Barton, N. 2022

IX Latin American Rock Mechanics Symposium
IX Simposio Latinoamericano de Mecánica de Rocas

Asunción-Paraguay
October 16-19, 2022 KEYNOTE PAPERS
Keynote Lecture: Continuum or Discontinuum – That is the

Question
N.R. Barton
Nick Barton and Associates, Oslo, Norway
ABSTRACT:
Several decades ago there was a strong focus on the need for discontinuum modelling to im-
prove upon the empirically based analysis of excavations in jointed rock. The remarkable codes
developed by Peter Cundall: UDEC and 3DEC were put to full use in the nineteen eighties and
nineties. For example, Q-system based cavern support could be verified or improved with such
analyses. Of course, these codes preferably require knowledge of rock mechanics and rock joint
behaviour, and perhaps familiarity with non-linear constitutive models as in UDEC-BB. Regret-
tably the classic textbooks of Hoek and Bray and Hoek and Brown in this period were subse-
quently followed by the suggestions for continuum modelling using a still not finalized GSI –
there are many attempts at improved quantification. JRC now reaching 50 years is also the sub-
ject of improved quantification, but it is not followed by the extraordinary page-wide equations
for ‘c’ and ‘φ’ so no software is needed. The incorrect addition of these components of shear
strength (as indeed in Mohr-Coulomb) in commercial continuum codes is the final source of er-
ror of so many analyses. So-called plastic zones are exaggerated around tunnels, and rock slopes
are given seldom observed deep spoon-shaped failure predictions, ignoring the frequent influ-
ence of major discontinuities, and the usual failures within the slope faces. Of course, lake-bed
open-pit slope deposits or extremely weathered rock will give spoon-shaped failures as for rock-
fill and soil, but competent jointed rock will not fail like this: major discontinuities will usually
be involved, and wedge or planar failures will be the usual reality.
1 INTRODUCTION
We were advised more than 50 years ago by Brace and Müller that cohesion is broken before
friction is fully mobilized. Gross errors are caused by adding these components of shear strength
when estimating the maximum height of cliffs and mountain walls. Since ‘c’ is not the lowest
component of strength, artificially lowered estimates are needed, or tensile strength and Pois-
son’s ratio are used (Barton and Shen, 2017). There is precious little empirical basis for the
Hoek-Brown equations for rock mass strength, but an excellent experimental basis of course for
the earlier intact rock H-B criterion. We may ask if is it logical to downgrade the strength of in-
tact rock to model rock masses (using opaque equations with joint roughness and number of
joint sets ignored) or better to apply the equations for the shear strength of joints and fractures
and estimate the initial cohesive contribution of intact bridges between the kinematically capa-
ble joint sets? In this lecture the author will be showing studies with UDEC, 3DEC, FLAC and
FLAC3D and FRACOD, and will be illustrating both discontinuum and continuum analyses for
tunnels, caverns and open-pit slopes. An earlier than UDEC phase, with fractured (2D) models
of underground excavations, will also be shown as an introduction.
2 PHYSICAL MODELS WITH FRACTURES PRE UDEC
The author was a student colleague of Cundall while he started to develop his first remarkable
computer code with blocks. However, we may start with what could be achieved with physical
Asunción-Paraguay
models of fractured media before Cundall’s codes became available, both pre-1970 and from
just prior to the Cundall/Itasca UDEC release in 1980. He was inspired to do better, as in Fig. 1.
Figure 1. The contrasting flexibility of Cundall’s intelligent computer code µDEC: two of four results of
varying angle φ from Cundall et al. (1975). A big contrast to the ‘fixed-fracture-sets’ fractured 2D models
developed some years earlier by the author in 1968. These physical models were utilized on two occa-
sions before UDEC was released. Several 2D ‘slab models’ with carefully loaded and excavated slope
models of 40,000 blocks were followed some years later with underground opening models with 20,000
blocks (Figs 2, 3 & 4). Individual slab models also allowed investigation of the relative effects of 4,000,
1,000 and 250 blocks in biaxial loading, which assisted in block-size scale effect understanding. The
smallest block sizes gave unexpected ‘linear’ stress-strain behaviour and some kink-band modes.
Figure 2. The principal mechanism of tension-fracture development and their roughness are illustrated top
left. A subsequent physical model series reported by Barton & Hansteen (1979) is shown on the right.
(My NGI colleague provided the continuum model comparisons). One year after publication of these
UNPP (underground nuclear power plant) studies, UDEC became commercially available. No regrets.
Some of these early pre-UDEC physical fracture models are reproduced in this lecture as they
still appear not to have been repeated by others and represent a ‘physical reality’ even though
not a ‘geological reality’. They may serve as 2D code verification objects. In the end the early
UNPP studies in Norway including site characterization, came to nothing, and Sweden (who
had part-funded these physical model studies through BeFo) built surface nuclear plants, while
Norway continued with more extensive underground hydroelectric power developments.
Asunción-Paraguay
Figure 3. A selection of the cavern models which were studied in the pre-UDEC period, in order to evalu-
ate the feasibility of underground nuclear power plant (UNPP) spans of 50m. The deformation vectors
show the importance of horizontal stress and demonstrate the inevitable inadequacy of isotropic continu-
um modelling. The inset joint (fracture) structures are at correct scale and show primary and secondary
fracture sets. All models had 20,000 blocks, meaning distant boundaries. Barton & Hansteen (1979).
3 SOME FUNDAMENTALS OF JOINT BEHAVIOUR
The physical fracture models suggested significant differences to continuum modelling, with
joint orientations being particularly important. In Figure 4 we can also see a more fundamental
feature of joint behavior. The four caverns were excavated one after the other in a high horizon-
tal stress situation, as in four of the cases in Figure 3. Of particular note is the hysteresis or ‘de-
formation set’ seen in the narrow pillars. The photogrammetry-determined deformations did not
reverse noticeably, when a new cavern was excavated, as would have happened in an elastic
continuum model.
Figure 4. A multi-cavern experiment with ultra-narrow pillars, prior to model earthquake loading moni-
tored with an accelerometer (giving a scaled 0.1 to 0.7g). Note the hysteresis in the cavern pillars. No re-
versal of deformations with successive cavern excavation. Fracture roughness and dilation explain the ap-
parent tensile strength. Note that primary (continuous) fracture set #1 dips to the right, and has no
Asunción-Paraguay
cohesive strength, just very high dilation and total friction angles at low normal stress. (Barton, 1973).
Figure 2, bottom right, shows the cavern degradation progress during the model ‘earthquake’.
In the next figure we see that three basic styles of rock mass deformation may occur with joint-
ed rock masses, and each have been recorded by large scale in situ loading tests. They are clear-
ly absent from conventional continuum modelling.
Figure 5. The N (normal) Type A concave deformation and the S (shear) Type C convex deformation,
with the combined mode Type B as seen in loading tests across columnar basalt. Chryssanthakis et al.
(1991) matched these three styles of deformation using UDEC-BB after this code became available in
1985. Note the normal stress-closure curves from Bandis et al. (1983) and the shear stress-displacement
curves (with scale effects) from Bandis et al. (1981). The N and S and N+S load-deformation trends and
the three rock mass assemblies are from Barton (1986).
To state the obvious there are no continuum model or GSI-based Hoek-Brown equations to
match the essentially discontinuous behaviour shown in any of the previous figures, nor the
shear strength details seen in Figure 6. GSI users can perhaps be happy that their ‘rock engineer-
ing’ activities are significantly easier thanks to RocScience software. But was the practice of
rock mechanics for rock engineering supposed to be so easy? In the opinion of the author the
use of GSI and continuum models has taken the realism out of the subject, which was not the
original intention of Hoek and Bray and Hoek and Brown. These were classically helpful text-
books that also addressed discontinuum behaviour, the first one especially so.
Asunción-Paraguay
Figure 6. Variations in the shear strength of unfilled, clay-coated, and clay-filled discontinuities as illus-
trated here from the two Q-system parameters Jr/Ja cannot be included in GSI in a ‘free manner’ inde-
pendent of joint roughness, because the GSI use of Bieniawski’s joint condition factor (along the x-axis)
does not make such differentiation possible. The Bieniawski (1989) ‘joint condition’ moves in one direc-
tion, so rougher joints cannot have clay coatings or fillings, and even more remarkably: slickensides must
be weathered. Figure from Barton (2002).
4 SOME EXAMPLES OF UDEC-BB MODELS SHOWING GEOLOGICAL DETAIL
In the following rather concentrated selection of discontinuum models the author will
show features of potential behaviour that can hardly be matched by continuum model-
ling. The exception – to a degree – is the redistribution of stress, from principle far-field
stresses to tangential and radial stresses. The rotation of the small ‘crosses’ are of course
seen clearly in continuum models, but the loading of bolts where they cross and support
wedges, and the unstressed unstable wedges are not seen in continuum models. The dis-
continuum model teaches us to look out for important behaviour that is absent from con-
tinuum modelling. In UDEC-BB we see the distribution of joint shearing and also the
stress- or depth-dependent physical apertures, and JRC-estimated hydraulic apertures,
for coupled H-M modelling. An introductory set of results with input is shown in Fig. 7.
Asunción-Paraguay
Figure 7. A glimpse of the principles of discontinuum modelling input data operations as applying to
UDEC-BB. On the left: joint orientations and frequencies, permeability and aperture estimation (but this
will be estimated by UDEC-BB) and joint index test results for JRC, JCS and φr. Boxes 5 and 6 symbol-
ize the shear stress-displacement-dilation-permeability and normal stress-closure operations of the Bar-
ton-Bandis model. (See Barton et al., 1985, Barton, 1993 and Barton and Bandis, 2017 for examples).
Asunción-Paraguay
Figure 8. Selected UDEC-BB studies of a 250m deep section and a 700m deep section of a planned TBM
spiral access tunnel through the interbedded St Bees sandstones and siltstones (top) and through the Bor-
rowdale welded tuffs-ignimbrites (bottom) at the planned Sellafield LLW/ILW nuclear waste repository,
from 1992. Thanks are due to NGI’s numerical modelling team, during our 6 years geotechnical consul-
tancy for UK Nirex, 1990-1996 which the author had the privilege to project-manage. All names later.
Figure 9. The discontinuum-modelling advantages of representing adversely steeply-dipping bedding.

