Uncertainty Quantification for High-Order, High-Performance Computational Fluid Dynamics

Beschreibung

Uncertainty quantification increases the reliability of numerical simulations. Numerous methods have been developed, but their application in computational fluid dynamics has so far been largely limited to simplified test cases, and their use in realistic engineering problems is cumbersome and has been little studied. In this thesis, the treatment of complex turbulent flow problems with different uncertainty quantification methods is explored. Intrusive and non-intrusive polynomial chaos methods as well as variance reduction techniques for Monte Carlo methods are employed. High-order discontinuous Galerkin methods are used for discretization in the physical space. Algorithmic improvements and implementational strategies for efficient simulations on high performance computing clusters are proposed. Two pieces of software are presented for implementation of the methods. Two large-scale application cases are considered. In cavity aeroacoustics, the effect of an uncertain cavity geometry on the emitted noise spectra is shown. In airfoil icing, the aerodynamic performance deterioration due to uncertain ice shapes is investigated. The applicability and efficiency of the employed methods are analyzed and compared. With its findings, the thesis aims to contribute to an informed and facilitated application of uncertainty quantification and thus to greater reliability in computational fluid dynamics.

Autor*in

Jakob Dürrwächter

Themen in »Uncertainty Quantification for High-Order, High-Performance Computational Fluid Dynamics«

Uncertainty quantification Fluid dynamics Numerical methods

Stimmen zu »Uncertainty Quantification for High-Order, High-Performance Computational Fluid Dynamics«

Details

Link teilen