CFD analysis of Lightning McQueen

[julien.decharentenay]

I have been playing a bit with 3D scanning. Small objects are easier to scan and play with than big one, so I tried scanning and running a CFD analysis of Lightning McQueen from Cars (the movie). It is still very much a work-in-progress. I made a short video of the process for discussion and giving ideas to other:

http://youtu.be/Cn7rKDp2xVY

[silente]

very interesting indeed!

could add some more words about the approach (above all about the 3D scanning techique) and about this website performing "CFD on demand" (assumptions, meshes features, results available, etc)?

I would see a potential to do that on real cars models, maybe accurate and big scale ones, something like what Suzuki has done with his Scale Wind Tunnel but in CFD...Could be interesting.

[MOWOG]

That is awesome! Loved the video. Kudos on doing a first class job! =D>

[silente]

http://www.youtube.com/watch?v=mpMOpyLb ... re=related

this was what i meant!

[andylaurence]

Building a little tunnel like that is a lot of effort and takes a lot of space. Then there's the models to make, although you can get those 3D printed pretty cheaply. CFD is hard for an amateur to get configured and running, so having a site to upload a file and have it processed is brilliant. I uploaded a basic representation of my car to test the site out and blogged about it. Quite a bit of that was transferrable to the actual car and I've started adding parts for the 2014 aero pack. I'll be uploading photos and some analysis of that soon.

[flynfrog]

I have been playing a bit with 3D scanning. Small objects are easier to scan and play with than big one, so I tried scanning and running a CFD analysis of Lightning McQueen from Cars (the movie). It is still very much a work-in-progress. I made a short video of the process for discussion and giving ideas to other:

http://youtu.be/Cn7rKDp2xVY

do you have a little more detail on your scanning.

[xxChrisxx]

CFD is hard for an amateur to get configured and running, so having a site to upload a file and have it processed is brilliant.

GIGO

[andylaurence]

CFD is hard for an amateur to get configured and running, so having a site to upload a file and have it processed is brilliant.

GIGO

Thanks for that considered constructive analysis. Garbage isn't an on/off situation. There's ideal (nothing is) and varying degrees from that. An F1 team might be looking at 0.1mm accuracy on their aero parts, but that's not true of club racing. There's a lot to learn from basic analysis. People still use 2D analysis to break down to which wing profile is most appropriate in a situation. Bear in mind that the target audience for this product is people whose cars were designed by a bloke in a garage with a pencil, not international race teams. Picking a direction is much easier with reduced accuracy when a simple change could lop half a second off your lap time, rather than the tiny increments a top level team would be looking at.

[xxChrisxx]

CFD is hard for an amateur to get configured and running, so having a site to upload a file and have it processed is brilliant.

GIGO

Thanks for that considered constructive analysis. Garbage isn't an on/off situation. There's ideal (nothing is) and varying degrees from that. An F1 team might be looking at 0.1mm accuracy on their aero parts, but that's not true of club racing. There's a lot to learn from basic analysis. People still use 2D analysis to break down to which wing profile is most appropriate in a situation. Bear in mind that the target audience for this product is people whose cars were designed by a bloke in a garage with a pencil, not international race teams. Picking a direction is much easier with reduced accuracy when a simple change could lop half a second off your lap time, rather than the tiny increments a top level team would be looking at.

You haven't done any CFD before have you? My issue isn't with the geometry.
This is a black box job. You put a model in, and get some pretty pictures out.

The first and most obvious question is: How do you know that what they have produced isn't a load of ---?
(put more nicely, how do your know their assumptions are in any way valid)

How do we know that their mesh is of any sort of quality?
How do you know they have decent boundary conditions?
How do you know that their turbulence model is suitable?
Are they running a suitable near wall model?
etc, etc, etc

With a model that takes 2 days from geometry to report that costs a few quid. You aren't getting a job that had involved any sort of thought. The results are nothing more than pictures with lines on.


As a more basic example:
You come to me asking me to calculate the area of your circle of radius 1. I stick it in my black box area-o-meter. The answer comes out as 3. Becucase I assumed pi = 3.
It's perfectly valid calculation, but the assumption is too far out to be considered useful.
This is simple enough to see where it's gone wrong. When assuming many many variables, it gets a little muddier.

The amateur would learn more about his car by sicking wool tufts to the surface and taking some photos.

[Lycoming]

There's a lot to learn from basic analysis.

Sure, if it's done correctly. What turbulence model do you use? How do you set your boundary conditions? How do you know that your solution has converged? How do you know that the mesh is detailed enough in the areas where it really needs to be? 2D or 3D? a basic analysis is really just less expensive in terms of computation time. The knowledge required to set it up doesn't drop to the point where anybody can upload a CAD file to the internet and get a useful result out of it.

