Differences of Mechanical Parameters and Rockburst Tendency Indices between Coal and Non-Coal Rocks and Modiﬁed Rockburst Tendency Classiﬁcation Criteria for Non-Coal Rocks

: Rockbursts represent hazardous dynamic disasters for underground coal mines and other underground rock engineering projects. Some bursting liability indices are put forward and applied to identify the likelihood of rock burst occurrence. The classiﬁcation criteria of the bursting liability indices are proved to be reasonable for coals, but they are still immature for non-coal rocks. Thus, it is uncertain that it is reasonable to use the classiﬁcation criteria of coal for evaluating the bursting liability of non-coal rocks. Hence, in this study, a large amount of data, such as the basic mechanical parameters, i.e., Poisson’s ratio µ , elastic modulus E , uniaxial compressive strength σ c , and uniaxial tensile strength σ t , and the bursting liability indices, i.e., elastic strain energy index W ET , bursting energy index W cf , dynamic fracture duration time DT , and brittleness index B , of different coals and non-coal rocks were collected in China. Then, the differences of mechanical parameters and rockburst tendency indices between coal and non-coal rocks were studied systematically, and apart from the Poisson’s ratio µ , the other three basic mechanical parameters of coal and non-coal rocks have great differences in data distribution and concentration scope, which proved that the non-coal rocks cannot share the same index system and classiﬁcation criteria of coals. In addition, the evaluation results of a single index for rock bursting liability of rocks were directly compared in pairs, and the inconsistency rate for coals is about 42–68%. It is necessary to build a comprehensive evaluation method to evaluate the bursting liability of rocks. At last, the modiﬁed rockburst tendency classiﬁcation criteria for non-coal rocks were put forward. It is reasonable to use the classiﬁcation criteria of the W ET and W cf to classify the bursting liability of non-coal rocks, while it is unreasonable to use that of the DT and σ c . It has been concluded that the index B are more suitable for non-coal rocks, and a new index, named strength decrease rate (SDR), was proposed to determine the bursting liability, which is the ratio of uniaxial compressive strength σ c to duration of dynamic fracture DT .


