Frequency slice graph spectrum model and its application in bearing fault feature extraction

Abstract

Complex electrical equipment in harsh environments can pose a major threat to the health of key components such as bearings. Weak features are often hidden in many interferences, which makes it very difficult to extract fault features of mechanical parts. This paper proposes a frequency slice graph spectrum model (FSGS Model), which aims to search for characteristic information that matches bearing faults from the enhanced data dimension. Firstly, the frequency slice groups in the time–frequency domain are used as graph structure vertices to construct a Laplacian matrix. On the basis of retaining the time domain features, the connection between the fault features in the spectrum is mined. Secondly, time–frequency graph Fourier clustering spectrum is established. The order of the clustering spectrum is tower-decomposed and reconstructed through a binary tree structure, providing different order combinations. In order to increase the recognition accuracy, the harmonic spectral kurtosis (HSK) is used to select the optimal reconstructed FSGS spectrum band. The feasibility of the proposed method is verified by constructing simulation signals, and it is applied to the fault diagnosis of the inner and outer rings of bearings. The effectiveness of this method is verified by comparison with three methods of Fast Kurtogram, Autogram, and Harmogram.