AI-powered warning system enhances capability to uncover hidden faults

A study, led by Dr. Wentao Huang, overcame a critical gap in five-phase permanent magnet synchronous motor (PMSM) diagnostics: conventional methods fail to assess inter-turn short-circuit (ITSC) severity. The method integrates two technologies: a real-time tracker that diagnoses faults, and an AI analyzer that processes signals to quantify damage while estimating short-circuit parameters.

For years, the challenge of quantifying inter-turn short-circuit severity in operating motors has stumped engineers, as traditional methods struggled to decouple complex fault parameters. Conventional diagnostic approaches fell short in real-time assessment, leaving critical risks like irreversible demagnetization undetected.

The method, based on extended state observer (ESO) and convolutional neural network (CNN), was developed at Jiangnan University and represents a fundamental leap forward. Critically, its ability to isolate short-circuit turn ratio from fault resistance eliminates a key roadblock in fault diagnostics, enabling precise real-time severity grading that dictates targeted protection responses.

The research is published in the journal CES Transactions on Electrical Machines and Systems.

This method delivers critical real-world protection: by enabling precise fault localization and real-time severity assessment, it provides sufficient information for implementing effective fault-tolerant measures. Furthermore, potential maintenance costs can be substantially reduced. For electric vehicles specifically, the technology serves as a crucial safeguard, preventing undetected motor short circuits from escalating into life-threatening electrical fires.

Looking to the future: Smarter, self-protecting motors

Next-phase development will equip motors with self-protection capabilities: automatically reducing power during fault detection to prevent damage, while seamlessly integrating with factory networks for live fleet health monitoring.

Beyond industrial use, this technology could be adapted for critical infrastructure—potentially hardening wind turbines against generator failures in harsh environments and incorporating protective systems into aerospace electric propulsion to mitigate in-flight hazards.

This evolution leverages our core breakthrough in real-time fault decoupling, enabling machines to autonomously respond to electrical faults before they escalate. Imagine wind farms autonomously reporting issues during storms, or electric aircraft proactively containing component overheating in flight—all enabled by advanced diagnostic intelligence.

Provided by
CES Transactions on Electrical Machines and Systems