Introduction
In recent years, a large amount of deep mineral resources has been mined in China [1,2]. With the increasing burial depth of the mineral resources, the geo-stress levels also increase [3], and a series of rockburst disasters have occurred, which are a serious threat to safe production in deep underground mines in China [4]. In fact, the rock burst is a complex rock failure process with two main influencing factors, which are the mechanical properties of rocks, and the geo-stress conditions and levels [5,6]. It is widely reported that Poisson's ratio µ, elastic modulus E, uniaxial compressive strength σ c , and uniaxial tensile strength σ t are the basic mechanical parameters of rocks [7,8]. The σ c and σ t jointly reflect tensile strength σt are the basic mechanical parameters of rocks [7,8]. The σc and σt jointly reflect the strength properties of rocks, and the ratio of the σc to σt is called as the brittleness index B of rocks [9][10][11]. The brittleness index B is often used to indicate a significant bursting liability index for non-coal rocks [12]. Poisson's ratio μ and elastic modulus E express the deformation capacity of rocks [13]. These basic mechanical parameters of rocks influence the occurrence of rockburst in underground engineering. Therefore, many bursting liability indices were put forward based on the lab tests of the basic mechanical parameters [14]. In conclusion, there are more than a dozen indices to evaluate the rock bursting liability of rocks [15]. Four indices, i.e., elastic strain energy index WET, bursting energy index Wcf, duration of dynamic fracture DT, uniaxial compressive strength σc, are the most used indices to evaluate the bursting liability of coals in China, as shown in Figure 1. The bursting liability classification criteria of coals are often divided into three bursting liability grades, i.e., heavy bursting liability, weak bursting liability, and no bursting liability. Neyman et al. firstly proposed the elastic strain energy storage index WET as a bursting liability evaluation index for coals in 1972 [16]. WET > 5, 2 < WET ≤ 5, and 0 < WET ≤ 2 are the classification criteria corresponding to heavy, weak, and no bursting liabilities of coals, which is proved to be reasonable for bursting liability prediction of coals by many testing data [17,18]. Sing [19] put forward a new classification criterion of WET for hard rocks in Canada, i.e., heavy bursting liability with WET > 15, weak bursting liability with 10 < WET ≤ 15, and no bursting liability with 0 < WET ≤ 10. In 1990, Hou et al. [20,21] wre the first to put forward the bursting energy index Wcf as another evaluation index of bursting liability, and defined that the values Wcf > 3.5, 1.5 < Wcf ≤ 3.5, and 0 < Wcf ≤ 1.5 are the classification criteria for heavy, weak, and no bursting liabilities, respectively. After the modified criterions of Wcf by Tan et al. [22] and MT [23], it is confirmed to be useful to predict the bursting liability of coals using the values of heavy bursting liability with Wcf >5, weak bursting liability with 1.5 < Wcf ≤ 5, and no bursting liability with 0 < Wcf ≤ 1.5. In the Classification and Laboratory Test Method on Bursting Liability of Coal (MT/T  [23], the fracture duration time DT in uniaxial compression tests was put forward, and DT > 500 ms, 50 < DT ≤ 500 ms, 0 < DT ≤ 50 ms were defined as the classification intervals for no, weak and heavy bursting liabilities of coals, respectively. Li.et al [15] defined DT > 2000 ms, 1000 < DT ≤ 2000 ms, 0 < DT ≤ 1000 ms as the classification criterions for no, weak, heavy bursting liabilities of non-coal rocks, respectively. In the Methods for Testing, Monitoring and Prevention of Rock Burst, Part 2: Classification and Laboratory The bursting liability classification criteria of coals are often divided into three bursting liability grades, i.e., heavy bursting liability, weak bursting liability, and no bursting liability. Neyman et al. firstly proposed the elastic strain energy storage index W ET as a bursting liability evaluation index for coals in 1972 [16]. W ET > 5, 2 < W ET ≤ 5, and 0 < W ET ≤ 2 are the classification criteria corresponding to heavy, weak, and no bursting liabilities of coals, which is proved to be reasonable for bursting liability prediction of coals by many testing data [17,18]. Sing [19] put forward a new classification criterion of W ET for hard rocks in Canada, i.e., heavy bursting liability with W ET > 15, weak bursting liability with 10 < W ET ≤ 15, and no bursting liability with 0 < W ET ≤ 10. In 1990, Hou et al. [20,21] wre the first to put forward the bursting energy index W cf as another evaluation index of bursting liability, and defined that the values W cf > 3.5, 1.5 < W cf ≤ 3.5, and 0 < W cf ≤ 1.5 are the classification criteria for heavy, weak, and no bursting liabilities, respectively. After the modified criterions of W cf by Tan et al. [22] and MT [23], it is confirmed to be useful to predict the bursting liability of coals using the values of heavy bursting liability with W cf >5, weak bursting liability with 1.5 < W cf ≤ 5, and no bursting liability with 0 < W cf ≤ 1.5. In the Classification and Laboratory Test Method on Bursting Liability of Coal (MT/T  [23], the fracture duration time DT in uniaxial compression tests was put forward, and DT > 500 ms, 50 < DT ≤ 500 ms, 0 < DT ≤ 50 ms were defined as the classification intervals for no, weak and heavy bursting liabilities of coals, respectively. Li.et al [15] defined DT > 2000 ms, 1000 < DT ≤ 2000 ms, 0 < DT ≤ 1000 ms as the classification criterions for no, weak, heavy bursting liabilities of non-coal rocks, respectively. In the Methods for Testing, Monitoring and Prevention of Rock Burst, Part 2: Classification and Laboratory Test Method on Bursting Liability of Coal (GB/T 25217.2-2010) [24] a new standard was put forward to evaluate the bursting liability of coals, which is the uniaxial compressive strength σ c , and σ c > 14 MPa, 7 < σ c ≤ 14 MPa, and 0 < σ c ≤ 7 MPa were the detailed classification criteria for no, weak and heavy bursting liabilities of coals. In 1996, Peng et al. [25] proposed the brittleness index B based on Griffith's strength theory to evaluate the bursting liabilities of non-coal rocks, and defined that the B > 26.7, 14.5 < B ≤ 26.7, and 0 < B ≤ 14.5 as the classification criterions for no, weak, and heavy bursting liabilities of non-coal rocks, respectively. In addition, the other two classification criterions were put forward by Li.et al [26] and Zhang et al. [27], which described "heavy bursting liability with B ≥ 18, weak bursting liability with 10 ≤ B < 18 and no bursting liability with 0 ≤ B < 10", and "heavy bursting liability with B > 22, weak bursting liability with 15 < B ≤ 22 and no bursting liability with 0 < B ≤ 15", respectively.
Nevertheless, rock burst disasters also frequently occur in deep underground non-coal mines and tunnels in China. Therefore, it is extremely vital to predict the bursting liability degree of deep non-coal rocks. The classification criterions of the bursting liability indices, i.e., W ET , W cf , DT, σ c , have been proved to be reasonable for coals by many field data in China. Most bursting classification criteria used for hard rocks often refer to those of coals and were not put forward based on the special mechanical properties of non-coal rocks. The rationality of using the bursting liability criteria of coals to predict the bursting liability of non-coal rocks needs to be studied in depth. Hence, in this study, a large amount of data, including basic mechanical parameters and bursting liability indices of different coals and non-coal rocks have been collected in China, and reported in the previous literatures have been used. The distribution range and distribution proportion of the mechanical parameters and rockburst tendency indices of coals and non-coal rocks are analyzed, and the rationality of using the bursting tendency indices for non-coal rocks is discussed. Finally, a series of comparatively reasonable modified rockburst tendency indices and classification criteria for non-coal rocks were put forward.

