Sei sulla pagina 1di 17

Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Contents lists available at ScienceDirect

Journal of Rock Mechanics and


Geotechnical Engineering
journal homepage: www.rockgeotech.org

Full Length Article

Revisiting rock classification to estimate rock mass properties


Robert Bertuzzi
Pells Sullivan Meynink, North Ryde, NSW 2113, Australia

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents the results of ongoing research carried out by the author exploring methods to
Received 28 July 2018 provide a more robust estimate of rock mass properties specifically for use in tunnel design. Data from
Received in revised form various large-scale rock mass failures are introduced, including coal pillars. The damage-initiation,
16 August 2018
spalling-limit approach is compared to the coal pillar database. New comparisons of estimating the
Accepted 31 August 2018
Available online 13 December 2018
geological strength index (GSI) and relationships to estimate the HoekeBrown failure criterion param-
eters, mb, s and a, are presented.
Ó 2018 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by
Keywords:
Rock mass properties
Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
HoekeBrown strength criterion licenses/by-nc-nd/4.0/).
Geological strength index (GSI)
Tunnelling

1. Introduction geological strength index (GSI) and the disturbance factor (D)
(Hoek et al., 2002) as follows:
Estimating the strength and deformation characteristics of a 9
rock mass is generally based on empiricism as it is simply not >
>
1 1  >
>
practical to test the rock mass at a large enough scale. The gener- a ¼ þ eGSI=15  e20=3 >>
>
>
alised HoekeBrown criterion (see Eq. (1)) (Hoek, 1994; Hoek and 2 6 =
Brown, 1997) has proved to be a practical method to estimate GSI100 (2)
mb ¼ mi e 2814D >
>
intact rock and rock mass strengths (Eberhardt, 2012). >
>
>
>
>
GSI100
s ¼ e 93D
>
;
 a
s03
s01 ¼ s03 þ sc mb þs (1)
sc where mi is a material parameter of the intact rock.
This paper presents the results of ongoing research carried out
where s01 and s03 are the major and minor principal effective by the author exploring methods to provide a more robust estimate
stresses, respectively; sc is the uniaxial compressive strength; and of rock mass properties specifically for use in tunnel design. The
mb, s and a are the parameters of the generalised HoekeBrown research carried out at University of New South Wales (UNSW),
criterion. Australia is the primary source of the material presented in this
Its popularity remains though many have suggested modifica- paper (Bertuzzi, 2017b). This paper starts with brief discussions on
tions (e.g. Pan and Hudson, 1988; Mostyn and Douglas, 2000; intact rock strength and classification systems before introducing
Sonmez and Ulusay, 2002; Serrano et al., 2007; Zhang and Zhu, an extensive database of large-scale rock mass cases, and then
2007; Benz et al., 2008; Saroglou and Tsiambaos, 2008; Dinc proposes new relationships to estimate the parameters, mb, s and a.
et al., 2011; Lee et al., 2012; Ismael et al., 2014). Zuo et al. (2015)
provided a theoretical basis for the criterion, yet its parameters, 2. Intact rock strength
mb, s and a, are typically subjectively estimated through the
Much of what is known about intact rock strength has come
from three types of laboratory tests: the uniaxial tensile strength
(UTS) test, which for practical purposes is almost always replaced
E-mail address: robert.bertuzzi@psm.com.au.
by the Brazilian indirect tensile strength test (Pells, 1993) though it
Peer review under responsibility of Institute of Rock and Soil Mechanics, overestimates the UTS by 10%e30% (Perras and Diederichs, 2014;
Chinese Academy of Sciences. Read and Richards, 2015); the uniaxial or unconfined compressive

https://doi.org/10.1016/j.jrmge.2018.08.011
1674-7755 Ó 2018 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510 495

strength (UCS) test; and the triaxial compressive strength test. True Zimmerman, 2005; Costamagna and Bruhns, 2007; You, 2009; Lee
three-dimensional (3D) triaxial or poly-axial tests are typically et al., 2012; Kwasniewski, 2013). Paraphrasing the study of
limited to specialised research. The difficulty in defining “rock Christensen (1997), many empirical criteria with two or more pa-
strength” stems from the inherent variability of the rock itself and rameters would likely fit some sets of test data very well.
the reliability of the testing methods (Pells, 1993), and also it is Bertuzzi et al. (2016b) added to this group by comparing the
unlikely that a general strength criterion is true for all rock types curve-fits of six strength criteria (the modified Weibols-Cook, Lade,
under all stress regimes (Fairhurst, 1971; Hardy et al., 1973). Yet in HoekeBrown and two of its modifications, and Christensen) against
rock mechanics, there is a need to predict the rock strength under a database comprising 128 data sets with over 7000 laboratory test
different confining stresses. results. The data sets include data obtained from several recent
Numerous intact rock strength criteria have been proposed, as tunnelling projects in Sydney and Brisbane, Australia, e.g. Sydney’s
elaborated in Edelbro (2003) and You (2011) for example. The Hawkesbury sandstone (Fig. 1) and Brisbane Tuff (Fig. 2). It would be
question which strength criterion is the best has been the subject of easy to ignore this tunnelling-sourced database as offering ill-
much research and many have compared various criteria with conditioned data as it is limited in number and more importantly
laboratory test data (e.g. Yoshinaka and Yamabe, 1980; Pan and in type, being mainly restricted to UCS tests. However, it was pur-
Hudson, 1988; Colmenares and Zoback, 2002; Al-Ajmi and posefully included to assess the capability of each of the criteria to

Fig. 1. Hawkesbury sandstone - Principal stress plot showing the experimental data with the predicted strength envelopes.

