IEEE DEI Schenectady Chapter, Machine-Learning Applications to Estimating Generator Useful Life

#WIE #DEI-Events #generators #diagnostics #aging #rotating #machine

Hydro-Québec is working on the development of a prognostic approach for hydrogenerators based on the combination of a physics-based model, which integrates all failure mechanisms in a graphical form, and data-based models adapted to the types of diagnostic tools. The failure mechanism graph, which has been defined by a group of experts, consists of several cause-and-effect chains formed by the grouping of sequential physical degradation states that ultimately lead to a failure state. Knowing the transition function between each pair of physical degradation states, it becomes possible to estimate the time required to reach the failure state for a given failure mechanism. The first essential step of this approach is to automatically identify the active failure mechanisms based on the analysis of observable symptoms extracted from diagnostic tools. For the stator winding of hydrogenerators, over 100 failure mechanisms have been consigned, most of which involve the presence of partial discharge (PD) activity. Analysis of each individual PD source is therefore essential to understanding the behavior and evolution of these failure mechanisms. This presentation will highlight the development carried out to automatically identify the active failure mechanisms for the stator winding, the diagnostic tools as well as the in-house integrated diagnostic application called DIAAA, and the methodology developed to recognize each active PD source using PDA and PRPD results.