Discrimination of Microseismic Events in Coal Mine Using Multifractal Method and Moment Tensor Inversion

Abstract

Discrimination of various microseismic (MS) events induced by blasting and mining in coal mines is significant for the evaluation and forecasting of rock bursts. In this paper, multifractal and moment tensor inversion methods were used to investigate the waveform characteristics and focal mechanisms of different MS events in a more quantitative way. The multifractal spectrum calculation results indicate that the three types of MS waveform have different distribution ranges in the multifractal parameters of ∆α and Δf(α). The results show that the blasting schemes also have a great influence on MS waveform characteristics. Consequently, the multifractal parameters of ∆α and Δf(α) can be used to discriminate different MS events. Further, the focal mechanisms of MS events were calculated by seismic moment tensor inversion. The results show that an explosion is not the dominant mechanism of deep-hole blasting MS events, and the CLVD and DC components account for an important proportion, indicating that some additional processes occur during blasting. Moreover, the coal-rock fracture MS events are characterized by compression implosion or compression/shear implosion mixed focal mechanisms, while the overburden movement MS events are tensile explosion or tensile/shear explosion mixed focal mechanisms. The focal mechanisms and nodal plane parameters have close relationships with the inducing factors and occurrence processes of MS events.

1. Introduction

With excavation activities moving to deeper strata in coal mines, the dynamic disaster of rock bursts is becoming increasingly complicated subjected to high in situ stress and strong disturbance stress. This poses a serious threat to mine production and miners’ safety [1,2]. Rock burst disasters are closely related to stress concentration and cumulative damage of coal and rock mass [3]. As a long-distance, real-time, dynamic, automated monitoring technology, microseismic (MS) monitoring is becoming an indispensable part of the evaluation system of dynamic disasters in coal mines [4,5]. It can record seismic waves emitted from rock micro-fractures and then provide source information of MS events, such as occurrence time, location and energy. The MS information is valuable for engineers to understand the stress state and mechanical behavior of rock mass [6,7]. A pure MS database is the prerequisite for the reliable evaluation and forecasting of rock burst disasters. However, deep-hole blasting technology has commonly been adopted for rock burst prevention in burst-prone coal mines, and blasting MS events occur frequently and are easily confused with mine MS events. Blasting MS events should be accurately discriminated and removed from the MS database. Otherwise, a confused MS database may lead to an inaccurate assessment of rock burst risk, such as a fictitious high-stress zone. Therefore, the discrimination of mining and blasting MS events in coal mines is a primary and challenging task.

In recent years, the problem of classifying seismic events has garnered a lot of research attention, and various methods have been proposed to discriminate seismic events from man-made blasting. The widely used methods consist of two steps: (1) extraction of seismic event characteristics and (2) classification using identifying models. Among them, the extraction of characteristic parameters of different seismic events is the primary task, which largely influences the discrimination accuracy of seismic events. At present, the extraction of MS event characteristics is mainly based on source parameters, waveform characteristics and spectrum characteristics, such as occurrence time, source energy, seismic moment, corner frequency, stress drop, seismic phases, time-frequency images and focal mechanism parameters. For instance, based on thirteen waveform characteristic parameters, Vallejos and Mckinnon [8] adopted two models to discriminate mine MS events and blasts, and both achieved good discrimination accuracy. However, the logistic regression method achieves a better generalization performance when more characteristic parameters are considered. Dong et al. [9,10,11] conducted a systematic study on the discrimination of blasts and MS events based on the waveform and seismic source parameters. Zhao et al. [12] extracted six waveform starting-up characteristic parameters to classify the blasts and mine MS events, including the first and maximum peak coordinates in the waveform and the corresponding two slope values. Rao et al. [13] first selected 22 seismic parameters as the discrimination feature parameters and conducted correlation analysis, then developed an artificial intelligence model consisting of an extreme learning machine and particle swarm optimization algorithm for discriminating MS events and blasting in the Fankou lead–zinc mine. Pu et al. [14] selected six characteristic parameters based on source parameters and waveform characteristics to discriminate the mine and blasting MS events and discussed the performance of ten recognition models. The results show that the logistic regression model has the best recognition performance, while the Gaussian process model performs worst. The above results suggest that the extraction of MS event characteristic parameters is of significance for the discrimination of mine and blasting MS events. Moreover, image recognition technology was applied to the recognition of mine and blasting MS events. Zhao et al. [15] proposed a hybrid model to recognize MS signals based on time-frequency RGB images of MS signals and discussed the performance of the proposed model. Wei et al. [16] proposed a waveform image method for discriminating MS events and blasts in underground mines. Song et al. [17] converted the MS signals into a two-dimensional spectrum image using Stockwell Transform and then recognized blasts from MS events based on a convolutional neural network.

