Data-Driven Input-Output Models for Reacting, High-Enthalpy Flows
Abstract
ABSTRACT (APPROVED FOR PUBLIC RELEASE)The extraction and description of principal structures and mechanisms in reacting, high-enthalpyflows is a key component in the understanding of these complex flows. The proposed effort aimsat developing data-driven and model-based methods for the input-output analysis of compressible,high-speed flows. An encompassing viewpoint, bring together advanced optimization methods,randomized algorithms and physics-tailored machine learning, will be taken to develop and benchmarktools for the effective analysis of these types of flows.Turbulent, reacting flows in hypersonic flight applications, both external to the vehicleand internalto the combustor, rank among the most complex in fluid dynamics, owing to a compoundinteraction of multiple physical processes (aerodynamics, turbulence, acoustics, shock-dynamics,aerothermochemistry, combustion, flame dynamics, nonequilibrium phenomena) over a very widerange of spatio-temporal scales. The reciprocal influence of these processes and the resulting emergenceof dominant flow features are crucial to a fundamental understanding of the flow physics.The difficulty in developing predictive models is compounded by the complexities inherent in obtainingexperimental data and providing high-fidelity simulations with detailed thermochemistryfor external and internal flows relevant to hypersonic vehicles. Recent advances in linear algebra,operator approximation, anddynamical systems approaches have provided exciting opportunitiesthat enable a detailed understanding of a range of multiscale physical phenomena, including turbulentflows. Reduced-order representations and input-output transfer function relationships derivedfromthe equations of motion have elucidated low-rank behavior (the naturally-amplified, dynamicallysignificant, hidden structure) in chaotic flows. In this proposal, our focus lies on reduced complexityunderstanding of the physics underlying the interaction of high-speed, turbulent flowswith chemistry using this novel approach to mathematical modeling.We distill the underlying mathematical foundations of these approaches and multiscale modelingto reimagine a modeling strategy with a focus on the specific challenges associatedwithhigh-speed flow and the complexity of the range of possible interactions between flow aerodynamics,turbulence and chemistry. Our ultimate objective is a framework to exploit the structure of thegoverning equations to generate reduced-order dynamical systems models for the interaction ofturbulence with chemistry in high-enthalpy flows. We expect that these models will inform futureturbulence models by characterizing the hidden variables regulating the high-dimensionality of thegoverning equations, with the potential toidentify the most propitious forms for sub-grid scalemodels and improved predictability of computational tools in the hypersonic regime.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 24, 2023
- Source ID
- N000142312646
Entities
People
- Beverley McKeon
Organizations
- Office of Naval Research
- Stanford University
- United States Navy