A roadmap to fault diagnosis of industrial machines via machine learning: A brief review

Abstract

In fault diagnosis, machine learning theories are gaining popularity as they proved to be an efficient tool that not only reduces human effort but also identifies the health conditions of the machines automatically. In this work, an attempt has been made to systematically review the progress of machine learning theories in fault diagnosis from scratch to future perspectives. Initially, artificial intelligence came into the picture which started to weaken the human effort whose efficiency relies on feature extraction which depends on expert knowledge. The introduction of deep learning theories has reformed the fault diagnosis process by realising the artificial aid, encouraging end-to-end encryption in the diagnostic procedure. The deep learning theories have also filled the gap between the large amount of monitoring data and the health conditions of industrial machines. The future of deep learning theories i.e. transfer learning which uses the knowledge of one domain to another related domain during fault diagnosis has been reviewed. In last, the research trends of the machine learning theories have been briefly discussed along with their challenges in fault diagnostics.