Anda di halaman 1dari 32

ADIF Seminar on Advanced Geotechnical Characterization for Tunnel Design

Madrid, Spain, 29 June, 2011



Richard Z. T. Bieniawski Preinl, DrHon.C (Madrid), DrHon.C (Cracovia)

Bieniawski Design Enterprises, USA

The things which create problems are not those we do not know;
but those we think we known for certain.
President Dwight Eisenhower

The quotation above is most pertinent to the theme of our Seminar because of
its importance to many fields, including engineering, as mistaken beliefs have
plagued mankind resulting in loss of lives and resources on a grand scale. Most
recently, errors in foreign policy in 2003 when the war in Iraq was based on wrong
intelligence accepted as 'sure' to the leaders of America, Britain and Spain, to the
financial crisis of 2008 when the stock market experts knew for 'sure' how to drive
the bank 'bubble' until it burst. In between, there were engineering disasters of coal
mine collapses in America, and those in Chile and China, which were caused by
errors in doing "what was always done the 'sure' way". Subsequently, in all these
events heroic efforts had to be made to correct the errors, but the lives lost or the
money squandered and minerals left behind, are still unaccountable.

In fact, the latest issue of Tunnels & Tunnelling International (February 2011)
lists no less that 41 major tunnel failures. These case histories, including two cases
from Spain, show that more than 85% were the result of unexpected geotechnical
conditions and mistaken interpretations.

This situation made me think about our own discipline, tunnelling in rock, and
of my own expertise, design and characterization of rock masses, when I was asked to
prepare an inaugural address in Poland eight months ago on the occasion of receiving
the distinction of Doctor Honoris Causa from the Universidad Politcnica de
Cracovia. I immediately remembered my inaugural lecture for an equally great honor
which I received in this city, as my first Doctor Honoris Causa from the
distinguished Universidad Politcnica de Madrid in 2001. Significantly, a decade has
lapsed between these two addresses in which I selected, in both cases, rock mass
characterization and classification as the topic requiring better understanding and
avoidance of errors because they provide the data which are used as input to
computer modelling - ever growing in sophistication and complexity. Clearly, as
reliable estimation of the mechanical properties of rock masses, primarily their
strength and deformability is the essence of engineering, I lamented at the time that
when input data are in error (like military intelligence or financial arrogance), the
output of the action that follows brings wrong results.

It is said that "to err is human" (Alexander Pope, 1688-1744) and that
"scientists and engineers learn though their errors" (Henry Petroski, 1991) but
there comes a time when we must take stock of what has been done wrongly and
bring about the necessary corrections, to avoid failures in the future.

For this purpose, I have compiled a significant amount of material

demonstrating the myths (or misconceptions) still existing when using rock mass
classifications and I would like to offer some of the solutions which emerged in the
past 10 years between my two Doctor Honoris Causa even though they need further
elaboration. Alas, while there are solutions, they still await action to implement them,
in spite of numerous projects done, publications written and endless discussions in
design offices and company boardrooms.

Common MYTHS in Classifications of Rock Masses

In the limited time at my disposal, I selected FIVE most glaring
misconceptions commonly occurring in rock engineering.



Not true. Doing either of these alone is a major error.

This myth concerns the design of
tunnels in general, which involves
three approaches which should go
hand in hand with one another
forming a part of an engineering
design process, such as the
Systems Design Methodology and
Principles, shown in Figure 1. It
is essential to avoid selecting just
only one of the methods involved,
justifying it by "we don't have the
time and the money" to try the
correct approach. The three
methods in question are:

empirical (for example RMR or

Figure 1. A part of the Systems Design Methodology Q classifications), analytical (for
and Principles for Rock Engineering (Bieniawski, 1992)
example, closed form solutions or
numerical computer modelling),
and observational (for example, measurements during construction or NATM).

The reason why all three approaches should be used on every tunnelling
project is that each method has its own important merit and role to play. Thus the
empirical rock mass classifications RMR or Q represent a wealth of accumulated
practical experience from case histories and provide the reality of current engineering
judgment. Theoretical, analytical criteria for rock mass strength and deformability are
also essential because they include an assessment of rock mass strength versus the
acting stresses, leading to determination of factors of safety. And the observational
methods featuring measurements of convergence and deformations during
construction are indispensible for comparing predictions with encountered behavior,
thus providing a check on design assumptions and estimations.

It is amazing how many publications have been written arguing against rock
mass classifications as the exclusive means for tunnel design, yet such exclusivity has
never been the intention when I developed the RMR system 38 years ago! I have
always emphasized that rock mass classifications should be used in conjunction with
the other two approaches. By the same argument, they should not be dismissed from
the process of design because they play a crucial role in rock mass characterization,
bridging qualitative geologic descriptions into quantitative engineering data.

Finally, a word of warning! When talking about empirical methods and
considering rock mass classifications as an example, one must realize that these
classifications are not equal; they were developed for different purposes and from
different data bases! In essence, RMR and Q are indeed in the same category of
assessing rock mass quality for tunnelling purposes and providing input for
engineering design and construction. Accordingly, they complement each other and
are correlated (Barton and Bieniawski, 2007). The GSI index is different from these
two; "it has no end use other than to provide input for the Hoek-Brown criterion"
(Hoek et al., 1995) as a rock mass characterization index. The NATM classification
is also different from RMR and Q; it is a part of the observational method of
tunnelling and its purpose is not to provide a geotechnical characterization of the
ground but to provide "an objective basis for calculating tunnel costs and advance
rates" (Galler, 2010), that is, specifying 'excavation classes' for the purpose of
contractual compensation.

