Vibration signals serve as an important parameter for diagnosing the mechanical fault diagnosis in high-voltage circuit breakers (HVCBs). Influenced by the complicated mechanical structures, transmission paths of mechanical vibrations, and signal collection through vibration accelerators at limited positions in HVCBs, it is challenging to fully utilize the information contained in vibration signals for fault diagnosing with high efficiency under small samples. To address this issue, this paper presents a hybrid method that accurately decomposes the transient vibration signals through the optimized tunable Q-factor wavelet transform and identifies the mechanical faults in HVCBs using the twin support vector machine (TWSVM). Meanwhile, D–S evidence theory is applied to fuse the diagnostic results from the TWSVM-based classifiers that exhibit a certain degree of inconsistency under the vibration signals from different positions. Vibration signals from a 35 kV-rated HVCB are measured under both the normal condition and typical mechanical faults for verification. The results have shown that the proposed method can efficiently capture discriminate features in transient vibration signals and has a high fault identification rate under small samples. Comparative analysis with different diagnosis models further validates the effectiveness of the proposed method in terms of diagnosis accuracy and efficiency.
Dong et al. (Sun,) studied this question.