Abstract:Bearings are one of the key components in rotating machinery. Therefore, it is important to assess the performance degradation degree of bearings for making maintenance plans and preventing unexpected defects and breakdowns during operation. In this paper, we present a novel bearing performance degradation assessment methodology based on wavelet packet transform (WPT) and hidden Markov models (HMMs). WPT is used to extract features from vibration signals of bearings, and the node energies and the total energy are selected as features. An HMM is trained using the data under normal condition and then the trained HMM is used to assess the performance degradation degree of bearings quantitatively. A bearing accelerated life test is performed to validate the proposed methodology. The experimental results show that the proposed methodology is feasible and effective