A CNN–LSTM and PSO–GRNN Prediction of the Rockburst Risk Level Based on the Data from the Combination of Microseismic Monitoring Variables with Electromagnetic Radiation Signals

Rockburst disaster is a kind of typical dynamic disaster phenomenon. In this work, the risk level of the rockburst disaster was predicted based on the convolutional neural networks and long short-term memory (CNN–LSTM), and particle swarm optimization and general regression neural network (PSO–GRNN) models. A CNN–LSTM deep learning model based on rockburst chaotic time series was proposed to predict the characteristic variables of rockburst state, with a method to quantitatively distinguish and predict the risk level of the rockburst disaster in the future, and thus the dynamic prediction of the rockburst activity was realized. As an example, the microseismic monitoring variables (i.e., indexes of the daily cumulative microseismic energies and daily maximum microseismic energy, angular frequency and concave-convex radius) and electromagnetic radiation signals (i.e., indexes of the daily average amplitude and daily maximum pulse) were used to predict the rockburst. The CNN–LSTM and PSO–GRNN models were confirmed to be the most suitable to predict the risk level of the rockburst. This work provides an important basis for timely mastering the future state of rockburst activities.