To make appropriate decisions in the evaluation phase of the exterior design of subway trains, an optimal selection method was proposed based on multi-level gray relational analysis. The exterior design factors of subway trains were analyzed to construct an index system for design evaluation. The significance of each index was compared through an analytic hierarchy process. The correlation coefficient of each index in the plan was calculated through gray relational analysis to obtain the weighted correlation degree of each design scheme. The optimal selection of the exterior design of Guangzhou Metro Line 6 in China was considered as an example. Four types of subjects were recruited: professional designers, students majoring in design, subway train design experts, and subway passengers in Guangzhou. The weight of each index in the evaluation system was calculated using questionnaire scoring. Virtual simulation software was applied to evaluate the human factors related to each scheme. The indices in each plan were then scored to calculate the correlation coefficient and the overall correlation degree; and finally, the optimal selection was obtained. The results showed that it was practical to evaluate and optimize the exterior design of subway trains based on multi-level gray relational analysis. In the evaluation index system, the weights of technology, human factors, aesthetics, and culture were 0.517, 0.297, 0.099, and 0.087, respectively, which showed that technology had the greatest impact on the system, while human factors, aesthetics, and culture were useful complements. Our results showed that Design Scheme 1 was unsuitable as an optimization scheme due to the high escape window. Meanwhile, Design Scheme 2 was optimal overall, from a technical perspective. Design Scheme 3 was the best in terms of the escape window index (a human factor). Design Schemes 3 and 4 were optimally assessed from aesthetic and cultural perspectives. This study is conducive to the optimization of the exterior design of subway trains, can be used to inform design iteration, and provides a reference for the optimal selection of design schemes for other urban rail trains.