Research on Fractional Lower Order Feature Extraction of Bearing Vibration Signals Under Alpha Stable Noise Conditions

Abstract

According to the performance degradation problem of feature extraction from higher-order statistics in the Alpha stable distribution noise, a new feature extraction method of rolling bearings under Alpha Stable Noise Conditions is proposed. Firstly, the non-stationary vibration signal of rolling bearings is decomposed into several product functions by Local Mean Decomposition (LMD) to realize signal stability. Then, the distribution properties of product functions in time domain is discussed by the comparison of heavy tails and characteristic exponent estimation. Fractional lower order p-function optimization is achieved by the calculation of the distance ratio based on K-means algorithms. Finally, faulty feature dataset is created by the optimal FLOS and lower dimensional mapping matrix to accurately and intuitively describe various faulty bearings. Since the Alpha stable noise is effectively suppressed, the presented method has shown nicer performance than the traditional methods in bearing experiments. Inking roller’s bearings are described precisely by fractional lower order feature extraction.