LASSO reconstruction scheme for radioactive source distributions inside reactor building rooms with spectral information and multi-radionuclide contaminated situations
Shi, W.*; Machida, Masahiko ; Yamada, Susumu ; Yoshida, Toru*; Hasegawa, Yukihiro*; Okamoto, Koji*
Clarification of radioactive source distributions is one of the most important steps in initial decommissioning of not only normally shutdown reactors but also damaged ones by accidents like Fukushima Daiichi Nuclear Power Plants (FDNPP). Generally, since radioactive hot spots are restricted into specific areas in normal operating conditions, the clarification scheme can be mapped onto the inverse estimation in sparse source distributions. On the other hand, the fact that radioactive hot spots are largely spread in unknown manner as seen in FDNPP motivates to construct an inversion scheme in non-sparse source conditions. Thus, a reconstruction scheme applicable to both sparse and non-sparse radioactive distributions is highly in demand. In addition, a variety of radionuclides is produced in reactors. Thus, we also need a scheme to distinguish each source distribution in mixed multi radionuclides. In this paper, we confirm that the inverse estimation scheme using Least Absolute Shrinkage and Selection Operator (LASSO) method with spectral information commonly shows excellent performance in the above all situations. The proposed LASSO scheme with the spectral information enables to reduce the number of measurement points in sparse conditions, while information proliferation by sensing the spectrum makes it possible to directly reconstruct source distribution as almost solvable problems in non-sparse ones. Moreover, the LASSO scheme allows to reconstruct the source distribution of each potential radionuclide in multi-radionuclide coexisting situations. Consequently, we confirm that the LASSO scheme to reconstruct radioactive sources is promising for the future nuclear decommissioning projects widely from normally shutdown reactors to damaged ones like FDNPP.