Estimating model-form uncertainty in Reynolds-averaged navier-stokes closures [electronic resource]
- Michael Alan Emory.
- Physical description
- 1 online resource.
All items must be viewed on site
Request items at least 2 days before you visit to allow retrieval from off-site storage. You can request at most 5 items per day.
|3781 2014 E||In-library use|
- Emory, Michael Alan.
- Iaccarino, Gianluca, primary advisor.
- Larsson, Johan, 1975- advisor.
- Lele, Sanjiva K. (Sanjiva Keshava), 1958- advisor.
- Stanford University. Department of Mechanical Engineering.
- Over the past several decades the role of numerical simulations has become increasingly important in the design of complex systems. In many such systems directly measuring or assessing the performance is prohibitively expensive, motivating the use of numerical simulations as a cost-effective alternative. As the role of simulations becomes more integral to the design and evaluation of such high risk systems, the ability to characterize and quantify the credibility or confidence in these simulation results becomes increasingly important. The subject of this thesis is one part of establishing credibility: uncertainty quantification (UQ). At its most basic level UQ tries to identify sources of uncertainty in the numerical simulation framework and characterize their influence on the solution. The focus of this work is a specific type uncertainty called model-form or "structural" uncertainty, which accounts for the assumptions, conceptualizations, or abstractions which the physics model relies upon, i.e. uncertainty in the mathematical structure of the model. The overarching goal of this thesis is to propose a general framework to address model-form uncertainty in complex numerical simulations. In the first portion of the thesis the framework is motivated and developed within the general context of computational analysis. A variety of analysis tools and concepts are explained in order to interpret the results provided in later chapters. The implementation of the framework to specifically investigate Reynolds-averaged Navier-Stokes turbulence model uncertainty is described in detail. The middle of the thesis shows application of the UQ framework to several turbulent flows, building a body of evidence for the utility of the approach. The cases are the incompressible developed channel and square duct, and the transonic flow over the Delery Case C geometry. These flows provide a hierarchy of unit level problems with increasing complexity and physical phenomena, allowing for investigation of the influence of the model-form framework in detail. The end of the thesis demonstrates the framework as part of a larger UQ analysis, within the context of establishing credibility for numerical simulations of the HyShot II scramjet engine. The various sources of uncertainty considered in this analysis are identified and described. The relative contributions to the overall uncertainty from these disparate sources is obtained, highlighting the importance of being able to independently analyze their influence.
- Publication date
- Submitted to the Department of Mechanical Engineering.
- Ph.D. Stanford University 2014
Browse related items
Start at call number: