Preliminary Study on High-Energy and Low-Energy Microfracture Event Evolution Characteristics in the Development Process of Rock Failure

School of Energy and Safety, Anhui University of Science and Technology, Huainan 232001, China State Key Laboratory of Deep Coal Mining and Environmental Protection, Huainan Mining (Group) Co., Ltd., Huainan 232001, China Center for Rock Instability and Seismicity Research, Northeastern University, Shenyang 110819, China State Key Laboratory of Coastal and Offshore Engineering, Dalian University of Technology, Dalian 116023, China


Introduction
In a study on the recurrence of large seismic events in Polish mines, Gibowicz and Kijko [1] found that the pattern of empirical distributions of the largest seismic events is more complex than expected based on the most general theoretical considerations (e.g., Gumbel distributions) [2]. The observed distributions possess bimodal distribution (BMD) characteristics, as shown in Figure 1 BMD has also been observed in the underground coal mines of Upper Silesia, Poland, at the The BMD results from the mixing of random variables generated by two different phenomena. The first phenomenon is responsible for the low-energy component and the second one for the high-energy component of the distribution. At present, the BMD is considered an important method for observing the evolution characteristics of highenergy and low-energy microseismic (MS) events.
Despite many valuable contributions to the literature, the precise mechanism of the BMD is not yet known. The traditional view is that the BMD obeys at least three statistical distributions. The first was proposed by Stankiewicz [5], who assumed that at any moment, the state of the rock mass can be described by a stress that increases linearly in time in the absence of a seismic event and drops to a certain value from the interval if an event occurs. The other two statistical distributions were proposed by Gibowicz and Kijko [1]. Under the second statistical distribution, high-and low-energy seismicity are independent of each other. Under the third statistical distribution, the high-energy seismicity is dependent on the low-energy seismicity. Monitoring data can be well fitted to these three statistical distributions, but none fully characterizes the BMD mechanism.
There are two hypotheses as to the nature of the two sets of events discussed above [6]. The first states that BMD is a result of the inhomogeneous and discontinuous structure of the rock mass. The second states that the low-energy component of the distribution is a result of mine-induced stress release, while the high-energy component is a result of the interaction between mine-induced and residual tectonic stresses in the area. The first hypothesis is acceptable in certain mining districts and the second hypothesis in others, where geologic factors play by far the most important role in the generation of seismic events in mines.
The early warning mechanism of mine disaster is closely related to the rock and fissures, and the evolution characteristics of high-energy and low-energy microfracture events monitored by microseismic early warning are of great significance to the failure mechanism of rocks [7,8]. Many previous researchers have conducted laboratory and in situ research by acoustic emission (AE) and MS monitoring technologies. In the field of MS monitoring, Xiao et al. [9], Feng et al. [10], Xu et al. [11], and Yu et al. [12,13] studied the rela-tionships between monitored MS activities and rockburst events in tunnels; rockburst events can be predicted by capturing MS events produced during the excavation of a rock. Zhang et al. (2017), Wang et al. [14], and Cheng et al. [15] studied the relationship between the MS event distribution and strata movement caused by underground mining. Cao et al. [16,17] and Li et al. [18] investigated the relationship between MS activities and rockburst events to predict possible rockburst locations in coal mines. Lu et al. [19] and Li et al. [20] studied the relationship between the distribution of MS events and coal-gas outburst in a coal mine. Li et al. [7], Zhang et al. [21], and Wang et al. [14] studied the relationship between MS monitoring results and water inrush. They found that fractured zones in rock masses, which may form water inrush channels, can be determined according to the spatial distribution of MS events.
Xu et al. [22], Dai et al. [23], and Salvoni et al. [24] carried out MS monitoring during the excavation of slopes and to find that ground deformation and MS data reflect rock damage at various depths as well as mechanisms of rock mass instability. Previous studies have also shown that the occurrence of MS events is closely related to failure rock mass and related disasters. AE evolution characteristics have been analyzed based on six main factors: number and energy [25][26][27], frequency [28,29], spatial location [30][31][32], spatial correlation length [33], fractal dimension [34,35], and moment tensor [36]. Dixon et al. [37,38] and Smith [39,40] used AE monitoring for slope deformation and stability predictions. These previous studies provide a workable basis for studying the evolution characteristics of high-energy and low-energy microfracture events related to the BMD mechanism.
Inspired by previous research results, MS monitoring was carried out in No. 22517 working face in the Dongjiahe Coal Mine in this study. A biaxial compression experiment was also performed with a granite specimen in the laboratory. The MS monitoring and experimental results were analyzed to determine the evolution characteristics of high-and lowenergy microfracture events. The relationship between the evolution characteristics and various failure scales were observed to determine the BMD mechanism. Finally, the similarities and differences between high-energy MS event and low-energy AE hit evolution characteristics were determined as discussed in detail below.

