Demagnetization Fault Diagnosis of Permanent Magnet Synchronous Motors Based on Stator Current Signal Processing and...

Area of Science:

• Electrical Engineering
• Machine Learning Applications

Background:

• Permanent magnet synchronous motors (PMSMs) are critical in high-reliability systems.
• Operational damage, particularly demagnetization of rotor permanent magnets (PMs), poses a significant risk.
• Effective fault diagnosis and condition monitoring are crucial for PMSM longevity.

Purpose of the Study:

• To develop and validate a machine learning (ML) based method for diagnosing demagnetization faults in PMSM drives.
• To compare the effectiveness of k-nearest neighbors (KNN) and multi-layer perceptron (MLP) models for this diagnostic task.

Main Methods:

• Utilizing time-frequency domain analysis via short-time Fourier transform (STFT) for feature extraction from stator phase current signals.
• Implementing and evaluating two distinct ML models: k-nearest neighbors (KNN) and multi-layer perceptron (MLP).
• Analyzing the impact of input vector elements, model parameters, and structures on diagnostic performance.

Main Results:

• The proposed ML-based method demonstrated very high effectiveness in detecting demagnetization faults.
• Experimental verification confirmed the reliability and accuracy of the diagnostic approach.
• Comparative analysis provided insights into the performance characteristics of KNN and MLP models for this application.

Conclusions:

• The presented ML-based demagnetization fault diagnosis method is highly effective for PMSM drives.
• STFT-based feature extraction combined with ML models offers a robust solution for condition monitoring.
• The study validates the potential of ML for ensuring the reliability of PMSM systems.