Using artificial intelligence to predict and validate nuclear data

使用人工智能预测和验证核数据

• 批准号: 2889665
• 金额: --
• 依托单位: University of Cambridge
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2889665
• 关键词: Using; artificial; intelligence; predict; validate

Nuclear data underpins all of nuclear science and technology. Even the most complex and concisely written nuclear data analysis tools are unreliable and untrustworthy if they use old or un-benchmarked nuclear data.Nuclear data are comprised of cross sections, angular scattering probabilities, outgoing energy probabilities, reaction product multiplicities, fission yield data, reaction products and more. All of which are vitally important in the design and safety cases of nuclear devices.- Predicting the functional form of (n,2n), thermal, scattering cross sections, with little or no measurements, based on learning (many papers could be written based on the same methodology)- AI cross section lookup table based on learning cross section forms- AI uncertainty analysis - learn to predict cross sections of well characterised cross sections within bounds of uncertainty. Use variation of synapse weights to create a cross section probability distribution function, which can be converted to a statistical uncertainty.

核数据是所有核科学和技术的基础。即使是最复杂和最简洁的核数据分析工具，如果使用旧的或未标杆的核数据，也是不可靠和不值得信任的。核数据由截面、角散射概率、传出能量概率、反应产物多重性、裂变产额数据、反应产物等组成。所有这些在核装置的设计和安全案例中都是至关重要的。-预测(n，2n)、热、散射截面的函数形式，几乎没有测量，基于学习(许多论文可以基于相同的方法撰写)-基于学习截面形式的人工智能截面查找表-人工智能不确定性分析-学会在不确定的范围内预测特征良好的截面的截面。使用突触权重的变化来创建横截面概率分布函数，该函数可以转换为统计不确定性。