Research on unknown fault diagnosis of rolling bearings based on parameter-adaptive maximum correlation kurtosis deconvolution

Maximum correlation kurtosis deconvolution (MCKD) is an effective means of identifying the periodic impulses of fault signals. However, the multiple input parameters required by MCKD complicate the process of fault diagnosis. To overcome this drawback, a new method for identifying fault types based on a parameter-adaptive MCKD method is proposed. First, an improved adaptive variational mode decomposition is developed to denoise the raw signal. The improved method adopts the weighted envelope entropy, which is constructed by combining the envelope entropy with the kurtosis, allowing the salience of the denoising performance to be evaluated. Furthermore, the mean maximum correlation kurtosis is constructed to allow the specification of fault types and the corresponding parameters. Finally, two rolling bearing test datasets are used to demonstrate the strong adaptability of this method compared with other adaptive techniques. Published under license by AIP Publishing.

Automated summary

This paper introduces a new method for identifying fault types based on a parameter-adaptive MCKD method. By improving the adaptive variational mode decomposition and constructing weighted envelope entropy, this method demonstrates strong adaptability in fault signal identification and denoising compared to other adaptive techniques.