A fault diagnosis method for analog circuits based on EEMD-PSO-SVM

Area of Science:

• Electrical Engineering
• Signal Processing
• Machine Learning

Background:

• Analog circuits are vital for electronic equipment reliability and safety.
• Traditional fault diagnosis methods struggle with nonlinear, non-stationary signals and parameter selection.
• Low diagnostic accuracy and model complexity hinder effective analog circuit fault detection.

Purpose of the Study:

• To propose a novel fault diagnosis method for analog circuits.
• To address the limitations of traditional methods in accuracy and parameter selection.
• To enhance the reliability and safety of electronic equipment through improved fault detection.

Main Methods:

• Ensemble Empirical Mode Decomposition (EEMD) for adaptive multi-scale feature extraction from fault signals.
• Pearson correlation coefficient and energy value analysis for initial feature vector construction.
• Maximum Information Coefficient (MIC) algorithm for optimized feature selection.
• Particle Swarm Optimization (PSO) to tune Support Vector Machine (SVM) hyperparameters for classification.

Main Results:

• The proposed method effectively extracts multi-scale fault features using EEMD.
• MIC algorithm successfully optimizes the feature vector, reducing complexity.
• PSO-optimized SVM achieves superior classification accuracy and model training efficiency.
• The method overcomes challenges associated with wavelet basis function selection.

Conclusions:

• The integrated EEMD-MIC-PSO-SVM approach offers a robust solution for analog circuit fault diagnosis.
• This method significantly enhances diagnostic accuracy and efficiency compared to traditional techniques.
• The proposed approach contributes to the improved safety and reliability of electronic systems.