This late 1980’s UDEC-BB model of a tunnel in Japan showed distinct signs of over-loading of the bolt-
ing. The bonding failure of the modelled S(fr) is also shown, bottom left.
Asunción-Paraguay
Figure 10. Top left: UDEC-BB to check TBM pilot tunnel break-out to full section. The 3DEC sections
show a maximum tangential stress of 7.9 MPa and a maximum displacement of 8.0mm. This motorway
project study is illustrated further in the next figure.
Figure 11. The unusual motorway pilot-TBM for the Tokyo-Osaka Tomei 2 Shimizu 3 motorway project.
The specially designed TBM could turn so as to provide a pilot bore for the parallel tube. It proved im-
portant to ‘turn’ prior to reaching highly weathered rock. Dr Juhn Itoh of the Fuji Research Institute was
responsible for this interesting NGI contract from Japan. NGI UDEC-BB and 3DEC modellers from the
late 1980’s and early 1990’s included Dr Mark Christianson (Itasca/NGI) who was responsible for putting
BB and UDEC ‘together’ in 1985 (with strong pleading to NGI’s director Kaare Høeg from Barton and
Bandis), Dr Axel Makurat (later Shell), Linda Hårvik, Dr Marte Gutierrez (later distinguished US profes-
sor), Dr Karstein Monsen (later Geoscan), Panos Chryssanthakis (later COWI), Dr Harald Hansteen, Dr
Asunción-Paraguay
Baotang Shen (later CSIRO in Australia), Lise Bacher, Dr Rajinder Bhasin, Dr Lloyd Tunbridge – and
since then several more.
A particularly notable discontinuum modelling contract for NGI was the design check for the
designers of the 62m span Gjøvik Olympic cavern. This was done on behalf of owner Fortifi-
kasjon, whose director Jan Rygh had drawn the ‘classic serviette sketch’ for the town’s chief
planning engineer some years earlier, before the winter games were finally awarded to nearby
Lillehammer, and ice hockey was played in this unique cavern. Modelled UDEC-BB defor-
mation 7 to 9mm, MPBX recorded deformation a constant 7 to 8mm after 5 years of monitoring.
Figure 12. Top: Horizontal stress sensitivity studies conducted by Lloyd Tunbridge, and bottom: subse-
quently more detailed UDEC-BB modelling by Chryssanthakis, for the 62m span Gjøvik cavern. Note
multi-author reference due also to Q-logging, cross-hole seismic tomography and rock stress measure-
ment by the NGI team. Barton, Chryssanthakis, Tunbridge, Kristiansen, Løset, Bhasin, Westerdahl, Vik
(1994).
Asunción-Paraguay
4 OVERBREAK AND TUNNEL AND STOPE STABILITY
A feature of excavation effects in rock masses that particularly marks the stark difference be-
tween continuum and discontinuum modelling is overbreak. In Barton (2007) it was suggested
that this could be quantified by the Q-parameters ratio Jn/Jr. Figure 13 (top photos and rough-
ness sketches) shows how this works. The red- coloured ratios show Jn/Jr < 6 (no overbreak ex-
pected) while the blue-coloured ratios show Jn/Jr ≥ 6 which is the suggested criterion. Extreme-
value Q-parameter statistics at the LKAB mine explain the overbreak experienced in drilling
drifts. Elevated values of Ja (clay coatings) obviously add to the likelihood of overbreak.
Figure 13. Top: the Q-parameter ratio Jn/Jr ≥ 6 suggests almost unavoidable overbreak if there are suffi-
cient degrees of freedom for block fall-out (higher Jn) and if joint roughness is low enough (low Jr). Bot-
tom: observations performed by the writer in 1988 in LKAB’s ‘Oscar’ long-hole drilling stope-
development project in Kiruna in northern Sweden. Explanations for the excessive overbreak were
sought. This was the first time that Q-parameter histograms were used by the author, and soon this be-
came a systematic way of logging individual Q parameters. Note the potentially serious large-scale J1/J2
deformation mechanism in the future stopes (shown in yellow). Problematic details for continuum model-
ers, or an interesting expansion of their possibilities to influence design decisions – in many areas.
Asunción-Paraguay
Figure 14. This ‘graphical matrix’ scheme developed by Dr Fraser of CSIRO, was shown to the writer
during a two-days rock engineering course given in Melbourne in 2005. The subject matter is stoping for
mineral extraction, and stable, caving, failure, and massive failure are the operative terms. The relative in-
fluence of various separate and combined Q-parameters are taken from the Potvin scheme of utilizing Q’
= RQD/Jn x Jr/Ja in mine-stope dimensioning. ‘Caving’ is seen bottom-left. Jn and Ja alone, and Jn/Jr are
seen to be important. They are absent in continuum modelling schemes.
5 INPUT DATA NEEDS FOR DISCONTINUUM MODELS ARE MORE THAN GSI
Figure 15. The six Q-parameters (seen as histograms on the left) that are used in characterizing rock
masses and tunnelling conditions more fully are obviously just part of the comprehensive data needed for
a thorough engineering geological description. Serious consultants do not / should not put a ‘blob’ on a
GSI chart and be satisfied that it will give the required input data for their (‘black-box’) modelling.
Asunción-Paraguay
Three of the most commonly used shear strength criteria for rock joints are illustrated in Fig-
ure 16. The classic Mohr-Coulomb involving a cohesion intercept ‘c’ and a stress-independent
friction angle ‘φ’ was not developed from testing rock joints, which actually do not display ac-
tual cohesion intercepts unless very steep (60-90º) steps are present in the joint plane, due for
instance to cross-joint influence. An improvement for lower stress was suggested by Patton,
1966 with his well-known ‘i-values’ where ‘φ + i’ will be the lower-stress estimate. The prob-
lem of course is which combination of ‘i-values’ one should use. This was the impetus for a new
student in 1966, and the resulting parameters described below took some years to materialize.
Figure 16. Mohr-Coulomb, Patton (1966) and Barton (1973) shear strength criteria for rock joints. The
first version of the JRC, JCS, φr criterion (in 1967) was developed from DST tests on 200 tension frac-
tures. One is shown in Figure 2: with respective values of 20, UCS (since no weathering) and 30º.
Figure 17. Shear box, tilt test, Schmidt hammer and roughness profiling methods (including a/L) used for
evaluating the JRC, JCS and φr data for UDEC-BB and other discontinuum modelling projects.
Asunción-Paraguay
In the case of NGI’s Sellafield project for UK Nirex (1990-1996) some 10 kilometers of core
were logged and numerous joint samples representing each joint set were prepared. Besides tun-
nel models (Figure 8) a large number of deep cavern models in welded tuff (ignimbrite) were al-
so performed using UDEC-BB to check on self-supporting abilities and EDZ depths. Note that
multi-stage shear testing of the same joint sample at increasing normal stress causes rotation of
the shear strength envelope. (Barton, 2013). This gives artificial ‘life support’ for ‘c’ and also
gives reduced ‘φ’. (Increased open pit costs if ‘c’ is ignored and set to ‘zero’ with ‘φ’ too low).
Figure 18. Left: examples of joint profiling and tilt tests. From Barton and Choubey (1977). Note that the
profile gage had four ‘shims’ per millimeter. There were no steps as some critics have assumed. Right:
tilt-tested joint samples (at correct angles) from TerraTek’s heated block test in Colorado. Barton, 1982.
6 WHEN CONTINUUM MODELLING CAN BE THE ONLY CHOICE
A suggested scheme for choosing which type of computer modelling is appropriate for which
type of rock mass, is shown in Figure 19. In the earlier examples of UDEC-BB given in this pa-
per, we were well within the suggested Q-range of 0.1 (very poor) to 100 (very good). It was
possible for engineering geologists (in the case of the UK Nirex project it was often Fredrik
Løset) to suggest a realistic and representative joint pattern, which was digitized so that it could
be reproduced as perhaps one of several UDEC model geometries. One can ‘move models
around’ in a big jointing ‘template’ and include faults in some models.
Figure 19. A scheme for helping in the selection of appropriate computer models, based on a ‘degree-of-
fracturing’ scheme with suggested Q-value scale for UDEC and 3DEC. Continuum at either end implied.
Asunción-Paraguay
If the rock mass is highly fractured as symbolized in the left side of Figure 19, then a contin-
uum approach may be needed. Logic would suggest that Q is stretched to 0.01 or worse to the
left side, and to an almost unjointed Q = 500 to 1000 on the right side. However, care is needed
in assuming that GSI and Hoek-Brown equations of assumed rock mass behaviour will give rel-
evant results. There is unfortunately evidence of grossly exaggerated ‘plastic zone’ prediction,
as revealed in an international court case some years ago. This can hardly be due to the com-
mercial FEM method used at that time though it might be partly due to adding the strength
components related with ‘c’ and ‘φ’. (c + σn’ tan φ).
Figure 20. Comparison of FLAC and UDEC-BB by a former NGI colleague Lise Backer (now BaneNor).
The same triple-tunnel motorway is modelled using the same boundary stresses and the same intact-block
moduli. The jointing makes the UDEC-BB model significantly more relevant than the continuum model.
Anisotropy and pillar distress is also seen.
In Figure 20 a continuum model using FLAC is compared with a UDEC-BB model. In this
particular case low Q-values are not being modelled, so the deformation is small in both cases,
but the UDEC-BB model teaches us much more than the continuum FLAC model.
In the next figure we see something that was a great surprise to the present author when first
seen in an SKB document with Prof Derek Martin as lead author. The date was 2002 or 2003.
Figure 21 shows part of the PhD study by an Iranian author: Dr. Hajiabdolmajid. In his 2000
publication he is supported by his presumed internal and external supervisors Prof Derek Martin
and Dr Peter Kaiser, both well known specialists in rock engineering and mining engineering.
The results they present, reproduced in Figures 21 and 22 were very important, but a huge ma-
jority of people working in rock mechanics have ignored the message and continue with GSI,
Hoek-Brown equations and FEM continuum software with ‘c + σn tan φ’.
Asunción-Paraguay
Figure 21. Mohr-Coulomb tri- and bi-linear elastic-brittle and elastic-plastic models of the well-known
URL (Manitoba) line-drilled mine-by tunnel. Magnitudes of the input parameters came from GSI and H-
B estimation. The objective of the modelling was to show the great improvement that results if the cohe-
sion is degraded while friction is mobilized, shown in Figure 22. The Mohr-Coulomb results are surpris-
ingly unrealistic, yet the criterion is so widely used. It is remarkable that so much incorrect modelling is
performed. The three authors are listed here Hajiabdolmajid, Martin and Kaiser (2000).
Figure 22. A greatly improved result using a CWFH (cohesion weakening friction hardening) approach.
Hajiabdolmajid et al. (2000). The over-stressed volumes do not fall out as the analysis is still a continuum
solution, but the reality of location is convincing. The author and a colleague (Barton and Pandey, 2011)
also used this CWFH approach in mine stope modelling using Q-based input data CC and FC (see Barton,
2002) in FLAC3D models of pre-instrumented stope areas.
IX Latin American-8 Rock
-7
Mechanics
-6 -5 -4
Symposium
-3 -2 -1
X Axis (m)
0 1 2 3 4 5 6 7
Pxx: 0E+0 Pyy: 0E+0
Asunción-Paraguay
7
Pxy: 0E+0
Max. Compres. Stress: 9.64219E+7
7
October 16-19, 2022

