Abstract:  The rolling bearing was one of the most prone to failure components inside the generator. In order to evaluate the operating status of the bearings and ensure the normal operation of the generator set, based on wavelet packet complementary aggregation empirical mode decomposition and variational Bayes hidden Markov model, a method for rolling bearing health status evaluation was proposed. First, the wavelet packet-complementary empirical mode decomposition for signal denoising and feature extraction was used. The extracted feature signals in a healthy state were input into the variational Bayes-Hidden Markov model for model training. Then the characteristic signal of the full-period bearing was input into the model. The model was tested with bearing degradation experimental data and experimental rig measured data. The research results show that the log-likelihood probability value output by the model can be used as an indicator for state evaluation. The changes in the log-likelihood probability curve can reflect the health, deterioration and failure of the bearing. Therefore, the experimental method has good applicability.

Key words:  rolling bearing; wavelet packet; complementary aggregation empirical mode decomposition(CEEMD); variational Bayes(VB); hidden Markov model(HMM)