Development of Radiation-Distribution Estimation-System with Higher Precision Using Deep Neural Networks and Photon Transport Simulator and Its Implementation on Edge Device

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K11991
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61040:Soft computing-related
• Research Institution: Kagawa University
• Principal Investigator: Fujimoto Ken'ichi 香川大学, 創造工学部, 教授 (20300626)
• Co-Investigator(Kenkyū-buntansha): 阪間 稔 徳島大学, 大学院医歯薬学研究部(医学域), 教授 (20325294)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 放射線強度分布推定 / 光子飛跡シミュレータ / 深層学習システム / 敵対的生成ネットワーク / データ拡張 / エッジデバイス実装

Outline of Final Research Achievements

This study aims to establish the basic technology for a portable radioactivity monitoring device for planning an efficient soil decontamination plan. We developed a machine learning system (deep neural network) that estimates the distribution of radioactive soil contaminants in the depth direction with high accuracy. We assumed a virtual situation in which radioisotopes are placed in the soil and simulated typical cases using a PHITS (Particle and Heavy Ion Transport code System). According to the data sets generated by the PHITS, a large amount of training data for the deep neural network was generated using generative adversarial networks. We created a deep neural network that can make high-precision estimates for the training data and implemented it on an edge device. These indicated that the basic technology for a soil radioactivity monitoring device with high-precision estimation was established.

Free Research Field

数理工学

Academic Significance and Societal Importance of the Research Achievements

本研究では，放射線計測分野における標準的ツール（光子飛跡シミュレータ）を用いて推定した放射線センサの応答特性に基づいて，土壌内の放射線強度深度分布を高精度に推定できる深層学習システムを構築した。本研究成果は，ソフトコンピューティングの立場から放射線計測技術の高度化に寄与する点において重要である。また，放射能汚染土壌の除染計画や安全評価の効率化に寄与できるだけでなく，仮置場・中間貯蔵施設における放射性物質の適正な管理など，社会全体の放射線リスク管理技術の向上にも貢献できる。