FLOW: Numerical experiments in a virtual wind tunnel
Being at the forefront of computational science in the coming years means effectively utilizing millions of processors for simulations of physical phenomena. Such capabilities will allow accurate large scale numerical simulations to take a role analogous to that of physical experiments. This paradigm shift has tremendous implications for many areas, including computation of aeronautical flows. The primary goal of the proposed project is to drive fluid simulation capabilities beyond those of typical laboratory experiments, to develop a virtual wind tunnel. This entails capabilities to accurately and simultaneously simulate a number of physical phenomena of flows over wings, such as laminar-turbulent transition, transitional separation bubbles, turbulence on the wing surface and turbulent wakes behind the wing. A virtual wind tunnel allows simpler, more accurate and less expensive wing testing. It also provides a wealth of new information for fluid researchers to tap, such as access to complete velocity fields, their sensitivity to important parameters and access to flow regimes not possible in physical experiments. These new capabilities will be a strong competitive advantage for the research group and KTH. Two ongoing physical experiments of relevance for laminar wing design provide an excellent opportunity for proof of concept simulations (see figures below). Laminar wings include the complicated flow physics described, regions of transitional, turbulent and separated flows. Both experiments are performed in projects where researchers from our group have key roles. The first is part of the EU-project RECEPT aiming at improved possibilities for laminar wing design. The second is a free-flight experiment at the University of Darmstadt on a glider wing-glove, where feed-back control is used to annihilate instability waves to increase the laminar part of the wing. Keeping the flow laminar reduces drag, which results in lower fuel consumption and lower carbon dioxide emissions, contributing to an environmentally sustainable air transport system.
Simulation techniques have been used in aerodynamics research for several decades, but always had to be validated with results from corresponding experiments. This is mainly due to inaccuracies in the engineering models describing laminar-turbulent transition, flow separation and the turbulence characteristics. Direct numerical simulation (DNS) methods avoid these models by resolving all of the relevant scales down to the smallest turbulent features. DNS has been used to accurately predict individual physical phenomena in simple geometries, but never combined on a complete wing. See the figure below for a state of the art simulation on part of a wing. Simultaneous capture of all the complicated flow physics in direct numerical simulation of a complete wing flow will be made possible through a combination of unique capabilities present in our research group. These include numerical and e-Science tools, solid expertise and a strong track record in simulating and analyzing relevant flow physics and access to large computational resources.
Our group has developed direct numerical simulation codes and been at the forefront of accurate simulations of fluid physics during the last two decades, performing some of the largest and the most spectacular simulations in recent years. We are working with the versatile, highly accurate and efficient computational software Nek5000 from Argonne National Laboratory. In collaboration with the main developer Dr. Paul Fischer, we are improving this software for use on future exascale computers within the EU-project CRESTA (Collaborative Research into Exascale Systemware, Tools and Applications). The leading role of our research group in the Swedish e-Science Research Centre (SeRC) and the Linné FLOW Centre is of paramount importance for the success of the proposed project. We are collaborating with the best researchers in Sweden in the areas of complex data management and visualization within SeRC and heading the research dealing with the relevant flow physics within FLOW. We also have extensive experience from a 20-year participation in the majority of the EU-funded laminar wing projects, from detailed analysis of the flow physics to development of laminar wing design methods. In addition we have substantial computer resources available. We have the largest shares of the VR/KTH funded Cray computer Lindgren and compete successfully for the largest European resources through the PRACE (Partnership for Advanced Computing in Europe) programme, at present with the first Swedish allocation (46 million core hours).
The virtual wind tunnel
The scientific methodology is based on large-scale computer simulations to obtain new data for the flows around the wings, but also to post process, visualize, synthesize, interact with the data and deduce new innovative results. The challenges are similar for both the proof of concept studies, and for the development of a virtual wind tunnel capability in general. First a high quality numerical discretization or grid of the area around the wing needs to be produced. The grid determines the number of degrees of freedom in the simulations, which we estimate will be around 100 billion for the studies proposed, more than one order of magnitude larger than we have previously used. We estimate that typical simulation cases will require 100-1000 million core hours, something available using the next generation fastest super computers. Such a simulation will generate about a peta-byte of data. We will utilize Swestore and the coming European EUDAT storage systems for making the data available online for our research group and other collaborators. Novel methods for visualization and interaction with data, such as augmented reality techniques, will be utilized through the e-Science center SeRC and its visualization and interaction studios, the C-center at the Linköping University, Norrköping campus and the KAW financed VIC-center at KTH.
The proof of concept studies have been chosen for the results of the simulations to contribute to the solution of major open questions, such as how turbulence in the air generate turbulence on the wing, how turbulence interact with separated flow and flow separation on a wing, and how one can enhance engineering predictions of such processes. In addition we will simulate flow control on full wings where the flow is manipulated to delay the transition to turbulence and decrease drag on the wing surface. The group has been involved in transferring systems control methods to aeronautical flows for the last decade.
Significance and strategic impact
The present project aims to take the first steps toward the use direct numerical simulation techniques to replace and improve typical aeronautical wind tunnel experiments regularly performed by aeronautical researchers. The future vision is to perform simulations that give completely new possibilities for full wing simulations, such as finding important flow sensitivities and gradients as part of so called adjoint solutions, have accessibility to flow regimes not possible in physical experiments and to tap the enormous amounts of data by new visualization and interaction possibilities. Such greatly improved turbulence simulation capabilities are also very important for research in biomedical flows, turbulent combustion, papermaking, atmospheric and ocean flows and in sustainable energy production such as wind and water power. For example, the Nek5000 code is used in our group to perform simulations of how turbulence generated by stenosis influences the circulatory system and to analyze the turbulence behind wind turbines.