摘要
针对现场实际记录中屡次发生故障的车载设备,为预测其下一次故障发生时间,提出基于岭回归算法的故障时间预测方法,整个研究过程通过随机森林算法结合人为经验的方式提取出6个有效故障特征,建立岭回归故障预测模型,并通过网格搜索结合交叉验证(Gridsearch CV)的方法优化模型超参数,在深圳某地铁线路实际数据中得到有效的验证,提出的方法可以较为准确的预测下一次故障,可以作为指导预防性维修的依据,实现故障的提前预测、提前感知、提前处理。
Aiming at the onboard equipment with repeated failures in actual field records,a fault time prediction method based on ridge regression algorithm is proposed in order to predict its next failure time.This paper extracts six eff ective fault features by random forest algorithm combined with human experience in the whole research process,and establishes ridge regression fault prediction model,optimizes the hyperparameters of the model by grid searching combined with cross validation(Gridsearch CV),which is effectively verified in actual data of a subway line in Shenzhen.The next fault can be predicted more accurately according to the proposed methods as a basis for guiding preventive maintenance,so as to realize the prediction in advance,perception in advance and processing in advance of the faults.
作者
孙超
Sun Chao(CRSC Research&Design Institute Group Co.,Ltd.,Beijing 100070,China;Engineering Research Center of Railway Industry of Intelligent and Autonomous Train Control,Beijing 100070,China)
出处
《铁路通信信号工程技术》
2024年第8期74-79,共6页
Railway Signalling & Communication Engineering
基金
国家重点研发计划资助项目(2022YFB4300600)
北京全路通信信号研究设计院集团有限公司科研项目(2300-K1240012)。
关键词
地铁车载设备
故障时间预测
岭回归算法
网格搜索和交叉验证
subway onboard equipment
fault time prediction
ridge regression algorithm
grid search and cross validation(Gridsearch CV)
