An artificial neural network approach for the discordance sensor data validation for SCRAM parameters

In Fast Breeder Reactor (FBR), shutdown system is envisaged by Safety and Control Rod Acceleration Movement by using (SCRAM) signals. These SCRAM signals are realized with redundant triplicate sensors, which are made available at different locations of reactor. In this case sensors should be in healthy condition to run the reactor in trouble free manner. To know the health status of sensors a monitoring system is necessary. For this purpose, discordance supervision system is envisaged, to monitor the discordance among the SCRAM signal sensors and generate the alarm when discordance occurs. If discordance occurs, the sensor data validation is necessary to justify the discordance. The sensor data validation by knowledge based approach is simple and reliable. The discordance data is obtained from SCRAM signals. To validate these sensors data value, a neural network based approach is used. The proposed technique is used the data obtained from the coolant temperature monitoring system and relevant application is reported in this paper. The results of this investigation are discussed in this paper. (authors)