Fig. 2. Brisbane tuff - Principal stress plot showing the experimental data with the predicted strength envelopes.
496 R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Fig. 3. Approximate relationship between Jr/Ja and apparent ‘shear strength’ (expressed as a friction angle) (Barton and Grimstad, 2014b).

provide a failure envelope with the type of limited data often comparisons of the numerous rock mass classification systems that
available. Several conclusions were drawn (Bertuzzi et al., 2016b): exist (e.g. Edelbro, 2003; Palmström, 2005).
The innate characteristics of the classification systems, namely
(1) Any of the assessed criteria can provide good fits to tight test the natural variability and subjectivity of a person who uses the
data. Conversely, correlations are poor across all the criteria system, suggest that they must be used cautiously. Perhaps
when the test data have considerable scatter (Figs. 1 and 2). counter-intuitively, it is argued that additional care is needed not to
(2) The variation in test data swamps the differences between mask the natural variability and complexity of geology when using
criteria. This is in keeping with Al-Ajmi and Zimmerman geological classification systems that facilitate use by inexperi-
(2005). enced professionals.
(3) Of the criteria assessed, the curves that best reproduce the
tensile and low confinement region are those of the Hoeke 3.1. Q
Brown criterion with parameters mi and a derived as rec-
ommended by Mostyn and Douglas (2000). Barton et al. (1974) in their seminal paper developed a corre-
lation between the permanent support installed in underground
excavations with rock mass quality, Q, which they defined as
3. Classification
RQD Jr Jw
Q ¼ (3)
The rock mass quality (Q) (Barton et al., 1974; Grimstad and Jn Ja SRF
Barton, 1993; Barton and Grimstad, 2014a) and the GSI (Hoek,
1994; Hoek and Brown, 1997; Marinos and Hoek, 2000) are dis-
cussed in this section specifically in relation to estimating rock “The RQD (rock quality designation) index, the number of joint sets
mass properties. The interested reader is referred elsewhere for (Jn), the roughness of the weakest joints (Jr), the degree of alteration
R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510 497

or filling along the weakest joints (Ja), and two further parameters criterion parameters (see Fig. 5). Several implications arise from
which account for the rock load (SRF) and water inflow (Jw)." this chart.
First, strength is not explicitly included in the chart though it is
the first parameter required when calculating RMR. Intact rock
While the terms “the weakest joints” and “least favourable joint
strength is however broadly implied by selecting a particular chart
set” were both used in Barton et al. (1974), the latter term is
as per Marinos and Hoek (2000) or in the matrix of Carter and
preferred (e.g. Grimstad and Barton, 1993; Barton and Grimstad,
Marinos (2014). Second, the user needs to consider the scale of
2014a). Though Q does not directly provide estimates of rock
the problem when selecting the vertical axis of the chart. The
mass strength, its constituent parameters are linked to a joint
introduction of this subjectivity addresses a major limitation of
frictional strength (f) and to the Barton-Bandis shear strength
RMR, in that its parameters, RQD and JCond89, are scale-dependent
criterion as indicated in Figs. 3 and 4 as follows:
(Priest and Hudson, 1976; Hudson and Harrison, 1997; Mostyn
  and Douglas, 2000). Third, the descriptions used to define the
JCS
f ¼ JR log10 þ fr (4) vertical axis of the chart also introduce the concept of interlocking
sn
rock blocks.
where fr is the residual friction angle for the general case. For flat, A further implication is that the GSI chart’s axes are not inde-
unweathered surfaces, Barton and Choubey (1977) claimed fr ¼ fb, pendent as the surface condition of geological structures (hori-
i.e. the basic friction angle. zontal axis) and rock mass blockiness and degree of interlocking
(vertical axis) are inter-related. This dependence is seen as a diag-
onal trend when overlaying various qualities of the same rock mass
3.2. GSI type on one chart. Marinos and Hoek (2000) and Marinos et al.
(2006) ‘blocked off’ the top right and bottom left cells in their GSI
Hoek (1994) introduced GSI to overcome limitations in the rock charts, noting that these combinations are not possible (Fig. 6).
mass rating (RMR) (Bieniawski, 1973, 1976, 1989) in determining Methods to re-quantify the GSI chart ‘to facilitate use of the
rock mass strengths. GSI was also a “system based more heavily on system especially by inexperienced engineers’ (Cai et al., 2004) have
fundamental geological observations and less on ‘numbers’” (Hoek, been suggested. Sonmez and Ulusay (1999) linked the vertical axis
2002). The GSI ranged from “about 10, for extremely poor rock of the GSI chart to ‘volumetric joint count’ (Jv) and its horizontal
masses, to 100 for intact rock” (Hoek, 1994) and was originally axis to components of the JCond89 parameter. Cai et al. (2004)
calculated on the basis of four of the RMR parameters: intact rock defined the vertical axis as ‘block volume’ (Vb) and the horizontal
strength, RQD, spacing of discontinuities, and the joint condition axis as ‘joint condition’ (Jc) following the RMi (rock mass index)
parameter (JCond89). classification system (Palmström, 1995). Russo (2009) did the same
Hoek (1994) paved the way to assess GSI directly from a chart but gave an explicit equation between GSI, Vb and Jc. Hoek et al.
rather than quantifying it through RMR by publishing a chart (2013) set the chart’s vertical axis to RQD/2 (limited to RQD  80)
showing a link between GSI and the HoekeBrown strength and its horizontal axis to 1.5JCond89 and provided the equation
GSI ¼ 1.5JCond89 þ RQD/2.
However, setting the vertical axis to a measure of blockiness re-
introduces the problem of scale. While Hoek et al. (2013) and
Cundall et al. (2008) recognised this and suggested subjectively
relating the vertical axis to the number of blocks at the required
scale, care is still needed as not all geological structures contribute
equally to the rock mass properties. Rock mass behaviour will be
increasingly controlled by persistent structures as the problem
scale increases.
Bertuzzi et al. (2016c) found that quantifying GSI following Hoek
et al. (2013) typically produces data points that are generally within
10 of those obtained directly from the GSI chart (Fig. 7), a similar
result to those of Russo (2009) and Pells et al. (2017). The excep-
tions are rock masses whose RQD does not capture the blockiness
or the degree to which rock blocks are interlocked as does the GSI
chart. For example, the very well interlocked yet blocky rock mass
in Fig. 8 has essentially a RQD ¼ 0 but plots between GSI ¼ 50
and 70.
To see whether block volume would offer a better indication of
blockiness, the method adopted by Cai et al. (2004) was followed to
calculate the GSI for the Hawkesbury sandstone and Ashfield shale
found in Sydney. The data used are those given in Bertuzzi
(2014a,b). The blocks in the Hawkesbury sandstone and Ashfield
shale are typically formed by the two pairs of sub-vertical joints
intersecting the sub-horizontal bedding partings. Cross-beds are
occasionally involved. Distributions of spacings for bedding plane
partings and joints are shown in Fig. 9. The typical values of the
calculated Jc and Vb are listed in Table 1 for Classes I to IV. There was
insufficient data for Class V materials.
The results suggest that the method of Cai et al. (2004) to
Fig. 4. Approximate relationship between Jr and JRC for 0.2 m and 1 m long defects quantify GSI produces results consistent with those assessed using
(Bandis, 1993). the chart for Hawkesbury sandstone and Ashfield shale. It is
498 R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Fig. 5. Chart published by Hoek (1994) linking GSI and the HoekeBrown criterion.

