Parameter multi-objective optimization of high-speed train suspension system based on combined approximation model
The parameters of the high-speed train suspension system are closely related to the system's dynamic perform-ance;the multi-objective optimization of these parameters is greatly helpful to improve the system's dynamic performance.In this article,according to the Pearson correlation,the analysis is conducted on the correlation between the system's parameters and its dynamic performance;then,four key parameters with the highest correlation to the system's dynamic performance are identified,with the help of a combination of Analytic Hierarchy Process and Fuzzy Comprehensive Evaluation.With the key parameters as the design variables,efforts are made to construct the Kriging approximation model,the radial basis function neural network(RBF)approximation model,and the second-order response surface(RSM)approximation model,which take aim at such factors as derailment coefficient,wheel load reduction rate,wheel-axle lateral force,comfort index,and nonlinear critical speed.Then,after the weight coefficients of the three single approximation models are calculated by means of the K-fold cross validation meth-od,they are fitted based on the weight coefficients,in order to set up a combined approximation model of the high-speed train's dynamic performance indicators,and then the accuracy of the combined approximation model is evaluated.With the combined ap-proximation model as the objective function,with such factors as derailment coefficient,wheel load reduction rate,wheel-axle lat-eral force,comfort index,and nonlinear critical speed as the objective response,the NSGA-Ⅱ optimization algorithm is used to optimize the system's parameters.The results show that the optimal solution has an optimization rate of over 10％for all five dy-namic performance indicators,thus effectively improving the high-speed train's dynamic performance.
combined approximation modelidentification of key parameterK-fold cross validationmulti-objective optimi-zationsuspension parameter
NETL
NSTL
万方数据
