A fault detection method for motors based on Local Polynomial Fourier Transform

In industry and research, most signals´ frequency is time-variant. Typically, the fault signals of the Brushless Direct Current Motors (BLDCM) are such signals. The commonly-used Fourier transform (FT) cannot be deployed here to detect this type of fault signals, as it cannot extract the changing distribution of the frequency with time. We have to analyze these signals within the time-frequency domain. So, we introduce one of the more efficient time-frequency representation (TFR)-the Local Polynomial Fourier Transform (LPFT) to address this problem. The properties and usage of the LPFT is briefly covered. And we conduct the Motor Current Signature Analysis (MCSA) with the LPFT to distinguish the fault signals collected from the BLDCM. The results of three different TFRs including the LPFT are then presented and compared. The results clearly demonstrate that the LPFT provides ideal resolution and concentration and avoids the influence of cross terms. Therefore, we draw the conclusion that the LPFT can be useful in the process of fault detection in motors.