基于RS-PSO-GRNN的埋地管道土壤腐蚀预测  

Soil Corrosion Prediction of Buried Pipeline Based on the Model of RS-PSO-GRNN

作　　者：骆正山[1] 王文辉 王小完[1] 张新生[1] LUO Zheng-shan;WANG Wen-hui;WANG Xiao-wan;ZHANG Xin-sheng(School of Management,Xi＇ an University of Architecture ＆ Technology,Xi＇ an 710055,China)

机构地区：[1]西安建筑科技大学管理学院,陕西西安710055

出　　处：《材料保护》2018年第8期47-52,79,共7页Materials Protection

基　　金：国家自然科学基金(61271278);陕西省重点学科建设专项资金资助项目(E08001);陕西省教育厅自然专项基金(16JK1465)资助

摘　　要：为克服埋地管道土壤腐蚀因素间的复杂性及传统方法预测精度低、适用性差等缺陷,提出基于粗糙集(RS)和改进粒子群算法(PSO)融合广义回归神经网络(GRNN)的埋地管道土壤腐蚀预测模型。通过属性约简,提取影响管道土壤腐蚀的主要因素,将其结果作为GRNN的输入,运用改进的PSO优化GRNN的参数,构建预测模型,并以中俄原油管道为例,进行土壤腐蚀实证分析。结果表明,与标准PSO相比,改进PSO的迭代收敛速度更快,稳定性更好,且该模型预测效果优于常规的误差反向传播(BP)模型和粗糙集融合支持向量机(RS-SVM)模型,为埋地管道土壤腐蚀研究提供了新思路,具有较好的借鉴意义。In order to overcome the complexity of soil corrosion factors of buried pipeline and the defects of low precision and poor applicability of traditional prediction methods, the soil corrosion prediction model of buried pipeline was proposed based on rough set theory （RS） and improved particle swarm optimization （ PSO）, which also fused generalized regression neural network （GRNN）. Through attribute reduction, the main factors affecting pipe corrosion were extracted, and then the results were taken as input of GRNN. The improved PSO was used to optimize GRNN parameters to construct the prediction model. Moreover, the China-Russia crude oil pipeline as an example was used for the empirical analysis of soil corrosion. Results showed that compared with the standard PSO, the improved PSO iteration converged faster and had better stability, and the predicting result was better than those of conventional BP model and RS- SVM model, which provided a new idea for the research on soil corrosion of buried pipelines and possessed a good reference value.

关 键 词：土壤腐蚀 埋地管道 粗糙集理论 粒子群算法 广义回归神经网络 

分 类 号：TG172.4[金属学及工艺—金属表面处理] TB115[金属学及工艺—金属学]