Estimation of HFIR [High-Flux Isotope Reactor] core flow rate using a back propagation network

The corrosion of aluminum dispersion fuel of Oak Ridge National Laboratory's High-Flux Isotope Reactor (HFIR) has recently been under review. The surface temperature of the fuel under high-flux conditions increases as the oxide layer grows or crud deposits on the fuel. This additional temperature rise in turn speeds the oxide buildup creating an unstable situation. Since the low thermal conductivity of the corrosion product film may compromise the thermal limits of the reactor, and also, erosion of the aluminum plates may lead to leakage of fission products from the plates, a careful monitoring of corrosion would be useful for the HFIR. In addition, an examination of HFIR operating data indicates that excessive corrosion has a noticeable effect on core flow rate and pressure drop. Typically, as the thickness of oxide layer increases, the flow rate gradually decreases and the pressure drop increases; therefore, it has been recommended that the core pressure drop and flow characteristics be monitored during each fuel cycle. The purpose of this study is to investigate the feasibility of applying neural networks to predict the flow rate in the HFIR core using a set of other HFIR operating data. The back propagation network (BPN) is a multilayer, fully connected heteroassociative network. During the training stage, the BPN learning algorithm computes the weights between pairs such that the difference between the actual output and the network output is minimized in a least-squares sense