Because of the misconceptions between the RMR (1989) and GSI (1995),
their history and their applications, the sections which follow will deal with these
aspects in some detail, while the NATM - not being a geotechnical characterization
approach, falls outside the scope of this presentation.


Not true, this is a misconception of facts. The fact is that RMR continues to be
used successfully even for 'very poor rock masses', Class V when RMR < 20,
when the input data are properly determined.

This myth stems from a number of misconceptions arising when rock mass
classifications are treated as a "black-box" or "cook book" expecting ready 'recipes'
for all design situations. In a publication in 1991, I cautioned engineers and
geologists that weak rock masses do require special attention for careful geotechnical
characterization because the accuracy of RMR, depending on expertise, maybe to
within 2 - 3 rating points (the same applies to other classifications). But this does not
mean that the RMR system cannot be applied to assess very poor rock masses. After
all, we have numerous case histories on record with RMR = zero to 3.

The most serious misconception arising in the literature in the past decade
was that it was not clearly understood that the ratings for the input parameters for
RMR amount to zero for their minimum values. It was overlooked that the original

RMR table of ratings - Table 1 below - depicted the average ratings for each range of
parameters, and not the minimum ratings, as implied by Hoek et al. (1995).

Yet graphic rating charts were available in the literature - prepared to

facilitate computer analyses - Figure 2 - that clearly show the curves starting at zero.
Hence, the worst rock mass quality has a rating value of RMR = 0, meaning that in
such a case one is dealing with soil and not rock.

This misconception arose when Hoek et al. (1995), acting in good faith to deal
with very poor rock masses, presented an example of RMR compilation by assuming
completely dry rock conditions and a very favourable discontinuity orientation for a
very poor rock mass with a negligible rock material strength c. In such a case, from
Table 1, the minimum RMR value was calculated erroneously as 8 (3+5), concluding
that the RMR system would not work for very weak rock masses. In order to
overcome this perceived limitation, a new purpose-built approach, the Geological
Strength Index (GSI) was introduced.

Figure 2a. Input charts for RMR parameters rock strength anf RQD (Bieniawski 1989)

Figure 2b. Input chart for RMR parameter discontinuity spacing (Bieniawski 1989).

Moreover, it was speculated that under these conditions, for RMR < 15, the
following approximate equation might apply:

GSI = RMR - 5,
because the RMR parameters of discontinuity density (RMR + joint spacing) and
discontinuity condition would be approximately the same as the two GSI parameters
of rock structure and surface condition, whose ratings were adopted from the RMR
classification anyway.

The problem with this reasoning is that such an equivalency only applied in
the lowest range of very poor rock mass conditions (RMR class V).

Unfortunately, this misconception was carried further by many geologists and

engineers when dealing with better quality rock masses, RMR >> 20 because it was

easier to apply a descriptive approach of GSI instead of a quantitative RMR system
with measured input parameters. In the process, inaccurate results were obtained and
passed on as 'reliable' input to sophisticated computer analyses.

Moreover, it should be recalled that when GSI was introduced in 1995, its
motto was: "do not try to be precise for quick estimates". The GSI developers pointed
out that it is an index of rock mass characterization, not meant for replacing a
classification system of the type of RMR or Q - but this was overlooked in practice,
even to this day. Note that the only function claimed for GSI was an estimation of the
rock mass strength using the Hoek-Brown criterion; specifically for very weak rock

Figure 3. Correlation between RQD and spacing of discontinuities (Priest and Brown 1983)

Another misconception was the main argument that the parameter RQD, used
as an input to RMR, is very difficult to obtain for very weak rock masses, as it may
approach the value of zero. This aspect was, in fact, anticipated back in 1989 and
published in a number of articles pointing out that the RQD was originally used in

both RMR and Q classifications because it had the advantage of having an extensive
collection of case histories in its database. So, it served as a back up, in a way, to
discontinuity spacing because there existed a correlation between the RQD and
spacing, Figure 3, leading to the concept of "discontinuity density" proposed for
RMR in 1989. The correlation of these two parameters in this concept allowed to use
it as an advantage in two situations: during exploratory drilling from the surface, the
RQD was used as the primary parameter to estimate discontinuity density as there
was no access for surface mapping of discontinuity spacing in the tunnel. On the
other hand, when inside a tunnel having an access for geological mapping but no
boreholes, the RQD could be ignored and the discontinuity spacing used to obtain the
density rating from Figure 3 based on discontinuity spacing. Clearly than, the
quantitative RMR and Q are equally or better suited for very weak rock masses than
the qualitative index GSI.

Finally, a proof for applicability of the RMR system for assessing the quality
of very poor rock masses can be best seen from the collection of many such case
histories in Figure 4 which shows a correlation between RMR and Q, published as
early as 1976. This chart also depicts a notable scatter of the results. Thus, both
classifications should be used on a given project to check on the applicability of the

RMR = 9 ln Q + 44

(same correlation as also shown in the literature as RMR = 9 loge Q + 44). This
expression is not valid when GSI is specified instead of RMR.

Figure 4. Correlation between RMR and Q (Bieniawski 1976)




Not true, the Coulomb-Mohr criterion, dating back to 1773 (!), serves different
purposes, particularly slope stability analysis, but other criteria equally effective
for peak strength estimation are available, for example, the Yudhbir-Bieniawski
criterion (1983) which is used to cross-check the results from the Hoek-Brown
criterion (Edelbro et al., 2006).

While the Hoek-Brown criterion is better known and commonly used in

computer analyses, the criterion of Yudhbir et al. (1983) based on Bieniawski (1974)
deserves attention for cross-checking the estimates of rock mass strength, which is
necessary because a direct measurement of this property is not practicable to achieve.
These two criteria are listed in Table 2 below.

