Multiobjective Optimization Shielding Design for Compact Accelerator-Driven Neutron Sources by Application of NSGA-II and MCNP
Baolong Ma, Lei Song, Mingfei Yan, Yujiro Ikeda, Yoshié Otake, Sheng Wang
Abstract
To find the optimal shielding design for compact accelerator-driven neutron sources (CANS) using multiobjective optimization, we developed a new method called nondominated sorting genetic algorithm-Monte Carlo method (NSGA-MC). NSGA-MC employs NSGA-II to optimize the shielding parameters based on calculations made by the Monte Carlo N-Particle Transport Code (MCNP). A layered shielding configuration with two materials of borated polyethylene (BPE) and lead (Pb) in the order of BPE/Pb/BPE/Pb for RIKEN Accelerator-driven Compact Neutron Source (RANS) was examined using this method, and two objectives were optimized simultaneously: equivalent dose rate and shielding structure weight. As a result, a tradeoff relationship between the objectives was finally obtained in the form of a Pareto front. The optimization results revealed significant improvements compared with the current RANS shielding configurations in terms of both dose and weight. The results indicate that a reduction in shielding weight of about 60% can be obtained by adopting the optimized shielding structure design, without sacrificing shielding performance. The performance of the method was discussed by showing advantages of NSGA-MC over the so-called weight sum method.