Bulletin of the American Physical Society
20th Biennial Conference of the APS Topical Group on Shock Compression of Condensed Matter
Volume 62, Number 9
Sunday–Friday, July 9–14, 2017; St. Louis, Missouri
Session E6: Focus Session: Uncertainty Quantification in Compressible High-Speed Flows II |
Hide Abstracts |
Chair: Thomas Jackson, University of Florida Room: Regency Ballroom E |
Monday, July 10, 2017 3:30PM - 3:45PM |
E6.00001: Multilevel UQ strategies for large-scale multiphysics applications: PSAAP II solar receiver Lluis Jofre, Gianluca Geraci, Gianluca Iaccarino Uncertainty quantification (UQ) plays a fundamental part in building confidence in predictive science. Of particular interest is the case of modeling and simulating engineering applications where, due to the inherent complexity, many uncertainties naturally arise, e.g. domain geometry, operating conditions, errors induced by modeling assumptions, etc. In this regard, one of the pacing items, especially in high-fidelity computational fluid dynamics (CFD) simulations, is the large amount of computing resources typically required to propagate incertitude through the models. Upcoming exascale supercomputers will significantly increase the available computational power. However, UQ approaches cannot entrust their applicability only on brute force Monte Carlo (MC) sampling; the large number of uncertainty sources and the presence of nonlinearities in the solution will make straightforward MC analysis unaffordable. Therefore, this work explores the multilevel MC strategy, and its extension to multi-fidelity and time convergence, to accelerate the estimation of the effect of uncertainties. The approach is described in detail, and its performance demonstrated on a radiated turbulent particle-laden flow case relevant to solar energy receivers (PSAAP II: Particle-laden turbulence in a radiation environment). [Preview Abstract] |
Monday, July 10, 2017 3:45PM - 4:00PM |
E6.00002: Adjoint-based ignition sensitivity in turbulent combustion David Buchta, Jesse Capecelatro, Jonathan Freund We demonstrate adjoint-based sensitivity calculations for large-scale turbulent combustion simulations, with the goal of identifying, quantifying, and reducing prediction uncertainties. It is demonstrated on a non-premixed turbulent shear layer, a reacting jet-in-crossflow, and ignition in decaying turbulence. We distinguish sensitivities between a detailed and a global one-step hydrogen-air mechanism. The primary model system is the ignition of a turbulent jet by a laser-induced optical breakdown (LIB). Ignition, defined by a space-time integral of temperature, is most sensitive to the modeled plasma kernel geometry and its energy deposited on the gas phase. Thus, combining the adjoint-based sensitivity with the LIB's aleatoric interspersed plasma kernels, these parameters dominate the propagated output uncertainty, which is local to the inputs. The present combustion sensitivity studies are a component of a multi-scale, multi-physics combustion application, which is also discussed for context. [Preview Abstract] |
Monday, July 10, 2017 4:00PM - 4:15PM |
E6.00003: Analysis of a Multi-Fidelity Surrogate for Handling Real Gas Equations of State Frederick Ouellet, Chanyoung Park, Bertrand Rollin, S. Balachandar The explosive dispersal of particles is a complex multiphase and multi-species fluid flow problem. In these flows, the detonation products of the explosive must be treated as real gas while the ideal gas equation of state is used for the surrounding air. As the products expand outward from the detonation point, they mix with ambient air and create a mixing region where both state equations must be satisfied. One of the most accurate, yet computationally expensive, methods to handle this problem is an algorithm that iterates between both equations of state until pressure and thermal equilibrium are achieved inside of each computational cell. This work aims to use a multi-fidelity surrogate model to replace this process. A Kriging model is used to produce a curve fit which interpolates selected data from the iterative algorithm using Bayesian statistics. We study the model performance with respect to the iterative method in simulations using a finite volume code. The model's (i) computational speed, (ii) memory requirements and (iii) computational accuracy are analyzed to show the benefits of this novel approach. Also, optimizing the combination of model accuracy and computational speed through the choice of sampling points is explained. [Preview Abstract] |
Monday, July 10, 2017 4:15PM - 4:30PM |
E6.00004: Uncertainty Quantification of the Reverse Taylor Impact Test and Localized Asynchronous Space-Time Algorithm Waad Subber, Alberto Salvadori, Sangmin Lee, Karel Matous The reverse Taylor impact is a common experiment to investigate the dynamical response of materials at high strain rates. To better understand the physical phenomena and to provide a platform for code validation and Uncertainty Quantification (UQ), a co-designed simulation and experimental paradigm is investigated. For validation under uncertainty, quantities of interest (QOIs) within subregions of the computational domain are introduced. For such simulations where regions of interest can be identified, the computational cost for UQ can be reduced by confining the random variability within these regions of interest. This observation inspired us to develop an asynchronous space and time computational algorithm with localized UQ. In the region of interest, the high resolution space and time discretization schemes are used for a stochastic model. Apart from the region of interest, low spatial and temporal resolutions are allowed for a stochastic model with low dimensional representation of uncertainty. The model is exercised on the linear elastodynamics and shows a potential in reducing the UQ computational cost. Although, we consider wave prorogation in solid, the proposed framework is general and can be used for fluid flow problems as well. [Preview Abstract] |
Monday, July 10, 2017 4:30PM - 5:00PM |
E6.00005: Uncertainty due to Imperfect Models of Chemically Reacting Systems Invited Speaker: Robert Moser In complex phenomena involving chemical reactions, such as detonations and hypersonic shocks, the kinetics of the chemical reactions are critical to the physics. Computational models of the phenomena must therefore reliably represent the chemical kinetics. However, even in relatively simple reactions, there are innumerable chemical reaction pathways involving large numbers of intermediate species and excited states. Development of chemical reaction mechanisms for use in computational models thus requires the modeler to identify and characterize a subset of reactions and species that are deemed most important. Further, in computational models of complex systems, chemical mechanisms may need to be simplified further to reduce computational costs. In either case, there are missing species and pathways, which raises the question of how they impact the reliability of simulations of the phenomena. To address this, we consider representations of the uncertainties due to imperfect kinetics models in terms of stochastic operators, which provide a probabilistic description of the missing species and reactions. The ``model inadequacy" representation is formulated to respect constraints imposed by conservation laws and thermodynamics. The approach will be demonstrated on a model of hydrogen-oxygen kinetics in a perfectly stirred reactor. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2021 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
1 Research Road, Ridge, NY 11961-2701
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700