Sei sulla pagina 1di 16

Rock Mech Rock Eng

DOI 10.1007/s00603-017-1271-6

TECHNICAL NOTE

A Benchmark Experiment to Assess Factors Affecting Tilt Test


Results for Sawcut Rock Surfaces
L. R. Alejano1 • J. Muralha2 • R. Ulusay3 • C. C. Li4 • I. Pérez-Rey1 •
H. Karakul5 • P. Chryssanthakis6,7 • Ö. Aydan8 • J. Martı́nez9 • N. Zhang4,10

Received: 11 May 2017 / Accepted: 25 June 2017


Ó Springer-Verlag GmbH Austria 2017

1 Introduction interlocking may contribute to the variability and the non-


reproducibility of tilt test results.
From the earliest studies on the topic, plane sliding tech- A number of authors have carried out simple tilt tests in
niques, usually known as tilt tests, have shown contradic- the past, and several examples illustrate the above-men-
tory features. On the one hand, they reflect on a small-scale tioned arguments. Thus, Hencher (2012) observed extreme
basic principles regarding definition of the friction angle variability in the results of tilt tests; Nicholson (1994)
and can reproduce conditions very similar to those of the found that friction angles for sawcut Berea sandstone in
sliding of blocks on rock slopes; furthermore, tilt angle direct shear tests varied by 12.5°, despite great attention
results agree with friction angles derived from shear and paid to sample preparation and reproducibility; Coulson
pull tests (Hencher 1977; Muralha 1996). On the other (1972) demonstrated that the friction angle of planar sur-
hand, the literature includes several reports of erratic faces of rock varied with surface finish; Krahn and Mor-
results and it is recognized that even apparently smooth genstern (1979) reported similar variation for surfaces
surfaces are actually rough at the microscopic level prepared in different ways and with different surface fin-
(Hencher and Richards 2015). Thus, adhesion and textural ishes; and Kveldsvik et al. (2008) found that the basic
friction angle for a rock slope, derived from tilt testing of
This manuscript contains interesting findings in rock mechanics core, varied between 21° and 36.4°.
practice, derived from large efforts to test the same rocks under These great differences in measured tilt angles are
different conditions and different laboratories. At the same time, this mainly attributed to different surface finishing and to wear
manuscript presents part of the work which will be used to prepare a
Suggested Method on Tilt Testing. It is published in the form of a of the rock contacts (Pérez-Rey et al. 2015, 2016), although
Technical Note, although its length exceeds what is generally other reasons, such as testing and ambient conditions,
accepted for this type of publication. cannot be ruled out. Mehrishall et al. (2016) revealed that
the residual friction coefficients of grinded joint surfaces
Electronic supplementary material The online version of this
article (doi:10.1007/s00603-017-1271-6) contains supplementary and of rough rock surfaces were almost identical.
material, which is available to authorized users.

& L. R. Alejano 5
Department of Petroleum and Natural Gas Engineering, İzmir
alejano@uvigo.es Kâtip Çelebi University, İzmir, Turkey
6
1 Norwegian Geotechnical Institute, Oslo, Norway
Department of Natural Resources and Environmental
7
Engineering, University of Vigo, Vigo, Spain COWI, Oslo, Norway
2 8
National Laboratory of Civil Engineering, Lisbon, Portugal Department of Civil Engineering and Architecture,
3 University of the Ryukyus, Okinawa, Japan
Department of Geological Engineering, Hacettepe
9
University, Ankara, Turkey Centro Universitario de la Defensa de Marı́n, Marı́n, Spain
4 10
Department of Geoscience and Petroleum, Norwegian North China Electric Power University, Beijing, China
University of Science and Technology (NTNU), Trondheim,
Norway

123
L. R. Alejano et al.

The benchmark experiment performed by the authors 8 9 792.5 cm using the own saw machines and to perform
intends to shed light on the observed variability in sliding tilt tests using the available tilting tables (Fig. 1b, c). Thus,
angle results from tilt tests. A large number of tilt tests in each slab-like specimen would have 2 faces with a surface
different conditions have already been carried out, of approximately 56 cm2 to carry out tilt tests. This area is
including those reported in Alejano et al. (2012) and in slightly over the recommended value of 50 cm2 established
some more recent studies (González et al. 2014; Ruı́z 2014, in previous experiments to ensure a large enough tilt sur-
Ulusay and Karakul 2016; Li et al. 2017), as well as several face and to avoid problems related to the curvature of the
other tests not presented in the literature. Results suggest a sawcut surfaces (Alejano et al. 2012). The slab specimens
reasonable degree of reproducibility of tilt testing under fulfilled the minimum recommended length-to-width ratio
controlled conditions. of 2 (in our case, it was, in fact, greater than 3) that ensures
The above-mentioned studies typically used different that contact stresses remained compressive when sliding
cutting machines, tilting tables and testing procedures. It occurs.
was therefore deemed interesting to check reproducibility Since 7 slab specimens with 2 surfaces—labelled ‘a’ and
by analysing tilt test results carried out in different labo- ‘b’ (Fig. 1b)—were obtained from each original rock
ratories in very similar controlled conditions, in order to block, it was possible to carry out tilt tests on numerous
analyse how and why results were different. The intention contacts. The benchmark experiment tests were carried out
is to contribute to the standardization of this very simple, on 7 of the possible contacts between surfaces 1a–2a, 1b–
yet somewhat controversial, rock mechanics test. 2b, 3a–4a, 3b–4b, 5a–6a, 5b–7b and 6b–7a. Tilt tests were
The initial idea was to test 7 contacts of 7 specimens of a repeated for each contact 7 times, with blocks held in the
particular rock type—granite—7 times, in the different same initial positions and tested along the same sliding
cutting and tilting conditions inherent to the equipment direction.
from each laboratory. However, 3 other rock types (lime- The following procedure for carrying out the tilt tests
stone, quartzite and basaltic andesite) were also tested was agreed:
under 3 or 4 different conditions to take into account
1. Clean specimen surfaces to remove dust and free
possible differences resulting from different mechanical
particles.
characteristics.
2. Place the specimen horizontally over the other spec-
Note that this benchmark experiment is not an in-depth
imen so that the upper block slides along the longer
study of the theoretical micromechanic issues that influ-
dimension (H).
ence the behaviour and mechanics of contacts in tilt tests.
3. Ensure that the sliding surface is positioned horizon-
It is, rather, an experimental study that aims to make
tally using (preferably) a bubble level.
information on a large number of controlled tilt tests
4. Tilt the table steadily until the upper slab slides along
available to the rock mechanics community. The main
the lower slab at a (preferable) tilt velocity of 10°/min
objectives of the experiment were to identify and single
(i.e. 3 min for 30° sliding).
out features that significantly affect tilt test outcomes and
5. Note the platform sliding angle value b (°) and tilt
to assess the reproducibility of tests performed in different
velocity (°/min), room temperature and relative
laboratories.
humidity.
6. Repeat points 1–6 a total of 7 times for the 7 surfaces.
2 Materials and Methods Ultimately, 7 tilt tests were carried out on each of the 7
surfaces, yielding 49 sliding angle values for each original
2.1 Experimental Approach rock block.

