Measurements of Turbulent Boundary Layers with Adverse Pressure Gradients and Curvature
Abstract
When sea-going vehicles maneuver, the flow can separate from the surface, yielding large forces that oppose the maneuver. Current computational techniques to predict separation are inaccurate, preventing the evaluation of new designs that could mitigate the problem. Machine learning is a promising avenue to generate new models that can make computational techniques more accurate in predictingflow separation, but machine learning requires data to learn from. This project will provide data to inform machine learning for computational prediction. The project will experimentally measure the shear stress at the wall of a curved surface and the velocity field of the fluid motion in the presence of adverse pressure gradients of varying strengths, up to those that generate separation. This project is designed to provide data sets for machine learning by measuring information that machine learning models need (wall shear stress and velocity near the wall) and by measuring at many conditions. Twenty pressure gradient configurations, five surface curvatures, and four Reynolds numbers will be studied, yielding a rich parameter space from which robust models can be developed through machine learning. These many configurations are made possible by a flexible experimental approach in which elastic sheets are bent into different configurations, allowing the same material to rapidly be reconfigured to generate different pressure gradients and surface curvatures. Wall shear stress will be measured using state-of-the-art MEMS devices, integrated alongside high-speed pressureprobes that will allow for careful calibration. Flow velocity will be measured using particle image velocimetry in both the streamwisewall-normal plane, providing information about the motion of the separation point, and in the spanwisewall-normal plane, providing information about the three-dimensionality of the flow near the wall. The statistics of the data will be made publicly availableand the raw data will be provided to machine learning experts and computationalists for their use in generating improved models.Theproject will also provide new insights into the physics of separation through the application of a novel data analysis technique that was developed in the laboratory of the principal investigator. Prior to separation, the flow near the wall beneath large-scale slow fluid motions often reverses directions, while the flow beneath large-scale fast fluid motions does not reverse. A new variant ofa commonly used analysis technique, proper orthogonal decomposition, will be used to quantify the structure of the flow near the wall in the presence of fast and slow large-scale motions separately, yielding accurate representations of reversed flow events.In total, the project will provide critical data on the flow of turbulent boundary layers on curved surfaces under pressure gradients. Thedata will be directly provided to computational collaborators to inform models, and statistics of the data will be published and made publicly available to enable model validation. A large parameter space will be studied to yield robust models that can improve the prediction of flow separation on Navy vehicles. New physical insights will be identified regarding the flow state just before separation, informing future modeling efforts.*Approved for public release*
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 05, 2021
- Source ID
- N000142112648
Entities
People
- Theresa Saxton-Fox
Organizations
- Office of Naval Research
- United States Navy
- University of Illinois Urbana–Champaign