6
KEYNOTE PAPERS
Max. Tensile Stress: 1.41005E+7
Elastic fracture 6
Open fracture
Slipping fracture
5 Fracture with Water
5
Compressive stress
Tensile stress 4.8
4 4
Fracom Ltd 4.6
3 Date: 23/12/2004 10:34:30 3
4.4
2 2 4.2
4.0
1 1
3.8
Y Axis (m)
Y Axis (m)
0 0 3.6
3.4
-1 -1
3.2
-2 -2 3.0
2.8
-3 -3
2.6
-4 -4
2.4
Figure 23. The BEM-5 fracture mechanics code FRACOD developed by Shen et al. (2013). -5 This shows a 2.2
realistic development
-6
of log-spiral failure surfaces when jointing is sparse (the right-hand
-6
side of Figure 2.0
19). Red represents low, and green represents high factor of safety against further fracturing. In situ stress 1.8
levels were > 40%-7 of UCS. These were scoping studies for the sub-Andean Olmos tunnel -7 where there 1.6
were rock-burst challenges, since a ‘stress-raising’ TBM had to be used in the new contract.
Tunnel failure mechanisms
-8 -8
Shen 2004. 8
-8 -7 -6 -5 -4 -3 -2 -1
X Axis (m)
0 1 2 3
1.4
4 5 6 7 8
8
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
X Axis (m) 7 7
6 6
5 5
4 4
3 3
2 2
1 1
Y Axis (m)
Y Axis (m)
0 0
-1 -1
-2 -2
-3 -3
-4 -4
Tunnel failure mechanisms
X Axis (m)
-5 -5
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
8 8
-6 -6
7 7
-7 -7
6 6
-8 -8
5 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 5
X Axis (m)
4 4
3 3
2 2
1 1
Y Axis (m)
Y Axis (m)
0 0
-1 -1
-2 -2
-3 -3
-4 -4
-5 -5
-6 -6
-7 -7
-8 -8
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
Figure 24. Left: FRACOD BEM fracture mechanics model of failure in a TBM tunnel (the English/Old X Axis (m)
Beaumont tunnel). Centre: physical model of a bored tunnel (or well) in a 3D loaded block showing inter-
secting log-spiral shear surfaces due to inclined boring. Right: FRACOD models of 1,000m deep TBM
tunnel in massive or jointed rock. Shen and Barton, 2018. This ‘starting as a continuum’ model
(FRACOD) is of course in an elevated category in relation to the usual ‘continuously’ continuum models.
7 A CRITIQUE OF GSI, HOEK-BROWN EQUATIONS, AND THEREFORE CONTINUUM

