Variable-fidelity and reduced-order models for aero data for loads predictions

Stefan Goertz, Ralf Zimmermann, Zhong Hua Han

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review


This paper summarizes recent progress in developing metamodels for efficiently predicting the aerodynamic loads acting on industrial aircraft configurations. We introduce a physics-based approach to reduced-order modeling based on proper orthogonal decomposition of snapshots of the full-order CFD model, and a mathematical approach to variable-fidelity modeling that aims at combining many low-fidelity CFD results with as few high-fidelity CFD results as possible using bridge functions and variants of Kriging and Cokriging. In both cases, the goal is to arrive at a model that can be used as an efficient surrogate to the original high-fidelity or full-order CFD model but with significantly less evaluation time and storage requirements. Both approaches are demonstrated on industrial aircraft configurations at subsonic and transonic flow conditions.

Original languageEnglish
Title of host publicationComputational Flight Testing : Results of the Closing Symposium of the German Research Initiative ComFliTe, Braunschweig, Germany, June 11th-12th, 2012
EditorsNorbert Kroll, Rolf Radespiel, Jan Willem Burg, Kaare Sørensen
Publication date2013
ISBN (Print)978-3-642-38876-7
ISBN (Electronic)978-3-642-38877-4
Publication statusPublished - 2013
Externally publishedYes
EventClosing Symposium of the German Research Initiative - Braunschweig, Germany
Duration: 11. Jun 201212. Jun 2012


ConferenceClosing Symposium of the German Research Initiative
SeriesNotes on Numerical Fluid Mechanics and Multidisciplinary Design


  • aerodynamics
  • computational fluid dynamics (CFD)
  • Kriging
  • loads
  • proper orthogonal decomposition (POD)
  • reduced order modeling (ROM)
  • Variable fidelity modeling (VFM)


Dive into the research topics of 'Variable-fidelity and reduced-order models for aero data for loads predictions'. Together they form a unique fingerprint.

Cite this