Cluster Analysis of Moment Tensor Solutions and its Application to Rockburst Risk Assessment in Underground Coal Mines

High-magnitude events (HMEs) are commonly observed in underground mines, and they can lead to violent rock failures, such as rockbursts. While moment tensors (MTs) have been widely used to investigate the triggering mechanism of HMEs, the cluster properties of MTs before HMEs have rarely been systematically studied. This hinders the application of MTs to predict HMEs and manage dynamic hazard risks. This study aims to characterize the cluster properties of MTs before HMEs and apply them to predict HMEs. A seismic clustering method suitable for the hybrid MT inversion was proposed to acquire reliable source mechanism solutions. Based on a case study, the MTs and decompositions were retrieved for seventeen HMEs and seismic events in the week preceding their occurrence. Cluster analysis results show that HMEs are close in location but still have entirely different source mechanisms. Pre-HME events show a distinct difference in strikes and dips in the cases dominated by roof movement or coal mass failure. The analysis of the cluster probabilities associated with source location errors, radius, and mechanisms was then conducted based on the seismic failure mechanisms; and a refined cluster possibility index, the number of possible clustered events (NPCE), was proposed to identify abnormal risk zones. A risk assessment index, namely the number of possible clustered events (I-NPCE), was proposed to evaluate the degree of a cluster of seismic events. A posteriori analysis of seventeen cases showed a prediction accuracy of up to 60% of HMEs, and a risk threshold of 0.4 was recommended in the case of acceptable completeness of seismic data. The low sensitivity of the monitoring system and the poor seismicity would be the two main limitations, and deep learning based automatic location and source mechanism inversion techniques could be used to improve the application.