Evolution Characteristics of High-Energy and
Low-Energy AE Hits 2.1. Experimental System. The biaxial compression experimental system used in this study is composed of a rock mechanic testing system, an AE monitoring system, and an observation system (Figure 2(a)). The experiment was conducted on a servocontrolled rock testing machine (RLW-3000). The loading was applied in both horizontal and vertical directions. The maximum axial loading was 3000 kN, the maximum horizontal loading was 1000 kN, and the measurement error of loading force was less than 1%. The AE activities were recorded on an eight-channel PCI-2 AE monitoring system (Physical Acoustic Corporation). The bandwidth of the AE acquisition card is 3 kHz-  The test granite specimen in this experiment was a cube with side length of 100 mm and a 15 mm-diameter circular opening in the middle (Figure 2(b)). Before AE monitoring began, the horizontal and vertical loads were gradually loaded to 10 kN by stress loading. The vertical load was stopped as the horizontal load continued to 100 kN and was held unchanged by the servocontrol system. AE sensors with 40 dB threshold and 1 MHz sampling frequency were then arranged on the specimens for AE monitoring. The loading applied at the upper side increased from 10 kN until failure of the granite specimen. The bottom end was prevented from displacement in the vertical direction. The sampling frequency of the high-speed video was 100 frames per second.

Axial Load-Displacement Curve and Fracture Evolution
Process. The axial load-displacement curve was determined as shown in Figure 3. Based on the biaxial compression load-displacement curve, three feature points (A, B, and C) were identified. Point A is the critical point from the compaction stage to elastic deformation, point B is the critical point from elastic deformation to inelastic deformation, and point C is the peak stress of the load-displacement curve. At Point A, the axial loading force of the granite specimen was equal to the lateral loading force. Li et al. [18] showed that when the stress increases to 78.2% of the peak loading value, critical cracks initiate in the tensile stress concentration zones under uniaxial loading conditions; similar results were also reached by Zhong et al. [41]. Based on previous research, the axial loading force at Point B was determined to be 80% of the peak stress under biaxial loading.
As shown in Figure 3, the axial force was 100.14 kN at Point A, 677.29 kN at Point B, and 846.59 kN at Point C. The ratio of the lateral loading force to the axial loading force was defined as the lateral pressure coefficient. During the experiment, the lateral pressure coefficient decreased as the axial load force increased continuously. In region OA, the lateral pressure coefficient was greater than 1. The axial loaddisplacement curve of the granite specimen shows deformation characteristics in the compaction stage. In region AB, the lateral pressure coefficient decreased from 1 to 0.148 and the axial load-displacement curve of the granite specimen exhibited elastic deformation characteristics. In region BC, the lateral pressure coefficient decreased from 0.148 to 0.120 and the axial load-displacement curve of the granite specimen showed plastic deformation characteristics.

