Weak fault detection of rolling element bearing combining robust EMD with adaptive maximum second-order cyclostationarity blind deconvolution

Abstract

To solve the difficulty in weak fault detection of rolling element bearing (REB), a fusion method by combining robust empirical mode decomposition (REMD) with adaptive maximum second-order cyclostationarity blind deconvolution (AMCBD) is proposed in the paper. The advantage of REMD in determining the optimal iteration number of a sifting process and the advantage of AMCBD in setting the key parameter (targeted cyclic frequency or fault period) appropriately are utilized comprehensively by the proposed method. Firstly, in view of the multi-component and modulation characteristic of the vibration signal of REB, REMD is used to extract the useful component from the multi-component and modulated signal. Then, AMCBD is used to process the selected useful component to further highlight the cyclostationary and impulse characteristics of the vibration signal of faulty REB. Compared with traditional maximum second-order cyclostationarity blind deconvolution (MCBD) method, AMCBD has the advantage of no needing prior knowledge of the faulty REB such as the targeted cyclic frequency or fault period. At last, envelope spectral (ES) is applied on the signal handled by AMCBD and satisfactory fault extraction feature result is obtained. Effectiveness of the proposed method is verified through simulated, experimental, and engineering signals, and its superiority is also presented through comparison study.