Aiming to explore the role and challenges of machine learning in predicting the remaining useful life of devices,this study conducted a systematic summary of machine learning in RUL prediction of devices in terms of application methods and characteristics of each category of methods.The research on the application of common machine learning to RUL prediction of stochastic degrading devices is divided into five categories according to the difference in model training methods and whether there is a label or not:supervised learning,unsupervised learn-ing,semisupervised learning,transfer learning,and reinforcement learning.The current applications of each cate-gory of methods in the field of RUL prediction are summarized.Typical cases are listed,where the advantages of various methods are combined and complement each other,and the roles of these different methods in the prediction are explained.This study also briefly introduces the characteristics and application fields of each method and analy-zes the advantages and shortcomings of different methods.The practical problems and requirements during equip-ment operation are focused on,and the challenges and difficulties of RUL prediction for stochastic degrading de-vices are discussed.
关键词
剩余寿命预测/随机退化设备/机器学习/监督学习/无监督学习/半监督学习/迁移学习/强化学习
Key words
remaining service life prediction/stochastic degrading device/machine learning/supervised learning/unsupervised learning/semisupervised learning/transfer learning/reinforcement learning
维普
万方数据