Study Of Blasting Vibration Effect Prediction In Xiannvyan Tunnel Under-traveling Oil And Gas Pipelines

The engineering background of this article is the exit section of Xian Nv tunnel driving blasting in Xian to Chengdu railway. AS the oil and gas pipelines are under the tunnel and they move towards to the intersection and the vertical distance was about 55 meters, so if we used the borehole-blasting method, it would be have an effect on the pipelines and the rock around the tunnel. In order to make the pipeline safe, we need to control the effect which the blasting caused.We can found that the different types of hole would have a different degrees of wave superposition in tunneling blasting process, when we analysis the surface blasting vibration characteristic through the monitoring data in the field in this paper. We can draw a conclusion though the analysis of frequency spectrum, that is to say:with the distance between gauging point and driving face increasing, the principal frequency vibration speed and the single explosive payload are increasing. however, the principal frequency of the peak vibration velocity is decreasing. The decomposition of EMD can selectively filter out unwanted waves and also it can analyze the actual detonating time and delay time between different detonators through reconstructing wave amplitude envelope. Through the regression on the measured data, we found that the project did not accord with the empirical formula of Steve Sadove, so we need to take other methods to predict the peak of blast vibration speed.We choose four forecasting methods in this essay, that are BP neural network GA-BP、 PSO-SVM、GA-SVM. Based on training of the measured data, it can predict results of each method. Compared with the forecasting results, the relative error percentage in the order of GA-BP、PSO-SVM、GA-SVM. BP neural network and the relative error percentage of GA-BP. PSO-SVM are all small, the topological structure of GA-BP is the same with BP neural network which has a certain instability. SVM conquer the defect that traditional neural network has and is used to predicting the small sample in many projects. In terms of comprehensive, it is better to take the method of PSO-SVM to predict the peak of blast vibration speed.