The authors prepared a good number of sample blocks of 2.2 Tested Rocks
the same rock with dimensions 30 9 897 cm and weight
4–5 kg per block (Fig. 1a) that could be easily mailed to Tested were 4 different rocks (igneous intrusive and vol-
the 4 laboratories that participated in the multi-laboratory canic, metamorphic and sedimentary) in at least 3 labora-
experiment: the University of Vigo (UVIGO) in Spain, the tories (Fig. 2):
National Laboratory for Civil Engineering (LNEC) in
(a) Granite from north-west Spain (locally known as
Lisbon (Portugal), Hacettepe University (HU) in Ankara
Blanco Mera). This bright white-coloured granite has
(Turkey) and the Norwegian University of Science and
a coarse-grained texture showing grains with sizes
Technology (NTNU) in Trondheim (Norway). The
typically ranging from 1 to 6 mm and unconfined
benchmark procedure was to ask each laboratory to cut 7
compressive strength (UCS) of roughly 130 MPa.
slab-like specimens with dimensions of about

123
A Benchmark Experiment to Assess Factors Affecting Tilt Test Results for Sawcut Rock Surfaces

Fig. 1 a Original block of


granite with its dimensions.
b Cutting process to produce 7
slabs (leaving some spare pieces
for use if needed. c 7 slabs cut
from the original block

Fig. 2 Surfaces of slab specimens (around 80 9 70 mm) before testing. a Spanish granite; b Portuguese limestone; c Portuguese quartzite.
d Turkish basaltic andesite

123
L. R. Alejano et al.

(b) Limestone from Portugal. This fossiliferous bioclas- quantity in carats per cm3 and other possibly relevant data
tic Cretaceous beige limestone (biosparite), mostly are also summarized.
made up of calcium carbonate with an abundance of On studying the surfaces of the granite slabs under a
fossils and fossil traces, is a hard, low-porosity binocular microscope, it was observed that those cut with
limestone with a UCS of about 100 MPa. jagged saw blades (A and B) had deeper and more dis-
(c) Quartzite from Portugal. This very hard rock, mainly continuous grooves than those cut with continuous rim saw
consisting of quartz and very small amounts of mica, blades (C and D), a difference which may affect tilt test
has rather regular quartz grains of sizes varying results.
between 1 and 3 mm and a UCS of over 200 MPa.
(d) Basaltic andesite from Turkey. This slightly altered,
2.4 Tilting Tables
vesicular, hypocrystalline and porphyritic rock,
mainly containing minerals such as biotite, clinopy-
Figure 4 shows the 4 tilting tables used to carry out the tilt
roxene and plagioclase and including abundant
tests in the 4 laboratories.
vesicles, has strength (indirectly Schmidt-hammer
The UVIGO tilting table (Fig. 4a) is motorized and has
measured) in the range 60–80 MPa.
a regulation system that allows tilt velocity to be set at
2.5°–36°/min with almost no vibration. For this experi-
2.3 Cutting Machines and Saw Blade Discs ment, a velocity of 10°/min was used for all the tilt tests.
The HU tilting table (Fig. 4b) is hand-operated: a wheel
Different cutting machines and saw blades or discs were connected to a screw is spun to make the platform tilt. Tilt
used to cut the specimens. Since the impact of the cutting tests for this experiment were carried out at an average
process was considered to be possibly relevant, the main velocity of 24°/min.
specifications of the discs used are presented in Fig. 3 and The LNEC tilting table (Fig. 4c), also hand-operated,
Table 1. has quite a robust system, consisting of a screw and a
The circular saw blades for hard rock are described in rotating system with various levers, for carrying out tilt
terms of blade diameter and blade thickness, as well as tests. Tests for this experiment were carried out at an
number of teeth and tooth width for jagged blades. The average tilting velocity of 10°/min.
bonding matrix of the cutting material, the observed size of The tilting table at NTNU (Fig. 4d), controlled by a
diamonds in the base material (tooth grit), the estimated lifting cylinder driven by compressed air, is quite accurate

Fig. 3 Saw blades used by 4 laboratories to cut rock specimens (detailed information in Table 1)

123
A Benchmark Experiment to Assess Factors Affecting Tilt Test Results for Sawcut Rock Surfaces

Table 1 Main saw blade features (see Fig. 3) used by 4 laboratories to cut rock specimens
Feature Blade type
A-UVIGO B-UVIGO C-UVIGO D-UVIGO C-HU E-LNEC D-NTNU

Blade diameter (mm) 350 600 300 230 300 500 250
Blade thickness 3.4 3.5 2.4 2 2 4 2
(mm)
Number of teeth 23 42 24 slots Cont. rim 22 slots 60 (2 9 30) Cont. rim
Tooth width (mm) 15 15 10 10 11 20 10
Bonding material Bronze alloy Bronze/cobalt alloy Bronze alloy Bronze alloy Bronze alloy Bronze alloy
matrix matrix matrix matrix matrix matrix
Diamond count (ct/ 0.60 0.97 0.79 0.62–0.79 0.6–0.8 0.62–0.79
cm3)
Tooth grit (US mesh) 50–60 50–60 80–100 60–80 80–100 60–100
estimate
Other features – Brand new saw – – – Much-used –
machine saw blade

