SNR-Based Detection of Static and Dynamic Eccentricity in Resolvers

Accurate position estimation is essential for electrical machine control and commutation, making resolvers critical components. Due to low manufacturing cost and robustness in harsh environments, resolvers are used but remain vulnerable to mechanical faults—especially static eccentricity (SE) and dynamic eccentricity (DE)—which degrade accuracy. Hence resolver health monitoring is vital. This study presents a robust signal-to-noise ratio (SNR)-based algorithm for detecting mechanical faults under noisy conditions. Applicable to various resolver types (wound-rotor, WR; variable reluctance, VR) and winding schemes (overlapping and non-overlapping), the method uses the same signals employed for position estimation, removing the need for extra sensing channels. Fault detection analyzes deviations of the signal envelope from a healthy baseline by subtracting resized data windows aligned via maximum sample cross-correlation (SCC). The SNR, defined as the ratio of fundamental signal power to fault-induced components, serves as the evaluation metric. Validation through finite element analysis (FEA) and experiments demonstrates 99.5% detection accuracy with suitable computational burden.