Influence des propnetes des diaclases sur la rnodelisatlon des masses roche uses diaclasees
Der EinfluB von Spaltbrucheigenschaften auf die Modellbildung von geklOfteten Felsmassen
NICK BARTON, Norwegian Geotechnical Institute, Oslo, Norway
ABSTRACT: Predictionof likelyresponsetoexcavation, andproductionof final designsfor therock reinforcement, requirerealistic
descriptions of the components of rock mass behaviour. This article explores some of the methods that have proved reasonably
successful indescribingandmodellingrockjoints androckmasses, despitethecomplexitiesinvolved. Index testingof rockjoints and
rock mass characterisation, including geophysical methods, are theessential activities inpreparation for two and threedimensional
distinct element modelling. Recent improvementsaredescribed.
RESUME: Laprevisiondelareponsevraisemblabled'un massif rocheux lors delarealisationd'une excavation, ainsi queIedimen
sionnement des renforcements necessalres, necessitent une description realistedu comportement des composants de ce massif. Cet
articleexplore quelquesunes des methodesqui sesont montrees raisonnablement satisfaisantespour ladescriptionet lamodelisatton
des massifsrocheux et deleursjoints, endepit delacomplexitequecelasuppose. Lesessaissurjoints et lacaracterisationdumassif
(y compris par lesmethodesgeophysiques) sont lestSltSments essentielsprealables aunemodelisationendeuxoutrois dimensionspar
elementsdiscontinus. Desdeveloppernentsrecents sont decrits.
ZUSAMMENFASSUNG: Die Vorhersageder wahrscheinlichen Gebirgsreaktion auf das Auffahren von Untertageriiumenund das
Design von Felsverstlirkungenverlangt die wirklichkeitsnaheBeschreibungder einzelnen Komponenten des Felsverhaltens. Dieser
Artikel beschreibt einige Methoden, welchetrotz ihrer Komplexitat, erfolgreich zur Kluft und Felsmodellierung und Beschreibung
angewandt werden. Das Indextesten von Kluften und die GebirgsklassifIzierung, geophysikalischeMethoden eingeschlossen, sind
wesentlicheBestandteilein der Vorbereitungsphasevon zwei und dreidimensionalen bestimmten Elemente Simulierungen. Neuere
Entwicklungenwerden beschrieben.
1INTRODUCTION
This article explores some of the methods which appear to be
havingsomesuccessinrealisticmodellinganddesignforjointed
rock masses. Key techniquesarejoint indextesting, rock mass
characterisation, seismic measurements and distinct element
modelling. At NGI, these methods can be represented by the
following basic symbols: J RC, J CS, q,r Q, V
p
, UDEC and
3DEC. The first three are the index parameters for thejoint
setsof concern (BartonandBandis, 1990). TheQvalues give
estimates for rock mass moduli and rock reinforcement,
following Grimstad and Barton, 1993. The two and three
dimensional distinct element models UDEC and 3DEC con
ceived by Cundall and refined by ItascaInc. provide the final
essential link to reality.
Spatial variability withintherock mass which is reflectedto
someextent by thestatisticsfor J RC, J CS, q,r andQ, isfurther
described by theseismic measurementswhichprovideameans
of extrapolation betweenmapping locations(i.e., exposures or
drill core). In its optimal form (crosshole seismic
tomography), it givesdetailed informationthat canbeapproxi
matelycorrelatedtoQvaluesandtodeformationmodulus, using
recent developments.
2 SHEAR BEHAVIOUR OF ROCK J OINTS
Direct shear testsof roughwalledtensionfracturesdevelopedin
weak model materials, that were performed many years ago
whentheauthor wasastudent, indicatedtheimportanceof both
thesurfaceroughness andtheuniaxial strength(O'J of therock.
Theempirical relationfor peak shear strengthgiveninEquation
1was essentially the forerunner of thesubsequent J RCJ CS or
BartonBandis model, where the joint roughness coefficient
(J Re) was equal to 20 for theserough tension fractures. The
joint walI strength (J CS) was equal to O'c (theunconfmedcom
pression strength).
