Validation of HiL Simulation for SVM-Based PMSM Fault Diagnosis

Le résumé fourni par la source

Acquiring sufficient and diverse real-world fault data for training machine learning models in Permanent Magnet Synchronous Motor (PMSM) Inter-turn Short-Circuit (ITSC) diagnosis is often prohibitively expensive, timeconsuming, and carries inherent safety risks. To address this, simulation techniques serve as a practical alternative. This study investigates and validates the effectiveness of Hardware-in-theLoop (HiL) simulation as a high-fidelity data generation source, comparing its performance against conventional offline simulation. A unified, high-fidelity PMSM model was used to generate two distinct datasets: one from a MATLAB/Simulink environment (Sim-Data) and another from a real-time HiL simulator (HiL-Data). A Support Vector Machine (SVM) classifier was then trained and evaluated on both datasets to diagnose ITSC fault severities (5%, 10%, 15%). The SVM trained on HiL-Data achieved a perfect 100% test accuracy, significantly outperforming the model trained on Sim-Data. This result strongly validates the fidelity of the HiL simulation approach, demonstrating its superiority and potential as a cost-effective, safe, and reliable methodology for developing advanced data-driven diagnostic systems.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.