Structure-borne and Air-borne Sound Data for Condition Monitoring Applications

Abstract

This paper provides a new machine learning dataset that contains labeled structure-borne and air-borne sound data for eight different operating conditions of a condition monitoring demonstrator. Our dataset is used to train and evaluate multiple classifiers in order to establish a baseline accuracy for classifiers on this dataset. It can be shown that both structure-borne and airborne sound data provide relevant information to train performant condition monitoring classifiers, which can be further increased by using a combination of both sound modalities.