Review of Rock-Mass Rating and Tunneling Quality Index Systems for Tunnel Design: Development, Refinement, Application and Limitation

Although rock-mass rating (RMR) and tunneling quality index (Q) systems are used in different rock engineering projects as empirical design tools, their application in tunnel design is widely accepted as these systems were developed and updated for this purpose specifically. This paper reviews the work conducted by various researchers since the development of these two systems with respect to tunneling only. Compared to other empirical classification systems, these systems received international acceptance and are still used as empirical design tools in tunneling due to their continuous updates in the form of characterization and support. As the primary output of these systems is the initial support design for tunnel, however, their use in the calculation for rock-mass properties is an essential contribution of these systems in rock engineering design. Essential for the tunnel design, these rock-mass properties include the modulus of deformation, strength, Poisson’s ratio, Mohr-Coulomb parameters and Hoek-Brown constants. Other application for tunneling include the stand-up time and rock load. The uses and limitations of these systems as empirical tunnel design tools are also included in this review article for better results. Research to date indicates that if the ground behavior is also taken into account, the application of these empirical systems will be more beneficial to the preliminary design of tunnels.


Introduction
In rock engineering design, the rock mass is considered as a complex natural geological building material and during tunnel construction, ground behavior is dependent on the ground condition and tunnel-related features [1]. The reason for the overall difficulty of modelling rock masses for engineering design is that the physical or engineering properties of rock mass must be established. This natural geological material is generally discontinuous, anisotropic, non-homogeneous and non-elastic [2] and produces uncertainties in modelling a rock mass. Modelling a rock-mass in a precise way allows scholars to decrease the ambiguity related with its characterization and replicate its inherent spatial inconsistency and heterogeneities [3]. For the purposes of rock engineering design, eight modelling methods have been categorized, including pre-existing standard methods, analytical methods, basic and extended numerical methods, precedent type analysis, empirical classification, the basic system approach and the combined system approach [4]. An important fact in rock mechanics is that having inadequate information for engineering design of structure in rock is a way of life, Aydan et al. [7] The RMR and Q systems are admitted currently as a compulsory adjunct for evaluating rock-mass environments for engineering purposes. The two main objectives of these classification systems are easier communication between different users and as a decision-making tool [20]. The later objective is the remarkable practice of classification systems in rock engineering. The RMR system was suggested in 1973, as a jointed classification system for rock-mass [8]. This system was established on experience in underground projects in South Africa [21]. Afterwards, this system went through substantial variations. These variations are commonly due to ratings added or modified for ground-water, joint spacing and their condition, rock alterability, excavation method and stress-strain behavior. This system was numerously refined numerous times and main modifications in its characterization and structure were made in 1989 and 2014 [22,23]. The major changes in the structure of Q system were made in 2002 [24]. This system was established on tunneling case records for hard and jointed rock masses [25]. It was initially developed for the grouping of rock masses for assessing the need for supports in tunnels and caverns. Using a set of tables, for the characterization of rock mass, along with footnotes, the rating for various parameters can be recognized. The parameters rating can be based on the geological interpretations in tunnels during its construction, or by core logging during exploration. Through time, the RMR and Q systems have been updated so as to increase their trustworthiness for tunnel support. Along with tunnel support, these systems are also used to calculate the rock masses' geomechanical properties, the stand-up time for the tunnel and support load. Although these systems were developed, refined and updated for tunnel support design, these classification systems also have applications in various engineering projects like mining [26][27][28][29][30][31][32], slopes [10,[33][34][35][36][37], dam foundations [38] and TBM (Tunnel Boring Machine) tunnels [17,19,39].
The RMR and Q systems have changed with time to reveal the apparent impact of several rock-mass factors on the stability of excavation. The focus of this review is on the historical progress and alteration of the two systems through rock-mass characterization and support recommendation with a particular emphasis on tunneling. The advanced versions of RMR and Q-system classification and the scope of its application in the field of tunneling are specified. The use of these systems in tunneling other than support recommendation is also discussed. The introduced changes have perhaps improved the applicability of RMR and Q systems in the field of tunneling but they have still limitations. The limitations of the two systems for use in the field of tunneling are also reviewed.

Development of the RMR System
To develop RMR system, the first three existing rock-mass classification systems of Table 1 were considered. Bienwiaski combined the best features of these available classification systems and supplemented them with the parameters that could be obtained from the data available from exploration, resulting in an RMR system, as shown in the Table 2 [8]. This RMR system, RMR 73 , comprised eight parameters for classification of a jointed rock mass, in which each parameter has five important ratings. To rate the state of weathering, three different classifications were used: the Geological Society of London, Task committee of the American Society of Civil Engineers and the South African section of the Association of Engineering Geologists. The engineering classification of intact rock, proposed by Deere [15], was used in the RMR system with a slight modification. Classifications for joint spacing were made as per the suggestion by Deere [40]. Ground water and joint orientation were included as separate governing parameters in the RMR structure as per the study of Wickham, Tiedmann and Skinner [41]. Rating for all eight parameters were proposed based on this study. The summation of all the parameters determines the rock-mass quality. This system classifies the ground into five rock-masses. Table 2. Characterization criteria in different versions of the rock-mass rating (RMR) system.

