The High-Speed Train Parameter Design And Optimazation Based On Radial Basis Function

In recent years, the high speed train has been developing rapidly in our country and around the world. Because of its safety and comfort, it has become one of the means of transportation which people choose first. Therefore, the development and application of high speed train technology has become a symbol of the country’s traffic. High speed train is a complex mechanical and electrical products, the design process is very complex, the use of the agent model technology in the complex mechanical and electrical design becomes more and more widely.According to the research results of previous scholars, the 16 key parameters of the CRH are used as design variables, the lateral stability, vertical stability, derailment coefficient, reduction rate of wheel load, wheel axle lateral load, overturning coefficient and critical speed of 7 main performance indexes are output variables. The Latin hypercube design method was used to sample the 16 design variables, and the 100 level of the test sample points were obtained by using SIMPACK simulation software, and the simulation analysis was carried out, and the output response values of 7 performance indexes were obtained. The input and output values are normalized, and based on the radial basis function, the 95 sample points are used to establish surrogate model, and the remaining 5 sample points are used as test samples, after repeated training, the test sample is verified, and the accurate agent model is constructed based on the radial basis function.The design optimization of high speed train contains 7 sub targets, which belongs to the optimization of multi objective design. At first, the model of high speed train is optimized by using multi-objective particle swarm optimization algorithm, finally get a series of better solutions, the 26 selected groups are used for simulation and validation in SIMPACK software.7 performance indicatorsof 2 groups are better than an optimal solution of a type of vehicle CRH in the original design. Then, the model is optimized by differential evolution algorithm. According to the mechanism of differential evolution, the program is compiled in MATLAB. The initial population and non-dominated solutions are generated randomly, and then the operation of the cross, variation and selection of the competition is followed. Generate 200 groups of Pareto solution.60 groups of solutions are selected for simulation in SIMPACK software.7 performance indicators of 5 groups are better than the original design of the optimum solution. Finally, the results of the two optimization algorithms are compared and analyzed. The analysis shows that the two optimization algorithms have their advantages.