Comparison of Microseismic Locations Scatter Reduction Methods and Unsupervised Machine Learning Algorithms for Automatic Detection of Seismogenic Structures in a Polish Hard Coal Mine. - RASIM2022
- Organization:
- Society for Mining, Metallurgy & Exploration
- Pages:
- 10
- File Size:
- 1216 KB
- Publication Date:
- Apr 26, 2022
Abstract
The earthquake locations are generally scattered mainly due to a simplified velocity model or seismic event picking errors. The aim of our research was to increase the resolution of the cloud of microseismic events recorded during mining activity in the Bobrek hard coal mine on the 3/503 coalface to delineate seismically active structures using unsupervised machine learning algorithms.
We presented results of applying two completely different algorithms for reducing scatter in microseismic location to all events recorded during mining activity on the coalface - the collapsing (proposed by Jones and Stewart in 1997) and condensation (proposed by Kamer in 2015). The collapsing method simplifies the seismic cloud by shifting events inside their error ellipsoids towards local source densities. The error ellipsoids are determined by errors in the measured parameters and seismic network configuration. The advantage of the collapsing method is the partial elimination of blurring of the microseismic cloud. The condensation method reduces the size of the set of seismic events without changing the geometrical information. The reduction of the number of seismic events in the catalog takes place through the condensation of localized events with larger uncertainty into events with lower uncertainty. The spatio-temporal analysis of microseismic events was performed separately on two processed clouds – collapsed and condensed one. We examined how collapsing and condensation methods affect the automatic detection of clusters of microseismicity using density-based spatial clustering of application with noise algorithms. We compared structures detected by the machine learning algorithms in both seismic clouds first created by collapsing and second by the condensation. We identified structures created by mining activity and structures related to the local tectonics activated by mining. The presented results are essential for the prediction of future tremors induced by the mining operation and tectonics and thus ensuring safety in the mine.
Citation
APA:
(2022) Comparison of Microseismic Locations Scatter Reduction Methods and Unsupervised Machine Learning Algorithms for Automatic Detection of Seismogenic Structures in a Polish Hard Coal Mine. - RASIM2022MLA: Comparison of Microseismic Locations Scatter Reduction Methods and Unsupervised Machine Learning Algorithms for Automatic Detection of Seismogenic Structures in a Polish Hard Coal Mine. - RASIM2022. Society for Mining, Metallurgy & Exploration, 2022.