Machine-Learning-Assisted Real-Time Inter-Turn Short-Circuit Fault Detection in Six-Phase Induction Machines

Multiphase drives are gaining traction in the heavy-duty transportation and aerospace industries due to their inherent fault tolerance and reduced per-phase power requirements, making them ideal for reliability-critical applications. However, harsh environmental and operating conditions can compromise stator winding insulation, leading to inter-turn short-circuit (ITSC) faults that, if undetected, may escalate into catastrophic failures. Extracting robust fault features while remaining computationally efficient is a challenge for fault detection methods. This paper proposes a novel fault feature extraction technique for Machine-Learning-Assisted real-time ITSC fault detection leveraging residual currents in harmonic subspaces. The proposed approach achieves over <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 9 \%}$</tex> detection accuracy and identifies faults within half a fundamental cycle, underscoring its robustness and suitability for fault-tolerant systems.