Bearing fault diagnosis method based on maximum noise ratio kurtosis product deconvolution with noise conditions

Abstract

The weak signal changes caused by early bearing failures and complex transmission path attenuation result in low signal-to-noise ratio of vibration status signals, which makes it difficult to effectively extract faults features. Blind deconvolution methods have excellent performance in fault impulses extraction. The factors affecting performance of deconvolution include selection of objective function, periodicity detection, and method of finding the optimal filter. To better solve these three problems, maximum noise ratio kurtosis product deconvolution (MNRKPD) is proposed. In this method, the maximization of noise ratio kurtosis product (NRKP) is proposed as objective function, and signal envelope cycle kurtosis (SECK) is constructed to estimate fault period. Artificial rabbits optimization (ARO) is applied to find optimal filter. The effectiveness of MNRKPD is verified by simulations and experimental validations. Compared with traditional deconvolution methods shows that MNRKPD can overcome strong noise and harmonic effect effectively and can extract periodic fault impulses more effectively.