Classification of Pressure Drop Devices of Proto Type Fast Breeder Reactor through Seven Layered Feed Forward Neural Network

摘要

This paper presents a method to analyze the quality of pressure drop devices used for flow zoning in Prototype Fast Breeder Reactor (PFBR) by analyzing the occurrence of cavitations. In this work artificial neural network (ANN) has been used to classify the pressure drop devices as cavitating or not cavitating, under given operating conditions. A multi layer feed forward network with resilient back propagation algorithm has been used. The magnitude of root mean square (RMS) of the time signal acquired from an accelerometer installed downstream of various flow zones (totally 15) are fed as feature to the network for training and testing. Once adequately trained, the Neural Network based cavitation detection system would serve as an automated scheme for predicting the incipient cavitation regime and cavitation characteristics of a pressure drop device for a particular flow zone.