Machine diagnosis using acoustic analysis: a review

Abstract

Diagnosis or fault identification in real industrial machines using audio or sound signals is a challenging task. Active research has been pursued to determine acoustic features, classification & clustering algorithms that could estimate the state of an industrial machine. Acoustic features & classifiers from different domains have been successfully implemented for fault identification in industrial machines. This paper is a comparative study of propositions, experiments, applications and systems developed by various researchers. Effort has been made to generate a collection of test benches developed, results observed and conclusion arrived. These insights suggest deep learning and anomaly detection techniques as a promising technology for preventive maintenance in real industrial machines.