Prediction and Analysis of Blast-Induced Vibration for Urban Shallow Buried Tunnel Using Various Types of Artificial Neural Networks

Abstract

Urban shallow buried tunnel excavated in mining method may produce a bad effect on constructions by blast-induced vibration, especially for the tunnel in complex environment. Based on Beijing metro line16 engineering which is beneath the gas pipeline, in soil and rocks mixing zone, close to buildings, comparative analysis was done between the blast-induced vibration velocity predicted by Sardolfski formula and normal back propagation neural network(BP-NN). The research shows that the average predict error of Sardolfski formula is larger than that of BP-NN because of influences of medium for seismic wave propagation, blasting technology and surrounding rock properties. Even though the BP-NN has a higher prediction accuracy, it can not meet the needs of precision blasting control. A new dynamic prediction model with local feedback characteristics called Elman neural network(Elman-NN) is proposed based on field data analysis. The prediction particle velocity accuracy of Elman-NN results is improved by 9.1 percentage. Therefore, the Elman-NN has profound guiding significance on urban shallow buried tunnel excavated safety and efficient.