Applied Research On Model-based Fault Diagnosis Method

Since Three Mile Island accident and Chernobyl accident, the problem of safely running of nuclear power plant(NPP) is more concerned by nuclear energy community.Many efficient studies have been done.The method study of NPP fault diagnosis is focus on knowledge-based method. Methods of fault diagnosis using Neural Network, Genetic Algorithms, pattern recognition and expert system are highlights of study now. The model-based fault diagnosis method needs very accurate model, and accurate modeling of NPP is out of the question, so the study of NPP fault diagnosis has no regard for the method. To this point, Robust fault detection technology is introduced. The idea of it is that the fault detection filter design for NPP is sensitive to sensor failure, actuator failure or component failure,and not sensitive to the uncertainty of model, disturbance and noise. Based on Unkown Inputs Observer(UIO) method and based on Strong Tracking Filter(STF) method have excellent robustness. Theory and design process will be intrduced in detail. Many simulation studies are made to illustrate the effectiveness of the both.In nuclear power station, the mission of steam turbine governing system is to insure the quality and quality of electrical power. Once it has catastrophic failure, the turbine will be possibly damaged and spread to the whole power network.For the jam fault, STF is used to detect and diagnose.In the future, depending on the study of Robust fault detection technology, model-based fault diagnosis method will be more applied to the fault diagnosis of NPP.