System Identification CFD-Based Reduced-Order Modeling for Hypersonic Vehicles

Abstract

System identification (SID) techniques were utilized to assemble reduced-order models purposed for estimating the aerodynamic coefficients of a hypersonic vehicle subjected to flight conditions of interest. The reduced-order models combined the accuracy of high-fidelity hybrid Reynolds Averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES) computational fluid dynamic (CFD) models with the computing speed of low-fidelity inviscid CFD models, efficiently capturing the effects of complex physics in a timely manner. The vehicle geometry utilized for this study was the High-Speed Army Reference Vehicle (HARV), which was simulated in training maneuver motions solved by HPCMP CREATETM-AV Kestrel, the high-fidelity CFD software. The resulting data was used and assessed in its information supply to the SID techniques, which were also performed in Kestrel as a post-processing operation. Many SID models with varying structures were built with the training maneuver data. The models were validated using a variety of different dynamic maneuvers and static configurations in an effort to understand the limits and capabilities of hypersonic SID modeling. The results suggested insufficient low-rate data information in the training maneuver hampered the SID model prediction accuracies the most. A single trajectory analysis revealed that simulation results using SID model prediction aerodynamic databases and using low-fidelity CFD model prediction databases did not drastically differ. Once constructed, the SID model expressed the capacity to predict much more complex databases in significantly less time. This emphasized substantial benefits toward utilizing SID reduced-order models in the design phase of hypersonic vehicles.