Fig. 4 Tilting tables used for tilt tests by 4 laboratories

and shows no signs of vibration. It has a constant tilting significantly affect results (Pérez-Rey et al. 2016), velocity
velocity of approximately 36°/min. for this benchmark experiment was not defined in a precise
LNEC’s tilt table used a blocking plate (Fig. 5b) to manner, although a value of 10°/min was recommended.
avoid full sliding of the upper block that stops after a small Due to the large number of tests, some laboratories (HU
run of approximately 1 cm for these tests. This issue was and NTNU) used higher angular velocities of 24° and 36°/
not accounted for in the other 3 laboratories, where the min, respectively.
upper slab is allowed to slide through its full length Since acceleration occurred in the platform of the hand-
(Fig. 5a). operated tilting device used in HU, caused by variations in
Since the authors found, in a previous study carried out manually controlled velocity, the impact vibrations were
with a motorized tilting table, that tilt velocity did not theoretically and experimentally investigated. According to

123
L. R. Alejano et al.

Fig. 5 Different tilt test sliding


displacements. a Equipment
allowing full sliding of the
upper slab (HU, UVIGO and
NTNU). b Equipment with a
blocking system to stop upper
slab after a short displacement
of around 1 cm (LNEC)

Fig. 6 Impact of horizontal acceleration on tilt tests associated with vibration of the tilt testing device

basic statics, the equilibrium of a block on an inclined applied to the potentially sliding block. Figure 6 also illus-
plane was computed considering upper block horizontal trates how to estimate the friction angle of the contact surface
acceleration of ±ag, where a is the horizontal acceleration /±a for several values of a, given the observed sliding angle
coefficient, that is, the ratio of horizontal acceleration to b. As shown in the table in Fig. 6, for a sliding value of
which the block is submitted due to vibration in relation to b = 38° and a vibration coefficient a of 0.10, the friction
gravity, denoted by g (Fig. 6): angle /?a will be 43.71° in the case of outwards acceleration
or /-a = 31.71° in the case of inwards acceleration.
/a ¼ b  arctan a ð1Þ
To check for the impact of vibration on determination of
This means that in case an outwards acceleration ag is the basic friction angle, tilt-induced accelerations were
applied to the potentially sliding block, the sliding angle will experimentally measured. These measurements were made
be the basic friction angle decreased in arctana degrees. for the hand-operated tilting devices (hard-running and
Analogously, the sliding angle observed will increase in smooth-running machines) used in HU and in the Univer-
tan-1a degrees when a horizontal inwards acceleration ag is sity of the Ryukyus (UR). A QV3OAM-SYC-type

123
A Benchmark Experiment to Assess Factors Affecting Tilt Test Results for Sawcut Rock Surfaces

accelerometer, based on smart monitoring technology, as for the 7 test repetitions on every surface are denoted b1–
developed by Aydan (2015), with a measuring capacity of b7.
2 g and storage and filter capacities of 2 GB and 100 Hz, The results obtained for 21 rock blocks (8 granite, 5
respectively, was used. limestone, 5 quartzite and 3 basaltic andesite) cut with
Two accelerometers were mounted on the tilting different discs in the 4 laboratories (HU, NTNU, UVIGO
table and the base. For the HU tilting device, vibrations and LNEC) are shown in Tables A.1–A.8 (granite), A.9–
were measured for tilting rates of 22, 32 and 39°/min, while A.13, (limestone), A.14–A.18 (quartzite) and A.19–A.21
for the University of the Ryukyus (UR) device, they were (basaltic andesite) in the form of electronic supplementary
measured for a tilting rate of 27°/min. Only acceleration material (Online Resource 1). In these tables, available at
responses for the tilting tables and tilting angle variations Online Resource 1, the sliding angles measured for the 7
over time were recorded. Maximum acceleration during repetitions of 7 tests for every original block are pre-
HU measurements was 0.030–0.055 g, and acceleration at sented. Table 3 summarizes the mean and standard devi-
the time of sliding was 0.002 g. According to the table in ation of all the measured angle values of every set of 49
Fig. 6, this level of acceleration may cause about 2° dif- tests (7 repetitions for 7 surfaces) performed in each
ference between the true friction angle and the sliding laboratory for the different rock types. Table 3 also pre-
friction angle. The maximum tilt-induced acceleration of sents the mean and standard deviation of the angles of
0.004 g for the UR tilting device was negligible. first slide of every test, when the contact surface has
Additionally, using a smartphone application called suffered no wear at all
Accelerometer Analyser, it was confirmed that maximum Graphs with the results for each series of 49 tests for the
measured acceleration for the UVIGO, LNEC and NTNU 21 specimens in groups of 7 repetitions were plotted,
devices was always less than 0.005, 0.006 and 0.003 g, allowing to easily detect outliers. For illustrative purposes,
respectively, and therefore practically negligible. Fig. 7 shows one such plot: the results of a set of 7 repe-
titions of a granite specimen test performed at HU. Other
2.5 Environmental Conditions plots are not reproduced here for reasons of brevity, but all
the original data are available in the Online Resource 1. In
No constraints related to environmental conditions were Table 3, the average slope of the decreasing sliding angle
stipulated for the benchmark tests, although temperature with repetitions, as shown in Fig. 7, is presented too, since
and relative humidity had to be recorded during testing. this is indicative of the wear of surfaces.
Average temperature and relative humidity values and An analysis of the graphs immediately allowed some
ranges recorded at each laboratory during tilt tests are conclusions to be drawn. Wear effects were more relevant
summarized in Table 2. While temperature is not deemed in the UVIGO tests, with varying levels evident on the
to significantly affect tilt test results, it is unclear whether surfaces of different specimens from the same rock block.
relative humidity might have affected contact behaviour in This suggests that it is convenient to try to avoid this effect.
particular cases. Particularly evident wear effects were observed for 2 sets
of the 7 tests of limestone cut with saw blade C
(Table A.11 at Online Resource 1, contacts 1b–2b and 6b–
3 Results 7a). This wear effect, accompanied by a significant
reduction in the sliding angle since the first repetitions, can
3.1 Raw Data be attributed to the small grain size, which can rapidly
produce a polished effect. In the case of UVIGO, this effect
Raw data resulting from some 1000 tilt tests of different was amplified by high relative humidity and, possibly,
rocks cut with different discs and tested in different labo- more thorough cleaning.
ratories are described in the following. Sliding angle values In the LNEC tests on quartzite, a single set of tests for
each rock block resulted in unusually low values
(Tables A.17, contact 3b–4b and Table A.18, contact 6b–7a
Table 2 Average temperature and relative humidity values and at Online Resource 1). This was attributed to the low
ranges recorded during tilt testing at 4 laboratories
planarity of surfaces, with grooves induced in this partic-
Temperature (°C) Relative humidity (%) ularly hard rock by the (purposely) much-used saw blade.
HU 28 (25–32) 45 (35–50) Defective contact between surfaces, which can actually be
NTNU 20 (19–21) 36 (30–40)
felt by touch, may have drastically reduced the sliding
UVIGO 23 (22–26) 53 (45–60)
angle. However, this phenomenon was only observed in 2
sets of tests (out of 14) and in the most strenuous condi-
LNEC 22 (20–25) 62 (50–75)
tions, that is, the hardest rock cut with the weakest disc.

