Stochastic framework for cavitating flows: hybrid model and simulation (White Paper Tracking #: 21-0
Abstract
Prediction of complex, turbulent multiphase and cavitating flows is a crucial challenge across multiple US Navy applications. Howeve,r, the extreme spatial and temporal scale separations inherent to these flows make ab initio simulations out of reach for even the l,atest exascale computers. Similar challenges have prompted the development of statistics-based computational methods for particulate, flows, bypassing much of the computational cost of fully resolved simulations. Our previous research applied these techniques to,dilute multiphase flows of spherical cavitation bubbles. Such techniques are a critical stepping-stone to more advanced stochastic r,epresentations of complex flows. However, they remain fundamentally limited in three important ways: (i) the sub-grid models they re,ly upon ignore essential physics that reduce their accuracy, (ii) they require sophisticated and sometimes fragile statistical techn,iques to render the equations computationally tractable, and (iii) they cannot reach flows that involve both large-scale, non-spheri,cal cavities and small-scale bubbles. The most critical naval applications fall into this latter category, though the three are clo,sely coupled.We propose two separate but interrelated efforts to develop a general framework that overcomes these limitations and en,ables predictive simulations. The Georgia Tech effort (PI Bryngelson) will formulate a fully general population balance model. This, approach affords a mesoscale representation of the physics associated with complex bubbly cavitating flows at a stochastic level.,The Caltech effort (PI Colonius) will develop a hybrid modeling paradigm that incorporates the mesoscale bubble dynamics as a sub-gr,id model that interacts with the interface-resolved multiphase flow via dynamic mass exchange between the continuous (resolved) and,dispersed (unresolved) gas phases. Both projects rely on a careful, building-block approach to solve the respective closure problems,, and feature traditional and modern aspects of data-driven science. These entail the generation of high-resolution ab initio valida,tion and training data, the use of sophisticated statistical and machine learning techniques, and portable, GPU-optimized simulation, algorithms for key model components. The framework will be developed and tested in our open-source flow solver, MFC, and ported to,capable solvers with strong community adoption like ECOGEN.Both projects lead to a common code with state-of-the-art stochastic meso,scale bubble models capable of functioning independently or in hybrid mode to simulate multiscale, multiphase flows with resolved an,d sub-grid cavities. A series of progressively more challenging model problems will be considered, including cavitating flows over w,edges, spheres, and back-steps, for which experimental data is available for validation. We will address key physical issues such as, sheet-to-cloud cavitation transition and its sensitivity to mesoscale stochastic inputs such as nuclei populations and their conten,ts. Our simulations will also generate data ideal for further refinement of the closure models. Despite being crafted for multiscal,e cavitating flows, our proposed efforts will generate a fundamentally general framework for multiscale disperse flows in general, a, first of its kind.Approved for Public Release
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 13, 2022
- Source ID
- N000142212518
Entities
People
- Tim Colonius
Organizations
- California Institute of Technology
- Office of Naval Research
- United States Navy