Errors with examples

[joel-simon]

Hello again, unfortunately more Ubuntu issues. Both examples are crashing for me.

Taken from the .bat files I have run both verbatim

FractureRB bowl.bullet bowl.csv -o out/bowl -n 500 -i 1e3 -f 5e6 --default-solver

as well as

FractureRB glass.bullet glass.csv -o out/glass -i 1e5 -f 1e7 -s 2

bowl one raises

... % have 1 fragments ... % \ 9 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % 64: I 1931 F 4.76e+06 (bunny) % 64: I 1931 F 9.717e+05 (Arma1) % ** % fracture simulation for bunny_64 ... 37 time steps % ... initializing BEM solver % ... regularizing Neumann problem % BEM initialized ... 24.0385s % setting new boundary data ... 0.3233s % fracturing (1/37) ... % computing SIFs ... 0.0045s % writing output ... 0.0158s % crack initiation ...terminate called after throwing an instance of 'openvdb::v2_1::IoError' what(): IoError: could not open file vdbmat\grains_cu_oct_nb10.vdb Aborted (core dumped)

This may be a \/ issue? should it be vdbmat/grains?

and the glass one raises

... % sub-stepping ... 0.1434s % total ... 10.1538s (have 1998 elements and 83 cracks) % fracturing (2/10) ... % computing SIFs ... 14.0205s % crack initiation ... % ... ... candidates listed % ... ... seeding done 1.1803s % sub-stepping ...Segmentation fault (core dumped)

These both crash at the same place if re-run.

I am running the version with the bug fixes you pushed this morning. I did not recompile openVDB with the fix but I I saw it causes crashes with stack limits and mine is 1,000,000 so that should be fine.

Thanks again for your help! I look forward to being able to use this. Joel

[david-hahn]

Hi!

For the first one it’s probably because the path to the VDB file is also written in Windows format in the CSV file. Try changing that to “vdb(vdbmat/grainscu...” with “/” (of course you might as well copy the material files into the same folder as the script/binary).

For the second one, I’m not sure at the moment, but I’ve seen some issues with certain versions of OpenVDB and GCC, in particular GCC 4.9. Maybe you can try it on GCC 4.7, which is what I’ve tested with (although by far not as much as I’ve tested on Windows, so I can’t be sure).

I hope this helps.

[david-hahn]

I've built the whole thing (without the lastest bugfix today, so the output is a bit less, but you'll still get the important parts) on Debian x64 g++-4.7.2 (that's the closest to an Ubuntu system I have atm.) and uploaded them here: http://pub.ist.ac.at/~dhahn/FractureRB_Debian_x64.zip Here are the libs (double-precision Bullet) I've linked against: http://pub.ist.ac.at/~dhahn/Libs_Debian_x64.zip

These seem to run fine on my system.

[joel-simon]

Hello David,

thank you for your suggestions and link. For whatever reason I still had issues when compiling with 4.7. Your build worked for me which I really appreciate and likely saved me a lot of time.

I have a few questions:

• Is a blender or maya plugin the easiest way to get .bullet from .obj files? are there stand alone scripts for this?
• Is there any easy way to set a starting height for objects? I am interested in dropping rather than collisions. Would a starting +Y velocity be a good way to currently achieve that?
• Which parameters would you recommend for large scale marble?
• Is there a way to produce a high res of all major parts but only at the final time step?

Thanks again for your time Joel

[david-hahn]

Hi!

I’m working with Maya 2015, so that’s what I’m referring to in the following. I think the relevant parts haven’t changed too much recently, and Blender should have similar capabilities but I don’t have much experience with it. Our simulation outputs a Maya MEL script that loads the fragments all motion-keys for all objects into the scene. I’ve once started to code up Blender compatible output but it is not up-to-date.

My typical workflow is as follows:

• create the scene in Maya
• create rigid bodies using Maya’s Bullet plugin (activate via the menu: Window > Settings/Preferences > Plug-in Manager > bullet.mll, once loaded you should see “Bullet” on the menu bar)
• set the collider shape type to mesh (unless the object is a box, sphere or a convex shape in which case you can use box, sphere or hull) – I know Maya says meshes should only be used for static objects (in particular mesh-mesh collisions won’t work in Maya, but we’ll handle them in our sim), this will make sure the whole mesh is exported to the bullet file. Btw. you probably want to make sure that “Collider Shape Offset” and “Center Of Mass” are all zeroed out, otherwise crazy things can happen.
• export the scene in Bullet format (menu: File > Export All > set type to “bullet”) this includes all the geometry and initial positions (but I couldn’t get the initial velocities to work which is why you have to specify them in the CSV parameter file)
• write up the simulation parameters in the CSV file (in particular the bullet export will generate objects named “collisionObjectX” and you’ll need to match that to the obect’s name in the Maya scene – I know this is annoying but I couldn’t find a better way to do this so far – the numbering is usually in the order you created the rigid bodies in, but sometimes it can get messed up when you edit the scene)
• run the simulation
• remove Bullet stuff from the scene (menu: Bullet > Delete Entire Bullet System) – don’t overwrite the scene file, save it under a new name!
• load the simulation output into the scene (I usually use a little Python script that loads the MEL script one line at a time, somehow that dramatically reduces Maya’s memory usage during import)
• at this point you’ll have the simulation result in Maya in a sort-of “preview” resolution (i.e. all the meshes are not full-res as in the VDB grids
• once you’re happy with the scene, you can use the FractureRB_vis tool to produce the high-res meshes, it will also write out a MEL script you can use to load the new meshes and parent them to the objects in the preview scene. Sometimes the numbering of fragments gets messed up and I never quite found out why it happens, so some hand-tuning might be required to get a renderable high-res scene.

Please understand that our implementation is “experimental code” and not very user-friendly at all. If you find ways to improve it please let me know.

About the drop-testing: you could either

• set up different scenes for different drop heights
• set a fixed drop height and compute equivalent initial velocities (there are no fancy aerodynamics in Bullet so this is a trivial calculation)
• use the FractureBEM tool (that does not use Bullet at all) and set the corresponding boundary conditions manually (all the examples from the old repo should still work (more or less) with the new version, but the new one also supports pure Neumann problems and the fast approximation method.

As for material parameters, you’d have to try and find some literature on the types of materials you need, here are some things to keep in mind though:

http://www.twi-global.com/technical-knowledge/faqs/structural-integrity-faqs/faq-where-can-i-find-a-database-of-fracture-toughness-values/

Finally, here’s how I load the MEL scripts and execute line-by-line in Maya (just in case Maya crashes when you try to load the results):

---

import os

import maya.mel

from pymel.core import sceneName

os.chdir(os.path.abspath(os.path.join(sceneName(), os.pardir)))

f = open("_out/myscene_sim.mel")

for line in f:

try:

maya.mel.eval(line)

except:

print "error: ", line

f.close()

---

Cheers,

David

[joel-simon]

Hey!

Its been a while since I had time to work on this, but I got around to it again yesterday and everything worked well :) Only trick was maya not allowing dynamic mesh bullet objects but once you create it you can then change it. One issue was sometimes the RB_vis scripts messes up numbering like you said, but its only happened once. Thanks again for your help.

If you're interested, I have been using it for a sculpture project. http://www.joelsimon.net/sculpture-drop.html. I hope to try 3d printing things soon.

Cheers, Joel

[david-hahn]

Looks cool!

I’m glad it (mostly) worked.

Cheers!