SSIM 2023 – PM Dight (ed.

)
© 2023 Australian Centre for Geomechanics, Perth, ISBN 978-0-6450938-6-5
doi:10.36487/ACG_repo/2335_20

Multistage triaxial testing of intact rock: volumetric

strain-based methods applied to rock slope design

K Koosmen PSM, Australia

M Serati The University of Queensland, Australia
B Craig Glencore, Australia

Abstract
Triaxial testing of intact rock is commonly undertaken to provide input parameters for rock slope design.
Multistage triaxial tests offer economies in terms of time, sample quantities and costs, and for these reasons
have found commonplace in engineering practice. Volumetric strain-based methods are sometimes employed
for multistage testing to minimise damage accumulation in early test stages. This is done by (1) terminating
early test stages at volumetric strain reversal, (2) straining the final stage until failure occurs, then (3) inferring
a peak stress for the earlier stages based on a correction factor derived from the final stage.
This paper provides a review of multistage triaxial testing along with fundamental rock mechanics theories
that underpin the volumetric strain-based multistage testing method. Results from single and multistage tests
are then used to demonstrate possible errors that may result if multistage tests are conducted over a range
of stresses where compressive failure of intact rock is controlled by different failure mechanisms. Finally, some
examples and discussion are provided to demonstrate the possible impact that these errors may have on rock
slope stability assessments.
Keywords: intact rock strength, triaxial testing, multistage methods, rock fracture mechanics

1 Introduction

1.1 Context
Triaxial testing of intact rock is used to provide input parameters for the design of rock structures. For open
pit slope design the intact rock strengths are primarily used as input parameters to estimate rock mass
strengths using continuum-based methods such as the generalised Hoek–Brown criterion (Hoek & Brown
2018). Less commonly, the intact strength may also be used to estimate the strength of discrete blocks of
intact rock in discontinuum-based models. Other parameters such as residual strengths, modulus and
Poisson’s ratio may also be of interest, depending on the application.
Single-stage triaxial tests provide a measure of the intact strength by applying a lateral confining stress (σ3)
to a core specimen before increasing the axial stress (σ1) until failure occurs. The σ1 value at failure is then
taken as the peak intact strength (σp). This process destroys the core specimen and, consequently, each test
only provides a single measurement of peak strength for the nominated confining stress.
Multistage techniques are fundamentally similar to single-stage techniques except the individual test stages
are terminated before failure is induced. This preserves the specimens so that testing can be undertaken
across a range of confining pressures. Benefits of this approach are:
• Multistage tests are cheaper, faster and require less samples to provide the same number of data
points when compared to single-stage tests.
• A failure envelope can be defined from a single specimen when using multistage techniques.

SSIM 2023, Perth, Australia 343

Multistage triaxial testing of intact rock: volumetric strain based methods K Koosmen et al.
applied to rock slope design

• Conducting several multistage tests can provide several failure envelopes to better understand
strength and parameter variability.
Multistage tests have historically employed ‘imminent failure’ methods based on the early work of Kovari
& Tisa (1975). This involves monitoring the stress-strain curve throughout the test to infer the point when
failure is imminent so the stage can be terminated to preserve the specimen for testing in later stages (see
stress paths on the left of Figure 1). Use of imminent failure methods is often met with scepticism as selecting
the point of imminent failure is subject to human error, and the growth of microcracks during earlier test
stages may induce some damage in the specimen for later test stages. Both factors may lead to an
underestimation of the peak strengths during later test stages.
An alternative method which seeks to overcome these issues is to terminate the early test stages at
volumetric strain reversal (VSR) where the stress (σVSR), is significantly lower than the peak stress at failure.
The VSR criterion allows for automation of stage termination while also limiting microcrack growth by not
testing to ‘imminent failure’. Failure is then induced during the final test stage where a stress ratio (Orilogi
2019; Venter et al. 2019) or stress magnitude (Pagoulatos 2004) is calculated, then applied to data from
earlier tests stages to infer peak strengths at the lower confining stresses. Stres paths for this method are
shown on the right of Figure 1, with equations to calculate peaks strengths shown in Equations 1 and 2.
VSR stress ratio
Ratio corrected . . (1)
. . from final stage
VSR stress magnitude
Magnitude corrected . . . . (2)
from final stage
where:
σp.nf = calculated peak strength in non-final test stage.
σp.f = measured peak strength in final test stage.
σVSR.nf = stress at VSR measured in non-final test stage.
σVSR.f = stress at VSR measured in final test stage.

σp.f

σVSR.f

(a) (b)

Figure 1 Stress paths for multistage triaxial test. (a) Imminent failure method; (b) Volumetric strain-based
method

To date, the literature and experience with volumetric strain-based methods are limited compared to
imminent failure methods. Furthermore, a fundamental assumption of the volumetric strain-based method
is that the correction factors derived from the final test stage are applicable to all earlier test stages despite
the changes in confining stress and the fracture mechanics which govern the brittle failure process.

