Abstract: The key problem of the loose parts monitoring system for primary loop of pressurized water reactor is how to extract loose parts impact signal from complex background noise. A blind source separation algorithm based on maximum signal to noise ratio was used to extract loose parts impact signal and background noise of power water reactor from complex mixed-signals, and an evaluate method using similarity coefficients was used to evaluate the separation effect of the algorithm. The maximum value of signal to noise ratio function was obtained according to the algorithm when separated signals were independent. The research results show that loose parts impact signal and background noise from primary loop of pressurized water reactor can be separated efficiently and accurately with the algorithm.