Impingement of supercooled large droplets via reduced order models

Marco Fossati, Wagdi G. Habashi, Guido Baruzzi

Research output: Contribution to conferencePaper

1 Citation (Scopus)

Abstract

The high computational cost of 3-D viscous turbulent aero-icing simulations is one of the main limitations to address in order to more extensively use computational fluid dynamics to explore the wide variety of icing conditions to be tested before achieving aircraft airworthiness. In an attempt to overcome the computational burden of these simulations, a Reduced Order Modeling (ROM) approach, based on Proper Orthogonal Decomposition (POD) and Kriging interpolation techniques, is applied to the computation of the impingement pattern of supercooled large droplets (SLD) on aircraft. Relying on a suitable database of high fidelity full-order simulations, the ROM approach provides a lower-order approximation of the system in terms of a linear combination of appropriate functions. The accuracy of the resulting surrogate solution is successfully compared to experimental and CFD results for sample 2-D problems and then extended to a typical 3-D case.

Original languageEnglish
DOIs
Publication statusPublished - 1 Dec 2011
EventSAE 2011 International Conference on Aircraft and Engine Icing and Ground Deicing - Chicago, IL, United States
Duration: 13 Jun 201117 Jun 2011

Conference

ConferenceSAE 2011 International Conference on Aircraft and Engine Icing and Ground Deicing
CountryUnited States
CityChicago, IL
Period13/06/1117/06/11

Keywords

  • aircraft
  • computational fluid dynamics
  • drop breakup
  • principal component analysis
  • three dimensional
  • computational burden
  • computational costs
  • full-order simulation
  • icing conditions
  • Kriging interpolation technique
  • linear combinations
  • proper orthogonal decompositions
  • reduced order models
  • reduced-order modeling
  • supercooled large droplets
  • turbulence

Fingerprint Dive into the research topics of 'Impingement of supercooled large droplets via reduced order models'. Together they form a unique fingerprint.

Cite this