煤矿冲击地压预测的PCA-GRNN神经网络方法  

Neural network method of PCA and GRNN for rock burst prediction in coal mine

作　　者：高峰[1] 陈连城[1] 王连国[2] 

机构地区：[1]山西大同大学煤炭工程学院,山西省大同市037003 [2]中国矿业大学深部岩土力学与地下工程国家重点实验室,江苏省徐州市221008

出　　处：《中国煤炭》2016年第4期48-52,共5页China Coal

基　　金：国家自然科学基金项目(50874103);山西省软科学研究计划项目(2014041068-4)

摘　　要：为了更合理有效地解决煤矿冲击地压危险性预测问题,引入主成分分析法对广义回归神经网络的输入样本进行信息压缩,得到冲击地压危险性影响因素的主成分因子,构建BPNN、GRNN、PCA—BP、PCA—GRNN 4种神经网络模型。预测结果表明所建PCA—GRNN模型较之其它3种模型整体工作性能优势明显,具有很好的预测能力和泛化能力,能较好解释冲击地压与各影响因素间的关系。In order to more reasonably and effectively solve the risk prediction problem of rock burst in coal mine,the PCA（Principal Component Analysis） was introduced to compress the information from input samples of the GRNN（Generalized Regression Neural Network）,and find the principal component of rock burst risk influencing factors,four neural network models were built,which included BPNN,GRNN,PCA-BP and PCA-GRNN.The prediction results indicated that the PCA-GRNN model showed more excellent network performances and higher prediction accuracy and generalization ability than the other three models,which was able to analyze preferably the relationship between the rock burst and each influencing factor.

关 键 词：冲击地压 主成分分析法 广义回归神经网络 电磁辐射预警技术 预测 

分 类 号：TD324[矿业工程—矿井建设]