Multi-fidelity model fusion and uncertainty quantification using high dimensional model representation

Martin Kubicek, Piyush M. Mehta, Edmondo Minisci, Massimiliano Vasile

Research output: Chapter in Book/Report/Conference proceedingConference contribution


High-fidelity modeling based on experiments or simulations is generally very expensive. Low-fidelity models, when available, typically have simplifying assumptions made during the development and hence are quick but not so accurate. We present development of a new and novel approach for multi-fidelity model fusion to achieve the accuracy of the expensive high-fidelity methods with the speed of the inaccurate low-fidelity models. The multi-fidelity fusion model and the associated uncertainties is achieved using a new derivation of the high dimensional model representation (HDMR) method. The method can provide valuable insights for efficient placement of the expensive high-fidelity simulations in the domain towards reducing the multi-fidelity model uncertainties. The method is applied and validated with aerodynamic and aerothermodynamic models for atmospheric re-entry.

Original languageEnglish (US)
Title of host publicationSpaceflight Mechanics 2016
EditorsMartin T. Ozimek, Renato Zanetti, Angela L. Bowes, Ryan P. Russell, Martin T. Ozimek
PublisherUnivelt Inc.
Number of pages16
ISBN (Print)9780877036333
StatePublished - 2016
Event26th AAS/AIAA Space Flight Mechanics Meeting, 2016 - Napa, United States
Duration: Feb 14 2016Feb 18 2016

Publication series

NameAdvances in the Astronautical Sciences
ISSN (Print)0065-3438


Other26th AAS/AIAA Space Flight Mechanics Meeting, 2016
Country/TerritoryUnited States

Bibliographical note

Funding Information:
For Piyush Mehta is provided by the European Commission through the Marie Curie Initial Training Network (ITN) STARDUST under grant number 317185. Partial support for Martin Kubicek is provided by 'OPTIMAD Engineering Srl'.


Dive into the research topics of 'Multi-fidelity model fusion and uncertainty quantification using high dimensional model representation'. Together they form a unique fingerprint.

Cite this