By: D. Milne, Department of Civil, Geological & Environmental Engineering, University of Saskatchewan, Canada.

Abstract 

Rock Mass Classification is one of the key tools used in stope design for both empirical / graphical design techniques, as well as for numerical / modelling approaches. Both design approaches consider different modes of failure, and require classification systems that look at different properties of the rock mass. The choice of the best classification system, hence the best general design approach, should be based on the experience of the on-site engineer and the observations and case histories collected at the mine. It has been suggested that empirical stope design is best suited for early preliminary mine design work and numerical approaches are best when more detailed data, as well as case histories for model calibration, are available. This paper suggests that both approaches should be used at all stages of mine life. Empirical approaches benefit from more detailed data on rock mass data and numerical and empirical approaches both benefit from a better understanding of the rock mass behaviour, as well as stope case histories. A rock mass characterization approach is suggested in this paper which involves measuring key properties of the rock mass from which classification values can be derived.

Introduction

The three main classification systems used for stope design that are discussed in this paper include: the Q system (Barton et al., 1974), RMR₇₆ system (Bieniawski, 1976) and the GSI Strength Index (Hoek et al., 1995). The MRMR approach for assessing joint sets and the RMR89 system for characterizing joint condition is also referred to (Laubscher, 1976); (Bieniawski, 1989).

Both the Q and RMR₇₆ classification systems were developed for tunnel design and need some modification before they can be applied to open stoping mining methods. The GSI system is called a strength index rather than a classification system and has some built in assumptions that must be met be- fore it can be used for design. These assumptions will be further discussed.

To use classification systems for tunnel design, terms to assess loading conditions are included in the Q system, and a term to assess joint orientation conditions is included in the RMR₇₆ classification. If only an estimate of the rock mass condition is needed, these terms are not used. Rock mass classification, without the loading or orientation terms, give a value representing rock mass characterization.

The application of rock mass classification or characterization systems to stope design methods are improved through experience and calibration to case histories. The design calibration approach for empir- ical and numerical design methods differ significant- ly and will be discussed in some detail.

Rock mass classification versus characterization is discussed in the next section.

Rock Mass Classification vs Characterization

Tunnel design based on the Q and RMR systems de- veloped with the classification systems. Without a design application, the systems would not have had such a wide following, which became the basis for their success. Much of the following discussion is taken from Milne (2007).

With a rock mass classification value and tunnel span, support requirements and estimated tunnel stability can be obtained (Barton et al., 1974; Bieniawski 1976). The Q systems includes a factor to assess stress conditions and the RMR₇₆ system contains a term that assesses the orientation of discontinuities relative to the engineered structure. One of the main differences between tunnelling and mining applications of rock classification is the large variation in orientation, depth and geometry of underground openings in mining. Civil engineering applications are generally applied to tunnels at a fairly constant depth, orientation, and geometry; these conditions are seldom constant in mining applica- tions.

If mining applications included joint orientation and stress conditions in rock classification, the same rock mass could have dozens of classification values throughout the mine depending on the drift orienta- tion, mining level, and the excavation history (Milne et al. 1998). This would lead to significant confusion and make the classification systems very difficult to apply.

Both the RMR and Q classification systems are frequently adjusted for mining applications and are denoted with the use of the prime symbol (´). The Q´ system is used in numerous empirical design tech- niques and differs from the Q system in that the stress reduction factor (SRF) is set to 1.0 (Potvin 1988; Clark 1998). The RMR´ system is often used for mining span design. The RMR´ system does not include the RMR correction for joint orientation. The Q´ and RMR´ systems give rock mass characterization estimates and are used with the understanding that the effect of stress and joint orientation interaction with engineering structures will be assessed later in the design process.

Palmström et al. (2001) discuss the difference be- tween rock classification and characterization. Rock mass characterization should consist of the intrinsic properties of the rock mass, which include intact rock properties, discontinuity spacing and pattern, as well as discontinuity properties. If rock characterization is used, loading or environmental factors such as stress or discontinuity orientation should be considered later in the design process. Rock classification systems, however, should be treated as complete design packages and are to be used with the appropriate empirical design charts.

