Innovations-based method for DC-signal failure detection. Application to NPP.

摘要

A state-space approach method for failure detection in DC-signals is described. The detection system receives data from sensors as input and estimates the DC state by means of Kalman filtering methodology, where the data assumed to be obeying a non-linear ARE model. The cumulative sum test (CUSUM) is implemented in real-time to the estimated DC state resulting from the Kalman filter as the test is in particular suitable for jump or pulse type anomalies in those signals subject to DC-level change. To test the effectiveness of the method Borssele NPP DC signals with operational changes were used. Real-time test observations showed operational effectiveness of the method. (author). 5 refs.; 5 figs. (Atomindex citation 23:075767)