基于PCA-MSSA-BP神经网络的列车车轮踏面磨耗预测模型  

Prediction Model of Train Wheel Tread Wear Based on PCA-MSSA-BP neural network

作　　者：王冬 杨钰鑫 WANG Dong;YANG Yuxin

机构地区：[1]国能铁路装备有限责任公司包头车辆维修分公司,内蒙古包头014060 [2]西南交通大学轨道交通运载系统全国重点实验室,成都610031

出　　处：《科技创新与应用》2024年第12期49-54,共6页Technology Innovation and Application

基　　金：国能铁路装备有限责任公司包头车辆维修分公司研究项目(TZKY-21-45)。

摘　　要：分析列车车轮踏面磨耗,预测车轮剩余寿命,对降低车辆运营成本、提高运行安全品质具有重要意义。该文以某公司某型车为例,分析轮对历史检修数据,建立基于PCA-MSSA-BP神经网络的车轮踏面磨耗模型,与传统方法相比,预测精度更高、速度更快。该文首先用主成分分析法从众多磨耗影响因素中提取4个主成分因子,接着建立BP神经网络模型,并针对麻雀优化算法进行改进,验证改进效果,将改进后麻雀算法对网络权值和阈值进行优化,实验结果表明,轮径磨耗、轮缘厚磨耗预测的平均绝对误差分别为0.1935、0.1215 mm。It is of great significance to reduce vehicle operation cost and improve operation safety quality to analyze train wheel tread wear and predict wheel residual life.Taking a certain type of car in a company as an example,the historical maintenance data of wheel sets are analyzed,and the wheel tread wear model based on PCA-MSSA-BP neural network is established.Compared with the traditional method,the prediction accuracy is higher and the speed is faster.In this paper,four principal component factors are extracted from many wear influencing factors by principal component analysis,and then the BP neural network model is established.The improved sparrow optimization algorithm is improved and the improved effect is verified.The improved sparrow algorithm is used to optimize the network weight and threshold.The experimental results show that the mean absolute errors of wheel diameter wear and flange thickness wear are 0.1935 mm and 0.1215 mm respectively.

关 键 词：车轮 磨耗预测 主成分分析 麻雀算法 BP神经网络 

分 类 号：U270.33[机械工程—车辆工程]