摘要
质子引发的散裂反应的产物截面数据是许多核反应应用中的关键基础,然而由于散裂反应涉及到的体系复杂且反应能量跨越3~4个数量级,现有的理论模型在截面预测上的准确性仍有不足。为了获取更高的预测精度,采用贝叶斯神经网络(Bayesian Neural Networks,BNN)方法,针对质子与铜(Cu)靶材的散裂反应截面进行建模和分析。与传统方法相比,BNN能够通过学习实验数据提供高精度预测,并且量化预测中的不确定性。基于EXFOR数据库的实验数据,构建的BNN模型在多个能量区间内准确预测了反应截面,并与实验数据进行了验证。结果表明,BNN模型在广泛能量范围内具有良好的预测能力,尤其在数据不足的区域,通过合理的置信区间量化了预测结果的不确定性。
The product cross-section data from proton-induced spallation reactions serves as a crucial foundation for many nuclear reaction applications.However,due to the complexity of the systems involved and the wide energy range of spallation reactions and the reaction energy exceeding 3-4 orders of magnitude,current theoretical models still face limitations in accurately predicting cross-sections.In order to obtain higher prediction accuracy,Bayesian Neural Networks(BNN)method was used to model and analyze the spallation reaction cross-sections of protons with a copper(Cu)target.Compared with traditional methods,BNN can provide high-precision predictions by learning from experimental data and quantify the uncertainty in predictions.Based on experimental data from the EXFOR database,the constructed BNN model accurately predicted reaction cross-sections across multiple energy ranges,and is verified with experimental data.Results show that the BNN model demonstrates excellent predictive capability across a broad energy spectrum and,especially in regions with sparse data,effectively quantifies the uncertainty through reasonable confidence intervals.
作者
黄永顺
黄美容
张苏雅拉吐
HUANG Yongshun;HUANG Meirong;ZHANG Suyalatu(College of Physics and Electronic Information,Inner Mongolia Minzu University,Tongliao 028043,China;Institute of Nuclear Physics,Inner Mongolia Minzu University,Tongliao 028043,China)
出处
《内蒙古民族大学学报(自然科学版)》
2025年第2期16-22,共7页
Journal of Inner Mongolia Minzu University:Natural Sciences Edition
基金
国家自然科学基金项目(12365018)
内蒙古自治区高等学校青年科技英才支持计划项目(NJYT23109)
内蒙古自治区高等学校创新团队发展支持计划项目(NMGIRT2217)
内蒙古自治区直属高校基本科研业务费项目(GXKY22061)
内蒙古自治区自然科学基金项目(2023MS01005)。
关键词
贝叶斯神经网络
余核截面
natCu
Bayesian Neural Networks
residual core cross-section
^(nat)Cu