't = all tan[20log(::) + 30
0
] (1)
Theoriginal formof Equation 1isthereforeperfectlyconsistent
with today's equation:
't =0'11 tan[J RC 10ge~:) + ~r] (2)
Equation1representsthethreelimitingvaluesof thethreeinput
parameters, i.e.,
J RC 20 (roughest possiblejoint without actual steps)
J CS = O'c (least possibleweatheringgrade, i.e., freshfrac
ture)
q,r ~ (fresh unweathered fracture with basic friction
angles intherange28~ to 31~0).
Bandis et al., 1981, 1983and Barton et al., 1985, have sub
sequently shown how these three index parameters J RC, J CS
and q,r can be used for modelling both the sheardilation and
normal closure behaviour of rock joints with estimation of
physical and hydraulic joint aperture, and with dueaccount of
scaleeffects and shear reversals, etc.
Figure 1illustratesthefirst versionof theconstitutivemodel
for shear anddilationbehaviour, whichwassubsequentlycoded
by Itasca for use in UDECBB (Christianson, 1985, personal
communication) and improved by NGI and Itasca (Gutierrez,
1995; Christianson, 1995, personal communication) for use in
an improvedversion UDECBB.
Although different degrees of joint weathering and mineral
1023
,
:1.0
+ L1," +
2 . 5
2.
30' IUlDUAI.
LA. INIlTV NATUIAl.
TIlT TIlT II.OCItI

II U
KI 110 10 ~
&
to
4. 0 6.1
"''''
'.
s s
s
,
,
s
Figure 4. Normal (N) and shear (S) components of joint
deformation determine the general form of stressstrain curves
for loading tests on rock masses. (Barton, 1985; Bandis' et
al., 1981; 1983)
The concave (N) component and the convex (S) component
are each dominant, or combine with each other, as the case may
be. (Types A, C and B respectively). Uniaxial (strain) loading
in simple UDECBB distinct element models of the same
problems are shown in Figure 5.
Despite the uniaxial loading with no lateral strain, the Type
C rock mass shows larger overall deformation and of course
joint shearing. Peak stresses were also higher in this model.
Physical model studies reported by Barton and Bandis (1982)
have indicated higher shear resistance for thejointed assemblies
of blocks that had the smallest block sizes. This finding is
shown schematically in Figure 6, where models with 4000, 1000
or 250 blocks were studied in biaxial shear. Reduced J RC and
lCS values have to be used for the larger block sizes (i.e., J RC
n
and J CS
n
for block sizes r...).
The two equations given below show how J RC and lCS given
in Equation 2 can be scaled down to allow for the lower shear
resistance expected at in situ block size.
J RC J RCo(~r02 me. (4)
(
L )0.03 me.
lCS.. res, r.: ( 5 )
Despite the potential for scale effects connected with block
size, in which the smaller blocks may give higher ultimate shear
resistance (for equal joint roughness), there is nevertheless a
general experience that the deformation modulus of more
heavily jointed rock masses is lower than for massive rock
masses.
Figure 7, which is an idealised UDEC study of tunnelling in
assemblages of 250 to 10,000 blocks using Mohr Coulomb joint
parameters (and no builtin joint strength scale effect), shows,
as expected, much larger disturbed zones (and deformation) as
1025
I
.0>1 .heol"' dtep 1. eeOE04 I
each I In. ~hl ck I.000E05
2.00

c:
QI
E
~ I.J O
o
Q.
" 'C
.0.'0
)(
o
~


I
MUIC shwor"' J 1er 218041
each J lne t.hick 1.00010'
Figure 5. Uniaxial strain loading of threehypothetical rock masseswith a2D UDECBB model.
(Chryssanthakis et al., 1991)
block size reduces. Figure 8 gives the distributions of joint
shearing caused by numerical tunnelling in what is a highly
anisotropic stress field (C7
y
=20 MPa, C7b =5 MPa). The
deformability of theclosely jointed model is clearly by far the
highest of the cases studied, with or without tunnel support
measures.
Physical models and UDEC models that were driven to a
stageof complete tunnel failure showed shear band formation
(block rotation) when the block size was sufficiently small
compared to the excavation dimensions or loaded boundary
dimensions (Shenand Barton, inpreparation).
