基于PSO-SVM的煤矿巷道爆破效果预测关键技术研究

Research on Key Technologies of Blasting Effect Prediction of Coal Mine Roadway based on PSO-SVM

煤矿巷道爆破效果受很多因素的影响,传统单一模式下的人工智能方法对煤矿巷道爆破效果预测不佳。因此以提高模型的预测精度为目的,通过建立PSO-SVM模型进行关键参数寻优。计算得出惩罚参数c与核函数参数g分别为14.0046和1.3622。以我国煤矿岩石巷道爆破工程为背景,从现场调研收集到的100余条巷道爆破工程实例中选取42组典型案例作为训练和测试样本,分别在RBF核函数的基础上应用传统SVM、Gridsearch-SVM和PSO-SVM模型对炮孔利用率进行预测对比,得到3种预测结果的准确率分别为66.67%、75%、91.67%;同时进一步验证了在PSO-SVM模型中4种不同核函数的预测准确率。结果表明:PSO-SVM模型中选取RBF核函数所得到的准确率最高,精度能够满足工程实际需求。

The effect of coal mine roadway blasting is affected by many factors.This effect cannot be predicted effectively enough by traditional single-mode artificial intelligence methods.Therefore,in order to improve the prediction accuracy of the model,the key parameters of PSO-SVM(Particle Swarm Optimization-support vector machine) model were optimized,and the penalty parameters c and kernel function parameters g were calculated as 14.0046 and 1.3622,respectively.Based on the rock tunnel blasting engineering of coal mines in China,42 typical cases were selected as training and testing samples from more than 100 tunnel blasting engineering cases collected from field investigations.Based on the RBF kernel function,the traditional SVM,Grid search-SVM and PSO-SVM models were used to compare the prediction accuracy of the blast hole utilization factor,and the results were 66.67 %,75 % and 91.67 %,respectively.At the same time,this paper further verified the prediction accuracy of four different kernel functions in PSO-SVM model.The results show that the RBF kernel function selected in PSO-SVM model has the highest accuracy,and the accuracy can meet the actual needs of engineering.