Intelligent cross-condition fault recognition of rolling bearings based on normalized resampled characteristic power and self-organizing map

Abstract

Intelligent bearing fault recognition under nonstationary conditions is still a challenge. This paper presents a novel intelligent cross-condition bearing fault recognition scheme. In this scheme, we propose a normalized resampled characteristic power (NRCP) feature, which is constructed based on the pulse-based order spectrums. Based on NRCP feature, the whole fault recognition strategy is developed. First, the resampled signals are obtained by pulse-based order tracking technique, and the order spectrums are produced by the joint application of Hilbert transform and fast Fourier transform. Second, the NRCP feature space is constructed based on the order spectrums. Then, the Laplacian score (LS) algorithm is applied to refine the NRCP features. Finally, the new features are fed into self-organizing map (SOM) to identify the health conditions of rolling bearings. The proposed method is experimentally validated to be able to differentiate health, outer race fault, inner race fault, and multiple fault bearings.