A health indicator construction method based on the information entropy of specific frequency band and GGRBM

摘要

The traditional physics health indicator(PHI) has poor monotonicity and is insensitive to the incipient fault. In order to solve this problem, we propose a virtual health indicator(VHI) construction model, which is based on the information entropy of specific frequency domain and restricted Boltzmann machine. This model consists of two layers: physics health indicator construction layer and features fusion layer. In the physics health indicator construction layer, we propose a method based on the information entropy of specific frequency domain to construct the physics health indicator. In the features fusion layer, the monotonicity metric is used to select parts of the PHIs to form the feature set, which is then fused by the Gaussian-gaussian restricted Boltzmann machine to get a VHI. The experiment results show that the proposed model could make a VHI, which can effectively enhance the monotonicity of the degradation assessment curve.