Rapid Fault-Tolerant MPC Strategy for Six-Phase PMSMs: Optimizing Torque Stability, Current Constraint Management During Phase Transition

Abstract

Six-phase permanent magnet synchronous machines (PMSMs) provide improved fault tolerance and reliability, making them well-suited for critical applications like aerospace and hybrid electric vehicle systems. Nonetheless, guaranteeing torque stability while keeping phase currents within safe limits during fault scenarios poses considerable challenges. Conventional control strategies struggle with current redistribution when phase faults occur, leading to torque oscillations and potentially damaging current levels in remaining healthy phases. This paper proposes a Fault-Tolerant Model Predictive Control (FT-MPC) strategy that optimizes current distribution in six-phase PMSMs to maintain smooth torque output while strictly adhering to peak current constraints. The proposed approach employs a predictive model to calculate optimal current references during the transition from six-phase to three-phase operation, implementing a cost function that balances torque maintenance with current limitation requirements. Simulation analysis and experimental testing on a 3 kW six-phase PMSM setup are conducted to validate the effectiveness of the proposed FT-MPC under various fault scenarios, comparing it with the conventional controllers. Compared to conventional controllers, the proposed method prevents current spikes during phase-switching transients while maintaining torque within reference values. Additionally, the controller successfully limits currents in healthy phases to remain below predetermined thresholds, preventing thermal damage while maximizing available torque. The comprehensive experimental results confirm that the FT-MPC approach significantly enhances system reliability and performance during fault conditions. It is particularly suitable for electric vehicle propulsion systems, aerospace applications, and other safety-critical industrial drives requiring fault-tolerant operation.