Scholarlink

Abstract <p>Estimation of the effective dose and unfolding the spectrum of neutrons at nuclear power facilities and charged particle accelerators is complicated due to absence of direct methods for detecting neutrons and the need to register secondary particles. The main difficulties are related to the wide energy range of neutrons from 1 meV to several hundreds MeV, complex dependence of the neutron interaction cross section on energy. One of the main devices used for neutron spectrometry is the Bonner multi-sphere spectrometer. The measurement results and the neutron spectrum, discretized on the energy grid (or decomposed into basis functions) are tabular data. However, due to the limited set of moderator spheres and correlations in its response functions, the number of input features is limited. In this paper, it is proposed to transform the original scalar continuous features into vectors. Then unfold the spectra for the transformed features using deep learning models included in the Mambular framework. The models quality metrics are compared with the automated machine learning (AutoML) frameworks that implements a set of linear and decision tree-based regression models (LightGBM, CatBoost, and random forest). For training and validation a set of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(5\times 10^{5}\)</EquationSource> <!--BPhysMGU2570283Chizhov-m1--> </InlineEquation> synthetic spectra was generated, modeled as a superposition of four weighted components describing the spectra of thermal, epithermal, fast, and high energy neutrons. A comparison was made with calculated and measured spectra from the IAEA compendium database and open access papers: 375 spectra in total. The uncertainty of spectra unfolding was estimated using the Monte Carlo method, in which random perturbations were introduced into the input data. The model was trained in the JINR Multifunctional Information and Computing Complex.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Neutron Spectrum Unfolding Using Deep Learning Models for Tabular Data

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • K. A. Chizhov,
  • A. A. Bely

摘要

Abstract

Estimation of the effective dose and unfolding the spectrum of neutrons at nuclear power facilities and charged particle accelerators is complicated due to absence of direct methods for detecting neutrons and the need to register secondary particles. The main difficulties are related to the wide energy range of neutrons from 1 meV to several hundreds MeV, complex dependence of the neutron interaction cross section on energy. One of the main devices used for neutron spectrometry is the Bonner multi-sphere spectrometer. The measurement results and the neutron spectrum, discretized on the energy grid (or decomposed into basis functions) are tabular data. However, due to the limited set of moderator spheres and correlations in its response functions, the number of input features is limited. In this paper, it is proposed to transform the original scalar continuous features into vectors. Then unfold the spectra for the transformed features using deep learning models included in the Mambular framework. The models quality metrics are compared with the automated machine learning (AutoML) frameworks that implements a set of linear and decision tree-based regression models (LightGBM, CatBoost, and random forest). For training and validation a set of \(5\times 10^{5}\) synthetic spectra was generated, modeled as a superposition of four weighted components describing the spectra of thermal, epithermal, fast, and high energy neutrons. A comparison was made with calculated and measured spectra from the IAEA compendium database and open access papers: 375 spectra in total. The uncertainty of spectra unfolding was estimated using the Monte Carlo method, in which random perturbations were introduced into the input data. The model was trained in the JINR Multifunctional Information and Computing Complex.