NICOP - Using a novel machine-learning framework for accurate flow and noise predictions at high Reynolds numbers

Abstract

The proposed research aims to exploit recent developments in data-driven and reduced- order modelling for near-wall turbulence to enable accurate and affordable flow and noise predictions for maritime and aviation applications of the U.S. Navy at high Reynolds num"bers. In particular, a novel machine-learning approach developed at the University of Melbourne will be employed to i) train RANS cl""osures with improved accuracy for non-equilibrium turbulent flows, and ii) allow the use of a novel modal representation of near-wal""l turbulence, called resolvent analysis, for prediction of the surface pressure field, based only on mean flow quantities from the t""rained RANS models. Better predictive capabilities are essential to future improvements in maneuverability, speed, and energy effici""ency, as well as survivability, e.g., pertaining to acoustic signature. Results will be presented at conferences, published in leadi""ng journals and turbulence and noise prediction models will be generated for use in other research.Specify relevance to ONR, the U""S Naval Research Enterprise (NRE), and/ or other US DoD.The proposed research aims to improve the predictive accuracy of low order"" models, specifically RANS closure models and noise prediction models, as used in a design context. The research is therefore highly" relevant to the US Navy as better predictive capabilities of affordable numerical approaches is essential for future improvements i"n maneuverability, speed, and energy efficiency, as well as survivability, e.g., pertaining to acoustic signature of naval applicati"ons.Figure 2 exemplifies potential naval applications that can be targeted by the proposed low-order modeling approaches. The present goals of developing an accurate and affordable flow and noise prediction framework also holds the longer term promise to use more" reliable and robust tools for technology development using design optimization in a number of areas, e.g. for drag reduction, or tu""rbomachinery noise. In that sense, the current work is well aligned with the Naval S&T Strategy~s focus areas ~Power & Energy~ and ~"Platform Design & Survivability~.Identify all US collaborators especially if there are any ONR Program Officers/ codes that are collaborators (even if the ONR PO is not actively contributing funds to the effort).N/AState the desired outcomes of this research" effort (e.g., conference presentation, journal article, algorithm, tool, additional research).It is anticipated that results will"" be presented at conferences, published in leading journals. Turbulence and noise prediction models will be generated that can be us""ed in other research, e.g. in future design optimization studies of more efficient platforms with improved survivability.

Document Details

Document Type
DoD Grant Award
Publication Date
Jun 09, 2017
Source ID
N629091712083

Entities

People

  • Richard Sandberg

Organizations

  • Office of Naval Research
  • United States Navy
  • University of Melbourne

Tags

Readers

  • Computational Fluid Dynamics (CFD)
  • Distributed Systems and Data Platform Development
  • Maritime and Naval Warfare Studies

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Neural Networks