Development of the Q System
This classification system was established in 1974 after the RMR system based on 212 tunnel cases. The development of this system was the result of the Seminar on Large Permanent Underground Openings held in Oslo in 1969. Two main breaches pointed out in the symposium were: (i) the acquisition of mechanical data and whether an excavation should be lined, or rock bolted, or unlined [47]; and (ii) the dilatant property of the rock had been unnoticed when designing rock-bolt spacing [48]. This system was independent from the RMR system but there are some parameters in were united within the term "discontinuity condition" with variations to the relative points [21]. Thus, the eight parameters creating the system was reduced to six. This revised RMR was again modified for intact rock strength (σ c ) and negative values were applied as adjustment parameters for the first time in 1975 [43]. The rating range for the first-class rock-mass (90-100) was changed to 81-100. The joint roughness was added as an additional parameter to the system and as a result, the assessment score was increased for the parameter "condition of joints." In 1979, modifications associated to the joint spacing, joint condition and ground water were made [44]. Main alterations to the RMR system were suggested in 1989 [22]. These comprised new graphs and tables for the score of σ c , RQD, spacing and situation of discontinuities. The refined system is called RMR 89 . In RMR 89 , the basic RMR (RMR b ) score is the sum of the scores of five stated parameters. It comprises the score for σ c (R 1 ), RQD (R 2 ), joint spacing (R 3 ), joint condition (R 4 ) and groundwater condition (R 5 ). The RMR 89 system is the outcome once adjusting RMR b for joint direction with respect to the tunnel alignment and direction of excavation score (R 6 ). The corresponding quality is achieved using Equation (2).
The minimum values of R 2 and R 3 for RMR 73 version of RMR are 3 and 5 respectively, as listed in Table 2. A continuous-rating idea through equations was presented by Sen and Sadagah for R 2 and R 3 values of RMR 73 system. According to RMR 89 , the R 2 and R 3 values are from 0 to 20, after modification [45] (as shown in Table 2). The continuous-rating concept through equations, which was introduced by Sen and Sadagah for RMR 73 , was extended for RMR 89 to characterize R 2 and R 3 easily and precisely [58]. Since its development [59], RQD was used in RMR and Q system because of its historical background. Because of some intrinsic limitations and ignoring the basic definition [60][61][62], RQD, along with joints spacing, is substituted by joint frequency in the excavation face or rock core [46]. Koutsoftas [63] analyzed the arguments about the different definitions of RQD in different parts of the world and the use of rock exposures for determining RQD which are considered as the main sources of error. In fact, these are modifications made in the RQD for intended purposes and adopted wrongly. However, it cannot be considered as a system fault and is not sufficient reason to abandon the use of RQD in classification systems. The modified criterion shows that RMR 89 can be stated in terms of five parameters and calculated using Equation (3): where R 2-3 shows the rating for joint frequency. After 25 years of use, an updated version of RMR 89 , known as RMR 14 , was proposed with new parameters, a revised rating and final structure [23]. The structure of RMR in RMR 14 version is represented by Equation (4).
where RMR b = RMR basic (without the effect of excavation), F 0 = An adjustment factor like R 6 in RMR 89 , F e = An adjustment factor associated to the method of excavation, and F s = An adjustment factor associated to the stress-strain behavior at the tunnel face. The discontinuity condition term is a function of discontinuity persistence, aperture, roughness, infilling and weathering. The corresponding score can be nominated for RMR 89 using Table 3 but some situations are mutually exclusive; that is, in the case of infilling material, the roughness is overshadowed by the infilling influence [22]. The score for the discontinuity condition was reviewed in RMR 14 and, because of the overshadowing effect, the aperture parameter is eliminated. The details are shown in Table 3. Additional parameters, namely the intact rock alterability, F e and F s , were also incorporated in RMR 14 . The slake durability index is an important parameter to estimate qualitatively the durability of shale or other similar rock in the service environment and the intact rock alterability is rated per the results of the slake durability test, as defined in the standard (ASTM D 4644) [64]. The score of the rock alterability factor were proposed in tabular form in RMR 14 . Mechanical excavation has a positive impact in tunneling [65] and this impact was introduced in the structure of the RMR 14 system related to the excavation method. It can be obtained from Equation (5) [23].
When the excavation is made by the drill and blast method, F e = 1.
The RMR score at the tunnel face during excavation is significantly lesser than forecast during the design stage because of the yielding of the excavation periphery. To overcome this consequence, F s is announced. This parameter was defined in terms of ICE (Índice de Comportamiento Elástico) [66], as stated in Equation (6). The Kirsch solution for stresses at the periphery of a circular tunnel was used for ICE calculation and has been prolonged to non-circular excavations using numerical modelling. For the derivation of Equation (6), Equation (7) [67] was used to calculate the rock-mass strength.
where σ c = intact rock uniaxial compressive strength (UCS) (MPa), K 0 = virgin stress ratio, H = tunnel depth (m), and F = shape coefficient. The value of ICE achieved from Equation (6) is used for the calculation of the adjustment factor, F s , using Equation (8).
Since the introduction, the Q system has not been tremendously updated like the RMR system in terms of characterization. The foremost modifications in characterization were suggested in relations of σ c and SRF [68][69][70]. In the rock-mass properties, the role of σ c is significant. Thus, a normalization factor is applied to Equation (1) for a modified Q c : The initial maximum value of SRF for the condition of competent rock with a rock-stress problem was 20 [9]. The first relationship for SRF characterization was based on the principal field-stress ratio (k), cover depth of the tunnel (H) and σ c [71]. The key changes were made for SRF characterization in hard massive rock in a high-stress environment and the rating for SRF increased from 20 to 400 [68]. These changes were based on the relation between RQD/J n and SRF in hard rock under stress and no real stress problem was experienced for low values of RQD/J n . The ratio RQD/J n is the relative block size that is appropriate for differentiating massive, rock-burst-prone rock and the rock-burst-prone rock has RQD/J n ratio from 25 to 200 whereas typical jointed rock mass has an RQD/J n of 10 [24]. Tunneling in jointed rock under high stress is likely to be less hazardous than in massive rock because the rock-bursting phenomenon is less severe [72]. For Australian mining, where the high stresses are either due to mining depth or the advancing mining front acting on jointed rock, a relation was proposed by Peck for a low SRF (SRF = f (strength stress ratio)) value [73], which was confirmed by Barton [74]. The continuous rating of SRF based on tunneling data under a highly stressed jointed rock-mass environment was modified such that SRF is a function of relative block size (RQD/J n ), intact rock strength (σ c ) and strength-to-stress ratio (σ c /σ 1 ) [70].

Development in Terms of Support
Keeping the basic purpose of the two systems, the RMR and Q systems were also updated in terms of support guidelines along with improved characterization for rock-quality determination.

RMR System
The initial guidelines were for drill and blast excavated tunnel support and were based on the rock-mass class in a tabulated form for a tunnel span from 5 to 12 m [8]. The length of fully resin-grouted rock bolts (25 mm in diameter) was purely based on the tunnel width, where the length of a rock bolt is half of the tunnel width. The other supporting materials were shotcrete, steel sets and wire mesh. In the 1976 version of RMR, the support guideline table was updated for a 10-m-wide tunnel with a horseshoe shape [75]. The rock bolts length and the thickness of shotcrete was reduced. In 1979, the length of the 20-mm-diameter fully grouted rock bolts was made a function of the RMR value [44]. The limit of vertical stresses was specified. Separate guidelines were suggested for support in walls and the crown of the tunnel. In 1989, the support chart remained unchanged and refinement of the system was made in characterization. Support technology upgraded for tunnels with time and thus, the designers used support for different tunnel spans [76][77][78]. The foremost update in support using RMR system was accomplished, after twenty four years of experience, in which the tunnel support is a function of the tunnel size and rock-mass score [46]. In this revision, the shotcrete thickness and rock bolt length (L b ) are functions of rock-mass score and tunnel span. The length of the rock bolt can be calculated using Equation (10).
Rock-bolt spacing (S b ) is the function of RMR score only. When RMR 89 > 85, spot bolting is required. In shotcrete design graph, the shotcrete thickness is greatly influenced by the tunnel size at low RMR 89 score, whereas shotcrete thickness is greatly influenced by RMR 89 for larger tunnel spans (Span > 10 m) [46]. The rock-bolt spacing can be obtained from Equation (11).
To classify a rock mass for its quality, Q system was updated by continuous re-analyzing of the tunnel case record until a reliable connection was achieved between Q value, the equivalent dimension (D e ) and the actual support used for the stability of the tunnel [9]. These three variables are organized by means of a support chart in the Q system of rock-mass classification. The D e is a function of the excavation size and its purpose and this purpose of the excavation was expressed in terms of the excavation-support ratio (ESR). The conclusion of the support recommendation chart was 38 box categories of permanent support. This support was for the crown of the tunnel. For the tunnel wall support, Q was increased up to 5 times and then the support was recommended. The recommended supports were bolting (spot or systematic), grouting, shotcrete (plain or mesh-reinforced) and cast-concrete arches (plain or steel-reinforced). In the 1970s, the wet process of shotcrete along with fiber reinforcement was a revolution in tunnel support. The innovation in shotcrete technology was summarized by Barrett and McCreath [79]. In 1993, mesh-reinforced shotcrete was replaced by fiber-reinforced shotcrete [68]. The 38 box categories are simplified to the graphical method (9 reinforcement categories) based on 1050 more tunnel case records. Rock-bolt spacing was correlated to the rock-mass quality value in both shotcrete and non-shotcrete area. Out of the 9 reinforced categories, reinforced ribs of shotcrete were also added for one category. Based on numerical modelling [80,81] and empiricism, the Q-system support chart was again updated by Grimstad for reinforced ribs of shotcrete [82]. The thickness of shotcrete was increased due to the widespread application of shotcrete in tunneling. The diameter of the rock bolt was specified as 20-25 mm in diameter. In the latest support chart [25], the bolts spacing is quantified for a 20-mm diameter and detailed adjustments are included for the tunnel wall support. It is also mentioned that the shotcrete thickness for excavation with D e less than 3 m is just supposed and is not empirically based. As in the RMR system, bolt spacing is a function of Q value and thickness of shotcrete is function of the D e along with the Q value.