The source rupture scale and impact area of MS events in coal mines are relatively small, and the underground geological conditions are complex, which makes the discrimination of mine and blasting MS events more challenging. Li et al. [18] investigated the time-frequency and non-linear characteristics of blasting and mine MS events in the Jixian coal mine using FFT, SPWVD and multifractal methods. Hilbert–Huang transform was used to analyze the time-frequency characteristics of mine MS events, blasting MS events, hydraulic fracturing MS events and natural earthquakes [19,20,21]. Ma et al. [22] proposed two discrimination methods based on waveform characteristics and source parameters, respectively, and discussed the discrimination accuracy of the two methods. Chen et al. [23] proposed an automatic identification model of micro-earthquakes and blasting events based on source rupture properties including corner frequency, source radius and stress drop. Ma et al. [24] and Simone Cesca et al. [25] studied the discrimination of induced seismicity by full moment tensor inversion and decomposition. Łukasz Wojtecki et al. [26] adopted the seismic moment tensor inversion method to analyze the focal mechanism of deep-hole blasting MS events. However, mine MS events can be divided into various types according to the failure mechanism of the coal and rock, and different MS events have different impacts on rock bursts [27]. The above studies did not further classify the types of MS events or conduct comparative analysis on their waveform and source characteristics. Moreover, existing studies have attached importance to the extraction of MS event characteristic parameters, while the triggering of MS events and the internal relationship between the source rupture model and the mining process were neglected.

Given the above, deep-hole blasting MS events and different mine MS events (coal-rock fracture MS and overburden movement MS) were selected as research objects in this study. First, the multifractal method was used to investigate the waveform characteristics. Based on this, multifractal parameters were used to quantitatively discriminate the above three types of MS events. At the same time, the influence of different types of MS events on rock burst risk was discussed. Then, the focal mechanism and parameters of different types of MS events were further investigated using the moment tensor inversion. Further, the triggering of MS events and the internal relationship between the source rupture model and the mining process were discussed in detail. The research results are meaningful for discriminating deep-hole blasting MS events from mine MS events and may help engineers further understand the rupture mechanisms of different MS events.

2. Site Characterization

2.1. Geological and Mining Conditions of the Coal Mine

The study area is in the northwest of the Binchang mining area in Shaanxi Province, China. In this coal mine, the No. 4 coal seam with strong burst liability is mainly mined, which has a thickness of 6 to 15 m and a maximum depth of over 980 m. As shown in Figure 1a, the fully mechanized caving technology and longwall retreat mining method were adopted in the working face. By December 2021, in the No. 3 mining district, longwall panel (LW) 301 was being mined and LW 302 had been mined. In the southeast of LW 301, LW 204 and 205 in the No. 2 mining district have been mined, and there is a panel coal pillar with a width of 200 m between the No. 2 and No. 3 mining districts. Moreover, there is a large syncline, named X3, in parallel with the strike of the ventilation roadway of LW 301. Influenced by great buried depth and syncline, the maximum principal stress in this area reaches 44.8 MPa. The stratigraphic column of bore 31-2 is demonstrated in Figure 1b. As can be seen, three layers of thick and hard sandstone cover the No. 4 coal seam, including 10.79 m of fine sandstone, 13.03 m of medium sandstone and 20.08 m of coarse sandstone, and the rock strata in the roof all have weak burst liability.

