Discrimination and location of seismic events for characterization of rock mass response near a dyke - RASIM 2022

Routine analyses of seismicity and event locations are essential for coal burst risk management, especially in the presence of unfavorable and complex geology. In an Australian longwall coal mine, a small-scale seismic monitoring network was installed to monitor a section of a major dyke intersected by the roadway, which was classified as a coal burst prone zone. The Deep Convolutional Neural Network (DCNN) technique was adopted for automatically discriminating seismic events to assess seismicity in this zone. The discriminated seismic events were located with the Shortest Path Method based on the Boundary Discretization Scheme (BDS-SPM) for temporal-spatial seismic event distribution analysis. The results demonstrate that the seismic events are discriminated with an average accuracy of 94%. Moreover, the seismic events are more accurately located using the BDS-SPM with a 3D heterogeneous velocity model compared with results obtained with a homogeneous velocity model. The temporal and spatial variances in seismic activities identify that the rock fractures in the dyke zone are closely related to the longwall mining process and a local dyke has a controlling influence on the stress concentration in its vicinity.