thought that this is largely a function of Vb being a better measure 4.1. Hawkesbury sandstone
of blockiness than RQD, in line with the conclusion reached by
Bertuzzi et al. (2016c), “the quantified GSI approach is used to sup- Finite element modelling (Fig. 10) was used to assess the likely
plement and check the visually assessed one”. magnitudes of the stresses presented at the time of documented
failures that occurred in nine tunnels excavated within Hawkes-
bury sandstone (Bertuzzi, 2015). The values of major and minor
4. Rock mass strength database
principal stresses were compared to the rock mass strength enve-
lopes of Hawkesbury sandstone derived using the HoekeBrown
Hoek and Brown (1980) started with the Griffith crack extension
criterion (Fig. 11).
theory and used trial-and-error method with an extensive intact rock
It should be highlighted that the inferred range of confining
data set and a modest collection of triaxial data of 6-inch (1
stress, s3, presented in the case studies is rather modest being only
inch ¼ 2.54 cm) diameter core and re-compacted samples of Panguna
up to 3 MPa. Yet the predicted HoekeBrown rock mass strength
andesite, and empirically derived a criterion that “adequately de-
criterion differentiates the ‘failed’ and ‘stable’ zones in the nine
scribes the response of an intact rock sample from uniaxial tensile stress
cases very well. The criterion did identify zones other than those
to triaxial compressive stress”. Bertuzzi et al. (2016d) added laboratory
mapped as failing but these tend to be in the floors of the tunnels,
test results and important data from large-scale underground exca-
i.e. zones which are not as readily picked up in mapping as having
vations, and thereby suggested improvements to the HoekeBrown
failed.
criterion parameters and GSI relationships given in Eq. (2).
R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510 499

Fig. 6. Rock mass qualities in ophiolitic complexes from Marinos et al. (2006), showing the typical diagonal trend from top left to bottom right for decreasing rock mass quality, from
Classes 1 to 4.

4.2. Coal from 13 m to 594 m with width to height (w/h) ratios between 0.6
and 16.7 was therefore collated by Bertuzzi et al. (2016a). Fig. 12
Estimates of coal pillar strength tend to be based on shows that there is a general trend of increasing w/h with
geometrically-based empirical methods. The intact strength of the increasing depth as evidenced by the arbitrary line, which can be
coal, the in situ stress and the geological structures of the coal rock written as w/h ¼ 2 þ 0.0125depth. The database contains 15
mass may be implicit in the methods, but their relative impact is collapsed cases with w/h > 5. All the collapsed cases have w/h  8.
largely unknown. A means of dealing with this was assessed in the This combined database may address a criticism of the UNSW coal
1990s by applying the HoekeBrown criterion to coal pillars, pillar design method (Galvin et al., 1999; Galvin, 2006), that its
particularly in the context of highwall mining (Medhurst and database only contains one collapse case with w/h > 5 (Colwell,
Brown, 1996, 1998; Medhurst, 1999). However, the typical 2010; Seedsman, 2012).
method of estimating coal strength is at odds with the methods for Using 2D (two-dimensional) and 3D finite element modelling,
non-coal rocks. Bertuzzi et al. (2016a) plotted the stresses predicted in the
A database comprising 162 collapsed and 507 un-collapsed coal collapsed and un-collapsed pillars of the database against a
pillars from Australia, South Africa, USA and India covering depths damage-initiation, spalling limit (DISL) envelope (Diederichs, 2000,
500 R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Fig. 7. Comparison of the GSI quantified following Hoek et al. (2013) and that assessed from the chart.

mb ¼ 1:47; s ¼ 0:07; a ¼ 0:5 (5c)

(4) Maximum confined strength:

s1  s3
¼ 0:6 ðin the range of 0:5 and 0:7Þ (5d)
sc
The extended DISL envelope differentiates between the
collapsed and the un-collapsed coal pillars very well (Fig. 13),
separating approximately 80% of the collapsed and 90% of the un-
collapsed pillars. Most of the collapsed pillars that the envelope
did not differentiate have w/h  4, which can be defined as slender
pillars (Mark, 2006) and almost half have w/h  2. Half of the un-
collapsed pillars that plot above the envelope are identified as be-
ing highly rectangular.
It is clear that geological structures need to be considered, partic-
ularly for slender pillars (w/h  4), but if failure is expected to occur
through the rock mass then the extended DISL envelope can be used to
Fig. 8. Quartzite rock mass downstream of the Ord Diversion Dam, Western Australia.
estimate rock mass strength (Bertuzzi et al., 2016a) (see Fig. 14).