2.2. MS Monitoring System

Since December 2017, an MS monitoring system called “SOS”, developed by the Central Mining Institute of Poland, Katowice, Poland, has been installed and applied in the coal mine. It can realize long-distance (maximum 10 km), real-time, dynamic and automatic monitoring of MS events and record the complete waveform of MS signals. This system mainly consists of a certain number of underground MS sensors, two ground signal acquisition stations, one ground signal recorder and a data processing system. The underground MS sensors and ground signal acquisition stations work together to record and transmit MS signals via electric transmission lines, and the MS signals are converted to digital form by the AS-1 signal recorder. Then, a series of software tools complete useful signal extraction, visualization and analysis to calculate MS source parameters (three-dimensional position and energy). Generally, the horizontal and vertical locating errors are within 20 and 30 m, respectively. The composition of the “SOS” system installed in the coal mine is shown in Figure 2. In total, four MS sensors, which will be moved as the working face advances, were installed in the haulage and ventilation roadway of the 301 panel, and eight fixed MS sensors were installed in the main entries.

2.3. Data Preparation and Analysis

In coal mines, the inducing factors of MS events can be divided into two types, including mining activities and human production operations. As shown in Figure 3a, micro-cracks would be generated in the coal and rock at the mining boundary under advanced abutment stress. Moreover, the overburden in the goaf will collapse, and the faults in the stope may slip under the mining stress. These mining activities occur with the release and propagation of elastic strain energy in the form of elastic waves, which are known as mine MS events. Pan et al. [28] proposed that the mine MS events can be further divided into three types of coal-rock fracture MS, overburden movement MS and fault dislocation MS based on MS source type. In addition, some human production operations in coal mines also induce MS events, such as destress measures for rock burst prevention. Figure 3b shows the commonly used destress measures for rock burst prevention in coal mines, including borehole destress, coal destress blasting, floor destress blasting, hydraulic fracturing and roof deep-hole blasting.

According to the geological and mining conditions of the coal mine, threatened by the deep cover, fold structure and thick hard roof, an extremely high risk of rock burst is present during the excavation of working faces. Therefore, multiple destress measures have been taken in working faces for the prevention of rock burst, such as destress blasting and drilling boreholes. Seismic waves generated by rock blasting and drilling can propagate in the rock medium and be recorded by the underground MS sensors. When destress measures are implemented, the operators will immediately inform the technicians who operate the monitoring system about the location, time and type of destress measures. Additionally, the corresponding MS events will be marked and saved in the database.

Figure 4 shows the spatial distribution and energy of mine and deep-hole blasting MS events during LW 301 mining from November to December 2021 in the coal mine. The source energy of the mine MS event generally varies from 1.0 × 10² to 1.0 × 10⁶ J, while that of deep-hole blasting MS events is 1.0 × 10⁴ to 1.0 × 10⁵ J. He et al. [29] pointed out that mine MS events provide very valuable information for rock burst prediction and early warning, especially for mine MS events with energy greater than 1.0 × 10⁴ J. However, deep-hole blasting and mine MS events have similar source information (location and energy) and are difficult to discriminate. The fact that the mine MS database is contaminated by the deep-hole blasting MS events with high source energy may mislead the risk assessment of rock bursts. Therefore, a challenge in MS monitoring is to discriminate deep-hole blasting MS from mine MS databases.

Existing studies have illustrated that different types of MS events have different focal mechanisms and propagation characteristics. The MS waveform is the carrier of the information of hypocenter and propagation medium property, and MS events can be discriminated by analyzing waveform characteristics and extracting information. Therefore, this study aimed to discriminate different mine and deep-hole blasting MS events based on waveform characteristics and focal mechanisms of MS events. For this purpose, this study first selected four typical waveforms of deep-hole blasting MS events (named DBMS-1 to DBMS-4) and eight typical waveforms of mine MS events (four coal-rock fracture MS events, named CFMS-1 to CFMS-4, and four overburden movement MS events, named OMMS-1 to OMMS-4) as research objects to analyze the waveform characteristics difference, as shown in Figure 5.

