Compound Defects Feature Extraction Method of Rotate Vector Reducers Based on Optimized Maximum Second-Order Cyclostationarity Blind Deconvolution

A compound defects feature separation method for rotary vector (RV) reducer is proposed to address the difficulty in separating compound defect characteristics. The method utilizes a combination of peak detection of envelope frequency energy and multiscale permutation entropy (MPE) to optimize parameters for maximum second-order cyclostationarity blind deconvolution (CYCBD). In this article, the means is referred to as optimized maximum second-order CYCBD (OCYCBD). First, the defect signals are analyzed using peak detection of envelope frequency energy to estimate the cyclic frequency set of CYCBD. Second, utilizing the estimated cyclic frequency set and guided by MPE, the filter length of CYCBD is adaptively chosen. Finally, the separated defect components are analyzed using the envelope spectrum analysis to determine the defect types. The proposed method is demonstrated to effectively separate defect features through analysis of simulated signals. In addition, the feasibility of the OCYCBD is further verified by analyzing experimental data from a custom-built RV reducer test platform.