Basic Mechanical Parameters
The basic mechanical parameters of the coals and non-coal rocks in China were collected from the previous literature, as shown in Tables S1 and S2 in the Supplementary Material files. The statistical parameters of the basic mechanical parameters of coals and non-coal rocks are shown in Table 1. The detailed location of the coals and non-coal rocks in this study are shown in Figure 2. The basic mechanical parameters of coals and non-coal rocks are shown as Figure 3. From the statistical results, it can be concluded that the distribution ranges of basic mechanical parameters of non-coal rocks are much wider than that of coals, except for the Poisson ratio. The Poisson's ratios of coals and non-coal rocks share a similar distribution range, while the other three parameters have great differences in the distribution ranges. The ranges of σ c , σ t and E are 0-50 MPa, 0-4 MPa and 0-46 GPa for coals, respectively, and 10-310 MPa, 0-20 MPa and 1-82 GPa for non-coal rocks, respectively. It can be easily found that the maximum uniaxial compressive strength of non-coal rocks is more than six times higher than that of coals. In addition, the maximum uniaxial tensile strength and maximum elastic modulus of non-coal rocks are about five times and two times higher than those of coals, respectively, and it is widely known that rocks with higher elastic modulus and uniaxial compressive strength generally have higher brittleness and stronger bursting liability [6,[28][29][30].
The basic mechanical parameters of coal and non-coal rocks were divided into four parts by the same division standard, and the proportion of data points in each interval was counted and plotted in Figure 4. It is obvious that the uniaxial compressive strength σ c , uniaxial tensile strength σ t , and elastic modulus E of coals are concentrated in the smallest interval, whereas, the above three parameters of non-coal rocks are mainly concentrated in the two medium intervals, as shown in Figure 4a-c. The Poisson's ratio proportion pie charts of non-coal rocks and coals have no significant differences, as shown in Figure 4d. These results indicate that there are considerable differences between the basic mechanical parameters of coals and those of non-coal rocks. It is thus questionable to evaluate the bursting liabilities of non-coal rocks by using the same classification criteria of coals. Taking the uniaxial compressive strength as an example, the proportions of heavy bursting liability, weak bursting liability and no bursting liability for coals are 59.6%, 29.8%, and 10.7%, respectively, while they are 97.5%, 2.5%, and 0 for non-coal rocks when the bursting liability of non-coal rocks was evaluated by the uniaxial compressive strength classification criterion of coals, as shown in Figure 5. Therefore, it is necessary to define more reasonable evaluation criteria for non-coal rocks.       The basic mechanical parameters of coal and non-coal rocks were divided into four parts by the same division standard, and the proportion of data points in each interval was counted and plotted in Figure 4. It is obvious that the uniaxial compressive strength σc, uniaxial tensile strength σt, and elastic modulus E of coals are concentrated in the smallest interval, whereas, the above three parameters of non-coal rocks are mainly concentrated in the two medium intervals, as shown in Figure 4a-c. The Poisson's ratio proportion pie charts of non-coal rocks and coals have no significant differences, as shown in Figure 4d. These results indicate that there are considerable differences between the basic mechanical parameters of coals and those of non-coal rocks. It is thus questionable to evaluate the bursting liabilities of non-coal rocks by using the same classification criteria of coals. Taking the uniaxial compressive strength as an example, the proportions of heavy bursting liability, weak bursting liability and no bursting liability for coals are 59.6%, 29.8%, and 10.7%, respectively, while they are 97.5%, 2.5%, and 0 for non-coal rocks when the bursting liability of non-coal rocks was evaluated by the uniaxial compressive strength classification criterion of coals, as shown in Figure 5. Therefore, it is necessary to define more reasonable evaluation criteria for non-coal rocks.

Bursting Liability Indices
The bursting liability indices often used in the bursting tendency evaluation of rocks are put forward on different theory principles, and the differences of rockburst tendency indices between coal and non-coal rocks are still unclear. In this chapter, the theory principles of each indices were summarized, and the differences of rockburst tendency indices between coal and non-coal rocks were analyzed.