Table 2. Rock mass strength criteria currently in use (Hoek et al. 2002, Bieniawski 1989)

Rock Mass Strength Criterion Parameters and Definitions

Hoek-Brown: mi : rock material constant
mi = f (rock type)
e.g. sandstone mi = 12-17
mb : reduced value of the
material constant mi
s and a: constants for the rock
D : disturbance factor

Yudhbir-Bieniawski: A : rock mass constant

A = f (rock mass quality)
e.g. A=1 for rock material
B : rock material costant
B = f (rock type)
e.g. sandstone B = 4.0

In order to apply the Hoek-Brown criterion, knowledge of RMR or GSI and

the rock material constant mi are required. Note that the original equations (1988) for
determinations of parameters mb (designated as m in 1980) and s for the Hoek-Brown
criterion, were based on RMR (after Priest and Brown, 1983) and many engineers
still consider them most appropriate. They were subsequently replaced (Hoek et al.,
1995) by equating GSI with RMR, but using the same form of expressions. The
original criterion also included a square root term with a power of 0.5 instead of the
variable and complex expression for a used currently (which has a maximum value of

Most recently, a detailed study of this aspect was made by Malkowski (2010).
He determined how the value of GSI affects the empirical constants mb and s as well
as the rock mass strength M. As shown in Table 3, Malkowski's calculations
demonstrate that a change of 5 points of GSI, from 35 to 40, leads to dramatic
increases in the values of the following: M by 37%, change in parameter mb by 20%
and in the modulus of deformation EM by 33%, while that of parameter s by 85%!

Table 3. Parameters mb and s, strength M and modulus of deformation EM for different values of
GSI with C = 50 MPa, mi = 10 and D = 0 (after Malkowski, 2010).

GSI mb s M (MPa) EM (GPa)

35 0.981 0.0007 1.204 2.982
40 1.173 0.0013 1.654 3.976
45 1.403 0.0022 2.241 5.303
50 1.677 0.0039 3.011 7.071
55 2.005 0.0067 4.022 9.429
60 2.367 0.0117 5.350 12.574
65 2.865 0.0205 7.099 16.768
70 3.425 0.0357 9.401 22.361

In the above calculations, the empirical rock material constant mi is also

required to establish the constant mb (as per Table 2) in the Hoek-Brown criterion.
The actual material constant mi depends on a number of factors and can range
considerably for a given rock so that a table of values was published by Hoek and
Brown (1995). Later, a software, RocLab (, was made
available which provided estimates of mi. In addition, laboratory triaxial tests on
intact rock were reported by many investigators but selection of mi still remains quite
subjective. Most recently, Kwasniewski conducted 625 triaxial tests (Malkowski,
2010) on 75 rock types and 46 types of coal from various parts of the world and
concluded that standard deviations were so large that, although there was a broad
agreement on the range of values for various rocks, the actual scatter of values
throws doubt on reliable practical applications, and this calls for cross-checking of
the results.

Due to the above uncertainties with quantifying GSI, yet the importance of
the Hoek-Brown criterion, a practice in Europe is to determine the RMR first and
then to convert it to GSI using the expression of GSI = RMR - 5, as recommended by
Hoek et al. (1995). But, as was shown earlier, this equation is only valid for very
weak rock masses.

Also quite recently, Brown (2008) re-assessed their Hoek-Brown criterion by
providing a review of its uses, abuses and limitations. Clarifying that the criterion
was intended for use in making approximate estimates of peak rock mass strengths
for use in preliminary studies of rock engineering projects involving underground
excavations, he pointed out three aspects deserving special caution, as follows:
"(1) the criterion should not be used in cases where there are only one or two sets of
(2) there is a limit to the range of GSI values over which the criterion may be applied
with confidence: care must be exercised when dealing with brittle fracture in strong,
massive rocks with high GSI values of over 75, and similarly, special care must be exercised
at low values of GSI of below about 30, and low c cases as well as tectonically disturbed,
heterogeneous rocks. The estimation of GSI values in such cases is a task for experienced
specialists because the Hoek-Brown criterion may not apply for very weak rocks with c <
15 MPa since then the index a can be greater than its maximum value of 0.666 and can
approach one, a value that is usually taken to apply to soil.
(3) The disturbance factor D is subject to significant errors if applied to the entire rock

mass instead of to the disturbed zone of a few meters only."

Therefore, a cross-check of the results when using the Hoek-Brown criterion

seems essential. This may be achieved in conjunction with the Yudhbir-Bieniawski
criterion listed above, as pointed out by a study by Edelbro at al. (2006), comparing
all available criteria with the measured values, as listed in Table 5.

The Yudhbir-Bieniawski criterion is of the form given in Table 2. The

constant A has also been given as A = exp(0.0765RMR-7.65). The constant B is
obtained from triaxial tests on rock materials, five of which are depicted below in
Figure 5 (Bieniawski, 1974).

The constant is shown here as = 0.75 but for some rock materials there
are exceptions, for example, coal from the Pittsburgh seam is best characterized with
= 0.65 while B = 4.4.

The disturbance factor D introduced in the GSI index is unnecessary when using the RMR system
because an easier adjustment factor Ab forms part of the RMR procedure to characterize the quality
of blasting.

Figure 5. Intact rock criterion for failure in triaxial compression for five rock materials
(Bieniawski 1974).

Table 5. Rock mass strength M estimated by the Hoek-Brown and Yudhbir-
Bieniawski criteria compared with in situ test results (after Edelbro et al., 2006).

