Vous êtes sur la page 1sur 2

Harmonising Rock Engineering and the Environment Qian & Zhou (eds)

2012 Taylor & Francis Group, London, ISBN 978-0-415-80444-8

A comparative study of mi , the Hoek-Brown constant for intact rock


material
S.A.L. Read
GNS Science, Lower Hutt, New Zealand

L. Richards
Rock Engineering Consultant, Canterbury, New Zealand

Subject: Rock material and rock mass property testing (laboratory and in situ)
Keywords: lab testing, rock failure, rock properties
1 INTRODUCTION
The original Hoek-Brown failure criterion (Hoek & Brown
1980) was defined in terms of major and minor total stresses
as follows:

where the relationship between the principal stresses ( 1 and


3 ) at failure is defined by two constants (m and s) and the
uniaxial compressive strength of intact rock ( ci ).
In the case of intact rock material where s = 1.0, Hoek &
Brown (1980) noted that m ci / | t | where t is the
uniaxial tensile strength of the intact rock.
The generalized failure criterion (Hoek & Brown, 1997)
in effective stress terms included three constants (mb , s and
a) in the equation as:

In the case of intact rock material (i.e. s = 1) the generalized


equation becomes:

the stress space would be 1 < 3 < 1/3.4 on the basis of


Mogis original definition of the brittle/ductile transition.
Subsequently Hoek & Brown (1997) recommended that the
range should be 0 < 3 < 0.5 ci .
Sheorey (1997) advocated the inclusion of tensile strength
data along with unconfined and triaxial compression data, as
better regression is obtained with data point(s) closer to the
3 axis of the principal stress plots.
The calculation method to analyse the testing data has
an effect on the mi values, with the RocData and RocLab
programs from RocScience (2010) using the non-linear
Levenberg-Marquardt method of least-squares curve fitting
as the default, with options for the simplex reflection and
linear regression methods.
2.1 Guideline values for mi
As part of the original failure criterion, Hoek & Brown
(1980) suggested that m for intact rock (mi ) reflected
rock type, with values increasing 7 for carbonate rocks to
25 for coarse-grained rocks. Some 15 years later, Hoek put
forward guideline mi values, based on a more detailed
lithological classification, with subsequent revisions
including ranges of values.
2.2 Values for mi calculated from test result
databases

The full paper first summarizes procedures used to calculate


intact material constant mi and then compares values
obtained from a large database of testing data with guideline
tables. After considering further testing results, suggestions
are made for a simple method of estimating mi by means of a
combina- tion of unconfined compression and indirect
Brazilian tensile tests.
2
CALCULATION
VALUES

OF

mi

The Hoek-Brown failure criterion is intended for use in the


brittle range of rock behaviour, and this controls the range
of minor principal stress (3 ) values over which triaxial testing to determine mi is carried out. Initially Hoek considered

2011 by Taylor & Francis Group, LLC

Valuable databases of laboratory test data have been


provided by Sheorey (1997) and Douglas (2002). In general,
the cal- culated database values show little correlation with
guideline values, with calculated values having a much
greater range than those in the guidelines.
Correlations of mi with other rock parameters, (such as
unconfined compressive strength) are mostly very poor.
2.3 Estimates of mi values using tensile testing
The only rock material parameter with which mi has any reasonable correlation is the ratio of the unconfined
compressive strength c to the tensile strength t . Sheorey
(1997) defined this ratio as R = c /| t .|.
The laboratory tensile data in the Douglas database comprised results from both pure uniaxial and indirect
(Brazilian) testing. Using these data (59 of the 521 sets in
the database), the intact rock parameter mi , shows a fairly
good relationship

364

Figure 1. Plot of laboratory compressive to tensile strength ratio


with Hoek-Brown constant mi using data from Douglas (2002).

The results are given with respect to measured values of


c , (as measured by the unconfined compressive (qu ) test)
and tb (as measured by Brazilian tensile test), and R(ratio
of ci to tb ), together with calculated values of ci , t and
mi , with or without the inclusion of tensile results.
The calculated mi values are generally outside the ranges
for guideline values (e.g., schists vary from 5 to 8, which is
below the lower end of the guideline range of 12 3).
While the ignimbrites at the two locations cover an
overlapping range of unconfined compressive strengths (5 to
25 and 14 to 40 MPa), the mi values at each site are
noticeably different (Figure 2).
The inclusion or exclusion of tensile results has a variable
influence on results. In the main, the mi values are reasonably consistent whereas others have greater differences,
more noticeably associated with higher mi values.
In most cases, there is reasonable correspondence
between the values of R and mi , reflecting good fits between
testing and calculated data from the full suite of triaxial tests,
in particular where the test and calculated tensile values are
similar. The R value is considered to give a good indication
of the value of the Hoek-Brown parameter mi .
4 CONCLUSIONS
Conclusions from our investigation of intact rock material databases, combined with further laboratory testing, to
determine the Hoek-Brown constant mi are:
1 The simple relationship between mi and rock type as
envis- aged by Hoek & Brown (1980) has not proven to
be appropriate, with the benefit of a much greater data
set than was available at that time.
2 The most accurate method of assessing mi values remains
as statistical analysis of data from a full set of laboratory
test results. These should include triaxial and unconfined
compression plus tensile tests within the recommended
stress range (i.e. t < 3 < 0.5 ci ).
3 In absence of a full suite of laboratory tests, R the ratio
of unconfined compressive strength to tensile strength, is
a useful indicator of mi values, particularly as the tests
for both properties are relatively straightforward to
perform.
REFERENCES

Figure 2. Principal stress plots for similar strength ignimbrites


from
Arapuni and Maraetai.

between R and mi , evenly distributed about the 1:1


correlation line for both test types (either Brazilian or direct)
Figure 1.
3 ASSESSMENT OF mi AND R WITH
ADDITIONAL DATA SETS
Four sets of tests have been performed on different rock
types (schist, conglomerate, two ignimbrites), including at
least two uniaxial and two Brazilian tensile tests, triaxial
tests with maximum
< 0.4 c to 0.6 c and calculated
3
using the Levenberg-Marquardt technique.
365
2011 by Taylor & Francis Group, LLC

Douglas, K. J. 2002. The shear strength of rock masses. PhD thesis, School of Civil and Environmental Engineering, University
of New South Wales, Sydney, Australia.
Hoek, E. 2010. Rock mass properties. Ch 11 in Practical Rock
Engineering.
http://download.rocscience.com/hoek/PracticalRock
Engineering.asp.
Hoek, E, & Brown, E. T. 1980. Empirical strength criterion for
rock masses. J. Geotech. Eng. Div., Proc. Am. Soc. Civil Eng.
106(GT9):
10131035.
Hoek, E. & Brown, E.T. 1997. Practical estimates of rock mass
strength. Int. J. Rock Mech. & Min. Sci. 34(8): 11651186.
Rocscience 2010. Product overview RocLab, RocProp and RocData. http://www.rocscience.com/products/overview.
Sheorey, P.R. 1997. Empirical rock failure criteria.
Rotterdam: Balkema.

Vous aimerez peut-être aussi