Transformer Fault Diagnosis Method Based on Quartz Vibrating Beam Accelerometer

Fault diagnosis is critical for ensuring the reliability and safety of power systems, minimizing downtime, and preventing catastrophic failures. This study investigates the application of a quartz beam accelerometer for fault state identification by analyzing vibration signals under varying load conditions. Both time-domain and frequency-domain features are extracted to characterize normal operation, winding looseness, and core looseness states. In the frequency domain, the normal operation state consistently lacks distinct secondary peaks, while winding looseness and core looseness exhibit such peaks, primarily in the ranges of 200 Hz to 400 Hz and 400 Hz to 800 Hz, respectively. To classify these states, several classification algorithms are employed, including Linear Support Vector Machine (SVM), Radial SVM, Logistic Regression, K-Nearest Neighbors (KNN), and Decision Tree. The Decision Tree and Linear SVM demonstrate the highest performance, achieving perfect and near-perfect accuracy, respectively. The findings highlight the effectiveness of combining time-domain and frequency-domain features for precise fault state identification, offering a robust framework for improving the reliability and safety of power transformers.