A neutron energy spectrum study of 241Am-Be based on improved Gravel and PSO algorithms

Abstract

In the study of neutron energy spectra, different methods have been proposed one after another. Due to their ability to provide efficient parameter optimization and accuracy, Gravel algorithm and PSO algorithm are applied to the deconvolution of ²⁴¹Am-Be neutron source. In this paper, we first obtain the response function of the multi-sphere spectrometer to the ²⁴¹Am-Be neutron source by Geant4 simulation on a platform that is highly consistent with the experimental environment. Then, by combining neutron experimental count rates and neutron response functions, an improved version of Gravel and PSO algorithms is employed to unfold the neutron spectrum of ²⁴¹Am-Be neutron source. In this case, the Gravel algorithm improves the divergence result by adding a scaling factor, and the PSO algorithm applies a Cauchy mutation strategy to escape the local optimal solution. Finally, the unfolded neutron energy spectrum of ²⁴¹Am-Be is compared with the international standard ²⁴¹Am-Be neutron spectrum, and error analysis of the deconvolution is performed using mean square error, average relative deviation, and neutron spectrum quality as indicators. The results show that the improved Gravel and PSO algorithms exhibit superior performance compared to their previous versions and are largely consistent with the international standard ²⁴¹Am-Be neutron energy spectrum curve. The MSE, ARD and Qs values for the improved Gravel algorithm are 0.004%, 1.22% and 1.31% respectively, while those for the improved PSO algorithm are 0.25%, 16.29% and 8.08% respectively. This demonstrates the reliability and applicability of both algorithms in reconstructing neutron energy spectra.