AE Hit and Energy Distribution
Characteristics. The number and total energy of AE hits in each time window were counted over the course of the experiment as shown in Figure 4. Both indicators increased until reaching a peak at about 30 s, then decreased until 100 s. From 100 s to 360 s, they remained basically unchanged. The selection of Point A could be revised at 100 s; the axial force was 15% of the peak stress, 130 kN.
As shown in Figure 4, the number of AE hits increased in a fluctuant manner while the total energy of AE dropped slightly from 560 s to 590 s, which is congruous with the characteristics of the number and total energy of AE hits. The number of AE hits decreased rapidly and the total energy released by AE hits increased rapidly after 590 s. The selection of Point B could be revised accordingly at 590 s; the axial force was 85% of the peak stress, 720 kN.
The region AB can be divided into three phases. In the first phase (from 100 s to 360 s), the number and total energy  3 Geofluids of AE hits was small and relatively constant. In the second phase (from 360 s to 510 s), the number and total energy of AE hits increased in a fluctuant manner. In the third phase (from 510 s to 590 s), the number of AE hits increased in a fluctuant manner while the total energy of AE slightly decreased. The region BC extended from 590 s to 699 s. Gao et al. [42,43] carried out similar experiments to obtain similar stress-strain curves and generation characteristics of AE, which confirms the repeatability of these experimental results.
Ai et al. (2011) examined the spatial-temporal evolution of AE locations under triaxial compression to find that the number of and energy released by the AE hits first increases and then decreases as axial stress reaches the confining pressure value. This is the so-called "confining pressure effect" of AE. The peak of the number and total energy of AE hits during the OA stage should be also attributed to the confining pressure.

Definition of High-Energy and Low-Energy AE Hits
(1) Average Energy Logarithm Evolution Characteristics of AE Hits. Ten seconds were selected as a time window and the AEL was calculated to analyze the AE hit energy evolution over the course of the rock specimen failure process. The change curve of AEL versus time during the biaxial compression experiment is shown in Figure 5. This curve can be divided into three stages. In the first stage (from 0 s to 360 s, corresponding to the region OA and the first phase in the region AB), AEL fluctuated around 3.75. In the second stage (from 360 s to 590 s, corresponding to the second and third phases in the region AB), the AEL fluctuated while trending upward. In the third stage (from 590 s to 705 s, corresponding to the region BC), the AEL began to increase rapidly from 4 to 8.23. The average energy corresponding to a given AE hit in the time window increased with some fluctuation after 360 s and significantly increased after 590 s.
(2) Energy Distribution Evolution Characteristics of AE Hits. The energy distribution evolution characteristics of AE hits were analyzed as shown in Figure 6. As shown in Figure 7, the density of AE hits versus the logarithm of AE energies obeys a Burr distribution from 0 s to 670 s. However, it did not obey the Burr distribution from 670 s to 705 s. The density function of the Burr distribution can be expressed as follows: where α is a scale parameter, c is a parameter of inequality, x is the logarithm of energy, and k is a shape parameter.
The peak density as per the AE hit energy distribution increased from 10E2:4 aJ to 10E3:0 aJ between 0 s and 670 s. Although the density of AE hits versus the logarithmic AE energies did not obey the Burr distribution from 670 s to 700 s, the peak density increased from 10E3:0 aJ to 10E5:6 aJ. The energy of AE hits ranged from 10E1:8 aJ to 10E5:6 aJ before 490 s (corresponding to the region OA and the first/second phases in the region AB). After 490 s (corresponding to the third phase in the region AB and the region BC), AE hits with energy larger than 10E5:6 aJ began to appear, then increased in proportion as the loading time progressed. The AEL evolution characteristics are closely related to the number of AE hits with energy larger than 10E5:6 aJ.
(3) Evolution Characteristics of Cracks in Granite Specimen. The evolution of cracks during the biaxial compression experiment is shown in Figure 8. As shown in Figure 8, no obvious microcracks were observed before 477 s. Between 477 s and 587 s, microcracks emerged and developed along the primary structural planes (e.g., interfaces between the groundmass and phenocryst) but did not connect with each other ( Figure 6). After 587 s, these microcracks expanded into tensile cracks (PTC), remote cracks (RC), and shear cracks (SC) around the circular opening due to the stress concentration around the opening. The SC continuously expanded and gradually connected with the RC and PTC over the loading time, as similarly observed by Dzik et al. [44], Martin et al. [45], and Zhu et al. [46].
By comparing the energy distribution evolution characteristics of AE hits and the evolution characteristics of cracks, the increase in AE hits with energy >10E5:6 aJ was found to be attributable to the evolution and growth of microcracks and cracks in the granite specimen. Thus, AE hits with energy < 10E5:6 aJ were defined as "low-energy AE hits" and those >10E5:6 aJ as "high-energy AE hits."

