Fault diagnosis in rotating machines based on transfer learning: Literature review

Abstract

With the emergence of machine learning methods, data-driven fault diagnosis has gained significant attention in recent years. However, traditional data-driven diagnosis approaches do not apply to engineering diagnosis problemssince they require that training and testing data have a consistent distribution. Transfer learning (TL) approaches for fault diagnosis are gaining popularity as a means of resolving this issue. These approaches aim to design models efficiently addressing target tasks by leveraging data from related but distinct source domains. The purpose of this study is to present a comprehensive survey of the recent progress made in applying TL techniques to diagnose faults in rotating machines. An overview of parameter-based, instance-based, feature-based, and relevance-based knowledge transfer is provided, followed by a summary of the various categories under which knowledge is transferred. These categories encompass various working environments, different machines, fault locations and their severity, imbalanced data, and more.This paper offers its readers a framework that can assist them in better understanding and recognizing the research status, problems, and future directions of transfer learning techniques for fault identification.