123
L. R. Alejano et al.

Table 3 Series of 49 tests (7


Granite Limestone Quartzite Basaltic andesite
repetitions for 7 surfaces)
performed in 4 laboratories (21 HU 1 1 1 1
specimens)
NTNU 1
UVIGO 4 (4 distinct discs) 2 (2 distinct discs) 2 (2 distinct discs) 1
LNEC 2 (same disc) 2 (same disc) 2 (same disc) 1

A possible reason for these low HU values may be the


acceleration detected in the tilting table. Another reason
may be the microroughness created by the saw blades. As
can be seen in Table 4, the sliding angles for the specimens
cut using continuous rim saw blades (Fig. 3) in HU (C) and
UVIGO (C and D) were smaller than those cut using jag-
ged saw blades. As also previously mentioned, jagged or
teethed discs (A and B in Fig. 3) produce deeper but more
discontinuous grooves than the continuous rim saw blades
(C and D in Fig. 3). Surfaces cut with the C-type blades
thus had less microroughness, which would suggest that
these specimens may slide at lower angles than specimens
cut with teethed discs.
Regarding the impact of the repetitions, the data clearly
indicate that the slope was systematically much greater for
Fig. 7 Results for a set of 7 tilt test repetitions for the same granite UVIGO than for the other laboratories (between -1° and
performed at HU -2° vs. between ?0.1° and –0.6°). The main explanation
for this difference is related to the distance that the dif-
As would be expected, for both limestone in UVIGO ferent equipment allowed the upper block to slide (Fig. 5).
and quartzite in LNEC, the results revealed greater scat- LNEC, where a blocking system was used, had the lowest
tering and less reliability. These 2 examples also illustrate slope values (an average of 0.17°/repetition); at NTNU
how small details or lack of due care may easily lead to also, blocks were not allowed to slide the full length since
unexpected test results, suggesting that the occurrence of they were stopped once the upper block began to move.
outliers is not rare in tilt testing. It is therefore evident that Other causes may relate to differences in the pretest
outliers should be excluded when interpreting tilt test cleaning procedures (with a brush or with a cloth) followed
results. in different laboratories.
To start analysing the data, Table 4 compiles the means
and standard deviations for each set of 49 results (7 test 3.2 Wear-Corrected Data
repetitions for 7 surfaces of the same rock tested under the
same conditions), as well as the means and standard Surfaces clearly wore when tests were repeated. This effect
deviations for the first 7 tilt tests from each surface. Bear in was particularly evident in the results obtained for UVIGO
mind that if wear deteriorated the surface, the first repeti- (half of all the results). To enhance the comparability of
tion values would be representative of sliding along the data, all sliding angle datasets were corrected in the sim-
undamaged surface. Table 4 also includes values for the plest possible manner for this wear effect; for each set of 7
average slope for the 7 test repetitions for each block (see repetitions for each contact, the slopes (as shown in Fig. 7)
Fig. 7). This slope is a simple statistic describing the were calculated using linear least squares and the values of
decrease of the sliding angle with consecutive repetitions, the repetitions were corrected according to this slope. Thus,
i.e. reflecting surface wear. the first-sliding value was retained and, for the rest of the
Overall, mean values from HU and UVIGO were lower results, the value of the slope multiplied by the number of
than those obtained in the other laboratories. Looking only previous repetitions was subtracted. The corrected results
at the first slide means (no wear effect is reflected), it can for the 21 rock blocks tested are presented in tables as
be observed that it was mainly the HU values which were electronic supplementary material (Online Resource 2).
lower than those of the other laboratories. For limestone, Table 5 includes the most important results for the
the differences were small (2°–3°), but for the basalt they corrected data: means, standard deviations and coefficients
were large (8°–9°). of variation. The first slide values are also included for

123
A Benchmark Experiment to Assess Factors Affecting Tilt Test Results for Sawcut Rock Surfaces

Table 4 Means (average) and


Rock type Laboratory All values First slide values Average slope
standard deviations for each set
of 49 tests conducted in 4 Mean SD Mean SD
laboratories, means and
standard deviation for each set Granite HU-C 27.18 1.00 27.63 1.22 -0.12
of 7 first slides performed and NTNU-D 30.69 2.97 31.07 3.08 -0.04
average slopes for the sliding
UVIGO-A 28.55 2.92 31.66 1.42 -1.03
angles with repetitions (see
Fig. 7) UVIGO-B 27.35 3.09 31.13 1.61 -1.31
UVIGO-C 27.28 3.29 32.14 0.86 -1.48
UVIGO-D 27.05 3.88 31.60 2.46 -1.28
LNEC-E1 30.74 2.05 31.40 2.93 -0.13
LNEC-E2 30.44 2.36 29.99 2.44 0.11
Limestone HU-C 35.67 1.52 35.69 1.50 -0.04
UVIGO-B 34.03 5.03 37.39 1.32 -1.74
UVIGO-C 30.50 8.04 36.74 1.54 -2.47
LNEC-E1 38.12 1.84 37.87 2.04 0.08
LNEC-E2 38.17 1.14 39.37 1.26 -0.29
Quartzite HU-C 22.29 1.79 24.00 1.25 -0.53
UVIGO-B 27.45 3.43 31.04 1.93 -1.14
UVIGO-C 29.94 2.91 32.96 1.81 -1.15
LNEC-E1 27.84 3.18 29.59 4.04 -0.37
LNEC-E2 27.12 3.21 29.21 3.09 -0.40
Basaltic andesite HU-C 26.10 2.83 27.64 2.07 -0.62
UVIGO-B 31.58 3.61 34.25 1.31 -1.55
LNEC-E1 34.71 2.09 34.94 2.04 -0.18

