The High-Speed Train Performance Parameter Design And Optimization Based On Surrogate Model Technology

High-speed trains is a complicated mechanical and electrical products. Because of a variety of design variables to meet dynamics performance of high-speed trains, the design space is great complex. There is a high complexity and difficulty to find the important design variable set of the performance, optimize them and achieve design variables which are better than those of the existing design parameter set. The simulation soft such as SIMPACK can train and analysis dynamics simulation models. However, there are so many design variables that the optimization and track are very difficult. It is critical technology that integrate the design analysis and optimization effectively and deal with optimized design based on the simulation in oder to combine CAD and CAE into a whole design process. The main research contents are as follows:(1) The study on design strategy of design and optimization of high speed train performance parameters based on the surrogate model technologyBased on MOP and surrogate model technology, combined with the engineering practice, determine the method of design and optimization of performance parameters of high speed trains, and choose the design strategy of Latin hypercube sampling strategy, neural network model and multi-MOP optimization methods.(2) Establishing a dynamic model of high speed train and reducing design spaceBased on the study of dynamics simulation models of multi-body systems of high speed trains, extract the topology relation among the components, and then abstract the expression form of physical system and simulation system, get all input and output variables involved in physical experiment and the simulation experiment. Based on the prior knowledge of field experts, after the integrated assessment for all design variables, acquire 29 design parameters which influence on vehicle dynamics (including range) greatly and the 7 response performance evaluation indexes which ensure the train operation based on performance analysis. The reduction of the design space is formed.(3)The improved neural network surrogate model scheme and the sensitivity analysisPut forward a kind of improved neural network surrogate model scheme based on design space pretreatment-LM-regularization neural network model to improve the neural network generalization ability and generalization accuracy. The surrogate model of high-speed train design parameters for the reduction design space is built, and its accuracy to meet the requirements is verified. A new way to identification of critical design parameters of the high-speed train based on the sensitivity analysis is proposed, and the application method is given.(4)Study on the method of MOP to solve high speed train design parametersBased on the PAC method, the multi-objective problem is changed into single objective problem, and the genetic algorithm and differential evolution algorithm is used to optimize the problem. The results show that the DE is better than the genetic algorithm and verify that the global searching ability of difference algorithm is better. In order to solve the results of the traditional multi-objective optimization algorithm is not so ideal, multi-objective heuristic intelligent optimization method based on differential evolution is employed. In the 200 set of Pareto non-dominated solutions, simulation shows that this method obtains there is one set of optimal solutions, of which indicators are not inferior to the original some type of CRH design solutions. It is showed that the intelligent multi-objective optimization method is better than the traditional multi-objective method. It is proved that the mathods of the sensitivity analysis and the identification of critical parameters are corret.(5) Improved differential evolution algorithm and its application in the intelligent optimization design of high speed trainsThis paper proposes an improved differential evolution algorithm, by which the reconstruction of initial population evolution with non-dominated solutions acquired from several hybrid optimization methods are obtained. By using the improved differential algorithm, after the design and optimization of high speed trains and the simulation verification,16 set of solutions are better than the 7 indicators of original some type of CRH solution set.In this paper, the research shows that the surrogate model technology can not only reduce design space and the simulation time greatly, and also achieve very good results of optimization design of high-speed train to solve the problem of the engineering field. In the end, results also indicated that the method is efficient and satisfied to improve the design of the high-speed train or similar complex electromechanical system, which means excellent engineering practicability.