Rockburst probability early warning method based on integrated infrared temperature and acoustic emission parameters

Abstract

Early warning of rockburst is a critical part of deep rock engineering; yet the selection of reliable indicators, the establishment of effective warning thresholds, and understanding of triggering mechanisms require further refinement. This paper presents a novel framework for early warning of rockburst, developed based on the integrated monitoring of infrared (IR) and acoustic emission (AE) data. Two distinct rockburst tests, simulating static-driven and dynamic-trigger scenarios, were conducted to obtain time series data of IR and AE parameters. The sliding window method was employed to segment the time series data, and statistical characteristics were extracted for each segment. Extreme differences in IR temperature and AE amplitude were identified as potential warning indicators. To enhance the lead time of early warning, three time series prediction models were applied to forecast the variation trends of the identified indicators. A comprehensive analysis of prediction error and generalization ability revealed that the Long Short-Term Memory (LSTM) model was the most suitable method for this prediction task. The cosine similarity was utilized to establish warning thresholds for each indicator, and a rockburst probability index was calculated using Bayesian theory. The effectiveness of the proposed framework was validated through laboratory-scale rockburst tests. The probability calculation revealed that the warning indicator exhibiting higher volatility, specifically AE amplitude, served as a more effective prior event.