Location error based seismic cluster analysis and its application to burst damage assessment in underground coal mines

https://doi.org/10.1016/j.ijrmms.2021.104784Get rights and content

Abstract

Coal bursts have emerged as the most critical mining hazard for underground coal mines around the world. Seismic monitoring and seismic clustering analysis are the cornerstones to develop an understanding of and to quantify coal burst hazards. The prerequisite to successfully detect seismic clustering behaviours is the accuracy of locating seismic events, which however can be a challenging task in underground coal mines. Therefore, the characteristics of location errors of seismic events and their impact on clustering results should be explicitly investigated and considered in seismic cluster analysis. Based on nine months of seismic data and 24 coal bursts from a longwall panel, this paper considers location errors and proposes a modified seismic clustering method to improve the results available for coal burst hazard assessment. The location errors in the area of interest were firstly assessed using the emulation testing method. Within the determined location errors, the clustering possibility between seismic events was calculated. The characteristics of the possible seismic clustering along with mining, named as “the Number of Possible Clustered Events” (NPCE), were investigated. The results showed that location errors presented large variations and strong anisotropic patterns in the longwall panel, with values ranging from 20 to more than 80 m. The NPCE result presented an improved detection on seismic clustering behaviour, and the high NPCE values also indicated a strong correlation with coal burst damages observed at the mine. Several intensive seismic clustering zones were observed in more than two months prior to the longwall passing these locations, which suggest that the method can be used for the medium to long term seismic hazard assessment.

Introduction

As a normal consequence of mining, mining-induced seismicity depicts the rock mass response to deformation and failure triggered by mining activities.1,2 A coal burst is a seismic event that causes consequent dynamic rock or coal failure in coal mines, resulting in a high-velocity ejection of the failed material into mine openings and posting a great threat to mine operators and equipment.3 As a typical seismic hazard, coal bursts and the induced damage increase the cost of mining, reduce mine productivity, and have a direct influence on workforce safety. Coal bursts have emerged as one of the most significant technical challenges for a large number of coal mines around the world, such as in Australia, US, China, and Poland.4, 5, 6, 7

Seismic monitoring provides a powerful means of detecting and evaluating mining-induced seismicity occurring around mine openings. It has become a standard practice to conduct seismic hazard management in burst-prone mines.8,9 Understanding the occurrence mechanism and the inner mechanisms of seismicity have become necessary for coal burst forecasting and control. Currently, several typical seismic hazard assessment methods have been developed using different characteristics of seismicity, including: seismic energy and frequency,10, 11, 12, 13, 14, 15, 16 source parameters,17 source mechanism,18, 19, 20 ground motions,21, 22, 23 velocity field reconstruction24, 25, 26, 27, 28 and seismic clustering methods.29, 30, 31, 32

Seismic cluster analysis is one of the most essential methods to evaluate the spatial distribution of mining-induced seismicity and represent various failure mechanisms.29 The cluster of seismic events is related to the mining method, mining sequence, and the geological and geo-mechanical environment of the mine. Therefore, seismic cluster analysis is an important approach to quantify the seismic hazards associated with mining.30 Multiple seismic clustering methods have been proposed and applied in underground mines. For instance, Falmagne31 proposed the cluster index function to define the weight of potential interaction between neighbouring events according to their distance and source radii. A density-based clustering algorithm modified from Density-based spatial clustering of applications with noise (DBSCAN) has been developed by Woodward et al.30 for spatial assessment of short-term seismicity in mines. Vasak et al.9 used the agglomerative hierarchical algorithm to cluster seismic data and applied principal component analysis (PCA) to identify potential seismically active planes from the clustered events. Hudyma29 developed the “comprehensive seismic event clustering methodology” (CSEC) involving two cycles of clustering to identify individual seismic sources and rock mass failure.

