Rockburst, due to its suddenness, randomness, and harmfulness, is one of the main factors restricting the safe construction of deep underground engineering. To achieve accurate prediction of rockburst occurrence probability, a dataset for drilling and blasting (D&B) tunnel construction was established using microseismic (MS) and geological data collected during excavation. A dual‐modality convolutional neural network‐long short‐term memory (DCNN‐LSTM) model is proposed to simultaneously extract MS and geological features. First, the impact of kernel size, number of convolutional layers, number of LSTM layers, and the number of neurons on the prediction accuracy of rockburst occurrence probability is analyzed to determine the model structure. Second, different learning rates and batch sizes are compared to obtain the best hyperparameters for the model. Then, according to the evaluation results of mean absolute error (MAE), mean absolute percentage error (MAPE), mean squared error (MSE), and R 2 , the model proposed in this study has higher prediction accuracy than other five common time series prediction models. Finally, the permutation importance method is used to explore the importance of features in the prediction process. The results show that the DCNN‐LSTM model can provide support for rockburst early warning in D&B underground engineering.
Xia et al. (Thu,) studied this question.