4.3. Others
2007) capped by the maximum confined strength, this latter
component is in keeping with Mogi (1966), Bieniawski (1968) and The following data were also added in addition to the Haw-
Barton (1976). This extended DISL envelope is defined in Eqs. (5a)e kesbury sandstone and coal pillar data: Ok Tedi hornfels and
(5d) and shown in Fig. 13. limestone core (Bertuzzi et al., 2016d), granite core (Alejano et al.,
2015), quartzite and phyllite core (Habimana et al., 2002), Luc du
(1) Damage-initiation threshold: Bonnet granite core (Martin, 1993), Stripa granite large-diameter
core (Thorpe et al., 1980), hard rock mines (Lunder, 1994), lime-
mb ¼ 0:1296; s ¼ 0:0081; a ¼ 0:25 (5a) stone mines (Esterhuizen et al., 2006), iron ore pillars (Hosni et al.,
2015), and Laisvall sandstone pillars (Soder and Krauland, 1989).
(2) Spalling limit:
4.4. Disturbance factor

mb ¼ 5:33; s ¼ 0; a ¼ 0:75 (5b) The preceding rock mass data sets were subjectively categorised
for the appropriate disturbance factor, D, following the guidelines
(3) Macro-scale shear failure: in Hoek et al. (2002). The resulting broad categories are shown in
R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510 501

Fig. 9. Distribution of logged spacing for bedding planes and joints for Hawkesbury sandstone (Classes I to IV) and Ashfield shale (Classes I to IV) (Bertuzzi, 2014a,b). Classes follow
those of Pells et al. (1998). Percentages are lengths of core.

Table 1
Comparison of the GSI quantified following the method of Cai et al. (2004) and that
estimated using the chart. that the relationships with GSI < 30 should be used cautiously as
recommended by Brown (2008).
Class Hawkesbury sandstone Ashfield shale
The relationships between mb/mi, s, GSI, and D, which are shown
Jc Vb GSI Jc Vb GSI
in Fig. 18, show a broad agreement between increasing D and lower
(m3) (m3)
Cai et al. (2004) Fig. 7 Cai et al. (2004) Fig. 7 mb and s. Fig. 18 also shows that the proposed relationships better
I 2.71 8.7 80 65 2.31 3.8 75 45 fits the data.
e85 e65
II 1.68 3 75 55 0.89 0.7 55 40
e75 e60 6. Modulus
III 1.02 1.4 60 45 0.56 0.2 45 30
e65 e50
IV 0.63 0.3 50 35 0.26 0.1 40 25
The modulus of a rock mass (Em) is at the core of almost any
e55 e35 form of analysis used for the design in rock (Brown, 2008). Para-
phrasing Brown (2008), there is now a 40-year history of attempts
to correlate rock mass modulus with rock mass classifications. The
empirical relationships between rock mass quality are that RQD
(Deere, 1968; Deere and Deere, 1988), RMR (Bieniawski, 1973,
Table 2 with the reasons for assigning the D values. It is stressed
1976,1989), Q (Barton et al., 1974; Grimstad and Barton, 1993;
that these categories are, by definition, imprecise.
Barton and Grimstad, 2014a), GSI (Hoek, 1994; Hoek and Brown,
1997; Marinos and Hoek, 2000; Hoek and Diederichs, 2006) or
5. GSI relationships other measures. However, the inherent scatter in measured Em
means that any curve fitting approach is expected to be a poor
The relationships between mb and GSI, s and GSI, and a and GSI, predictor.
proposed by Bertuzzi et al. (2016d), are shown in Figs. 15e17, Bertuzzi (2017a) plotted a rock mass moduli database that was
respectively, and in Table 3. While the first two relationships differ collated from published data of reliable rock mass moduli values
marginally from Eq. (2), the proposed relationship between the (Fig. 19). As can be seen in the figure, much of the data lies about the
parameter a and GSI allows a to vary between 0.5 for GSI ¼ 100 and value predicted by Hoek and Diederichs (2006) for D ¼ 0. The range
1 for GSI  10. The area GSI < 30 is shaded in Figs. 15e17 to indicate shown by the dashed lines in the figure is generated by replacing
502 R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Fig. 10. Example of the results from the Phase2 (RocScience Version 8.01) finite element analyses. Contours of major principal stress are shown (top). Lines along which the major
and minor principal stresses (s1 and s3) were queried from the results are also shown. The s1 and s3 pairs are plotted in the graph (bottom) which also shows the strength envelope
predicted by the HoekeBrown criterion.
R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510 503

Fig. 11. Principal stress plots (differential vs. minor) showing the results of the numerical modelling compared with proposed strength envelopes. Solid symbols are used to identify
the inferred stresses in “failed zones” and open symbols for “stable zones”. The graphs show that the strength envelope predicted by the HoekeBrown criterion differentiates the
stable and failed zones quite well.

the constant in the numerator of the exponential with 70 and 80, routinely recorded during tunnelling. The procedure requires highly
respectively, as follows: redundant data to counter its inherent high variability.
Back-analyses from convergence monitoring data obtained
  during tunnelling in Sydney suggested isotropic conditions for
1  D=2
Em ¼ 100 Bþ25DGSI
(6) Hawkesbury sandstone, with Eh ¼ Ev ¼ 5500 MPa and K ¼ 3.0,
1þe 11
and slightly anisotropic conditions for Ashfield shale with
Eh ¼ 720 MPa, Ev ¼ 1000 MPa, and K ¼ 1.7 (Bertuzzi, 2017a),
where Em is the modulus of a rock mass (GPa); B is a parameter with
where Eh, Ev and K are the rock mass modulus in the hori-
values of 70, 75 or 80 for the upper bound, mean or lower bound,
zontal and vertical directions and the in situ stress ratio,
respectively.
respectively.
Fig. 20 plots the data from Palmström and Singh (2001) and
Nejati et al. (2014), which is likely to represent rock mass disturbed
by drill-and-blast and stress-relief methods. The curves are 7. Conclusions
generated from Eq. (6) with D ¼ 0.5 and suggest that the value of
D ¼ 0.5 is appropriate for rock masses disturbed by drill-and-blast This paper presented the results of ongoing research exploring
and stress-relief methods. methods to provide a more robust estimate of rock mass properties
Most tunnel projects typically do not have the extensive and specifically for use in tunnel design. Several issues are addressed:
intensive monitoring schemes that have been used to back-analyse in
situ stresses and rock mass modulus, such as that carried out at the (1) Many strength criteria can provide good fits to tight test data
Atomic Energy of Canada Limited’s Underground Research Laboratory as it is the variation in test data that swamp the curve-fitting.
(Read, 1994). Following Amadei and Savage (1991), Bertuzzi (2017a) The HoekeBrown criterion can reproduce the tensile and
developed a relatively simple procedure to check the in situ stress low confinement region with due care in selecting parame-
ratio and rock mass modulus from the type of displacement data ters mi and a.
504 R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Fig. 12. (a) Collapsed and (b) un-collapsed databases plotted as depth (m) against pillar width (w) to pillar height (h) ratio. The circled cases are the yielding pillars from Jitpur, India
and the highly rectangular pillars from Utah, USA. The bottom plots (c) and (d) are subsets of the database. The line shown is w/h ¼ 2 þ 0.0125depth.