Picking a direction is much easier with reduced accuracy when a simple change could lop half a second off your lap time, rather than the tiny increments a top level team would be looking at.

This is absolutely correct... but it is very, very easy to end up with a result that looks plausible, yet is complete garbage; too far away from reality to be useful. With CFD, if you want a useful result, you need to know what you're doing. Uploading a geometry to the internet and getting an interpretation e-mailed to you is... not quite that. If you don't know what's going on, all you've done is make some colorful pictures. I can do that too, and for free! so could any 4 year old.

[andylaurence]

As a more basic example:
You come to me asking me to calculate the area of your circle of radius 1. I stick it in my black box area-o-meter. The answer comes out as 3. Becucase I assumed pi = 3.
It's perfectly valid calculation, but the assumption is too far out to be considered useful.
This is simple enough to see where it's gone wrong. When assuming many many variables, it gets a little muddier.

Your assumption is that the assumption is too far out to be considered useful. What if you had no idea what the area was? You'd be very happy to know that it was somewhere near 3. Yes, you'd be happier if someone told you 3.1 or 3.14, but you're closer to your goal.

The amateur would learn more about his car by sicking wool tufts to the surface and taking some photos.

I agree entirely, but testing is expensive for the amateur. Most competitors I know don't test at all, often because every penny goes into competition. Modifications they do are based on what's discussed in the paddock, seen on other cars or read about in forums. Being able to try a few ideas and get finger-in-the-air answers is a lot better than copying what someone else has (incorrectly) done. Remember that when you know nothing, something has proved useful. Most importantly, when that new part is fitted, it can be validated in practice at the race meeting rather than renting a track for a test session only to find out your uneducated guess was completely wrong. Wool tufts, manometers or pressure sensors on a data logger can be used and all are cheaper than a single test session.

[Lycoming]

You misunderstand. Partly because it was a bad analogy, but let's see if I can rectify that.

The problem is not so much that liberal assumptions are made; we all know that air is in fact compressible but with racecars, we usually assume incompressibility. The problem is that you don't understand what the black box is doing. you don't know what assumptions the box is making, therefore you don't know how valid your result is. You cannot place an error bound on your result.

In the above example of area calculation, the box could tell you that a radius of 1 gives an area of 9 while a radius of 2 gives an area of 3. Is that still a useful result? Obviously you can tell that this is wrong because it's a simple algebraic equation, but would you be able to tell if the black box was numerically solving the navier-stokes equations? Sure if it tells you your single element NACA0012 has a lift coefficient of 125 at 4 degrees angle of attack, something is probably wrong, but not every incorrect result is that obvious, and not every situation is that simple.

The bottom line is, in the process of using a black box CFD program, you don't know what you're doing. You don't know that you're off by 0.14, as in the above example; In fact, you have no idea how far off you are. You could easily be orders of magnitude off. Because of that, it doesn't bring you any new information, it does nothing to refine your guess of what ball park the result should be in. indeed if you don't know enough about fluid mechanics, you won't even know how large the ballpark is. It doesn't narrow anything down, and it doesn't tell you anything new. And that is why the result is useless.

[xxChrisxx]

I couldn't have put it better myself.... we know this becauase I didn't! Those are exactly the points I wanted to make.
You can use a model, even a poor one, if you know the assumptions (and therefore the limitations).

Not knowing what was done and why, makes the entire thing an unknown quantity.

[andylaurence]

I think you've missed my point too. 9 is closer than "I dunno"...

To illustrate my point about what an amateur could learn from this, if you asked most people in the paddock what direction the air was moving on the side edge of the bonnet, they would answer incorrectly. You don't need to know how big the ball park is to know that the results will show that air isn't moving fore-aft over the side edge of the bonnet. The CFD will show this. That might bring a whole new level of understanding to an amateur competitor. Have you seen some of the stuff that appears at the very bottom of the ladder? Vents that open into an obviously high pressure area, intakes from a low pressure zone. If you've literally no idea what's high pressure and what's low pressure, this will help. That the top of the wheel arches are low pressure would be a revelation to some.

[Lycoming]

CFD will not necessarily show any of the things you just mentioned if it isn't being done correctly. And with a black box program, you have no way of knowing if it was done correctly.

Back in the above sample, what if it told you the area of the circle was -6.5?

If you've literally no idea what's high pressure and low pressure, and you run CFD with incorrect mesh/boundary conditions/any of a billion things that you could input incorrectly... you still have no idea what's high pressure and low pressure. If you don't know what the program is doing, you're error bounds are practically infinite. to know the pressure coefficient at some point is -1 plus or minus infinite isn't helpful.

I don't think you realize just how badly CFD can go wrong.