Calculation Principle
The calculation principles of bursting liability indices are summarized in Figure 6. The elastic strain energy index WET is the ratio of elastic strain energy ΦSE to plastic strain energy ΦSP, which is determined from uniaxial loading-unloading compression tests. The elastic strain energy index WET is a parameter related to the elastic hysteresis loop of rocks, and the unloading point is set approximately to 80-90% of the uniaxial compression strength σc, as shown in Figure 6a.
With the increase of the permanent (irreversible) strain εd, the elastic strain energy index WET decreases, and the accumulation ability of elastic strain energy ΦSE of rocks decreases, which makes the bursting liability decrease. The bursting energy index Wcf is the ratio of the accumulated deformation energy AS to the dissipate deformation energy AX, which is determined from the complete stress-strain curve of rocks in uniaxial compression tests.

Bursting Liability Indices
The bursting liability indices often used in the bursting tendency evaluation of rocks are put forward on different theory principles, and the differences of rockburst tendency indices between coal and non-coal rocks are still unclear. In this chapter, the theory principles of each indices were summarized, and the differences of rockburst tendency indices between coal and non-coal rocks were analyzed.

Calculation Principle
The calculation principles of bursting liability indices are summarized in Figure 6. The elastic strain energy index W ET is the ratio of elastic strain energy Φ SE to plastic strain energy Φ SP , which is determined from uniaxial loading-unloading compression tests. The elastic strain energy index W ET is a parameter related to the elastic hysteresis loop of rocks, and the unloading point is set approximately to 80-90% of the uniaxial compression strength σ c , as shown in Figure 6a. The area of pre-peak curve is defined as the accumulated deformation energy AS, and the area of post-peak curve is defined as the dissipate deformation energy AX, as shown in Figure 6b. The duration time of post-peak segment of rocks in uniaxial compression tests is defined as the dynamic fracture duration time DT, as shown in Figure 6c. Peng et al. [25] defined the strength brittleness index B as the ratio of uniaxial compressive strength σc to uniaxial tensile strength σt, as shown in Figure 6e. . Calculation methods of bursting liability indices. Notes: ΦSE is the elastic strain energy, and it is the area under the unloading curve; ΦC is the total strain energy, and it is the area under the loading curve; ΦSP is the plastic strain energy, and it is the area enclosed by the loading curve and the unloading curve; εc is the total stain in loading-unloading curve; εd is the permanent nonelastic strain; εP is the total stain in comprehensive loading curve; εF is the stain before pre-peak strength in comprehensive loading curve; AS is the accumulated deformation energy; AX is the dissipate deformation energy; σc is the uniaxial compressive strength; σt is the uniaxial tensile strength.

Distribution Range
In order to compare the bursting liabilities of coals and non-coal rocks, the distribution ranges of bursting liability indices of coal and non-coal rocks are calculated and shown in Figure 7. The bursting liability data are basically collected from the underground engineering data in China (Figure 2), as shown in Tables S3 and S4 in the Supplementary Material files. The statistical parameters of the rockburst indices of coals and non-coal . Calculation methods of bursting liability indices. Notes: Φ SE is the elastic strain energy, and it is the area under the unloading curve; Φ C is the total strain energy, and it is the area under the loading curve; Φ SP is the plastic strain energy, and it is the area enclosed by the loading curve and the unloading curve; ε c is the total stain in loading-unloading curve; ε d is the permanent non-elastic strain; ε P is the total stain in comprehensive loading curve; ε F is the stain before pre-peak strength in comprehensive loading curve; A S is the accumulated deformation energy; A X is the dissipate deformation energy; σ c is the uniaxial compressive strength; σ t is the uniaxial tensile strength.
With the increase of the permanent (irreversible) strain ε d , the elastic strain energy index W ET decreases, and the accumulation ability of elastic strain energy Φ SE of rocks decreases, which makes the bursting liability decrease. The bursting energy index W cf is the ratio of the accumulated deformation energy A S to the dissipate deformation energy A X , which is determined from the complete stress-strain curve of rocks in uniaxial compression tests.
The area of pre-peak curve is defined as the accumulated deformation energy A S , and the area of post-peak curve is defined as the dissipate deformation energy A X , as shown in Figure 6b. The duration time of post-peak segment of rocks in uniaxial compression tests is defined as the dynamic fracture duration time DT, as shown in Figure 6c. Peng et al. [25] defined the strength brittleness index B as the ratio of uniaxial compressive strength σ c to uniaxial tensile strength σ t , as shown in Figure 6e.

