Research on Ekman at the Linné Flow Center, KTH MechanicsDan Henningson, Director
Research on Ekman at the Linné Flow Center, KTH Mechanics
Dan Henningson, Director
I will briefly introduce the Linné Flow Centre, the center that will be responsible for the turbulence simulations on the Ekman computer. Then I will give a couple of examples of research that we plan to perform on the new computer.
http://www.flow.kth.se Funded by VR as an one of the 20 original Linné centers of excellence
Here are almost all of the 60 participating researchers in FLOW. The picture is taken at the recent yearly meeting at Villa Söderås, January 2008. FLOW is one of the 20 original centers of excellence funded by the VR (Swedish Research Council) Linné grants.
Linné FLOW Centre
Vision
“FLOW as an outstanding environment for fundamental research in fluid mechanics, where innovative research is born and future research leaders are fostered”
We have a vision of FLOW as … We emphasize both research and fostering a new generation research leaders.
Activities and infrastructure
Research projects
Graduate school
Summer programs
Workshops/conferences
Seminars and visiting scientists
Infrastructure
World class wind-tunnels and acoustic measurement facilities
Climate and Turbulence computer Ekman
The activities in FLOW involves funding research projects. We also have a new graduate school in fundamental fluid mechanics recently funded by VR with 1.5MSEK/year. We plan summer programs, workshops and conferences. During the last year we had about 40 seminars, including many from visiting scientists, so called Linné visitors. In terms of infrastructure we have world class wind tunnels and acoustic measurement facilities. We have recently had an experimental turbulence measurement Jamboree where several international groups came to KTH to measure in our large-Reynolds number facility. And as you know we were recently part of the consortium that was awarded a grant of 25 MSEK for the “Climate and Turbulence” computer Ekman.
Research areas
Why do flows become turbulent?
How does turbulence behave in the ocean and atmosphere?
Why does turbulence generate noise?
How can we make flows behave the way we want to?
Why do fibers bundle?
We have five main research areas. The first one deals with why flows become turbulent, the second one deals with turbulent flow, for example how flows behave in the ocean and atmosphere. The third also relates to turbulence, but the interest is on how turbulence generate noise. We have an area of dealing with flow control where we are interested to manipulate and optimize flows, and a last area relating to micro and complex flows, where we for example are interested in why fibers bundle in a suspension.
Turbulent flows are everywhere, and they can be described by …
From www.efluids.com
Our computations on Ekman will mainly be of turbulent flow. Such flows can be found everywhere, in the atmosphere, in the eruption of the volcanoes, associated with aircraft, combusion, etc. All of these flows can be described by a single set of partial differential equations, … called the Navier-Stokes equations. Numerical simulations of turbulent flows means solving these equations on supercomputers, like the Ekman computer.
How do we perform numerical experiments?
Cray-1 (1976)
100 Mflops
1 processor Ekman Dell cluster
(2008) 100 Tflops
12000 processors Solve the Navier-Stokes equations for the velocity on grid points using super computers
So, how do we perform these numerical experiments? We solve the Navier-Stokes equations for the velocity on grid points. Here we see an example of a grid around a wing with the grid points where we calculate the velocity depicted in red. With the Ekman computer we can do very fast simulations, about 1.000.000 times faster than on a Cray-1 30 years ago. We do that by using around 10.000 processors at the same time in parallel.
Efficient simulations on many processors
# processors # processors speed-up SIMSON Nek5000 How much faster does the simulation run on many processors? linear scaling measurements (IBM BG/L)
It is important that the parallel computations are efficient and that the computer codes run twice as fast if we double the number of processors. The Nek5000 and SIMPSON are two of our main codes and they perform very well on parallel computers, with almost linear speed-up.
How can simulations help make modern aircraft more environmentally friendly?
EU NACRE project: Concept for quiet, light fuel efficient aircraft Suppressing turbulence on the wings (laminar flow control) improve fuel efficiency
Better models of turbulence on wing surfaces improve engineering design
Example: Airbus green aircraft concept
How can simulations help make modern aircraft more environmentally friendly? This is one question that may be addressed by the simulations we will perform on Ekman. An example is the Airbus green aircraft concept for quiet, light and fuel efficient aircraft we are working on in the EU-project NACRE. One way of making an aircraft fuel efficient is to reduce the amount of turbulence on the wings and engine nacelles by so called laminar flow control. Another is to improve the understanding of turbulence with the aim of reducing the drag on surfaces with turbulence. This require among other things improved engineering models of turbulent flows. These things can be expected from analysis of the simulations performed on Ekman.
Direct Numerical Simulations of all scales in the turbulent flow (no models)
Turbulence close to the surface
Friction Drag Fuel consumption
Let us consider what we can do today in terms of simulating real turbulence, not using engineering models. Turbulence close to the surface of the wing is important because it is responsible for the friction and thus a large part of the drag of the aircraft. This directly translates to the fuel consumption. The largest simulations of turbulence on a wing surface is marked in red, and looking down at a contour plot of the velocity it looks something like this. As you can see it contains a lot of structures. Let us look closer at the turbulent flow …
Direct Numerical Simulations of turbulent flow
Simulations: streamwise velocity in wall-parallel plane 20 cm x 0.8 cm, simulation on BlueGene at PDC Large range of scales require huge number of gridpoints 6144 x 385 x 576 = 1.4 billion grid points
This is results from a simulation of high Reynolds number turbulence using the IBM Blue Gene computer at PDC, it is one of the fastest computers today in Sweden. This simulation uses 1.4 billion grid points, remember the red dots on the grid of a wing I showed you earlier. The whole computational area corresponds to 20x0.8 cm^2, not large compared to the whole aircraft. If we look closer at an area towards the end of the simulations we see that it contains even more structures, some larger ones, like the one in the blue circle and some even smaller ones. We can zoom in once again to see the smallest scale structures in this simulation. It contains streaks of low and high velocity. This is the backbone of turbulence responsible for most of the friction drag of the aircraft.