(2) The natural variability of rock masses means that classifica- capture the degree to which rock blocks are interlocked. It
tion systems must be used cautiously. Perhaps counter- was observed for the major rock types found in Sydney that
intuitively, additional care is needed in using systems that the approach of Cai et al. (2004) matched the chart estimated
claim to be more accessible to inexperienced professionals, GSI very well.
so that the natural variability and the complexity of geology (4) A database that comprises 162 collapsed and 507 un-
are not masked. collapsed pillars was collated. The combined database con-
(3) The GSI assessed from its chart and that quantified following tains 15 collapsed cases with w/h > 5. All the collapsed cases
Cai et al. (2004) or Hoek et al. (2013) typically produce data have w/h  8. A damage-initiation, spalling limit envelope
points that are generally within 10 of each other. The ex- capped by the maximum confined strength, was shown to
ceptions to this are rock masses whose blockiness may not separate the collapsed and un-collapsed pillars very well.
R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510 505

Fig. 13. Normalised average principal stress plots of the coal pillar database; solid symbols represent the collapsed pillars - (a) and (b) - and open symbols the un-collapsed pillars -
(c) and (d). Further, square symbols represent the average stresses, triangles represent the stresses at the edges of pillars, and circles are the stresses at the centre of the pillars. The
stresses are along a horizontal plane at the mid-height through the coal pillars. The yielding pillars from Jitpur, India are circled as well as the highly rectangular pillars and slender
pillars. The right-hand plots are a subset of the data, showing the cases with s3/sc  0.5. The extended DISL strength curve is also shown with three maximum confined strength
(MCS) values. It differentiates well the un-collapsed and collapsed pillars.
506 R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Fig. 14. The extended DISL envelope for coal pillars.

Table 2
Broad categories of the disturbance factor, D (Bertuzzi et al., 2016d).

D Database

0 Sandstone pillar test e controlled blasting


Habimana sandstone and phyllite e drill core
Lac du Bonnet granite from URL (underground research laboratory) e controlled blasting
Sedimentary e TBM (tunnel boring machine)
Ok Tedi core samples
(0, 0.8) Hawkesbury sandstone e mainly mechanical excavation but some blasting
Coal pillars e mainly mechanical excavation or controlled blasting
Stripa granite e blasted tunnel but slot drilled core
Artificially jointed granite e man-made fractures
0.8 Hard rock pillars e production blasting
Limestone pillars e production blasting
Mines - metasediments, gneiss e production blasting
Iron ore pillars e production blasting

Fig. 15. Proposed relationship between mb/mi and GSI. The shading represents the values as the disturbance factor, D, varies from 0 to 1.
R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510 507

Fig. 16. Proposed relationship between s and GSI. The shading represents the values as the disturbance factor, D, varies from 0 to 1.

Fig. 17. Proposed relationship between a and GSI.

Table 3
Comparison of proposed and current equations between HoekeBrown parameters
and GSI.

Current equations (Hoek et al., 2002) Proposed equations (5) Data from laboratory testing and large-scale underground
GSI  100 GSI  100 cases were added to the Panguna andesite database that has
mb ¼ mi e 28  14D mb ¼ mi e 38  24D been the basis for the relationships between GSI and the
GSI  100 GSI  100 HoekeBrown criterion parameters mb, s and a. The larger
s ¼ e 9  3D s ¼ e 12  6D
  database has enabled improvements to the relationships.
1 1 1
a ¼ þ ðeGSI=15  e20=3 Þ a ¼ min 1; ð1 þ eðGSI10Þ=15 Þ (6) Upper and lower bounds are suggested for using GSI to
2 6 2
predict rock mass modulus.
508 R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Fig. 18. Proposed relationship between mb/mi and GSI (top) and between s and GSI (bottom) for the three broad categories of disturbance factor, D. Refer to legend in Fig. 17.

Fig. 19. Range of in situ modulus versus GSI showing the typical bounds for undis-
turbed rock masses. The mean curve is that of Hoek and Diederichs (2006), while the Fig. 20. Range of in situ modulus versus GSI for rock masses disturbed by drill-and-
range is that proposed in Bertuzzi (2017a). blast and stress relief methods. The mean curve is that of Hoek and Diederichs
(2006), while the range is that proposed in Bertuzzi (2017a).

(7) A relatively simple procedure using highly redundant Conflicts of interest


displacement data routinely recorded during tunnelling to
check the assumed in situ stress ratio and rock mass modulus The author wishes to confirm that there are no known conflicts
is presented. of interest associated with this publication and there has been no
R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510 509