As a seismic cluster represents the characteristics of the spatial distribution of seismic events, the accuracy of locating seismic events is the prerequisite to successfully detect its clustering behaviour. However, the location results of seismic events can be highly ambiguous because of the flatness of the geophone array, complex geological structures and intense extraction in underground coal mines.33 Gibowicz and Kijko8 stated that the location error of seismic events is of the order of 20–50 m or 50–100 m, depending on the number of geophones and the network size and geometry. In the Upper Silesian Coal Basin (USCB) in Poland, the location error of horizontal coordinates is 50 m, while it can be up to 100 m for vertical coordinates.34 The epicentre error of monitoring systems in one Canadian mine and one South African mine ranges from 4 m to 61 m, with an average of 15 m–23 m35 respectively. Gong et al.36 optimised the number of geophones for seismic monitoring in underground mines and analysed the spatial location errors of a Chinese coal mine, which appeared to be approximately 50 m. Since the error sizes are non-negligible compared to the mining excavations, it is highly likely to cause false results for seismic clustering analysis if significant location errors are not considered. Therefore, how to evaluate the location errors during underground mining and eliminate their impacts on seismic clustering analyses have become an urgent problem that needs to be addressed.

During underground mining, the configuration and position of a seismic monitoring system are regularly updated, which gives rise to changes in location errors in the area of interest. The routine analysis of seismic hazards must account for these time-dependent variations. Similar to mining-induced seismicity, the performance of a seismic monitoring system is also unique because of different mining and geological conditions. Therefore, the industry-wide solutions for seismic analysis in mines tend to be highly generalised and nonspecific. Therefore, developing a detailed seismic analysis procedure for coal burst risk evaluation, using locally collected seismic data, could be a more practical means of risk management and burst control in coal mines.29

This paper proposes an enhanced seismic clustering method to reduce the influence of location errors and improve the accuracy of seismic cluster analysis in coal burst hazard assessment. To refine the method, nine months of seismic data and coal burst damage records from a burst-prone longwall panel in a Chinese coal mine are used. The location errors in the area of interest are assessed using the emulation testing method. The location error characteristics of the seismic monitoring system in different time periods are separately discussed. Based on the fracture sizes and the horizontal locations of seismic events, the seismic clustering criterion considering location errors is established. An index, named “Number of Possible Clustered Events” (NPCE), is introduced to investigate the seismic clustering distribution and evolution along with mining. The proposed method and the index are then back analysed against the monitored coal burst damage zones recorded in the mine.

Section snippets

Geological and mining conditions

The case study site, Huating Coal Mine, is a typical burst-prone coal mine in Gansu Province, China, where more than 200 coal bursts have been reported since 2008. The target coal seam is the No. 5 Seam, which is 550 m–800 m deep with a dip angle of between 1° and 15°. As the coal seam has an average thickness of about 40 m, the mine uses a multi-slice mining method to extract the seam in three 13 m thick layers using the longwall top coal caving (LTCC) method. In each slice, the first pass

Location error emulation test

The classic least-square method was used to locate the events.38 In a homogeneous-isotropic velocity model with a constant velocity, the arrival time observations of a seismic event from n geophones are t1, …, tn. Find the occurrence time and the location (x0,y0,z0) at any spatial point in the Cartesian coordinate to make the sum of squared time residuals riΦ(t0,x0,y0,z0)=i=1nri2as the minimum. ri equalsri=tit0Ti(x0,y0,z0)Ti(x0,y0,z0)=(x0xi)2+(y0yi)2+(z0zi)2vpwhere Ti(x0,y0,z0) is the

Source radius of seismic events

Seismic events are often spatially clustered. The clustered seismic events are evidence of the energy release of unstable fracturing in highly stressed rock mass and failure of geological structures.9,29,31,32,42 In laboratory tests, the onset of unstable crack growth and fracture interaction, represented by the clustering of acoustic emission signals, is usually at about 70%–80% of the uniaxial compression strength (UCS) of the rock sample.43,44 The strength of the rock mass is commonly far

Seismic cluster and coal burst damage

To study the relationship between seismic clusters and coal burst hazards, coal burst damage and the NPCE results in LW250105 during the study period were back analysed. NPCE analysis was conducted on a weekly basis, using the procedure shown in Fig. 10. Based on the seismic data of a week, a NPCE scatter plot was first derived by using the method in Section 4 (see Fig. 10a). Then, by using a linear interpolation algorithm with grid space of 50 m, the NPCE scatter plot was transformed to a

Conclusions

In underground coal mines, seismic clustering analysis is a powerful tool to assess coal burst risks. However, the low accuracy of seismic monitoring system in locating events can lead to false results of seismic clustering analysis, which significantly limits the hazard prediction performance. How to assess the location errors during longwall mining and eliminate their impacts on seismic clustering analyses have become an urgent issue that needs to be addressed. Therefore, this paper

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

The first author would like to acknowledge the financial support of the China Scholarship Council–UNSW, Sydney Project (no. 201706420062).