4 ROCK MASS DEFORMABIUTY FROM ROCK MASS
CLASSIFICAnON
It is reasonably certain that the idealisedrock mass depicted as
Model No.4 (Figure7) would (inthereal world) havereduced
rock mass quality (RMR or Q), reduced deformation modulus
(M) andreduced seismic Pwavevelocity (Vp)' ascompared to
themore massivecases with less rock blocks. In reality there
might also be reduced joint roughness or even slickensiding
(i.e., J , = 0.5) andmineralisation(i.e., J . = 4) (i.e., lowIRC,
J CS andq,r) inthecaseof therock masswithsmall block sizes.
Theabovedifferences inbehaviour wouldbeaccentuatedby the
combination of lower deformation modulus and lower shear
resistance.
TheQsystemof rockmassclassification(Bartonet al., 1974)
is designed to provide greater levels of rock reinforcement and
tunnel support insuchcases. Thefact that theQvaluecanvary
from0.001 to 1000is also a reflection of theenormous range
of rock mass deformation moduli (i.e., 0.05 to 50 GPa) and
shear strengths (i.e., 0.1 to 20MPa) that may beencountered,
andwhichmayhaveanaccumulativeeffect ontheneedfor rock
reinforcement inthecaseof tunnelling.
TheQvalueof arock massisbuilt up fromanassessmentof
relative block size (RQD/J.J, interblock shear strength (1/J .)
andactivestress(1w/SRF); it thereforehascloseparallels tothe
processes demonstrated in Figure 8.
It is therefore logical that the widerange of rock reinforce
~LANI snus
AXIAL TlSTS
~H ~
::B: < > .,.
< > .,.
. ( > . , .
< > .,.
:I
!
,
I
,
/
,t
i 'I
I It_th I"
" It lei
,/ ts
tttnt
0,
4000,1000 . 2
III.,.,. 1t1.,1u
NORMAL
Figure 6. Schematic of physical model tests of fractured
rock, indicating block size dependence. (Barton and Bandis,
1982)
ment solutions shown in Figure 9 should also have some
relationshipwiththedeformationmodulusof theparticular rock
mass. The samecouldbe saidof theBieniawski (1989) RMR
value, whichhas anapproximaterelationship to theQvalue.
Correlations between the RMRvalueand theQvalue show
significant trends but quite widescatter, particularly for lower
qualities of rock. This ispartly dueto theabsenceof theSRF
termintheRMR method. Nevertheless, becauseof thesignifi
cant sets of data on rock mass deformation modulus in the
literature related to thetwo methods, it is convenient to fmd a
workablecorrelation betweenQ andRMR.
In Figure 10, data on rock mass deformation moduli (M)
reported by Bieniawski (1978) and Serafimand Pereira (1983)
are reproduced, together with these authors' linear and non
linear relationships between M andRMR. On thesamefigure
asuggestedcorrelation betweenRMR andQ isgiven, basedon
thefollowingapproximation:
RMR lSlogQ + SO (6)
On thebasis of this Q rating scale, theapproximation
I
M 10Q; (OPa)
(7)
1026
~
x x x '> < :"'x"'Y 'x x x x X
x~xxx
~
XJ < x x < ,x X X X X
x XY)I 1 \ X > < x > < x
~~
:xy< h x 0N> )<
xx x x xxy:
Xx~
> 6 < > 6 < > &~ v:
Figure 7. Idealised UDEC models of tunnels within 2D
assemblies of 250 to 10,000 blocks. (Shen and Barton, in
preparation)
is proposed for estimating themean valueof rock mass defor
mationmodulus.
The dotted curve in Figure 10shows good correlation with
reported results and extends into theregion of lowrock quali
ties, very closeto theSerafimandPereira (1983) relation.
Equation 7is also shown within a larger set of higher
deformationmodulusdataasthenonlinear curveinFigure 11.
For fair, goodand very goodrock qualities, it provides avery
similar estimateof modulus to that recommendedearlier (M =
25 log Q, Barton, 1983), a correlation that has been used
successfully inearlier verificationstudieswith UDECBB.
