방사능재난 시 작업자 선량평가에 신속 활용 가능한 AI 기반 내부피폭선량평가체계 개발 = Development of a System for Estimating Internal Dosimetry Based on Artificial Intelligence Rapidly Applicable to Dosimetry of Workers Under Radiological Emergency

다국어 초록 (Multilingual Abstract)

An increase of frequencies occurring the internal exposure situations of radiationworkers is concerned due to the environmental change, which are the enlargement of usingradiation sources and the introduction of decommissioning of nuclear power reactor, of nuclearindustries in Korea. This paper aims at developing a system, which can be impossible for thepresent commercial codes, for internal dosimetry based on artificial intelligence capable forrapidly estimating and processing much measurement information into a bundle under theradiological emergency situations. For defining the assesment model using artificial intelligence,an automatic system for generating database for intake scenarios and input values applicable tothe artificial neural network learning has been constructed by applying with the recommendationof ICRP, OIR and IDEAS. The artificial neural networks have been classified with two model, thatis, the case of knowing intake time and unknowing it. And, architectures for these models havebeen constructed for assessing the committed effective dose, and committed effective dose andintake time, respectively. Loss functions for two models have converged and their over-fitting hasnot occurred, and a validity of the system for internal dosimery based on artificial intelligence hasbeen achieved. And, the validity of the program for internal dosimetry has been also performedby using learning results of the artificial neural network. The accuracy of R2 score is to be about0.998 and this system based on artificial intelligency can be then reliable for internal dosimetry.