ANALYSES THAT DEPEND ON THEM
In the opinion of the author it is rather remarkable and somewhat discouraging for the past
and future of our subject ‘rock engineering’, that so many young (and not so young) people
have adopted the RocScience-promoted and of course Hoek-promoted GSI, with the associated
Hoek-Brown equations. The following three pages (by others) came as a shock to this author.
Asunción-Paraguay
Figure 25. The GSI-rating selection scheme based on a brief encounter with ‘geology’ (the sketches) is
followed by what could truly be called black-box modelling. Where does the number of joint sets actually
appear? Why does slickensiding have to be ‘highly weathered? Why cannot rougher joints have clay fill-
ing and be weathered? GSI is actually the least logical and least ‘geological’ method in use in rock engi-
neering, and the sequence of Hoek-Brown equations are much more a priori than a posteriori (Barton,
2011). Yet one researcher recently expressed the opinion that finally one has a ‘geological’ classification
method. The reality is that any chance of ‘geology’ is immediately smoothed in a ‘continuum blender’
giving different viscosities i.e. shear strength. The result is not a geological representation of rock masses.
(Hoek et al. 2002, Renani & Cai, 2021).
Asunción-Paraguay
Figure 26. The widely varying deformation modulus estimates connected with GSI. From: Renani & Cai
(2021): Forty-Year Review of the Hoek–Brown Failure Criterion for Jointed Rock Masses. One may seri-
ously ask how continuum modellers are choosing deformation modulus. Perhaps by waiting for the results
of the modelling and adjusting the disturbance factor D? Several private communications have stated
such, and if this is so it seems hardly defensible for presenting to a client.
Figure 27. Surprisingly creative additions for ‘better’ quantifying GSI. Renani and Cai, 2021. The better-
known methods are listed in Table 1. It is a somewhat surprising list. With all the proposed improvements
what happens to the predicted H-B shear strength results? Is this a form of empiricism or just guessing?
Asunción-Paraguay
Table 1. A selection of some equations used to improve the quantification of GSI. Why a published and
widely used method should require so much improvement, or indeed be used by so many with such ap-
parent confidence, is not clear. Ván & Vásárhelyi (2014).
GSI1 = RMR – 5 = R1+R2+R3+R4+R5(=15) – 5 (1)

