Dissertation Defense: Enhancement of CFD Surrogate Approaches for Thermo-Structural Response Prediction in High-Speed Flows

Kirk Brouwer, PhD Candidate, Aerospace Engineering

All dates for this event occur in the past.

E525 Scott Lab
E525 Scott Lab
201 W. 19th Ave.
Columbus, OH 43210
United States

Committee Members

  • Professor Jack McNamara, Chair (MAE)
  • Professor Jen-Ping Chen (MAE)
  • Professor Sandip Mazumder (MAE)
  • Benjamin Smarslok (Research Aerospace Engineer at AFRL)


Abstract

Robust and expedient prediction of the aerothermodynamic loads is critical to the development of modern reusable high-speed platforms. However, the presence of complex flow physics, including strong inviscid-viscous interactions, impinging shocks, and intense three-dimensionality, poses a significant challenge to the fielding of these platforms. Furthermore, the compliant nature of high-speed structures in combination with the extreme environments result in the potential for path-dependent loading conditions and deformed configurations that evolve over long-duration trajectories. While computational fluid dynamics (CFD) provide high accuracy solutions, computational costs limit application for online loads prediction. In contrast, basic engineering-level approximations are efficient but lack broad accuracy. These issues have motivated the development of CFD surrogates that harness the predictive accuracy of high-fidelity models while retaining the computational efficiency required for online predictions. While a significant body of work has demonstrated the capabilities of the CFD surrogate method, open questions remain regarding the viability of the approach for shock-dominated environments and systems with complex structural responses. This dissertation seeks to address these questions through the identification and development of required improvements. Specific objectives include characterizing the accuracy of the CFD surrogate approach for aeroelastic loads prediction in the presence of shock impingements and the development of modeling strategies to account for arbitrary structural deformations. For quasi-steady stationary and oscillating shock impingements, the CFD surrogate yields reasonable to excellent agreement with unsteady CFD at a fraction of the online computational cost. However, the accuracy of the model degrades as the relative length between the shock-induced separation and deformation increases. Through analytical and numerical studies, these errors are tied to the breakdown of quasi-steady interactions. Specifically, the interplay between shock-induced separation and surface motion induces unsteady variations in the separation length that are not accounted for in the surrogate model. This variation is linked to the dependency of separation on both the surface curvature and spatial gradient of surface velocity. In order to capture these unsteady effects, the CFD surrogate is supplemented with a recurrence model (Surrogate-Based Recurrence Framework or SBRF). Results demonstrate that the SBRF captures aeroelastic loading in the presence of shock-induced separation with a minimal increase in the online cost of the surrogate model. Finally, a new modeling strategy for steady aerothermodynamic loads prediction is developed in order to better accommodate structural deformation. The approach relies on the identification of key parameters in theoretical models, which are then used to construct pointwise surrogate models for the steady pressure and heating loads. In contrast to conventional surrogates that require a priori information on the structural response, the pointwise models capture arbitrary deformations with fewer steady CFD training solutions overall. Results highlight that the pointwise surrogates exhibit balanced performance in terms of accuracy and computational expense relative to several conventional modeling strategies.