33 Maldonado
33 Maldonado
)
© Australian Centre for Geomechanics, Perth, ISBN 978-0-9876389-7-7
https://papers.acg.uwa.edu.au/p/2025_33_Maldonado/
Abstract
Laboratory direct shear test results for natural bedding and sawcut defects have been statistically analysed
in shales and banded iron formation materials across BHP Western Australian Iron Ore deposits (BHP WAIO)
to prove the hypothesis of similarity of shear strength of defects.
The following paper provides typical Hencher corrected friction angles based on statistical methods to group
data sharing similar characteristics. This paper justifies the existence of similar populations of shear strength
of defects based on rigorous statistical analysis of valid direct shear test samples collected from multiple
projects of the Pilbara. Spatial distribution bias effects and normality tests have been checked. Surface
‘geotechnical’ weathering grades do not explain the reduction of the shear strength of defects in shales. The
use of base of completed oxidation and spectral signatures have provided a better explanation for the
variability observed on the shear strength of defects. Where weathered shale samples from the Brockman
column have greater frictional strength than fresh samples, a paradoxical phenomenon that can be attributed
to chemical alteration generating additional textural roughness.
Keywords: shear strength, shale, bedding partings
1 Introduction
Laboratory direct shear test results for natural bedding and sawcut defects have been statistically analysed
in shales and banded iron formation (BIF) materials across BHP Western Australian Iron Ore deposits
(BHP WAIO) to prove the hypothesis of similarity of shear strength of defects.
This paper provides typical Hencher corrected friction angles based on inferred statistical methods to group
data sharing similar characteristics. The paper aims to firstly demonstrate that the knowledge of the shear
strength of bedding planes from laboratory testing is sufficient to provide the engineer with reliable Hencher
corrected friction angles for formations and stratigraphic units encountered across BHP WAIO and secondly,
to justify the use of the precision index as a way to minimise laboratory testing.
The paper demonstrates that Hencher corrected friction angles within the dominant lithological units are
consistent across the BHP WAIO, are similar in magnitude and with sufficient precision to be used as typical
values for slope design with minimal error, using an interval of confidence of 95% about the sample mean.
2 Geology context
The iron ore deposits of Western Australia are located in the Pilbara region of the state. Mining operations
extract the iron-enriched deposits through numerous open pit excavations within the Hamersley province,
which covers an area of 80,000 km2 and contains late Archaean–Lower Proterozoic age sediments.
The province contains three constituent groups: the Fortescue, Hamersley and the Turee Creek Groups.
The BIF of the Hamersley Group is characteristically fine-grained and layered sedimentary rocks comprising
alternating cherts, iron oxide, shales and carbonates. BIFs were formed by the rhythmic deposition of
dominantly deep-water chemical sediments, which then underwent some form of mineralisation or
enrichment to create the current orebodies.
The sedimentary rocks within the Hamersley Group are further sub-divided into stratigraphical formations
including, from youngest to oldest and as follows:
• Brockman iron formation.
• Mt McRae formation.
• Mt Sylvia formation.
• Wittenoom formation.
• Marra Mamba iron formation.
The BHP exploration geology team commonly segregates the geology of the Pilbara in terms of Brockman
and Marra Mamba deposits, based on the orebody mineralogy type. The upper stratigraphical portion of the
Hamersley Group comprises the Brockman deposits (Joffre, Brockman, Mt McRae, and Mt Sylvia units), which
are characterised by dominant hematite content, whilst the bottom of the Hamersley Group comprises the
Marra Mamba deposits (Wittenoom formation and Marra Mamba iron formation), are characterised by
dominant goethite content (Kepert, 2001). From the practical point of view, this major subdivision of
Brockman and Marra Mamba deposits is applied to segregate the direct shear test data for hypothesis
testing.
The main objective of testing was the determination of the dilation corrected basic friction angle of defects
per rock type (i.e. Shale and BIF lithologies in the Pilbara environment), which is one of the main inputs for
slope stability analysis. The sampling was biased to shale defects, which are more critical for slope stability
in open pits, whilst lesser sampling and testing occurred in the competent iron-rich materials (BIF).