344 SSIM 2023, Perth, Australia

 Constructing models

1.2 Purpose
The purpose of this paper is to provide a review of volumetric strain-based methods for multistage triaxial
testing to assess their suitability for use in rock slope stability assessments. This is done by comparing results
from multistage triaxial tests with results from single-stage triaxial tests to understand what type of errors
may be incurred due to the multistage test methodology. The experimental data is then used to draw
conclusions on the validity of applying a single correction factor over a wide range of confining stresses to
calculate peak strength in the non-final test stages. Lastly, two examples are provided to illustrate how errors
associated with the peak intact strength may impact on slope stability assessments for open pit slope designs.

2 Rock mechanics theory

2.1 Hoek–Brown criterion

The Hoek–Brown failure criterion provides a non-linear relationship for peak strength of intact rock subject
to brittle failure processes. The criterion was developed by Hoek & Brown (1980a; 1980b) using Griffith’s
criterion as a theoretical basis, then applying trial and error to empirically fit a parabolic envelope to a wide
range of intact rock testing data (Hoek et al. 1995). The criterion is written in the principal stress space as
shown in Equation 3:
$
Hoek–Brown criterion n "# (3)
!

where:
σ1 & σ 3 = major and minor principal stresses.
σci = is the unconfined compressive strength.
mi = fitting parameter, usually varying between 5 and 40 depending on rock type.
s and a = fitting parameters, usually taken as 1 and 0.5, respectively, for intact rock.
Curve fitting of triaxial data can be undertaken by manually modifying σci and mi until a suitable fit is obtained
with the experimental data, or more rigorous regression analyses can be used such as those listed by Mostyn
& Douglas (2000). Figure 2 illustrates how the Y-intercept of the Hoek–Brown envelope is controlled by σci
while the gradient is controlled by mi. In this respect, σci and mi are analogous to the cohesion and friction
angle of the linear Mohr–Coulomb criterion (Hoek 1983). It should also be noted that the Hoek–Brown
criterion is only applicable to brittle, and not ductile, failure of intact rock. The transition from brittle to
ductile failure can be approximated by Mogi’s line (Mogi 1966; σ1 = 3.4σ3).

Figure 2 Effect of input parameters on the slope and intercept of the Hoek–Brown criterion

SSIM 2023, Perth, Australia 345

Multistage triaxial testing of intact rock: volumetric strain based methods K Koosmen et al.
applied to rock slope design

2.2 Microcracking of intact rock during compression

Modern concepts of brittle rock fracture are founded in the works of Griffith (1920, 1924) who studied the
fracture mechanics of glass windshields. Griffith proposed that (1) brittle solids contain randomly orientated
microcracks (named Griffith cracks), and (2) that tensile fractures (named wing cracks) nucleate from the tip
of Griffith cracks in a direction that is parallel to the major principal stress axis. For the case of intact rock,
the wing cracks may nucleate from pre-existing microcracks in the rock, or along other flaws such as grain
boundaries, cleavage within minerals, stiff or soft inclusions, and low-aspect ratio cavities (Horii
& Nemat-Nasser 1985; Tang & Hudson 2011; Kaiser & Kim 2014). Griffith’s original theoretical criteria are
now recognised to be of limited practical value in rock engineering (Hoek & Martin 2014), however, the
mechanism of tensile wing cracks propagating from a flaw in the microstructure remains a fundamental
concept in the brittle rock failure.
Under low confining stress the wing cracks propagate extensively in a direction parallel to the major principal
stress. This leads to axial splitting failures, for example, as commonly observed in the laboratory during
uniaxial compressive strength (UCS) testing, or in the field as rockburst or spalling at the boundaries of
underground excavations. Under higher confining stresses the wing cracks may still develop but their growth
is supressed in the axial direction due to the effects of confinement (Einstein & Dershowitz 1990; Gupta
& Bergstrom 1998; Horii & Nemat-Nasser 1985; Kaiser & Kim 2014). This supresses the growth of individual
wing cracks but facilitates the growth of wing cracks from neighbouring flaws which form clusters and
preferentially align in an en echelon arrangement to form a shear zone, leading to ultimate shear failure
(Gupta & Bergstrom 1998). Shear failure is therefore the dominant mode of failure during triaxial
compression at higher confining stresses. The reader is referred to Hoek & Martin (2014) for further
discussion of these concepts.

