Fault Diagnosis Method for Excitation Dry-Type Transformer Based on Multi-Channel Vibration Signal and Visual Feature...

Area of Science:

• Electrical Engineering
• Condition Monitoring
• Machine Learning

Background:

• Existing fault diagnosis methods for excitation dry-type transformers struggle with multi-axis vibration data utilization, accuracy in complex conditions, and computational efficiency.
• There is a need for robust and efficient methods to ensure the operational safety and reliability of these transformers.

Purpose of the Study:

• To develop a lightweight fault diagnosis approach for excitation dry-type transformers.
• To enhance the utilization of multi-channel vibration data and improve recognition accuracy and computational efficiency.

Main Methods:

• Developed a multi-physics field coupling simulation model for excitation dry-type transformers.
• Extended the Symmetrized Dot Pattern (ISDP) method to create 2D feature maps from multi-axis vibration data, optimized using Particle Swarm Optimization (PSO).
• Employed Oriented FAST and Rotated BRIEF (ORB) for feature extraction and an Adaptive Boosting Support Vector Machine (Adaboost-SVM) for fault classification.

Main Results:

• The proposed method achieved a fault diagnosis accuracy of 94.00% for typical faults like winding looseness, core looseness, and eccentricity.
• Outperformed existing signal-to-image techniques (GAF, RP, MTF) and deep learning models (CNN) in accuracy, training, and testing time.
• Demonstrated superior stability and robustness in repeated trials, suitable for resource-constrained environments.

Conclusions:

• The lightweight fusion approach effectively addresses limitations of existing methods for excitation dry-type transformer fault diagnosis.
• The method offers a robust, efficient, and accurate solution for online monitoring and rapid diagnosis, enhancing transformer safety and reliability.