GSI2 = 1.5 R4 + 0.5 RQD (2)
8 EXAMPLES OF ROCK SLOPES: REAL, DISCONTINUUM, CONTINUUM
With the exception of highly trafficked road-cuttings and railway-cuttings, and steep urban
slopes above busy pavements, there will generally not be much enthusiasm for bolting or cable
anchors for ensuring rock slope stability in general. This is because there are millions and mil-
lions of rock cuttings and they cover a huge area all around the world. Of course, this support
reluctance will not apply to some critical high-dam valley slopes as especially seen in deep and
already naturally over-steepened valleys in China,
Temporary and more permanent open-pit mine slopes suffer the same no-bolting fate for ob-
vious economic reasons. Since there may be no rock bolts, the alternative selection of stable
slope angles will usually have to depend on the shear strength properties of the local rock joints,
and especially joint orientations. This is the philosophy behind Qslope. The first client wanted a
20km dam access road in a steep valley without bolts. (Barton and Bar, 2015). Some five hun-
dred design-chart case records mostly collected by Bar are shown later. (Bar and Barton, 2017).
Of more concern in this final section is the misinformation promoted by commercial software
companies, who hold courses showing the popular but hardly existing curved (‘spoon-shaped’)
potential failure surfaces that are relevant for soil slopes and rockfill, but not for slopes in joint-
ed rock unless this is extremely weak. (In fact, in about 1968 the author had sketched a curved
failure surface for completely disintegrated slates in an open pit, but such weakness is the ex-
ception as most of our civil and mining slopes are in jointed rock with enough strength to cause
deformation along discontinuities as opposed to ignoring their existence.) It seems that the as-
sumed ‘rock mechanics’ modelling performed by hundreds or perhaps thousands of young engi-
neers has actually been based on old-fashioned soil mechanics continuum principles, and such is
regrettably the case when adding c and σn tan φ as in misleading Mohr-Coulomb and Hoek-
Brown methods. Figure 28 shows some alternatives that address discontinuities, whether joints
or filled discontinuities or fault planes., from Hoek and Bray (1974) and Barton (1971).
Asunción-Paraguay
Figure 28. Reality and representation – as discontinua. Central right diagram: Hoek and Bray.
Asunción-Paraguay
Figure 29. The Qslope method is based on Q parameters with almost unchanged RQD, Jn, Jr and Ja. How-
ever wedge stability is considered so pairs of Jr/Ja apply to different sides of potential wedges, with rele-
vant orientation weightings. Since slopes are under lower stress than many tunnels, SRF has three appro-
priate categories, and Jw is now called Jwice. The relative effects of tropical rain and ice-wedging can be
roughly assessed. Co-author Neil Bar has been a very active collector of case records. Barton & Bar
(2015), Bar & Barton (2017). Qslope : 0.01, 0.1, 1, 10, 100. Critical slopes: 25°, 45°, 65°, 85°, ‘half tun-
nels’ meaning over-hang is possible. Not advised but likely to be possible due to cohesive strength.
Returning to continuum models, according to Styles & Vakili (2020), the recently developed
Improved Unified Constitutive Model (IUCM) ‘takes the best parts of Mohr-Coulomb and
Hoek-Brown, with integrated confinement-controlled softening/hardening of cohesion and fric-
tion, confinement-controlled changes in the dilation angle, and porosity-controlled modulus sof-
tening’. However, the slope failure predictions shown in Figure 30 represent something usually
only seen in rockfill or hard soil. The ’spoon’ mode shown apparently has a limited actual rela-
tion to geology or jointed/faulted rock. Open pit slopes fortunately seldom if ever fail like this.
Figure 30. FLAC3D slice of a 200m high open-pit slope, using the IUCM constitutive model. Such
spoon-shaped failure is possible in rockfill and in hard soil, but the writer does not believe it has been
seen in competent jointed and sometimes faulted rock. Perhaps with this relatively low UCS = 20MPa in-
put, such an unusual (for rock) failure mode is possible. Styles & Vakili (2020).
Asunción-Paraguay
The circular failure beliefs of Mohr-Coulomb and Hoek-Brown modelers, with Carranza-
Torres (2021) a presumed chief contributor and probably largely responsible for the smooth
curves of the H-B rockmass criterion of Hoek et al. (2002) and the associated most complex
equations, has given a very detailed recent treatise concerning ‘rock slope’ stability. Figure 31 is
his clear but apparently misleading introduction, as it does not and cannot correspond to slope
failure in competent, jointed, and perhaps faulted rock because the UCS and actual cohesion are
too high in relation to the artificial ‘lowering’ in the H-B shear strength beliefs.
Figure 31 According to Carranza-Torres, 2021 this figure represents ‘a section of a rock slope in a rock
mass that obeys the Mohr-Coulomb or Hoek-Brown failure criteria’. Slopes in competent jointed rock
cannot actually fail as in this figure. A continuum myth is presented, only valid for very weak saprolite or
soil-like material.
Figure 32. Further ‘circular’ failure analyses using both FLAC and SLIDE computer modelling, apparent-
ly with input of c and φ derived from combined GSI and Hoek-Brown. Carranza-Torres (2021). These are
basically resembling failure in hard soil, perhaps also very weak rock like saprolite, and rockfill.
Asunción-Paraguay
Figure 33 The ‘elegance’ of the relative ‘c’ or ‘φ’ dominance given by Carranza-Torres (2021) in his
many impressive ‘design charts’ cannot be denied. The relative depths of the various spoon-shaped pre-
dicted failures are interesting. But they do not apply to competent jointed rock. Joints sets, intact bridges,
and possible faults are each going to bring back reality.
Figure 34. Four major open-pit failures reproduced by Carranza-Torres (2021).