Laisval Case, M = 19.8 MPa1.4MPa, C = 180 MPa, RMR = 4718

M min max
(MPa) (MPa) (MPa)
Hoek-Brown-RMR (1980) 34.5 12.0 74
Hoek-Brown-GSI (2002) 15.0 2.5 50
Yudhbir-Bieniawski 16.3 4.2 53
Sheorey-RMR (1989) 37.0 1.5 81

Edelbro et al. (2006) concluded from the above study that the Hoek-Brown
and Yudhbir-Bieniawski criteria provided "reasonable agreement with the measured
strengths and (in spite of a large span between the minimum and maximum values)
are the best candidates for realistic strength estimation, provided that care is taken
when choosing values for each of the parameters in each method." Nevertheless, the
agreement with measured strengths is still relatively poor by all these methods
implying that precise estimates cannot be expected with any one criterion.

Finally, all the above criteria assume a homogeneous isotropic rock mass,
apply to relationships between the peak principal effective stresses, and, as pointed
out by the developers, "should not be used for the analysis of anisotropic or strongly
structurally controlled rock mass behaviour". Alas, as Brown (2008) observed, until
a more fundamentally-based approach that explicitly accounts for the factors that
influence the mechanical properties of rock masses is developed, the empirical
criteria remain the only alternative. Therefore, this writer believes strongly that it is
important to always cross-check the results, and not rely on just one method.



Not true. Some relationships are better substantiated than others, but there is a
great difference between "determining" and "estimating" rock mass
deformability: determining is highly desirable, estimating is done in the absence
of reliable in situ data for preliminary designs.

This myth concerns the in situ modulus of deformation, which indeed is
needed for tunnel design to determine deformations and displacements in a tunnel
under the load of the overburden and induced stresses. This type of input data is best
obtained by such in situ tests as plate bearing tests or large flat jacks, but these are
very expensive and time consuming, and accordingly seldom used nowdays, except
in very special cases (underground disposal of nuclear waste). It is correct that this
knowledge is best derived from a rock mass classification proved suitable for this
purpose. In fact, it was the RMR system that first proposed (in 1978) a direct
correlation between rock mass quality and the field modulus of deformation EM, as
depicted in Figure 6, which is preferable to using a ratio of the laboratory-obtained
modulus of elasticity EC to EM. The correlation was based on numerous in situ large
scale tests, carefully monitored and analysed, and the data obtained formed the bases
of further studies.

Figure 6. Correlation between RMR and the modulus of deformation of rock masses EM
(Palmstrm & Singh 2001)

Some of these studies were very useful, for example extending the validity of
the original relationship to lower quality rock masses (Serafim and Pereira, 1983);
others were not helpful by unnecessarily introducing another variable, intact rock
strength c, (which only added more uncertainly inherent in laboratory testing
procedures). Furthermore, the matter was complicated the by adaptations of the GSI
qualitative index to the field-oriented quantitative relationship featuring RMR, its
input parameters being measurable and hence preferable for engineering purposes, in

place of descriptive estimates. The argument that qualitative data are easier and
cheaper to obtain is clearly erroneous, as it was a step backwards adding more
empiricism to an already empirical approach. Nevertheless some designers and
planers accepted such short-sighted reasoning - therefore an equation which includes
intact rock c should be avoided. In addition, since the GSI index has the only
function to provide input for the Hoek-Brown criterion, its role for determination of
the modulus of deformation is secondary and should be left to the RMR for easier
use and much better accuracy. For example, the disturbance factor D introduced in
the GSI index is unnecessary when using the RMR system because an easier
adjustment factor Ab forms part of the RMR procedure to characterize the quality of
Accordingly, a practice sometimes observed of substituting GSI in place of
RMR in Figure 6 is incorrect because there is no equivalence between the RMR
classification and the GSI index, except for exceptionally poor rock conditions, as
was evident from the previous section.

In summary, the equations listed in Figure 6 above, for the two ranges of
RMR, are recommended as the best fit to experimental data and representing a
realistic approach which settled the matter by the work of Palmstrm and Singh,
2001. Using two ranges of RMR has the advantage that modulus values are NOT
overestimated at the higher range nor underestimated at the lower range. This is more
realistic than when relying on one sigmoidal equation.

An exception is permissible if reliable laboratory data are available on the

modulus of elasticity of intact rock material; they may be included in an expression
for the in situ modulus of deformation, because laboratory modulus determination
procedures are better standardized than laboratory strength c determinations.

For this purpose, to estimate the modulus of deformation of the rock mass, an
alternative expression is one derived by Galera et al. (2005):

EM (GPa) = EL

where EL is the modulus of intact rock determined in laboratory.

Galera (2007) also derived another useful expression involving both rock
mass modulus EM and rock mass strength M, namely

EM / EL = { M / C }2/3

where EL and C are laboratory determined values of the modulus and the
uniaxial compressive strength, respectively.

The above expression has the merit of a useful cross-check and it conforms to
an old practice proposed by Deere and Miller (1966): a strength - deformation
representation featuring the concept of the "modulus ratio (MR)", shown in Figure 7,
the idea being that, in the absence of measured values, EL may be estimated from the
EL = MR C.
Palmstrm and Singh (2001) and Hoek and Diederichs (2006) updated the MR
values for different rock types.

Figure 7. The concept of the Modulus Ratio (MR): a strength-deformation representation

for rock materials and rock masses (Deere and Miller 1966).

Most recently, Vn and Vsrhelyi (2010) derived two similar expressions:

E M / EL = M / C = and EM / M = MR
where MR is the modulus ratio representative of a given rock type, as tabulated by
Palmstrm and Singh (2001), for example for sandstone, MR = 257 (C = 109 MPa, EL=28

Clearly, the above expressions are yet to be tested with experimental data and
empirical relationships should not replace in situ tests for final designs. Indeed, the
current practice of avoiding field tests, even on major projects, is lamentable and
quoting budget and time constrains too often can jeopardize project safety.



