Research on Three-Phase Asynchronous Motor Fault Diagnosis Based on Multiscale Weibull Dispersion Entropy.

Area of Science:

• Engineering
• Machine Learning
• Signal Processing

Background:

• Three-phase asynchronous motors are crucial in the machinery industry, necessitating reliable fault diagnosis for operational health.
• Existing fault diagnosis methods may lack accuracy and generalization capabilities.

Purpose of the Study:

• To propose an improved fault diagnosis method for three-phase asynchronous motors.
• To enhance the accuracy and generalization of fault diagnosis through a novel feature extraction and classification approach.

Main Methods:

• Vibration signals were processed using Weibull distribution (WB) for linearization and multiscale dispersion entropy (MDE) for feature extraction.
• A support vector machine (SVM) classifier was optimized using particle swarm optimization (PSO) to classify motor states.
• The method was validated using experimental data, including noisy signals and a public dataset, with piezoelectric acceleration sensors.

Main Results:

• The proposed WB-MDE and PSO-SVM method achieved 100% accuracy in fault classification and identification.
• The model demonstrated strong noise resistance and generalization capabilities when tested with varying signal-to-noise ratios and the CHIST-ERA SOON dataset.
• Experimental verification confirmed the effectiveness and superiority of the proposed fault diagnosis technique.

Conclusions:

• The combination of WB-MDE for feature extraction and PSO-SVM for classification offers a highly accurate and robust solution for three-phase asynchronous motor fault diagnosis.
• The method exhibits excellent performance in noisy environments and generalizes well to unseen data.
• This approach significantly contributes to ensuring the healthy operation of motors in industrial applications.