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the essential parts of rock engineering design. To consider the scale effect for rock masses, field tests are desired, as laboratory testing on rock masses is not easy and is very unwieldly. However, field tests to obtain these properties are time-consuming, difficult to perform and expensive. Therefore, correlation between the RMR and Q with mechanical properties of rock masses have been made. The mechanical properties include not only the deformation modulus (E m ) and UCS (σ cm ) of rock masses but also Poisson's ratio (ν m ), frictional angle ( The determination of rock mass mechanical properties is one of the essential parts of rock neering design. To consider the scale effect for rock masses, field tests are desired, as laboratory g on rock masses is not easy and is very unwieldly. However, field tests to obtain these erties are time-consuming, difficult to perform and expensive. Therefore, correlation between RMR and Q with mechanical properties of rock masses have been made. The mechanical erties include not only the deformation modulus (Em) and UCS (σcm) of rock masses but also son's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile strength , among the others, which are critical for tunnel design. Correlations between empirical rockclassification systems and mechanical rock-mass properties are a common practice observed in engineering, as they create an easy approach for assessing the rock-mass conditions.

eformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to blast damage, experiment od and procedure and therefore, a good rock mass characterization may give equivalent, or ibly better Em values [83]. In the literature, several correlations have been suggested for the Em the rock-mass quality (RMR and Q) values since the first empirical equation between RMR and y Bienwiaski [84] and the first empirical equation between Q and Em by Barton [85]. In these tions, the intact rock properties were missing. Later, it was determined that, for the calculation ck-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (Ei, σc, νi, фi, d σt respectively) and rock-mass classification indexes must be involved [86]. The proposed lations between the deformation modulus of rock mass and the RMR or Q values can be divided two groups: The Em is calculated independently from the Ei, Table 4. The Em is calculated dependently from the Ei, Table 5.
Hoek-Brown constant and tensile strength (σ tm ), among the others, which are critical for tunnel design. Correlations between empirical rock-mass classification systems and mechanical rock-mass properties are a common practice observed in rock engineering, as they create an easy approach for assessing the rock-mass conditions.

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ E m calculations due to blast damage, experiment method and procedure and therefore, a good rock mass characterization may give equivalent, or possibly better E m values [83]. In the literature, several correlations have been suggested for the E m and the rock-mass quality (RMR and Q) values since the first empirical equation between RMR and E m by Bienwiaski [84] and the first empirical equation between Q and E m by Barton [85]. In these equations, the intact rock properties were missing. Later, it was determined that, for the calculation of rock-mass mechanical properties (E m , σ cm , ν m , based. As in the RMR system, bolt spacing is a function of Q value and thickness of shotcrete is function of the De along with the Q value.

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the essential parts of rock engineering design. To consider the scale effect for rock masses, field tests are desired, as laboratory testing on rock masses is not easy and is very unwieldly. However, field tests to obtain these properties are time-consuming, difficult to perform and expensive. Therefore, correlation between the RMR and Q with mechanical properties of rock masses have been made. The mechanical properties include not only the deformation modulus (Em) and UCS (σcm) of rock masses but also Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile strength (σtm), among the others, which are critical for tunnel design. Correlations between empirical rockmass classification systems and mechanical rock-mass properties are a common practice observed in rock engineering, as they create an easy approach for assessing the rock-mass conditions.

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to blast damage, experiment method and procedure and therefore, a good rock mass characterization may give equivalent, or possibly better Em values [83]. In the literature, several correlations have been suggested for the Em and the rock-mass quality (RMR and Q) values since the first empirical equation between RMR and Em by Bienwiaski [84] and the first empirical equation between Q and Em by Barton [85]. In these equations, the intact rock properties were missing. Later, it was determined that, for the calculation of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (Ei, σc, νi, фi, ci and σt respectively) and rock-mass classification indexes must be involved [86]. The proposed correlations between the deformation modulus of rock mass and the RMR or Q values can be divided into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5.

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the ess engineering design. To consider the scale effect for rock masses, field tests are d testing on rock masses is not easy and is very unwieldly. However, field t properties are time-consuming, difficult to perform and expensive. Therefore, the RMR and Q with mechanical properties of rock masses have been ma properties include not only the deformation modulus (Em) and UCS (σcm) of r Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant (σtm), among the others, which are critical for tunnel design. Correlations betw mass classification systems and mechanical rock-mass properties are a common rock engineering, as they create an easy approach for assessing the rock-mass co

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to blast d method and procedure and therefore, a good rock mass characterization may possibly better Em values [83]. In the literature, several correlations have been s and the rock-mass quality (RMR and Q) values since the first empirical equation Em by Bienwiaski [84] and the first empirical equation between Q and Em by B equations, the intact rock properties were missing. Later, it was determined tha of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock pr ci and σt respectively) and rock-mass classification indexes must be involved correlations between the deformation modulus of rock mass and the RMR or Q v into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5.
i , c i and σ t respectively) and rock-mass classification indexes must be involved [86]. The proposed correlations between the deformation modulus of rock mass and the RMR or Q values can be divided into two groups: • The E m is calculated independently from the E i , Table 4.

•
The E m is calculated dependently from the E i , Table 5.

Strength of Rock Masses
Construction materials used in mining and civil construction are categorized according to their strength properties. This basic quality information of the material is used in engineering and design for various construction purposes. Although rock is considered a construction material, in rock engineering, no such precise strength characterization of the rock mass is applied. Although the various utilizations of rocks and rock masses have different purposes and are subjected to various problems, the suitability and quality of the rock mass depends largely on its strength properties. Reliable determination of rock-mass strength is crucial for underground excavation [11,57,99,100]. The development [12,101] and update of the GSI (geological strength index) system [102] and its incorporation in the Hoek-Brown empirical failure criterion [103] is mainly for the reliable calculation of σ cm . Like the empirical failure criteria, the use of empirical classification systems for the calculation of σ cm is a common practice in rock engineering through correlation. The proposed correlations between the σ cm and the RMR or Q values can also be divided into two groups:

•
The σ cm calculated independently of the σ c , Table 6.

