Condition Analysis of a High-Speed Train Based on Similarity Ratio and MDBN

Abstract

In order to make full use of the complementary information of multichannel vibration signals, this paper proposes a method of channel selection based on the similarity ratio and constructs a multichannel deep belief network (MDBN) for condition analysis of a high-speed train. First, fast Fourier transform (FFT) coefficients of the signals of all channels are extracted. Then, the similarity ratio of FFT coefficients of each channel is calculated, and a number of channels with a large similarity ratio are selected. Finally, the MDBN model is constructed to learn the features of the selected multichannel data and recognize conditions, and the feature fusion of multichannel data is realized in the common layer. The experimental results show that the similarity ratio is effective to select the channels with rich feature information, the feature learning ability of MDBN is better than DBN, and the condition recognition rate of MDBN is higher than other state-of-the-art methods.