A diagnostic system for identifying accident conditions in a nuclear reactor

The objective of this study is to develop a system, which assists the operator in identifying an accident quickly using ANNs that diagnoses the accidents based on reactor process parameters, and continuously displays the status of the nuclear reactor. A large database of transient data of reactor process parameters has been generated for reactor core, containment, environmental dispersion and radiological dose to train the ANNs. These data have been generated using various codes e.g., RELAP5—thermal-hydraulics code for the core. The present version of this system is capable of identifying large break LOCA scenarios of 220 MWe Indian PHWRs. The system has been designed to provide the necessary information to the operator to handle emergency situations when the reactor is operating. The diagnostic results obtained from ANNs study are satisfactory.

Research highlights▶ Neural networks based diagnostic system has been developed to identify transients quickly, estimate the source-term and assist the operator to take corrective actions during abnormal situations in 220 MWe PHWRs. ▶ The transient data for the break scenarios ranging from 20% to 200% has been generated using RELAP5 and CONTRAN codes. ▶ 32 break scenarios of large break LOCA in inlet and outlet reactor headers with and without ECCS have been analyzed using artificial neural networks. ▶ A few break scenarios were directly predicted without being trained earlier. Test results obtained from ANN are within the acceptable range.