•
The σ cm calculated dependently of σ c , Table 7. Table 6. Equations for calculation of σ cm without σ c .

S. No. Equation Equation
No. Reference Where γ is the unit weight in t/m 3 .

Poisson's Ratio Value
Compared to the other basic mechanical properties of the rock mass, Poisson's ratio is an elastic constant whose importance is generally underrated [110]. The behavior of rock masses is influenced by the mechanical behavior and properties of discontinuities and of the intact rock bounded by discontinuities. The UCS test on a rock mass (σ cm ) indicated that ν m has an inverse relation with the σ cm [111], as expressed in Equation (42). As RMR and Q systems are used for calculation of σ cm , these systems can be used indirectly for the calculation of ν m .
A direct method of determining ν m from the RMR value was proposed [93,112] and is expressed in Equation (43).
This equation was modified and according to this modification, the ratio of ν m and ν i is a function of the corresponding RMR value [93], as shown in Equation (44).

Mohr-Coulomb Parameters
Mohr-Coulomb model is an elastoplastic model based on the Mohr-Coulomb failure criterion and is the most common model in the context of geomaterials. When using this model, it is essential to estimate the c m and the based. As in the RMR system, bolt spacing is a function of Q value and thickness of shotcrete is function of the De along with the Q value.

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the essential parts of rock engineering design. To consider the scale effect for rock masses, field tests are desired, as laboratory testing on rock masses is not easy and is very unwieldly. However, field tests to obtain these properties are time-consuming, difficult to perform and expensive. Therefore, correlation between the RMR and Q with mechanical properties of rock masses have been made. The mechanical properties include not only the deformation modulus (Em) and UCS (σcm) of rock masses but also Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile strength (σtm), among the others, which are critical for tunnel design. Correlations between empirical rockmass classification systems and mechanical rock-mass properties are a common practice observed in rock engineering, as they create an easy approach for assessing the rock-mass conditions.

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to blast damage, experiment method and procedure and therefore, a good rock mass characterization may give equivalent, or possibly better Em values [83]. In the literature, several correlations have been suggested for the Em and the rock-mass quality (RMR and Q) values since the first empirical equation between RMR and Em by Bienwiaski [84] and the first empirical equation between Q and Em by Barton [85]. In these equations, the intact rock properties were missing. Later, it was determined that, for the calculation of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (Ei, σc, νi, фi, ci and σt respectively) and rock-mass classification indexes must be involved [86]. The proposed correlations between the deformation modulus of rock mass and the RMR or Q values can be divided into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5. R system, bolt spacing is a function of Q value and thickness of shotcrete is ng with the Q value.

termination of Rock-Mass Properties
ion of rock mass mechanical properties is one of the essential parts of rock To consider the scale effect for rock masses, field tests are desired, as laboratory ses is not easy and is very unwieldly. However, field tests to obtain these onsuming, difficult to perform and expensive. Therefore, correlation between th mechanical properties of rock masses have been made. The mechanical ot only the deformation modulus (Em) and UCS (σcm) of rock masses but also frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile strength ers, which are critical for tunnel design. Correlations between empirical rockstems and mechanical rock-mass properties are a common practice observed in they create an easy approach for assessing the rock-mass conditions.

ulus of Rock Masses
es are associated to in-situ Em calculations due to blast damage, experiment re and therefore, a good rock mass characterization may give equivalent, or lues [83]. In the literature, several correlations have been suggested for the Em ality (RMR and Q) values since the first empirical equation between RMR and 4] and the first empirical equation between Q and Em by Barton [85]. In these rock properties were missing. Later, it was determined that, for the calculation ical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (Ei, σc, νi, фi, ) and rock-mass classification indexes must be involved [86]. The proposed the deformation modulus of rock mass and the RMR or Q values can be divided ted independently from the Ei, Table 4. ted dependently from the Ei, Table 5.  [69] m in a discrete manner [44]. This discrete approach is used in literature for the calculation of the cohesion and internal friction angle [78,113,114]. The correlation between the σ cm and the Appl. Sci. 2018, 8, x FOR PEER REVIEW based. As in the RMR system, bolt spacing is a function of Q function of the De along with the Q value.

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties i engineering design. To consider the scale effect for rock masses, testing on rock masses is not easy and is very unwieldly. H properties are time-consuming, difficult to perform and expen the RMR and Q with mechanical properties of rock masses properties include not only the deformation modulus (Em) and Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-B (σtm), among the others, which are critical for tunnel design. C mass classification systems and mechanical rock-mass propertie rock engineering, as they create an easy approach for assessing

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculation method and procedure and therefore, a good rock mass chara possibly better Em values [83]. In the literature, several correlat and the rock-mass quality (RMR and Q) values since the first em Em by Bienwiaski [84] and the first empirical equation between equations, the intact rock properties were missing. Later, it was of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), t ci and σt respectively) and rock-mass classification indexes m correlations between the deformation modulus of rock mass and into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5.

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the essential parts of roc engineering design. To consider the scale effect for rock masses, field tests are desired, as laborator testing on rock masses is not easy and is very unwieldly. However, field tests to obtain thes properties are time-consuming, difficult to perform and expensive. Therefore, correlation betwee the RMR and Q with mechanical properties of rock masses have been made. The mechanica properties include not only the deformation modulus (Em) and UCS (σcm) of rock masses but als Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile strengt (σtm), among the others, which are critical for tunnel design. Correlations between empirical rock mass classification systems and mechanical rock-mass properties are a common practice observed i rock engineering, as they create an easy approach for assessing the rock-mass conditions.

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to blast damage, experimen method and procedure and therefore, a good rock mass characterization may give equivalent, o possibly better Em values [83]. In the literature, several correlations have been suggested for the E and the rock-mass quality (RMR and Q) values since the first empirical equation between RMR an Em by Bienwiaski [84] and the first empirical equation between Q and Em by Barton [85]. In thes equations, the intact rock properties were missing. Later, it was determined that, for the calculatio of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (Ei, σc, νi, ф ci and σt respectively) and rock-mass classification indexes must be involved [86]. The propose correlations between the deformation modulus of rock mass and the RMR or Q values can be divide into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5. based. As in the RMR system, bolt spacing is a function of Q value and thickness of shotcrete is function of the De along with the Q value.

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the essential parts of rock engineering design. To consider the scale effect for rock masses, field tests are desired, as laboratory testing on rock masses is not easy and is very unwieldly. However, field tests to obtain these properties are time-consuming, difficult to perform and expensive. Therefore, correlation between the RMR and Q with mechanical properties of rock masses have been made. The mechanical properties include not only the deformation modulus (Em) and UCS (σcm) of rock masses but also Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile strength (σtm), among the others, which are critical for tunnel design. Correlations between empirical rockmass classification systems and mechanical rock-mass properties are a common practice observed in rock engineering, as they create an easy approach for assessing the rock-mass conditions.

