基于GRA-GEP的爆破峰值速度预测

摘要

针对在爆破施工中爆破振动危害严重、爆破振动峰值速度难以预测的问题，通过灰色关联度理论和MyEclipse开发工具，建立基于灰色关联度分析(GRA)和基因表达式编程算法(GEP)的GRA−GEP爆破峰值速度预测模型。以湖北铜录山现场露天台阶爆破实测数据进行模拟预测，通过灰色关联度分析，认为最大段药量、总装药量、水平距离、高程差、前排抵抗线长度、测点与最小抵抗线方向夹角等与爆破峰值速度存在相关性，进而为了实现爆破峰值速度进行预测，根据GEP计算思路，采用MyEclipse软件进行Java语言编程模拟运算。研究结果表明：GRA−GEP模型预测结果最大相对误差为14.4%，平均相对误差为7.8%，远低于萨道夫斯基经验公式(平均相对误差30.6%)与BP神经网络预测模型(平均相对误差13.3%)。%In order to predict the peak particle velocity of blasting vibration, the measured data of an open pit bench blasting was selected, and a prediction model of peak particle velocity of blasting vibration was established based on grey relational analysis and gene expression programming (GRA−GEP) with the theory of grey correlation degree and MyEclipse development tool. Based on blasting data of open pit bench in Tonglushan in Hubei Province, the maximum explosive charge, total charge, horizontal distance, height difference, the front line of resistance, measuring point and the minimum resistance line angle were associated with peak particle velocity of blasting vibration for sure. Then, according to GEP calculation ideas, and the Java language, blasting-vibration-peak-speed was predicted through MyEclipse software platform. The results show that maximum prediction error of forecast results is 14.4%;the average error is 7.8%, which is much lower than forecast value of experience formula (the average error is 30.6%) and the BP neural network (the average error is 13.3%).