Research on signal-noise separation algorithm of loose part monitoring signals based on independent component analysis

When the signal-noise ratio of the mixed signals detected by the nuclear reactor loose part monitoring system is very low, the frequency characteristics of weak impact signals is difficult to get by traditional denoising method. In this paper, some relevant theory and algorithm of Independent Component Analysis is studied, and ICA technology is introduced into the signal-noise separation of the reactor loose parts. The impact signal of steel ball is collected from the simulated experimental apparatus. It is constructed into simulation mixed signals with a White Gaussian Noise and a coloured noise. Kurtosis, based on Gaussian function, is chosen as the target function of ICA. The algorithm is applied to the signal-noise separation of the simulation mixed signals after preprocessed, and is proved effective. (authors)