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to blast damage, experiment method and procedure and therefore, a good rock mass characterization may give equivalent, or possibly better Em values [83]. In the literature, several correlations have been suggested for the Em and the rock-mass quality (RMR and Q) values since the first empirical equation between RMR and Em by Bienwiaski [84] and the first empirical equation between Q and Em by Barton [85]. In these equations, the intact rock properties were missing. Later, it was determined that, for the calculation of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (Ei, σc, νi, фi, ci and σt respectively) and rock-mass classification indexes must be involved [86]. The proposed correlations between the deformation modulus of rock mass and the RMR or Q values can be divided into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5. m using Equation (45).
Later, a direct relation between the rock-mass internal friction angle and the RMR value was developed [115], as shown in Equation (46), taking the intact rock friction angle ( Appl. Sci. 2018, 8, x FOR PEER REVIEW based. As in the RMR system, bolt spacing is a function of Q value and t function of the De along with the Q value.

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the engineering design. To consider the scale effect for rock masses, field tests ar testing on rock masses is not easy and is very unwieldly. However, fiel properties are time-consuming, difficult to perform and expensive. Therefo the RMR and Q with mechanical properties of rock masses have been properties include not only the deformation modulus (Em) and UCS (σcm) o Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown consta (σtm), among the others, which are critical for tunnel design. Correlations b mass classification systems and mechanical rock-mass properties are a comm rock engineering, as they create an easy approach for assessing the rock-mas

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to bla method and procedure and therefore, a good rock mass characterization m possibly better Em values [83]. In the literature, several correlations have be and the rock-mass quality (RMR and Q) values since the first empirical equa Em by Bienwiaski [84] and the first empirical equation between Q and Em b equations, the intact rock properties were missing. Later, it was determined of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock ci and σt respectively) and rock-mass classification indexes must be involv correlations between the deformation modulus of rock mass and the RMR or into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5.
β is a constant and its value varies between 0.1 and 3. The revised value for β is 1.0 and Equation (46) is simplified to Equation (47) [112].

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the essential parts of rock engineering design. To consider the scale effect for rock masses, field tests are desired, as laboratory testing on rock masses is not easy and is very unwieldly. However, field tests to obtain these properties are time-consuming, difficult to perform and expensive. Therefore, correlation between the RMR and Q with mechanical properties of rock masses have been made. The mechanical properties include not only the deformation modulus (Em) and UCS (σcm) of rock masses but also Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile strength (σtm), among the others, which are critical for tunnel design. Correlations between empirical rockmass classification systems and mechanical rock-mass properties are a common practice observed in rock engineering, as they create an easy approach for assessing the rock-mass conditions.

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to blast damage, experiment method and procedure and therefore, a good rock mass characterization may give equivalent, or possibly better Em values [83]. In the literature, several correlations have been suggested for the Em and the rock-mass quality (RMR and Q) values since the first empirical equation between RMR and Em by Bienwiaski [84] and the first empirical equation between Q and Em by Barton [85]. In these equations, the intact rock properties were missing. Later, it was determined that, for the calculation of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (Ei, σc, νi, фi, ci and σt respectively) and rock-mass classification indexes must be involved [86]. The proposed correlations between the deformation modulus of rock mass and the RMR or Q values can be divided into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5.  [116] as cited by [117].

Application in Determination of Rock-Mass Properties
The determination of rock mass mechanical properties is one of the essential part engineering design. To consider the scale effect for rock masses, field tests are desired, as la testing on rock masses is not easy and is very unwieldly. However, field tests to obt properties are time-consuming, difficult to perform and expensive. Therefore, correlation the RMR and Q with mechanical properties of rock masses have been made. The m properties include not only the deformation modulus (Em) and UCS (σcm) of rock masses Poisson's ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile (σtm), among the others, which are critical for tunnel design. Correlations between empir mass classification systems and mechanical rock-mass properties are a common practice ob rock engineering, as they create an easy approach for assessing the rock-mass conditions.

Deformation Modulus of Rock Masses
The uncertainties are associated to in-situ Em calculations due to blast damage, ex method and procedure and therefore, a good rock mass characterization may give equiv possibly better Em values [83]. In the literature, several correlations have been suggested f and the rock-mass quality (RMR and Q) values since the first empirical equation between Em by Bienwiaski [84] and the first empirical equation between Q and Em by Barton [85] equations, the intact rock properties were missing. Later, it was determined that, for the ca of rock-mass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (E ci and σt respectively) and rock-mass classification indexes must be involved [86]. The correlations between the deformation modulus of rock mass and the RMR or Q values can b into two groups:  The Em is calculated independently from the Ei, Table 4.  The Em is calculated dependently from the Ei, Table 5.  (48) indirectly uses the empirical classification systems for the rock-mass cohesion (c m ). Later, a direct relation between c m and the RMR value was developed and, according to this relation, the c m is a function of intact rock cohesion (c i ) and RMR [112], as expressed in Equation (49).
Sen and Sadagah also suggested a continuous system for the calculation of cohesion and the frictional angle of rock masses [18] using Equations (50) and (51).
Barton split the modified Q system from Equation (9) into two components and designated them the c m and ation in Determination of Rock-Mass Properties determination of rock mass mechanical properties is one of the essential parts of rock ng design. To consider the scale effect for rock masses, field tests are desired, as laboratory rock masses is not easy and is very unwieldly. However, field tests to obtain these s are time-consuming, difficult to perform and expensive. Therefore, correlation between and Q with mechanical properties of rock masses have been made. The mechanical s include not only the deformation modulus (Em) and UCS (σcm) of rock masses but also ratio (νm), frictional angle (фm), cohesion (cm), Hoek-Brown constant and tensile strength ng the others, which are critical for tunnel design. Correlations between empirical rocksification systems and mechanical rock-mass properties are a common practice observed in neering, as they create an easy approach for assessing the rock-mass conditions.

ation Modulus of Rock Masses
uncertainties are associated to in-situ Em calculations due to blast damage, experiment nd procedure and therefore, a good rock mass characterization may give equivalent, or etter Em values [83]. In the literature, several correlations have been suggested for the Em ck-mass quality (RMR and Q) values since the first empirical equation between RMR and nwiaski [84] and the first empirical equation between Q and Em by Barton [85]. In these , the intact rock properties were missing. Later, it was determined that, for the calculation ass mechanical properties (Em, σcm, νm, фm, cm and σtm), the intact rock properties (Ei, σc, νi, фi, respectively) and rock-mass classification indexes must be involved [86]. The proposed ns between the deformation modulus of rock mass and the RMR or Q values can be divided roups: m is calculated independently from the Ei, Table 4. m is calculated dependently from the Ei, Table 5.  [24], as shown in Equations (52) and (53), respectively.

