Surrogate-based aerodynamic shape optimization for train geometry design

Abstract

An enhanced parametric construction method of train model is presented to deal with shape optimization problems in this paper. The dimensional parameters of the train shape are sufficiently implemented in the global optimization for the first time. The size of the cross-sectional area and the vehicle volume are considered as nonlinear constraints thus making the optimization as close to the specification as possible. An optimization algorithm and a surrogate evaluation procedure are employed to facilitate design optimization for large-scale numerical simulations. To demonstrate the effectiveness of the designed surrogate-based optimization method, the aerodynamic shape optimization problems of the leading vehicle of a typical high-speed train ICE 3 is considered in terms of drag reduction and crosswind stability. Insights into the initial variable design space are provided by the results of computational fluid dynamics (CFD) simulations and the analysis of variance (ANOVA) test. These results indicate that the length of a uniform section of the train significantly contributes to the drag coefficient, while the train's height is the most critical design parameter for the side force coefficient. Concerning the drag coefficient without crosswind and the side force coefficient with crosswind, two single-objective optimization procedures are developed separately to minimize the two objectives, resulting in reductions in drag and side force coefficients of approximately 22.30% and 24.31%, respectively. On this basis, a multi-objective optimization procedure is performed to obtain a Pareto front. Although the optimization strength of the side force coefficient is obviously higher than that of the drag coefficient, the relative changes indicate that both objectives are in the same order of significance.