Applications of Microseismic Monitoring Technique in Coal Mines: A State-of-the-Art Review

Abstract

China’s coal mines have to extend to greater depths for the exploitation of more mineral resources, and they have suffered catastrophic mining-induced disasters, such as rockbursts, water inrushes, coal and gas outbursts, and roof fall accidents. The microseismic monitoring technique is a practical tool for mine safety management, which is extensively utilized in many Chinese coal mines. Microcracks of coal/rock masses are recorded as microseismicities in the field, and the potential mining-induced instabilities can be assessed by in-depth analysis of the microseismic parameters. This study provides a state-of-the-art review of the achievements and developments of the microseismic monitoring technique in coal mines. It also presents some prospects for improving the location accuracy of microseismicity, efficient and intelligent processing of the microseismic data, comprehensive assessment of coal/rock instabilities, and development of new microseismic monitoring equipment. This study is valuable for mine safety management and may contribute to improving the deep mining production.

1. Introduction

According to the International Energy Agency’s 2023 Annual Coal Report, coal remains the largest energy source for producing electricity, steel, and cement, and maintains an important role in the world economy. China, India, and Indonesia are the three largest coal producers, and the International Energy Agency expects China and India to account for more than 70% of global coal consumption [1]. In China, approximately 2950 billion tons of coal resources are stored at a depth of more than 1000 m, and underground mining outputs more than 90% of the coal production [2]. The mining industry has to exploit deep coal resources because several decades of high-intensity mining have exhausted the mineral reserves in the shallow layers [3]. The mining operation depth increases at a rate of 8–25 m per year in China, and the deepest coal mine has extended to approximately 1500 m [4,5]. The geological scenarios become complicated with high original stresses and high stress concentrations due to deep coal extraction [6]. Additionally, the mechanical properties of deep-buried rock mass are evidently different from that of superficial rock mass [7,8]. High-intensity deep coal mining inevitably causes disasters, e.g., rockbursts, water inrush, coal and gas outbursts, and roof-fall accidents [9,10,11,12,13,14,15,16]. Mining-induced disasters destroy underground structures and equipment and threaten the safety of underground workers.

Table 1. Statistics for the deaths caused by different coal mine accidents [17].

Accident Type	Deaths	Accident Type	Deaths
Gas explosion	23,866	Electromechanical accident	1496
Roof accident	23,769	Fire	1092
Haulage accident	6381	Blasting accident	1053
Mine water inrush	5210	Others	4502

lists the deaths caused by different accidents in Chinese coal mines from 1999 to 2018, and the mining-induced disasters associated with gas, roof, and mine water led to 52,845 deaths, accounting for 78.4% of the total deaths.

The microseismic monitoring technique is one of the most effective methods for analyzing, monitoring, and warning of deep coal mining-induced disasters. Though microseismic monitoring has gained increasing popularity regarding the applications in studying mining-induced disasters, it faces numerous technique limitations, e.g., inadequate location accuracy and difficulties in the accurate assessment of coal/rock instabilities. This study aims to highlight the recent achievements of microseismic monitoring in coal mining-induced disasters and propose some possible solutions to the technique limitations, which could be meaningful to promote the development of the microseismic monitoring technique in coal mines.

Mining-induced disasters have been extensively studied by researchers using various methods [18,19,20,21]. Field investigation is crucial for studying the coal/rock failure processes in various conditions. Therefore, an efficient field monitoring method for coal/rock fracturing is essential to analyze, warn, and control these disasters. The real-time microseismic monitoring technique is commonly employed in coal mines, which use seismic signals generated by coal/rock masses to investigate the fracture and failure processes [22]. Microseismic monitoring in mines provides a reliable tool to warn and mitigate coal/rock mass instabilities [23]. Obert and Duvall first found the phenomenon that stressed rocks in deep mines emit micro-level sounds in the late 1930s [24]. Microseismic monitoring was initially employed to identify the source location of rockburst in South African gold mines in the early 1960s [25,26]. Since then, the microseismic technique has been rapidly developed for various geotechnical engineering aspects, e.g., mines, tunnels, caverns, slopes, petroleum and natural gas reservoirs, geothermal reservoirs, and nuclear waste repositories [27,28,29,30,31,32,33]. In recent decades, the microseismic monitoring technique has become a practical tool for mine safety management and has been extensively employed in the mines of many countries, e.g., China, Canada, Australia, Poland, America, and India [34,35,36,37,38,39]. Studies on the microseismic monitoring technique in coal mines mainly focus on source location methods for microseismicities in complex geological scenarios, quantitative interpretation of coal/rock fracturing, early warning of mining-induced disasters, and geotechnical investigations to determine the causes of the disaster. Wang et al. [40] proposed an innovative method to locate microseismicities in jointed rock mass of deep coal mines and integrated improved particle swarm optimization and the double-difference location method. Khan et al. [41] used the temporal parameters, energy induces, and frequency of microseismic data to quantify the increased rock pressure caused by the extraction of an ultra-thick coal seam. Di et al. [42] proposed an early warning method for rockbursts using microseismic, acoustic emission, and electromagnetic radiation parameters based on long short-term memory recurrent neural networks. Roof-fall accidents in Indian coal mines, such as the Churcha mine, the Kothadi mine, and the GDK-11A incline, are closely related to the presence of a hard roof. To enhance the safety of longwall coal mining in India, Mondal and Roy [43] incorporated the spatial distribution and magnitude of microseismicities and parameters of surface blasting to forecast roof falls. Driad-Lebeau et al. [44] studied the causes of a violent rockburst that occurred at a French coal mine based on microseismic monitoring and in situ stress measurements. Li et al. [45] used microseismic monitoring as the primary tool to study the water inrush disaster at the Laohutai coal mine and found some abnormalities in the microseismic parameters prior to the incident, i.e., decreases in the P-wave to S-wave velocity ratio and dominant frequency to abnormal low values, an increase in the S-wave to P-wave amplitude ratio to abnormal high values, the presence of a high-amplitude and long-period waveform, and a decrease in microseismic events.