A big mistake! There is a wealth of pertinent information to learn from "our

cousins" the mining engineers, for applications in civil engineering.

Both civil engineering and mining engineering have rich traditions and
achievements to their credit in the design and construction of civil engineering
tunnels and underground mining drift, caverns and stopes. Yet, there is remarkable
little interaction between the two disciplines, and this is particularly evident involving
rock mass classifications. Since I am involved in both fields and have struggled for
years to integrate the teams of civil engineers, mining engineers and geologists
working on my projects, I am convinced of many excellent opportunities lost for lack
of an exchange of ideas. For example, the applications of RMR, called mining RMR,
in Chile and Australia in block caving operations when mining hard rock deposits,
and in coal mining drifts in America, called Coal Mine Roof Rating, representing
weak rock mining, provided major contributions to design of support measures, large
underground spans and stability of coal and rock pillars, including the effect of depth
and induced stresses due to mining, Figures 8 and 9.

Figure 8. RMR input Parameters and Adjustments for Factors (Bieniawski 1989)

Figure 9. RMR classification scheme for mining applications (Kendorski et al. 1983).

The design of rock pillars in mining is a special area where civil engineers can
benefit from the research and experience of their mining colleagues. This concerns
dimensioning of rock pillars to account for the shape effect and determination of the
in situ strength of a rock mass, M. The well known equation (Bieniawski, 1984) for
this purpose, used extensively in mining in the USA, is:

pillar = M [ 0.64 + 0.36 w/h ]

where pillar : overall strength of a pillar (MPa)
M: rock mass (1m3) unit strength in situ (MPa) - from Figure 10a w: pillar
width (m)
h: pillar height (m).

The uniaxial compressive strength of a rock mass, M is determined from the

Kalamaras-Bieniawski criterion of 1995 (Figure 10a) which is also very useful for
civil engineering design applications involving tunnels and chambers.

Figure 10. A strength criterion for rock masses as a function of RMR and uniaxial compressive strength of
rock material (Kalamaras and Bieniawski 1995). Experimental data are from in situ shear strength tests in
Japan reported by Aydan and Dalgic (1998).

Figure 10b. Direct correlation between rock mass strength and RMR (Aydan & Dalgic (1998).

The RMR and Q rock mass classifications were independent developments in
1973 and 1974, respectively, whose common purpose was to quantify rock mass
characteristics previously based on qualitative geological descriptions. The value of
thorough geological exploration was always emphasized. In addition, it was
repeatedly stated that these classification systems were not "cookbooks" but had to be
used for the purpose for which they were developed, as part of the engineering design
process. This means an iterative procedure in the case of underground works, where
detailed knowledge of the ground is obtained from day to day.
At the time of the development of RMR and Q, geologists often worked in
separate teams from those of engineers, leading to potential misunderstanding of what
was required by whom, for engineering purposes. In fact, the advent of our rock mass
classifications seems to have stimulated an opportunity to combine the efforts of
engineers and geologists to act as one team, with clear statements of basic tunnel
engineering needs and some carefully selected and quantitative geological data
requirements. As seen in Table 6, this led to developments of many classification

variants for different purposes, such as tunnelling, mining, slope stability and dam

The scope of RMR and Q classification systems

The RMR and Q systems are particularly well suited in the planning stage of a
tunnelling project when a preliminary assessment of the most likely tunnel support
requirements is required, based on core logging, field mapping, and refraction seismic
surveys. During construction, application becomes even more essential, as the
appropriate support classes are checked on a day-by-day basis. The reasons for this
are as follows:

1) RMR and Q originated, and have been specifically updated, for estimating
tunnel support. Later they were extended for assessing rock mass properties, such as
the modulus of deformation, interpreting seismic velocities, and for assisting with
analyzing of monitoring data during construction, via convergence-quality-tunnel-
dimension links.

2) Estimating of rock mass properties for numerical modelling has turned out
to provide competitive alternatives to expensive and complex in situ tests, which rely
on a number of assumptions for interpretation of the data. Significantly, plate bearing
tests, large flat jack tests and pressure tunnel tests are nowadays rarely used because
of their expense, and because of difficulties with disturbed zone phenomena.

Table 6. Major Engineering Rock Mass Classifications

Country of
Name Originator and Date Applications
1. Rock Load Terzaghi, 1946 USA Tunnels with steel support
2. Stand-up Time Lauffer, 1958 Austria Tunneling
3. New Austrian Tunneling Pacher, Rabcewicz, 1964 Austria Tunneling
Method (NATM)
4. Rock Quality Deere et al., 1967 USA Core logging, tunneling
Designation (RQD)
5. Rock Structure Rating Wickham et al., 1972 USA Tunneling
(RSR) concept
6. Rock Mass Rating Bieniawski, 1973 South Africa & Tunnels, mines, slopes,
(RMR) system modified in 1989 USA foundations
Weaver, 1975 South Africa Rippability
RMR system extensions Laubscher, 1976 South Africa Hard rock mining
Olivier, 1979 South Africa Weatherability
Ghose and Raju, 1981 India Coal mining
Moreno Tallon, 1982 Spain Tunneling
Kendorski and Cummings, 1983 USA Hard rock mining
Nakao et al., 1983 Japan Tunneling
Serafim and Pereira, 1983 Brazil Foundations