It is important to mention that Hencher, S & Richards, LR (2015) prefers the testing of natural defect surfaces
as a way to estimate the basic friction angle, whilst Hoek (2014) suggests that testing of artificial saw cuts are
similarly reliable. In this study, the results of direct shear testing on both natural and artificial saw cuts are
statistically compared to validate the hypothesis of any difference or similarity of shear strength.
Hencher (2012) describes that the determination of the basic friction angle through the analysis of laboratory
direct shear testing data is not straightforward since the results of laboratory samples are affected by the
contribution of minor asperities (over short profile lengths of 100 mm) that add a dilation component to the
basic friction angle.
Hencher (2012) recommends the application of a dilation correction to the stresses directly derived from
laboratory test results in order to subtract the effect of the small-scale geometry and enable a better
estimation of the basic friction angle (i.e. without dilation). The principle used for dilation correction is based
on the Patton equation:
τ = σ tan (φb + i) (1)
where:
τ = peak shear strength.
σ = normal stress.
φb = basic friction angle.
I = laboratory scale dilation angle.
According to this concept, during shearing, rough surfaces will cause progressive overriding of asperities
throughout the test, and then increase the shear strength to a peak value. Thus, the basic friction angle is
assumed approximately constant whilst the dilation friction angle varies along the test progression.
The dilation component is instantaneously measured during testing by applying the relation: dilation angle =
i = arctangent (vertical displacement/horizontal displacement) then the dilation correction is applied to the
stress data using the below relationships:
τcorrected = (τ cos i − σ sin i ) cos I (2)
σcorrected = (σ cos i + τ sin i ) cos I (3)
Occasionally, the linear regression fitting causes some negative cohesion values, which are a product of
uncontrolled contraction of the test during the initial stage. Whilst this negative cohesion does not detract
from the value of the test, the laboratory has reported the cohesion as being zero, without performing
re-calculation of the friction angle for post-contraction stages. Therefore, the first author was required to
re-calculate the dilation corrected peak friction angle by disregarding the stage of contraction and conducting
the linear regression with a zero intercept (this is a secant friction angle).
6 Statistical methods
For simplicity, the shear strength of defects has been calculated using a fixed normal stress of 350 kPa, in
order to combine the effects of both friction and apparent cohesion in just one variable. Before commencing
the statistical analysis, the following conditions were checked:
• Individual samples are well distributed across multiple projects of the Pilbara (spatial distribution).
• A minimum number of five samples per stratigraphy were considered for hypothesis testing
(sample size).
• All samples when grouped by lithology are approximately normally distributed (normality testing).
The analysis of statistical data was conducted using the MINITAB software (2019). The significant results of
the data analysis are presented within the paper.
In order to prove the existence of similar populations, statistical hypothesis testing methods were applied to
compare the similarity of the mean and variance of either two or more combined samples. These were the
F-test and T-test, and the analysis of variance test (ANOVA) with Tukey’s procedure, described below:
• F-test and T-test: when comparing two samples, the Student T-test method was applied to verify
the similarity of mean values, whilst the Fisher F-test to check the similarity of variances.
• Analysis of variance test (ANOVA) in conjunction with Tukey’s procedure: when comparing more
than two samples; ANOVA was employed to verify the similarity of variances in conjunction to
Tukey’s ‘pairwise’ procedure to check the similarity of the mean values, as presented by Maldonado
& Haile (2015).
The application of the hypothesis testing methods required a minimum sample size of five elements for each
group to ensure fair comparisons (Levine et al. 2013). As varying sample sizes limit the use of simple
techniques, the application of ANOVA becomes a more robust method when aided by the Tukey’s procedure
that transfers the effects of different sample sizes into ‘pairwise’ comparisons.
The hypotheses are assessed on the basis of a 95% interval of confidence using the p-value of the statistic,
where the p-value has to be greater than the level of significance for acceptance (i.e. > 5%). More detail on
these methods can be found in standard statistical books (Levine et al. 2013).
Note that the application of the above methods relies on three assumptions: the data is normally distributed,
each sample has been taken at random from its respective populations, and the variance within each
population must be similar to all populations (McKillup & Dyar 2010).
Figure 1 Spatial distribution of all shale samples across the BHPB WAIO
The shale sample population (without differentiation of weathering grade) is presented in Table 1 by
stratigraphic unit and formation per BHP WAIO project area. The data was used to create the spatial
distribution map in Figures 1 and 2. From Table 1, it is shown that for the most significant stratigraphic
units – which are known to control the stability of slopes within open cut pits (i.e. W, D, R, MM and N) – a
sufficient sample size for most project areas is available for testing.
