Optimization of multi-layered beam shaping assembly for boron neutron capture therapy based on NSGA-III algorithm

Abstract The design of a high-performance Beam Shaping Assembly (BSA) for accelerator-based BNCT is a complex multi-objective optimization problem. We present an automated framework integrating the NSGA-III algorithm with Monte Carlo simulations to optimize the material and geometry of a multi-layered BSA for a 2.8 MeV RFQ neutron source. An entropy-weighted TOPSIS method objectively selects the optimal compromise from the Pareto set. The optimized BSA delivers an epithermal flux of 2.024 × 10 9 n·cm -2 ·s -1 at 20 mA, while reducing fast-neutron contamination by ∼ 44% below the IAEA limit. Dosimetry in a head phantom demonstrates promising therapeutic potential, with an 8.6 cm advantage depth and a 4.78 advantage ratio. This work provides a robust, intelligent design framework for supporting the preliminary evaluation of AB-BNCT facilities.