Fission product yields (FPY) constitute critical infrastructure data for a wide range of nuclear applications. Accurately predicting FPY remains a significant challenge both experimentally and theoretically. Drawing inspiration from recent advancements in Bayesian neural networks within nuclear physics, as demonstrated by Wang et al. [Phys. Rev. Lett. 123, 122501 (2019)], we propose a physics-guided approach utilizing a multilayer perceptron that achieves highly accurate predictions of nuclear fission yields. The engineering and analysis of the network validate the robustness of this physics-driven neural network methodology for FPY. This approach also provides uncertainty quantification for the predicted yields. These findings underscore the effectiveness of machine learning in nuclear FPY prediction, enhancing the accuracy and reliability of theoretical models, and offering valuable insights for experimental scientists.

Close

PrevNext

View figure in article

• Fission
• Models & methods for nuclear reactions
• Nuclear astrophysics
• Nuclear binding
• Nuclear reactions