Surrogate modeling for realistic early-stage landing gear design load estimates

• Gaussian Process surrogate models accurately predict nonlinear landing gear dynamics. • PCA reduces dimensionality enabling efficient optimization of landing gear parameters. • Per-sample uncertainty optimization reduces conservatism in early-stage design loads. • Sobol sensitivity analysis identifies aircraft pitch and mass as key load drivers. • Framework integrates uncertainty quantification with subsystem optimization seamlessly. The work presented in this paper focuses on surrogate modeling for early-stage landing gear design to enable rapid load prediction and optimization under uncertainty. We propose an approach that integrates Gaussian-Process regression with Principal Component Analysis to approximate the dynamic behavior of a simplified landing gear model. The work presented in this paper focuses on surrogate modeling for early-stage landing gear design to enable rapid load prediction and optimization under uncertainty. We propose an approach that integrates Gaussian-Process regression with Principal Component Analysis to approximate the dynamic behavior of a simplified landing gear model featuring nonlinear damping and stiffness characteristics. A key feature of the methodology is the management of design uncertainty combined with global sensitivity analysis using Sobol indices. These indices assess the contribution of key parameters, such as aircraft mass, and tire stiffness, to the peak landing loads. These analyses not only identify the most influential factors, but also guide targeted subsystem optimization. The optimization process minimizes the reaction factor, defined as the ratio of the peak ground-load to the static-load by adjusting the landing gear’s stiffness and damping properties. Comparisons between design-only and per-sample uncertainty optimization highlight the advantages of directly incorporating uncertainty within the optimization loop. Results show that the surrogate model provides accurate load estimates that closely align with higher-fidelity simulations, while the optimization reduces peak loads and enhances the dynamic response of the landing gear. The framework offers practical insights into the effects of varying design parameters on system performance, thereby aiding robust design decision-making.