【摘要】 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。更多还原

【Abstract】 In order to solve pressure bump prediction problem caused by coal mining more reasonably and effectively, the Xinzhouyao coal mine was taken as an example, the PCA method, a data dimensionality reduction algorithm was used to compress information contained in input sample to GRNN, a PCA-GRNN model was built for predicting pressyre bump. Nine factors affecting intensity of pressure bump were identified such as thickness and angle of coal seam. Four principal components factors expressions of pressure bump were obtained. Models of BPNN, GRNN and PCA-BP were built to predict pressure bump too. The results demonstrate that the average training error of the PCA-GRNN model is 3.5%, the average prediction error is 3.6%, which has good predictive ability and generalization ability, and can be used to ensure the safety production of coal mine.更多还原