Page 1 of 1

Paper that could be useful for fire and smoke simulation

Posted: Tue Mar 14, 2006 5:55 pm
by Don_Kisciotte
See this paper that could be useful for the simulation of the fire and the smoke:
http://graphics.ucsd.edu/%7Ehenrik/papers/fire/
http://www.cs.huji.ac.il/labs/cglab/research/tdsmoke/
cheers

Posted: Tue Mar 14, 2006 10:48 pm
by epat
That fire simulation paper looks cool - but as I'm just a newbie to blender's source code,
don't expect anything from me.

I might give it a go though :D

Posted: Wed Mar 15, 2006 3:29 pm
by Toon_Scheur
I didn't open the links but I assume you mean Prof. Henrik Jensen's papers about fire and smoke. Unless there is a a good mathematics freak with something to prove out there, we won't be seeing those these coming years in Blender. Fore those simulations we need a voxel system in Blender to begin with. Prof. Jensen's papers are an enigma on their own (for me anyhow). For example it took me a year to understand his fast SSS paper. I very much like too implement it in the Blender source for this is a elegant solution, but I realy don't have the time to study the Blender code. I guess think I have more time once my kid is properly toilet trained :lol:

I've been reading also papers about anisotropic remeshing, clothing your objects, hair and cloth self collisions, all very interesting....but.....

Posted: Wed Mar 15, 2006 9:32 pm
by LetterRip
Toon_Schuer,
Fore those simulations we need a voxel system in Blender to begin with.
the fluid dynamics is all voxel based, so voxel code already exists.

LetterRip

Posted: Wed Mar 15, 2006 10:22 pm
by Toon_Scheur
Indeed it was specified that the fluid sim does start out with voxels, but is it realy voxels? Because it is all converted to derived meshed is it not?
Are those voxels displayable? Is there a marching cube algorithm in place for the voxels? Could they be filtered? Are there visibilty and compression algorithms for those voxels? Or are those 'voxels' merely of a grid placeholder for the domain to execute the fluid sim on?

Of course I could answer these questions myself if I delve in the code, or maybe if you know the voxel part of the code you could answer it?

Posted: Fri Mar 17, 2006 5:07 pm
by poutsa

Posted: Fri Mar 17, 2006 5:38 pm
by Toon_Scheur

Posted: Fri Mar 17, 2006 8:40 pm
by epat
I read the 'Fast SSS' paper also but I don't think that this would be very
easy to implement into blender without a recode of some of the more
basic features of blenders api. :(

Posted: Fri Mar 17, 2006 9:00 pm
by Toon_Scheur
In any case, there are a lot of material out there to implement for SOC'ers :D (if there would be any this year again :roll: )

Posted: Sat Mar 18, 2006 12:17 am
by LetterRip
Hi,

I'm not familar with the code itself but I do have some idea of the implemenation,
Indeed it was specified that the fluid sim does start out with voxels, but is it realy voxels? Because it is all converted to derived meshed is it not?
Are those voxels displayable? Is there a marching cube algorithm in place for the voxels?
They are converted to a mesh after all of the calculations are done. Marching cubes is the method he uses to convert from voxels to the mesh representation.
Could they be filtered? Are there visibilty and compression algorithms for those voxels? Or are those 'voxels' merely of a grid placeholder for the domain to execute the fluid sim on?
I'm not sure, you could email Nils and ask him. (email me LetterRip at gmail dot com for his email address....)

LetterRip

Posted: Sat Mar 18, 2006 6:35 am
by Surt
I've looked a bit into Jos Stam's fluid method, and it looks quite promising (used as the basis of MAYA Fluid Effects), but again the math is somewhat beyond me.

Stam's papers are here: http://www.dgp.toronto.edu/~stam/realit ... h/pub.html

Posted: Sat Mar 18, 2006 5:08 pm
by n_t
Hi LetterRip&All,

the existing code is voxel based but creates a triangle mesh for displaying the water surface. on itself, a smoke simulator is simpler than a free surface one, but I think the main problem in Blender currently is that there's no way to display volume data such as smoke... It would be cool to e.g. have the volume rendering capabilites of PBRT in Blender, but I probably wont have time for that in a while.

Regards,
-> Nils

Posted: Sat Mar 18, 2006 9:24 pm
by Xtra
Well, a fire simulation could be another kick ass feature for Blender, of course. Who's the guy who developed the fluid dynamics? Maybe he's able to code somthing like that?

Posted: Sun Mar 19, 2006 1:27 am
by Rhs_CG
The guy who coded the fluids engine is Nils, the post before.

Smoke and fire would be sweet, I agree. BTW, Nils, a million kudos and thank you's - the simulator is sweet!

Posted: Tue Apr 04, 2006 8:39 pm
by epat
has anyone seen bfast: http://sourceforge.net/projects/bfast
I don't know whether its any good but its open source so might be useful.