Different MS events show significant differences in waveform parameters, including duration, dominant frequency and vibration amplitude. First, the average duration of the deep-hole blasting MS waveforms is 1158 ms, which is far longer than that of mine MS waveforms (825 ms of coal-rock fracture MS waveforms and 598 ms of overburden movement MS waveforms). During deep-hole blasting, the explosive in one borehole was triggered by two detonating cords connected to the bottom and top of the charge section, respectively, resulting in two independent vibration waves. Then, influenced by the propagation distance and medium, the two vibration waves were received successively by MS sensors, and there was a time difference. Therefore, two obvious fluctuations with similar amplitude can be clearly observed in the DBMS-1 to DBMS-3 waveforms. Particularly, the DBMS-4 MS event was induced by double hole blasting; the superposition of vibrations generated by four detonation points results in the large amplitude and complex propagation of the waveform, and more than two evident fluctuations can be observed in the waveform. Moreover, with the increase in MS source energy, the boundary between two obvious fluctuations in the waveforms becomes blurred. The results indicate that the blasting intensity and scheme have a great influence on the waveform characteristics. Then, except for the DBMS-4 waveform, the dominant frequency of deep-hole blasting MS waveforms was generally 36 Hz, lower than that of mine MS waveforms (68 Hz). For mine MS waveforms, only one evident fluctuation can be clearly observed, which can be divided into three segments, namely pre-peak, peak duration and post-peak. Among them, compared with overburden movement MS waveforms, the pre-peak segment of the coal-rock fracture MS waveform was shorter, while the peak duration was longer, and the coda wave was more prominent. The results indicate that the coal-rock fracture MS events are characterized by strong intensity and slow propagation attenuation. The main reason is that many macroscopic cracks exist in the overburden of the goaf, and the structural integrity of the rock mass is poor. Reflection, refraction and diffraction phenomena occur in the propagation process of vibration waves, resulting in fast vibration attenuation. However, the coal-rock fracture MS events are generally located in the unmined area of the advanced working face, and the vibration propagation medium has better integrity, leading to slow vibration attenuation. Moreover, the source energy also has a great influence on the waveform characteristics. With the increase in source energy of MS events, the peak duration increases, and the coda becomes more developed.

In summary, different types of MS events differ from each other in waveform fluctuation due to different source types, source energies and propagation paths. However, the above conclusion is a more qualitative description, and the wave fluctuation characteristic parameters of MS waveforms need to be further extracted for discriminating different MS events.

3. Multifractal Parameters of MS Waveform

3.1. Multifractal Method

Fractal theory is a mathematical method to describe the local-scale and global self-similarity of objective things. Fractals are usually divided into two categories, one is monofractal, which can be fully described by fractal dimension D. However, for non-linear and nonstationary signals, monofractal is not sufficient to describe the local singularity feature. The other is the multifractal, which needs to be described by the multifractal spectrum F(α)-α. The fractal theory has been widely used in biological, financial, transportation, mining and seismic fields. For example, Fan et al. [30] distinguished the different heart rate variability signals using the multifractal method. Laudani et al. [31] investigated the multifractal property of the streamwise velocity spectra based on generalized Cauchy and Dagum models. Moreover, they studied the fractal and Hurst effect of the fracture of beams with random field properties [32]. Kantelhardt et al. [33] put forward the multifractal detrended fluctuation analysis method (MF-DFA) to analyze the multifractal property of signals. Xu et al. [34] and Fu et al. [35] illustrated the superior applicability of the MF-DFA method in the analysis of mine and blasting MS signals. Therefore, this research used the MF-DFA method to analyze the fluctuation characteristics of different mine MS signals and deep-hole blasting MS signals.

For one-dimensional time series of MS signal x_k with length N, the calculation procedure of the MF-DFA method was given as a flow chart in Figure 6. In total, seven steps were successively performed to calculate the multifractal spectrum of MS signals [36]. To be specific, a cumulative sum sequence Y(_i) was constructed first by removing the mean value from signal x_k. Subsequently, 2N_s segments were obtained by dividing the sequence Y(_i) and its reverse sequence into segments with length s, which can make full use of sequence data. Then, the least-squares fit was adopted to local trends on each segment, and the variance was calculated. The average value of the fluctuation function with q order was calculated. If there is an exponential relationship between q order fluctuation function F_q(s) and time scale s, it indicates that the signal has self-similar characteristics. The logarithmic transformation was performed for F_q(s) and s, and the slope of curve log F_q(s)-log s is the q order generalized Hurst exponent h(_q). When h(q) changes with q, the time series has multifractal characteristics, and then the multifractal spectrum can be calculated using the formula in Step 7.

