Application of deep neural network for generating resonance self-shielded cross-section

Abstract In this paper, the deep learning based on the artificial neural network (ANN), which is referred to as the deep neural network (DNN), is adopted to build a new model for the generation of the resonance self-shielded cross-sections (XSs). In this model, using the dataset generated from the pin-based ultra-fine-group (UFG) calculations under a multi-dimensional parameter table, the multi-layer DNN is trained to learn the underlying relationship between resonance self-shielded XSs and correlated parameters. Then the trained DNN is used for further practical calculations, which takes a negligible computing time. The computing accuracy of this model is tested through the generated datasets and practical PWR problems, and numerical results show that the new model is a promising approach for the generation of the resonance self-shielded XSs.