What can we learn from even larger simulations?
Large scale
experiments Re Domain size Ekman Previous simulations Ciclope project: large pipe flow facility in 130 m Italian tunnel 12288 x 721 x 1152 = 11 billion grid points
Overlap with large scale experiments
What can we learn from even the even larger simulations we can perform on Ekman? We expect to increase our simulations almost a factor of 10 in size, to at least 11 billion grid points. This will make our simulations for the first time overlap with large scale experiments, and not suffer from what scientists call low-Reynolds number effects. Experiments can still go higher in complexity and size, like the ones planned in the large pipe flow facility in an 130 meter long tunnel in Italy, one of the tunnels Mussolini made to secure the production of aircrafts during the second world war. The overlap with large scale experiments can …
What can we learn from even larger simulations?
12288 x 721 x 1152 = 11 billion grid points
Overlap with large scale experiments
Dynamics of large scale turbulent structures and their interaction with small near-wall structures
Provide quantities difficult to measure, input in engineering models of turbulence
Independent prediction of quantities like drag
How to best suppress turbulence on the wing
…
How will the ocean circulation respond to global warming?
Ocean large heat regulator of climate
Great conveyor belt transport warm surface water to north pole and cold water back along bottom
Circulation affected by global warming?
Another example of the simulations we will perform on Ekman is simulations of turbulence in the ocean. How will the ocean circulation respond to global warming? The ocean acts as a large heat regulator of the climate, storing and transporting enormous amounts of heat. Here we see what is called the great conveyor belt. Warm surface water is transported to the north pole where it sinks to the bottom and cold water is transported along the bottom towards the equator. One of the great challenges to science today is to predict how this circulation pattern will respond to global warming. To be able to predict this, it is essential to realize that ...
The ocean is turbulent!
Simulation: ECCO code using MITgcm.
JPL. NASA Ames
… the ocean is turbulent! This is a simulation of the global ocean circulation using the ECCO code, based on the MIT general circulation model, a numerical model designed for study of the atmosphere, ocean, and the climate. The movie is created by by Chris Henze of NASA AMES from a simulation carried out by Dimitris Menemenlis and others at JPL (Jet Propulsion Laboratory) with help from core MITgcm team members and staff from NASA AMES. I want to point out the gulf stream and the western boundary currents and the general turbulent and chaotic nature of the global circulation in the ocean.
The global circulation is very sensitive to the turbulent diffusivity (Nilsson et al., MISU)
There is even
turbulence at
centimeter scale! Smaller scales determine turbulent diffusivity, how fast is the cold water cooling the warmer water above ?
As has been shown by several groups, for example a group Swedish colleagues, ocean circulation is also very sensitive to small scale turbulence at centimeter scales. The cold bottom water is mixed and lifted up by vertical turbulent mixing. To be able to predict how the general circulation will respond to the warming we need to know the turbulent diffusivity, i.e. how fast the cold water is cooling the warmer water above. But what is the turbulent diffusivity?
What are we doing now?
Direct Numerical Simulation of ocean turbulence 2048 x 2048 x 384 = 1.6 billion grid points
This question may be answered by direct numerical simulations. This is an example of what we have done so far, a simulation with about 1.6 billion grid points. It is a little bit too small to get the turbulent diffusivity correctly.
What will we do? Larger DNS of ocean turbulence 4096 x 4096 x 1024 = 17 billion grid points
Ekman may give answer on turbulent diffusivity in ocean Simulation: Kelvin-Helmholz breakdown by Colorado Res. Ass. bridge gap between larger scales affected by density differences and smaller isotropic scales
This is an example of what we can do on Ekman. (4096 x 4096 x 1024 = 16 billion grid points ). Ekman may give us the answer on the turbulent diffusivity, because we will bridge the turbulent scales that are affected by the stratification (density differences in the ocean) and smaller isotropic turbulent scales. The visualization is taken from a simulation of the breakdown of a Kelvin-Helmholz instability in a stratified ocean turbulence, carried out by Joe Werne at Colorado Research Associates, Boulder, Colorado.
Simulations on Ekman will help fill the gap between smallest scales accounted for in climate and engineering models and those of importance in nature
Examples …
Simulations of ocean turbulence to predict turbulent diffusivity influencing how the ocean responds to global warming
Simulations of turbulence in boundary layers on aircraft and ground vehicles to improve engineering predictions of drag
Simulations of the onset of turbulence on wings to develop laminar flow control and lower fuel consumption
As a summary we can say that the Simulations on Ekman will help fill … Examples are simulations of ocean and atmospheric turbulence and simulations of boundary layers such as those on aircraft
Comments