迁移学习在机械设备智能诊断中的应用研究综述
Application of Transfer Learning in Mechanical Equipment Intelligent Diagnosis:Literature Review
刘韬 1王振亚 1伍星 2李孟航1
作者信息
- 1. 昆明理工大学机电工程学院,云南昆明 650500;云南省先进装备智能制造技术重点实验室,云南昆明 650500
- 2. 云南省先进装备智能制造技术重点实验室,云南昆明 650500;滇西应用技术大学,云南大理 671000
- 折叠
摘要
随着智能制造进程的加快与工业新质生产力的需求,机械设备的运行条件变得愈发严峻,作为保生产过程稳定运行的重要环节,设备的状态监测与故障诊断也变得愈发重要.实际生产中,设备的故障诊断常面临变工况、数据分布差异大、标签样本匮乏等问题的挑战,传统故障诊断方法在这些复杂环境下往往难以取得理想效果.迁移学习作为一种新兴技术,能够有效利用已有知识和数据,提升诊断性能.首先,分析了机械设备故障诊断的趋势,阐述了迁移学习的基本概念;其次,从基于参数的迁移学习,基于特征的迁移学习,基于实例的迁移学习以及领域自适应方法4个方面对现有的迁移诊断方法进行归纳分析;最后,总结了当前迁移学习研究中面临的问题,指出了迁移学习在机械故障诊断领域的发展趋势.本综述旨在帮助相关领域研究人员了解迁移学习的最新进展,促进迁移学习在机械设备诊断中的应用和发展.
Abstract
Accelerating the process of intelligent manufacturing and the demand for new industrial productivity,the operating conditions of machinery and equipment have become ever more severe.As an important link to en-sure the stable operation of the production process,the condition monitoring and fault diagnosis of equipment have become equally important.The fault diagnosis of equipment in actual production is often challenged by variable working conditions,large differences in data distribution,and lack of labeled samples,etc.Traditional fault diag-nosis methods are often difficult to achieve ideal results in these complex environments.Transfer learning(TL)as an emerging technology can effectively utilize existing knowledge and data to improve the diagnostic performance.Firstly,this paper analyzes the trend of mechanical equipment fault diagnosis and explains the basic concept of TL.Then TL based on parameters,TL based on features,TL based on instances and domain adaptive(DA)meth-ods are summarized and analyzed in terms of existing TL methods.Finally,the problems faced in the current TL research are summarized and the future development trend is pointed out.This review aims to help researchers in related fields understand the latest progress of TL and promote the application and development of TL in mechani-cal equipment diagnosis.
关键词
机械设备/迁移学习/变工况/故障诊断/样本分布差异Key words
mechanical equipment/transfer learning/variable operating conditions/fault diagnosis/sample distri-bution differences引用本文复制引用
基金项目
National Natural Science Foundation of China(52065030)
Key Scientific Research Projects of Yunnan Province(202202AC080008)
出版年
2024
NETL
NSTL
万方数据