Real-time simulations to enhance distributed on-line monitoring and fault detection in Pressurized Water Reactors

Human error is one of the leading factors in nuclear accidents. However, on-line monitoring may help operators to detect and locate the abnormality on time, and thus it has the benefit of reducing misoperation, improve the safety of nuclear power plants and guide the maintenance plan. This work presents a distributed on-line monitoring and fault detection methodology based on modeling and simulation of nuclear plant's sub-units. This methodology utilizes quantitative calculation from the simulation model and combines the measurements gotten from instrumentation and control system to detect incipient faults. Under normal operation, the method allows real-time tracking to set dynamic thresholds. When a failure occurs, the corresponding parameters can be compared to detect the fault. Furthermore, the root cause, fault type and degree of the fault could be identified and calculated while considering compensation from the control system and trends of measurements. The performance of this method was evaluated and verified by applying it to the Reactor Coolant System of a PWR and comparing its result with the one obtained from Principal Component Analysis (PCA) method. The result of simulation analysis under steady operation and transients separately show the effectiveness and accuracy of this methodology when compared with PCA method; therefore, laying an essential foundation for the prediction of fault trends. (C) 2017 Published by Elsevier Ltd.