References (48)

  • T. Chen et al.

    In situ identification of high vertical stress areas in an underground coal mine panel using seismic refraction tomography

    Int J Coal Geol

    (2015)
  • G. Si et al.

    Seismic monitoring and analysis of excessive gas emissions in heterogeneous coal seams

    Int J Coal Geol

    (2015)
  • K. Woodward et al.

    A spatially focused clustering methodology for mining seismicity

    Eng Geol

    (2018)
  • A. Leśniak et al.
  • A. Kijko et al.

    Optimal spatial distribution of seismic stations in mines

    International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts

    (1995)
  • E. Eberhardt et al.

    Quantifying progressive pre-peak brittle fracture damage in rock during uniaxial compression

    Int J Rock Mech Min Sci

    (1999)
  • K. Larsson

    Seismicity in Mines: A Review

    (2004)
  • C. Mark

    Coal bursts in the deep longwall mines of the United States

    Int J Coal Sci Tech

    (2016)
  • S.J. Gibowicz et al.

    An Introduction to Mining Seismology

    (1994)
  • P. Vasak et al.

    Hazard map approach using space-time clustering analysis of mining-induced microseismicity

    Canadian Institute of Mining and Metallurgy Annual General Meeting, Edmonton

    (2004)
  • J. Wesseloo

    Evaluation of the spatial variation of b-value

    J S Afr Inst Min Metall

    (2014)
  • T. Urbancic et al.

    Space-time correlations of b values with stress release

    Pure Appl Geophys

    (1992)
  • C. Wang et al.

    A new method to assess coal burst risks using dynamic and static loading analysis

    Rock Mech Rock Eng

    (2019)
  • A. Cao et al.

    Microseismic precursory characteristics of rock burst hazard in mining areas near a large residual coal pillar: a case study from Xuzhuang Coal Mine, Xuzhou, China

    Rock Mech Rock Eng

    (2016)
  • Cited by (28)

    • Numerical modelling of coal and gas outburst initiation using energy balance principles

      2023, Fuel
      Citation Excerpt :

      Later, Fan, et al. [6] developed a fully coupled model using COMSOL Multiphysics by considering gas migration, stress and damage evolution based on the Mohr-Coulomb and maximum tensile stress failure criteria. In their study, the size of the damaged area was considered as the indication of the risk and intensity of an outburst event, and by definition, geological structures and mining disturbance were categorized as the internal and external drivers of outburst occurrence[6,7]. Zhao, et al. [8] also used COMSOL Multiphysics to couple the gas diffusion, gas flow in fractures, mechanical deformation, and damage evolution.

    • A theoretical goaf resistance model based on gas production analysis in goaf gas drainage

      2022, International Journal of Coal Geology
      Citation Excerpt :

      A similar coupled geomechanics and gas flow modelling approach was also applied in thick seam mines (Si et al., 2015a; Si et al., 2015b). Other indirectly methods using microseismic monitoring to infer mining-induced permeability changes were also explored (Zhao et al., 2019; Wang et al., 2021; Duan et al., 2021; Wang et al., 2021). Diamond et al. (1994) and Diamond (1995) reported high permeability near the edge of the longwall panel as the strata were supported by adjacent pillars, and goafholes near the edge produced 80% more gas than that near the centreline.

    • Experimental study on butterfly shape of failure zone and fractal characteristics of rock burst

      2022, Engineering Failure Analysis
      Citation Excerpt :

      The location result of the event can reflect the specific location of the failure [57], and the failure must have experienced the process of stress exceeding the rock yield strength, so it is reasonable to use it to reflect the shape of the failure zone. The accurate location of AE events usually depends on appropriate and accurate algorithms [58], and these algorithms mainly include cluster analysis algorithm [59], arrival time difference algorithm [60], machine learning algorithm [61], source scanning algorithm [62,63], etc. Most algorithms depend on the time gap between take-off points between different channels, so they have high requirements for accurate denoising method [64] and pick-up of take-off points.

    View all citing articles on Scopus
    View full text