Temperature-Insensitive Irreversible Demagnetization Fault Diagnosis Based on Dual-State Flux Estimation for PMSMs

This paper proposes a temperature-insensitive irreversible demagnetization (ID) fault diagnosis based on dual-state flux estimation for permanent magnet synchronous motors (PMSMs) considering temperature influence. The methodology employs dual-state ESOs to accurately capture system disturbances and compute the real magnetic flux through full-rank fourth-order discrete equations, leveraging the particle swarm optimization (PSO) algorithm. This approach ensures precise flux estimation unaffected by other parameter disturbances. Additionally, a novel temperature estimation strategy utilizing the multi-precision physics-informed neural network (MP-PINN) is proposed, which synthesizes physical knowledge from a lumped parameter thermal network (LPTN) and empirical data, achieving high accuracy and generalization with limited high-precision data. The ID faults are diagnosed by correlating the estimated flux and temperature, with a threshold set considering the temperature's effect on the permanent magnet's remanence characteristics, thus avoiding misdiagnosis of reversible changes as irreversible faults and enhancing diagnostic accuracy. The paper systematically elaborates on the dual-state ESO-based flux estimation method, the MP-PINN-based temperature estimation approach, and the comprehensive ID fault diagnosis process, supported by experimental validations that underscore the effectiveness of the proposed solution