Hoek-Brown Constants and Tensile Strength of Rock Mass
For the subsurface excavations design in hard rock, Hoek and Brown introduced their empirical failure criterion, known Hoek-Brown failure criterion [99,118]. The criterion for rock masses is accepted widely in a huge number of excavation projects worldwide [103]. The development history of this empirical failure criterion [119] shows that RMR served this criteria before replacing it with the GSI system [12] for the calculation of the rock-mass Hoek-Brown constants m and s. The original Hoek-Brown failure criterion is shown in Equation (54) for intact rock and the generalized Hoek-Brown failure criterion for rock mass is expressed in Equation (55) [103].
where m b in the above equations is a reduced value for the rock mass of the material constant m i and s and a are constants for the rock mass. The rock-mass constants m b and s were expressed in terms of the basic RMR before the introduction of GSI in the Hoek-Brown criterion [22,120], as shown in Equations (56)- (59). Disturbed rock masses: Undisturbed or interlocking rock masses: These rock-mass constants were also expressed in Q' (Q without SRF and J w ) value, as expressed in Equations (60) and (61) [104].
This equation indirectly yields the σ tm from the RMR and Q values, as m b and s are functions of these systems.
The direct method for the calculation of the σ tm from the RMR value is given in Equation (63) [121].

Stand-Up Time
In subsurface openings, time-dependent loss of the rock mass is normal, causing either accelerated damage or slow damage [122]. A practical consequence of time dependency in tunneling is the amount of time the excavated rock mass can remain unsupported before instabilities occur, resulting in failure and collapse. This period of time to collapse is known as the stand-up time. The stand-up time was established as an empirical classification system [13]. According to this classification, the rock mass was classified into seven classes and a relation between the rock-mass classes, stand-up time and active span was established. The stand-up time classification was one of the three classification systems that were considered during the development of the RMR system and a chart was developed to correlate the predicted stand-up time for the unsupported span of the tunnel and the RMR score [8]. Along with the update of the RMR system for tunneling, the stand-up time chart was also updated for drill and blast tunnel [22] and was also revised for TBM tunnel [123].
In the Q system, instead of providing the stand-up time, the system suggests the span of the unsupported tunnel based on the tunnel-quality index value, that is, Q. The unsupported span can be calculated using Equation (64) [9].
The general condition requirement for the unsupported span was expressed for different ESR values [85]. Based on the correlation between the RMR and Q systems, the stand-up time for an unsupported span was extended for the Q system [53], as shown in Figure 1.

Stand-Up Time
In subsurface openings, time-dependent loss of the rock mass is normal, causing either accelerated damage or slow damage [122]. A practical consequence of time dependency in tunneling is the amount of time the excavated rock mass can remain unsupported before instabilities occur, resulting in failure and collapse. This period of time to collapse is known as the stand-up time. The stand-up time was established as an empirical classification system [13]. According to this classification, the rock mass was classified into seven classes and a relation between the rock-mass classes, stand-up time and active span was established. The stand-up time classification was one of the three classification systems that were considered during the development of the RMR system and a chart was developed to correlate the predicted stand-up time for the unsupported span of the tunnel and the RMR score [8]. Along with the update of the RMR system for tunneling, the stand-up time chart was also updated for drill and blast tunnel [22] and was also revised for TBM tunnel [123].
In the Q system, instead of providing the stand-up time, the system suggests the span of the unsupported tunnel based on the tunnel-quality index value, that is, Q. The unsupported span can be calculated using Equation (64) [9].
The general condition requirement for the unsupported span was expressed for different ESR values [85]. Based on the correlation between the RMR and Q systems, the stand-up time for an unsupported span was extended for the Q system [53], as shown in Figure 1.

Rock Load
In underground excavations, generally, the load on installed support is known as rock load. It indicates the rock pressure that results from the rock-load height above the underground excavation. Reliable forecast of rock load is one of the toughest jobs in rock engineering. Terzagi was the first whose rock-mass classification system,-that is, rock-load classification-was used to estimate the rock load to be carried by the steel arches installed to support a tunnel [5], as shown in Figure 2.

Rock Load
In underground excavations, generally, the load on installed support is known as rock load. It indicates the rock pressure that results from the rock-load height above the underground excavation. Reliable forecast of rock load is one of the toughest jobs in rock engineering. Terzagi was the first whose rock-mass classification system,-that is, rock-load classification-was used to estimate the rock load to be carried by the steel arches installed to support a tunnel [5], as shown in Figure 2. Since then, different methods, including empirical, analytical, numerical and experimental approaches, were used for prediction of the rock load [9,[124][125][126]. According to Barton [9] and supported by Verman [88], the rock load is free of the tunnel size in rock and Barton expressed this using Equation (65).
The correlation of Equation (65) has proven useful, except in cases of squeezing-ground conditions. For squeezing ground, three adjustment factors were applied to Barton's original equation, including the tunnel depth factor, tunnel radius factor and correction factor for time [127,128]. The application of the RMR system was introduced for the support load by Unal [124] based on the old version of RMR and extended for RMR89 using Equation (66). He defined the rock load as unstable zone height above the excavation crown, which tends to collapse ultimately.
where p is the support load (kN), B is the tunnel width (m) and γ is the rock density (kg/m 3 ).
Since the above equation is based on a coal mine roadway, the evaluation of this equation for rock tunnel applications shows that the estimated support pressure is unsafe in squeezing conditions [128]. Furthermore, the estimates for the non-squeezing condition are unsafe in the case of a small Since then, different methods, including empirical, analytical, numerical and experimental approaches, were used for prediction of the rock load [9,[124][125][126]. According to Barton [9] and supported by Verman [88], the rock load is free of the tunnel size in rock and Barton expressed this using Equation (65).
The correlation of Equation (65) has proven useful, except in cases of squeezing-ground conditions. For squeezing ground, three adjustment factors were applied to Barton's original equation, including the tunnel depth factor, tunnel radius factor and correction factor for time [127,128]. The application of the RMR system was introduced for the support load by Unal [124] based on the old version of RMR and extended for RMR 89 using Equation (66). He defined the rock load as unstable zone height above the excavation crown, which tends to collapse ultimately.
where p is the support load (kN), B is the tunnel width (m) and γ is the rock density (kg/m 3 ).
Since the above equation is based on a coal mine roadway, the evaluation of this equation for rock tunnel applications shows that the estimated support pressure is unsafe in squeezing conditions [128]. Furthermore, the estimates for the non-squeezing condition are unsafe in the case of a small tunnel and are overestimated for a large tunnel. The following correlations (Equations (67) and (68)) have been proposed for a rock tunnel, taking the tunnel depth and tunnel radius as two additional parameters with RMR for non-squeezing and squeezing conditions [128].
For non-squeezing ground: For squeezing ground: where P is the ultimate support pressure, f (RMR 89 ) is the correction factor for the tunnel closure, H is the tunnel depth (m) and a is the radius of the tunnel (m). Due to the widespread acceptance of GSI in rock engineering, Equation (66) is considered the core origin for the Equation (69), which is based on GSI for the calculation of support pressure [129].
GSI defines the quality of rock-mass, D is the disturbance factor, σ cr is the residual compressive strength of the rock mass in the broken zone around the tunnel, D e is the equivalent diameter of the excavation, γ is the unit weight of rock mass, C s is the correction factor for the horizontal-to-vertical field stress ratio and S q is the correction factor for the squeezing-ground condition.