There has been some discussion concerning the assessment of groundwater factors in rock mass classification and characterization. Palmström et al. (2001) suggested that groundwater be excluded from rock mass characterization and added later in the design process since water conditions can vary significantly in the same rock mass.

The author feels that there are too many empirical and numerical design approaches that do not intrinsically account for the presence of water. To ignore water in the rock characterization stage of design risks the possibility that it will be ignored in the design process. Laubscher and Taylor (1976) incorporated water as a factor reducing the strength properties on the discontinuity surfaces in their modified RMR system (MRMR). There is some merit to this approach since it links the presence of water to its effect on the frictional properties of the rock mass.

There is also some confusion as to the application of water conditions with the stability graph design method for underground openings (Potvin 1988). Hoek et al. (1995) state the following concerning the application of Q´ for the stability graph method: “The system has not been applied in conditions with significant groundwater, so the joint water reduction factor Jw is commonly 1.0.” The groundwater term in the Q´ classification is sometimes ignored when using the stability graph design method. This is not a conservative approach because there is nowhere else to assess groundwater conditions in this design method. Similar confusion exists with determining “m” and “s” failure criteria for design. The original “m” and “s” factors were based on RMR76 classifica- tion values, with water considered (Hoek and Brown, 1980). The m and s values are now based on GSI values which do not include an assessment of groundwater. Before using this approach for design in areas where water is a concern, the engineer must consider if the modelled assessment of water pres- sure adequately reflects the drop in stability that could be due to the decrease in frictional properties of the rock mass, as well as the effect on stress.

Factors Used in Rock Characterization and Their Weighting

The rock characterization methods used in this paper have significant differences, however, they all quan- tify basic properties of a rock mass and combine them to represent overall rock mass properties. Each system selects what it considers the most significant properties of a rock mass to which a rating is as- signed. The addition or multiplication of these individual ratings provides a unique value representing the quality of the rock mass. Table 1 summarizes some of the key properties used in the rock mass characterization systems discussed in this paper.

It is important to note the weightings given to rock mass properties in these characterization systems. Before choosing an appropriate method for characterizing the rock mass for stope design, the engineer or geologist must use their judgement to determine if the rock mass properties that appear to have the greatest influence on stope stability, are the same properties assessed in the characterization method. It is interesting to note the large range in assessments for the influence of rock mass block size. This is discussed in the next section.

Rock Classification, Strength Factor and Scale Effects

The scale effect in empirical design is the interac- tion of the size of the engineering structure being de- signed, with the size of the intact blocks of rock bounded by discontinuities. The factor describing the intact block size in a classification system gives an indication of sensitivity of the classification system at different ranges of block size.

There is a difference between the RMR₇₆´ and Q´ assessment of block size. In the Q´ system, the RQD term looks at block size, coupled with Jn (joint sets factor) that partially takes into account block shape with the factor representing number of joint sets present. The RQD term is sensitive to varied joint spacing over a limited range. Figure 1 compares the effective range of RQD compared to the joints per cubic metre term, and block volume. Figure 1 shows that RQD can differentiate between a rock mass with 4 to 30 joints per cubic metre. RQD cannot differentiate between joint spacings below 30 joints per cubic metre or above 5 joints per cubic metre. There is a term in the Q system where 1.0 is added to the joint roughness term Jr, if the relevant joint set spac- ing is greater than 3.0 metres, but this does not com- pletely address the lack of sensitivity to large or small joint spacings.

For the RMR₇₆´ system, the RQD term is also used, but it is coupled with a term for joint spacing which greatly increases the range over which the RMR₇₆´ system will be sensitive to changes to the joint spacing (Figure 1).

The GSI system does not consider block size, however, both the Q´ and RMR₇₆´ give block size significant weight. The GSI system produces a value referred to as a Strength Index, which is linked to an estimate of the strength of the rock mass. The Q´ and RMR₇₆´ values are linked to the properties that influence rock mass behaviour. This is an important distinction and is best described by comparing rock mass strength to the unconfined compressive strength of intact rock. When testing an intact rock, the following guidelines are provided for sample size (ASTM Standard D7012-10, 2010): “The diameter of rock test specimens shall be at least ten times the diameter of the largest mineral grain. For weak rock types, which behave more like soil (for example, weakly cemented sandstone), the specimen diameter shall be at least six times the maximum particle diameter.” 