comparison purposes, as also the median of the first 3 (ANOVAs) were implemented for the different scenarios
corrected values, so as to simplify standardization of the tilt considered for this experiment, to determine the (statisti-
testing. cal) belongingness of the corrected results (specimens) to
Correlations are plotted in Fig. 8a, b. The correlation the same population.
between the mean first-sliding angles—expected to be the All data corrected for wear effects were represented as
most reliable—and mean corrected values as plotted from boxplots (Fig. 10). For each dataset, median and 25th and
Table 5 was very satisfactory (Fig. 8a), as also was the 75th percentiles are indicated by the central mark and box
correlation between mean first-sliding angles and the edges in each boxplot, respectively. Any values located out
median of the first 3 wear-corrected repetitions (Fig. 8b). of the whiskers were considered to be outliers (indicated by
These results would suggest that any of these parameters red ? signs). For all studied datasets, the fact that the
could be a good estimator of first-sliding angle values. boxplots were found to be reasonably symmetric in terms
Results for potential normal distributions obtained from of edges and whiskers was an indicator of the normality of
the means and standard deviations for the 49 wear-cor- the distributions.
rected values for each rock block and each laboratory are Referring to Fig. 10, for tilt tests carried out with granite
shown in Figs. 9a (granite), 9b (limestone), 9c (quartzite) (G) specimens, medians were generally similar except for the
and 9d (basaltic andesite). A more detailed statistical HU dataset, whose value for the median was lower—as
analysis is presented in the next section. already detected for the case of means analysed above. For
quartzite (Q), median values were found to be more variable,
ranging from a low of 24.23° (HU dataset) to a high of around
4 Statistical Considerations 34° (Q-UVIGO-C). In terms of scattering, the UVIGO and
HU datasets showed less dispersion than the LNEC dataset.
To better understand the results obtained from this exper- These last datasets (Q from HU and UVIGO) were more left-
imental study, a number of statistical tests and techniques skewed than the others, attributable to the polishing of sur-
were applied. Boxplots and probability distribution func- faces. For limestone (L), similar medians were obtained for
tions fitted to histograms were created to study data after UVIGO and LNEC, and a somewhat smaller median was
correction for wear effects. One-way analyses of variance obtained for HU. Limestone was the rock for which the

123
L. R. Alejano et al.

Table 5 Means and standard deviations for each set of 7 first slides performed in each laboratory, means and standard deviations for each set of
49 wear-corrected tests performed in each laboratory and medians for the first 3 wear-corrected repetitions of the 7 contacts for each block
Rock type Laboratory First slide values (7) All wear-corrected values (49) Median of first 3 repetitions
(wear-corrected)
Mean SD Mean SD CV (%)

Granite HU-C 27.63 1.22 27.54 1.00 3.63 27.60


NTNU-D 31.07 3.08 30.82 2.54 8.24 30.97
UVIGO-A 31.66 1.42 31.65 2.11 6.67 31.70
UVIGO-B 31.13 1.61 31.27 1.70 5.44 31.15
UVIGO-C 32.14 0.86 31.71 1.28 4.04 31.90
UVIGO-D 31.60 2.46 30.88 2.24 7.25 30.72
LNEC-E1 31.40 2.93 31.14 2.31 7.42 31.14
LNEC-E2 29.99 2.44 30.12 2.43 8.07 30.33
Limestone HU-C 35.69 1.50 36.03 1.52 4.22 36.25
UVIGO-B 37.39 1.32 39.24 2.20 5.61 39.11
UVIGO-C 36.74 1.54 37.91 4.44 11.71 38.70
LNEC-E1 37.87 2.04 37.89 2.30 6.07 38.61
LNEC-E2 39.37 1.26 39.06 2.27 5.81 39.10
Quartzite HU-C 24.00 1.25 23.94 1.32 5.51 24.30
UVIGO-B 31.04 1.93 30.88 2.02 6.54 30.80
UVIGO-C 32.96 1.81 33.38 1.87 5.60 33.30
LNEC-E1 29.59 4.04 28.94 4.03 13.93 30.30
LNEC-E2 29.21 3.09 28.32 3.57 12.61 27.90
Basaltic andesite HU-C 27.64 2.07 27.95 2.44 8.73 28.70
UVIGO-B 34.25 1.31 36.24 1.88 5.19 36.05
LNEC-E1 34.94 2.04 35.51 2.22 6.25 34.90
CV coefficient of variation

highest sliding angle was determined and where most vari- the original data are presented in the form of electronic
ability—and outliers—was observed. As for the basaltic supplementary material (Online Resource 2).
andesite (BA) tilt test results, the HU dataset again rendered ANOVA is a particular form of statistical hypothesis
the lowest median. The LNEC dataset was the most sym- testing used to analyse experimental data. A statistically
metric, and little left-skewness could be observed for the HU significant result—when a probability (p value) is below a
and UVIGO datasets. specific threshold—justifies rejection of the null hypothe-
Considering in conjunction the boxplot results and the sis, which in this case was that all sets of results are random
mean values shown in Table 5, it can be observed that the samples drawn from the same population. ANOVA is quite
sliding angle values for granite obtained by NTNU, useful to check the repeatability of tests and, in this
UVIGO and LNEC were quite similar and were also research, to assess the influence of cutting with different
comparable to the average wear-corrected sliding angles discs and of using different tilting tables and procedures.
for limestone, quartzite and basaltic andesite obtained by ANOVA was applied to raw data, and results were
UVIGO and LNEC. Values for HU tended to be lower than grouped according to the repetitions. Thus, for each set of
those for the other laboratories—a fact which is attributed results for the same rock type produced by a given labo-
to vibrations in the tilting table. Typical coefficients of ratory, 7 groups were considered. Each of these groups is
variation ranged between 3 and 14%—within the range reflected in the wear-corrected b1–b7 values displayed in
typically observed when testing other geomechanical the tables (Online Resource 1). Table 6 summarizes the
properties such as UCS. results for raw data and wear-corrected data. In both cases,
The wear-corrected data were represented in terms of the null hypothesis was that the groups belonged to the
histograms, and normal distributions for granite, limestone same population, the significance level was p [ 0.05 (re-
and quartzite and basaltic andesite were estimated as well flecting a probability of 0.05 that the null hypothesis would
as means and standard deviations for each dataset. Data be rejected) and the critical value of Fisher-Snedecor’s
and graphs are not reproduced in the interest of brevity, but F was Fcrit = 2.32.

