Generalized Treatment of Surface Deformation for High-Speed Computational Fluid Dynamic Surrogates

Abstract

A new surrogate modeling approach is developed that augments the loads from an analysis over an undeformed surface in order to account for the effects of static surface deformations in high-speed flows. The approach relies on pointwise data-driven surrogates to approximate the nonlinear flow mechanics and theoretical expressions to identify the surrogate input parameters. The pointwise nature of the model enables efficient construction with minimal assumptions on the surface deformation. The accuracy of the modeling strategy is evaluated relative to steady Reynolds-averaged Navier–Stokes analyses for prescribed deformations and quasi-static aerothermoelastic response prediction. Basic fluid loads models and conventional kriging interpolation models, where deformation is parameterized using global characteristic shapes, are also considered. The new surrogate approach yields normalized root mean square errors of less than 5% relative to steady Reynolds-averaged Navier–Stokes predictions of pressure and heat flux over prescribed deformations. Results for aerothermoelastic simulations highlight that the approach outperforms all other models in terms of predicting the thermo-structural response and fluid loads. In regard to computational requirements, the new approach affords a 93% reduction in the offline costs compared with conventional kriging models and is several orders of magnitude faster than Reynolds-averaged Navier–Stokes simulations for online predictions.