.822
0.482 lead foil
0.502 lead foil
0.489 lead foil
35mm diameter samples:
Polished platens
1.065
1.12
1.063
0.799
0.743
0.838
0.862
0.845
0.505
0.908*
0.837*
1.409*
0.637 lead foil*
0.818 lead foil
0.524 lead foil
1.552 lead foil
0.790 lead foil
* One sample end strain gauge failed for these tests – the remaining sample mid to end
circumferential strain ratio is shown
300
Sample Barrelling
Sample Hourglassing
250
UCS (MPa)
200
Samples Tested with Lead Foil
150
100
High Velocity Small Samples
High Velocity Small Samples - 1 End Gauged
Low Velocity Large Samples
High Velocity Large Samples
50
0
0
0.5
1
1.5
2
Strain Ratio
Figure 4. Sample UCS versus the minimum mid-point to end-point circumferential strain ratio.
Figure 5. Post-failure photograph of a sample that showed relatively high strength with pronounced sample barrelling before failure.
Circumferential
Gauges
Figure 6. Post-failure photograph of a sample that showed relatively low strength with sample hourglassing before failure.
Strain
Figure 7. Strain gauged sample showing circumferential
gauge locations.
5 CONCLUSIONS
Testing has been conducted on relatively uniform
medium grained granitic rocks. The samples tested
were divided into two groups based on compressional wave velocity, and the lower velocity samples appeared to have a slightly lower strength. Circumferential strain gauges applied at centre of the sample
and 10 mm from the each sample end provided an
indirect measure of the friction end confinement due
to the contact between the platens and the rock.
Sample hourglassing resulted in sample strength
dropping to less than 50% of the maximum UCS
values obtained. In half the cases of sample hourglassing, lead foil had been placed between the loading platens and the sample ends. Hawkes and Mellor
(1970) reported similar testing procedures with the
application of paper or Teflon between the loading
platens and samples. It was reported that thicker layers, in excess of 0.5 mm, extruded under loading and
the failed samples exhibited axial cleavage. It was
thought that the axial cleavage was evidence that the
material placed between the sample ends and platens
was inducing a tensile stress at the samples ends and
this approach was not recommended. The tests reported in this paper suggest that rock failure by axial
cleavage can be obtained simply by reducing end
friction during testing and that this can significantly
influence the reported unconfined compressive
strength of the rock. Additional testing is planned at
the University of Saskatchewan to test other rock
types and to better quantify sample end constraint
and friction effects.
REFERENCES
ASTM D2938-86. 1987. Standard test method for unconfined
compressive strength of intact core specimens, ASTM,
Philadelphia, Pennsylvania, USA.
ASTM D3967-95a. 1995. Standard Test Method for Splitting
Tensile Strength of Intact Rock Core Specimens, ASTM,
Philadelphia, Pennsylvania, USA.
Eberhardt, E. 1998. Brittle rock fracture and progressive damage in uniaxial compression. Ph.D. thesis, Department of
Geological Sciences, University of Saskatchewan, Saskatoon.
Griffith, A.A. 1924. Theory of Rupture, First Intern. Congr.
Appl. Mech., Delft, pp. 55-63.
Hawkes, I. And Mellor, M. 1970. Uniaxial testing in rock mechanics laboratories, Engineering Geology, 4, pp 177-285.
Hebert, M., 2004, “Get the Roll Surface Right”, Plastics Technology,Website, http://www.plasticstechnology.com/articles
Szczepanik, Z., Milne, D., Kostakis, K. and Eberhardt, E.
2003. Long term laboratory strength tests in hard rock, International Society of Rock Mechanics, Gauteng, South Africa.
Szczepanik, Z., Milne, D., Hawkes, C. and Greenlay, K. 2005.
The influence of end effects on unconfined compressive
strength, CGS - AGM, Saskatoon, September, (CD-ROM).