There is some sampling bias towards the Whaleback (WH) project area as this is one of the deepest open cut
pits (covering the Marra Mamba and Brockman formations) and has multiple pushbacks that have all been
extensively drilled over the years. The other projects have limited sample size for some stratigraphic units
due to their much shallower depth and limited development with less diamond drilling.
The subsequent testing is based on a ‘comparison’ of stratigraphic units that have a minimum sample size of
five within each project area. An exception is applied at the South Flank (SF) and North Flank (NF) project areas,
where the samples of MacLeod (MM), Nammuldi (MU) and Mt Newman (N) were combined due to limited
data. This is justified by the fact that the NF and SF projects are part of same orebody boundary, with similar
geology including stratigraphical units, weathering profile, structural setting and mineralisation, which provide
geological support to the decision of combining the samples of both projects as presented in Table 1.
8 Normality testing
Before commencing the statistical hypothesis, testing it is important to check the ‘normality’ of sample
distribution, for which the Kolmogorov-Smirnov (KS) test is used. The results of the KS test for normality of
shale samples is presented in Figure 2. Note that all samples are lumped together as a single population and
the p-value of the test is ~9%, therefore the single population pass the normality test, i.e. p-value test of
normality is greater than 5%. Applying the KS test to BIF also resulted in a normal population.
It is not considered practical to test the normality of further subdivisions of the dataset, i.e. sub-groups based
on permutations of weathering grades and stratigraphies, as the test of normality requires large sample sizes.
60
50
40
30
20
10
5
0.1
0 1 00 200 300 400
Tau (350kPa)_Statistical Study
Figure 2 Screenshots from MINITAB for the test of normality for all of shales samples
Bedding planes Sawcut 0.967 0.968 N/A Natural bedding planes and artificial saw cuts is
MW-HW MW-HW very similar when surface weathering is MW and
HW.
Bedding planes Sawcut 0.219 0.233 N/A There is sufficient statistical evidence to accept
(all weathering) (all weathering) that the shear strength of bedding planes and
artificial saw cuts are similar.
All bedding planes and sawcut samples (all N/A N/A 0.001 Stratigraphy column is an important factor to
weathering grades), stacked by define separate populations
stratigraphical unit.
Stratigraphy units across projects area
D (between projects) N/A N/A 0.290 Similar
W (between projects) 0.084 0.107 N/A Similar
R (between projects) N/A N/A 0.935 Similar
MM (North Flank, Whaleback) 0.07 0.07 N/A Similar
N (North Flank, Whaleback) 0.496 0.522 N/A Similar
Stratigraphy units within project area
D, J, R (Eastern Ridge) N/A N/A 0.861 Similar
Y, R (Shovelana) 0.664 0.665 N/A Similar
D, W (Whaleback) 0.483 0.49 N/A Similar
D, W, R, S (Whaleback) N/A N/A 0.363 Similar
Processing of geotechnical data and limit design
557
558
Lithology Comparison F-Test (p- T-Test (p-value, ANOVA (p-value, Conclusion
value, one-tail) two tails) two tails)
D, W (Packsaddle) 0.651 0.64 N/A Similar
MM versus N (Whaleback) 0.64 0.66 N/A Similar
MM, N, WA (Whaleback) N/A N/A 0.012 WA is different
MM, N (North Flanck) 0.125 0.127 N/A Similar
Formations across project area
Brockman Iron formation N/A N/A 0.107 Similar
Mt McRae formation N/A N/A 0.935 Similar
Brockman versus Mt McRae v. Mt Sylvia N/A N/A 0.205 Similar
Marra Mamba Iron formation N/A N/A 0.144 Similar
Marra Mamba versus Wittenoom N/A N/A 0.087 Similar
All bedding planes and sawcut samples (all N/A N/A 0.000 Two major stratigraphical formations can be
weathering grades), stacked by defined:
stratigraphical Formations (Brockman,
of non-dilatational angles, mineralogy relationships, and nominal roughness
All bedding planes and sawcut samples (all N/A N/A 0.243 Similar
weathering grades), stacked by project
area
Brockman formation (BOCO: 1 versus 0) 0.032 0.043 N/A Similar Variance but different mean
Marra Mamba formation 0.601 0.561 N/A Similar Variance and mean
(BOCO: 1 versus 0)
P-values should be compared against α=0.025 for a two-tail test, and α=0.05 for a one-tail test.