Distribution Range
In order to compare the bursting liabilities of coals and non-coal rocks, the distribution ranges of bursting liability indices of coal and non-coal rocks are calculated and shown in Figure 7. The bursting liability data are basically collected from the underground engineering data in China (Figure 2), as shown in Tables S3 and S4 in the Supplementary Material files. The statistical parameters of the rockburst indices of coals and non-coal rocks are shown in Table 2. Figure 6. Calculation methods of bursting liability indices. Notes: ΦSE is the elastic strain energy, and it is the area under the unloading curve; ΦC is the total strain energy, and it is the area under the loading curve; ΦSP is the plastic strain energy, and it is the area enclosed by the loading curve and the unloading curve; εc is the total stain in loading-unloading curve; εd is the permanent nonelastic strain; εP is the total stain in comprehensive loading curve; εF is the stain before pre-peak strength in comprehensive loading curve; AS is the accumulated deformation energy; AX is the dissipate deformation energy; σc is the uniaxial compressive strength; σt is the uniaxial tensile strength.

Distribution Range
In order to compare the bursting liabilities of coals and non-coal rocks, the distribution ranges of bursting liability indices of coal and non-coal rocks are calculated and shown in Figure 7. The bursting liability data are basically collected from the underground engineering data in China (Figure 2), as shown in Tables S3 and S4 in the Supplementary Material files. The statistical parameters of the rockburst indices of coals and non-coal rocks are shown in Table 2.   As shown in Figure 7, the distribution ranges of the bursting indicators, such as elastic strain energy index W ET , bursting energy index W cf and brittleness index B, are 0-26, 0-22 and 0-52 for coals, respectively, and which are 0-26, 0-35 and 0-55 for non-coal rocks, respectively. It is easily concluded that there are few differences for W ET , W cf and B between coal and non-coal rocks in the distribution range. However, there are large discrepancies for DT and σ c between coal and non-coal rocks, as shown in Figures 3a and 7c. The distribution ranges of DT and σ c are 0-3000 ms and 0-50 MPa for coals, respectively, and 0-4000 ms, and 20-310 MPa for non-coal rocks, respectively. The distribution ranges of σ c and DT for coals are much smaller than these for non-coal rocks.

Classification Standards of Bursting Liability Indices
Based on the aforementioned analyses, the classification criteria of the bursting liability indices for coals in Methods for Testing, Monitoring and Prevention of Rock Burst Part 2: Classification and Laboratory Test Method on Bursting Liability of Coal (GB/T 25217.2-2010) [24] are reasonable, and it is necessary to discuss the classification criteria of bursting liability indices for non-coal rocks. In addition, it is well known that the different bursting liability indices are established by different theory foundations. When the different indices were applied to predict the bursting liability of rocks, it is urgent to study the uniformity of the bursting liability results by the classification criterions of different indices.

Bursting Liabilities of Non-Coal Rocks Using Classification Criterions of Coals
The bursting liability classification results of coals and non-coal rocks based on the classification standards in GB/T 25217.2-2010 [24] were obtained, and the percentages of different bursting liability types (heavy, weak, and none) were drawn in the pie charts, as shown in Figure 8. In Figure 8a-c, taking σ c = 50 MPa (the σ c maximum value of coals) as the split point, the non-coal rocks with σ c > 50 MPa and those with 0 < σ c < 50 MPa were discussed separately. The ratios of the three indicators' bursting liability classification results are presented in Table 3. The bursting liability classification results of coals and non-coal rocks based on classification standards in GB/T 25217.2-2010 [24] were obtained, and the percentages different bursting liability types (heavy, weak, and none) were drawn in the pie charts shown in Figure 8. In Figures 8a-c, taking σc = 50 MPa (the σc maximum value of coals the split point, the non-coal rocks with σc > 50 MPa and those with 0 < σc < 50 MPa w discussed separately. The ratios of the three indicators' bursting liability classification sults are presented in Table 3. Overall, the W ET bursting liability degrees (heavy, weak, and none) of coals and noncoal rocks, were 42.8%, 46.7%, 10.5%, and 44.7%, 45.5%, 9.8%, respectively, as shown in Table 3 and Figure 8a. The W cf bursting liability degrees (heavy, weak, and none) of coal and non-coal rocks, were 32.8%, 49.8%, 17.4%, and 31.3%, 45.6%, 23.3%, respectively, as shown in Figure 8b. It can be clearly concluded that there are obvious similar distribution properties between W ET and W cf for coal and non-coal rocks. The percentage ratio among heavy, weak and no bursting liability for elastic strain energy index W ET , bursting energy index W cf and dynamic fracture duration time DT of coals is approximately 9:9:2, 3:5:2, and 3:5:2, respectively, and those of non-coal rocks are 9:9:2, 6:9:5, 1:3:6, respectively, as shown in Table 1. For W ET , W cf and DT of coals, the weak bursting liability occupies the largest proportion, followed by heavy bursting liability, the least is for no bursting liability. For W ET , W cf and DT of non-coal rocks, the distribution lows of W ET , W cf are similar with that of coals, but that of DT is different with that of coals. The no bursting liability of non-coal rocks occupies the largest proportion, about 57.5%, so the classification criteria of W ET and W cf of coals can be used to determine the bursting liabilities of non-coal rocks, while that of DT cannot be used.
When the uniaxial compressive strength σ c of non-coal rock is lower than 50 MPa, the percentage of non-coal rocks with heavy bursting liability is only 7.7%, and these with weak and no bursting liabilities are both 46.2%. Therefore, the bursting liability classification results of coals and non-coal rocks with σ c < 50 MPa using W ET are significantly different. However, the bursting liability classification results of coals and non-coal rocks with σ c > 50 MPa using W ET share a similar distribution characteristic. Similarly, the bursting liability classification results of coals and non-coal rocks with σ c < 50 MPa using W cf are significantly different, these with σ c > 50 MPa using W cf share a similar distribution characteristic, as shown in Figure 8b. Thus, it can be concluded that the classification criterions of W ET and W cf for coals are suitable for the bursting liability classifications of non-coal rocks. For the index DT, however, the proportion distributions of bursting liability evaluation results of coals and non-coal rocks are quite different, similar to the results evaluated by σ c (Figure 5). For non-coal rocks, 57.5% are divided to the rocks with no bursting liabilities using DT classification criterions of coals, which is 16.6% for coals. Therefore, the classification standards of DT and σ c for coals are unreasonable to classify the bursting liabilities of non-coal rocks.

