Dissertation/Thesis Abstract

An Integrated Bi-Fidelity Approach to Parametric Studies of Finite Wing Geometries
by Bieri, Kevin M., M.S., University of Colorado at Boulder, 2019, 156; 27662961
Abstract (Summary)

An approach to computational fluid dynamics (CFD) studies of the aerodynamic performance of finite wings based on parameter sweeps of the wings' geometries is developed. To reduce computational cost while retaining prediction accuracy, the approach leverages a bi-fidelity surrogate modeling technique which uses the skeletonization of the solution space from a low fidelity model to map a small sample of high fidelity model results throughout the parameter space. To facilitate automatic generation of the discrete realizations of the geometries, a mesh transformation technique integrated into the CFD solver uses linear elastic mesh deformation and mesh improvement post-processing to map a single baseline discretization of the flow problem to each target geometry within the parameter space while preserving the original mesh connectivity and structure in addition to any initial and boundary conditions.

The integrated bi-fidelity approach was evaluated by considering the specific application to predicting the sectional lift and drag coefficients distributions for the wing of a medium altitude, long endurance (MALE) unmanned aerial vehicle (UAV) given variations in the wing's span, taper ratio, and sweep angle. Consideration of the mesh transformation technique showed it produced a mesh of a target geometry with a 20% increase in span, a 20% change in taper ratio, and 15° of sweep with comparable quality to both the initial mesh and a mesh manually generated for the target geometry. Consideration of the bi-fidelity technique showed a surrogate model formed with eight high fidelity samples had an average root-mean-square error in its lift prediction relative across eight test cases of 0.5% compared to 14.9% for the low fidelity model alone. The drag prediction of the bi-fidelity model was slightly less accurate due to physics not captured by the high fidelity model but still had a relatively low error of 7.2% compared to 203.5% for the low fidelity model.

Indexing (document details)
Advisor: Jansen, Kenneth
Commitee: Doostan, Alireza, Evans, John
School: University of Colorado at Boulder
Department: Aerospace Engineering
School Location: United States -- Colorado
Source: MAI 81/8(E), Masters Abstracts International
Subjects: Aerospace engineering, Engineering, Fluid mechanics
Keywords: Bi-fidelity, CFD, Mesh Transformation, Multi-fidelity, Surrogate modeling
Publication Number: 27662961
ISBN: 9781392540879
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy