CEWAF - Curvature Effects in Wall-bounded Flows
Department of Engineering Mechanics, KTH, Stockholm, Sweden
Local Project ID:
HPC Platform used:
JUWELS of JSC
In many engineering applications, fluids or gases flow over curved surfaces. Some examples are the flow over wings and over wind and gas turbine blades. The streamlines of these flows are curved and this streamline curvature has a large impact on the flow and flow drag.
For developing and designing efficient engineering applications like pumps and turbines it is important to have models that can accurately predict flows over curved surfaces. However, engineering flow models often do not correctly capture streamline curvature on flows and therefore cannot accurately predict the flow and drag over such curved surfaces. The underlying reason is that the influence of streamline curvature on flows is complex and not yet well understood.
In the project, large-scale so-called direct numerical simulations are carried out of turbulent flows in mildly to strongly curved channels. The turbulent fluid motions are fully resolved in time and space on a very fine mesh with up to 3 billion grid points and small time steps with a high-order numerical algorithm. These large simulations were performed on the HPC system JUWELS at JSC and are highly accurate since the fundamental governing equations for fluid motions are solved exactly without using any modelling approximations.
The performed simulations give us a much better physical understanding of the influence of streamline curvature on turbulent flows over curved surfaces. The simulations show that the turbulent eddies are much larger and strongly amplified on the outer concave side of the curved channel, whereas the turbulent eddies on the inner convex channel side are smaller and much less intense. As a consequence, the flow drag is much larger on the concave than on the convex side of the channel and this difference grows when the curvature of the channel increases. In the simulation of the strongest curved channel the streamline curvature effects are so large that the turbulence ceases to exist locally on the convex side and the flow there becomes laminar and more regular. The large turbulent eddies or vortices on the concave side of the curved channel form very elongated roll cells parallel with the flow direction. Figure 1 shows a visualization of the flow field in one of the simulations at one time instant in a plane parallel to the curved wall on the concave channel side. The imprint of the large roll cells is obvious in the visualization.
The large eddies or roll cells on the concave channel side are very efficient at transporting momentum as well as heat and this results in high drag and heat transfer rates. By contrast, the damped turbulence and eddies on the convex channel side results in reduced drag and heat transfer rates. All these observations highlight the strong and complex influence of streamline curvature on turbulent flows over curved surfaces.
Besides insights into the streamline curvature effects, the simulations have also produced valuable and highly accurate data that can be used to develop and validate new and better engineering models for turbulent flows over curved surfaces. A promising novel approach to develop better and more advanced engineering models is machine learning. The numerical data produced in this project are very suitable for training and testing new models constructed via machine learning techniques. Ultimately, this work will contribute to the development and design of smarter and more efficient engineering applications involving flows over curved surfaces.
G. Brethouwer, Turbulent flow in curved channels, submitted for publication.
Dr. Geert Brethouwer
Department of Engineering Mechanics
KTH Royal Institute of Technology
SE-100 44 Stockholm, Sweden
e-mail: geert [@] mech.kth.se
NOTE: This simulation project was made possible by PRACE (Partnership for Advanced Computing in Europe) allocating a computing time grant on GCS HPC system JUWELS of the Jülich Supercomputing Centre (JSC). GCS is a hosting member of PRACE.
Local project ID: PRA108