基于PSO-LSSVM模型的露天矿爆破振动效应预测

Prediction of blasting vibration effect in open-pit mine based on PSO-LSSVM model

为寻找一种适合露天矿爆破振动效应的预测模型,结合内蒙古康宁爆破公司德隆和华武煤矿爆破工程的50组实测数据,建立最小二乘支持向量机(LSSVM)模型分别对爆破峰值振动速度v和振动主频f两个评价爆破振动效应的指标进行预测,并结合粒子群算法(PSO)寻优获得LSSVM中正则化参数γ和核函数宽度系数σ的最佳参数组合。结果表明:PSO-LSSVM模型对爆破峰值振动速度和振动主频预测的MRE为3.31%和6.38%,RMSE为0.98%和2.02%。与BP神经网络、LSSVM模型对比,此模型具有更好的泛化能力和更高的预测精度,为多因素影响下爆破振动效应预测提供了一种新的思路。

In order to find a suitable model for predicting the blasting vibration effect of open-pit mines,the least square support vector machine(LSSVM)model is established to predict the blasting peak vibration velocity and blasting vibration main frequency which evaluate the blasting vibration index combining with 50 sets of measured data of the blasting engineering of Delong and Huawu Coal Mine of Inner Mongolia Corning Blasting Company.In addition,particle swarm optimization(PSO)is used to optimize the optimal combination of regularization parametersγand kernel function width coefficientsσin LSSVM.The results show that the average relative error of the PSO-LSSVM model to predict the blasting vibration velocity and blasting vibration main frequency are 3.31%and 6.38%,and the mean square error is 0.98%and 2.02%.Compared with BP neural network model and LSSVM model,PSO-LSSVM model has better generalization ability and higher prediction accuracy.The research provides a new idea for the prediction of blasting vibration effect under the influence of multiple factors.