Mechanical Fault Feature Extraction and Diagnosis of SF6 High-Voltage Circuit Breakers Based on Shell Vibration Signals

Abstract

To address the issues of sensor loosening and significant signal interference when vibration sensors are installed near the operating mechanism of SF₆ circuit breakers, this study proposes relocating sensors to the breaker enclosure. To overcome signal attenuation at the enclosure, the ZF series 252 kV SF₆ circuit breaker was investigated, with sensor placement optimised according to contact engagement positions and moving direction. Linear and nonlinear features were extracted across time-domain, frequency-domain, and time-frequency-domain analyses to characterise vibration signals under typical mechanical faults, including inadequate closing spring energy storage, pin detachment, and insufficient contact finger gripping force. A fault diagnosis model based on feature fusion of enclosure vibration signals was constructed using the Random Forest algorithm. Experimental results demonstrate that compared to mechanism-mounted solutions, this approach significantly improves recognition accuracy for diverse fault types, providing an innovative solution for circuit breaker condition monitoring and intelligent maintenance.