A Small-Sample Fault Diagnosis Method for High-Voltage Circuit Breaker Spring Mechanisms Based on Multi-Source...

Area of Science:

• Electrical Engineering
• Mechanical Engineering
• Data Science

Background:

• High-voltage circuit breakers are critical for power system reliability.
• Limited fault data and single-feature limitations hinder effective diagnosis of spring operating mechanisms.
• Existing methods struggle with capturing static spring fatigue characteristics.

Purpose of the Study:

• To develop a small-sample fault diagnosis method for high-voltage circuit breaker spring operating mechanisms.
• To address challenges of limited fault samples and insufficient feature representation.
• To improve the accuracy and reliability of fault diagnosis.

Main Methods:

• Constructed a multi-source sensing system (MEMS pressure/travel, coil/motor current) for comprehensive monitoring.
• Created a high-dimensional feature space using Savitzky-Golay filtering and physical feature extraction.
• Designed a two-layer Stacking ensemble learning model with SVM, RF, KNN as base classifiers and Logistic Regression as meta-learner.

Main Results:

• Achieved an average diagnostic accuracy of 96.1% under normal and four typical fault conditions.
• Demonstrated superior performance compared to single base models (e.g., RF at 94.2%).
• Identified closing and opening pressures as critical features for mechanical fault distinction.

Conclusions:

• The proposed multi-source feature fusion and Stacking ensemble learning method is effective for small-sample fault diagnosis.
• This approach provides a robust theoretical basis and technical support for condition-based maintenance of high-voltage circuit breakers.
• Non-invasive MEMS sensors offer a solution for capturing static spring fatigue characteristics.