3.2. Key Parameters Setting of MF-DFA Method

The key parameters of the multifractal category mainly include signal length N, fitting order m, time scale s and weight factor q. In this research, the signal series should intercept effective vibration segments of the MS waveform to eliminate the side effects of interference noise, and the signal length N should be kept uniform to discriminate the fractal characteristics of different types of MS signals. According to the effective vibration duration of signals, it was set as 1400 ms. For fitting order m, Mao et al. [37] believed a value of 3 was more reasonable for analyzing the MS signals, which could consider both calculation accuracy and time. The basic principles of determining time scale s need to be followed, the minimum segment size must be much larger than the fitting order m to avoid excessive fitting, while the maximum segment size should be small enough to provide a sufficient number of segments. Therefore, the minimum and maximum time scale s values were preliminarily determined to be 16 and 256, respectively. The choice of weight factor q should avoid large negative and positive values to reduce the numerical errors in the tails of the multifractal spectrum, and it ranged from −5 to 5 in this study. In order to verify that MS signals have the multifractal characteristics and it is feasible to analyze MS signals using the MF-DFA method, based on the above parameters, a time series of the mine MS signal was taken as an example, and the logarithmic fitting trend of q-order fluctuation function F_q(s) and time scale s was obtained first, as shown in Figure 7a. The results show that a good power-law relationship occurs in the F_q(s) and s, which illustrates that the MS signal has scale invariance and fractal characteristics. Moreover, the q-order Hurst exponent is also given in Figure 7b, and h(q) presents a decreasing trend with the increase in q. This means that the MS signal has multifractal characteristics. Then, the multifractal spectrum can be further calculated based on the q-order Hurst exponent using the formulas in Step 7 of Figure 6, and the result is given in the following section.

3.3. Results

As described above, MS signals were verified to have multifractal characteristics. Then, the multifractal spectrums of the selected waveforms were obtained using MF-DFA methods, as shown in Figure 8. Two crucial parameters of ∆α and ∆f(α) in multifractal spectrums were calculated. Spectral width ∆α (∆α = α_max − α_min) describes the distribution uniformity of the waveform fluctuation, and the higher the value is, the more severe the local fluctuation of the waveform; otherwise, it indicates that the waveform fluctuates stably. Another parameter ∆f(α), calculated by the formula ∆f(α) = f(α_max) − f(α_min), was used to measure the proportion of large and small peaks in the MS waveform. To be specific, when ∆f(α) < 0, the large peaks have prominent advantages in the waveform, and when ∆f(α) > 0, the small peaks occupy a large proportion of the waveform [19].

As can be seen from Figure 8, the spectrum width Δα of the deep-hole blasting MS waveform (DBMS-1 to DBMS-3) was generally lower than that of the mine MS waveform, indicating that the local fluctuation degree of deep-hole blasting waveforms was relatively moderate. In particular, the DBMS-4 waveform was induced by double hole blasting, the vibrations were superimposed on each other in the propagation process, and the vibration intensity was large. As a result, the waveform was more complex, and the spectrum width Δα was significantly higher than that induced by single hole blasting. Compared with the mine MS, the deep-hole blasting MS waveform has a longer duration, but its attenuation was regular, and its wave coda was not developed. The mine MS waveform had clear pre-peak and post-peak segments, and the local fluctuation was strong, indicating that the spectrum width Δα of the mine MS waveform was large. Further, there is no discernible difference in spectrum width Δα between coal-rock fracture and overburden movement MS waveforms. Moreover, Li et al. [18] pointed out that the parameter Δα is most closely related to the dominant frequency of the waveform; to be specific, the higher the dominant frequency, the higher Δα. In this research, the dominant frequency of the mine MS waveform is higher than that of the deep-hole blasting MS waveform. Correspondingly, the parameter Δα of the mine MS waveform is higher, which is consistent with their conclusions.

