Basically I was hoping someone here would know where to find or how to determine the drag of a 3D mesh. I am trying to write a program that would accept a 3d mesh, and tell you the drag, ect. I am almoast possitive that this is a very complex topic, but it cant hurt to ask right? And I need to knwo how to get fairly acurate numbers or the project is pretty much useless for me. Thanx alot ahead Zach Dwiel ~ Tazzel3d

It is a complex subject. The aerodynamic coefficients due to the shape of the model are generally obtained experimentally, though basic shapes are known

Drag = (1/2)*(air density)*(speed^2)*(Section)*(aerodynamic coefficient).


Reference (from OR''A "Physics for game developpers") :

F. Sighard Hoerner "Fluid Dynamic Drag".
Bakersfield, Ca : Hoerner Fluid Dynamics, 1992

Hey,
Thanx for the info. So you don''t think I could use an algo to generate the drag of a 3d mesh algorithmically? I was kinda afraid of that. Are there ways to predict the drag by combining many known objects and their know drag?

Thanx again ~ Tazzel3d

You could have an algorithm that calculates a faces angle relative to a plane in front of the object, the more parallel, the more it adds to a resistance factor. then do this for each polygon. ( ignore back facing polygons).

Presumably a resistance would be calculated before run-time.

modification: this could allow for material resistances, but there are limitations. Fruny is right, the only way to truly know accurate values for air resistance is by sticking it in a wind chamber.

Edited by - Xgkkp on February 8, 2002 7:19:52 PM

quote:

---

Original post by Tazzel3D
Thanx for the info. So you don''t think I could use an algo to generate the drag of a 3d mesh algorithmically? I was kinda afraid of that. Are there ways to predict the drag by combining many known objects and their know drag?

---

Well, there is a way, but that involve actually building a full-fledgedfluid dynamics simulator, which is a non-trivial (read extremely hard) task.

You would only be able to sum the coefficients of distinct objects if you were able to guarantee that they are not going to interfer one with the other. Just think about how one part can mask a second part and deflect the flow onto a third one. Or how the actual direction of the flow matters (a cone point or base first)...

I am personally glad I am through with fluid mechanics

you can''t ignore the back-facing polygons... they determine how "smoothly" the air can return to it''s normal flow behind the object, which affects drag greatly (which is why they started making bicycle helmets pointy in the back, among other things)...
i agree with fruny, that you would need to use a fluid dynamics simulator to get a realistic number (and remember, it will change based on the direction the objecdt is travelling in)...


--- krez (krezisback@aol.com)

Thanx for the info again guys. I assumed that was going to be the case, but just thought I would ask.

Zach Dwiel ~ Tazzeld

Just to give you an impression of how much calculations are needed for that, I''m a physics student working on some simulations for coral growth and i need to calculate flow fields around those objects (if you know that you can also easily compute the drag), and run 64^3 grids on a parallel machine (54 800Mhz machines glued together) and typical runtimes on 4 nodes are 1 hour just to get one flow field , the real experiaments take over a week on 16 nodes.
So your best bet at calcualting drags is to play with the numbers a bit and see what "looks" good. Perhaps calculating the fractal dimension of your object might be a nice idea (find a good textbook on how to do this, basically it''s relating the surface of your object against it''s volume).

There are quick computational fluid dynamics methods that are not too difficult to write for computing drag. Unfortunately, the quick methods tend to be quite POOR at generating reasonable drag estimates. The simple methods such as parabolic drag polars using fixed cross section area as the reference area provide much better estimates of drag. Which is interesting. The simpler, faster methods produce results that are closer to reality than more detailed, CFD calculations!

Search the forum archives for "drag polar". There have been other posts, and some of the responses from myself and from others are quite detailed---enough that you could easily implement from the information posted.

Graham Rhodes
Senior Scientist
Applied Research Associates, Inc.

quote:

---

Original post by Anonymous Poster
Just to give you an impression of how much calculations are needed for that, I''m a physics student working on some simulations for coral growth and i need to calculate flow fields around those objects (if you know that you can also easily compute the drag), and run 64^3 grids on a parallel machine (54 800Mhz machines glued together) and typical runtimes on 4 nodes are 1 hour just to get one flow field , the real experiaments take over a week on 16 nodes.
So your best bet at calcualting drags is to play with the numbers a bit and see what "looks" good. Perhaps calculating the fractal dimension of your object might be a nice idea (find a good textbook on how to do this, basically it''s relating the surface of your object against it''s volume).

---

Well for quick calculations you could use a boundary element (panel) method, which runs MUCH faster than your method (probably an iterative method?) I mean panel methods can often compute one time step in a fraction of a second for reasonably complex surface meshes. But they still don''t give good drag predictions. The drag polar method was derived out of wind tunnel experiments and gives good results for miniscule computational cost---as long as your mesh is somewhat streamlined.

Regarding your performance....1 hour on 4 machines for one flow field (one time step?)...I am guessing that you are not using multigrid or conjugate gradient? Or even preconditioning to accelerate convergence? What governing equations are you using? Full Navier-Stokes? RANS? TLNS? Euler? Potential? There may be somethings you can do to speed the calcs up. Do you deal with variations in properties due to variable salt content or chemical reactions? Or temperature variations. This is a serious engineering calculation, certainly nothing appropriate for games at this point.

Graham Rhodes
Senior Scientist
Applied Research Associates, Inc.