Asunción-Paraguay
The four major open-pit failures shown in Figure 34 were reproduced again by the present au-
thor, but for different reasons compared to Carranza-Torres. Cases a) and b) are strictly ‘inter-
nal’ failures that hardly extend to the crest of the slopes. Only case b) looks like jointed rock.
Case c) is lake-bed sediments and saprolite, so not jointed rock. Case d) which certainly extends
beyond the pit crest is actually an example of a wedge failure (Lorig et al. 2009), as was also the
case for the record Bingham Canyon mine failure shown in Figure 35.
Case d) was a major slope failure in northern Chile. Looking closer to the source, Chapter 10
of the invaluable ‘Guidelines for Open Pit Slope Design edited and partly written by John Read
and Peter Stacey, we can see reference to the failure as a ‘wedge failure’ example (Lorig, Stacey
and Read, Chapter 10: their Figure 10.12 caption reads: ‘Wedge failure disrupting the entire
slope’. So it is not an example of ‘circular’ (or ‘spoon-shaped’) failure. An even larger (and
world-record) failure involving a major planar discontinuity is shown in Figure 35.
Figure 35 Bingham Canyon mine with its 70,000,000 m3 failure. From top-to-toe the slide measured a
surprising 3km, and the debris looks like a frozen ‘liquid’. The adverse (tangential stress eliminating)
‘nose’ and a major discontinuity (a fault?) of many 100’s of meters length seem to be the main culprits.
9 CONCLUSIONS
1. In this lecture the writer has suggested ‘putting back jointing and geology’ into the
practice of rock mechanics modelling and understanding. Following the limitations but
useful lessons from physical fracture models, a number of UDEC-BB models were re-
Asunción-Paraguay
produced to emphasize the possibility of representing some details of ‘geology and

