Fault Diagnosis of PMSM Stator Winding Based on Continuous Wavelet Transform Analysis of Stator Phase Current Signal and Selected Artificial Intelligence Techniques

High efficiency, high reliability and excellent dynamic performance have been key aspects considered in recent years when selecting motors for modern drive systems. These features characterize permanent magnet synchronous motors (PMSMs). This paper presents the application of continuous wavelet tran...

Full description

Saved in:
Bibliographic Details
Published in:Electronics (Basel) 2023-04, Vol.12 (7), p.1543
Main Authors: Pietrzak, Przemyslaw, Wolkiewicz, Marcin
Format: Article
Language:English
Subjects:
Algorithms
Artificial intelligence
Automation
Classification
Continuous wavelet transform
Effectiveness
Failure
Fault detection
Fault diagnosis
Fourier transforms
Machine learning
Magnets, Permanent
Methods
Multilayer perceptrons
Neural networks
Parameter estimation
Permanent magnets
Phase current
Short circuits
Signal processing
Stators
Support vector machines
Synchronous motors
Wavelet analysis
Wavelet transforms
Winding
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:High efficiency, high reliability and excellent dynamic performance have been key aspects considered in recent years when selecting motors for modern drive systems. These features characterize permanent magnet synchronous motors (PMSMs). This paper presents the application of continuous wavelet transform (CWT) and artificial intelligence (AI) techniques to the detection and classification of PMSM stator winding faults. The complex generalized Morse wavelet used for CWT analysis of three different diagnostic signals—the stator phase current, its envelope and the space vector module—is used to extract the symptoms most sensitive to the interturn short circuits (ITSCs) at the incipient stage of the damage. The effectiveness of automatic stator winding fault classification is compared for three selected ML algorithms: multilayer perceptron, support vector machine and k-nearest neighbors. The effect of the ML models’ hyperparameters on their accuracy is also verified. The high effectiveness of the proposed methodology is confirmed by the results of the experimental verification carried out for different load torque levels and supply voltage frequency values.
ISSN:2079-9292
2079-9292
DOI:10.3390/electronics12071543