Evolution of High-Energy and Low-Energy AE Hits
(1) Evolution Characteristics of High-Energy and Low-Energy AE Hits. The energy distribution of AE hits over the course of the experiment is shown in Figure 9(a). The energy of AE hits was mainly distributed in the low-energy range during the experiment. The energy distribution of high-energy AE hits  Geofluids is shown in Figure 9(b). There were fewer high-than lowenergy hits, but some did remain in the later stages of the experiment.
AE hits with energy ranging from 10E2:4 aJ to 10E2:6 aJ were selected to analyze the low-energy hit evolution characteristics during the experiment; hits ranging from 10E5:8 aJ to 10E6:0 aJ were selected to analyze the highenergy hit evolution characteristics during the experiment. As shown in Figure 10(a), the energy evolution characteristics of low-energy AE hits was approximately the same as that of all AE hits. As shown in Figure 8(b), the energy evolution characteristics of high-energy AE hits were quite different from that of low-energy hits and can be divided into three stages: basically negligible in the first loading stage (before 360 s), increasing slightly in the second loading stage (from 360 s to 490 s), and increasing rapidly in the third loading stage (after 490 s).
The distribution of AE hits and energy monitored during the experiment were mainly dependent on the lowenergy AE hits. There also were some similarities among the evolution characteristics of high-energy AE hits and the AEL evolution characteristics of the hits; an increase in the number of the high-energy AE hits is the root cause of an increase in AEL. The evolution characteristics of high-energy AE hits also are in close agreement with the evolution characteristics of cracks in the granite specimen.
(2) Evolution Mechanism of High-Energy and Low-Energy AE Hits. The granite specimen used in the experiment has a  5 Geofluids porphyritic texture and is composed of two different sizes of granules. Large granules (or "phenocrysts") were scattered among the small granules (groundmass). The groundmass is composed of microlite, crystallite, cryptocrystalline, or glassy material. In the local region, there were several closely packed, interconnected phenocrysts. Phenocrysts and their interfaces constituted the framework of the granite specimen. The diameter of phenocrysts ranges from 1 mm to 5 mm while that of groundmass granules is less than 1 mm. The evolution mechanism of high-energy and low-energy AE hits can be summarized as per the microscopic structure of the granite specimen, evolution characteristics of cracks in the granite specimen, and the evolution mechanism of highenergy and low-energy AE hits.
In the early stage of the loading process (before 490 s, corresponding to the region OA and the first/second phases in the region AB), low-energy AE hits caused by failure of the small groundmass occurred in the granite specimen. Though groundmass failure had little effect on the damage of the granite specimen, it led to the stress concentration in the specimen's framework (which did not fail in this stage). As the loading increased, the interfaces of one phenocryst to another phenocryst or to the groundmass failed and produced high-energy AE events. The granite specimen entered the midstage of the loading process at this point (from 490 s to 590 s, corresponding to the third phase in the region AB), where the number of low-energy AE hits was much higher than that of high-energy AE hits.
As the number of the failed interfaces continued to increase, stress transferred to the phenocrysts in the granite specimen and caused them to fail alongside the emergence of high-energy AE events. Failure of both the interfaces and phenocrysts damaged the granite specimen. Once a certain amount of damage had accumulated, the granite specimen entered the later stage of the loading process (from 590 s to 705 s, corresponding to the region BC). Phenocrysts inside the specimen failed and quickly penetrated the primary structural planes, which ultimately led to the failure and instability of the whole specimen. A rapid increase in the number of high-energy AE hits (with still relatively few low-energy hits) is an important indicator that this stage has begun. The evolution mechanism of high energy and low energy AE hits explains this phenomenon that the closer to failure AE data are used as initial input into the AE datadriven model to simulate the rock failure process, the better the model predictions of failure are by Wei et al. [47].
(3) Relationship between Evolution Mechanism and Gutenberg-Richter b Value. The Gutenberg-Richter b value was calculated over the loading process. Ten seconds was selected as a time window representative of the whole biaxial compression experiment. The b value in each time window was calculated to determine its variations with time as shown in Figure 11. The b value fluctuated around 0.83 before 490 s (corresponding to the region OA and the first/second phases in the region AB) and decreased slightly from 490 s to 590 s (corresponding to the third phase in the region AB), then   9 Geofluids declined rapidly after 590 s (corresponding to the region BC). At 490 s, the axial loading force increased to 72% of the peak loading value. At 590 s, the axial loading force increased to 85% of the peak loading value (corresponding to Point C in the axial load-displacement curve).
The high-energy AE hit evolution characteristics were inversely proportional to the Gutenberg-Richter b value of the granite specimen during the loading process. The increase in low-energy AE hits did not cause the Gutenberg-Richter b value to change, but the increase in high-energy hits caused the Gutenberg-Richter b value to decrease. A larger b value decreasing rate created a larger increasing high-energy AE hit rate. This further validates the high-and low-energy AE hit evolution mechanism presented above.

