基于数据驱动的微小故障诊断方法综述

能否及时诊断出微小故障是保障系统安全运行并抑制故障恶化的关键，本文针对微小故障幅值低、易被系统扰动和噪声掩盖等特点，从数据驱动的角度对现有研究进行综述.并将其分为三大类: 基于统计分析的微小故障诊断技术、基于信号处理的微小故障诊断技术和基于人工智能的微小故障诊断技术，进而对不同方法的基本研究思想、研究进展、应用以及局限性予以介绍.最后不仅指出复杂系统微小故障诊断研究中的现存问题，而且从增加新的信息、挖掘未利用的隐含信息和采用新的数学工具三个角度进行展望，提出基于关联性分析、基于多源信息融合、基于机器学习和基于时频分析四个值得探究的微小故障诊断思想.

A Review of Data Driven-based Incipient Fault Diagnosis

As timely incipient fault diagnosis is the key to guarantee operation safety and suppress fault deterioration, this paper gives a review of data driven-based researches for incipient faults, which have low amplitude and may be covered by system disturbance and noise easily. Data driven-based incipient fault diagnosis can be divided into three parts, i.e., statistical analysis-based technology, signal processing-based technology, artificial intelligence-based technology. Their basic ideas, research progresses, application and limitations are discussed in detail. Furthermore, this paper not only points out the existing problems about complex systems, but also looks forward to the advance of this area by means of adding new information, mining unused implied information, using new mathematical tools. Finally, four thoughts worth exploring are proposed: diagnosis based on correlation analysis, multi-source information fusion, machine learning and time-frequency transform.