This guideline is provided because mineral grains will often have different properties than the surrounding matrix. By ensuring the sample size is much greater than the grain size, the sample behaves like a more homogeneous material. This same logic can be applied to block size which suggests that as long as the sample size or engineering structure under load, is at least 10 times the diameter of the average block size, the block size will not influence the overall rock mass strength. This could be treated as a guideline to determine if the application of the GSI Strength Factor is appropriate to use in a given situation. It is, however, difficult to know if this same ratio of mineral grain size to sample size pertains to the block size of a rock mass. It may be challenging to estimate when the GSI system is appropriate to use. Work has been done to augment the GSI system with the addition of RQD to assess block size (Cai et al., 2004); however, it appears that GSI was specifically developed to avoid this approach.

Consider the ASTM guideline for sample size exceeding ten times the mineral grain size; if this guideline is met, the sample strength will not be affected if the grain size is reduced. The application of this analogy to intact block size in a rock mass has some validity. A large pile of angular gravel may form to an angle of repose of approximately 38^o. If the average sample size was cut in half, and maintained the same angular character, the angle of repose would not change significantly. This suggests that the GSI strength factor can only be applied where the loaded volume of rock is very large relative to the block size. If a reduction in intact block size would influence the stability of the engineering structure under design, then the GSI term should not be used for design since it is not sensitive to block size or discontinuity spacing. The author does not know how an appropriate ratio of engineering structure size to block size could be determined, but suggests it is a determination experienced users of the GSI Strength make. The following provides some guidelines for the application of GSI for failure analysis; “The Hoek-Brown failure criterion is only applicable to intact or to heavily jointed rock masses which can be considered homogeneous and isotropic.” (Hoek et al., 1995).

Qualitative versus Quantitative Rock Mass Characterization

The Q´ and RMR₇₆´ systems consist of assessments of the size and perhaps shape of intact blocks of rock bounded by discontinuities, the discontinuity surface condition or frictional properties, intact rock strength, and groundwater conditions.

The method of assessment of these categories has evolved from mainly subjective assessments of factors to more qualitative assessments. Intact rock strength, joint spacing, and groundwater are assessed in fairly quantitative terms. The discontinuity assessment term is more subjective and contains descriptions such as “very rough surfaces” and “slight- ly rough surfaces,” which require experience to differentiate between and do not provide a very precise assessment. The RMR classification system has evolved to be more quantifiable (Bieniawski, 1989), and others have attempted to improve rock mass characterization by improving our ability to measure rock mass properties such as discontinuity surface properties. The joint assessment for RMR₈₉ is more quantifiable and could be used with RMR₇₆´ if the RMR89 rating, which is out of a maximum of 30, was scaled down to the RMR₇₆´ value out of 25.

The Q´ system is possibly the least subjective method currently in common use. The more analytical and quantitative descriptions used in the Q´ system are coupled with an assessment of more rock mass parameters, and these assessments are divided into many more categories. For instance, the RMR₇₆´ system describes the condition of discontinuities with five broad categories. The Q´ system, with its assessment of smalland largescale roughness, alteration, and infilling can differentiate between more than 60 conditions of joint surfaces. The Q´ system can give very precise rock classification values; however, this results in making repeatability more difficult to achieve and also increases the time required to obtain an estimate of rock classification.

Descriptions of joints and block size in the Q´ system can be somewhat subjective and research has been conducted to make these categories more quantifiable (Milne et al., 1991; Hadjigeorgiou et al. 1994) and intact block size distributions (Hadjigeor- giou et al., 1998).

The GSI system is the newest commonly used rock mass classification system. It makes a conscious attempt to move away from classification systems that quantify or rate individual properties of the rock mass. In his discussion of the development of the GSI classification system, Hoek (2004) states: “It was also felt that a system based more heavily on fundamental geological observations and less on ‘numbers’ was needed.” The GSI classification system consists of six categories describing the size and shape of intact rock blocks and five categories describing the surface condition of discontinuities. This system is based on geological observations and avoids an engineering approach of dividing the properties of a rock mass into components and measuring these components as accurately as possible.

