基于BP神经网络的桩基爆破振动速度预测

摘要

为研究桩基爆破振动对邻近埋地天然气管道的影响,以荣乌高速公路第7标段红泉村3号大桥桩基爆破振动为研究对象,以水平距离、雷管段数、总装药量、最大单段装药量、桩基深度和爆心距作为主要因素,建立桩基爆破振动BP神经网络预测模型,以现场测试的15组数据为学习样本对模型进行训练,以5组数据为检测样本进行预测,并将预测结果与萨道夫斯基公式和高程修正公式进行对比.结果表明:BP神经网络、萨道夫斯基公式和高程修正公式预测平均相对误差分别为7.90%、27.68%和24.30%,BP神经网络比萨道夫斯基公式和高程修正公式预测精度分别提高71.43%和67.49%.%To study the impact of blasting vibration of pile foundation on nearby buried natural gas pipeline, blasting vibration of the pile foundation of No.3 Bridge of Hongquan Village on the 7th section of RongWu express-way was taken as an example.Horizontal distance,detonator segments number,total charge,maximum charge amount per delay,pile depth and burst center distance were taken as the main factor to build BP neural network prediction model for pile foundation blasting vibration,where 15 sets of data tested in the filed were used to train the study sam-ples and 5 sets of data to predict the samples,and the predicted results with the Sadov′s formula and the elevation cor-rection formula were compared.The result shows that BP neural network and Sadov′s formula and elevation correction formula in predicting the average relative error is 7.90%,27.68% and 24.30%,and the prediction accuracy of BP neural network is 71.43% and 67.49% higher than Sadov′s formula and elevation correction formula respectively.