Within this study, a state-of-the-art review is presented in the scope of applications of the microseismic monitoring technique in studying coal mining-induced disasters. Some fundamental aspects of microseismic monitoring and the quantitative interpretations of microseismicities are first introduced in Section 2 and Section 3. Then, the associated achievements of the microseismic monitoring technique in studying mining-induced instabilities are described in Section 4 based on an extended literature review. Finally, the prospects of the microseismic monitoring technique in future studies are depicted in Section 5. This paper aims to familiarize mining managers with the microseismic monitoring technique, which is currently extensively employed in multiple countries, and improve both safety management and mining production during deep coal extraction.

2. Fundamental Aspects of Microseismic Monitoring Technique

2.1. Microseismic Monitoring Principle

Rock fracturing and the development of existing rock fractures generate microseismic events and radiates seismic waves from the fracturing source that can be detected by sensors [46]. Figure 1 shows the seismicity frequency spectrum at different scales of rock fractures in various fields. The frequency of mining-induced seismicities mainly ranges from several to a few thousand Hertz, and these events can be detected by a geophone or accelerometer array featuring an excellent frequency bandwidth.

Figure 2 illustrates a general microseismic monitoring configuration in underground coal mines. The microseismicity generated by coal/rock fracturing can be recorded by a geophone or an accelerometer as an analog signal series and is digitalized by the data acquisition system of underground substations. The digital seismograms recorded by the substations are sent to the surface server by an Ethernet ring network. Furthermore, the seismic signal series can be manually processed in the surface office, and the parameters of microseismicities, e.g., the energy, source radius, and stress drop, are calculated by the supported software. The microseismicities can be intuitively displayed on the screen at the monitoring center as balls of different diameters and colors, and the risk of mining-induced instabilities is able to be assessed via an analysis of an extensive microseismic parameters dataset.

2.2. Microseismic Source Location and Sensor Layout

There are two common approaches to locating microseismicity, namely, the arrival time approach and the triaxial sensor approach, and the former is widely used in most underground location cases [48]. Seismic waves radiating from the microseismic source can be detected by sensors installed at suitable positions, and the arrival time function for each sensor is listed as follows:

$$\sqrt{{(x_i-x_0)}^2+{(x_i-x_0)}^2+{(x_i-x_0)}^2}=v_p(t_i-t_0)$$

(1)

where, $(x_0,y_0,z_0)$ and $t_0$ are the coordinates and onset time of the microseismic source, $(x_i,y_i,z_i)$ and $t_i$ are the coordinates and arrival time of the i-th sensor, and $v_p$ is the P-wave velocity. Classical methods, e.g., the Geiger method, the double-difference method, and the simplex method, have been commonly employed to identify the source location of microseismicity. However, these classical methods are heavily dependent on the sensor layout, the precision of arrival time picking, and the velocity model of seismic waves, which leads to the low accuracy of the microseismic source location [47]. The simplified elastic wave velocity model used to interpret seismic wave propagation in layered grounds of coal mines results in a reduction in the accuracy of identifying the microseismic location. Cong et al. [49] proposed a novel inversion method for a stratified velocity model in layered ground based on the time difference location algorithm of microseismicities, and the method showed good performance at improving the accuracy of the microseismic location. The source location is of great importance for risk assessment of impending mining-induced instabilities. Several location methods have been developed to improve the microseismic location accuracy, e.g., the nonlinear method, methods without the pre-measured wave velocity or arrival time picking, the collaborative method based on analytical and iterative solutions, and the optimized relative location method [50,51,52,53,54].

The sensor layout is a key contributing factor influencing microseismic location accuracy. Longwall extraction is the most popular mining method in China, and microseismic sensors in coal mines are usually distributed in a plane around the longwall panels (Figure 3) [55]. The planar layout of the microseismic network may make it difficult for high-precision identification of the microseismic location. Some optimization theories of the microseismic network are developed, such as calculating network monitoring capability based on the Monte Carlo method [56] and designing the microseismic network based on the degree of seismic danger (D value) and spatial concentration degree of seismicities (C value) [57] to enhance the accuracy and the reliability of the microseismic location in coal mines.

2.3. Microseismic Parameters

Microseismic parameters are closely related to coal/rock fracturing at the source and are used to characterize the development of microcracks, variation in the rock stress regime, and the risk of catastrophe instabilities. Energy and the event count are two basic microseismic parameters, which reflect the intensity and frequency of coal/rock fracturing [58]. However, utilizing the basic parameters from limited microseismic data makes it difficult to characterize rock fracture development and assess the risk of mining-induced instabilities. Other microseismic source parameters and statistical parameters are also employed to study the rock fracturing and failure process during coal mining. Lu et al. [58] characterized the rockburst caused by a hard roof fall based on microseismic parameters, e.g., energy, event count, Z value, lack of shock b_L, and dominant frequency. Cheng et al. [59] used source parameters and statistical parameters, e.g., energy, seismic potency, and b value, to study the potential precursory indicators of coal mining-induced water-conducting channels. Li et al. [60] integrated microseismic parameters, e.g., energy, the apparent volume, and the Schmidt number, to study the progressive failure process of the longwall working face. Ghosh and Sivakumar [39] employed microseismic parameters, e.g., event rate, energy, and seismic viscosity to understand rock fracture development and identify the abnormal stress concentration prior to a roof fall.

