Failure detection in power stations. Application to nuclear reactors

The development and application of a technique for the automatic diagnosis system of a nuclear power station is presented. The aim of this study is to design a computerized system to detect a malfunction in the core of a nuclear power plant. After a review of anomaly detection methods, a diagnosis method based upon internal representation is chosen. This method is characterized by the use of a physical modeling of the nuclear core. A systematical investigation is achieved in order to retain the convenient structure of the mathematical model. A digital simulation of several identification algorithms is performed to obtain the most efficient one. An on-line identification method using sensitivity equations and generalized recursive least squares technique is developed. The diagnosis of a malfunction is obtained by comparing in real time the reference parameters value corresponding to a normal functioning to the actual value of an on-line identification scheme. An on-line computerized system, able to detect incipient failures, is designed. Applications, using real data, are given to demonstrate the efficiency of the proposed technique