ERS: An Adaptive Spectral Analysis Method for Fault Diagnosis

Abstract

The development of spectral analysis methods is very rapid, but these methods rarely take into account the difference of feature extraction under strong and weak random noise. In this article, a new adaptive spectral analysis method called enhanced Ramanujan spectrum (ERS) is proposed to strengthen the ability of feature extraction and noise robustness. First, hybrid Ramanujan Fourier transform is used to improve the calculation accuracy and period recognition ability of discrete Fourier transform. Second, generalized Ramanujan spectrum (GRS) is used to obtain features in the frequency domain. Finally, the ERS can be adaptively constructed by the optimal GRSs in each segment to reduce the influence of random noise. The analysis results of rolling bearing fault signals show that ERS is an effective feature extraction method and can be used in fault diagnosis field.