Table 2. Microseismic parameters for characterization of coal mining-induced instabilities.

Parameter	Description
Event count/rate	The number of microseismic events, and is used to reflect the frequency of rock fracture.
Seismic potency P	The seismic potency represents the co-seismic inelastic deformation at the microseismic source, and is written as $P_P{,}_S=4\pi \rho v_P{,}_{S}R{\displaystyle {\int }_0^{t_s}{u}_{\mathrm{corr}}}(t)dt$, where, vP, S is P-wave or S-wave velocity, R is the distance from the microseismic source, ucorr(t) is the square of the seismic wave displacement pulse corrected for the far-field radiation pattern, and ts is source duration [61].
Seismic energy E	The energy of the microseismic event reflects the intensity of rock fracturing. Seismic energy of the P-wave or S-wave can be calculated as $E_P{,}_S=\frac{8}{5}\pi \rho v_P{,}_S{R}^2{\displaystyle {\int }_0^{t_s}{\dot{u}}_{\mathrm{corr}}}(t)dt$, where, ρ denotes rock density [61].
Seismic moment M0	A parameter measures the inelastic deformation at the microseismic source, and can be calculated as M0 = μP, where, μ denotes rock shear modulus.
Apparent stress σA	Apparent stress is the ratio of the seismic energy E to the seismic potency P [61].
Apparent volume VA	Apparent volume, often used together with the energy index, scales the rock volume with the inelastic change, and is written as VA = M02/(2μE) [62].
Energy index EI	Energy index is the ratio of seismic energy E of a microseismic event to the average energy $\overline{E}(P)$ of microseismicities with potency P, and can be calculated as $EI=E/\overline{E}(P)$ [63].
Schmidt number Sc	A parameter measures the spatio-temporal complexity of seismic flow and reflects the potential instability, and can be calculated as $Sc=\frac{4{\mu }^2\Delta V\Delta t(\overline{t}){\displaystyle {\sum }_{t_1}^{t_2}E}}{\rho {(\overline{X})}^2{({\displaystyle {\sum }_{t_1}^{t_2}{M}_{ij}})}^2}$, where, ΔV denotes the volume of interest, Δt is the time step, $\overline{t}$ is the mean time of microseismicities, t1 is the start time of a time period, t2 is the end time of a time period,$\overline{X}$ denotes the mean distance between interacting microseismicities, Mij is the seismic moment [64].
Fault total area A(t)	Fault total area is defined as $A(t)={\displaystyle \sum _{k=k_0}^{k-1}N(k)L^{k=k_0}}(L=4.5)$, where, k0 and k are the lower limit energy level and energy level of the microseismicity, respectively, N(k) denotes the event count of a given energy level k [58].
b value	The slope of the Gutenberg and Richter relation written as log N = a-bM, where M is the magnitude, N is the number of seismicities with magnitude larger than M, a is a constant reflecting the level of seismicities [65].
Lack of shock bL	A statistical parameter defined as $b_L=\frac{0.4343}{\overline{M}-M_0}$, where, $\overline{M}$ is the average energy level, and M0 is the initial energy level [66].
Z value	Z value is a parameter to describe the rockburst risk defined as $Z=\frac{\overline{M}-\overline{m}}{\sqrt{{\sigma }_M^2/N-{\sigma }_m^2/n}}$, where, $\overline{M}$ and $\overline{m}$ are the arithmetic mean of average magnitude samples in the statistical period and the observation period, respectively, N and n are the numbers of $\overline{M}$ and $\overline{m}$ samples, σM and σm are their standard deviations, respectively [58].

lists the above-mentioned microseismic parameters in the published literature.

3. Interpretation of Mining-Induced Microseismicity

3.1. Microseismic Focal Mechanism

The microseismic focal mechanism is correlated to the fracturing behavior of rocks, which can be used for the inversion of the rock fracture type [67]. The ratio of the S-wave energy component Es to the P-wave energy component Ep (E_s/E_p) is found to be an indicator for characterizing the focal mechanism of microseismicities [68]. It is generally accepted that if E_s/E_p > 10, the tensile component is dominant in the rock-fracturing process [69,70]. If E_s/E_p < 20 to 30, the shear component dominates the rock-fracturing process [71]. Duan et al. [72] studied the focal mechanism of mining-induced seismicities near an intrusive dyke via analysis of the E_s/E_p ratios (Figure 4), and the large-magnitude microseismic events characterized by high E_s/E_p ratios revealed the shear slip rock fracture process near the geological discontinuity in an Australia coal mine. Mu et al. [73] employed E_s/E_p ratios to characterize the failure of a grouted floor and found that the grouting slurry significantly influences the failure mechanism of the grouted floor and the proportion of tensile fracture decreases with the tensile strength enhancement by grouted cement. Wojtecki et al. [74] found that the focal mechanisms of 26 tremors induced by destress blasting in a Polish coal mine differ from that of tremors induced by coal mining, and the former exhibits a decreased E_S/E_P ratio.

