Modified central frequency mode decomposition for the fault diagnosis of rotating machinery

Abstract

Abstract Central frequency mode decomposition (CFMD) is a promising tool for complex mechanical signal processing. Some characteristics of CFMD are disclosed by performing a detailed discussion on its decomposing theory in this study. As a result, three deficiencies of CFMD are found through the characteristic analysis, including the low accuracies of detected central frequencies (CFs), the filters with too wide bandwidth, and the excessive number of the decomposed modes. To address these issues, a modified CFMD (MCFMD) method is proposed for enhancing its performance of fault diagnosis. First, an exchange weighting function is defined to improve the accuracies of detected CFs. Then, the combination of the detected CFs and the separating information of modes is used to optimize the structures of filters for excluding the interferential information as much as possible. Furthermore, the number of the decomposed modes are compressed based the properties of the fault-related modes to facilitate the implementation of the diagnosis tasks. Lastly, a product envelope spectrum is introduced for further enabling the fault characteristic frequency in the fault-related modes to be more prominent while suppressing other noises. Analysis results verified the effectiveness of the MCFMD and its superiority over some existing advanced methods in the fault diagnosis of rotating machinery.