2.3 Crack thresholds for intact rock

The brittle fracture of intact rock occurs from the development, growth and coalescence of microcracks.
Based on the works of numerous authors (Hoek 1965; Brace et al. 1966; Bieniawski 1967; Lajtai 1974;
Tapponier & Brace 1976; Gramberg 1989; Lockner et al. 1992; Martin 1993; Martin & Chandler 1994; Martin
1997; Eberhardt et al. 1998), the microcrack development is characterised by three distinct stress thresholds
which divide the stress-strain curve into four distinct zones. The stress thresholds are (see Figure 3):
• Crack closure (σcc) – the stress when most of the pre-existing microcracks are closed. The
stress-strain curve is non-linear leading up to crack closure and linear-elastic immediately following.
• Crack initiation (σci) – the stress where small and randomly distributed microcracks begin to develop
throughout the rock, corresponding to the end of linear-elastic behaviour.
• Crack damage (σcd) – defining the onset of uncontrolled crack growth when individual cracks
lengthen and/or coalesce with other cracks. Macro-scale failure then occurs by axial splitting or
shear when the critical crack density is exceeded.
The σci threshold depends on rock type and which method is used to determine or infer σci, however, typical
values are 30 to 60% of the peak strength during unconfined compression, and 40 to 80% of the peak strength
during confined compression (Nicksiar & Martin 2013; Mutaz et al. 2021).
The σcd threshold also varies for different rock types and the determination method, and corresponds with
the point of VSR on the stress-strain curve as seen Figure 3 (Brace et al. 1966; Bieniawski 1967; Martin 1993).
Typical values of σcd are between 65 to 90% of the peak strength during unconfined compression and 60 to
100% of the peak strength in confined compression (Diederichs et al. 2009; Taheri et al. 2020).

346 SSIM 2023, Perth, Australia

 Constructing models

Figure 3 Idealised stress-strain curve illustrating the stages of microcracking and stress thresholds

3 Laboratory testing

3.1 Methodology
Single and multistage triaxial tests were conducted under varying conditions so that the single-stage tests
could be used as a control group for validation of the multistage tests. All tests were conducted on sandstone
specimens that were collected from a coal deposit in Northern Queensland, Australia. All samples were
collected from the same unit within the stratigraphy and with the same lithology (sandstone), with a
field-estimated strength of R2 to R3 and with a density in the range of 2.54 to 2.61 t/m3. The triaxial testing
was conducted using a stiff-framed GCTS RTR2500 triaxial machine fitted with a closed-loop servo-controlled
loading system. This allowed for automation of the testing procedure according to instantaneous
stress-strain measurements. Linear variable differential transformers were used to measure the
circumferential and axial strains during each test.
A summary of the testing schedule is provided in Table 1, with additional details as follows:
• The dataset for the single-stage triaxial tests (Group 2) included data points from 33 single-stage
tests plus nine data points from the final stage of the multistage tests. These multistage results
were included in the single-stage dataset assuming minimal ‘damage’ accumulation in early stages
of the multistage tests. Therefore, the final stages are effectively the same as a single-stage test.
• The multistage test results were conducted across two stress ranges, where the Group 3 tests were
conducted at lower confining stresses of 2 to 6 MPa, and the Group 4 tests were conducted at
higher confining stresses of 6 to 24 MPa.
• All multistage test stages were terminated at VSR in the non-final stages, then tested to failure in
the final test stage (i.e. per the stress paths in Figure 1a).
• The peak strength for non-final stages of the multistage tests were then calculated using both the
ratio and magnitude correction factors as shown by Equations 1 and 2.

SSIM 2023, Perth, Australia 347

Multistage triaxial testing of intact rock: volumetric strain based methods K Koosmen et al.
applied to rock slope design

Table 1 Schedule of triaxial testing

Group Test type Confining stress (MPa) # of data points

1 Uniaxial compressive strength 0 29
2 Single-stage triaxial 0.5, 2, 6, 10, 15, 24 33 + 9 [2] = 42
3 Multistage triaxial [1] 2, 4, 6 (low stress range) 15 (= 5 tests × 3 stages)
4 Multistage triaxial [2] 6, 12, 18, 24 (higher stress range) 16 (= 4 tests × 4 stages)
[1] Using volumetric strain reversal as stage termination criteria in the non-final stages
[2] The single-stage dataset includes nine results from the final stage of each multistage test. See further discussion in text.

3.2 Single-stage test results

Results from the single-stage triaxial tests (Group 2) are plotted in Figure 4 where a Hoek–Brown envelope is
fitted to the average stress points, giving σci of 39 MPa and mi of 9. The σci value of 39 MPa from curve fitting
compares very well with the median UCS value of 38 MPa. The VSR stress ratios and VSR stress magnitudes
are included in Figure 5 and discussed in more detail in the following sections.

(a) (b)

Figure 4 (a) Cumulative distribution from uniaxial compressive strength (UCS) testing; (b) peak strength
envelope based on UCS and single-stage triaxial testing

(a) (b)

Figure 5 Results from single-stage triaxial testing showing (a) Volumetric strain reversal (VSR) stress ratio;
(b) VSR stress magnitude

