Identification of Cold-Leg Break Size in LOCA Accident using Artificial Neural Networks and Simulation Database

Abstract

The most widely studied LOCA (Loss of Coolant Accidents) is a rupture of a cold leg pipe causing the Reactor Cooling System to depressurize first, with different break sizes corresponding to the change in trigger signal from the Instrument and Control System (I&C System) such as pressure, temperature, power, pressure vessel water level, etc. is different. Therefore, the response of nuclear power plant varies considerably with the size of break. To mitigate the consequence of LOCA with a given break size, it is necessary to design the emergency core coolant systems so that the fuel is cooled efficiently during all phases of the accident. Therefore, the size of rupture needs to be detected and identified as soon as possible right after reactor scram. To achieve this goal, this study is conducted to investigate the applicability of artificial neural networks (ANN) for recognizing LOCAs, especially identifying the rupture sizes of the LOCAs according to the changes of operational parameters of VVER-1000 nuclear power plant. This study mainly focuses on building, training, and optimizing the artificial neural networks using simulation databases obtained from the RELAP5 simulation program for VVER-1000 reactor technology. Results clearly showed the potential application of ANN-based model for detecting the break size even with uncertainty of input parameters added.