structure’ at least in 2D and occasionally in the more demanding 3DEC.
2. GSI with the ‘G’ in its initials has actually removed the geology from our subject,
once users leave the six ‘sketches’ and progress to the Hoek-Brown equations and
commercial software with the impressive but misleading curves of shear strength. Re-
grettably the uniquely opaque ‘black-box’ analyses effectively become homogenized as
in a food blender, but with variable viscosities. Continuum behaviour is clearly mis-
leading many and, in the opinion of this author, is hardly to be considered as rock engi-
neering or rock mechanics since missing engineering geology input. Looking at sketch-
es and guessing a representative ‘joint condition’ (and modulus) is not defensible.
3. We have been warned for more than 5 decades – historians in our subject would claim
even longer – that we should not add ‘c’ and the frictional component of strength based
on ‘φ’ (σn tan φ). At the very least it should be ‘c then sigma n tan phi’. This separation
in time and deformation is logical and urgently needed.
4. A newer proposal for a more comprehensive (progressive) failure criterion is the CcSs
‘crack’ ‘crunch’ ‘scrape’ ‘swoosh’ concept applied in Barton (2021). This has the abil-
ity to add the progressive contributions of the broken cohesion of rock bridges, the re-
sulting rough, fresh fracture surfaces, the joints (if one or more sets are suitably orient-
ed), and finally the lower resistance of faults or clay-filled discontinuities.
10 REFERENCES
Bandis, S. C., Lumsden, A. & Barton, N. 1981. Experimental studies of scale effects on the shear be-
haviour of rock joints. Int. J. of Rock Mech. Min. Sci. and Geomech. Abstr. 18, 1-21.
Bandis, S. C., Lumsden, A.C. & Barton, N. 1983. Fundamentals of rock joint deformation. Int. J. Rock
Mech. Min. Sci. and Geomech. Abstr. Vol. 20: 6: 249-268.
Bar, N. and N. Barton, 2017. The Q-slope method for rock slope engineering. Rock Mechanics and
Rock Engineering. Springer (2017) 50: 3307-3322.
Barton, N. 1971. A model study of the behaviour of steep excavated rock slopes. Ph.D. Thesis, Univ. of
London.
Barton, N. 1971. Progressive failure of excavated rock slopes. 13th US Symp. of Rock Mech. Stability
of rock slopes, Illinois, 139-170.
Barton, N. 1973. Review of a new shear strength criterion for rock joints, Engineering Geology, Else-
vier, Amsterdam, Vol. 7, 287-332.
Barton, N. & Hansteen, H. 1979. Very large span openings at shallow depth: Deformation magnitudes
from jointed models and F.E. analysis. 4th RETC; Atlanta Georgia, Vol. 2: 1131-1353. Eds Maevis
and Hustrulid. Am. Inst. of Min. &, Metall., and Petr. Engrs, Inc. New York.
Barton, N. 1982. Modelling rock joint behaviour from in situ block tests: Implications for nuclear
waste repository design. Office of Nuclear Waste Isolation, Columbus, OH, 96 p., ONWI-308, Sep-
tember 1982.
Barton, N., Bandis, S. & Bakhtar, K. 1985. Strength, deformation and conductivity coupling of rock
joints. Int. J. Rock Mech. & Min. Sci. & Geomech. Abstr. 22: 3: 121-140.
Barton, N. 1986. Deformation phenomena in jointed rock. 8th Laurits Bjerrum Memorial Lecture, Os-
lo. Geotechnique, Vol. 36: 2: 147-167.
Barton, N. 1993. Physical and discrete element models of excavation and failure in jointed rock. Key-
note lecture, ISRM Int. Symp. on Assessment and Prevention of Failure Phenomena in Rock Engi-
neering, 5-7 April, Istanbul, Turkey.
Barton, N., By, T.L., Chryssanthakis, P., Tunbridge, L., Kristiansen, J., Løset, F., Bhasin, R.K., West-
erdahl, H. & Vik, G. 1994. Predicted and measured performance of the 62m span Norwegian
Olympic Ice Hockey Cavern at Gjøvik. Int. J. Rock Mech, Min. Sci. & Geomech. Abstr. 31:6: 617-
641.
Asunción-Paraguay
Barton, N. 1999. General report concerning some 20th Century lessons and 21st Century challenges in
applied rock mechanics, safety and control of the environment. Proc. of 9th ISRM Congress, Paris,
3: 1659-1679, Balkema, Netherlands.
Barton, N. 2002. Some new Q-value correlations to assist in site characterization and tunnel design.
Int. J. Rock Mech. & Min. Sci. Vol. 39/2:185-216.
Barton, N. 2007. Near-surface gradients of rock quality, deformation modulus, Vp and Qp to 1km
depth. First Break, EAGE, October, 2007, Vol. 25, 53-60.
Barton N. 2011. From empiricism, through theory, to problem solving in rock engineering. ISRM
Cong., Beijing. 6th Müller Lecture. Proceedings. Harmonising Rock Engineering and the Environ-
ment, (Qian & Zhou ed), Beijing Taylor & Francis, (1):3-14.
Barton, N. and S.K. Pandey, 2011. Numerical modelling of two stoping methods in two Indian mines
using degradation of c and mobilization of φ based on Q-parameters. Int. J. Rock Mech. & Min.
Sci., 48,7: 1095-1112.
Barton, N. 2013. Shear strength criteria for rock, rock joints, rockfill and rock masses: problems and
some solutions. J. of Rock Mech. & Geotech. Engr., Wuhan, Elsevier 5(2013) 249-261.
Barton N and N. Bar. 2015. Introducing the Q-slope method and its intended use within civil and min-
ing engineering projects. In Schubert & Kluckner (ed.), Future Development of Rock Mechanics;
Proc. ISRM symp. Eurock 2015 & 64th Geomechanics Colloquium, Salzburg: 157-162.
Barton, N.R. and S.C. Bandis, 2017. Characterization and modelling of the shear strength, stiffness and
hydraulic behaviour of rock joints for engineering purposes. Rock Mechanics and Engineering, Ed.
Xia-Ting Feng, Vol. 1, Principles. Ch. 1, 3-40. Taylor & Francis.
Barton, N. and Shen, B. 2017. Extension failure mechanisms explain failure initiation in deep tunnels
and critical heights of cliff faces and near-vertical mountain walls. US Rock Mech. Symp. San
Fransisco, ARMA17-686, 20p.
Barton, N. 2021. New ideas about failure modes in rock masses – from tunnels to Prekestolen to El
Capitan to Everest. Bergmekanikkdag, Oslo. 33.1-33.20.
Bieniawski Z.T. 1989. Engineering Rock Mass Classifications: A Complete Manual for Engineers and
Geologists in Mining, Civil, and Petroleum Engineering, New York, Wiley, 272p.
Carranza-Torres, C. 2021. Computational tools for the analysis of circular failure of rock slopes. Key-
note Lecture. In Proceedings of EUROCK 2021. Mechanics and Rock Engineering from theory to
practice. Turin. Italy.
Chryssanthakis, P., Monsen, K. & Barton, N. 1991. Tunnelling in jointed rock simulated in a comput-
er. (In Norwegian). Tunneller og Undergrunnslanlegg, 1989-1991, NTNF, Oslo, pp. 23-28.
Hajiabdolmajid,V., C. D. Martin and P. K. Kaiser, 2000. Modelling brittle failure. Proc. 4th North
American Rock Mechanics Symposium, NARMS 2000 Seattle, J. Girard, M. Liebman, C. Breeds
and T. Doe (Eds), 991–998. Balkema.
Hoek. E. and J. Bray, 1974. Rock slope engineering. Institute of Mining.
Hoek, E, C. Carranza-Torres & B. Corkum, 2002. Hoek-Brown failure criterion – 2002 Edition. Proc.
NARMS-TAC Conference, Toronto, 2002, 1, 267-273.
Lorig, L., P. Stacey and J. Read 2009. Slope design methods, Ch 10 in Guidelines for Open Pit Slope
Design, Eds. Read and Stacey, CSIRO Publishing.
Martin, C. D. and P. F. Stacey, 2013. Pit slopes in weathered and weak rocks. Slope Stability 2013.
Brisbane. P. M. Dight (ed), Australian Centre for Geomechanics, Perth, 29p.
Patton, F. D. 1966. Multiple modes of shear failure in rock. Proc. 1st Cong. of Int. Soc. Rock Mech.
Lisbon. Vol. 1, 509-513.
Potvin, Y. 1988. Empirical open stope design in Canada, Ph.D. thesis. Univ. of British Colombia,
Canada, 350p.
Renani, H.R. and M. Cai, 2021. Forty-Year Review of the Hoek–Brown Failure Criterion for Jointed
Rock Masses. Rock Mechanics and Rock Engineering 55(4):1-23.
Shen, B., O. Stephansson, and M. Rinnie, 2013. Modelling Rock fracturing Processes – A Fracture
Mechanics Approach Using FRACOD. Springer, 173p.
Styles, T.D. & A. Vakili, 2020. Slope Angle Versus Slope Height - The Basis of an Empirical Tool for
Slope Design Within Fractured Rock Masses using IUCM. EUROCK 2020, Trondheim
Ván, P. and B. Vásárhelyi, 2014. Sensitivity analysis of GSI based mechanical parameters of the rock
mass. Periodica Polytechnica, Civil Engineering. 58/4 (2014) 379–386.

