数据驱动故障预测和健康管理综述  

Data-driven prognostics and health management: A review of recent advances

作　　者：彭宇[1] 刘大同[1] 

机构地区：[1]哈尔滨工业大学电气工程及自动化学院自动化测试与控制系,哈尔滨150080

出　　处：《仪器仪表学报》2014年第3期481-495,共15页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61301205);高校博士基金(20112302120027);部委预先研究课题(51317040302);中央高校基本科研业务费专项基金(HIT.NSRIF.2014017)资助项目

摘　　要：着重介绍数据驱动故障预测和健康管理(PHM)方法的研究现状。通过对数据驱动PHM方法的分类阐述,逐步说明面向复杂系统数据驱动PHM的方法体系和流程,并重点对构成数据驱动PHM方法体系的核心环节进行分析和总结。在此基础上,采用一个锂离子电池循环寿命预测实例综合分析了数据驱动PHM的实现过程。最后,分析了数据驱动PHM方法的发展趋势和研究挑战。The data-driven prognostics and health management （PHM） approaches are focused in this review.The methodologies and categories for data-driven PHM approaches are firstly introduced.Then,the data-driven PHM framework and system architecture for complex system are discussed in detail.The health state monitoring,feature identification and extraction,data-driven prediction algorithms,prognostic uncertainty and hybrid prognostic approach in data-driven PHM framework are systematically described.Based on above,the life cycle prediction of a lithium-ion battery is taken as the example to synthetically analyze the implementation process of data-driven prognostics and health management.Finally,with summarizing the research hot issues,the challenges and the developing trend of data-driven PHM are analyzed.

关 键 词：故障预测和健康管理 数据驱动故障预测 融合方法 故障预测不确定性 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]