AI-Assisted Multidisciplinary Design Optimization of Hydrogen-Electric Aircraft Integrating Aerodynamics, Propulsion, Thermal Management, and Structural Mass

Hydrogen-electric aircraft represent a promising pathway to zero-carbon aviation, but their conceptual design is governed by strong interdisciplinary couplings among aerodynamics, electric propulsion, fuel-cell thermal management, and cryogenic structural mass estimation. Existing multidisciplinary design optimization (MDO) studies address subsets of these disciplines, typically aero-structural or thermal-propulsion coupling, without an integrated, AI-accelerated framework that simultaneously captures all four. This paper proposes a novel MDO architecture in which physics-informed neural network (PINN) surrogates replace expensive high-fidelity disciplinary solvers inside an NSGA-II multi-objective optimization loop, enabling rapid exploration of the Pareto trade space between maximum take-off weight (MTOW), mission range, and thermal management system (TMS) drag penalty. The aerodynamic module is based on Reynolds-Averaged Navier–Stokes (RANS) solutions computed with the open-source SU2 solver; the propulsion module couples a PEM fuel-cell electrochemical-thermal model with electric motor efficiency maps; the thermal module integrates ram-air heat-exchanger sizing; and the structural module employs semi-analytical wing-box mass estimation (validated against three aircraft with < 5% error) combined with ASME Boiler and Pressure Vessel Code cryogenic tank sizing. A benchmark reference dataset is assembled from published specifications of existing hydrogen-electric demonstrators (DLR/H2FLY HY4, ZeroAvia Do228, and Airbus ZEROe) to anchor validation. The methodology includes mesh-independence studies, turbulence-model selection, convergence-criteria documentation, sensitivity analysis, and end-to-end uncertainty quantification. The results framework specifies the required comparison variables, contour plots, statistical tests, and baseline cases without fabricating numerical outcomes. A critical peer review identifies likely reviewer concerns about surrogate generalization, validation fidelity, and data-leakage risk, and a revised strategy addresses each. The framework is designed for reproducibility using open-source tools (SU2, OpenMDAO, OpenVSP, Python/PyTorch) and is targeted at Q1 journals in aerospace engineering and applied energy.