This study aimed to propose an equal-integral-bandwidth feature extraction method based on fast Fourier transform (FFT) to solve the problem of cumbersome processing and a large amount of calculation in the common feature extraction algorithm for vibration signals of on-load tap changer (OLTC). First, the vibration signals of OLTC were preprocessed in segments, which highlighted the status features and avoided the shortcomings of the FFT spectrum that lacked time axis information. Second, the vibration signal segments were analyzed with FFT, and the generated signal spectrum was divided into several segments according to equal integral. The bandwidth coefficient obtained in each segment was the characteristic value. Third, this study proposed that adding appropriate time domain features and further improving the algorithm could improve the accuracy of fault diagnosis. Finally, the main mechanical faults of OLTC were simulated, and the vibration signals were collected to carry out the fault diagnosis experiment of OLTC. The results showed that the FFT-based equal-integral-bandwidth feature extraction method was simple in processing, small in calculation, easy to implement in an embedded system, and had a high accuracy of fault diagnosis.
Keywords: OLTC; equal-integral-bandwidth; fault diagnosis; feature extraction; vibration signal.