Abstract: In order to solve the problems of low efficiency and high error in traditional radiation shielding design, an intelligent optimization method for reactor shielding design based on the coupling of the fully connected neural network (FCNN) and the third generation non dominated sorting genetic algorithm (NSGA-III) was proposed. Taking a molten salt reactor as an example, the reactor shielding optimization model is established and Monte Carlo software is used to calculate a large number of samples. FCNN is used to machine learn the calculation data, and the multi-dimensional nonlinear mapping relationship between input layer parameters and output layer parameters is established. The neural network prediction results are used as the basis for calculating the fitness function. Based on NSGA-III, multi-objective optimization is carried out to obtain the Pareto optimal solution for multi-objective optimization of radiation shielding design. The results show that the FCNN coupled NSGA-III method performs well in solving multi-objective optimization problems and can be applied to reactor shielding design.