financial support for this work, from any source, that could have Bieniawski ZT. Rock mass classifications in rock engineering. In: Bieniawski ZT,
editor. Proceedings of the symposium on exploration for rock engineeringvol. 1.
influenced its outcome.
Johannesburg, South Africa: A.A. Balkema; 1976. p. 97e106.
Bieniawski ZT. Engineering rock mass classifications. New York: John Wiley & Sons;
1989. p. 1e125.
Acknowledgements Brown ET. Estimating the mechanical properties of rock masses. In: Potvin Y,
Carter J, Dyskin A, Jeffrey R, editors. SHIRMS 2008. Perth: Australian Centre for
Geomechanics; 2008. p. 3e21.
The author is indebted to Kurt Douglas and Garry Mostyn for Cai M, Kaiser PK, Uno H, Tasaka Y, Minami M. Estimation of rock mass defor-
guiding his PhD thesis research program. Not having suffered mation modulus and strength of jointed hard rock masses using the GSI
enough, Garry also reviewed the draft manuscript. Though they system. International Journal of Rock Mechanics and Mining Sciences
2004;41(1):3e19.
greatly helped with the research, the errors and omissions are the Carter TG, Marinos V. Use of GSI for rock engineering design. In: Proceedings the 1st
authors alone. Many thanks also to Professor Ted Brown for his international conference on applied emprical design methods in mining. Lima,
encouraging comments and his incredible commitment to our Peru; 2014. p. 1e19.
Christensen RM. Yield functions/failure criteria for isotropic materials. Proceedings
profession. He continues to set too high a standard for mortals to of the Royal Society A: Mathematical Physical and Engineering Sciences
emulate. 1997;453(1962):1473e91.
Colmenares LB, Zoback MD. A statistical evaluation of intact rock failure criteria
constrained by polyaxial test data for five different rocks. International Journal
Appendix A. Supplementary data of Rock Mechanics and Mining Sciences 2002;39(6):695e729.
Colwell MG. Pillar design procedures and research methodologies - can there or
should there be a unified approach?. In: The second Australian ground control
Supplementary data to this article can be found online at in mining conference. Sydney: AusIMM; 2010. p. 67e77.
https://doi.org/10.1016/j.jrmge.2018.08.011. Costamagna R, Bruhns OT. A four-parameter criterion for failure of geomaterials.
Engineering Structures 2007;29(3):461e8.
Cundall PA, Pierce ME, Mas Ivars D. Quantifying the size effect of rock mass
strength. In: Potvin Y, Carter J, Dyskin A, Jeffrey R, editors. The 1st southern
References hemisphere international rock mechanics symposium. Perth: Australian Centre
for Geomechanics; 2008. p. 3e15.
Al-Ajmi AM, Zimmerman RW. Relation between the Mogi and the Coulomb failure Deere DU. Geological considerations. In: Stagg KC, Zienkiewicz OC, editors. Rock
criteria. International Journal of Rock Mechanics and Mining Sciences mechanics in engineering practice. John wiley & Sons; 1968. p. 1e20.
2005;42(3):431e9. Deere DU, Deere DW. The rock quality designation (RQD) index in practice. In:
Alejano L, Arzua J, Perez-Rey I. Effect of scale and structure on the strength and Kirkdalie L, editor. Rock classification systems for engineering purposes. Cin-
deformability of rocks. In: Hassani F, Hadjigeorgiou J, Archibald J, editors. The cinnati: American Society for Testing Materials; 1988. p. 91e101.
13th international symposium on rock mechanics. Montreal, Canada: ISRM Diederichs MS. Instability of hard rockmasses: the role of tensile damage and
Congress; 2015. ISRM-13CONGRESS-2015-190. relaxation. Waterloo, Ontario, Canada: University of Waterloo; 2000. p. 566.
Amadei B, Savage WZ. Analysis of borehole expansion and gallery tests in aniso- Diederichs MS. The 2003 Canadian Geotechnical Colloquium: mechanistic inter-
tropic rock masses. International Journal of Rock Mechanics Mining Sciences pretation and practical application of damage and spalling prediction criteria
and Geomechanics Abstracts 1991;28(5):383e96. for deep tunnelling. Canadian Geotechnical Journal 2007;44(9):1082e116.
Bandis SC. Engineering properties and characterization of rock discontinuities. In: Dinc OS, Sonmez H, Tunusluoglu C, Kasapoglu KE. A new general empirical
Hudson JA, editor. Comprehensive rock engineering. Pergamon; 1993. p. 155e approach for the prediction of rock mass strengths of soft to hard rock masses.
83. International Journal of Rock Mechanics and Mining Sciences 2011;48(4):650e
Barton N, Lien R, Lunde J. Engineering classification of rock masses for the design of 65.
tunnel support. Rock Mechanics 1974;6(4):189e236. Eberhardt E. The HoekeBrown failure criterion. Rock Mechanics and Rock Engi-
Barton N. The shear strength of rock and rock joints. International Journal of Rock neering 2012;45(6):981e8.
Mechanics and Mining Sciences and Geomechanics Abstracts 1976;13(9):225e Edelbro C. Rock mass strength: A review. Lulea, Sweden: Lulea University of
