论文部分内容阅读
为更合理有效地解决煤矿开采引起的冲击地压危险性预测问题,以忻州窑煤矿冲击地压事故为工程背景,采用一种数据降维算法—主成分分析法(PCA),对广义回归神经网络(GRNN)的输入样本进行信息压缩,构建冲击地压危险性预测的PCA-GRNN模型。通过PCA法提取影响冲击地压强度的煤层厚度、倾角等9个因素,得到冲击地压危险性影响因素的前4个主成分因子表达式,并构建BPNN,GRNN和PCA-BP等另外3种模型,验证PCA-GRNN法预测冲击地压危险性的智能性和泛化能力。结果表明,所建PCA-GRNN模型平均训练误差为3.5%,平均预测误差为3.6%,有很好的预测能力和泛化能力。
In order to solve the problem of rock burst risk prediction reasonably and effectively, taking Xinzhou kiln coal mine rock burst as engineering background, a data dimension reduction algorithm (PCA) Network (GRNN) input samples for information compression to build a PCA-GRNN model of rock burst risk prediction. PCA method was used to extract the thickness and dip angle of coal seam that affect the rock burst strength, and the expressions of the four principal component factors influencing the risk factors of rock burst were obtained. Three other BPNN, GRNN and PCA-BP Model to verify the intelligence and generalization ability of PCA-GRNN method to predict the risk of rock burst. The results show that the average training error of PCA-GRNN model is 3.5% and the average prediction error is 3.6%, which shows good predictive ability and generalization ability.