Improving the Accuracy of Coal-Rock Dynamic Hazard Anomaly Detection Using a Dynamic Threshold and a Depth Auto-Coding Gaussian Hybrid Model

A coal-rock dynamic disaster is a rapid instability and failure process with effects huge disastrous consequences that occurs in mass under mining disturbance. Disasters lead to catastrophic consequences, such as mine collapse, equipment damage, casualties. Early detection can prevent the occurrence of disasters. However, due low accuracy anomaly detection, disasters still occur frequently. In order improve for disasters, this paper proposes an method based on threshold deep self-encoded Gaussian mixture model. First, pre-mining data were used initial threshold, subsequent continuously arriving flow dynamically update solve impact artificially setting accuracy. Second, feature dimensionality reduction reorganization carried out, low-dimensional representation reconstruction error modeling difficulty extracting features from high-dimensional real time. Finally, through end-to-end optimization calculation energy probability distribution between different categories problem key abnormal information may be lost was solved accurate monitoring realized. Experimental results showed has better performance than other methods.