High dimensional uncertainty propagation for hypersonic flows and entry propagation

  • Martin Kubicek

Student thesis: Doctoral Thesis

Abstract

To solve complex design problems, engineers cannot avoid to take into account the involved uncertainties. This is important for the analysis and design of hypersonic objects and vehicles, which have to operate in extreme conditions.In this work, two approaches for a high dimensional uncertainty quantification (UQ) are developed. The first approach performs a single-fidelity non-intrusive forward UQ, while the second one performs a multi fidelity UQ, as an extension of the first approach. Both methods are focused on real engineering problems and, therefore, appropriate heuristics are included to achieve an optimal trade-off between accuracy and computational costs.In the first approach, the stochastic domain is decomposed into domains of lower dimensionality, and, then, each domain is handled separately. This is possible due to the application of the HDMR, which is here derived in a new way. This new derivation allowed to deduce important conclusions about the high dimensional modelling, which are used in the prediction scheme.This novel approach for the selection of the higher order interaction effects drastically reduce the required number of samples. In order to have optimally distributed samples for the problem of interest, the adaptive sampling scheme is introduced. Moreover, the multi-surrogate approach is introduced in order to improve the robustness of the method. The single-fidelity approach is tested on a debris re-entry case and the method is validated with respect to the MC simulation method.In the second approach, the multi-fidelity approach has been developed. In order to have the optimal combination of the low fidelity models, the power ratio approach is introduced. To correct the low fidelity model, the classical additive correction, adapted to work within the HDMR approach, is used. The multi-fidelity approach has been tested on the GOCE re-entry case, where the performed tests demonstrate the potentialities of the method.
Date of Award1 May 2016
LanguageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsUniversity of Strathclyde
SupervisorEdmondo Minisci (Supervisor) & Massimiliano Vasile (Supervisor)

Cite this