Barton, N. 2022

Uploaded by

Copyright:

Available Formats

Barton, N. 2022

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Barton, N. 2022

Uploaded by

Copyright:

Available Formats

IX Latin American Rock Mechanics Symposium

IX Simposio Latinoamericano de Mecánica de Rocas

Keynote Lecture: Continuum or Discontinuum – That is the

2 PHYSICAL MODELS WITH FRACTURES PRE UDEC

3 SOME FUNDAMENTALS OF JOINT BEHAVIOUR

4 SOME EXAMPLES OF UDEC-BB MODELS SHOWING GEOLOGICAL DETAIL

Figure 9. The discontinuum-modelling advantages of representing adversely steeply-dipping bedding.

4 OVERBREAK AND TUNNEL AND STOPE STABILITY

6 WHEN CONTINUUM MODELLING CAN BE THE ONLY CHOICE

October 16-19, 2022

7 A CRITIQUE OF GSI, HOEK-BROWN EQUATIONS, AND THEREFORE CONTINUUM

GSI1 = RMR – 5 = R1+R2+R3+R4+R5(=15) – 5 (1)

8 EXAMPLES OF ROCK SLOPES: REAL, DISCONTINUUM, CONTINUUM

Figure 34. Four major open-pit failures reproduced by Carranza-Torres (2021).

produced to emphasize the possibility of representing some details of ‘geology and

You might also like