MULTI-OBJECTIVE OPTIMIZATION OF VEHICLE/TRACK PARAMETERS BASED ON RBF NEURAL NETWORK SURROGATE MODEL

Abstract

The RBF（ Radial Basis Function） neural network surrogate model that employed to explore the multi-objective optimization problems of vehicle and track parameters is to improve the dynamic performance of vehicles. The sensitivity of dynamic performance on vehicle and track parameters was analyzed by constructing a vehicle-track coupling dynamic simulation model of high-speed train and using the UM and Isight joint simulation technology. The eight parameters with the highest sensitivity ratio were used as the design variables,and a surrogate model of RBF neural network was established on the response of the dynamic performance. Then the model was performed to optimize the vehicle/track parameters. The results show that the optimization rate of the optimal solution for the derailment coefficient is 13. 14%,and the optimization rate of the wheel load reduction rate is 14. 63% after the vehicle and track parameters are optimized,which demonstrates that the optimization effect is remarkable,and the dynamic performance of the vehicle has been significantly improved.

Keywords

• High speed train;Agent model;Radial Basis Function（RBF）;Multi-objective optimization;Sensitivity analysis