No abstract

No abstract

No abstract

With the development of China's transportation industry, more and more tunnel projects will pass through the gas geological area, resulting in the tunnel construction will face a huge hidden danger. The gas will flow out rapidly after the excavation of the working face. The distribution and concentration of gas in the tunnel are the key factors affecting the gas disaster. In time, grasping the gas concentration near the excavation face is the key to avoiding safety accidents. Therefore, predicting the gas concentration near the tunnel excavation face is necessary after blasting excavation. This paper relies on the NEW Chengdu-Kunming Railway Like Tunnel project to conduct research. According to the collected historical gas concentration data, based on the back-propagation neural network and long and short-term memory neural network, a new gas concentration prediction model based on the fusion algorithm of back-propagation and long and short-term memory (BP-LSTM) is proposed to realize the long-term prediction of gas concentration near the face of the tunnel after blasting excavation. This paper used the Pearson correlation method to improve model prediction accuracy and analyze the correlation between geological parameters and environmental factors and gas concentration occurrence. This model predicts the gas concentration of four monitoring points (LAFEF monitoring point, RAFEF monitoring point, EFV monitoring point, and AFEF monitoring point) in actual tunnel engineering. And compared with the prediction results of the back-propagation and long and short-term memory models, the model was evaluated according to the error. The results indicate that the prediction result of this model has high accuracy, and it can be successfully applied to predict the gas concentration after excavating the face of the gas tunnel, which can ensure the safety of construction.