348 SSIM 2023, Perth, Australia

 Constructing models

3.3 Multistage test results

Hoek–Brown failure envelopes were determined from the multistage tests and are shown in Figure 6 for the
ratio corrected peak strengths, and in Figure 7 for magnitude corrected peak strengths. The Hoek–Brown
intact strength parameters (σci and mi) for all failure envelopes are summarised in Table 2.
Across both datasets the average strength that was measured in the test final stage (6 and 24 MPa) is
comparable with the strength envelope defined from the single-stage tests. This indicates that the multistage
specimens have a similar average strength to the single-stage specimens, and that any significant deviations
between the multistage and single-stage failure envelopes are therefore not caused by strength variability.
When the ratio correction factor is used for the low stress tests (Figure 6a) it is seen that there is an increase
in the apparent error as the confining stresses reduce. Similar trends are also observed for the higher stress
tests (Figure 6b), although the apparent error is less pronounced than what is seen for low stress tests in
Figure 6a.
In both datasets the apparent error is attributed to variability of the VSR stress ratio across different confining
stresses, which is evident from the single-stage test data that is shown in Figure 5a. If a higher ratio is derived
in the final stage, then applied at lower confining stress where a lower ratio is appliable, the net effect will
be that the calculated peak strengths underestimate the true peak strength at lower confining stresses. As
an example, consider the data from the single-stage tests where average ratios of 65% and 80% are measured
at 2 and 6 MPa. If peak strengths at 2 MPa are calculated using Equation 1 and the 80% ratio then the peak
strength will be 1.25 x σVSR. If the ‘correct’ ratio of 65% is considered for 2 MPa then the peak strength will
be equal to 1.54 x σVSR. The apparent errors in the multistage results from Figure 6 are therefore caused by
assuming that the ratio correction factors derived from the final test stages are constant across all stages
where, in fact, the ratio is variable across a range of confining pressures. At this stage it is hypothesised that
the change in ratio with confining pressure may be related to a change in failure mechanism from tensile to
shear-dominated failure processes for intact rock, although further work is required to confirm this theory.

(a) Ratio correction factor, (b) Ratio correction factor,

low stress tests high stress tests

Figure 6 Multistage test results where the ratio correction factor was used to calculate peak strengths.
(a) Lower stress tests with σ3 = 2–6 MPa; (b) Higher stress tests with σ3 = 6–24 MPa

The magnitude corrected multistage results are shown in Figure 7 and give a better comparison with the
single-stage results than what was seen for the ratio corrected results in Figure 6. Tests conducted at lower
confining pressures (Figure 7a) still show considerable variability from the single-stage envelope, while tests
at higher confining pressures (Figure 7b) align very closely with the single-stage envelope. These outcomes
are explained using the results from the single-stage triaxial tests in Figure 5b, where the average stress
magnitude values for this rock type show minimal variability across a wide range of confining pressures and
with the greatest variation in the low stress range.

SSIM 2023, Perth, Australia 349

Multistage triaxial testing of intact rock: volumetric strain based methods K Koosmen et al.
applied to rock slope design

(a) Magnitude correction (b) Magnitude correction

factor, low stress tests factor, high stress tests

Figure 7 Multistage results where the magnitude correction factor was used to calculate peak strengths.
(a) Lower stress tests with σ3 = 2–6 MPa; (b) Higher stress tests with σ3 = 6–24 MPa

Table 2 Hoek–Brown parameters determined from single-stage and multistage triaxial testing

Test type Confining stress Peaks strengths calculated from σci (MPa) mi See figure
Single-stage σ3 = 0–24 MPa – 39 9 4b
Multistage σ3 = 2–6 MPa VSR stress ratio 6 81 6a
Multistage σ3 = 6–24 MPa VSR stress ratio 21 16 6b
Multistage σ3 = 2–6 MPa VSR stress magnitude 25 17 7a
Multistage σ3 = 6–24 MPa VSR stress magnitude 38 8 7b

3.4 Main findings from laboratory testing

The main findings from the laboratory testing are:
1. The VSR stress ratio (σVSR/σP) and VSR stress magnitude (σP – σVSR) can vary with confining pressure.
The maximum variability appears to occur in the low stress range and may be related to a transition
between tensile and shear-dominated fracture processes.
2. Use of a ratio or magnitude correction factor calculated from the final stage of a multistage test
may introduce errors if used to calculate peak strengths in the non-final stages of a multistage test.
3. For the sandstone material considered in the experiments above, it was found that:
a. The magnitude correction was more accurate than the ratio correction factor for calculating
peak strengths from multistage tests when compared to the measured peak strengths from
single-stage testing.
b. Peak strengths calculated from the multistage tests using the ratio and magnitude correction
factors incurred less error when tests were conducted at higher confining stresses.

4 Other rock types

The laboratory testing and results discussed in Section 3 have only considered a single rock type. Additional
data was therefore sourced from the literature, then analysed to provide a better understanding on how the
ratio and magnitude correction factors perform for predicting peak strengths across different rock types. The
process was as follows:

350 SSIM 2023, Perth, Australia

 Constructing models

1. Single-stage triaxial test data was sourced for various rock types where σVSR and peak strength were
reported across a wide range of confining pressures. Average data points were calculated where
datasets contain several data points at a single confining pressure.
2. A ratio correction factor and magnitude correction factor were determined from the test or data
point that corresponds to the maximum confining pressure.
3. The ratio and magnitude correction factors were then applied to σVSR recorded for all other stage
tests to calculate peak strengths at lower confining pressures.
4. Hoek–Brown envelopes were then fitted to the calculated peak strengths and measured peak
strengths to enable a comparison of the datasets.
The peak strength envelopes and intact parameters (σci and mi) are shown in Figure 8 for non-sedimentary
rocks and Figure 9 for sedimentary rocks. A summary of how the calculated peak strength envelopes compare
with the measured peak strength envelopes is provided in Table 3. A good fit is obtained where the calculated
peak strength envelope compares well with the measured peak strength envelope, and where the maximum
error is less than a nominal value of around 15%. Where these criteria are satisfied it is implied that the ratio
or magnitude correction factor is reasonably constant across the applied stress range.

