TY - JOUR
T1 - Multilayer design and multi-objective optimization of neutron shielding composites by means of MCNP simulation and machine learning
AU - Liu, Benben
AU - Gu, Yizhuo
AU - Guo, Ruiqi
AU - Wang, Shaokai
AU - Li, Min
N1 - Publisher Copyright:
© 2025 Elsevier Ltd.
PY - 2026/2/8
Y1 - 2026/2/8
N2 - To meet neutron shielding and lightweight requirements, fiber-reinforced polymer matrix composites offer significant advantages as multifunctional materials with both structural and shielding capabilities. Owing to their inherent multicomponent and multilayered configurations, selecting suitable reinforcements and optimizing multilayer structure remains challenging. This study addresses the design and multi-objective optimization of multilayer composite shielding structures for neutron radiation protection. Monte Carlo N-Particle (MCNP) simulation method is adopted to predict radiation shielding property of various composites. A homogeneous model is first employed to examine the effects of typical shielding fillers (B4C and WO3) on the effective neutron dose in an epoxy resin matrix across the full neutron energy spectrum. Subsequently, an idealized layered structure model is used to clarify material composition strategies and multi-layer design principles for epoxy resin matrix composite. The results show that for fast neutron protection, a bilayer configuration with a high-Z material as the front layer and a hydrogen-rich matrix as the rear layer is optimal. For slow neutron protection, multilayer configurations demonstrate significant advantages: a 128-layer structure can reduce the effective dose of slow neutrons by up to 30 % compared with a bilayer structure. Furthermore, a multi-objective optimization strategy is proposed for multilayer structures by integrating MCNP simulations with machine learning, which can optimize shielding efficiency, structural thickness, and overall mass. Among six regression algorithms, a three-layer neural network model is chosen, which achieves high prediction precision. This approach optimizes both the minimum-dose configuration at fixed thickness and the minimum-weight configuration at fixed dose, providing efficient design guidelines for multilayer composite shielding.
AB - To meet neutron shielding and lightweight requirements, fiber-reinforced polymer matrix composites offer significant advantages as multifunctional materials with both structural and shielding capabilities. Owing to their inherent multicomponent and multilayered configurations, selecting suitable reinforcements and optimizing multilayer structure remains challenging. This study addresses the design and multi-objective optimization of multilayer composite shielding structures for neutron radiation protection. Monte Carlo N-Particle (MCNP) simulation method is adopted to predict radiation shielding property of various composites. A homogeneous model is first employed to examine the effects of typical shielding fillers (B4C and WO3) on the effective neutron dose in an epoxy resin matrix across the full neutron energy spectrum. Subsequently, an idealized layered structure model is used to clarify material composition strategies and multi-layer design principles for epoxy resin matrix composite. The results show that for fast neutron protection, a bilayer configuration with a high-Z material as the front layer and a hydrogen-rich matrix as the rear layer is optimal. For slow neutron protection, multilayer configurations demonstrate significant advantages: a 128-layer structure can reduce the effective dose of slow neutrons by up to 30 % compared with a bilayer structure. Furthermore, a multi-objective optimization strategy is proposed for multilayer structures by integrating MCNP simulations with machine learning, which can optimize shielding efficiency, structural thickness, and overall mass. Among six regression algorithms, a three-layer neural network model is chosen, which achieves high prediction precision. This approach optimizes both the minimum-dose configuration at fixed thickness and the minimum-weight configuration at fixed dose, providing efficient design guidelines for multilayer composite shielding.
KW - Fiber-reinforced polymer matrix composites
KW - MCNP simulation
KW - Machine learning
KW - Neutron radiation shielding
UR - https://www.scopus.com/pages/publications/105021817928
U2 - 10.1016/j.compscitech.2025.111451
DO - 10.1016/j.compscitech.2025.111451
M3 - 文章
AN - SCOPUS:105021817928
SN - 0266-3538
VL - 274
JO - Composites Science and Technology
JF - Composites Science and Technology
M1 - 111451
ER -