Stratified analysis of the magnetic Barkhausen noise signal based on wavelet decomposition and back propagation neural network

Because the wavelet transform has the characteristic of multi-scale analysis and it can characterize local feature of signals, this article uses the wavelet decomposition method to investigate the sensitivity of different time-frequency components of the magnetic Barkhausen noise (MBN) signal with changes in temperature and stress. After using the db5 wavelet with six layers to decompose the MBN signal and reconstructing low-frequency signal at each layer, the mean and RMS value are extracted and followed by a discussion on the relationship between features and the variation of the applied temperature and stress. It is found that within the elastic range of the sample, both of the mean and RMS values of the low-frequency reconstructed signal at each layer have inverse relationship with as compressive stress. The mean and RMS value of the low-frequency reconstructed signals decrease at the first to fourth layer, remain constant at the fifth layer and increase at the sixth layer as temperature increases, respectively. A new neural network model is built by taking the temperature, the mean and RMS values of MBN signals and the decomposition coefficients as the input and the stress as the output. It is shown that the proposed neural network model for stress measurement has higher accuracy than the former BP neural network models in which the wavelet decomposition is not used.