Table 3 Comparison of calculated peak strength envelopes with measured peak strength envelopes

Maximum error [1] Goodness of fit between failure envelopes [2]

Error for Error for Reasonable Reasonable Reasonable Poor fit
See Rock ratio magnitude fit for ratio fit only for fit only for for ratio
Figure type corrected corrected and ratio magnitude and
failure failure magnitude corrected corrected magnitude
envelope envelope corrected failure failure corrected
envelopes envelope envelope envelopes
8a B. Granite -24% 39% ✓
8b LdB Granite 15% 74% ✓
8c Marble -9% 33% ✓
8d Basalt -13% -2% ✓
9a Sandstone -46% -5% ✓
9b Sandstone -55% 51% ✓
9c B4 Shale 28% 41% ✓
9d Shale 8% 27% ✓
[1] Max error is obtained in all cases where σ3=0 MPa, and is the percentage difference between the σci value from the calculated and measured
peak strength envelopes. Negative/positive errors indicate an under/overestimate of the measured peak strength.
[2] Refers to a comparison between the measured peak strength envelope from the single-stage tests and the calculated peak strength envelope/s
from the multistage tests.

The observed trends for the different rock types are as follows:
• Variable results are observed where calculated failure envelopes compare well with the measured
failure envelopes for some rock types, but not for others.
• The ratio corrected peak strength envelopes provide a reasonable fit to the measured peak strength
envelope in four out of the eight cases (Beishan granite, marble, basalt and shale). Otherwise there
is generally an underestimate of peak strength (LdB granite, and two sandstones). There is only one
instance where the ratio corrected envelope significantly overestimates the measured peak
strength envelope (b4 shale).

SSIM 2023, Perth, Australia 351

Multistage triaxial testing of intact rock: volumetric strain based methods K Koosmen et al.
applied to rock slope design

• The magnitude corrected peak strengths tended to overestimate the measured peak strengths in
most cases, with a maximum error of up to 74% for Beishan granite (Figure 8b). There were only
two of eight instances where the magnitude corrected envelope provided a reasonable fit to the
measured peak strength envelope (basalt, Figure 8d; and Berea sandstone, Figure 9a).
• It is concluded that neither of the methods are universally reliable for calculating peak strengths
across the range of rock types that are considered.
• Where large errors are observed, this outcome is attributed to the ratio correction factor or
magnitude correction factor not being constant over the applied range of confining pressures.

(a) (b)

(c)

(d)

Figure 8 Results from single-stage triaxial tests sourced from the literature for non-sedimentary rocks
where peak strengths have been calculated using the ratio and magnitude correction factors,
then compared with measured peak strengths

352 SSIM 2023, Perth, Australia

 Constructing models

(a) (b)

(c) (d)

Figure 9 Results from single-stage triaxial tests sourced from the literature for sedimentary rocks where
peak strengths have been calculated using the ratio and magnitude correction factors, then
compared with measured peak strengths

5 Implications for rock slope stability assessments

Data from Sections 3 and 4 have demonstrated that errors of varying degree can be incurred when using
volumetric strain-based methods to calculate peak intact rock strengths from multistage triaxial tests. The
magnitude of the error depends on (1) the rock type, (2) the range of confining stress over which the tests
are conducted and (3) whether the ratio correction or magnitude correction factors are applied to calculate
the peak strengths. But what do these errors mean with respect to rock slope stability estimates? Are the
impacts significant or trivial?
Generally speaking, the most significant impacts can be expected for failure mechanisms where rock mass
strengths significantly contribute to the stresses that resist slope failure. Two simplified cases are used to
illustrate this where 2D limit equilibrium techniques have been used to estimate to the Factor of Safety (FOS).

SSIM 2023, Perth, Australia 353

Multistage triaxial testing of intact rock: volumetric strain based methods K Koosmen et al.
applied to rock slope design

Case 1 considers a highwall slope where the failure mechanism is an active-passive wedge with basal sliding
along a weak horizontal plane and with the back scarp extending through the rock mass. Case 2 considers a
footwall slope with sliding along a weak inclined plane with breakout through the rock mass in the toe region.
Model geometries are shown in Figure 10, with additional assumptions as follows:
• Mohr–Coulomb strength parameters for the weak structure were c’ = 0 kPa and ɸ’ = 15°.
• Pore pressures were estimated using a steady state flownet.
• Rock mass strengths were based on the generalised Hoek–Brown criterion (Hoek & Brown 2018).
• The geological strength index (GSI) for the rock mass was 65.
• The blast disturbance factor (D) was zero.
• Different combinations of intact strength parameters (σci and mi) are adopted based on the
combinations derived from the single-stage and multistage triaxial testing described in Section 3.

