ABSTRACT
In order to effectively predict the time and energy of the loading failure process of brittle rock, uniaxial compression acoustic emission tests were conducted on two kinds of small-sized brittle rocks, granite and red sandstone. Based on the empirical statistical method, a prediction model for predicting rock failure time and energy is established by using acoustic emission count data and stress-strain curve. Based on the machine learning method, the Long Short-Term Memory (LSTM) neural network is constructed. The acoustic emission amplitude data is used as input to predict the amplitude signal at the future moment and determine the rock failure time. The prediction model and LSTM are used to predict the failure time and energy of large-scale granite and sandstone. Compared with the actual results, the error of failure time predicted by empirical method and machine learning method is about 200 s, and the error of failure energy predicted by empirical method is about 100 J. The error is within an acceptable range.
