Multi-objective optimization of structural parameters for new box-type subgrade based on genetic algorithm

• A highly feasible multi-objective optimization method has been proposed by integrating genetic algorithms and BP neural networks. • Using sensitivity analysis and harmonic response analysis, optimization parameters for the new box-type subgrade structure have been determined, which enhances the structural strength and rigidity. • The optimized box-type subgrade structure achieves lightweight and excellent dynamic performance, with good optimization effects and significant cost benefits in practical applications. The new box-type subgrade structure possesses significantly high strength and rigidity. The present study conducts a multi-objective dynamic optimization for the lightweight and dynamic performance enhancement of the box subgrade structure. Sensitivity analysis methods, combined with finite element theory, are utilized to determine the design variables and dynamic performance optimization objectives of the box subgrade structure. These variables are employed as input and output variables for a BP neural network model, which is trained and validated to serve as the fitness function for the NSGA-II algorithm. The dynamic design optimization of the box subgrade is achieved through the collaborative optimization of the NSGA-II algorithm and the BP neural network. A design scheme that realizes the lightweight structure and optimal dynamic performance is selected, and compared with the results before optimization, the feasibility and superiority of the optimization scheme are verified.