CFD - Computational Fluid Dynamics, Motorsport, Formula 1

[marekk]

[quote="hollus"
How heat moves from one cell to the next is exactly understood.
How mass moves from one cell to the next is exactly understood.
How internal molecular energy is traded for external kinetic energy for the mass of air is exactly understood. And so on.
[/quote]

I wouldn't be so generous in using the word " exactly".
Currently working on new compute cluster for quantum chemistry guys on our local uni (small one, 1TB RAM, about 400 cores, infiniband interconnect) - working full throttle now for 2 weeks.
Problem size - less then 100 atoms. Not even exact solution, still using models and assusmptions. Should be ready by the end of April. Then one need anaother few weeks of skilled people's hard work to evaluate and understand the results.

[Th0R]

In less words, CFD works just like FEA, only it's applied to fluids.

That's not exactly true. In CFD mostly finite volume methods are used while FEA mostly uses finite element methods.

Well FEA is by definition.. the finite element way of doing things. And yes, some CFD packages do use FVM rather than FEM. In either event, to say that CFD works like FEA is not particularly accurate since you could do CFD by FE.

Sorry that was my mistake. For me usually is FEA for structure mechanics cause that's the field it was normally used in my studys. Of course, you can use the different discretisation concepts for different problems. I just wanted to state which way it is normally done.

[Just_a_fan]

How do the CFD programs account for the various theories of lift generation? Those explanations based directly on Newton’s laws of motion and explanations based on Bernoulli’s principle.

Is it one or the other or are their values additive?

Brian

Newton's laws of motion. CFD tries to solve the Navier-Stokes equations (or simplified versions thereof) and these are effectively Newton applied to a fluid which results in a series of approximated conservation equations (momentum, mass, energy) which are then solved.

I think.

[Jersey Tom]

My understanding is that the "Newtonian" (momentum transfer) and "Bernoulli" (equal transit) theories of lift are both BS. It's all about flow turning. Then again, I'm not an aero guy.

In any event.. point is you're just solving for something at individual discrete points.. not a closed form analytic solution. No worry of various theories, it just is what it is.

[hardingfv32]

CFD tries to solve the Navier-Stokes equations (or simplified versions thereof)

Yes, I see the use of Navier-Stokes in most the papers I review. Not having any understanding for these equations, I move on to the next section.

Jersey Tom...

I have not sure I have run into 'flow turning'. Would you say this is an accurate descriptor?

Brian

[Lycoming]

Everything that I've read says that what JT says is more or less correct, though his explanation of it is not what you would call "rigorous". That said, I unfortunately cannot produce a more rigorous definition. The best I can give is that it relates to the way the airflow follows the curve of the airfoil.

[KeithYoung]

CFD tries to solve the Navier-Stokes equations (or simplified versions thereof)

Yes, I see the use of Navier-Stokes in most the papers I review. Not having any understanding for these equations, I move on to the next section.

Jersey Tom...

I have not sure I have run into 'flow turning'. Would you say this is an accurate descriptor?

Brian

Let me know if this helps.

http://racingtech.wordpress.com/2011/11 ... on-take-2/

[Pierce89]

That's not exactly true. In CFD mostly finite volume methods are used while FEA mostly uses finite element methods.

Well FEA is by definition.. the finite element way of doing things. And yes, some CFD packages do use FVM rather than FEM. In either event, to say that CFD works like FEA is not particularly accurate since you could do CFD by FE.

Sorry that was my mistake. For me usually is FEA for structure mechanics cause that's the field it was normally used in my studys. Of course, you can use the different discretisation concepts for different problems. I just wanted to state which way it is normally done.

wasn't trying to be the most "correct" just trying to give people a visualization of how cfd works on a cell by cell basis.

[Just_a_fan]

JT is sort of right. But lift generation is Newtonian in that it deals with the conservation of momentum (one of Newton's laws, of course) and it is Bernoullian in that it deals with conservation of energy (Bernoulli's equation does this). But there are other things going on which these two can not deal with because they don't include them in their equations.

That is to say that the Newtonian and Bernoullian views are, in effect, merely partial explanations of what's going on (one might describe them as approximations). No different to using the Rutherford-Bohr atomic model to describe what an atom is to a 10 year old. It's wrong (because it depicts electrons as little balls in orbit around the nucleus rather than as wave functions (shells of probability surrounding the nucleus)) but it gets the basic idea across without all of that horrid quantum mathematics stuff.

