Detection of Sensor Drifts using a Standardized Sum of Innovation Test for a Pressurizer in a Nuclear Power Plant

A statistical test method for detection of the sensor drift is proposed in this paper. The method generates the standardized sum of the innovations for a test statistic. The test statistic passes through a decision function defined by the test to decide whether the processed signal is drift or not. The distribution of the standardized sum of the innovations and the average sample number function of the test are derived. The average sample number function is compared using the sequence data simulated by a pressurizer model of a nuclear power plant with those of the Shewhart control chart, sequential probability ratio test, and generalized likelihood ratio test to analyze the performance. The results of the comparison show that the proposed test is a most powerful test among the algorithms under the unknown small change magnitude condition.