Uniformity of Bursting Liability Classification Results Using Different Indices
For the same rock, when different bursting indexes are used, the rating results may be inconsistent. For example, when the dynamic fracture duration time DT and the elastic strain energy index W ET were used to evaluate the bursting tendency of coals, 32% of the rating results were consistent, while 68% of the rating results were inconsistent. The consistency and inconsistency rate of bursting liability evaluation results of coals using two different indices are shown in Figure 9. The consistency rate of the bursting liability evaluation results by using elastic strain energy index W ET and uniaxial compressive strength σ c is 58%, which is higher than that of any other two indexes. The inconsistency rate of the bursting liability evaluation results by using elastic strain energy index W ET and dynamic fracture duration time DT is 68%, which is higher than that of any other two indexes. Overall, for any two indices, the consistency rate of the bursting liability evaluation results of coals is about from 32-58%, and the inconsistency rate is about from 42-68%. Therefore, it is necessary to build a comprehensive rating method to evaluate the bursting liabilities of rocks. bursting liability evaluation results by using elastic strain energy index WET and fracture duration time DT is 68%, which is higher than that of any other two Overall, for any two indices, the consistency rate of the bursting liability evaluati of coals is about from 32-58%, and the inconsistency rate is about from 42-68%. T it is necessary to build a comprehensive rating method to evaluate the bursting of rocks.

Bursting Liability Criteria of Elastic Strain Energy Index WET and Bursting Index Wcf
As shown in Figure 8, there is a similar WET and Wcf distribution and percent acteristics of coals and non-coal rocks, so it is feasible to use classification criteria ing liability indices WET and Wcf of coals to determine the bursting liabilities of rocks.

Bursting Liability Criterion of Brittleness Index B
According to the statistical results in Figure 7d, the distribution ranges of th ness index B of coal and non-coal rocks have no obvious differences. From the a the sections above, it has been proved that the indexes WET and Wcf are reasonab dict the bursting liabilities of non-coal rocks. Thus, the comparison screening me statistical analysis have been used in this paper to establish the classification sta brittleness index B. The bursting classification results of rocks determined by sistent with that of WET and Wcf. If the bursting classification results of WET an inconsistent, the higher bursting category was chosen as the classification result determined by B. The data of brittleness index B and the bursting rating results h shown in Figure 10. The concentration area has been framed by a dotted black the cutoff point of three bursting liability grades also given in Figure 10, which for no bursting liability and weak bursting liability, and 17 for weak burstin and heavy bursting liability, so the new classification criteria are heavy burstin with B ≥ 17, weak bursting liability with 9 ≤ B < 17, and no bursting liability w which is similar to the scale put forward by Li et al. [26].

Bursting Liability Criteria of Elastic Strain Energy Index W ET and Bursting Energy Index W cf
As shown in Figure 8, there is a similar W ET and W cf distribution and percentage characteristics of coals and non-coal rocks, so it is feasible to use classification criteria of bursting liability indices W ET and W cf of coals to determine the bursting liabilities of non-coal rocks.