79. Technology; 2003.
Barton NR, Choubey VD. The shear strength of rock joints in theory and practice. Esterhuizen GS, Iannacchione AT, Ellenberger JL, Dolinar DR. Pillar stability issues
Rock Mechanics and Rock Engineering 1977;10(1e2):1e54. based on a survey of pillar performance in underground limestone mines. In:
Barton N, Grimstad E. Forty years with the Q-system in Norway and abroad. In: The 25th international conference on ground control in mining. Morgantown.
Fjellsprengningsteknikk, Bergmekanikk/Geoteknikk. Tapir Trondheim; 2014a. West Virginia: West Virginia University; 2006. p. 354e61.
p. 4.1e25. Fairhurst C. Fundamental considerations relating to the strength of rock. In: Col-
Barton NR, Grimstad E. An illustrated guide to the Q-system. Oslo, Norway: Nick loquium on rock fracture 1971. Instisution Bodenmechanik und Felsmechanik
Barton & Associates; 2014b. (Karlsruhe); 1971. p. 1e56.
Benz T, Schwab R, Kauther RA, Vermeer PA. A HoekeBrown criterion with intrinsic Galvin JM, Hebblewhite BK, Salamon MDG. UNSW pillar strength determinations
material strength factorization. International Journal of Rock Mechanics and for Australian and South African conditions. In: The 37th US rock mechanics
Mining Sciences 2008;45(2):210e22. symposium. Vail: NIOSH; 1999. p. 63e71.
Bertuzzi R. Sydney sandstone and shale parameters for tunnel design. Australian Galvin JM. Considerations associated with the application of the UNSW and other
Geomechanics 2014a;49(1):1e40. pillar design formulae. In: The 41st US symposium on rock mechanics. Golden,
Bertuzzi R. Sydney sandstone and shale parameters for tunnel design - Corri- Colorado: ARMA/USRMS; 2006. p391e9.
gendum 1. Australian Geomechanics 2014b;49(2):95e104. Grimstad E, Barton N. Updating of the Q-system for NMT. In: Kompen C,
Bertuzzi R. Forming a view of the rock mass strength of Hawkesbury sandstone Opsahl SL, Berg SL, editors. Proceedings of the international symposium on
from tunnelling case histories. In: Hassani FP, Hadjigeorgiou J, Archibald J, ed- sprayed concrete - modern use of wet mix sprayed concrete for underground
itors. The 13th international congress of rock mechanics. Montreal, Canada: support. Fagernes. Oslo, Norway: Norwegian Concrete Association; 1993.
ISRM; 2015. Paper 421. p. 46e66.
Bertuzzi R, Douglas K, Mostyn G. An approach to model the strength of coal pillars. Habimana J, Labiouse V, Descoeudres F. Geomechanical characterisation of cata-
International Journal of Rock Mechanics and Mining Sciences 2016a;89:165e75. clastic rocks: experience from the Cleuson-Dixence project. International
Bertuzzi R, Douglas K, Mostyn G. Comparison of intact rock strength criteria for Journal of Rock Mechanics and Mining Sciences 2002;39(6):677e93.
pragmatic design. Journal of Geotechnical and Geoenvironmental Engineering Hardy MP, Hudson JA, Fairhurst C. The failure of rock beams: Part I - theoretical
2016b;143(4):06016029. studies. International Journal of Rock Mechanics and Mining Sciences
Bertuzzi R, Douglas K, Mostyn G. Comparison of quantified and chart GSI for four 1973;10(1):53e67.
rock masses. Engineering Geology 2016c;202:24e35. Hoek E, Brown ET. Empirical strength and criterion for rock masses. Journal of
Bertuzzi R, Douglas K, Mostyn G. Improving the GSI Hoek-Brown criterion re- Geotechnical Engineering 1980;106(GT9):1013e35.
lationships. International Journal of Rock Mechanics and Mining Sciences Hoek E. Strength of rock and rock masses. ISRM News Journal 1994;2(2):4e16.
2016d;89:185e99. Hoek E, Brown ET. Practical estimates of rock mass strength. International Journal of
Bertuzzi R. Back-analysing rock mass modulus from monitoring data of two tunnels Rock Mechanics and Mining Sciences 1997;34(8):1165e86.
in Sydney, Australia. Journal of Rock Mechanics and Geotechnical Engineering Hoek E, Carranza-Torres C, Corkum B. Hoek-Brown failure criterion - 2002 edition.
2017a;9(5):877e91. In: 5th North American rock mechanics symposium. Toronto, Ontario, Canada:
Bertuzzi R. Rock mass properties for tunnelling. PhD Thesis. Sydney, Australia: University of Toronto Press; 2002. p. 267e73.
University of New South Wales (UNSW); 2017b. Hoek E. A brief history of the development of the Hoek-Brown failure criterion. 2002.
Bieniawski ZT. In situ strength and deformation characteristics of coal. Engineering https://www.rocscience.com/assets/resources/learning/hoek/The-Development-
Geology 1968;2(5):325e40. of-the-Hoek-Brown-Failure-Criterion.pdf.
Bieniawski ZT. Engineering classification of jointed rock masses. Transactions South Hoek E, Diederichs MS. Empirical estimation of rock mass modulus. International
African Institution of Civil Engineers 1973;15(2):335e42. Journal of Rock Mechanics and Mining Sciences 2006;43(2):203e15.
510 R. Bertuzzi / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 494e510