High-Energy and Low-Energy MS Event
Evolution Characteristics  Figure 12. The advancement schedule of the No. 22517 working face is given in Table 1.
According to the drill histogram of No. 22517 (Figure 13), the rock mass strata around the working face is, according to the ages of formation from old to new, as follows: middle Ordovician System, Fengfeng Formation; upper Carboniferous System, Taiyuan Formation; lower Permian System, Shanxi Formation; lower Permian System, Xiashihezi Formation; and Quaternary System overburden soil.
Geophones with 43.3 V/m/s sensitivity were adopted in the MS monitoring system used in this study. The response frequency of geophones ranges from 15 to 1000 Hz. The geophone placement scheme for MS monitoring is given in detail by Cheng et al. [15]. The Geiger location method was used here to localize the MS. The P-wave velocity was estimated at 2800 m/s and the S-wave velocity at 1800 m/s based on artificial blasting tests. The ESG microseismic monitoring system is used for signal acquisition, including Paladin underground digital signal acquisition system, Hyperion surface digital signal processing system, and MMVTS-3D visualization software. The system can monitor continuously 24 hours in real time. Many focal parameters such as time-   space coordinates, errors, magnitude, and energy of microseismic events can be obtained. The collected data can be filtered, and the complete waveform and spectrum analysis chart of source information can be provided. At the same time, it can automatically identify the type of microseismic events and eliminate noise events by filtering, threshold setting and bandwidth detection.

Evolution Characteristics of High-Energy and Low-
Energy MS Events. Basic MS monitoring results are given in the previous publication by Cheng et al. [15]. The energy and moment magnitude of MS events over coal seams caused by mining activities were analyzed based on this information; the results are shown in Figure 14. As shown in Figures 14(a) and 14(b), the energy distribution and the moment magnitude distribution of MS events conformed in this case to the BMD. The energy of low-energy MS events was less than 158.5 J, while the energy of highenergy MS events was greater than 158.5 J (Figure 14(a)). The moment of low-energy MS events was less than -0.1, while the energy of high-energy MS events was greater than -0.1 (Figure 14(b)). The average energy logarithm (AEL) was defined to further analyze the difference between high-and -low-energy MS event components. The AEL is a logarithm of the ratio of the total energy to the number of all MS events in a oneday period (Formula (2)). Its physical meaning corresponds to the average energy of MS events in a given day. The curve of AEL versus time is given in Figure 14 where AEL is the average energy logarithm, N w is the number of MS events in a day, and E w is the sum of energy of MS events in a day. Cheng et al. [15] introduced time-, location-, and energyrelated evolution characteristics of MS events in the roof of the Dongjiahe Coal Mine. Based on their monitoring results, new MS events added monthly fell into the region from 30 m behind the mining face to 220 m in front of the mining face along the horizontal direction and in the region from the elevation of +320 m to +500 m along the vertical direction after Mar. 6, 2015 [15]. Most of the new MS events added monthly were concentrated in the coal pillar supporting zone, the support and bed-separation zone along the horizontal direction, and from the top of the fractured zone to the bottom of the continuous zone along the vertical direction. Based on the conceptual model proposed by Qian et al. [48], the rock mass was in a state of near failure or had already failed in the zone where these new monthly MS events occurred. By Feb. 25, 2015, the working face had advanced 360 m (about twice the width of the working face) from the open cut to the stopping line.
The MS monitoring data suggests that the working face was in the insufficient mining stage (in the stage, only the collapse zone is formed but the fracture zone is not formed in the process of rock failure; the maximum subsidence value of the strata continues to increase) before Feb. 25 and that the height of the fracture zone caused by mining did not develop to the final height. After Feb. 25, the working face was in the sufficient mining stage (in this stage, the size of the goaf is large enough and the maximum subsidence value of the strata does not increase) and the fracture zone developed to the final height. Thus, the high-energy MS events occurred beginning on Feb. 25, because the failure scale of the overlying rock mass increased upon the working face entering the sufficient mining stage.