Y = Yandicogina, J = Joffre, W=Whaleback, D = Dales Gorges, R = McRae Shale, S = Sylvia, WA = West Angela Shale, N = Newman, MM = McLeod, MU=Nammuldi.
Packsaddle,
Brockman Below 22 23 7 17 4.9% 1.22 Detailed
Eastern
(Y, J, W, D, Ridge,
R, S) Shovelanna
Above 26 26 6 22 2.5% 1.10 Detailed
Whaleback
Marra
North Flank
Mamba
South Flank All 28 29 7 25 3.4% 1.14 Detailed
(WA, N,
Whaleback
MM, MU)
Y = Yandicogina, J = Joffre, W = Whaleback, D = Dales Gorges, R = McRae Shale, S = Sylvia, WA = West Angela Shale, N = Newman, MM = McLeod,
MU = Nammuldi.
The friction angle of Brockman shales is additionally divided by BOCO, which is based on distinct chemical
compositions from fresh (below BOCO) to weathered (above BOCO). In fresh conditions, natural defects can
approach the lower bound of 17° (25th percentile) when surfaces are polished, whilst the average friction
angle of 23° is representative of planar smooth conditions (not polished). The recent geochemical analysis of
shale composition supports that the oxidation process has altered the original mineralogy of shales by
changing the proportions of ferric oxides, quartz and clayey mineral contains. This chemical transformation
provides an explanation for the relative increase in the basic friction angle in weathered shales of the
Brockman column as presented in Figure 3 and Table 3.
The majority of friction angle data for Marra Mamba shales are in weathered (above BOCO). The basic friction
angles for defects can be represented by the mean value of 29° and the lower bound value of 25°. These
friction angles are applicable for the stratigraphic units presented in Figure 4 and Table 3.
12 Nominal roughness
The Hencher correction is applied to the peak stresses of DST in order to remove the effects of dilation caused
by either the overriding of roughness (small-scale geometry component) or the work carried out by
mobilisation of debris during progressive shearing (asperities failure). Either way, the dilation phenomenon
adds a nominal roughness angle to the basic friction angle, which should be accounted for in the estimation
of the peak friction angle at the laboratory scale.
The nominal roughness angle is empirically related to Joint Roughness Coefficient (JRC) and Joint
Compressive Strength (JCS) and the normal stress as defined by Barton (1973, 1976), Barton &
Choubey (1977) and Barton (1995). Interestingly, the dilation angles determined from DST data used in this
study are lower than the values derived by the equation of Barton (1973). For pragmatism, it is only possible
to estimate the nominal roughness angle based on JRC alone, at the lab scale (profile with a length between
10 and 20 cm). Table 4 presents the relationships between material type, dilation angle and JRC that are
recommended to estimate the nominal roughness angle for design purposes.
Nominal roughness
Rock type Relevant stratigraphy Studied projects BOCO R-square
angle (d°)
Packsaddle, Below 0.7 d° = 0.4 JRC
Brockman Eastern Ridge,
(Y, J, W, D, R, S) Shovelana, Above N/A d° = 0
Shale
Whaleback
North Flank,
Marra Mamba
South Flank, Either N/A d° = 0
(WA, N, MM, MU)
Whaleback
Y = Yandicogina, J = Joffre, W=Whaleback, D = Dales Gorges, R = McRae Shale, S = Sylvia, WA = West Angela Shale, N = Newman, MM = McLeod,
MU=Nammuldi.
Histogram of Hencher_Friction
Normal
0.08 BOCO
0
0.07 1
Mean StDev N
0.06 23.23 7.354 41
26.1 6 6.071 84
0.05
Density
0.04
0.03
0.02
0.01
0.00
8 16 24 32 40
Hencher_Friction
Figure 3 Histogram of Hencher corrected friction angle for shales in Brockman iron formation
differentiated by above and below BOCO
Mean
Median
27 28 29 30 31
Figure 4 Histogram of Hencher corrected friction angle for shales in Marra Mamba iron formation
14 Conclusion
The current precision index of the Hencher’s friction angle presented in Table 3 exceeds the precision needed
for practical engineering design. The level of knowledge approached a research level (Gill et al. 2005).