123
A Benchmark Experiment to Assess Factors Affecting Tilt Test Results for Sawcut Rock Surfaces

limestone, and 28.94 and 28.32 for quartzite). It could


therefore be concluded that the tests were reasonably
repeatable.
In order to check whether results for the same rock type
tilted with the same table but cut with different saw blades
were repeatable, the 4 UVIGO granite datasets were
grouped in different manners to test the null hypothesis.
Grouping these datasets two by two (A and B, C and D) or
all together (A, B, C and D) the null hypothesis was
accepted, indicating that all the groups belonged to the
same population (Table 7). Since mean values were also
similar (31.65, 31.27, 31.71 and 30.88; see Table 5), the
tests can also be considered reasonably repeatable. This
means that, for UVIGO, the type of saw blade used
apparently had no significant effect on the results.
Finally, to check that results for the same rock type cut
with different machines and tested in different laboratories
were similar, granite results for NTNU, UVIGO and LNEC
were grouped in twos and in a single group. Again, the null
hypothesis was accepted (Table 7). Mean values were also
similar, indicating that the tests performed in groups of 7 in
3 different laboratories using 6 different saw blades and 3
different tilting platforms were considered repeatable.
However, when HU test results for granite were included,
Fig. 8 a Correlation between average first slide angles and means for the null hypothesis was rejected, indicating that groups of
all wear-corrected values for each block. b Correlation between results did not belong to the same population. The null
average first slide angles and medians for the first 3 slide angles (wear hypothesis was also rejected when results from HU were
corrected) for each block combined with any other results from NTNU, UVIGO and
LNEC.
Table 6 shows that raw data results for all contacts
tested at UVIGO in groups of 7 (first slide, second slide,
etc.) and those for LNEC-E2 for limestone and HU-C for
quartzite do not belong to the same original population. 5 Discussion
This means that the tilt angle reduction induced by surface
wear from previous tests in these tests is relevant and so Although tilt tests applied to the cut surfaces of rocks are
has to be accounted for. The ANOVA results for the wear- simple tests, the mechanical behaviour of surface contacts is
corrected values support this conclusion, since after cor- a complex physical phenomenon in which different issues
rection, almost all data (except for those corresponding to 2 may significantly affect results. Advantages of this test
UVIGO cases) can be considered as random samples drawn reported in the literature typically include simplicity,
from the same populations. observability and repeatability (Hencher 1976). However,
To check whether results for groups of specimens cut drawbacks regarding complex contact micromechanics
with the same saw blade and tilted with the same tilting issues that may affect test results when test conditions are not
table were similar or reasonably repeatable, we analysed well controlled are also reported (Hencher and Richards
LNEC results for granite, limestone and quartzite (for 2015).
which there were 2 groups of results), finding that the block To ensure that contact stresses were compressive when
tests comprehending 7 groups of 7 tests belonged to the sliding occurred and to avoid smaller contact surfaces
same population. If the null hypothesis was accepted and associated with low planarity, rock specimens with sur-
means and standard deviations were similar, then it would faces of around 56 cm2 and a length-to-height ratio greater
be possible to conclude that the tests were reasonably than 3 (80/25 mm) were used in this benchmark experi-
repeatable. Table 7 shows that the null hypothesis was ment. According to stress calculation approaches (Hencher
accepted for LNEC-E1 and LNEC-E2 for granite, lime- 1976; Muralha 1996; Kim et al. 2016), these geometric
stone and quartzite, as the groups yielded similar mean conditions ensure that only sliding (and not any other
values (31.14 and 30.12 for granite, 37.89 and 39.06 for phenomenon, such as toppling) occurs in the tests.

123
L. R. Alejano et al.

Fig. 9 Normal distributions for


the 21 specimens of the
different rocks tested: a granite
(8), b limestone (5), c quartzite
(5) and d basaltic andesite (3)

123
A Benchmark Experiment to Assess Factors Affecting Tilt Test Results for Sawcut Rock Surfaces

Fig. 10 Boxplot representations of wear-corrected results for granite (G), quartzite (Q), limestone (L) and basaltic andesite (BA) for 4
laboratories

Tilt test results may be affected by surface finishing, the the impact of wear is largely reduced when a blocking
outcome of the hardness and grain size of the rock and of system is used to reduce the sliding distance. It can
the type and velocity of the saw blades that are used. The therefore be concluded that restricting complete sliding of
presented results would suggest that average sliding angle the upper block contributes to controlling the wear impact
values are not much affected by the cutting procedure and on tilt test results.
the type of disc used. Rather, the scatter in results may be Although angular velocity does not affect results for tilt
more affected by the type of saw blade used, particularly tests conducted under mechanically controlled conditions
for harder rocks. (Pérez-Rey et al. 2016), results shown in this study would
A complex mechanical issue that may affect results is suggest that it can significantly affect sliding angles when a
surface wear and rock dust or powder between tilted sur- hand-operated tilting platform is used, due to vibrations
faces. The residual sliding angle observed for more pol- associated with an imperfectly controlled testing procedure
ished surfaces tends to be much smaller when powder or to mismatching in the mechanical tilting system. Results
produced by previous slides is removed (Hencher from the benchmark tests corroborate USBR (2009) in
1976, 1977; González et al. 2014). Better knowledge of recommending a low velocity when using hand-operated
these mechanical issues could enhance understanding of machines to avoid vibrations during tilting. Tilting veloc-
the mechanisms affecting tilt test results. It was observed ity, when controlled, does not affect sliding angle results.
that avoiding full sliding of the upper slab by blocking the For hand-operated tilting tables, a low tilting velocity of
process after a limited run (10% of the contact length; see 5°–10°/min should be used to control acceleration and
Fig. 5b) avoids damage to the surface and ensures less ensure comparable results. In view of the acceleration
variable and more representative results when carrying out response of 0.05 g measured for its tilting table, vibrations
repetitive testing. may account for the lower values from HU. Note that the
To further check the effect of restricting sliding distance vibration effect is intensified for hard rocks.
to 10% of specimen length, a further series of 49 tilt tests Relative humidity may also affect results, although not
was carried out at UVIGO (using the same granite and the results of the benchmark experiment here described.
testing procedures as described above). Results, summa- Aydan (1998) and Ulusay and Karakul (2016) have shown
rized in Table A.22 in the Online Resource 1, indicate that that water has no impact if rocks do not absorb it in full-