Slope A
Highwall

240 m

Rock mass
in back scarp

Weak plane at base

Slope B
Benched Footwall 150 m

Weak plane
along base

Rock mass through

toe breakout zone

Figure 10 Highwall and footwall slopes with failure mechanisms considered for stability analysis. Contours
show pore pressures estimated from flownet via finite element analysis

The intact strength parameters used for rock mass strength estimates are summarised in Table 4 along with
the FOS values that were calculated for each case. For both slopes types the FOS is calculated to be around
1.5 where the single-stage triaxial test results are used as the basis for rock mass strength estimates (scenario
1). The single-stage tests are considered to provide the most reliable estimate of intact strength. Therefore,
the scenario 1 results are used as the basis to calculate the error for the other scenarios.

354 SSIM 2023, Perth, Australia

 Constructing models

Table 4 Intact rock strengths used to define rock mass strengths with calculated FOS values for the two
different slope configurations

Triaxial test data from Section 3 used to define rock mass strengths Slope A – Slope B –
Scenario Test Correction method used σ3 range σci mi highwall footwall
type to calculate peak strength (MPa) (MPa) FOS [1] FOS [1]

1 SS TXL N/A (measured directly) 0.5 – 24 39 9 1.48 1.54

2 MS TXL Stress ratio 0–6 6 81 1.22 (18%) 1.09 (29%)
3 MS TXL Stress ratio 6 – 24 21 16 1.32 (11%) 1.23 (20%)
4 MS TXL Stress magnitude 0–6 25 17 1.38 (7%) 1.30 (16%)
5 MS TXL Stress magnitude 6 – 24 38 8 1.46 (1%) 1.53 (1%)
[1] Value in brackets shows relative ‘error’ compared to the FOS that is calculated for scenario 1.

When the rock mass strengths are estimated using the intact strength parameters from the multistage tests
(scenarios 2 to 5), it is seen that the FOS values are between 1% and 29% lower than what was calculated for
scenario 1 based on the single-stage tests. Not surprisingly, the largest errors for both slopes were observed
for scenario 2, where the multistage test results provided the largest underestimate of the peak intact
strengths (Figure 6a). Furthermore, it was also seen that the footwall slope recorded larger errors than the
highwall slope. This is because the strength errors associated with multistage tests tend to compound at
lower confining stresses, and the average stress is lower in the rock mass component in the footwall slope
than in the highwall slope. The error of 1% that is obtained for both slopes for scenario 5 is also to be
expected, given that σci and mi are nearly identical to values that are used for scenario 1. However, and as
illustrated in previous sections, it may not always be the case (and more likely is seldom the case) that there
is such a close match between the single-stage and multistage test results. It is only with the benefit of
single-stage testing that the results from scenario 5 are able to be validated, and without the single-stage
tests it could not be confidently predicted whether scenario 2, 3, 4 or 5 is the valid case.
A final point to note is regarding the effects of GSI on rock mass strength estimates. The assessments above
have utilised a GSI of 65 which corresponds to a reasonably good rock mass. When utilising the generalised
Hoek–Brown criterion to estimate rock mass strengths it is found that rock mass strengths become less
sensitive to the intact strength parameters as the GSI decreases. This is shown by Figure 11 where the intact
strength parameters from scenarios 1 to 5 are used to calculate rock mass strengths for GSI values between
40 and 70. Maximum variability between the rock mass strength envelopes is observed when the GSI is 70
and the failure envelopes converge as the GSI reduces, despite the variable intact strength parameters.
The implications of this were tested using the footwall scenario from Table 4 (case B) and reducing the GSI
from 65 to 42 for the rock mass strength estimates. For scenario 1 where the intact strength parameters are
calculated from the single-stage triaxial tests, the FOS value was calculated to be 0.98. For scenarios 2 to 5,
the FOS values ranged from 0.92 to 0.98 with a maximum error of 6% (relative to the FOS for scenario 1). This
is a significant reduction in the maximum error of 29% that was recorded with an assumed GSI of 65, even
though the same intact strength parameters are used.
The analyses presented above have demonstrated that:
• Errors with intact strength estimates resulting from multistage test methods may have significant
impacts on the reliability of slope stability assessments.
• The impact of these errors is more pronounced for higher quality rock mass with higher GSI values
if the generalised Hoek–Brown criterion is used to estimate rock mass strengths.