5 GEOPHY SICAL CLASSIFICATION OF ROCK MASSES
Inmany countries withdeep surfaceweatheringandsoil cover,
the use of seismic refraction, crosshole seismic or step fre
quencyradar measurements, maybetheonlywayto extrapolate
G
Exceptlonall,
poor
100
so
e
.5
~ '" 20
..!l !3
~ 10
1
0.001 0.0040.01
Figure 8. Extent of joint shearing zones caused by widely
different block sizes. (ShenandBarton, inpreparation)
rockmasscharacterisationdatabetweenmappedrock exposures
or betweenavailablecored drill holes. Unless drill holes are
sufficiently deep (and close), there may also be uncertainty
concerning the rock mass quality at tunnel depth since the
refraction measurements have limited penetration. There are
other complicationsconnectedwiththeinfluenceof stresslevel
(i.e., depth) and rock density and porosity effects, each of
whichwill influencetheinterpretationof seismicvelocityandits
relation to rock mass quality.
International interest in potential correlations between rock
mass quality, rock mass deformation modulus and seismic
velocity has been considerable for many years, and various
correlationshavebeensuggested, includingwell knowncorrela
tionswithRQD. Theadvent of crossholeseismictomography
in the last ten years or so, and concerns with nuclear waste
E D C B
20
!} J
'" 2.4 II
i::>
1.5
0.04 0.1 0.4 I 4 10 40 100
. ~ J r J w
Rock mass quality Q= In x 18 x SRF
400 1000
Figure9. TheQsystemof classificationandreinforcement selection. (GrimstadandBarton, 1993)
1027
90
10
10 20 30 40 50 60 70 60 90 100
Geomec:hanic:a rock mass rating IRMA)
0.001 0.01 0.1 1.0 10 100 1000
arating
Figure 10. Approximations to deformation modulus based on
RMR andQ; covering the lower ranges of rock qualities.
NGI Classification (Q)
1 4 10
80
80
~
60
e
.
1
50
I
40
30
M (min) ..
..
..
i
..
20
.5
10
0
50 60 70 80 90 100
Fair
I
Good
I
Very good
C
RMR Cla.slfleatlon
Figure 11. Approximation to deformation modulus covering
thehigher ranges of rock qualities.
repository design has heightened this interest.
In 1991, NGI performed crossholeseismicmeasurements at
the sitefor Norway's Olympic rock cavern at Gjevik, Results
(redrawnfor clarity of reproduction) arepresentedinFigure 12.
These measurements which are described more fully in Barton
et a1. (1994), gave the opportunity for detailed correlation
between QIogging of the core and the adjacent velocity
calculations.
Thegeneral trend observed at this siteand for hard rocks at
other shallowsites (i.e., 25 to 50mdepth) inother countries is
as follows:
Vp logQ + 3.5 (km/s) (8)
(mil) 3238 3449 3661 3872 4083 429' 4'06 47174929
.
oi
!
oJ
IOl
~
oJ
<
IOl
17'
."
IOl
~
III
<
!i:
~
::
I"
BORE
HOLE
NO.3
LONGITUDINAL SECTION (m)
(mil) 3086
209
,,,
BORE
HOLE
NO.3
16 24 32
CROSS SECTION (m)
40
Figure 12. Crosshole seismic tomography at the Olympic
cavern site, Gjavik.
i.e., V
p
3.5 km/s for Q =I, V
p
4.5 km/s for Q =10,
etc.
Additional studyof resultsfromother sitesaroundtheworld,
including weak and porous rocks such as chalk, sandstoneand
tuff, anddeep locationsaccessedby crossholetomography has
resulted inthesuggestedcorrelations betweenquality, velocity
andmodulus given inFigure 13.
Essential features of theseismic correlation chart are:
1) correction for increasedstress or depth (causingincreasein
velocity anddeformation modulus),
2) correction for increased porosity (n%) or reduced uniaxial
compressionstrength(u
c
) (causingreductionsinvelocityand
deformation modulus).
1028
a _+ v
p
4_~ M
Rock mass quality Seismic velocity Deformation modulus
V,"logQ+3.5(kmisec) M"10.Q'~(GPa) M"W.l0("f!) (GPa)
Extremely
poor
Very
poor
Poor
6.0
~ 5.0
~
~ 4.0
s
~ 3.0
.\1
.~
c? J 2.0
1.0
0. Q1
0.1
Approx.
ra~pe
'deform.
moduli
M M
.... 
(GPa)
100 100
6 ,0 53 6 6
30 46
5,0 17 32
9 22
4.0 5 15
3 10
3,0 2 7
1 5
2.0 0.5 3
0.3 2
1,0 0,2 1.5
0.1
1.0
4 10 40 100 400 1000
Q =(RQDx~x~)~
C J , J . SRF 100
Figure 13. Rockmassquality, seismicvelocityanddeformation
modulus correlations for design.
