利用者:Sebbas/GSoC 2017/Reports/Week 8

提供: wiki
移動先: 案内検索

Week 8

In this eighth week I mainly worked on secondary particles again. Some bugs, such as the velocity extrapolation bug, were also on my list.

What's been done

  • What last week was just possible in Mantaflow itself is now possible inside Blender as well: Drop, bubble and float particles can be enabled from the "Fluid Particles" tab. (rB354b4969722f, rBd9ba9993d8b4, rB6ed2fe0229c1)
  • I improved drop and bubble particles in the sampling function a little more. Bubbles now take surrounding fluid velocity into consideration (and so don't just rise straight up). Drop particles are only generated if the angle between velocity and surface normal doesn't exceed a given threshold (the "Spray Foam Bubbles" paper from Ihmsen et al. [2] suggested this nice approach). (rB44f49e765b1b)
  • Setting outflow cells is now done with initDomain(). Previously by using setOpenBound(), we got a one cell margin between FLIP particles and actual outflow (and thus particles never got reset!). (rB766572cd0a7a)
  • Regarding the velocity extrapolation bug: I didn't fix it yet but made some progress in terms of what's causing it (check out Github [1] for more info). I also fixed a related issue in this context. The outermost outflow cells, which now get set with initDomain(), weren't updated when one manually placed an obstacle object near the border. A small change in updateFromLevelset() fixed this issue. (rB25170baeab5b)
  • Smaller bug fixes (including a Windows easter egg :) The define PFLOAT, which I used for the float particle type, had already been defined in a Windows header ...) (rBf22f84ea3220)

Next up

  • Revisit the particle sampling function: Expose more arguments to the UI (tracer amount field still needs some work) and optimize a bit (currently each particle type has its own loop - it should be possible to combine some.)
  • Take a closer look at Mantaflows fluid guiding setup. I'll start very simple and try to run one of the existing sample scripts in Blender.
  • The velocity extrapolation bug remains. Nils pointed out some parts in one of Mantaflows velocity extrapolation functions that might cause problems. What I'll now do is, since I couldn't fix the bug in Blender directly, setup a Mantaflow scene that also produces the issue.

Questions

  • No

References

  1. https://github.com/sebbas/BlenderMantaflow/issues/7
  2. https://cg.informatik.uni-freiburg.de/publications/2012_CGI_sprayFoamBubbles.pdf