基于KPCA-PSO-RBF-SVM的矿井突水水源识别模型

Mine water inrush source identification model based on KPCA-PSO-RBBF-SVM

为对矿井突水水源进行识别以减少矿井突水事故的发生,提出了粒子群(PSO)结合RBF核参数优化的SVM模型,并使用核主成分分析法(KPCA)对选取水源特征指标进行高效降维.根据水源离子敏感性选取了8种水化学指标(K+、Na+、Mg2+、Ga2+、HCO3-、Cl-、F-、SO42-)作为突水水源识别特征参数.使用基于最大方差关联度准则的核参数选择方法并结合粒子群算法构造参数优化算法,使用参数优选后的支持向量机模型对90组突水水源识别训练数据进行模型训练,用其余32组数据进行测试,模型实测效果与Logistic模型、PCA-Fisher模型以及PSO-SVM模型进行对比,结果表明:采用径向基核函数优化的支持向量机模型能够选取较优参数,模型实测平均准确率为93.75%,误差明显低于其他模型,证明了该模型能精准且高效地识别矿井突水水源.

In order to identify mine water inrush sources to reduce the occurrence of mine water inrush accidents,a SVM model based on particle swarm optimization(PSO)and RBF kernel parameter optimization was proposed,and KPCA was used to efficiently reduce the dimension of selected water source characteristics.According to the ion sensitivity of water source,8 kinds of water chemical indexes(K+、Na+、Mg2+、Ga2+、HCO3-、Cl-、F-、SO42-)were selected as the identification characteristic parameters of water inrush source.Using kernel parameter selection method based on maximum variance correlation criterion and combining the structure parameter optimization algorithm of particle swarm optimization(PSO),90 set of water inrush source identification training data was used to simulate the mode by using parameter optimization of support vector machine(SVM)model,and the remaining 32 groups of data were used for testing.The model test results compared with that of the Logistic model,PCA-Fisher model and the PSO-SVM model.The results show that the radial basis kernel function optimization of support vector machine(SVM)model can select the optimal parameters,and the model measured the average accuracy is 93.75%,and the error is obviously lower than other models.It is proved that this model can identify mine water inrush sources accurately and efficiently.