基于贝叶斯神经网络分析质子引发的散裂反应产生的同位素截面

Isotopic cross section predictions based on bayesian neural network method in proton-induced spallation reaction

质子散裂反应产物截面数据是许多核应用领域的关键基础数据.不论在实验还是理论上,获得精确、完整的能量依赖的散裂产物截面都具有很大的挑战.运用贝叶斯神经网络方法(Bayesian Neural network,BNN)学习现有的质子散裂反应产生的同位素截面,并做出预测.BNN预测结果与实验测量截面吻合的很好,并展现出较好的模型泛化能力,可为核天体物理、加速器驱动的亚临界系统、质子治疗等领域提供参考价值.

Fragments productions in Spallation reactions are key infrastructure data for various applications.A Bayesia-n-n eural-network(BNN)approach is established to predict the fragment cross sections in the proton induced spallation reaction for systems.The predicted results of BNN show a good agreement to the measured results.The established method is suggested to benefit the related researches of nuclear astrophysics,nuclear radioactive beam source,ADS,proton therapy,etc.