Weak fault detection for wind turbine bearing based on ACYCBD and IESB

Abstract

To detect the rolling bearing fault of a wind turbine at incipient injury phase, a novel method combining adaptive cyclostationary blind deconvolution (ACYCBD) with instantaneous energy slice bispectrum (IESB) is proposed. Cuckoo search algorithm (CSA) is fused with cyclostationary blind deconvolution (CYCBD) to optimize cyclic frequency and filter length parameters, and the satisfied processing result is able to adaptively use this developed ACYCBD method. Besides, fusing the complementary superiorities of frequency weighted energy operator and slice bispectrum, a new frequency domain analysis method named IESB is put forward to identify the characteristic frequency components. During fault diagnosis, the deconvolution signal with higher signal to noise ratio is first separated from the collected signal by applying ACYCBD method. However, owing to the intense background noises, external interferences still remain in the deconvolution signal. Then the IESB is further used as a post-processing method, the residual noises are suppressed effectively, and the characteristic spectral lines are highlighted immensely. In the end, the incipient fault injury of wind turbine bearing can be diagnosed by analyzing the acquired instantaneous energy slice bispectrum. The analyzed results of wind turbine bearing fault signals from the experimental rig and the actual engineering testing field demonstrate the feasibility and the superiority of this novel detection method.