The current state of accuracy on shear strength is sufficient to justify limited direct shear testing for the
specific stratigraphy and lithology types listed in Table 3.
The use of Hencher friction angles for defining the bedding shear strength of defects increase the accuracy
on modelling and promote a stream-line design process.
Table 4 presents the most accurate relationship to estimate nominal roughness for shale and BIF from logging
data, from which the ‘most likely’ nominal roughness can be determined from JRC distribution (i.e. the
median of JRC).
Acknowledgement
The authors thank the BHP geotechnical engineering team for supporting this work. In addition, the authors
thank to Ted Brown as the external reviewer of the project in 2017. Similarly, the authors express their
gratitude to Andy Haile and Ken Mercer for encouraging the PhD work since November 2015.
References
Barton, N 1973, ‘Review of a new shear strength criterion for rock joints’, Engineering Geology, vol. 7, pp. 287–332.
Barton, N 1976, ‘The shear strength of rock and rock joints’, International Journal of Rock Mechanics & Mining Sciences, vol.13,
pp. 255–279.
Barton, N 1995, ‘The influence of joint properties in modelling jointed rock masses’, Proceedings of the 8th International Society for
Rock Mechanics Congress, International Society of Rock Mechanics, Salzburg, pp. 1023–1032.
Barton, N & Bandis, SC 1990, ‘Review of predictive capabilities of JRC-JCS model in engineering practice’, in N Barton & O Stephansson
(eds), Proceedings of the International Symposium on Rock Joints, Balkema, Rotterdam, pp. 603–610.
Barton, N & Choubey, V 1977, ‘The shear strength of rock and rock joints in theory and practice’, Journal of Rock Mechanics and Rock
Engineering, vol. 10, pp. 1–54.
Gill, DE, Ro, T & Leite, MH 2005a, ‘Determining the minimal number of specimens for laboratory testing of rock properties’,
Engineering Geology, vol. 78, vol. 1–2, pp. 29–51.
Gill, DE, Corthesy, R & Leite, MH 2005b, ‘A statistical approach for determining practical rock strength and deformability values from
laboratory tests’, Engineering Geology, vol. 78, issue 1–2, pp. 53–67.
Hencher, S 2012, Practical Engineering Geology, CRC Press, Boca Raton.
Hencher, S & Richards, LR 2015, ‘Assessing the shear strength of rock discontinuities at laboratory and field scales’, Rock Mechanics
& Rock Engineering, vol. 48, issue 3, pp. 883–905.
Hoek, E 2004 Rock Engineering, Course notes, Canada, 313 p.
International Society for Rock Mechanics 1978, ‘Suggested methods for the quantitative description of the discontinuities in rock
masses’, International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, vol. 15, issue 6,
pp. 319–368.
Kepert, DA 2001, The mapped stratigraphy and structure of the mining Area C region, an eclectic synthesis of geological mapping by
BHPBIO exploration 1994-2001, internal report.
Levine, DM, Stephan, DF & Szabat, KA 2013, Statistics for Managers, Pearson Higher Ed USA, USA.
Mc Killup, S & Dyar, MD 2010, Geostatistics Examined, An Introductory Guide for Earth Scientists, Cambridge University Press,
Cambridge.
Maldonado, A & Haile, A 2015, ‘Application of ANOVA and Tuckey-Cramer, statistical analysis to determine similarity of rock mass
strength properties across Banded Iron Formations of the Pilbara region in Western’, in TR Stacey (ed.), Proceedings of the
2015 International Symposium on Slope Stability in Open Pit Mining and Civil Engineering, The Southern African Institute of
Mining and Metallurgy, Johannesburg.
Mercer, K 2013, ‘Linear Anisotropic Model History of development Part 1’, Australian Centre for Geomechanics Newsletter,
https://acg.uwa.edu.au/wp-content/uploads/2014/12/NL38_web_ref.pdf
Minitab Software 2019, https://www.minitab.com/en-us/
RocScience Inc. 2018, Slide2, computer software, https://www.rocscience.com/software/slide2