基于机器学习的辅助逆变器进风滤网堵塞诊断  

Machine Learning-based Diagnosis of Air Inlet Filter Blockage in Auxiliary Inverters

作　　者：曲涛 景宁 张仁义 洪希仁 常伟 王亚茹 QU Tao;JING Ning;ZHANG Renyi;HONG Xiren;CHANG Wei;WANG Yaru(Jiangsu CRRC Digital Technology Co.,Ltd.,Nanjing 210008,China;Thales China Enterprise Management Co.,Ltd,Beijing 100125,China;Shanghai Jueyun Technology Co.,Ltd,Shanghai 200030,China;Guangdong Yuxiu Technology Co.,Ltd.,Guangzhou 510623,China)

机构地区：[1]江苏中车数字科技有限公司,江苏南京210008 [2]泰雷兹(中国)企业管理有限公司,北京100125 [3]上海觉云科技有限公司,上海200030 [4]广东毓秀科技有限公司,广东广州510623

出　　处：《铁道车辆》2024年第3期80-83,共4页Rolling Stock

摘　　要：辅助逆变器是地铁牵引系统的重要组成部分,为了优化辅助逆变器的运用性能,提高地铁运行的稳定性,提出一种基于机器学习判断辅助逆变器进风滤网是否堵塞的方法。首先,在原有数据的基础上提取对辅助逆变器模块温度产生影响的特征;其次,建立模型进行模型训练。结果表明:采用该方法进行滤网堵塞诊断时,相较于支持向量机(SVM)模型方法,基于多层感知机(MLP)模型的辅助逆变器进风滤网堵塞诊断方法在数据计算和预测中诊断精度较高、鲁棒性较强。同时,研究也为后续地铁项目中辅助逆变器滤网在开展制造、维修时的优化换型提供了重要依据。Auxiliary inverter is an important part of metro traction system.In order to optimize the operation performance of auxiliary inverter and improve the stability of metro operation,a method based on machine learning is proposed to judge whether the inlet filter of auxiliary inverter is blocked.Firstly,on the basis of the original data,the characteristics that affect the temperature of the auxiliary inverter module are extracted.Secondly,a model is built for model training.The results show that compared with support vector machine(SVM)method,the multilayer perceptron(MLP)based auxiliary inverter inlet filter blocking diagnosis method has higher diagnostic accuracy and robustness in data calculation and prediction.At the same time,the research also provides an important basis for the optimization and replacement of auxiliary inverter inlet filter in the manufacture and maintenance of subsequent metro projects.

关 键 词：地铁 辅助逆变器进风滤网 多层感知机 支持向量机 模型训练 

分 类 号：U270.381[机械工程—车辆工程]