Development and Model-in-the-Loop Validation of a Neural-Network Algorithm for the Detection and Isolation of Faults in PMSM Drives for UAV Propellers

The increase in reliability and safety levels of electrical power devices is key to the transition toward aviation electrification. Safety-critical equipment, such as propulsion motors, must comply with stringent requirements to obtain airworthiness certification. Permanent Magnet Synchronous Motors (PMSMs) have emerged as a suitable solution for high-performance flight control actuation, with promising extension to propulsive functions. Nevertheless, the reliability of PMSM power electronics converters remains below certification requirements, so fault-tolerant capabilities are essential for flightworthiness, making prompt and effective Fault Detection and Isolation (FDI) algorithms essential. This paper presents the Model-in-the-Loop (MIL) validation of a Neural Network (NN) algorithm for the detection and isolation of MOSFET faults and open-phase faults in PMSM power drives. The approach combines a feedforward NN, trained on simulation-generated data for a multi-class classification problem, with a counter-based decision logic ensuring fast and robust isolation, using a compact architecture. A feature selection analysis reduces the network size at a modest accuracy cost. For MIL validation, the NN-based algorithm is integrated within a high-fidelity dynamic model of an electrically driven UAV propeller, comprising a three-phase PMSM with Field-Oriented Control (FOC), digital signal processing, and aerodynamic loads derived from Blade Element Momentum Theory (BEMT). Faults are injected at different operating speeds, during speed transients, and under combined parametric uncertainty to evaluate the FDI performance in terms of detection latency, isolation capability, and robustness. The results demonstrate the effectiveness of the approach, which can be extended to other fault categories and support the development of fault-tolerant PMSM control.