A WAV-ICA based on-line fault diagnosis method for redundant sensors in nuclear power plants is proposed in the paper.In order to eliminate the effect of the electric noise and transmission noise,the high-frequency noise in redundant sensors signal is filtered with wavelet decomposition method at first.Then the Independent Component Analysis(ICA) is performed and the component with interest is selected for reconstructing the estimated signal,so as to detect the fixed bias and abrupt faults of the sensors,as well as their drift failures.A history data set from five redundant pressurizer pressure sensors in a NPP is adopted to validate the proposed method.The result shows that compared with the Simple Analysis(SA) method and direct ICA method,the proposed method can detect the fixed bias and drift faults of the sensors on-line effectively,with lower rate of misdiagnosis and better robustness.