In 1970, Gilbert first proposed the concept of a moment tensor to study the earthquake mechanism and total seismic moment based on the normal mode theory [75]. Since then, many researchers have used different moment tensor inversion methods to understand the focal mechanisms of seismicities [67]. The moment tensor inversion has been extensively employed in Chinese coal mines to analyze the fracture behavior of coal/rock masses. Song et al. [76] revealed the focal mechanisms of microseismic events of irregular coal pillars in the Dongtan coal mine through a moment tensor, and the result suggested that the tensile failures were dominant at the coal pillar edge with the focal mechanism solution of reverse fault sliding, while the coal pillar core area experienced compression failures with the focal mechanism solution of normal fault sliding (Figure 5). Ma et al. [77] inverted the focal mechanisms of microseismic events in the floor of a panel of the Dongjiahe coal mine by moment tensor analysis and found that the hidden reverse fault has a dip angle and a rupture length of approximately 70° and 21 m, respectively. Wu et al. [78] developed a preliminary criterion to determine the rock fracture mechanism utilizing source rupture parameters obtained by moment tensor analysis and revealed that the fracture and movement of the overlying rock structure is the main reason for mining-induced strong seismicities in the Dongtan coal mine. Cheng et al. [79] conducted moment tensor analysis to invert the focal mechanism of microseismicities with energy exceeding 50 J, determining the dip angle and the azimuth angle of the concealed reverse fault to be 73° and 307.5° in relation to the orbital roadway of the Dongjiahe coal mine, respectively.

3.2. Stress Inversion

Deep coal mining inevitably leads to stress adjustments in the surrounding rocks: the abutment stress generally increases from the virgin stress regime with the progress of excavation and decreases to residual stress when the highly stressed coal mass exceeds its strength limit [2]. The redistributed stress is influenced by numerous contributing factors, e.g., the mechanical properties of coal/mass, the geometric characteristics of overlying strata, mining height and intensity, and geological structures. The stress around the longwall working face can be measured by stress meters embedded in the coal/rock masses [80]. However, describing the stress distribution using limited data from a few stress meters is exceedingly difficult. Understanding the high stress concentration zone is critically important to evaluate the risk of mining-induced instabilities. Seismic velocity tomography is an effective method for stress evaluation, which can be used to identify the impending coal/rock failure in coal mines.

Rock deformation generates the ultrasonic P-wave and S-wave with different transmit velocities [81]. The variation of seismic wave velocity is often utilized to characterize the stress-level changes caused by mining [82]. Seismic velocity tomography is dependent on the relation that the seismic wave velocities in the rock medium (divided into discrete units) equals the transmission distance divided by the traveling time, as follows [83,84]:

$$V=\frac{L}{T}\to VT=L$$

(2)

$$T_i={\displaystyle {\int }_{L_i}\frac{dL}{V\left(x,y,z\right)}}={\displaystyle {\int }_{L_i}S\left(x,y,z\right)dL}$$

(3)

$$T_i={\displaystyle \sum _{j=1}^M{d}_{ij}{S}_j} \left(i=1,\cdots ,N\right)$$

(4)

where $V(x,y,z)$ is the seismic ray velocity, L denotes the transmission distance of the seismic ray, T denotes the traveling time of the seismic ray, $S(x,y,z)$ denotes the slowness, d_ij is the distance of the i-th seismic ray through the j-th rock unit, and N and M are the numbers of the seismic ray and rock units, respectively.

Cai et al. [85] characterized the ability of seismic velocity tomography concerning the detection of geological discontinuity, the identification of stress change, and the assessment of rockburst hazard by applications of active and passive seismic velocity tomograms in longwall mining panels of the Yuejin coal mine, and a conclusion was drawn that the active seismic velocity tomography detects the geological discontinuity accurately, while the passive seismic velocity tomography works better in the cases of identifying the stress change and assessing the rockburst risk. Lu et al. [81] applied compressional wave velocity tomography to invert the stress regime consisting of static and dynamic stresses and found that a sharp increase in the stress concentration factor is the stress precursor of the rockburst in the Junde coal mine. Wang et al. [86] used seismic velocity tomography to recognize the high-stress region in the Xing’an coal mine, and they further employed large-diameter hole and deep-hole blasting as countermeasures to mitigate the rockburst risk in the high-velocity region (red region in Figure 6).

3.3. Spectral Analysis

Microseismic waveforms are complicated and unstable time series, and these signals contain abundant information related to the coal/rock rupture process [87]. The frequency-spectrum characteristics of microseismic waveforms are robust due to their consistencies with physics and mathematics [88]. The commonly used waveform analysis methods for microseismicities consist of frequency spectrum analysis and time-frequency analysis methods based on the Fourier transform. The Fourier transform can transfer the seismic waveform into the signal series in the frequency domain conveniently, as follows:

$$F\left(f\right)={\displaystyle \underset{-\infty }{\overset{+\infty }{\int }}x\left(t\right)}{e}^{-j2\pi ft}dt$$

(5)

where $x(t)$ is the signal series of the seismic waveform, f is the frequency, t is the time, and j denotes the imaginary unit. The Fast Fourier transform is the most famous method for frequency-spectrum analysis. However, the Fast Fourier transform cannot realize frequency localization and is not able to provide information in the time domain. The time-frequency analysis methods, e.g., the Wavelet transform, the S transform, and the Hilbert–Huang transform, are utilized to study the time-frequency characteristics of mining-induced microseismicities. Lu et al. [88] studied the frequency-spectrum characteristics of microseismic signals recorded in the laboratory coal/rock compression test and coal mine, and they concluded that the increase in the vibration velocity of the lower-dominant-frequency microseismic signal is an important precursory indicator for rockbursts. Shu et al. [89] studied the frequency and time-frequency characteristics of microseismic signals generated in different construction procedures in the heading face, and some frequency parameters were employed in the long short-term memory network to warn of the outburst disaster. Wang et al. [90] proposed a dual-channel convolution neural network to classify different seismic signals in coal mines, and the new network improved the classification performance by taking advantage of wavelet packet decomposition coefficients in terms of signal enhancement and noise cancellation. Wang et al. [55] used the frequency parameters extracted by the S transform and Manifold Learning methods to identify the microseismic signals in the Xiashijie coal mine with a high accuracy of 94 percent. Li et al. [87] extracted microseismic features such as the waveform energy (Figure 7) using the Hilbert–Huang transform to study the time-frequency and energy characteristics of microseismic waveforms prior to the rockburst in the Qianqiu coal mine and drew some valuable conclusions for the improvement of rockburst prediction accuracy concerning the dominant frequency variation and energy accumulation.

