A maximum cyclic impulses energy ratio deconvolution fused with enhanced envelope derivative operator frequency spectrum for early fault diagnosis of rolling bearing

Abstract

Abstract Blind deconvolution is a powerful tool for rolling bearing fault diagnosis. As one of deconvolution methods, maximum second-order cyclostationarity blind deconvolution (CYCBD) is proved to be effective in extracting bearing fault characteristics. However, the performance of CYCBD method is greatly compromised by setting of fault characteristic frequency (FCF) in advance. Moreover, its performance decreases dramatically under the interference of random shocks and strong noise. To address these issues, a new deconvolution method, named as maximum cyclic impulses energy ratio deconvolution (MCIERD) fused with enhanced envelope derivative operator frequency spectrum (EEDOFS) is proposed in this research. In this method, the EEDOFS is proposed to estimate the FCF. Furthermore, the cyclic impulses energy ratio (CIER) is employed as the deconvolution indicator. Moreover, the hybrid firefly and particle swarm optimization algorithm is used to solve the optimal filter coefficients by maximizing the CIER. Simulation results show that EEDOFS exhibits a greater robustness in estimating FCF accurately under strong interferences and MCIERD performs well in extracting fault cyclic impulses under the interference of heavy noise and random shocks. Finally, three run-to-failure bearing datasets are employed for experimental validation, whose results demonstrate the effectiveness of EEDOFS in accurate estimating FCF and identifying the early bearing fault. Meanwhile, MCIERD fused with EEDOFS is proved to have greater advantages in extracting early bearing fault feature.