Abstract:

This paper presents an observer-based detection framework for permanent magnet synchronous motors (PMSMs) that can detect both uniform and localized demagnetization while explicitly handling inductance uncertainties. First, a sliding-mode observer (SMO) is developed to estimate the permanent-magnet flux linkage and to generate a flux-variation-based fault indicator sensitive to demagnetization-induced changes. To mitigate inductance uncertainties, two adaptive observers are developed to estimate the d- and q-axis inductances online, and their estimates are incorporated into the SMO-based detection framework. Unlike existing approaches that rely on accurate prior knowledge of motor inductance, prior knowledge of fault signatures, or computationally intensive adaptation, the proposed observers directly reconstruct inductances from measured current signals, which reduces computational load while improving robustness. In addition, a fault-tolerant control strategy based on the estimated flux linkage is implemented to mitigate the effects of demagnetization faults. Simulation and hardware experiments with virtual fault injection demonstrate that the proposed method achieves accurate fault detection across various operating conditions and for both uniform and localized demagnetization faults.