The multifractal spectrums of deep-hole blasting and mine MS waveforms were leftward hook-like (Δf(α) < 0), demonstrating that large fluctuations have a prominent advantage in the waveforms. The main reason for this was that the selected waveforms were received by the MS sensor nearest to the source, and the vibration amplitude was large. Further, the Δf(α) values of the deep-hole blasting and overburden movement MS waveform were similar, ranging from −0.46 to −0.14. The Δf(α) of the DBMS-4 waveform was -0.69, which was different from other deep-hole blasting MS waveforms, and the reason for this was explained in the above analysis. The parameter Δf(α) of the coal-rock fracture MS waveform was the smallest (Δf(α) = −0.61 to −0.40), and the results show that the proportion of large peaks in the coal-rock fracture MS waveform was higher than the other two types of MS waveforms. The reasons for this are as follows: First, compared with mine MS, the deep-hole blasting MS waveforms have a smaller dominant frequency and faster attenuation, and small peaks occupy a considerable proportion, which is related to the focal mechanism of deep-hole blasting MS events. Further, the selected coal-rock fracture MS waveforms were mainly distributed in the advanced area of the working face. Compared with the goaf with many macroscopic fractures, the structural integrity of coal and rock in this area was better. As a result, the attenuation of coal-rock fracture MS waveforms was smaller, and the vibration intensity was stronger. This was one of the key reasons why the coal-rock fracture MS waveform is more likely to induce a rock burst disaster.

To summarize, different types of MS waveforms present different multifractal parameters. Fifteen waveforms of each type were randomly selected for further multifractal parameter statistical analysis, as shown in Figure 9. To summarize, the parameter ∆α of the deep-hole blasting MS waveform was generally less than 1.57, while that of the mine MS waveform was opposite—it was greater than 1.57. Therefore, parameter ∆α can be used to discriminate deep-hole blasting and mine MS waveforms. Moreover, the parameter Δf(α) of coal-rock fracture MS waveforms varies from −0.75 to −0.40, which is lower than that of overburden movement MS waveforms (from −0.43 to −0.14). The parameter Δf(α) can be used to discriminate the coal-rock fracture and overburden movement MS waveforms. Therefore, it is feasible to discriminate different types of MS events using multifractal parameters. In particular, there are three MS waveforms induced by double hole blasting among the fifteen deep-hole blasting MS waveforms, and they have considerable large ∆α (greater than 2.43) and small Δf(α) (less than −0.69). The multifractal parameters can also be used to discriminate deep-hole blasting MS waveforms with different blasting schemes.

4. Focal Mechanism of Different Types of MS Events

4.1. Seismic Moment Tensor Inversion

The waveform characteristics were generally influenced by the focal mechanism, source intensity and propagation medium properties [38]. Understanding the failure mechanism of MS events was of great significance for the discrimination of different types of MS events. The moment tensor was used to describe equivalent forces acting at a seismic source; therefore, seismic moment tensor inversion was considered a powerful tool to study the rupture behavior of the seismic source and was widely applied for calculating the focal mechanism of natural earthquakes and induced seismic events. The seismic moment tensor was represented by nine couples of equivalent dipole forces M_ij acting at the seismic source and can be expressed as:

$$M=\left(\begin{array}{ccc}{M}_{11}& M_{12}& M_{13}\\ M_{21}& M_{22}& M_{23}\\ M_{31}& M_{32}& M_{33}\end{array}\right)$$

The far-field displacement caused by a seismic source can be described as a convolution of the moment tensor and Green’s functions, and the mathematical expression was listed as follows [39]:

$$u_k(x,t)=M_{ij}\ast \frac{\partial G_{ki}}{\partial x_j}=M_{ij}\ast G_{ki,j}$$

where G_ki is Green’s function, M_ij is moment tensor components of the force couples acting along the x_i axis with an arm on the x_j axis, and the symbol “*” represents convolution.

