Lab Home | Phone | Search | ||||||||
|
||||||||
This talk presents some of the methodologies in use and under development in the Stanford University/DoE PSAAP center to quantify the uncertainties that are present in the operation of hypersonic propulsion systems (scramjets) so that the appearance of the unstart phenomenon can be avoided. Uncertainties of various kinds and sources must be well understood so that future systems can be designed and operated safely. Firstly, aleatory / irreducible uncertainties arising from the natural variability of the system and its environment must be propagated from input to output in a computationally efficient fashion. Various intrusive and non-intrusive approaches are discussed, especially some focusing on the augmentation of function values by first derivative information (obtained inexpensively through the use of adjoint methods). Epistemic (or model-form) uncertainties arising from the use of models that fail to capture all of the physics in the real system are harder to quantify. The talk will outline our current efforts that focus on the quantification of the model-form errors introduced by the use of Reynolds-Averaged Navier-Stokes turbulence models. The talk will conclude with an introduction to a system-level modeling project that involves the design of the national air transportation system (NextGen) that can leverage many of the UQ techniques we are pursuing in PSAAP and, in addition, includes the added complication of strategic actors whose interactions largely influence the evolution of the system. Host: David Wolpert, CCS-3, 665-7914, Game Theory Seminar Series |