Mechanical fault diagnosis of high-voltage circuit breakers based on EEMD energy entropy and BO-RF

Abstract

The vibration signal is composed of impact sub-waves generated by friction and collision among mechanical components during the operating process of the circuit breaker. It contains key operational information such as spring energy release, mechanical transmission, closing latch, and braking. This paper proposes a circuit breaker fault identification method based on vibration signals. First, the denoising performance of wavelet-processed vibration signals is evaluated using signal-to-noise ratio and root mean squared error. Then, ensemble empirical mode decomposition is applied to decompose the denoised vibration signal with complex frequency structure in a systematic manner. A probability density function is constructed to calculate energy entropy. Finally, a random forest-based mechanical fault identification method optimized by Bayesian optimization is proposed. This method effectively addresses the overfitting problem under small sample conditions. Experimental results show that the proposed approach improves the diagnostic accuracy of typical circuit breaker faults and meets the requirements of condition monitoring and maintenance in substations.