Recommended Procedure for Using RMR and Q Systems and their Correlation
After RMR and Q systems had been used in the field of tunneling worldwide, Nick Barton and Z. T. Bieniawski provided the "ten commandments" for the proper use of RMR and Q systems [53] as follows: • Use standard procedure of measurement for the parameters instead of only description. Include the characterization information of rock-mass along with the design procedure, specifications and assumptions in the Geotechnical Baseline Report.

•
In the construction phase, calculate RMR and Q for design verification and modification.
A back-analysis methodology for RMR and Q calculation, as shown in Figure 3, was proposed using the installed support and used to evaluate and extend the application of these classification in the field of tunneling [70]. The variation of rock-mass quality in underground excavation is always significant. However, the forecast of RMR ahead of the excavation face in tunneling, up to certain distance, is reliable using the rock-mass quality of the current working face of the tunnel [130]. It is worth emphasizing that use at least two empirical classification systems in the tunnel-support design process [131]. When using both systems, they should be checked through correlation [53]. After the development of RMR and Q systems, a linear correlation based on the regression analysis of RMR and Q values was presented by Bienwiaski [75], as shown in Equation (70), where A = 9 and B = 44. Figure 3. Back-calculation technique for RMR (left) and Q system (right) using tunnel span along with rock-bolt spacing and shotcrete thickness [58].
Since then, several researchers have presented other correlations but there is no scientific basis to assume a universally valid regression between the two systems. The limitations in correlation is due to the differences between the rock-mass characterizations and rating rules of RMR and Q system [132]. Due to this constraint of correlation, some scholars propose local, project-based, or geologicalbased correlation between the two systems [132][133][134][135][136]. Another suggested approach for the correlation is between the truncated RMR and Q values [54,128]. The latest version of RMR, that is, RMR14, is not as popular as RMR89; however, an excellent correlation between the two versions exists [23,58,137], as shown in Figure 4.

Limitations of RMR and Q Systems
Rock-mass classification systems are frequently used in rock engineering and design and have gained wide attention in this field; however, these classification systems have limitations. Nevertheless, if applied appropriately, they are valuable design tools [138]. One of the many reasons for updating the two systems is to point out their limitations. These updates kept the two systems  Since then, several researchers have presented other correlations but there is no scientific basis to assume a universally valid regression between the two systems. The limitations in correlation is due to the differences between the rock-mass characterizations and rating rules of RMR and Q system [132]. Due to this constraint of correlation, some scholars propose local, project-based, or geological-based correlation between the two systems [132][133][134][135][136]. Another suggested approach for the correlation is between the truncated RMR and Q values [54,128]. The latest version of RMR, that is, RMR 14 , is not as popular as RMR 89 ; however, an excellent correlation between the two versions exists [23,58,137], as shown in Figure 4. Since then, several researchers have presented other correlations but there is no scientific basis to assume a universally valid regression between the two systems. The limitations in correlation is due to the differences between the rock-mass characterizations and rating rules of RMR and Q system [132]. Due to this constraint of correlation, some scholars propose local, project-based, or geologicalbased correlation between the two systems [132][133][134][135][136]. Another suggested approach for the correlation is between the truncated RMR and Q values [54,128]. The latest version of RMR, that is, RMR14, is not as popular as RMR89; however, an excellent correlation between the two versions exists [23,58,137], as shown in Figure 4.

Limitations of RMR and Q Systems
Rock-mass classification systems are frequently used in rock engineering and design and have gained wide attention in this field; however, these classification systems have limitations. Nevertheless, if applied appropriately, they are valuable design tools [138]. One of the many reasons for updating the two systems is to point out their limitations. These updates kept the two systems

Limitations of RMR and Q Systems
Rock-mass classification systems are frequently used in rock engineering and design and have gained wide attention in this field; however, these classification systems have limitations. Nevertheless, if applied appropriately, they are valuable design tools [138]. One of the many reasons for updating the two systems is to point out their limitations. These updates kept the two systems viable to date and also expanded their applications in rock engineering. Some limitations found were inherent to the two systems, which have been proven from their use over time, whereas the others were overcome with improvements. The inherent limitations of these systems formed basis for new classification systems. Stille and Palmstorm pointed out six necessities for a true rock mass classification system and, according to them, none of these systems fulfill these requirements [20]. However, Hand distinguished between the two main types of classification, that is, unsupervised and supervised classification [139]. The RMR and Q systems belong to the supervised classification because the rock-mass class structure is logical and the principles of separation are expressed so as to allow one to distribute objects to their suitable class [20].
Many empirical systems in rock engineering have been suggested with the respect of a specific rock-mass structure and/or particular purposes. Rock mass related professionals incline to the empirical method over the theoretical and numerical methods due its simplicity [128]. However, these empirical approaches are unable to characterize complex rock-masses. This is probably one of the causes why rock engineers have continued to develop new systems or modify and extend current ones [7]. Each classification has its own capabilities and therefore they are used in parallel as much as possible. The limitations can be broadly divided into three types, as described in the following sections.

Limitations Related to RMR System
The key benefit of the RMR classification is that it is easy to practice. However, this system is comparatively insensitive to slight changes in RMR value. Taking ground behavior in consideration, adequate support measure be selected with clear understanding in tunnel design [1]. Based on ground behavior and the available tools for underground excavation in rock engineering, the application of RMR systems are the lowest among the classification systems [56], as shown in Table 8. This system is based on a comparatively small case record, which makes the RMR system less appropriate as an empirical design technique for rock-mass support, particularly when in-situ stresses or time-dependent rock mass properties are of importance for the rock engineering design [20]. In the RMR system, the rock-mass quality is divided into five classes but these classes are scarcely comparable due to the wide variation of rock-mass quality within a single class [140]. As the classification and design systems develop, the old forms of rock-mass classification systems are not always well-matched with new design approaches. Presently, usual mistakes of RMR system applications primarily deal with the use of old versions of the system [141], for example, the use of RMR 76 for hard rock-stability analysis for a span in entry-type excavation [142]. Although this system is improved tremendously in terms of characterization and support, it is noticeable that the application of RMR 89 , RMR 14 systems for tunnel-support design in high-stress locations is still a limitation of the system [58]. Although the positive impact of mechanical excavation is incorporated in the latest version of RMR, that is, RMR 14 [23], the mechanical excavation impact is negative in high-stress environments and drill and blast excavation has a positive impact against rock bursting [143,144].

