Generative inverse design of aircraft aerodynamic shapes for optimal trajectory performance

Deep generative models have introduced data-driven capabilities to aerodynamic design, enabling the rapid exploration of complex design spaces without the prohibitive costs of iterative traditional high-fidelity simulations. However, existing methods often focus on static aerodynamic performance at isolated design points, neglecting the strong coupling between aerodynamic shapes and time-varying trajectory parameters across the full flight mission. To address this limitation, this article proposes a generative inverse design framework for aircraft aerodynamic shapes oriented toward optimal trajectory performance. First, a coupled geometry-aerodynamics-trajectory simulation workflow is established to construct a high-dimensional dataset linking parameterized shapes to optimal flight ranges. Second, to ensure strict satisfaction of performance constraints, a conditional information-enhanced conditional Wasserstein generative adversarial network with gradient penalty (CI-CWGAN-GP) model is developed. A critical innovation of this model is the integration of a differentiable trajectory performance consistency loss, which explicitly enforces physical constraints within the generative process, thereby significantly reducing the deviation between the generated and target trajectory performance. Finally, validation using the NASA common research model demonstrates that the proposed method significantly outperforms standard CWGAN-GP and physics-guided CGAN (PG-CGAN) models. Specifically, the CI-CWGAN-GP reduces generation errors by 45.89% and 11.74%, respectively, achieving a relative performance deviation of 0.5773% for specified range targets. This study establishes a novel paradigm for conceptual aircraft design, shifting the focus from static-point optimization to full-mission performance-driven generation.