Thecompressionstrengthcorrectionisappliedby "normalising"
theQvaluestoanominal hardrock compressionstrengthvalue
of 100MPa.
Q., = Q x 2 (9)
100
This correction is to allowtheQ
c
valueto reflect theinfluence
of rock compression strengthon seismic velocity.
The standard Qvalue is reduced by SRF when the ratio of
rock strength to major principal stress (uc!Ut) implies rock
failure problems and need for increased rock reinforcement.
Although V and M values will be expected to reduce in the
excavationdisturbed zone (EDZ) (as shownin Figure 14), the
correlations given in Figure 13shouldbe appliedwith caution
intheEDZ around atunnel.
Anexamplewill beused to illustratehowtousetheseismic
correlation chart (Figure 13):
AssumeQ =4and U
c
=25 MPa, thereforeQ
c
=1
AssumeH =250mand n% =5%
Thelatter areexpectedto causeapproximately (+) 1.2and ()
. 0.6km/s change in seismic velocity for Q
c
=I, compared to
the shallow (25m) nominal value. Therefore V
p
"" 4.0 km/s
andthemeandeformationmodulus (at250mdepth) "" 15GPa.
tONE J
ICl ,
Figure 14. Seismic measurements incircular tunnels showing
effect of stress concentration. (Plichon, 1980)
These correlations will usually be applied in reverse order,
i.e., by measuring V
p
at depth H, with estimated n% and U
c
(MPa) values, an approximate Qvalue could be selected for
preliminary assessment of rock support needs. For design
purposes the seismic measurement would allow the rock mass
deformation modulus to beestimated, prior to in situmeasure
ment or direct classificationof core.
6PRESENTATION OF J OINT AND ROCK MASS DATA
Thegeotechnical loggingchart prepared asaLotusspreadsheet
inFigure 15showshowthepreviously describedjoint androck
mass logging and index test data can be assembled for rapid
reference. Eachchart might represent thestatisticsfromseveral
core boxes, from several kilometres of surface or tunnel
mappingor fromacompletedproject.
TheloggingstatisticsshowninFigure 15havedataarrangedas
inTable2. This means that:
the upper third of the chart gives geometrical properties of
therock mass (for buildingthenumerical models),
themiddlethird of thechart givesjoint strength and rough
ness(for strengthanddeformabilityinput tothemodels), and
the lower third of the chart gives approximate ranges of
permeability, rock strength and major stress (for defining
boundary conditions inthemodels).
7 UTIUSATION OF LOGGED DATA IN UDECBB AND
3DEC
Utilisationof joint and rock mass loggingdata (Figure 15) for
numerical distinctelementmodellingisillustratedbyanexample
UDECBB model in Figure 16. The inset below the figure
shows thejoint index data for laboratory scalevalues of J RC,
J CS andtP
r
[A largescalejoint roughness (i) valueof 6 was
also assessedat this site.]
Thedeformationmoduli of 20, 30and40GPaat depthsof 0
45m, 4565mand65125mshowninFigure 16wereestimated
from QIogging and fromevaluation of the seismic measure
ments. It will be noted from Figure 12 that the seismic
tomography shows Vp values in the range 4to 5 km/s in the
1029
I J
40.0 I 12.0'
10.0I ".0
87.2, e.e .
I J
1.5 4.0
3.0I 0.'*
2.3 1.'
a All
1/ 1.0 1.25 obMMtIonl
1I 1.0 10 1981/92
1 1.0 '.549
:::
1" ~~
': leore plecel I
> _10cm
".
_1;1
o ao I' " I , 01
8
L
o
C
K
8
I
Z
E
8
8
H
E
"
,.
::
le
::
o 1 0.' 1 I.' ,.I 2 ,
~
'I ", ".R
9
T
"E
N
G
T
H
C
T
I
V
E
? 
..
.
o 0.01 0.1 0.' o.~ 0.' 0.. 1
~
20'5'0,20,5'0,'7,552,520,052.5,2,5
9Ol. FRESH
'1[ $'I:=;~='
8
T
"E
8
8
E