Hoek E, Carter TG, Diederichs MS. Quantification of the geological strength index Pells PJN, Bieniawski ZT, Hencher SR, Pells S. Rock quality designation (RQD): time
chart. In: 47th US rock mechanics/geomechanics symposium. San Francisco, CA, to rest in peace. Canadian Geotechnical Journal 2017;54(6):825e34.
USA: ARMA; 2013. ARMA 13-672. Perras MA, Diederichs M. A review of the tensile strength of rock: Concepts and
Hosni A, Hadadou R, Josien JP, Piguet JP, Baroudi H. Subsidence kinetics above room testing. Geotechnical and Geological Engineering 2014;32(1):525e46.
and pillar mines & development of kinetic criterion based on retro-analysis of Priest SD, Hudson JA. Discontinuity spacings in rock. International Journal of Rock
subsidence cases. In: The 13th international symposium on rock mechanics. Mechanics and Mining Sciences and Geomechanics Abstracts 1976;13(5):135e
Montreal, Canada: ISRM Congress; 2015. ISRM-13CONGRESS-2015-126. 48.
Hudson JA, Harrison JP. Engineering rock mechanics: an introduction to the prin- Read RS. Interpreting excavation-induced displacements around a tunnel in highly
ciples. Oxford, England: Pergamon; 1997. stressed granite. PhD Thesis. Winnipeg, Manitoba, Canada: University of Man-
Ismael MA, Imam HF, El-Shayeb Y. A simplified approach to directly consider intact itoba; 1994.
rock anisotropy in Hoek-Brown failure criterion. Journal of Rock Mechanics and Read SAL, Richards LR. Guidelines for use of tensile data in the calculation of the
Geotechnical Engineering 2014;6(5):486e92. Hoek-Brown constant, mi. In: The 13th international symposium on rock me-
Kwasniewski M. Mechanical behaviour of rocks under true triaxial compression chanics. Montreal, Canada: ISRM Congress; 2015. ISRM-13CONGRESS-2015-149.
conditions - A review. In: Kwasniewski M, Li X, Takahashi M, editors. True Russo G. A new rational method for calculating the GSI. Tunnelling and Under-
triaxial testing of rocks. CRC Press; 2013. p. 99e138. ground Space Technology 2009;24(1):103e11.
Lee YK, Pietruszczak S, Choi BH. Failure criteria for rocks based on smooth ap- Saroglou H, Tsiambaos G. A modified HoekeBrown failure criterion for anisotropic
proximations to MohreCoulomb and HoekeBrown failure functions. Interna- intact rock. International Journal of Rock Mechanics and Mining Sciences
tional Journal of Rock Mechanics and Mining Sciences 2012;56:146e60. 2008;45(2):223e34.
Lunder PJ. Hard rock pillar strength estimation: an applied empirical approach. MS Seedsman RW. The strength of the pillar-floor system. In: The 12th coal operators’
Thesis. British Columbia, Canada: The University of British Columbia; 1994. conference. Univeristy of Wollongong and the Australian Institute of Mining &
Marinos P, Hoek EGSI. A geologically friendly tool for rock mass strength estimation. Metallurgy; 2012. p. 23e30.
In: ISRM international symposium. International Society for Rock Mechanics; Serrano A, Estaire J, Olalla C. Extension of the Hoek-Brown failure criterion to three
2000. http://www.geoplanning.it/test/wp-content/uploads/2012/02/GSI.pdf. dimensions. In: The 11th congress of the ISRM. Leiden Taylor & Francis; 2007.
Marinos P, Hoek E, Marinos V. Variability of the engineering properties of rock p. 289e92.
masses quantified by the geological strength index: the case of ophiolites with Soder P, Krauland N. Determination of pillar strength by full scale pillar tests in the
special emphasis on tunnelling. Bulletin of Engineering Geology and the Laisvall Mine. In: Kidybinski A, Dubinski J, editors. Proceedings of the 11th
Environment 2006;65(2):129e42. plenary scientific session of the international Bureau of strata mechanics, strata
Mark C. The evolution of intelligent coal pillar design: 1981-2006. In: 25th inter- control in deep mines. Novosibirsk: Balkema; 1989. p. 39e59.
national conference on ground control in mining. Morgantown, West Virginia, Sonmez H, Ulusay R. Modifications to the geological strength index (GSI) and their
USA;; 2006. p. 325e34. applicability to stability of slopes. International Journal of Rock Mechanics and
Martin CD. The strength of massive Lac du Bonnet granite around underground Mining Sciences 1999;36(6):743e60.
opening. PhD Thesis. Winnipeg, Manitoba, Canada: University of Manitoba; Sonmez H, Ulusay R. A discussion on the Hoek-Brown failure criterion and sug-
1993. gested modifications to the criterion verified by slope stability case studies.
Medhurst TP, Brown ET. Large scale laboratory testing of coal. In: Jaksa MB, Yerbilimleri Bulletin of Earth Sciences 2002;26(1):77e99.
Kaggwa WS, Cameron DS, editors. Geomechanics in a changing world, pro- Thorpe R, Watkins DJ, Ralph WE, Hsu R, Flexser S. Strength and permeability tests
ceedings, 7th Australia e New Zealand conference on geomechanics. Adelaide, on ultra-large Stripa granite core. Berkeley, CA: Lawrence Berkeley Laboratory,
Australia: Institution of Engineers, Australia; 1996. p. 201e8. University of California; 1980.
Medhurst TP, Brown ET. A study of the mechanical behaviour of coal for pillar Yoshinaka R, Yamabe T. Strength criterion of rocks. Soils and Foundations
design. International Journal of Rock Mechanics and Mining Sciences 1980;20(4):113e26.
1998;35(8):1087e105. You M. True-triaxial strength criteria for rock. International Journal of Rock Me-
Medhurst TP. Highwall mining: practical estimates of coal-seam strength and chanics and Mining Sciences 2009;46(1):115e27.
design of slender pillars. Mining Technology IMM Transactions Section A You M. Comparison of the accuracy of some conventional triaxial strength criteria
1999;108:A161e71. for intact rock. International Journal of Rock Mechanics and Mining Sciences
Mogi K. Pressure dependence of rock strength and transition from brittle fracture to 2011;48(5):852e63.
ductile flow. Bulletin of the Earthquake Research Institute 1966;44:215e32. Zhang L, Zhu H. Three-dimensional Hoek-Brown strength criterion for rocks. Jour-
Mostyn GR, Douglas KJ. Strength of intact rock and rock masses. In: ISRM Geo- nal of Geotechnical and Geoenvironmental Engineering 2007;133(9):1128e35.
Eng2000. Melbourne, Australia: Technomic, Lancaster; 2000. p. 1389e421. Zuo J, Liu H, Li H. A theoretical derivation of the Hoek-Brown failure criterion for
Nejati HR, Ghazvinian A, Moosavi SA, Sarfarazi V. On the use of the RMR system for rock materials. Journal of Rock Mechanics and Geotechnical Engineering
estimation of rock mass deformation modulus. Bulletin Engineering Geology 2015;7(4):361e6.
and the Environment 2014;73(2):531e40.
Palmström A. RMi e a rock mass characterisation system for rock engineering
purposes (PhD Thesis). Oslo, Norway: University of Oslo; 1995. Robert Bertuzzi graduated from the University of Sydney
Palmström A, Singh R. The deformation modulus of rock masses. Tunnelling and in 1985 with a bachelor degree in mining engineering and
Underground Space Technology 2001;16(3):115e31. the University Medal. He completed a master degree at the
Palmström A. Measurements of and correlations between block size and rock University of Sydney in 1989 while working in under-
quality designation (RQD). Tunnelling and Underground Space Technology ground and open cut coal mining. His first tunnelling job
2005;20(4):362e77. was on the English Channel Tunnel with Halcrow in the
Pan XD, Hudson JA. A simplified three-dimensional Hoek-Brown yield criterion. In: early 1990s. Robert was a senior geotechnical engineer
Romana MR, editor. Rock mechanics and power plants. Rotterdam: A.A. Bal- with Coffey Partners before joining PSM in January 1994.
kema; 1988. p. 95e103. He has been there ever since. He is fortunate in having
Pells PJN. Uniaxial strength testing. In: Hudson JA, editor. Comprehensive rock worked on many tunnelling projects in Australia and is
engineering. Oxford, England: Pergamon; 1993. p. 67e85. currently involved with several rail and road tunnels in
Pells PJN, Mostyn GR, Walker BF. Foundations on sandstone and shale in the Sydney Sydney. Robert completed his PhD in 2017 at University of
region. Australian Geomechanics 1998;33(2):17e29. New South Wales where he also teaches subjects for the
Master of Engineering Science course.

Potrebbero piacerti anche