continuum mechanical model for cloth simulation

Please don't post Bullet support questions here, use the above forums instead.
Post Reply
guernika
Posts: 18
Joined: Fri May 07, 2010 2:20 pm

continuum mechanical model for cloth simulation

Post by guernika »

Dear all,

I have just developed a cloth simulation engine. The mechanical model is not particle based but it use continuum mechanics.
It is quite stable (for the same physics based scene I can use the same integration timestep of the default bullet solver), and also fast. I have to extend the mechanical model in order to account for friction and bending force.
IMHO simulation looks quite better than the particle based approach.
I posted on youtube two comparison videos of the two approaches, using bullet as reference for particle based simulation.
You can find the links in the following lines.

continuum mechanics with static cylinder:
http://www.youtube.com/watch?v=s1iJkAJh ... e=youtu.be

particle based with static cylinder:
http://www.youtube.com/watch?v=3V_4gvfmgZA

continuum mechanics with moving cylinder:
http://www.youtube.com/watch?v=dEKL1ugXIzk

particle based with moving cylinder:
http://www.youtube.com/watch?v=f0W_qj7Pw5o


Any feedback will be appreciated.

Regards,

Francesco
mobeen
Posts: 122
Joined: Thu May 05, 2011 11:47 am

Re: continuum mechanical model for cloth simulation

Post by mobeen »

A couple of questions to ask you?

1) What were the resolutions of these meshes and what were the timing results for these videos.
2) How different is your model from the standard continuum model where the mesh is represented in the form of tetrahedra/hexahedra and then FEM is solved on these elements?
3) Does your model consider self collision also and how does your model perform during self collision? The real test is the ribbon free fall from a set of feeders when the cloth settles on top of itself in the form of folds.
guernika
Posts: 18
Joined: Fri May 07, 2010 2:20 pm

Re: continuum mechanical model for cloth simulation

Post by guernika »

Hi Mobeen,

1) The cloth it is a triangle mesh of 1024(32*32) vertices, and the cylinder is a triangle mesh made up by 760 vertex. I used maya to create the scene.
The timing results are displayed on the video: elapsed time and simulation time. Elapsed time is the time to compute the simulation, while simulation time is the time of the numerical simulation.

2) The solver handles only triangle meshes. I compute the stress of each triangle from the green deformation tensor. The approach is quite similar to the works of Volino. Further, as I previously said, my implementation still does not account for bending forces.

3) In my previous post I forgot to say that I used bullet for collision detection. At the moment there is no check for self collisions. My investigation is not about collisions, but in cloth mechanical model. I will use bullet for collision detection also in the future implementations accounting for self collisions.

Regards,

Francesco


quote="mobeen"]A couple of questions to ask you?

1) What were the resolutions of these meshes and what were the timing results for these videos.
2) How different is your model from the standard continuum model where the mesh is represented in the form of tetrahedra/hexahedra and then FEM is solved on these elements?
3) Does your model consider self collision also and how does your model perform during self collision? The real test is the ribbon free fall from a set of feeders when the cloth settles on top of itself in the form of folds.[/quote]
mobeen
Posts: 122
Joined: Thu May 05, 2011 11:47 am

Re: continuum mechanical model for cloth simulation

Post by mobeen »

OK thanks for the clarifications Francesco. Its actually a different story altogether when u introduce collision and in particular self-collision into the performance stats. And I suspect when u add that in, particle based approach will certainly show better performance but still its an interesting comparative study.
Post Reply