4. Applications of Microseismic Monitoring in Coal Mine

4.1. Roof Strata Behavior

Longwall coal mining causes the deformation, fracture, and movement of roof strata, which influences the surface deformation and the stability of underground excavations. It is generally accepted that the caved zone, the fractured zone, and the continuous deformation zone of the overburden movement in the vertical distribution are distinctly distributed above the excavation panels in the longwall extraction method (Figure 8) [91]. Cheng et al. [92] proposed a zoning model that divided the overburden movement into six zones in the vertical direction based on the quantity and energy distributions of microseismic events in the Dongjiahe coal mine. The roof generally fractures and caves periodically with the progress of excavation, which releases the excess accumulated stress. Mondal et al. [93] used the fractal patterns of a microseismic event’s three-dimensional distribution to divide the roof-fall risk into five threat levels and two hazard levels and further designed an artificial neural network for automatically identifying the roof-fall threat level in real time. The longwall top coal caving mining method is effective and cost-efficient for extremely thick coal seams, and the large mining-height excavation inevitably induces more intensive movement and failure of overburden. Yu et al. [94] studied the influence of a longwall top coal caving operation on the overburden movement above a longwall panel of the Tashan coal mine and divided three zones of the overburden movement in the vertical direction based on the microseismic event density.

The fracturing and movement of massive and hard strata induced by longwall coal mining is different from that of other overlying strata. The massive hard rock reduces the cavability of roof strata and delays the caving of the large-area hard roof, which releases excess accumulated stress and strain energy violently and may induce catastrophic hazards, e.g., rockbursts and coal and gas outbursts. Ning et al. [95] studied the fracturing and movement of the double-layer hard and thick sandstone main roofs in the Xinhe coal mine using microseismic monitoring, and the severe strata behavior was characterized by increases in the microseismic event count and energy. Lu et al. [96] integrated the spatio-temporal distribution and the frequency characteristics of microseismicities to study the strata behavior of hard and thick igneous layers in the Yangliu coal mine and evaluated the hard roof fracturing intensity using the ratio of spectral energy of different frequency bands.

4.2. Rockburst

Rockburst hazard becomes increasingly significant in Chinese coal mines as coal extraction goes deeper. Intense rockbursts not only destroy underground structures but also trap or kill working miners, and

Table 3. Some intense rockburst cases in Chinese coal mines [86,100].

Occurrence Time	Location	Description of the Rockburst
19 June 2007	25080 working face of Yuejin coal mine	destructed a nearly 300-m-long roadway.
5 June 2008	21201 working face of Qianqiu coal mine	killed 13 miners and damaged a 565-m-long roadway
1 March 2011	25110 working face of Yuejin coal mine	damaged a 200-m-long roadway
15 October 2012	Xing’an coal mine	damaged a 104-m-long section of the roadway
15 March 2013	Junde coal mine	trapped 25 miners, and destructed the tailgate and headgate
26 July 2015	3302 working face of Xingcun coal mine	damaged a 200-m-long roadway
22 December 2015	13230 working face of Gengcun coal mine	damaged more than 160-m-long roadway
15 September 2016	Open-off cut of Junde coal mine	trapped and suffocated five miners
20 October 2018	Longyun Coal Industry Co., Ltd., Yuzhou, China	killed 21 miners

lists some intense rockburst cases and associated descriptions in Chinese coal mines. Microseismic monitoring is an effective method to characterize the rockburst generation process, which has been extensively employed to predict and mitigate rockburst hazards. Lu et al. [97] analyzed frequency-spectrum characteristics to study the microseismic precursors for rockburst triggered by both the high original stress (static stress) and the roof fall disturbance (dynamic stress) based on laboratory compression tests and field monitoring of an intense rockburst in the Huating coal mine. Cao et al. [98] utilized microseismic multidimensional information, such as the microseismicity evolution, the event counts and energy release, and the passive seismic velocity tomography, to assess the rockburst risk and determine the high-risk region near a large residual coal pillar in the Xuzhuang coal mine. Wang et al. [99] integrated the microseismic event count, the energy density clouds, and the seismic velocity tomography to characterize the adjustment of the stress regime during rockburst generation processes, and further proposed some effective countermeasures to control rockburst disasters in subsequent mining activities.

Rockburst mechanisms and the generation process are complicated and influenced by many contributing factors, and it is difficult to identify the comprehensive characteristics of a rockburst and predict rockburst hazards accurately using a single microseismic parameter. Therefore, extraction of the microseismic multi-parameter information is of great significance in recognizing rockburst precursors and warning of rockbursts [101]. Lu et al. [58] used microseismic multi-parameter information, e.g., the energy, event count, Z value, and dominant frequency, to study the rockburst hazard in the Junde coal mine. Dou et al. [101] developed a microseismic multi-parameter index system including the bursting strain energy index and 14 indices associated with time, space, and magnitude information, and the critical values of these indices were determined to warn of coal/rock failures. Cai et al. [102] proposed a fuzzy comprehensive evaluation model to quantitatively evaluate the likelihood of rockburst occurrence and forecast rockbursts in coal mines, selecting ten microseismic indices based on the acoustic emissions of coal samples.

