矿井涌水水源的主成分分析和BP神经网络判别

为准确判别矿井涌水水源,针对矿井各主要含水层的水化学特征数据样本,利用主成分分析法消除变量中的重复信息,采用BP算法对网络进行训练,实现对随机挑选样本的判别,并与Bayes判别结果进行比较.结果表明：主成分分析与BP神经网络相结合的方法判别涌水水源的正确率为82.35％,优于Bayes判别法.该研究为有效开展矿井防治水工作提供了参考.

Research on mine water inrush source discrimination based on combination of PCA and BP neural network

This paper is a response to a more accurate discrimination of swallet water.The discrimination-focused research comprises eliminating the duplicate information in variables,drawing on the data samples associated with water chemistry characteristics related to the major aquifers in a mine and using principal component analysis ; training the network using BP algorithm to realize the discrimination of randomly selected samples; and performing a comparison between the discrimination and the results derived from Bayes discriminant method.The results demonstrate that,when used to discriminate gushing sources,the combination of the principal component analysis and BP neural network ensures the correct rate of 82.35％ and thus boasts an advantage over Bayes discriminant method.This study may serve as a reference for an effective prevention and control of gushing water in a coal mine.