Fault classification of rolling bearing based on reconstructed phase space and Gaussian mixture model

Rolling bearings are common and vital elements in rotating machinery and vibration signal is a kind of effective mean to characterize the status of rolling bearing fault and its severity. In this paper, a novel method is introduced to realize classification of fault signal without extracting feature vector preliminarily. By estimating the time delay and embedding dimension of time series, vibration signal is reconstructed into phase space and Gaussian mixture model (GMM) is established for every kind of fault signal in the reconstructed phase space. After these models are built, classification of fault signal is accomplished by computing the conditional likelihoods of the signal under each learned GMM model and selecting the model with the highest likelihood. By testifying of vibration signal under different kinds of bearing status, it is proved that this method is effective for classifying not only fault types but also fault severity. Moreover, all parameters needed in this method could be obtained by analyzing the time series directly so it is very suitable for industry application.