The deformation and rupture of coal/rock masses inevitably release some energy, which radiates detectable seismic waves and electromagnetic radiation [103]. The microseismic system records coal/rock mass ruptures globally, while the electromagnetic radiation equipment monitors the coal /rock fractures in a limited area. Microseismic monitoring and electromagnetic radiation can be used together to enhance the prediction accuracy of rockbursts in coal mines. Dou et al. [104] established a multi-parameter system relying on microseismic monitoring, electromagnetic radiation, and acoustic emissions for classification in forecasting rockburst, and the proposed method has been validated to be effective in more than ten burst-prone coal mines. Li et al. [105] investigated the rockburst generation process and mechanism via an analysis of microseismic monitoring and electromagnetic radiation data and identified a quiet period of microseismicity accompanied by an increase in electromagnetic intensity to the peak value of rockburst precursors. Mu et al. [106] established a rockburst monitoring and early warning method, integrating microseismic monitoring and electromagnetic radiation, which uses microseismic indices to predict the dynamic stress in the regional space and electromagnetic radiation indices to evaluate the concentration of static stress in the local mining space.

4.3. Water Inrush

Chinese coal mines face the most complicated hydrogeological conditions worldwide, and nearly 900 coal mines have suffered the serious threat of water inrush [19]. Water inrush is a catastrophic disaster frequently encountered in Chinese coal mines, caused by high-intensity mining in complex hydrogeological scenarios. As the mining operation goes deeper, the increasing water pressure and ground pressure result in a sharp increase in the probability of water inrush. The water-conducting channel is a necessary condition for water inrush in deep coal mines, which involves the evolution of coal/rock micro-fractures under a high-stress regime and water pressure [59]. Figure 9 illustrates a typical water inrush disaster from various water source locations, and the water-conducting channel connecting the excavation face and the water source is one of the key factors inducing water inrush. Therefore, identification of the micro-fracture precursor is essential to study the generation of the water-conducting channel and to warn miners of water inrush in coal mines.

Microseismic monitoring is capable of capturing and characterizing micro-fractures in coal/rock mass, which has been extensively utilized to study the generation process of water inrush during coal mining in recent years. Sun et al. [107] estimated the failure depth of an inclined floor above a confined aquifer using the planar and vertical distribution of microseismic events at the Taoyuan coal mine and successfully identified the floor near two mining roadways as dangerous areas regarding water inrush. Ma et al. [77] used the contour of microseismic event density to identify the high-risk area of water inrush in the Dongjiahe coal mine; furthermore, they inverted the focal mechanism of microseismicities in the potential water inrush region by executing the moment tensor analysis, which revealed the mixed failure mechanism of the floor ruptures. Cheng et al. [59] identified the location and generation of water-conducting channels in the Dongjiahe coal mine by utilizing the contour of microseismic energy density, and the variation in microseismic parameters, e.g., the energy release, seismic potency, and cumulative apparent volume, were analyzed to study the precursors for the water flow channels. Furthermore, they established an early warning system for water inrush based on geophysical methods and a geological survey (Figure 10). Ma et al. [108] used the improved theoretical model considering the confined aquifer and numerical modeling considering the coal/rock heterogeneity to investigate the stress regime variation and floor failure process; additionally, they estimated the floor failure depth using microseismic monitoring data. Zuo et al. [109] used theoretical analysis, finite element modeling, and microseismic monitoring to reveal the generation mechanism of the water-conducting channel in the floor of the deep mining face of the Xingdong coal mine and assess the performance of countermeasures for disaster control employing multi-level cooperative grouting reinforcement. Li et al. [110] used the variation in microseismic event counts and spatial location to study the fracture and development of crack structures in the floor of a longwall panel of the Xingdong coal mine; furthermore, they established a warning system for floor water inrush based on monitoring technologies of crack structures.

4.4. Coal and Gas Outburst

The generation and mechanism of coal and gas outbursts during deep coal mining are complicated, and a catastrophic disaster suddenly ejects a significant quantity of coal and gas to the excavation face, which destroys underground structures and threatens worker safety [111]. Approximately 510 coal and gas outburst accidents occurred in China between 2001 and 2017, which resulted in 3576 deaths [112].

The micro-fracture of coal/rock masses is the precursory characteristic of coal and gas outbursts; therefore, microseismic monitoring, which is capable of capturing the precursor, is extensively utilized to study and warn of coal and gas outburst disasters. Lu et al. [113] studied the generation process and the mechanism of a coal and gas outburst in the Haizi coal mine by analyzing the spectral characteristics of seismic waveforms and the evolution of the event count and energy release and revealed that the hazard was caused by external shock wave disturbance and highly stressed igneous rock. Si et al. [114] employed seismic tomography and microseismic monitoring programs to assess the dynamic response of the coal seam of Coal Mine Velenje and found a direct relationship between microseismic energy and gas emissions, in which the gas emission rate tends to reach its peak with a sharp increase in microseismic energy. Ding et al. [115] analyzed the seismic waveform, time-frequency spectrum, and fractal of microseismic energy to identify the differences between tremors and coal and gas outbursts in coal mines. They further divided the coal and outburst process into the pre-shock stage, the secondary shock stage, and the main shock stage. Shu et al. [89] utilized the characteristic parameters of the frequency spectrum and microseismic waveform as inputs for the long short-term memory network to recognize the microseismicities generated in different operation stages in the heading face of the Xinyuan coal mine with a low loss value and high accuracy rate, and these events can be used to study the gas concentration process and accurately warn miners of the outbursts. Zhang et al. [116] divided the coal and gas outburst process into the preparation stage, the start stage, the development stage, and the stop stage based on the variation in cavity gas pressure in a simulation experiment (Figure 11), noting that coal and gas outbursts occur at stage III. Furthermore, the microseismic characteristics during the generation process of coal and gas outbursts were studied, which quantitatively describe the coal and gas outburst process and provide insight into warnings of coal and gas outbursts.

