The Comfort Optimization Design Of High-speed EMU Based On Surrogate Model

With the sustainable development of China’s high-speed EMU and high-speed railway technology,China’s national railway network is changing with time and high-speed EMUs have become the most beautiful image embodying "China Speed".The high-speed train is a complex giant system,each performance indicator of train operation will change with the increase of speed.However,the comfort of train operation is based on the performance of train operation safety,stability and smoothness.Therefore,in order to ensure that the two major advantages of high-speed trains,"high speed" and "comfort",are more prominent,it is necessary to further study the factors that affect the performance indicators of high-speed trains,which includes nonlinear critical speed,derailment coefficient,wheel load reduction rate,comfort index and so on.Firstly,the vehicle-track coupling multi-body dynamic model of high-speed EMU is established and its correctness is verified.Due to the development of computer technology and the maturity of multi-body dynamics simulation software,it has become a reality to use modeling and simulation methods to study the dynamics of high-speed EMUs.For the high-speed EMU system,the multi-body dynamics model is established with the UM software by combining the multi-body dynamics theory with the topological structure of the high-speed EMU,and the specific indicators are simulated and calculated that affect the dynamic performance of the train.The results show that the performance index values all meet the requirements of the "High-speed EMU Vehicle Test Specification",that is,the built high-speed EMU model has correctness.Secondly,it is established a surrogate model of high-speed EMUs and it is verified that the accuracy of the surrogate model meets the requirements.Because it has the extremely low efficiency of directly using the multi-body dynamics model of high-speed EMUs to optimize design parameters,it is necessary to establish a data-driven approximate model(RBF-NN surrogate model)with a high accuracy to achieve the goal of subsequent optimization of design parameters.When using the Isight software to establish the surrogate model,the design parameter samples obtained by the optimal Latin hypercube(abbreviated as OLHC)method are selected as the input of the surrogate model,the train dynamics performance indexes are used as the output of the surrogate model,and the radial basis neural network is used to build the surrogate model,which are in order to obtain several RBF surrogate models which meet the accuracy requirements.Finally,the comfort optimization function of high-speed EMUs is established,and the optimization results are analyzed.This paper selects 10 train structure parameters as optimization design variables,selects the established multiple RBF surrogate models as the objective function,selects NSGA-Ⅱ as the multi-objective optimization algorithm which sets the population size of 40,the genetic algebra of 200,and the crossover probability of 0.9.In the end,the Isight optimization software is used to establish the comfort optimization function of the high-speed EMU,and the optimization results of four schemes in Pareto solution are evaluated.The result shows that all indexes have been improved and the optimization purpose is realized.