基于SSA-BP的爆破振动峰值速度预测研究

Prediction of Peak Velocity of Blasting Vibration Based on SSA-BP

为了精准预测爆破振动峰值速度(PPV),有效降低爆破振动的危害,以星光一号露天矿山爆破工程为依托,选取爆心距、堵塞长度、最小抵抗线、炸药单耗、最大单孔装药量、总延期时间、最大单响药量等7个影响因素作为输入变量,采用灰色关联分析法评估各因素与PPV之间的相关性,构建麻雀搜索算法(SSA)优化BP神经网络的爆破峰值振速预测模型,对三向峰值振动速度进行预测,并与BP神经网络模型预测结果进行对比分析,得到SSA-BP神经网络模型预测结果的平均误差分别为6.08%、7.34%、1.91%,BP神经网络模型预测结果的平均误差分别为22.19%、54.01%、25.29%。研究结果表明:SSA-BP神经网络模型全面考虑了多种爆破设计参数对振动峰值速度的影响;麻雀搜索优化算法有效解决了传统BP神经网络模型容易陷入局部最优的问题,预测结果更精确,与振速监测值吻合度更高、误差更小;并且极大地缩短了样本数据的学习训练时间,加快BP神经网络预测模型的收敛速度,可为类似露天爆破工程质点峰值振速的预测提供借鉴。

To accurately predict the peak particle velocity(PPV)and effectively reduce the hazards of blasting vibration,a prediction model was built by BP neural network based on the blasting project of Xingguang No.1 open-pit mine.Seven influencing factors as core distance,plugging length,minimum resistance line,explosives unit consumption,maximum single-hole charge,total extension time,and maximum single-delay charge,were selected as input variables,and the correlation between each factor and PPV was evaluated by using the grey correlation analysis method.The Sparrow Search Algorithm(SSA)optimized the BP neural network to predict the three-way peak vibration velocity.By comparing and analyzing the prediction results of the BP neural network model,the average errors of the prediction results of the SSA-BP neural network model were 6.08%,7.34%,and 1.91%,respectively,and that of the prediction results of the BP neural network model was 22.19%,54.01%,and 25.29%,respectively.The results show that the SSA-BP neural network model comprehensively considers the influence of multiple blasting design parameters on the peak vibration velocity.The sparrow search optimization algorithm can effectively solve the problem of the traditional BP neural network model,which quickly falls into the local optimum.The prediction results are more accurate,and the vibration velocity monitoring value is more consistent with smaller errors.Meanwhile,it can significantly shorten the learning and training time of the sample data to speed up the convergence speed of BP.Additionally,it can also significantly shorten the training time of sample data and accelerate the convergence speed of the BP neural network prediction model.