Performance improvement of artificial neural networks designed for safety key parameters prediction in nuclear research reactors

Abstract

The present work explores, through a comprehensive sensitivity study, a new methodology to find a suitable artificial neural network architecture which improves its performances capabilities in predicting two significant parameters in safety assessment i.e. the multiplication factor k eff and the fuel powers peaks P max of the benchmark 10 MW IAEA LEU core research reactor. The performances under consideration were the improvement of network predictions during the validation process and the speed up of computational time during the training phase. To reach this objective, we took benefit from Neural Network MATLAB Toolbox to carry out a widespread sensitivity study. Consequently, the speed up of several popular algorithms has been assessed during the training process. The comprehensive neural system was subsequently trained on different transfer functions, number of hidden neurons, levels of error and size of generalization corpus. Thus, using a personal computer with data created from preceding work, the final results obtained for the treated benchmark were improved in both network generalization phase and much more in computational time during the training process in comparison to the results obtained previously.