Abstract: Traditional design methods of rail plate carrying vehicle tend to be conservative.The structural strength had a large surplus.To reduce weight and the manufacturing cost of the structure,a new structure optimization method was proposed.This method combines the finite element software ANSYS and the optimization software ISIGHT.First of all,choosing the design variables which have large influence on constraint condition and objective function by the method of Latin Hypercube Design(LHD).Then,the model of the neural network radial basis approximation is set up.Finally,the approximate model is optimized by using the Multi Island Genetic Algorithm(MIGA).The results show that the method can greatly improve the efficiency of optimization,and the structure weight of rail plate carrying vehicle reduce 9.9 % by the optimization design which is of important significance for the design of rail plate carrying vehicle.

Key words: rail plate carrying vehicle, optimization design, Radial Basis Function(RBF) neural network, Multi Island Genetic Algorithm(MIGA)