Gonzalez de Vallego, 1983 Spain Tunneling
RMR system extensions nal, 1983 USA Coal mine roof bolting
Romana, 1985 Spain Slope stability
Newman, 1985 USA Coal mining
Sandbak, 1985 USA Boreability
Smith, 1986 USA Dredgeability
Venkateswarlu, 1986 India Coal mining (CMRS)
Robertson, 1988 Canada Slope stability
Thiel, 1985 Poland Carpathian flysch
nal, 1996 Turkey Weak rock, coal
Pakalnis et al., 2007 Canada Weak rock mining
7. Q-system Barton et al., 1974 Norway Tunnels, chambers
Kirsten, 1982 South Africa Excavatability
Q - system extensions Kirsten, 1983 South Africa Tunneling
Barton, 2000 Norway, Brazil TBM tunneling
8. Strength-size Franklin, 1975 Canada Tunneling
9. Unified classification Williamson, 1984 USA General, communication
10. Coal Mine Roof Rating Molinda and Mark, 1994 USA Coal mining
11. Geological Strength Hoek et al., 1995 Canada Rock mass characterization
Index (GSI)
12. Rock Mass index Palmstrm, 1995 Norway Rock engineering
13. Deutsche Steinkohle Witthaus, 2006 Germany Coal mining
14. Rock Mass Bieniawski et al., 2007 Spain TBM tunneling
Excavability (RME)

RMR and Q systems provide realistic estimates for modelling purposes, and through
seismic measurement and interpretation can assist in the interpretation of the
disturbed zone characteristics.

3) Appropriate monitoring and recording of one or both rock mass

classifications during construction is essential to quantify the encountered rock mass
conditions, select the appropriate support class, and is useful in case of contractual
disputes, arbitrations and design changes.

4) Technology has changed much in 38 years, since the RMR system was
developed, hence support materials and methods must be modified. Therefore major
updates have been made from time-to-time, such as the shift from mesh-reinforced to
fibre-reinforced lining.

5) RMR and Q were found to be equally effective in very poor rock masses
and in very good rock masses and it is incorrect to state that alternative descriptive
methods might be preferable in poor rock mass conditions. As engineering geological

techniques improve with advancing technology, our quantitative rating systems will
always be preferable to any qualitative descriptive assessments.
6) Both Q and RMR now form the basis of new TBM performance prognoses,
in the shape of QTBM and RME, which are developing both supporters and critics, as
is only to be expected, in our challenging work place.
7) Finally, as rock mass classifications are a part of an empirical approach, care
should be exercised so that the relationships involved, which are estimates, are not of
a 'black-box' level complexity, pretending accuracy beyond the meaning of the
empirical approach. Barton (2007) gave an example of GSI-based Hoek-Brown
formulations shown in Table 7; he called this "extraordinarily complex formulae
proposed for developing input data for some recent continuum models; on the left,
there is no possibility to have any feel for the influence of local rock quality on the
rock mass strength, deformation, friction angle and cohesion, when formulations
require software, rather than estimation for their evaluation. The formulae on the left
can no longer be considered "empirical". On the right in the table, the RMR and Q
formulations are much simpler and more suitable for practical rock engineering.

Table 7. The extraordinarily complex formulas (left) for input to continuum models and
comparison (right) with equivalent expressions (modified after Barton 2007)


Perhaps as a result of time and budgetary pressures, there has been a distinct
trend for using "convenient" continuum codes, which have particularly good graphics
representation of results. Simple software packages for handling the complex
calculations in Table 7 are also available, so that a user might need only limited
understanding of rock mechanics principles to use the codes "successfully". As a
result, a consultant's report might contain endless colourful stress distribution and
deformation patterns; but does all this 'color' represent anything real? What
continuum approximations were made? The fact is that using sophisticated codes
correctly, with realistic input data, needs experience, time and thus budgets to match.
Ironically, input data for some continuum models seems now to be considerably
more complex than for discontinuum codes, as suggested by Table 7.

Most of all, a pretence of accuracy in the complex expressions on the left in

Table 7 do not make any sense when, in any case, safety factors (also known as
"factors of ignorance"!) form a part of any tunnel design. They "cover up" errors,
large standard deviations and assumptions, usually allowing a considerable margin of
safety, with factors typically 1.5 for tunnels, 2.0 for large chambers and 3.0 and over
for large dams!


To avoid confusion, I would like to summarize the "Ten Commandments" of

broad principles for proper use of the two major rock mass classifications (Barton and
Bieniawski, 2008):

I. Ensure that the classification parameters are quantified (measured, not just
described), from standardized tests, for each geologically designated structural
region, employing boreholes, exploration adits, and surface mapping, plus seismic
refraction for interpolation between the inevitably limited numbers of boreholes.

II. Follow the established procedures for classifying the rock mass by RMR
and Q and determining their typical ranges and the average values;

III. Use both systems and then check with well-known published correlations
of Bieniawski (1976) and Barton (2008);

IV. Estimate rock mass properties, particularly the rock mass modulus (for
modelling purposes) - see Figure 6 - and the stand up time, according to Figure 11.
Note to include a correction for TBM constructed tunnels shown in Figure 12.

Figure 11. The stand up time of tunnels constructed by drilling and blasting as a function
of RMR (Bieniawski 1989). Black squares represent mining cases.

Figure 12.
Correlation between RMR
ratings from drill and blast
excavations and an
adjustment for TBM
excavated tunnels
(Alber 1993).

V. Estimate support requirements for primary rock reinforcement (Figure 12),
noting the two correlations for selection;

Figure 13. Chart for selection of tunnel reinforcement (Barton & Bieniawski 2007)

VI. Perform numerical modelling and, checking if sufficient information is

available, determine zones of factors of safety. Use at least two rock mass strength
criteria for comparison and cross-checking the results obtained from the Hoek-Brown

VII. If sufficient information is not available, recognizing the iterative design

process, request further geological exploration and parameter testing, e.g. stress
measurements, if necessary;

VIII. Consider the construction process, and in the case of TBM feasibility
studies, estimate the rates of advance, using the QTBM and RME methods;

IX. Ensure that all the rock mass characterization information is included in the
Geotechnical Baseline Report which discusses design procedures, assumptions and

X. Perform RMR and Q mapping as the construction proceeds so that

comparisons can be made of expected and encountered conditions, leading to design
verification or appropriate changes.

