Abstract This paper deals with reconstruction of reactor states from measurements of fixed in-core Self-Powered Neutron Detectors (SPNDs) by Artificial Neural Networks (ANNs). Reactor states are defined by position of… Click to show full abstract
Abstract This paper deals with reconstruction of reactor states from measurements of fixed in-core Self-Powered Neutron Detectors (SPNDs) by Artificial Neural Networks (ANNs). Reactor states are defined by position of the control rods and burn-up of the fuel for Tehran Research Reactor. Position of control rods during fuel life-cycle should be monitored, because of their strong impact on neutron flux distributions and operational safety margins of the reactor. In the literature, position of control rods is reconstructed using axial distributions of neutron flux measured by in-core detectors in the vicinity of the control rods. In this research, ANNs are employed to determine position of the control rods and burn-up of the fuel, using 2D neutron flux measurements of the previously optimized SPNDs at central elevation of the reactor core. Also, sensitivity studies are performed on mapping performance of the ANNs with different architectures and training algorithms. The results indicate that a specific architecture of feed-forward neural networks, which are trained by Bayesian regularization back-propagation algorithm, has the best performance for the reactor states reconstruction. In addition, noise resistance capability of the developed ANN is evaluated with noisy data sets. The developed neural network has satisfactory noise resistance response, especially when trained by noisy data sets.
               
Click one of the above tabs to view related content.