用Fisher判别法确定矿井突水水源

Identifying of mine water inrush sources by Fisher discriminant analysis method

运用Fisher判别分析(FDA)理论,根据含水层的标型组分和涌水点水样的化学成分,针对简单的两类和复杂的多类突(涌)水水源识别,分别建立Fisher的线性判别函数模型和典则判别的函数模型,并对它们进行判别分析,将所建的矿井突(涌)水水源识别的Fisher判别模型应用于华北某矿井予以验证。研究结果表明:该模型利用回代估计法所得到的误判率小,并具有较强的判别能力。运用该模型进行判别分析,简易方便,分类效率高,对研究矿井突(涌)水水源的快速、有效判别意义明显;该模型适用性强,有广阔的应用前景。

A usual discriminant analysis method, i.e., Fisher discriminant analysis （FDA） was used to identify mine water inrush sources. On the basis of Fisher discriminant analysis theory, Fisher＇s linear discriminant functions was applied in identifying simple mine water inrush sources and canonical discriminant functions was applied in complicated situations. The results indicate that the FDA method can be used to reduce the dimension effectively and extract the features based on the classification by fusing information in training data, the mine water inrush prediction model of FDA has excellent discriminant performance and the resubstitution error rate is very low. It is easy and efficient to make discriminant analysis using this model and it can provide efficient method to identify mine water inrush source in order to deal with water inrush disasters correctly. It has good prospect for the identification and classification of mine water inrush.