基于BP神经网络不同输入变量的爆破振速预测

Prediction of Blasting Vibration by BP Neural Network Based on Different Input Variables

以某隧道爆破开挖为实例,利用BP神经网络解决复杂非线性函数逼近问题的能力,以最大段药量、爆心距、爆破分段数、泊松比、岩石基本质量指标作为影响爆破振动速度的主要因素,选取不同维数的输入变量建立BP神经网络模型来预测爆破振动速度。对比分析各组预测值与实测值之间的相对误差,选取合理维数的输入变量建立了爆破振动危害预测的BP神经网络模型。

Based on the tunnel blasting excavation,the approximation capability of BP neural network is used to solve complex problems of nonlinear function,maximum charge weight per delay interval,explosive distance,the number of blasting segment,Poisson ratio,and rock basic quality indicators are considered as the main factors of impacting blasting vibration velocity,and the blasting vibration velocity is predicted by establishing BP neural network model of different dimension input variables.Comparing the relative error between the predictive values and the measured values,the prediction model of BP neural network for blasting vibration hazards is established by selecting a reasonable dimension of input variables.