Bursting Liability Criterion of Brittleness Index B
According to the statistical results in Figure 7d, the distribution ranges of the brittleness index B of coal and non-coal rocks have no obvious differences. From the analysis in the sections above, it has been proved that the indexes W ET and W cf are reasonable to predict the bursting liabilities of non-coal rocks. Thus, the comparison screening method and statistical analysis have been used in this paper to establish the classification standards of brittleness index B. The bursting classification results of rocks determined by B is consistent with that of W ET and W cf . If the bursting classification results of W ET and W cf are inconsistent, the higher bursting category was chosen as the classification results of rocks determined by B. The data of brittleness index B and the bursting rating results have been shown in Figure 10. The concentration area has been framed by a dotted black line, and the cutoff point of three bursting liability grades also given in Figure 10, which is about 9 for no bursting liability and weak bursting liability, and 17 for weak bursting liability and heavy bursting liability, so the new classification criteria are heavy bursting liability with B ≥ 17, weak bursting liability with 9 ≤ B < 17, and no bursting liability with B < 9, which is similar to the scale put forward by Li et al. [26].

Bursting Liability Criterion of Strength Decrease Rate SDR
As shown in Figures 7c and 8c, the distribution range of index DT and the associated proportion of coal and non-coal rocks have clear differences, so the DT was not suitable to be used in evaluating the rockburst tendency of non-coal rocks. A new index, the strength decrease rate (SDR), is defined as the ratio of uniaxial compressive strength σ c and dynamic fracture duration time DT. The units of SDR are MPa/ms and indicates the drop speed of the bearing stress of rocks in the post-peak stage. Supposed that the bursting classification results of coals determined by SDR is consistent with that of uniaxial compression strength σ c and dynamic fracture duration time DT of coals. The SDR cutoff point between heavy and weak bursting liability is very close to 0.3 MPa/ms, and the SDR cutoff point between weak and no bursting liability is close to 0.015 MPa/ms, as shown in Figure 11. The cutoff points were also applied in the bursting tendency rating of non-coal rocks, so the classification criteria of SDR can be set as heavy bursting liability with SDR > 0.3 MPa/ms, weak bursting liability with 0.015 < SDR ≤ 0.3 MPa/ms and no bursting liability with 0 < SDR ≤ 0.015 MPa/ms. Thirteen types of rocks were used in uniaxial compression tests, and the SDR values of the rocks were obtained, as shown in Table 4. The bursting classification results of the rocks were consistent with the field situation.

Bursting Liability Criterion of Strength Decrease Rate SDR
As shown in Figures 7c and 8c, the distribution range of index DT and the associate proportion of coal and non-coal rocks have clear differences, so the DT was not suitab to be used in evaluating the rockburst tendency of non-coal rocks. A new index, th strength decrease rate (SDR), is defined as the ratio of uniaxial compressive strength and dynamic fracture duration time DT. The units of SDR are MPa/ms and indicates th drop speed of the bearing stress of rocks in the post-peak stage. Supposed that the burs ing classification results of coals determined by SDR is consistent with that of uniaxi compression strength σc and dynamic fracture duration time DT of coals. The SDR cuto point between heavy and weak bursting liability is very close to 0.3 MPa/ms, and the SD cutoff point between weak and no bursting liability is close to 0.015 MPa/ms, as shown i Figure 11. The cutoff points were also applied in the bursting tendency rating of non-co rocks, so the classification criteria of SDR can be set as heavy bursting liability with SD > 0.3 MPa/ms, weak bursting liability with 0.015 < SDR ≤ 0.3 MPa/ms and no burstin liability with 0 < SDR ≤ 0.015 MPa/ms. Thirteen types of rocks were used in uniaxial com pression tests, and the SDR values of the rocks were obtained, as shown in Table 4. Th bursting classification results of the rocks were consistent with the field situation.   Figure 11. The cutoff points were also applied in the bursting tendency rating of non-coa rocks, so the classification criteria of SDR can be set as heavy bursting liability with SDR > 0.3 MPa/ms, weak bursting liability with 0.015 < SDR ≤ 0.3 MPa/ms and no bursting liability with 0 < SDR ≤ 0.015 MPa/ms. Thirteen types of rocks were used in uniaxial com pression tests, and the SDR values of the rocks were obtained, as shown in Table 4. Th bursting classification results of the rocks were consistent with the field situation.

Discussion
For two different bursting indices, the inconsistent rate of the bursting liability evaluation results is about from 42-68%, so some comprehensive methods are necessary to be used to predict the bursting tendency of rocks, which we can summarize as follows: the geo-stress is a non-negligible factor for the occurrence of rock bursts, and the differences of the relationship between the uniaxial compression strength and geo-stress in coal bursting area and non-coal bursting area are discussed.

