A method for extracting weak impact signal in NPP based on adaptive Morlet wavelet transform and kurtosis

The monitoring and detecting of loose parts in the reactor coolant system is vital for the safe operation of a NPP (Nuclear Power Plant). The impact signal stimulated by the loose part hitting on other parts is used to analyze the mass and the position of the loose part. However, the impact signal is always interfered by the strong background noise caused by the vibration of primary circuit and other components. In order to remove the interfering noise and enhance the weak impact features, a modified denoising and extracting method based on adaptive Morlet wavelet transform and kurtosis is proposed in this paper. Firstly, the Morlet wavelet parameters are optimized using a modified algorithm based on the Shannon entropy. Then, the wavelet coefficients that contains the most information of impact are selected using the kurtosis index. Finally, after denoising the selected coefficients using the adaptive soft-thresholding method, the impact signal is reconstructed through the ICWT. The proposed method is tested through simulation experiment which takes steel balls as the real loose parts, and results show that it maintains good performance under low SNR. In addition, the proposed method is compared with another two denoising methods and shows much greater performance.