SSIM 2023, Perth, Australia 355

Multistage triaxial testing of intact rock: volumetric strain based methods K Koosmen et al.
applied to rock slope design

Figure 11 Rock mass failure envelopes calculated using the generalised Hoek–Brown criterion, with the
same five sets of intact strength parameters (σci and mi) and considering different GSI values

6 Conclusion
The conclusions from this study are as follows:
• Single-stage triaxial tests provide the most reliable method for determining intact rock strength.
While multistage techniques may be preferred in some instances due to constraints imposed by
testing budgets and sample quantities, their use will incur a reduction in the reliability of the results.
• Volumetric strain-based techniques for multistage testing provide an alternative to the more
conventional imminent failure techniques. These tests terminate the non-final test stages at VSR
then calculate the peak strengths in these stages based on a VSR stress ratio or VSR stress
magnitude correction factor which is determined from the final test stage.
• A fundamental assumption when applying these correction factors to calculate peak strength in the
non-final test stages is that the correction factor is constant across the applied range of confining
stresses.
• Results from this study have demonstrated that the correction factors may not be constant across
a wide range of confining stresses for different rock types. The maximum rate of variability appears
to occur in the low stress range and may be related to a transition between tensile and
shear-dominated fracture processes.
• The experimental data reviewed in this study indicates that use of the VSR stress ratio or VSR stress
magnitude as a correction factor can provide a reliable estimate of peak strength for some rocks,

356 SSIM 2023, Perth, Australia

 Constructing models

but not for others. Moreover, it was demonstrated for some rock types that neither of the methods
provide a reliable strength estimate. The maximum errors that are observed between the
calculated and peak strengths tend to compound for lower confining stresses.
• There is no clear trend from the current results to provide an indication in advance on which
correction factor may give a reliable results for a particular rock type. For example, it is not the case
that use of the VSR stress ratio will always give reliable results for sandstone, igneous rocks or some
other rock type, and likewise with VSR stress magnitude. The only way to confirm the reliability of
the multistage methods is by undertaking parallel single-stage tests to enable a direct comparison
of peak strengths from the single and multistage datasets.
• Intact rock strengths are commonly used to estimate rock mass strengths for assessing slope
stability where key failure mechanisms include some element of failure through the rock mass. In
these cases, significant errors with the calculated intact strengths from multistage tests may then
introduce significant errors in slope stability estimates. All else being equal, the impact of intact
strength errors on slope stability estimates will be more pronounced for higher quality rock masses
with higher GSI values if the generalised Hoek–Brown criterion is used to estimate rock mass
strengths. This is because rock mass strength estimates based on the Hoek–Brown criterion are
more sensitive to variation of intact rock strength for higher GSI values.
• If volumetric strain-based techniques are to be used for multistage testing then it is recommended
that (1) a series of UCS tests be included in the dataset to help constrain σci on the Y-intercept,
(2) that variable confining stresses are utilised for the final stage of different tests, and (3) that
larger weightings are given to the final test stages where peak strength is directly measured and
not calculated using a correction factor.
• This paper has demonstrated that large errors of more than 50% may occur when estimating peak
intact rock strengths using volumetric strain-based techniques for multistage triaxial testing (see
Table 3). Volumetric strain-based techniques are usually employed to minimise the perceived errors
associated with the more conventional imminent-failure-based techniques. Given potential errors
of more than 50%, the question must be asked: Are the volumetric strain-based techniques more
or less reliable than the conventional imminent failure techniques? Further work is required to
provide a broad assessment on the reliability of imminent failure techniques, however, it is the
authors’ opinion that maximum potential errors would probably be less than 50% with reasonable
testing procedures. Undertaking a large suite of single-stage testing would help to eliminate the
potential for errors associated with either of the multistage techniques.

References
Brace WF, Paulding, BW & Scholz, C 1966, ‘Dilatancy in the fracture of crystalline rocks’, Journal of Geophysical Research, vol. 71,
no. 16, pp. 3939–3953.
Bieniawski, ZT 1978 ‘Determining rock mass deformability: experience from case histories’, International Journal of Rock Mechanics
and Mining Sciences and Geomechanics Abstracts, vol. 15, no.5, pp 237–247.
Diederichs, MS, Carter, T & Martin, D 2009, ‘Practical rock spall prediction in tunnels’, in Proceedings of the International Tunnelling
Association World Tunnel Congress, pp. 1–8.
Eberhardt, E, Stead, D, Stimpson, B & Read, RS 1998, ‘Identifying crack initiation and propagation thresholds in brittle rock’, Canadian
Geotechnical Journal, vol. 35, no. 2, pp. 222–233.
Einstein, EW & Dershowitz, WS 1990, ‘Tensile and shear fracturing in predominantly compressive stress fields – a review’, Engineering
Geology, vol. 29, pp 149–172.
Gramberg, J 1989, A Non-Conventional Review on Rock Mechanics and Fracture Mechanics, CRC Press, Boca Raton.
Griffith, AA 1920, ‘The phenomena of rupture and flow in solids’, The Philosophical Transactions of the Royal Society of London,
vol. 221, pp. 163–198.
Griffith, AA 1924, ‘The theory of rupture’, in CB Biezeno & JM Burgers (eds), Proceedings of the First International Congress for Applied
Mechanics, Delft Tech, pp 55–63.
Gupta, V & Bergstrom, JS 1998, ‘Compressive failure of rocks by shear faulting’, Journal of Geophysical Research: Solid Earth, vol. 110,
no. B10, pp. 23875–23895 .

