A novel adaptive blind deconvolution algorithm: application to feature extraction of weak faults in RV reducer gears

Abstract

Aiming at the problem that the non-stationary and nonlinear weak fault signal of RV (rotate vector) reducers is hard to extract fault features due to the influence of noise and transmission paths, as well as the selection of parameters for maximum correlation kurtosis deconvolution (MCKD) relies heavily on manual experience, this article proposes a fault feature extraction method based on parameter adaptive MCKD for the gear faults of RV reducers. Firstly, the sparrow search algorithm combining sine-cosine and Cauchy mutation（SCSSA）is used to adaptively search for the input parameters of MCKD and obtain the signal after deconvolution with the optimal parameters. Secondly, the deconvoluted signal is subjected to ensemble empirical mode decomposition (EEMD) to obtain modal components on different frequency bands. Finally, calculate the multi-scale fuzzy entropy (MFE) of each component, constructing a MFE feature set vector, and input the feature vector into the support vector machine (SVM) for fault classification and recognition. The experimental analysis and verification results both indicate that the proposed method can adaptively enhance the weak impact components in the gear signals of the RV reducer, effectively extracting weak fault features disturbed by noise. Compared with minimum entropy deconvolution (MED), multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) and MCKD, the proposed method has improved identification rate by 17.50%, 10.63% and 15.63%, respectively. In addition, in comparison to multiverse optimization (MVO) and particle swarm optimization (PSO) algorithms, the SCSSA exhibits superior performance when optimizing MCKD parameters, offering faster convergence speed, higher accuracy, and greater robustness.