Compressed Sensing Artificial Neural Network for Reactor Core Flux Mapping

Abstract

This paper presents a novel technique of signal recovery from sparse measurements based on compressed sensing (CS) and artificial neural network (ANN) and its application to core flux mapping in Advanced Heavy Water Reactor (AHWR). In large size nuclear reactors, neutron flux distribution undergoes continuous variation due to routine perturbations, nonuniform burn up at different locations, xenon oscillation, etc. An online flux mapping system (FMS) is needed to continuously monitor neutron flux distribution and display to the operator. FMS employs a suitable algorithm to estimate the core flux distribution from the measurements of only a few in-core detectors. The proposed method called CS-ANN methodology first uses CS technique for flux mapping along the vertical direction at the detector housing locations (1-D) and subsequently ANN technique in several horizontal planes (2-D) for estimating the 3-D neutron flux profile under different operating conditions. Error in the estimation using the proposed CS-ANN methodology has been compared with other existing methods and found to be significantly lower.