SSIM 2023, Perth, Australia 357

Multistage triaxial testing of intact rock: volumetric strain based methods K Koosmen et al.
applied to rock slope design

Hoek, E 1965, Rock Fracture Under Static Stress Conditions, report for Council for Scientific and Industrial Research, MEG 383,
Pretoria.
Hoek, E & Brown, ET 1980a, ‘Empirical strength criterion for rock masses’, Journal of the Geotechnical Engineering Division, ASCE,
vol. 106, no. 9, pp 1013–1035.
Hoek, E & Brown, ET 1980b, Underground Excavations in Rock, Institute of Mining and Metallurgy, London.
Hoek, E 1983, ‘The strength of jointed rock masses’, Geotechnique, vol. 33, no. 3, pp. 187–222.
Hoek, E, Kaiser, PK & Bawden, WF 1995, Support of Underground Excavation in Hard Rock, A.A. Balkema, Rotterdam.
Hoek, E & Brown, ET 2018, ‘The Hoek-Brown failure criterion and GSI – 2018 edition’, Journal of Rock Mechanics and Geotechnical
Engineering, vol. 11, is. 3, pp.445–463.
Hoek, E & Martin, CD 2014, ‘Fracture initiation and propagation in intact rock – a review’, Journal of Rock Mechanics and Geotechnical
Engineering, vol. 6, no. 4, pp. 278–300.
Horii, H & Nemat-Nasser, S 1985, ‘Compression-induced microcrack growth in brittle solids - axial splitting and shear failure’, Journal
of Geophysical Research, vol. 90, no. B4, pp. 3105–3125.
Kaiser, PK & Kim, BH 2014, ‘Characterization of strength of intact brittle rock considering confinement-dependent failure processes’,
Rock Mechanics and Rock Engineering, vol. 48, pp. 107–119.
Kovari, K & Tisa, A 1975, ‘Multiple failure state and strain controlled triaxial tests’, Rock Mechanics, vol. 7, pp. 17–33.
Lajtai, EZ 1974, ‘Brittle fracture in compression’, International Journal of Fracture, vol. 10, pp. 525–536.
Lockner, DA, Moore, DE & Reches, Z 1992, ‘Microcrack interaction leading to shear failure’, paper presented at the 33rd U.S.
Symposium on Rock Mechanics, Santa Fe, New Mexico.
Martin, CD 1993, Strength of Massive Lac du Bonnet Granite Around Underground Openings, PhD thesis, University of Manitoba,
Winnipeg.
Martin, CD & Chandler, NA 1994, ‘The progressive fracture of Lac du Bonnet granite’, International Journal of Rock Mechanics and
Mining Sciences and Geomechanics Abstracts, vol. 31, no. 6, pp. 643–659.
Martin, CD 1997, ‘The effect of cohesion loss and stress path on brittle rock strength’, Canadian Geotechnical Journal, vol. 34, no. 5,
pp.698–725.
Mogi, K 1966, ‘Pressure dependence of rock strength and transition from brittle fracture to ductile flow’, Bulletin Earthquake Research
Institute, vol. 44, no 215–232.
Mostyn, GR & Douglas, KJ 2000, ‘Strength of intact rock and rock masses’, paper presented at the GeoEng2000 Conference,
Melbourne.
Mutaz, E, Serati, M, Bahaaddini, M & Williams, B 2021, ‘On the evolution of crack initiation stress threshold’, paper presented at The
55th U.S. Rock Mechanics/Geomechanics Symposium, American Rock Mechanics Association, Alexandria.
Nicksiar, M & Martin, CD 2013, ‘Crack initiation stress in low porosity crystalline and sedimentary rocks’, Engineering Geology,
vol. 154, pp. 64–76.
Orilogi, E 2019, Evaluating Volumetric Strain as a Predictor of Yield and Peak Strength for the Multistage Triaxial Test: A Case Study
with Utah Coal Specimens, MSc thesis, Montana Technological University, Butte.
Pagoulatos, A 2004, Evaluation of Multistage Triaxial Testing on Berea Sandstone, MSc thesis, The University of Oklahoma, Norman.
Taheri, A, Zhang, Y & Munoz, H 2020, ‘Performance of rock crack stress thresholds determination criteria and investigating strength
and confining pressure effects’, Construction and Building Materials, vol. 243.
Tang, C & Hudson, JA 2011, Rock Failure Mechanisms, Taylor & Francis, London.
Tapponnier, P & Brace, WF 1976, ‘Development of stress-induced microcracks in Westerly granite’, International Journal of Rock
Mechanics and Mining Sciences and Geomechanics Abstracts, vol. 13, pp. 103–112.
Venter, J, Hammah, ECF & Purvis, C 2019, ‘New revelations in intact rock strength through automated triaxial testing’, paper
presented at the 53rd U.S. Rock Mechanics/Geomechanics Symposium, American Rock Mechanics Association, Alexandria.

358 SSIM 2023, Perth, Australia