123
L. R. Alejano et al.

Table 6 ANOVA for results


Rock Laboratory Raw data values Wear-corrected values
for 4 rock types for groups of 7
tests (first slide, second slide, p F Null hyp. p F Null hyp.
etc.) (if p [ 0.05) A (if F \ Fcrit) A
Else R Else R

Granite HU-C 0.364 1.125 A 0.736 0.590 A


NTNU-D 0.983 0.170 A 0.966 0.226 A
UVIGO-A 0.000 7.539 R 0.996 0.102 A
UVIGO-B 0.000 23.528 R 0.438 1.001 A
UVIGO-C 0.000 39.883 R 0.153 1.667 A
UVIGO-D 0.000 6.049 R 0.808 0.496 A
LNEC-E1 0.981 0.178 A 0.999 0.040 A
LNEC-E2 0.993 0.119 A 0.999 0.053 A
Limestone HU-C 0.856 0.427 A 0.922 0.320 A
UVIGO-B 0.000 9.393 R 0.000 5.904 R
UVIGO-C 0.001 4.627 R 0.938 0.290 A
LNEC-E1 0.969 0.218 A 0.995 0.105 A
LNEC-E2 0.001 4.560 R 0.606 0.758 A
Quartzite HU-C 0.000 5.259 R 0.989 0.143 A
UVIGO-B 0.000 6.125 R 0.957 0.247 A
UVIGO-C 0.000 13.137 R 0.917 0.330 A
LNEC-E1 0.691 0.648 A 0.991 0.133 A
LNEC-E2 0.672 0.672 A 0.990 0.138 A
Basaltic andesite HU-C 0.082 2.036 A 0.925 0.315 A
UVIGO-B 0.000 41.319 R 0.003 3.996 R
LNEC-E1 0.791 0.519 A 0.949 0.267 A
A and R indicate that the null hypothesis is accepted or rejected, respectively

Table 7 ANOVA results by group


Rock Datasets: rock type, laboratory and disc type One-way ANOVA results
F p Fcrit Null hyp.

Granite LNEC (E1, E2) 0.354 0.9801 1.839 A


Limestone LNEC (E1, E2) 0.901 0.5554 1.839 A
Quartzite LNEC (E1, E2) 0.170 0.9995 1.839 A
Granite UVIGO (A, B) 0.467 0.9372 1.839 A
Granite UVIGO (C, D) 1.074 0.3925 1.839 A
Granite UVIGO (A, B, C, D) 0.735 0.8256 1.552 A
Granite UVIGO (A, B, C, D) ? LNEC (E1, E2) 0.739 0.8775 1.438 A
Granite NTNU (D) ? UVIGO (A, B, C, D) 0.639 0.9402 1.486 A
Granite HU (C) ? UVIGO (A, B, C, D) 6.067 3.5 9 10-17 1.486 R
Granite NTNU (D) ? LNEC (E1, E2) 0.303 0.9984 1.655 A
Granite HU (C) ? LNEC (E1, E2) 3.733 2.8 9 10-6 1.655 R
Granite HU (C) ? NTNU (D) 5.190 1.1 9 10-6 1.839 R
Granite NTNU (D) ? UVIGO (A, B, C, D) ? LNEC (E1, E2) 0.643 0.9677 1.401 A
Granite HU (C) ? NTNU (D) ? UVIGO (A, B, C, D) ? LNEC (E1, E2) 2.824 5.5 9 10-9 1.373 R
A and R indicate that the null hypothesis is accepted or rejected, respectively

123
A Benchmark Experiment to Assess Factors Affecting Tilt Test Results for Sawcut Rock Surfaces

immersion conditions. If humidity is an issue, water pres- planar to make good contact. A straightforward way of
sure distribution across the discontinuity surface and its removing outlier values is to use the median to calculate
lubrication effect may play a role in the apparent friction the final sliding angle value.
angle of planar discontinuities. Based on the above observations and conclusions, it
Although a number of complex phenomena require seems appropriate to try to advance towards standardiza-
further research, the results of this benchmark experiment tion of tilt tests by controlling the most important factors
point to little scattering of sliding angle values obtained in affecting results. This would ensure comparable results and
different laboratories for tilt tests conducted under rela- would facilitate further studies of the nature of sliding
tively controlled conditions. Indeed, the range of scattering along rock cut surfaces, the mechanics of contact and the
is in the same order of magnitude as for other rock relationship with the basic friction angle.
mechanic properties tested in laboratories such as UCS or
tensile strength. Therefore, given simple guidelines and Acknowledgements Co-authors L.R. Alejano and I. Pérez-Rey
acknowledge the Spanish Ministry of the Economy, Industry and
following similar procedures, a reasonable degree of Competitiveness for funding their research studies under Contract
reproducibility can be expected from results from different Reference No. BIA2014-53368P, partially financed by ERDF funds
laboratories. from the EU. The persons working at the laboratories cited in the
paper are also acknowledged for their help in performing the testing.
Ailish M. J. Maher is acknowledged for English edition of a version
of the manuscript.
6 Conclusions