Limitations Related to Q System
Empirical design methods have been accepted for a long time, owing to the lack of appropriate design methods and undeveloped theory, which show quite adverse results during construction [145]. An important benefit of the Q system is that this system is based on well-documented case records for its original development. However, compared to RMR, this system is not covering new fields in its use. Therefore, the support chart of Q system gives only a suggestion of the support to be applied and this support system should be tempered by practical and sound engineering judgement [138]. In this system, RQD/J n is an indication of relative block size, however, this is not understandable [146], which was supported by Palmstorm [60]. The effect of water in the Q system was discussed by Palmstorm through example, which shows that in exceptionally high-flow cases, the Q system is unable to recommend suitable support [138].
The application of SRF is unclear for buckling, rock-burst, and/or squeezing conditions, or for weakness zones [56]. The SRF values were adjusted in 1993, in overstressed massive brittle rocks based on limited data [68]. This increase is from a value of 20 to 400 based on the strength-stress ratio but no criteria have been selected for the rating selection of SRF even though SRF was a fine tuning parameter [9]. In case of squeezing rock, Barton selected the range of SRF from 5 to 20 [9]; however, this SRF range does not effectively characterize the squeezing influence due to the deficiency of enough case records and resulting in unsafe tunnel practice when using the Q system for large underground excavations in squeezing ground [128]. The true nature of the rock-mass is not explicitly reflected in the Q system, although it is crucial for the prediction of the support measures (e.g., popping ground, squeezing, or swelling) [20]. According to Loset [147], the Q system is not applicable in weak rock and weak zones, which was also confirmed by comparison of Q-system support with that of NATM (New Austrian Tunneling Method) support in swelling rock [148]. Bhawani Singh and his co-researchers [127] have presented an empirical approach (Equation (71)) for critical overburden calculation when squeezing may happen.
In the calculation of H expressed above, the method of calculating Q is uncertain. This was also pointed out by Palmstorm [138].
In the Q-system support chart, the determination of support is based on Q value and D e . The D e is the function tunnel span and ESR. The ESR value should be based on the safety of the working team according to the country's safety criteria because each country has its own criteria [138]. Application of the Q system is more prevalent in jointed rock without overstressing [138] and displays optimal performance for a tunnel with D e between 2.5 and 30 and Q values between 0.1 and 40 [149].

Limitations Related to the Two Systems
The different excavation (size, shape and purpose) and corresponding rock-supporting practices and procedures in different countries, as well as the permanent support requirements, result in approaches and amounts of support that fluctuate widely from tunnel to tunnel. The users of empirical support charts ignore on which averaged and imprecise basis the rock-support given in the support chart is based. This feature is common for quantitative rock mass classification systems used for tunnel support. It is the opinion of the experts in tunneling that rock-mass classifications as a designing tool on their own should be used for preliminary design of underground openings, that is, for planning purposes and not for final tunnel support [99,150,151].
The two systems have duplications of some parameters, such as RQD, joint number and their spacing, which doubles the effect of the joints on the final score. Furthermore, the influence of water on clay-bearing rocks plays a significant role in reducing rock-mass strength, however, this influence is not effectively reflected in these classification systems [7]. RQD is used in both systems as a mandatory parameter but their correlation with other jointing measurement parameters is very difficult [60].
The difficulty in correlation is that RQD is a one-dimensional measurement based only on core pieces longer than 0.1 m, which means that the application of RQD in rock engineering may lead to inaccuracy. The incorporation of this parameter within the two systems was a matter of historical development and its use in these classification system is no longer essential [61]. In rock-mass classification systems, parameters and ratings are applied irrespective of the project, specific characteristics and failure mechanisms [151].
Three of the basic limitations in using these systems are the risk of over-simplification by summarizing a complex rock-mass with a single digit, the scale effect and accounting for anisotropy and heterogeneity in tunnel and cavern designs [151][152][153]. The same rock mass quality value (RMR or Q) can be attained by a number of groupings of parameters score, even though the behavior of rock-mass could be dissimilar [151]. In rock-mass classification systems, there is no way of assessing the safety margin of a certain support under certain conditions, as the classification was assigned based on a rule and not on the engineering analysis [153]. These systems do not allow the user to quantify the degree of safety achieved by the design [56]. According to Pell and Bertuzzi [154], rock-mass classification systems include a significant degree of interpretation, as they are based on a particular structure at a particular depth.
One concern with empirical systems is that the true class of the rock mass is unidentified, which results in additional uncertainties and therefore the risk of misclassification [20]. The hazards and associated consequences due to misclassification are not taken into consideration. Thus, they will have a limited understanding of the mechanical processes associated with tunneling [153]. When excavated, a rock mass is the location of complex phenomena that rely upon mechanics, physics, thermodynamics, chemistry and so forth. Therefore, although it is comparatively easy to quantify the rock matrix properties. However, the same is not true for the rock mass at the scale of our works and a single score of this rock mass can clearly not pretend to achieve a description of the large variety properties.
In rock engineering, with time, original rules and temporary clarifications become rooted as exercise and the original rules are hardly reconsidered or questioned [155]. These systems are misused continuously due to the limitations used and assumptions made in developing them have been ignored by users. For this reason, empirical classification systems contain a hidden critical risk [156]. The assigned parameters are neither independent nor directly connected with rock behavior [157]. Therefore, the users only consider the geological characterization of the rock mass followed by ground support recommendation without reflection of the probable ground behavior. It is essential to examine the available rock-mass behavior information whether they are addressed by the empirical systems or not. A number of potential modes of failure are shown in Table 8 which reveal that every failure mode are not covered by the empirical methods and they must be considered individually [151]. As these systems are based empirically for the design and no division is made between durability, serviceability and structural resistance, this is a severe deficiency, particularly when linked to the design codes [56]. In heterogeneous and poor ground conditions, these systems may deliver false results, whilst their other inadequacies comprise a deficiency of consideration for different rock-mass failure modes and for the ground-support interaction [151].

Conclusions
Classification systems of rock masses have been developed significantly and serve as transmitters of their characteristics. Specifically, with respect to tunneling, RMR and Q systems describe and divide the ground into different classes and are also part of the engineering design methodologies. These RMR and Q systems are numerical classifications and are established on jointed rock-mass case records. These have a wide-range of application in rock mechanics but are established for tunnel-support and for this particular purpose they are updated empirically. The key purpose behind their worldwide recognition is that these systems have been continuously updated both in the form of characterization and support. These systems on their own can be used for the preliminary design and planning of tunnel supports. They are unreliable during tunnel construction for rock-support, as local geometry of the underground excavation and geological features may dominate the rock-mass quality defined by these systems. Along with preliminary design of tunnel support, these systems are also good alternatives for the estimation of rock masses geomechanical properties (deformation modulus, strength, Mohr-Coulomb parameters and Hoek-Brown constant), stand-up time for the tunnel and support load. Along with the wide use of these systems in rock engineering in general and tunneling in particular, there are limitations of using RMR and Q systems that will be encountered when using them in the engineering applications. The classification system accuracy and the hazard resulting from misclassification must be weighed. It is concluded that both systems are suitable for jointed rock-mass where the ground behavior is led by rock fall. However, other types of behavior, like swelling, squeezing, raveling, or popping ground, are not effectively covered by these systems. These systems should be applied with great care for such conditions and, preferably, should be supported by other design tools. Rock-mass classifications were never suggested as the decisive solution to design problems but only a way towards this end. For a true classification, these empirical design methods need further improvements, as it is difficult to define their accuracy. There are a number of pitfalls in using these classification systems in the design and construction of tunnels, which must be considered by the users. Considering the limitations of the said systems and when used along with other suitable design tools, the RMR and Q systems are valuable tools for tunnel design and also serve as a kind of checklist.