泊松克里金算法在核辐射场重构中的应用研究

为验证泊松克里金法在核辐射场重构中的应用效果，本文对泊松克里金算法在辐射场剂量率插值计算中的计算原理进行了研究，结合代理模型对泊松克里金算法进行了求解。通过SuperMC软件建立虚拟辐射场，对泊松克里金算法的参数影响进行了分析，筛选出二次多项式基函数和GAUSS误差相关模型。采用6150AD-b闪烁体探测器对实际放射源进行测量，并利用实测辐射场剂量率数据和福岛周边空间剂量率数据对泊松克里金算法在核辐射场重构中的可行性进行了验证。结果表明：小范围简单辐射场中，样本数据量为网格点数的1/10时，平均相对误差在10%以内；大范围辐射场中，辐射热点区域略有偏移，平均相对误差为141.69%。以上结果表明，该方法在快速、低成本解决未知放射源辐射场的重构问题中有一定的潜力。

Reconstruction of Nuclear Radiation Dose Field Using Alternative Poisson Kriging Algorithm

The investigation of radiation field distribution is very important in radiation protection, and rapid searching for radiation sources is necessary to minimize radiation damage. Using interpolation algorithm, the radiation dose field distribution of an unknown radioactive source can be reconstructed from a small amount of radiation monitoring data. In order to verify the application effect of Poisson-Kriging algorithm in nuclear radiation field reconstruction, the calculation principle of Poisson-Kriging algorithm in radiation field dose rate interpolation was studied, and the Poisson-Kriging algorithm was solved by surrogate model. Firstly, the parameters of the Poisson-Kriging algorithm were analyzed and selected, and the data of the parameter analysis came from the virtual radiation field dose rate data were constructed by SuperMC software. Based on this data, 100 dose rate data were randomly selected as samples, and the influence of the polynomial basis function and error correlation coefficient model in the Poisson-Kriging algorithm was analyzed. Then, a quadratic polynomial basis function and GAUSS correlation coefficient model suitable for constructing the radiation field were screened out. Finally, the algorithm was verified experimentally in two radiation fields with different ranges. The small-range radiation field was constructed by placing two sets of 137Cs radiation sources indoors, and using 6150AD-b instrument to measure the dose rate data. In this radiation field, as the number of samples increases, the hot spot area of the radiation field gradually converges, and the average error gradually decreases. When the number of samples is 30 (about 1/10 of the grid points), the average error is about 10%. The large-scale radiation field data came from the 2016 JAEA dose rate monitoring results at the 80 km range of the Fukushima Daiichi Nuclear Power Plant. The distribution of the dose rate peak in the dose rate distribution reconstructed by the algorithm interpolation is consistent with the original dose rate distribution. In this radiation field, the average relative error of the radiation field reconstruction based on 600 sample data is 141.69%, indicating that the location of the radiation hot spot is not as accurate as in the small-scale radiation field. The above results show that Poisson-Kriging algorithm has certain potential in solving the problem of reconstruction of the radiation field of unknown radioactive sources quickly and at low cost.