5. Some Prospects of the Microseismic Monitoring Technique in Coal Mines

Though coal demand and output will fall in almost all advanced economies, for example, Poland will stop coal mining by 2049 and the coal production of the United States will decline to below 400 million tons in 2026 [1], coal will still play an important role in developed economies in the following decades due to the subdued hydropower and nuclear electricity generation. The coal mining industry has no choice but to exploit deep-seated coal resources due to the depletion of shallow coal resources. The mechanical properties of coal/rock mass in deep coal mines and their engineering responses present significant differences compared to those in a shallow mining environment. As a coal mine goes deeper, the influences of high ground stress, high temperature, high gas pressure, and mining disturbances become serious, which increases the frequency of mining accidents and makes the disaster mechanism more complicated. Though microseismic monitoring has been commonly employed as a practical tool for safety management in many Chinese coal mines, some problems should be solved to make the technique more sophisticated and efficient for the monitoring and warning of mining-induced disasters.

5.1. Improving the Location Accuracy of Microseismicity

The location accuracy of microseismicity directly affects the accuracy of the source parameters, which will constrain the application of microseismic monitoring in the study of coal mining-induced dynamic disasters. The coal/rock fracture emits seismic waves traveling in layered rocks, and a complicated anisotropic velocity model should be adopted to improve the location accuracy of microseismicity instead of the isotropic velocity model. Furthermore, new methods should be developed to improve the precision and efficiency of locating the microseismic source in the coming years. Lagos and Velis [117] used Very Fast Simulated Annealing and Particle Swarm Optimization to optimize the location of microseismicity generated in hydraulic fracturing scenarios, which solves the non-linear optimization problem for travel time differences in the seismic waves and results in a remarkable increase in the speed of calculations for microseismic source location. Hassani et al. [118] accurately located the microseismic source in a uranium mine executing the Kirchhoff Prestack Depth Migration approach, and the localization algorithm enhanced the location of microseismicities by employing the P-wave arrival time if an anisotropic 3-D velocity model was adopted. Huang et al. [119] proposed a new location method employing interferometric imaging and cross-wavelet transform to overcome the shortcomings of the cross-correlation calculation method and diminish location error, which provides higher source location accuracy compared to the traditional method.

With the development of deep learning methods, the massive microseismic data generated by coal mining can be applied for deep learning to improve the efficiency and location accuracy of microseismicity. Chen et al. [120] proposed a deep learning model of a convolution long short-term memory network to pick the first arrival time of microseismic waves in a coal mine, which directly improves the accuracy of microseismic localization. Zhu et al. [121] trained quality control models for P-phase arrival picks using a convolutional neural network and five other traditional deep-learning methods, and the result suggested that the convolutional neural network has the best performance for recognition of P-phase arrival picks. Furthermore, the automation of locating microseismicities in coal mines based on a dynamic, in-depth analysis model for microseismic signals and artificial intelligence will be a major trend in the future.

5.2. Efficient and Intelligent Processing of Microseismic Data

Microseismic monitoring not only records coal/rock fracture signals but also noise signals such as the blast, mechanical vibration, and current interference. Engineers have to identify and locate useful microseismic events among the huge volume of seismic data, making it difficult to assess the risk of mining-induced instabilities in real time and implement countermeasures in a timely manner, as data processing is time-consuming. On the other hand, the effectiveness of microseismic data and the associated risk assessment is highly dependent on the experience of the engineers, presenting challenges for promoting the microseismic monitoring technique in coal mines. Therefore, the efficient and intelligent processing of microseismic data is critically important for the development of the microseismic monitoring technique in the future.

Intelligent analysis of microseismic data will play a pivotal role in processing the vast seismic signals in coal mines and will avoid the high-intensity labor of manually identifying useful microseismic signals and the human factor in data-processing errors. Tang et al. [122] developed a novel depthwise spatial and channel attention module, and when integrated with a convolutional neural network, the module can emphasize useful information and mitigate noise interference. Additionally, the depthwise spatial and channel attention module and modified residual connections were embedded in a normal deep convolutional neural network, which improves the accuracy of microseismic waveform recognition in a deep tunnel. Li et al. [123] employed various deep learning models to automatically recognize and classify microseismic waveforms with high accuracy rates of no less than 0.96 based on computer vision, and the image learning configuration and flow chart are demonstrated in Figure 12. With further development of the neural network algorithm, these data-processing methods can be effectively applied in coal mines for intelligent monitoring.

5.3. Comprehensive Assessment of Mining-Induced Instabilities

The mechanisms of disasters in deep coal mines are complicated and influenced by many contributing factors, e.g., the stress regime, geological defects, excavation geometry, and rock mechanical properties. It is very challenging to monitor, analyze, and warn of mining-induced instabilities using a single research method such as theoretical analysis, a laboratory test, numerical modeling, or a field survey. Both theoretical analysis and numerical modeling have to simplify the complicated boundary conditions of deep excavations, and laboratory tests, limited by model dimensions, make it difficult to study the effect of geological defects. On the other hand, the accuracy of field monitoring data greatly influences disaster mitigation in coal mines. Therefore, an integrated approach that combines features and leverages the strengths of various methods is prospective for more accurate assessment and control of mining-induced instabilities.

