LASSO reconstruction scheme to predict radioactive source distributions inside reactor building rooms; Theory & demonstration

Shi, W.*; 町田 昌彦; 山田 進; 吉田 亨*; 長谷川 幸弘*; 岡本 孝司*

Proceedings of Waste Management Symposia 2023 (WM2023) (Internet), 8 Pages, 2023/02

Clarifying hot spots of radioactive sources inside reactor building rooms based on monitoring air dose rates is one of the most essential steps in decommissioning of nuclear power plants. However, the attempt is regarded as a rather difficult task, because information obtained by air dose rate measurements is generally not enough to inversely estimate contaminated distribution among a tremendous number of potential distributions inside complex reactor building rooms as far as one uses the conventional ways. Then, in order to successfully perform the inverse estimations on source distributions even in such ill-posed circumstances, we suggest that a machine learning method, least absolute shrinkage and selection operator (LASSO) is a promising scheme. Subsequently, we construct a simple room model and employ Monte Carlo simulation code, Particle and Heavy Ion Transport Systems (PHITS) to numerically test feasibility of LASSO inverse estimation scheme. Consequently, we confirm high reconstruction performance of the LASSO scheme in successfully predicting radioactive source distributions. In addition, we carry out uncertainty analysis for the inverse estimation and derive an error function describing uncertainty of the inverse estimation as a useful error estimator. Finally, we find that additional use of spectral information in the measurements can significantly decrease the number of measurement points for the present inverse estimation. In conclusion, LASSO scheme is a quite useful way to explore radioactive hot spots toward the future decommissioning of nuclear power plants.