From http://www.grc.nasa.gov/WWW/K-12/airplane/bernnew.html :

The real details of how an object generates lift are very complex and do not lend themselves to simplification. For a gas, we have to simultaneously conserve the mass, momentum, and energy in the flow. Newton's laws of motion are statements concerning the conservation of momentum. Bernoulli's equation is derived by considering conservation of energy. So both of these equations are satisfied in the generation of lift; both are correct. The conservation of mass introduces a lot of complexity into the analysis and understanding of aerodynamic problems. For example, from the conservation of mass, a change in the velocity of a gas in one direction results in a change in the velocity of the gas in a direction perpendicular to the original change. This is very different from the motion of solids, on which we base most of our experiences in physics. The simultaneous conservation of mass, momentum, and energy of a fluid (while neglecting the effects of air viscosity) are called the Euler Equations after Leonard Euler. Euler was a student of Johann Bernoulli, Daniel's father, and for a time had worked with Daniel Bernoulli in St. Petersburg. If we include the effects of viscosity, we have the Navier-Stokes Equations which are named after two independent researchers in France and in England. To truly understand the details of the generation of lift, one has to have a good working knowledge of the Euler Equations.

See here: http://www.grc.nasa.gov/WWW/K-12/airplane/right2.html

And here: http://www.grc.nasa.gov/WWW/K-12/airplane/wrong1.html (follow the links at the bottom to see the other theories considered to be incorrect).

It's a complex subject. Which might be why some want its influence on the sport reduced...

[hardingfv32]

Generally speaking would you say that all the CFD code is generated by the above stated formulas or are there some 'magic numbers' added to allow the CFD code match reality?

Brian

[Pierce89]

Generally speaking would you say that all the CFD code is generated by the above stated formulas or are there some 'magic numbers' added to allow the CFD code match reality?

Brian

they end up fudging some factors to get the results to correlate with the tunnel,track etc.

[Just_a_fan]

All modelling contains assumptions (magic numbers) of some form.

[MadMatt]

My understanding is that the "Newtonian" (momentum transfer) and "Bernoulli" (equal transit) theories of lift are both BS. It's all about flow turning. Then again, I'm not an aero guy.

In any event.. point is you're just solving for something at individual discrete points.. not a closed form analytic solution. No worry of various theories, it just is what it is.

So can you shut up maybe rather than saying REAL bs ?

[volarchico]

Generally speaking would you say that all the CFD code is generated by the above stated formulas or are there some 'magic numbers' added to allow the CFD code match reality?

Brian

The "magic numbers" for CFD are used to correlate "real life" track data to wind tunnel models to CFD results. With the wind tunnel, there are approximations being made for scale and boundary conditions (wall effects, mounting equipment, etc). With CFD there are approximations being made for grid size, turbulence models, and computational convergence.

As others have said, the basis of all CFD is the Navier-Stokes equations. If you could simultaneously solve this set of equations at every single point in space for a sufficiently large volume surrounding an object you could get an exact representation of the "real world" (assuming your CAD model is also exactly matching the real world).

The problem is created because computers are currently not powerful enough to solve the equations at a high enough resolution. Some estimates say at the current rate of increase in computational power, it could still be 100 years before a full aircraft can be modeled by Direct Numerical Simulation (the technical term for using the N-S equations for all scales of computation) in a useable time frame (days instead of years). I'll assume it will be a similar wait for an F1 car because they are fairly complex, even if the Reynold's number is much less than most aircraft. So for the really small scales, they have developed approximation models for the flow and these are called turbulence models. They are a big part of the "magic numbers" you refer to.

Now imagine that every "solution" is at one instant in time, and most everything at an F1 scale fluctuates with time. The flow itself fluctuates with time but so does the geometry (wing flex, steering angle, tire deflection, ride height, etc). So now the CFD needs to be run to find average values across a given time and configuration.

Makes me want to live to be 200 yrs old so I can be around when they start being able to simulate the whole vehicle at high accuracy.

[Kiril Varbanov]

I cascaded this one through Twitter this morning, but here it is again:
CFD Simulation of Vortex Shedding Behind a Cylinder - even though the geometry of the cylinder is supposedly simple, the vortex isn't smooth at all.
Video by Symscape - http://www.youtube.com/watch?v=_IL-K9_5XmY

P.S. Sorry if it's a bit out of context. I'm fairly tied to aerodynamics and CFD, so I decided to post this simple, yet interesting video.