The mechanical parameters of rocks deteriorate continuously during coal mining, while the mechanical parameters of fresh rock specimens are usually employed in numerical modeling to study the evolution of rock damage, which means the numerical results do not align well with actual scenarios. Zhao et al. [124] developed a damage model of rocks concerning the relationship between microseismic source parameters and releasable strain energy, which used microseismic source parameters (e.g., seismic energy, apparent volume, source dimension, and seismic efficiency) as inputs to update the mechanical parameters of rocks in a numerical model dynamically and analyze the evolution of surrounding rock damage in the Shirengou iron mine. Ma et al. [108] integrated the improved theoretical model considering confined aquifer pressure and RFPA numerical modeling to investigate the stress regime variation and floor failure process, and further estimated the floor failure depth using microseismic monitoring data. The flow chart for assessing the risk of the water inrush is shown in Figure 13.

The microseismic monitoring technique can continuously capture coal/rock fracturing at a wide range of scales and is commonly used in conjunction with other monitoring methods, e.g., the stress meter, extensometer, and borehole televiewer. It is difficult to interpret the stress and deformation field of coal/rock limited by the quantity of data from a few pre-defined monitoring points, which will affect the accuracy of assessment and early warning of mining-induced disasters based on microseismic monitoring and other monitoring data. Laser scanning can be used in underground mines for convergence monitoring, change detection, and deformation tracking [125], as it is able to acquire the deformation field of coal/rock mass. Mobile laser scanning, e.g., vehicle- and drone-mounted [126], is highly suitable for coal mining environments due to its ease of operation and the ability to quickly scan large areas. Therefore, microseismic monitoring and laser scanning are effective in detecting the microcracking field and deformation field of coal/rock, and they can be used in combination for a comprehensive assessment of mining-induced disasters.

5.4. Development of Microseismic Monitoring Equipment

The microseismic monitoring technique has been extensively utilized in Chinese coal mines, and the microseismic monitoring equipment is mainly imported from abroad. The SOS microseismic system from Poland, the IMS microseismic system from South Africa, and the ESG microseismic system from Canada are famous in Chinese coal mines. The core technology of the imported microseismic equipment is closed, which inevitably limits the user to embedding the latest research results into the microseismic configuration and improving the equipment according to new requirements under different mining conditions. The Hubei Seaquake Technology Co., Ltd. (Hubei, China), in collaboration with the Institute of Rock and Soil Mechanics, Chinese Academy of Sciences and Northeastern University, started to research and develop microseismic monitoring equipment for rock fracturing in 2008. A high-precision microseismic monitoring system named SinoSeism was successfully created and applied to various deep rock excavations [127]. The SinoSeism system uses a newly developed time arrival picker for seismic waves emitted from the stressed rocks and a sectional velocity model for positioning microseismicities.

Due to the complicated environment of underground coal mining, microseismic monitoring equipment characterized by a simple structure, high precision, high sensitivity, anti-interference, and a wide transmission frequency band is needed in terms of long-term stability, accurately detecting coal/rock fracturing, and the efficient transmission of the microseismic data with low signal attenuation in coal mines. The all-fiber microseismic monitoring system meets the abovementioned requirements, which comprises the hardware of microseismic sensors based on fiber optic interferometry and signal transmission fiber, and the software of intelligent data processing and analysis methods. The microseismic sensors based on fiber optic interferometry are sensitive, have a wide bandwidth and dynamic range, and feature great performances in complicated coal mining environments in comparison with the traditional piezoelectric sensors. In addition, they are cheaper than sensors imported from abroad. Therefore, the development of the all-fiber microseismic monitoring system is meaningful for the mitigation of mining-induced disasters and improvements in safety conditions during deep coal mining.

6. Conclusions

Deep coal mining-induced disasters, e.g., rockbursts, water inrush, coal and gas outbursts, and roof-fall accidents, seriously threaten the safety of miners, equipment, and underground structures. The study of the coal/rock fracturing process plays a pivotal role in the prediction and control of mining-induced disasters. The microseismic monitoring technique is commonly used for the assessment of coal/rock instabilities and safety management in many Chinese coal mines. This study reviews the achievements, development, and applications of the microseismic monitoring technique in coal mines, and certain conclusions are drawn:

• The microseismic monitoring technique is capable of capturing coal/rock fractures at different scales, which is challenging for the traditional stress and deformation monitoring methods. Microseismicities contain abundant coal/rock fracturing information, and the quantitative interpretation of mining-induced microseismicities contributes to studying the generation process of disasters in coal mines. Furthermore, the microseismic monitoring technique is effective for the assessment of coal/rock instabilities and early warning of disasters.
• Some source parameters and statistical parameters of microseismicities are effective and sensitive in evaluating the stress regime and characterizing the damage of surrounding rocks. These microseismic parameters can be used in conjunction with other monitoring methods, e.g., laser scanning and electromagnetic radiation, which are meaningful for the mitigation and control of mining-induced disasters.
• Though fruitful achievements have been made in coal mines using the microseismic monitoring technique, there are certain problems to be addressed. To better monitor, warn, and control the disasters in deep coal mines, future research should focus on a precise microseismic source location method, an efficient and intelligent data processing method, a comprehensive assessment method for mining-induced instabilities, and the development of new microseismic equipment.