Multisignal Joint HVCB Fault Diagnosis Research Based on Adaptive Framing MFCC Feature Extraction Method

Mechanical fault diagnosis of high-voltage circuit breakers (HVCB) is of great significance for grid safety. However, as evaluation data for diagnosis, most relevant studies considered only a single-signal type with one-sided information. Furthermore, when using the Mel frequency cepstrum coefficient (MFCC) algorithm to extract features from fault signals, the traditional equally spaced framing method does not match the characteristics of the HVCB action mechanism. Therefore, this study combines travel, vibration, and sound signals to propose an adaptive framing MFCC (AF-MFCC) feature extraction method. First, the signal is subjected to endpoint detection according to the HVCB action mechanism and corrected according to the travel curve, and a series of action moments are adaptively acquired as the basis of MFCC frame splitting. The fault signal is subjected to two frame-splitting refinement processes to extract the AF-MFCC features of the signal, which solves the problem that the traditional MFCC cannot be adapted to the HVCB fault diagnosis application scenarios. Simulation examples show that the AF-MFCC has higher accuracy and more stable performance. Furthermore, this study proposes a two-stage diagnosis mechanism for multisignal fusion. The travel signal is used for primary abnormality diagnosis, and then, the AF-MFCC fusion features of vibration and sound signals are used for final fault diagnosis. The diagnostic accuracy is 93.37%, a significant improvement compared with conventional single-signal fault diagnosis accuracy. This verifies the superiority of the multisignal fusion diagnosis mechanism and solves the problem of single data type, which can be applied to HVCB fault diagnosis.