Research on mine earthquake disaster identification based on improved SimCNN model
Mine microseisms generated during coal mining can cause significant damage to the mine.In order to identify microseisms in mines and improve accuracy,convolutional neural networks were used in the study,and similar feature layers were introduced to improve it.Due to the small sample size of the microseismic dataset,it is difficult to meet the training requirements of the model.Therefore,the study also adopted transfer learning to transfer the features of seismic data to microseismic data,and used this dataset to pre train the improved convolutional neural network.In order to further enhance the feature acquisition capability of the improved convolutional neural network,research introduced long short-term memory networks and attention mechanisms on the basis of existing improvements to obtain temporal features of data and a complete mining microseismic recognition model.The results showed that on the test set,the average time taken to study and design the mining microseismic identification model was 8.216 seconds,with the maximum and minimum recognition accuracy values of 99.87％and 93.29％,respectively,which were significantly better than the comparison mining microseismic identifi-cation model.The mining microseismic identification model can accurately predict the longitudinal and transverse wave velocities of frac-tured coal and rock during mining microseismic events.The design of a mining microseismic identification model has good performance and can provide technical support for microseismic identification in current coal mining processes.
万方数据
维普
