Surrogate-based multi-objective nose shape optimization for 400 km/h-class high-speed trains under open-air, crosswind, and tunnel operations

Abstract

Abstract This study presents a surrogate-based multi-objective nose shape optimization for a 400 km/h-class high-speed train, considering normal open-air, crosswind open-air, and tunnel-related scenarios. The KTX-Cheongryong nose shape is parameterized using Bézier curves and a section box approach to satisfy dimensional constraints specified in relevant technical regulations. The objective functions are the tail car drag coefficient under normal open-air conditions, the leeward rail rolling moment coefficient under a 40 m/s crosswind, and a micro-pressure wave-related proxy based on the circumferential integration of wall-normal velocity on an imaginary tunnel wall. A trackside pressure variation limit of 700 Pa at 250 km/h is imposed as a constraint. A dataset comprising 544 designs was generated using three-dimensional compressible steady Reynolds-averaged Navier–Stokes simulations and used to construct Gaussian process regression surrogate models. These models are coupled with the non-dominated sorting genetic algorithm II to obtain Pareto-optimal solutions, all of which dominate the KTX-Cheongryong across the three objectives. Extreme designs on the Pareto front, selected as boundary solutions for each objective, are compared with the corresponding single-objective optima to clarify objective trade-offs. The representative compromise design achieves an 8.9% reduction in the total drag coefficient compared to KTX-Cheongryong, primarily due to a 22.9% decrease in the tail car drag coefficient. It also reduces the micro-pressure wave-related proxy and the leeward rail rolling moment coefficient by 4.3% and 3.6%, respectively, while satisfying the constraint. The results demonstrate that the proposed optimization process mitigates degradation of non-target objectives and develops high-performance nose designs for 400 km/h operation.