The seismic source fracture type cannot be obtained directly from the moment tensor, and the moment tensor needs to be decomposed reasonably to establish the relationship between the moment tensor and seismic source fracture type. The moment tensor decomposition method proposed by Knopoff and Randall has been widely accepted [40]. This method decomposed the moment tensors into three components: the isotropic (ISO), linear compensated vector dipole (CLVD) and double couple (DC) parts. The ISO part denotes volume changes in the seismic source, that is, explosion (“+”) and implosion (“−”). The CLVD part was used to describe the tensile crack (“+”) or compressive crack (“−”), and the DC part describes the shear mechanism. As shown in Figure 10, the corresponding relationship between seismic source type and moment tensor component is displayed in a diamond plot, and the color intensity denotes the DC component. When the seismic source was located at the origin of coordinates, the focal mechanism of the seismic source was pure shear. When the seismic source was located at the top or bottom of the diamond, it was explosion and implosion, respectively. Pure tensile and compressive cracks were located at the boundary of the diamond.

In this research, focal mechanisms of three types of MS events were calculated using the moment tensor inversion, and the seismic moment tensor was calculated using FOCI software [42]. These calculations were based on the P-wave initial motion inversion theory. After the decomposition of the seismic moment tensor, the percentage of each moment tensor component was calculated according to Vavryčuk [43]. The focal mechanism was the shear failure when the DC component exceeds 50%. If the ISO component was dominant, the focal mechanism was assumed as the implosion or explosion. The seismic focal mechanism solution was usually represented by a beachball diagram. Moreover, two node-plane parameter information can be obtained, including the strike angle Φ, dip angle δ and rake angle λ.

4.2. Results

The seismic moment tensor inversion was performed for the MS events of each type selected in the previous section. The solution is shown in Figure 11, the percentage of individual components (ISO, CLVD and DC) is given in a form of a diamond plot and the parameters of nodal plane A in a form of a polar diagram. The spatial location and beachball diagram of selected MS events are shown in Figure 12.

For deep-hole blasting MS events, the percentages of ISO, CLVD and DC components are well-balanced, indicating that an explosion was not the dominant mechanism at the source but a mixed focal mechanism of the explosion, tension and shear. As mentioned above, deep-hole blasting was conducted in a high-stress environment, and accumulated energy in the rock mass was released accompanied by an explosion. The share of CLVD and DC components reflects the release of accumulated energy in the rock mass. For this reason, the DC component percentage can be considered as a possible index for destress effectiveness evaluation, and researchers have carried out relevant studies on this. Moreover, the nodal plane strike angle Φ of the deep-hole blasting MS event was about 150° or 300°, and the plane dip angle δ varied from 45° to 75°. According to the mining direction of the longwall panel shown in Figure 1, it can be concluded that the source fracture plane was perpendicular to the mining direction of the longwall panel and parallel to the direction of the blasting hole. The results show that the focal mechanism correlates well with the blasting process.

For the mine MS events, the CLVD component is the dominant mechanism at the source, while the DC component has a lower percentage. To be specific, among the thirty mine MS events, only four (two each of coal-rock fracture and overburden movement MS events) have a DC component percentage above 40%. However, 80% of the mine MS events have a CLVD component percentage above 40%. Moreover, most coal-rock fracture MS events have negative ISO and CLVD components, while the overburden movement MS events have an opposite tendency. These results indicate that the two types of mine MS events have different focal mechanisms. Further, the overburden movement MS events are mainly tensile explosion or tensile/shear explosion mixed focal mechanisms, while the coal-rock fracture MS events are mainly compression implosion or compression/shear implosion focal mechanisms. Moreover, the nodal plane strike angle Φ of coal-rock fracture MS event varies from 60° to 75° or 210° to 270°, and the dip angle varies from 60° to 90°. The source fracture plane is approximately parallel to the mining direction of the longwall panel. As described in Section 2.1, the maximum horizontal stress in the mine MS distribution area reaches 44.8 MPa, and the direction angle is from 210° to 240°. The strike angle of the fracture plane was consistent with the direction of maximum principal stress. Therefore, the coal-rock fracture MS events were mainly induced by compression failure of the coal and rock mass subjected to horizontal stress.