The GSI system has been developed to provide rock mass properties for numerical modelling, which may account for the different approach taken for assessing the rock mass. Practitioners are encouraged to avoid precise estimates of classification, but rather to give a range representative of the highly variable properties found in natural materials such as rock masses. This approach is also well suited to numerical modelling, where more precise estimates of rock mass properties may rely on back analysis of observed rock mass behaviour.

Design Calibration Using RMR76’, Q’ and GSI For Design

The calibration of design methods is a wellestablished practice to verify or refine a design approach and is used in both empirical and numerical design methods. In some instances, design calibration consists of adjusting the least well defined parameter in the analysis until the design matches observed conditions. This can be done for slope stability studies where groundwater conditions are calibrated to fit an observed failure. Groundwater conditions may be chosen if they are the least well known input parameter. In empirical design approaches, the design line or design prediction is often adjusted to fit observed rock mass behaviour.

Figures 2 and 3 show dilution measured from a Uranium Mine in Northern Saskatchewan plotted on the Modified Dilution Graph (Capes et al., 2005); (Forster, 2013). Figure 2 shows the data from stopes with between 0.5 and 1 metre of dilution, while Figure 3 shows data between 1 and 2 metres of dilution. Calibration of this dilution prediction design tool could consist of shifting the blue line that separates the two groups of data lower. Shifting the dilution prediction line lower for this data suggests that the dilution which has occurred is less than the value suggested by the design method, possibly due to improved blasting, reduced borehole deviation or some other factor typical to this mine that is not built into the design method or rock characterization technique.

In numerical modelling techniques, stress driven failure is often assessed based on rock mass proper- ties such as m and s, which are based on rock type and the estimate of the GSI Strength Factor. Since the GSI value is based on subjective assessments of the rock mass, it is often given as a range. Signifi- cant adjustments to m and s values can be justified based on varying the GSI value used for design to fit with observed rock mass behaviour. This is an effec- tive approach for model calibration, however, as rock conditions vary, additional model calibration may be required.

Guidelines for Stope Design

Collecting data for rock mass characterization is one of the first steps in the stope design process. The en- gineer or geologist collecting this data should not be limited to data for a single characterization method, as this will tie them to a design method that may not be appropriate. As mentioned, the design method used tends to determine the failure mode that will be interpreted. Table 2 shows an example of the data that should be collected for rock characterization for open stope design. The Q’ value assesses data from the properties of the joint set critical to design. The RMR₇₆´ value uses an average joint set spacing and the GSI value is the range within 1 segment of a GSI chart. The top half of Table 2 shows the data and ob- servations that need to be made. The correlations linking joint condition and spacing to the Q’ classi- fication values are from literature (Milne et al., 1998).

Estimating the mode of stope instability before conducting your design is critical. With extensive calibration, most design methods will often predict rock mass behaviour correctly, even if the failure mode is not appropriate. Changing conditions will require frequent re-calibration if the failure mode designed for is not appropriate.

In high stress or bursting conditions, strength – stress based modelling using the GSI system is needed. In cases of low stressed / relaxed stope hanging walls, where the stope size to block size is likely critical, empirical design methods like the RMR₇₆´ span design, the stability graph and the dilu- tion graph approaches will likely be key (Potvin, Y., 1998); (Clark, L., 1998); (Capes, 2009); (Wang et al., 2003).

An additional approach that should be followed is to estimate the properties that influence stope failure based on case histories and observations. If stope de- sign appear sensitive to changes in the maximum in- duced stress, concentrate on numerical modelling approaches looking at stress. If RQD appears to in- fluence stability, use a characterization method that directly measures RQD and joint spacing. If rock mass alteration is the key factor, any of the three rock mass characterization methods may be appropriate. If stope stability is sensitive to the HW dip, empirical methods may be the best approach. If secondary stopes give you a lot more trouble than primary stopes, stress is likely your concern and you need to model it.

Most numerical modelling approaches are sensitive to the shape of an excavation in terms of the relative length width and height. The extent of any predicted failure could be expressed as a fraction of the minimum excavation dimension, independent of actual size. If depth of failure only occurs when the opening geometry reaches a critical size and then increases rapidly with opening size, empirical methods that consider opening extent may be more appropriate (Figures 2 and 3).