Of course, it goes without saying that laboratory tests must be included and
performed diligently according to standardized procedures and with a sufficient
budget. The engineers and geologists should act as a team and communicate regularly
among themselves and with the client.


Errors in rock mass characterization - in general - and those in rock mass

classifications and usage of strength and deformation criteria - in particular - require
special attention as numerous cases of recent tunnel failures have demonstrated

With reference to rock mass strength and deformation criteria, this presentation
is not against the use of the Geological Strength Index or the Hoek-Brown criterion;
indeed they are useful methods if used correctly. I simply believe that its limitations
are often overlooked when estimating rock mass properties and one must be careful
in accepting commercially available software without checking its assumptions and
approximations. Hence, at a minimum, one should use two strength criteria to cross
check the results.

The purpose of this presentation was to stimulate discussion of the most

relevant aspects in question, to emphasize the available solutions for engineers and
geologists, and to urge an appropriate action on the projects in planning and design,
as well as those already in construction where a cross-check on design
methodologies, the assumptions involved and methods used is highly recommended.

Alber, G. 1993. Classifying TBM contracts. Tunnels & Tunnelling, v. December, p.41-43.
Aydan, O, and Dalgic, S, 1998. Prediction of deformation behavior of 3-lanes Bolu tunnels through
squeezing rocks of North Anatolian fault zone (NAFZ). Proc. Reg. Symp. on Sedimentary
Rock Engng. Taipei, Taiwan, 228233.
Barton, N, 2000. TBM Tunnelling in Jointed and Faulted Rock. Balkema, Rotterdam, 173p.
Barton, N, 2007. Future directions for rock mass classification and characterization. Proc. Rock
Mechanics Challenges. Taylor & Francis, London, 179-188.
Barton, N and Bieniawski, ZT, 2008. Setting the record straight about RMR and Q. Tunnels &
Tunnelling International, February, 26-29.
Bieniawski, ZT, 1974. Estimating the strength of rock materials. S. Afr. Inst. Min.Met., 74, 312-320.
Bieniawski, ZT, 1976. Rock mass classifications in rock engineering. Exploration For Rock
Engineering, A.A. Balkema, Cape Town, 97-106.
Bieniawski, ZT, 1993. Engineering Classification of Rock Masses: The RMR System and Future
Trends. Comprehensive Rock Engineering, Pergamon Press, New York, v.3, 553-573.
Bieniawski, ZT, 1978. Determining rock mass deformability. Int J Rock Mech Min Sci, 15, 237-247.
Bieniawski, ZT, 1984. Rock Mechanics Design in Mining and Tunneling. A A Balkema, Rotterdam.
Bieniawski, ZT, 1989. Engineering Rock Mass Classifications: a Complete Manual. John Wiley
and Sons, New York.
Bieniawski, ZT, 1992. Design methodology in Rock Engineering: Theory, Education and Practice.
A A Balkema, Rotterdam.
Bieniawski, ZT, 2001. Construyendo puentes para el siglo XXI entre la Ingeniera, la Geologa y la
Sociedad. IngeoPres, Madrid, no.100, Deciembre, 10-22.
Bieniawski, ZT, Celada, B, and Galera, JM, 2007. Predicting TBM excavability. Tunnels &
Tunnelling International, September, 25-29.
Bieniawski, ZT and Grandori, R, 2008. Predicting TBM excavability-Pt 2. Tunnels & Tunnelling
International, January, 15-20.
Brown, ET, 2008. Estimating the mechanical properties of rock masses. Proc. SHIRMS 2008,
Australian Centre or Geomechanics, Perth, 3-22.
Celada, B, Fernndez, M, Rodrguez, A, and Tardguila, I, 2010. Definicin preliminary de las
secciones tipo de sostenimiento en los tunnels proyectado por mtodos convencionales.
IngeoPres, Mayo, nr 192, 16-22.
Deere, DU and Miller, RP, 1966. Engineering Classification and Index Properties for Intact Rock.
Rock Mechanics in Engineering Practice, eds Stagg K & Zienkiewicz O, Wiley, London, 1-20.
Edelbro, C, Sjlberg, J, and Nordlund, E, 2006. A quantitative comparison of strength criteria for
hard rock masses. Tunnelling & Underground Space Technology, 22, 57-68.
Galera, JM, lvarez, M and Bieniawski, ZT, 2005. Evaluation of the deformation modulus of rock
masses: comparison with dilatometer tests with RMR predictions. Proc. ISP5-PRESSIO 2005,
eds Gambin, Magnan, Mestat and Baguelin, LCPC Publications, Paris, ISBN 2-7208-0405-3.
Galera, JM, lvarez, M and Bieniawski, ZT, 2007. Evaluation of the deformation modulus of rock
masses using RMR: Comparison with dilatometer tests. Proc.11th ISRM Congress, Workshop
W1, Taylor & Francis, Madrid, 71-77. ISBN 97804-1545-0287.
Galler, R, 2010. NATM - The Austrian practice of conventional tunnelling. Proc. Conf. on
Underground Construction, Prague, Invited Lecture, 3-13.
Gonzles de Vallejo, LI, 2002. Ingeniera Geolgica, Clasificaciones Geomecnicas, Prentice Hall,
Madrid, 199; 229-239.