This benchmark study demonstrated that sliding angles for


sawcut rock surfaces determined by tilt tests performed in References
different laboratories tend to be similar when conditions
Alejano LR, González J, Muralha J (2012) Comparison of different
are sufficiently controlled. Moreover, the statistical analy- techniques of tilt testing and basic friction angle variability
ses indicate that sliding angle results for contact surfaces of assessment. Rock Mech Rock Eng 45:1023–1035
the same rock type, cut with different saw blades and tested Aydan Ö (1998) Simple tests on the validation of the effective stress
law for rock discontinuities. In: Thimus et al (ed) Proceedings of
in different tilting devices, are similar, that is, the results
poromechanics, Belgium, Balkema, Roterdam, pp 539–544
can be attributed with a reasonable degree of confidence to Aydan Ö (2015) Lessons from the recent great earthquakes on the
the same original population. performance of buildings under different geotechnical environ-
Surface wear as affected by the cutting procedure, ments. In: Firat S, Aruntas Y, Gultekin AB (eds) Proceedings of
the 2nd international sustainable buildings symposium. Gazi
cleaning practices and rock hardness, does affect results.
Univ.y, Ankara, Turkey, pp 3–20
To mitigate the impact of wear and reflect non-worn Coulson JH (1972) Shear strength of flat surfaces in rock stability of
behaviour in results, it seems wise to carry out a limited rock slopes. In: Cording EJ (ed) Proceedings of Symposium in
number of tests with each pair of rock pieces and to limit Rock Mechanics 13th, Urbana, Ill., 1971. American Society of
Civil Engineers, New York, pp 77–105
sliding of the upper block to 10% of the specimen length.
González J, González-Pastoriza N, Castro U, Alejano LR, Muralha J
Vibrations may prove to be relevant in reducing sliding (2014) Considerations on the laboratory estimate of the basic
angles and may account for HU’s lower values, in view of friction angle of rock joints. In: Alejano LR, Perucho A, Olalla
the acceleration response of 0.05 g measured for its tilting C, Jimenez R (eds) Rock mechanics and rock engineering:
structures on and in rock masses. EUROCK 2014, Vigo.
table. To ensure that vibrations do not affect tilting
Balkema, Rotterdam
equipment and results, it is suggested performing simple Hencher SR (1976) Discussion of the paper ‘‘a simple sliding
acceleration measurements using an easily available apparatus for the measurement of rock joint friction’’ by Cawsey,
smartphone accelerometer application. D. C. and Farrar, N. S. 1976. Géotechnique 26:382–386
Hencher SR (1977) The effect of vibration on the friction angle
The cutting process and the ensuing surface finishing
between planar rock surfaces. Unpublished Ph.D. Thesis,
may also affect interlocking or microtextures of rock sur- Imperial College of Science and Technology, London University
faces and influence sliding angle results. In any case, in this Hencher SR (2012) Discussion of the paper ‘‘comparison of different
benchmark experiment, various laboratories using different techniques of tilt testing and basic friction angle variability
assessment’’. Rock Mech Rock Eng 45:1141–1143
cutting tools produced reasonably similar results. The
Hencher SR, Richards LR (2015) Assessing the shear strength of rock
cutting process appears to affect the scattering of results discontinuities at laboratory and field scales. Rock Mech Rock
(standard deviations) more than average values. Eng 48:883–905
Although outliers were not frequent in this benchmark Kim DH, Gratchev I, Hein M, Balasubramaniam A (2016) The
application of normal stress reduction function in tilt tests for
experiment, they are likely to appear in less controlled test
different block shapes. Rock Mech Rock Eng 49:3041–3054
conditions. It is recommended to remove outliers which Krahn J, Morgenstern NR (1979) The ultimate frictional resistance of
may be associated with rapid polishing or poor matching rock discontinuities. Int J Rock Mech Min Sci Geomech Abstr
(due to poor cutting) of surfaces that are insufficiently 16:127–133

123
L. R. Alejano et al.

Kveldsvik V, Nilsen B, Einstein HH, Nadim F (2008) Alternative discontinuities. In: Schubert W (ed) EUROCK 2015 & 64th
approaches for analyses of a 100,000 m3 rock slide based on geomechanics colloquium, ÖGG, Salzburg, Austria (on CD)
Barton-Bandis shear strength criterion. Landslides 5:161–176 Pérez-Rey I, Alejano LR, Arzúa J, Muralha J (2016) The role of tilting
Li ChC, Zhang N, Ruiz J (2017) Measurement of the basic friction rate and wear surfaces on basic friction angle testing. In: Ulusay
angle of a planar rock discontinuities with three rock cores. Eng R, Aydan Ö, Gercek H, Hindistan MA, Tuncay E (eds)
Geol Environ Bull. doi:10.1007/s10064-017-1045-0 EUROCK2016 symposium, Cappadocia, Turkey, vol 1, pp 29–31
Mehrishall S, Sharifzadeh M, Shahriar K, Song JJ (2016) An Ruı́z J, Li CC (2014) Measurement of the basic friction angle of rock
experimental study on normal stress and shear rate dependency by three different tilt test methods. In Alejano et al (eds) Rock
of basic friction coefficient in dry and wet limestone joints. Rock mechanics and rock engineering: structures on and in rock
Mech Rock Eng 49:4607–4629 masses. In: Alejano LR, Perucho A, Olalla C, Jimenez R (eds)
Muralha J (1996) Probabilistic approach to the mechanical behaviour EUROCK 2014, Vigo. Rotterdam, Balkema (on CD)
of discontinuities in rock masses (in Portuguese). Ph.D. disser- Ulusay R, Karakul H (2016) Assessment of basic friction angles of
tation, Instituto Superior Técnico, Lisbon various rock types from Turkey under dry, wet and submerged
Nicholson GA (1994) A test is worth a thousand guesses—a paradox. conditions and some considerations on tilt testing. Bull Eng Geol
In: Nelson PP, Laubach SE (eds) Proceedings of 1st NARMS Env 75:1683–1699
symposium, Austin, USA, pp 523–529 USBR (2009) Procedure for determining the angle of basic friction
Pérez-Rey I, Alejano LR, González-Pastoriza N, González J, Arzúa J (static) using a tilting table test (Designation USBR 6258-09),
(2015) Effect of time and wear on the basic friction angle of rock www.usbr.gov

123

Potrebbero piacerti anche