ch3 Posted December 16, 2015 Share Posted December 16, 2015 Hi, has anyone here in the forums used solids and the finite element solver to do large scale, production level destruction? I am doing some tests to determine whether we will be using it in a series of shots, but so far my simulations have been rather slow and unstable. My current test is a moving sphere moving through a section of a wall that has 4-5 different layers to describe the different materials (plaster, wooden beams, cement). I understand that sandwiching multiple objects together and forcing another object through is a very demanding task for the solver, so I am wondering it it is possible to simulate something like that with a reasonable turn around times? All the objects together have around 100k tets in total, I've set it to 25 substeps and 10 collision passes, in an attempt to make the simulation stable. Some frames took more than 45 minutes and even at these calculation times the simulation became unstable and ended up exploding a few frames after the impact. Is there anything obvious I am missing out in terms of simulation efficiency/stability, or is it just a matter of keep increasing the substeps? As a straight comparison I have Kali, the DMM engine from pixelux that I was using in MPC, where we were able to simulate a couple of hundred frames overnight with more than a million tets. Do you think there is such a huge difference between the two solvers? thank you georgios Quote Link to comment Share on other sites More sharing options...
Guest tar Posted December 16, 2015 Share Posted December 16, 2015 It'd be worth posting this at the SideFx.com forums. The devs are usually very good at answering there for FEM questions. Also if you post your test file we may be able to help problem solve it. Quote Link to comment Share on other sites More sharing options...
Diego A Grimaldi Posted December 17, 2015 Share Posted December 17, 2015 Yea it'd be good to know if you come to any good results Georgios and share it with the community. Quote Link to comment Share on other sites More sharing options...
ch3 Posted December 17, 2015 Author Share Posted December 17, 2015 Ok, here is my test file. I will post it on the sideFX forum as well. Everything is under the 'diner' geo node and all yellow nodes are what goes into the purple DOP network. It takes about a minute to generate all the tet meshes. I could optimize the generation to just the small section I am simulating, but that's irrelevant to the performance of the simulation itself. Let me know if you think I could optimize the simulation in any way. thanks a lot Georgios FEM_scene.hip Quote Link to comment Share on other sites More sharing options...
Guest tar Posted December 17, 2015 Share Posted December 17, 2015 thanks! Initial testing shows your 'concrete' and 'wallThinBoard' blowing up when tested individually with a fresh finite element solver. That's the first place to optimise Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.