改进MGM(1,1)的管道腐蚀预测模型构建  

Construction of pipeline corrosion prediction model based on improved MGM(1,1)

作　　者：柏淘雨 余华云[1] 宋文广[1] Bai Taoyu;Yu Huayun;Song Wenguang(School of Computer Science,Yangtze University,Hubei Jingzhou,434023,China)

机构地区：[1]长江大学计算机科学学院,湖北荆州434023

出　　处：《机械设计与制造工程》2022年第4期69-73,共5页Machine Design and Manufacturing Engineering

基　　金：国家“十四五”重大专项课题(2021DJ1006);2020年新疆维吾尔自治区创新人才建设专项自然科学计划(自然科学基金)面上项目(2020D01A132);湖北省科技示范项目(2019YYD016);中国高校产学研创新基金(2021ALA01004)。

摘　　要：针对管道腐蚀预测分析因素复杂且具有较大随机性与模糊性的问题,建立了一种基于集成学习与机器学习算法的PSO-MGM管道腐蚀预测模型。在管道腐蚀检查数据的基础上,以MGM(1,1)为预测模型,采用PSO算法选出最优白化因子,实现了管道腐蚀速率预测。仿真结果表明,所提模型预测值与实际值拟合效果良好,可较为理想地反映管道腐蚀发育情况,相较于标准GM(1,1)预测模型和MGM(1,1)预测模型,其平均相对误差更小,满足预测精度需求,并可具体预测某一管道腐蚀发育情况,具有一定的实际应用价值。Aiming at the problem of complex factors,randomness and fuzziness in pipeline corrosion prediction and analysis,a PSO-MGM pipeline corrosion prediction model based on integrated learning and machine learning algorithm is proposed.Based on the pipeline corrosion inspection data,MGM(1,1)is selected as the prediction model,and the optimal whitening factor of the model is selected by PSO algorithm to realize the prediction of pipeline corrosion rate.The results show that the predicted value of the proposed model fits well with the actual value,and can better reflect the corrosion development of the pipeline.Compared with the standard GM(1,1)prediction model and MGM(1,1)prediction model,the average relative error is smaller,which meets the demand of prediction accuracy,and can specifically predict the corrosion development of a certain pipeline,which has a certain practical application value.

关 键 词：集成学习 机器学习 管道腐蚀预测 MGM(1 1)模型 PSO算法 

分 类 号：TH122[机械工程—机械设计及理论]