Discussion
There are some similarities among the evolution characteristics of high-energy MS events and low-energy AE hits. Both have bimodal characteristics and can be divided into the high-energy and low-energy components. An increase in the failure scale is the root cause of an increase in the highenergy component. There is also a significant difference between the two: the bimodal distribution of AE hits is far less obvious than that of MS events. There are two potential reasons for this.
The first reason involves differences in the monitoring equipment. The frequency range of the R6a sensor in the AE monitoring system extends from 35 kHz to 100 kHz while that of the geophone in the MS monitoring system is from 15 Hz to 1000 Hz. The bandwidth of the AE monitoring system is much larger than that of the MS monitoring system. In addition, the scale of the AE monitoring system monitoring object is far smaller than that of the MS monitoring system monitoring object. Therefore, the AE monitoring system serves to detect AE events caused by the full-scale microcracks of rock specimens during the experiment while the MS monitoring system cannot detect the MS events caused by small-scale microcracks, which are the main components of low-energy MS events.
The second reason involves differences in the monitoring object. In this study, the object of the AE monitoring system is a granite specimen while that of the MS monitoring system is sandstone/mudstone. Granite is an igneous rock; sandstone and mudstone are sedimentary. The diagenesis of igneous rock material differs from that of sedimentary rock materials, so there are substantial differences between the microstructures of granite and sandstone and mudstone. The rock specimen used in the laboratory was not designed with consideration of its structural plane, though the in situ rock mass contains a large number of original structural planes. These factors may have made the BMD of AE hits less obvious than that of MS events.

Conclusions
Microseismic and acoustic emission monitoring are always carried out in engineering rock masses and experimental rocks, and the evolution characteristics of high-energy and low-energy events occurring in the deformation of the rock are instructive for optimizing monitoring and understanding the failure mechanism. In this study, combining the granite specimen biaxial compression experiment and the MS monitoring results of the 22517 working face in the Dongjiahe   (P1s)   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   (2) Biaxial compression experiments were conducted on a granite specimen to observe various characteristics of low-and high-energy AE hits. The whole loading process was divided into three stages. In the first loading stage, only low-energy AE hits occurred.
In the second loading stage, high-energy AE hits began to appear and slowly grew more common.
In the third loading stage, the quantity of highenergy AE hits increased rapidly until the specimen failed. Low-energy AE hits were more common than high-energy AE hits throughout the whole loading process, but the change characteristics of the high-energy hits appeared to be positively correlated with the evolution of cracks in the granite specimen and negatively correlated with changes in the Gutenberg-Richter b value. The high-energy and low-energy AE hit evolution mechanism was defined here based on the microscopic structure of the granite specimen (3) This paper discussed the similarities and differences between the evolution of high-energy MS events and low-energy AE hits. Both were found to have bimodal characteristics and can be divided into high-energy and low-energy components. An increase in the failure scale is the root cause of the increase observed in the high-energy component. The bimodal distribution of AE hits was far less obvious than that of MS events, which is likely attributable to differences in the monitoring equipment and monitoring object

Data Availability
As for the data availability statement, some or all data, models, or code generated or used during the study are available from the corresponding author by request (list items).

Conflicts of Interest
No conflict of interest exits in the submission of this manuscript.