An Unsupervised Approach to Partial Discharge Monitoring in Rotating Machines: Detection to Diagnosis With Reduced Need of Expert Support

Abstract

Condition monitoring and maintenance in rotating machine are generally managed through expert evaluation of the diagnostic properties and relevant harmfulness of the underlying aging mechanisms. In order to prompt broad diffusion of condition monitoring systems, especially in large assets and in the rising field of electrification transport, however, ways to get rid of the time-consuming and expensive support of experts must be prompted. A straightforward solution is transiting towards automatic and unsupervised diagnostics and condition assessment methodologies. This paper proposes and applies algorithms for the automatic detection of partial discharges, which is the property most often associated to the fastest accelerated aging mechanisms in electrical insulation, including noise rejection, identification of the type of partial discharge sources and estimation of an health condition index in rotating machines fed by AC sinusoidal voltage. The goal is to obtain a self-assessment of the health condition by each rotating machine of an asset, which can thus interact with the asset or maintenance manager when needed, that is, when reliability is at risk.