Grandori, R, 2007. TBM performance and RME classification. Proc. Jornada sobre Tuneladoras,
CEDEX, Madrid.
Grandori, R, Bieniawski ZT, Vinzzino, D, Lizzadro, L, Romualdi, P and Busillo, A, 2011. Hard
rock extreme conditions in the first 10 km of TBM-driven Brenner Exploratory Tunnel.
Proc. RETC, San Francisco, SME, Denver, 667-685.
Hoek, E and Brown, ET, 1988. The Hoek-Brown failure criterion - a 1988 update. Proc. 15th
Canadian Rock Mech. Symp., Univ. of Toronto, Toronto, 31-38.
Hoek, E, Kaiser, PK, and Bawden, WF, 1995. Support of Underground Excavations in Hard Rock.
A.A. Balkema, Rotterdam, 215p.
Hoek, E, Carranza-Torres, C, and Corkum, B, 2002. Hoek-Brown failure criterion - 2002 edition.
Proc. 5th North Amer. Rock Mechanics Symp., Toronto, 1, 267-273.
Hoek E and Diederichs MS, 2006. Empirical estimation of rock mass modulus. Int. J. Rock Mech.
Min. Sci., 43, 203-215.
Kalamaras, G and Bieniawski, ZT, 1995. A rock mass strength concept incorporating the effect of
time. Proc. 8th ISRM Congress, Tokyo, Sept., 295-302.
Kendorski, FS, Cummings, RA, and Bieniawski, ZT, 1983. A rock mass classification scheme for
mining. Proc. RETC, New York, AIME, 193-223.
Malkowski, P, 2010. Application of studies in mines and laboratories for selection of the constants
in the Hoek-Brown criterion, Achiwum Gornictwa, November.
Pakalnis, R, Brady, TM, Hughes, P, Caceres, C, Ouchi, AM, and MacLaughlin, MM, 2008. Weak
rock mass (RMR<45) design for underground mining. Proc. Int. Workshop on Rock Mass
Classification in Mining, NIOSH Information Circular 9498, Pittsburgh, PA, 119-127.
Palmstrm, A and Singh, R, 2001. The deformation modulus of rock masses - comparisons between
in situ tests and indirect estimates. Tunnelling & Underground Space Techn. 16 (3), 115 - 131.
Palmstrm, A and Broch, E, 2006. Use and misuse of rock mass classification systems. Tunnelling &
Underground Space Technology, 21, 21-30.
Pells, PJN, 2007. Limitations of rock mass classifications. Tunnels & Tunnelling Int., April, 33-38.
Pells, PJN, 2011. Against limit state design in rock. Tunnels & Tunnelling Int., February, 34-38.
Petroski, H, 1991. Paconius and the pedestal of Apollo: a case study of error in conceptual design.
Research in Engineering Design, 3, 123-128.
Priest, SD and Brown, ET, 1983. Probabilistic stability analysis of variable rock slopes. Transaction
Institution of Mining and Metallurgy, 92, January, A112.
Ramamurthy, T, 2008. Joint factor concept in soliving rock engineering problems. Proc. 5th Asian
Rock Mechanics Symposium, IRSMR, Tehran, Iran, 147-156.
Romana, MR, 1993. Slope Mass Rating (SMR) - La aplicacin de la clasificacin de Bieniawski a
los taludes. Comprehensive Rock Engineering, 3, Oxford, 575-599.
Romana, MR, 2003. Dam Mass Rating (DMR) - An adaptation of RMR classification for use in
dam foundations. Proc. ISRM 2003 Symposium, SAIMM, Johannesburg, 972-980.
Stacey, TR, 2004. Design - A Strategic Issue. Proc. 2nd Int. Seminar on Strategic versus Tactical
Approaches in Mining. Section 4. Australian Centre for Geomechanics, Perth.
Tzamos, S and Sofianos, AI, 2007. A correlation of four rock mass classification systems through
their fabric indices. Int J Rock Mech Min Sci, 4, 477-495.
Vn, P and Vsrhelyi, B, 2010. Relation of rock mass characterization and damage. Proc. Rock
Engineering in Difficult Ground Conditions, Taylor & Francis, London, 399-404.
Yudhbir, RK, Lemanza, W and Printzl, F, 1983. An empirical failure strength criterion for rock
masses. Proc. ISRM Congress, Melbourne, 1, B1-B8.


In preparation of this address, the Author requested and received valuable

comments from a number of his colleagues and friends, which enhanced the ideas
expressed in this presentation and informed him of the latest developments. I
specifically wish to thank the following persons:

Prof. Dr. Hasan Gercek, Zonguldak, Karaelmas University, Turkey

Dr Nick Barton, Norway
Prof. Marek Caa, AGH University, Krakow, Poland
Dr-ing. Benjamn Celada, Spain
Dr Evert Hoek, Canada

The Author

Professor Richard Z.T. Bieniawski, Ph.D-DSc (Eng), M.ASCE, is

Professor Emeritus of Mineral Engineering at Pennsylvania State
University, having retired to Prescott, Arizona, in 1996. Since then he
consulted on international projects in many countries as part of his
consultancy, Bieniawski Design Enterprises. From 2001, he is also
Distinguished Professor of Geotechnical Engineering at the University
of Madrid, Spain, where he received Doctor Honoris Causa and in
2003 an auditorium was named after him, La Aula del Prof.
Bieniawski, which is still in use by the Madrid Superior School of
Mines. Last year, he was awarded his second Doctor Honoris Causa
from the AGH University of Science and Technology in Krakow,
Poland, the city of his birth.
The author of 12 books, over 200 research papers (some translated into Spanish, German, Polish,
Russian, Chinese and Korean), he guest-lectured in many countries and universities and held
Visiting Professorships at University of Karlsruhe, Germany, at Stanford University, at Harvard
University and at the University of Cambridge, England.