Abstract

One of the most critical challenges in rolling bearing diagnosis is dealing with weak fault signals that are buried in noises arising from environmental effects. To overcome this problem, the research described in this paper aims to develop a noise-robust feature extraction method, namely linear power-normalized cepstral coefficients (LPNCC), inspired by speech recognition based on auditory physiology. In this approach, for the cepstra from a feature extraction process, the squared envelope spectra are computed to find bearing characteristic frequencies. The performance of the proposed method is examined by studying simulation data in the presence of various levels of Gaussian background noises and through study of two experimental cases from the Case Western University dataset in the presence of impulsive noise and with a low signal-to-noise ratio (SNR), respectively. It can be concluded from the results that the proposed method has the potential to be utilized for robust bearing diagnosis in various noisy environments.