Data-driven estimation of multiple fault parameters in permanent magnet synchronous motors

This paper presents symbolic analysis of time series data for estimation of multiple faults in permanent magnet synchronous motors (PMSM). The analysis is based on an experimentally validated dynamic model, where the flux linkage of the permanent magnet and friction in the motor bearings are varied in the simulation model to represent different stages of degradation. The fault magnitudes are estimated from the time series of the instantaneous line current. The behavior patterns of the PMSM are compactly generated as quasi-stationary state probability histograms associated with the finite state automata of its symbolic dynamic representation. The proposed fault estimation method is suitable for real-time execution on a limited-memory platforms, such as those used in sensor network nodes.