Cooperation on radiation measurements for cross-border accidents, 2; Application of machine learning conversion method using monitoring data from Fukushima as teaching data

Sasaki, Miyuki  ; Sanada, Yukihisa   ; Lee, E.*; Joung, S.*; Ji, Y.-Y.*

Since the Fukushima Daiichi Nuclear Power Plant accident, some types of radiation monitoring have been conducted in Japan. Walking surveys and unmanned helicopter surveys have been conducted around the FDNPP to the ambient dose equivalent rates (air dose rate). The airborne radiation survey (ARS) by unmanned helicopter has the advantage of measuring large areas including forests. However, ARS has lower measurement resolution than walking surveys. Sasaki et al. constructed an artificial neural network (ANN) to convert ARS data into air dose rates at 1 m above ground level using data accumulated after the accident (Ref). It has been reported that the conversion using ANNs can convert to values closer to those measured on the ground than the conventional method. Showing that the ANN constructed from the Fukushima experience can be applied to detectors other than the radiation detector used to construct the ANN will greatly contribute to the future development of ARS. In this study, we investigated how to apply radiation measurement data acquired with a detector different from the radiation detector used to acquire the training data to an ANN that has already been constructed.