Surrogate Model of Solved Poisson Kriging Method for Radiation Field Reconstruction

Abstract

Reverse reconstruction methods for the radiation field do not require information on the radioactive source and are capable of constructing the radiation field using a small amount of monitoring data, showing huge significance for radiation protection. However, in previous studies, inverse reconstruction methods have given less consideration to variations in the time dimension. Herein, the principle of the Poisson Kriging method solved by the surrogate model has been analyzed, and the Poisson Kriging method has been applied to the inverse reconstruction of two-dimensional radiation fields at different moments. On this basis, this work also investigated the effects of the principal function and correlation coefficient model on the objective function, the results of which demonstrate that the quadratic polynomial principal function and the Gaussian model correlation coefficient have good stability and convergence. Compared with the inverse distance weighting methods and the radial basis function methods, the Poisson Kriging method has smaller errors, showing that it is more suitable for reconstructing complex radiation fields. Finally, the Poisson Kriging method was applied to the Fukushima nuclear accident radiation field calculation. The Pearson correlation coefficient of its results was r = 0.49, reflecting the validity of this method. Our work provides a calculation method for the spatial distribution and trend of the radiation field in the early stages of a nuclear accident, which is helpful for furthering radiation protection and emergency responses to nuclear accidents.