Using machine-learning to develop a novel selective scale-resolving framework for accurate flow and noise predictions of Naval applications
Abstract
The proposed research aims to exploit recent developments in simulation and data-drivenmodelling for near-wall turbulence at high Reynolds numbers to enable accurate and affordableflow and noise predictions for maritime and aviation applications of the U.S. Navy. Building onprevious work that focused on steady Reynolds-averaged Navier-Stokes (RANS) modeling, theunique machine-learning approach developed at the University of Melbourne will be employedto train a new breed of non-traditional selective scale resolving simulations, employing machinelearntclosures, achieving accurate predictions of non-equilibrium turbulent flows, in particularstatistically three-dimensional and adverse pressure gradient flows. Initially, models will bedeveloped and tested for building block flows, e.g. stern flows, followed by an approachautomatically selecting the appropriate model for different flow regions. Ultimately, the overallframework will be demonstrated on a submarine hull with appendages configuration. Theunsteady flow predictions, along with information contained in the custom models will helprealize improvements in propulsion and powering. Future coupling of the accurate flow fieldpredictions with acoustic analogies also will enable reliable predictions of turbulence ingestednoise, improving survivability, e.g., pertaining to acoustic signature. Results will be presented atconferences, published in leading journals and turbulence and noise prediction models will begenerated for use in other research.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 31, 2020
- Source ID
- N629092012046
Entities
People
- Richard Sandberg
Organizations
- Office of Naval Research
- United States Navy
- University of Melbourne