Edge Sample Enhancement Method for Bearing Fault Recognition

To improve the accuracy of fault diagnosis, this paper introduces an enhancement method for edge sample recognition in the time-frequency domain. First, the short-time Fourier transform is applied to obtain the samples of spectral content at each time index. The semi-supervised model, restricted Boltzmann machine, is utilized to identify the presence of the fault feature. The information entropy weight is then calculated by normalized exponential estimation, relying on the matching of the peak value between the information entropy and the initial classification. Finally, to smoothly enhance the probability of the edge sample recognition, the activation probabilities of the hidden layer neurons are adjusted by the information entropy weights. Its performance is validated on synthetic signals and real vibration signals, the results show that the proposed method can significantly enhance the accuracy of edge sample recognition.