A feasibility study on application of redundancy sensors estimation technique to nuclear power plants

Traditional periodic maintenance techniques being employed in nuclear power plants usually fail to detect the potential degradation in performance of a sensor timely, and may increase workload and radiation exposure of the maintenance staff. The Redundancy Sensors Estimation Technique (RSET) to be presented in this paper is a non-invasive and in-situ monitoring technique based on measurement theory. The technique allows staff to monitor redundant sensors on-line and to assess their performance instantly. If such a REST can be applied to a nuclear power plant, it is expected that it will improve the safety of the plant and reduce the costs of operation and maintenance. This paper is concerned with a feasibility study on the REST's application to a nuclear power plant. The results show that the RSET, featuring a strong generalization ability, can offer state estimation and fault diagnosis to signals from redundant sensors accurately. By adding drift data to the test dataset, the RSET can determine the signal drift accurately. The core algorithm of RSET can be explained by mathematical formulas and has high prediction accuracy. In conclusion, RSET can detect the performance degradation of redundant sensors in advance during the operation of plant. (author)