Conclusions

1. Empirical and numerical approaches should both be used at all stages of mine design. Basic rock properties of the rock mass should be measured in a quantifiable repeatable fashion, coupled with general observations, as shown in Table 2. This will allow most classification systems to be obtained from the collected data so that inaccurate correlations to link classification systems can be avoided.

2. It is important to continually make observations and assess case histories that provide data on the rock mass response to mining activity. This provides the engineering or geologists the tools they need to conduct appropriate design.

References

ASTM Standard D7012010. 2010. Standard test method for compressive strength and elastic moduli of intact rock core specimens under varying states of stress and temperatures. West Conshohocken, PA: Author. https://doi.org/10.1520/ D7012-10

Barton, N., R. Lien R. & J. Lunde, 1974. Engineering Classification of Jointed Rock Masses for the Design of Tunnel Support. Rock Mechanics, 6, pp. 189-236.

Bieniawski, Z.T. 1976. Rock Mass Classification of Jointed Rock Masses. Exploration for Rock Engineering. Z.T. Bieniawski Ed. Balkema, Johannesburg, pp. 97-106.

Bieniawski, Z.T. 1989. Engineering Rock Mass Classifications. John Wiley & Sons p. 251.

Cai, M., Kaiser, P., Uno, H., Tasaka, Y. and Minami, M. 2004. Estimation of rock mass deformation modulus and strength of jointed hard rock masses using the GSI system. International Journal of Rock Mechanics and Mining Sciences, vol. 41 p. 3-19.

Capes, G., Milne, D. and Grant, D. 2005. Stope hangingwall design approaches at the Xstrata Zinc, George Fisher Mine, North Queensland, Australia, American Rock Mechanics Symposium, Fairbanks, USA.

Clark, L. 1998. Minimizing dilution in open stope mining with a focus on stope design and narrow vein longhole blasting, MSc. Thesis, University of B.C., 316p.

Forster, K. 2013. Inferred weak rock mass classification for stope design, MSc Thesis, University of Saskatchewan, p. 194.

Hadjigeorgiou J, Germain P, Potvin Y. 1994. In situ estimation of rock joint alteration. Can Tunnelling, pp. 155–162.

Hadjigeorgiou J, Grenon M, Lessard J. 1998. Defining in-situ block size. CIM Bull 91(1020):72–75.

Hoek, E. and Brown, E.T. 1980. Underground excavations in rock. London, Inst. Min. Metall., p. 527.

Hoek, E., Kaiser, P.K. and Bawden, W.F. 1995. Support of underground excavations in hard rock. Rotterdam, Balkema, p. 215.

Hoek E. 2004. A brief history of the development of the Hoek- Brown failure criterion. Discussion paper #7. [http://www.rocscience.com/hoek/pdf/history]. March 2007.

Laubscher, D. and Taylor, H. 1976. The importance of geome- chanics of jointed rock masses in mining operations, Proceedings of the Symposium on Exploration for Rock Engineering, S. Africa, pp. 119-128.

Milne, D., Germain, P., Grant, D. and Noble, P. 1991. Field observation for the standardization of the NGI classification system for underground mine design. International Congress on Rock Mechanics, A.A. Balkema, Aachen, Germany.

Milne, D., Hadjigeorgiou, J. and Pakalnis, R. 1998. Rock mass characterization for underground hard rock mines, Tunnelling and Underground Space Technology, vol. 13, Nº. 4, pp. 383-391.

Milne, D. 2007. Problems with rock classification for empirical and numerical design, Proceedings of the International Workshop on Rock Mass Classification

Palmstrom, A., Milne, D. and Peck, W. 2001. The reliability of rock mass classification used in underground excavation and support design, ISRM News Journal, Vol. 6, Nº. 3, pp. 40-41.

Potvin, Y. 1988. Empirical open stope design in Canada, Ph.D. thesis. The University of B.C., p. 350.

Wang, J., Milne, D., Yao, M., Allen, G. and Capes, G. 2003. Open stope exposure time and stope dilution, CIM AGM, Montreal, 2003.