Comprehensive Bursting Liability Rating of Rock
It is widely known that the occurrence of rockbursts is a complex process and influenced by many factors. Therefore, if only a single bursting tendency index is used, the authenticity of evaluation results is unclear. Therefore, many comprehensive evaluation approaches were proposed to determine the bursting liability of rocks. Then, the process of establishing comprehensive evaluation methods has been drawn, as shown in Figure 12. The first step is distinguishing the rock types, i.e., coal or non-coal rock. After that, the comprehensive methods (i.e., fuzzy mathematical methods, linear or non-linear classification methods, simulation methods, support vector machine algorithms and so on) were used to rating the bursting liability rating of rocks. If the comprehensive rating results deviate for the actual situation, one has to go back to previous steps and re-establish the comprehensive method until the rating results are in accord with the actual situation.

Discussion
For two different bursting indices, the inconsistent rate of the bursting liability evaluation results is about from 42-68%, so some comprehensive methods are necessary to be used to predict the bursting tendency of rocks, which we can summarize as follows: the geo-stress is a non-negligible factor for the occurrence of rock bursts, and the differences of the relationship between the uniaxial compression strength and geo-stress in coal bursting area and non-coal bursting area are discussed.

Comprehensive Bursting Liability Rating of Rock
It is widely known that the occurrence of rockbursts is a complex process and influenced by many factors. Therefore, if only a single bursting tendency index is used, the authenticity of evaluation results is unclear. Therefore, many comprehensive evaluation approaches were proposed to determine the bursting liability of rocks. Then, the process of establishing comprehensive evaluation methods has been drawn, as shown in Figure  12. The first step is distinguishing the rock types, i.e., coal or non-coal rock. After that, the comprehensive methods (i.e., fuzzy mathematical methods, linear or non-linear classification methods, simulation methods, support vector machine algorithms and so on) were used to rating the bursting liability rating of rocks. If the comprehensive rating results deviate for the actual situation, one has to go back to previous steps and re-establish the comprehensive method until the rating results are in accord with the actual situation.

Geo-Stress Factors
With the depth of underground engineering increasing, the geo-stress levels are also increasing rapidly [31][32][33][34]. The geo-stress is an unignored factor affecting the occurrence of rockbursts, especially for non-coal rock engineering. As shown in Table 5, whether in a deep coal mine or in a shallow coal mine, the geo-stress values all exceed the uniaxial compressive strength of coals, and rockbursts occur in coal mines with buried depths of 230-1089 m, while in non-coal rock engineering, the non-coal rocks with a high uniaxial compressive strength do not experience rockbursts under low geo-stress levels. With the increase of the geo-stress levels, rockbursts occur in non-coal rock engineering, but the geo-stress values do not exceed the uniaxial compressive strength of non-coal rocks. In the Jinping II Hydropower Station, a rockburst did not occur in the marble with σ c of 44.6-150. 6 MPa under geo-stress of 11.2-22.3 MPa, and a rockburst occurred in the same marble under a geo-stress of 38.6-44.8 MPa, so the influence of geo-stress on rockbursts in non-coal rocks is worthy of being studied in depth.

Conclusions
The basic mechanical parameters (i.e., Poisson's ratio µ, elastic modulus E, uniaxial compressive strength σ c , and uniaxial tensile strength σ t ) and bursting liability indices (i.e., elastic strain energy index W ET , bursting energy index W cf , dynamic fracture duration time DT, and brittleness index B) of coals and non-coal rocks are compared in detail in this study. The main conclusions are as follows: (1) The basic mechanical parameters (σ c , E, and σ t ) of coal and non-coal rocks are quite different in distribution range. The σ c , E, and σ t maximum values of coals are 50 MPa, 46 GPa, 4 MPa, respectively. The maximum value of σ c of non-coal rocks is six times higher than that of coals, and the maximum values of E and σ t of non-coal rocks are about two and five times higher than that of coals, respectively. The µ values of coal and non-coal rocks share a similar distribution range. (2) The W ET , W cf and B of coals and non-coal rocks share similar distribution ranges, and the distribution ranges of σ c and DT for coals are much smaller than these for non-coal rocks. It can be concluded that the classification criteria of W ET and W cf for coals are suitable for non-coal rocks, while that of σ c and DT are not suitable. (3) The consistency rate of the bursting liability evaluation results of coals is only about 32-58%, and it is necessary to define a comprehensive method to evaluate the bursting liability of rocks.
(4) For non-coal rocks, the classification criterions of brittleness index B and strength decreases rate SDR (the ratio of uniaxial compressive strength σ c and fracture duration time DT) have been redefined in this study as follows: • The classification criterion of the brittleness index B: heavy bursting liability with B ≥ 17; weak bursting liability with 9 ≤ B < 17; no bursting liability with B < 9.

•
The classification criterion of the strength decrease rate SDR: heavy bursting liability with SDR > 0.3 MPa/ms; weak bursting liability with 0.015 < SDR ≤ 0.3 MPa/ms; no bursting liability with 0 < SDR ≤ 0.015 MPa/ms.