However, the nodal plane strike angle of the overburden movement MS event varied from 30° to 270°, and there was no dominant direction. For instance, for the MS events distributed in the central area of the goaf (OMMS-4, 9, 11, 13 and 15), the strike angle of the fracture plane varied from 120° to 180°, which was perpendicular to the direction of the longwall panel. For the MS events at the boundary of the goaf (OMMS-3, 5, 7 and 10), the strike angle ranged from 210° to 270° or 90 to 120°; in other words, it is parallel to the direction of the longwall panel. According to the O-X fracture characteristics of overburden in the goaf (see in Figure 12) [44,45], the fracture plane strike of overburden on both sides of the goaf boundary is parallel to the mining direction, while it is perpendicular to the mining direction in the middle of the goaf. This suggests that the nodal plane strike of the MS event at different locations in the goaf is consistent with the fracture plane strike of overburden at the corresponding position.

To summarize, the focal mechanisms and parameters of MS events are different due to the different fracture modes of coal and rock. The percentage of individual components of ISO, CLVD and DC and nodal plane parameters can be considered as characteristic parameters for discriminating three types of MS events in this research.

5. Conclusions

In this paper, three types of MS events collected in a coal mine were the research object, and multifractal and moment tensor inversion methods were used to investigate the waveform parameters and focal mechanisms of the MS events. Based upon the results, the following conclusions are drawn:

(1)

There are great differences in duration, dominant frequency and wave fluctuation features between different MS waveforms. In general, two obvious fluctuations with similar amplitude can be clearly observed in deep-hole blasting MS waveforms, and they have a longer duration, lower dominant frequency and regular fluctuation compared to mine MS waveforms. In particular, the MS waveforms induced by double hole blasting have more than two strong fluctuations and higher frequency. For mine MS waveforms, only one evident fluctuation with three segments of pre-peak, peak duration and post-peak can be clearly observed, and the amplitude reaches a peak slowly, and its coda wave is developed. Moreover, compared with the overburden movement MS waveform, the pre-peak segment of the coal-rock fracture MS waveform is shorter, while the peak duration is longer, and the coda wave is more prominent.

(2)

The multifractal parameters ∆α and ∆f(α) of different MS waveforms were calculated to quantitatively describe the waveform fluctuation characteristics. First, the multifractal parameter ∆α of the deep-hole blasting MS waveform was generally less than 1.57, while that of the mine MS waveform was opposite—it was greater than 1.57. Consequently, parameter ∆α can be used to discriminate the deep-hole blasting and mine MS waveforms. Then, the parameter Δf(α) of the coal-rock fracture MS waveform varies from −0.75 to −0.40, which is lower than that of deep-hole blasting and overburden movement MS waveforms (range from −0.46 to −0.14). Therefore, Δf(α) can be used to discriminate the coal-rock fracture and overburden movement MS waveforms. Especially, the double deep-hole blasting MS waveforms have considerable large ∆α (greater than 2.43) and low Δf(α) (less than −0.69), which is different from the above types of MS waveforms.

(3)

The moment tensor inversion results indicate that the three types of MS events differ from each other in the focal mechanisms and parameters due to the different fracture modes of coal and rock. For deep-hole blasting MS events, an explosion was not the dominant mechanism, but the CLVD and DC components account for an important proportion. This indicated that some other processes occur during blasting. The moment tensor inversion solution of the mine MS events showed that the CLVD component is the dominant mechanism at the source, while the DC component has a lower percentage. The coal-rock fracture MS events were characterized by compression implosion or compression/shear implosion mixed focal mechanisms, while the overburden movement MS events were tensile explosion or tensile/shear explosion mixed focal mechanisms. The focal mechanisms and nodal plane parameters have close relationships with the inducing factors and occurrence processes of MS events. Moreover, the percentages of individual components of ISO, CLVD and DC and nodal plane parameters can be considered as characteristic parameters for discriminating the three types of MS events.