Construction and Evaluation of a Surrogate Model for Predicting Radiation Dose Rates for an Indoor Space with a Square Pillar

Abstract

In this paper, we construct a surrogate model for radiation dose rate predictions using simulation results and deep learning, specifically for an indoor space containing a square pillar, and verify its accuracy. We also demonstrate, based on the principle of superposition, that the surrogate model can predict the distribution of radiation dose rates in a space with multiple radiation sources. Furthermore, we propose a method to predict the radiation dose rates in a space with multiple square pillars and sources by using a corrected surrogate model. Based on these findings, we assess the feasibility of predicting the radiation dose rates in the reactor building with complex structures in real time and with high accuracy.