An optimization method of shielding structure in marine reactor based on neural networks and genetic algorithms is studied in this paper,and developed a visual interface based on the optimization method. First a model of four shielding structure in marine reactor is given. The forecasting accuracy of neural network is verified by comparison of normalized neutron transmittance in different thickness between neural network''s forecasting and Monte Carlo calculation. Neural network''s predictive result is considered as the fitness function of genetic algorithm to find the best shielding parameter...