充填管道磨损风险的KPCA-PSO-GRNN评估模型及应用

KPCA-PSO-GRNN Evaluation Model and Application of Filling Pipeline Wear Risk

针对影响充填管道磨损风险因素的复杂性和不确定性,为更加科学准确的预测充填管道的磨损状况,减少由管道磨损带来的矿山损失,构建了一种基于核主成分分析(KPCA)和粒子群算法(PSO)优化广义回归神经网络(GRNN)的充填管道磨损风险评估模型。建立充填管道磨损风险评估指标体系,运用KPCA,对数据进行特征提取,将其结果作为GRNN的输入,采用PSO算法优化选取GRNN的光滑因子,完成GRNN的训练和学习,将该模型应用于国内某矿山充填管道的磨损风险评估中,并与其它模型进行对比。结果表明,利用该模型可以准确的预测充填管道磨损风险等级,与实际情况相一致。工程应用实例表明,所建模型预测精度更高,适用性更好,对充填管道的磨损评估具有较好的借鉴意义。

In order to predict the wear conditions of filling pipelines more scientifically and accurately,and reduce the mine losses caused by pipeline wear,this study builds a risk assessment model for filling pipeline wear based on Kernel Principal Component Analysis(KPCA)and Particle Swarm Optimization(PSO)to optimize Generalized Regression Neural Network(GRNN).An assessment index system for wear risk of filling pipelines was constructed,using KPCA extracts characteristics of the data,whose results are as the input of GRNN,and utilizing PSO algorithm to optimize the selection of GRNN smoothing factors to complete the training and learning of GRNN.The model was applied to the wear risk assessment of a domestic mine filling pipeline and compared with other models.The results confirm that the model can accurately predict the wear risk level of filling pipelines,which is consistent with the actual situation.The engineering application example shows that the model built in this paper has higher prediction accuracy,better applicability and it has a good reference for wear assessment of filling pipeline.