Differential Erosion / Melting simulation

[sb526]

Hi all! Loads of cool stuff here.

I'm looking for ideas/suggestions from Houdini veterans on how to recreate a process of melting or a differential erosion of the material. I have attached a screen cap of the closest example I've found in a video by Christoph Bader/Dominik Kolb called Area Contraction. Also attached a macro shot & GIF of my physical experiment of creating similar structures via melting. Sadly, I haven't been able to find any papers or simulation breakdown examples of Area Contraction or how to replicate it in Houdini.

There are so many potential paths to take that I feel a bit lost - is it best to operate in volumes or VDBs? Or should I simply stick to particle solvers? FLIP liquid with attractors? FLIP liquid driven by temperature values (lava, snow, ice shelf tools)?

Edited November 8, 2017 by sb526

[ParticleSkull]

Hey JR, have you seen this file? I guess it can give you a hint on where to start:
http://fx-td.com/content/misc/recursive_growth_v2.hiplc

 

[Rival Consoles]

A possible approach would be to convert the mesh to volume fog vdb, apply any noise you like to it and convert it back to mesh.

[sb526]

On 09/11/2017 at 12:03 AM, ParticleSkull said:

Hey JR, have you seen this file? I guess it can give you a hint on where to start:
http://fx-td.com/content/misc/recursive_growth_v2.hiplc

 

Thanks! I'll take a look at this.

[sb526]

On 09/11/2017 at 2:58 PM, Marcola said:

A possible approach would be to convert the mesh to volume fog vdb, apply any noise you like to it and convert it back to mesh.

Yes, that's exactly what I've been using as a temporary solution. It creates a similar effect yet the "strands" of matter are hard to achieve using only noise. Maybe I should try Voronoi/Cellular noise inside the pop solver but I haven't been able to map it into 3D, 2D only...

[f1480187]

@sb526, did you try VDB Reshape SDF? I think the video used pretty close stuff on erode part. Convert to fog, add noise, convert to SDF, erode, convert to polygons.

sdf_erode.hipnc

[sb526]

12 hours ago, f1480187 said:

@sb526, did you try VDB Reshape SDF? I think the video used pretty close stuff on erode part. Convert to fog, add noise, convert to SDF, erode, convert to polygons.

sdf_erode.hipnc

Wow, thanks! That's a super elegant and computationally cheap solution. And it seems to give a good amount of control over the process through the volume wrangle remap and VDB reshape itself.

I have been using a Voronoi Fracture with VDB Morph which is slower to compute and isn't technically an erosion. But it does in a way produce a more consistent result.

erosion_test_v1.hip

Do you think it is possible to generate a less randomised Worley/Cellular noise to drive the density / VDB Reshape? Some sort of a middle ground between the complete regularity of a Voronoi diagram and irregularity of cellular noise?

 

 

Edited November 13, 2017 by sb526
word

[f1480187]

I like your scene, I would try to use it too, since it generates nice strands.

Not sure if I understand the question. Worley/Cellular/Voronoi noise are same thing. They make some points automatically, by randomizing 3D grid (hence maintaining some uniformity). So, they are based on Voronoi diagrams themselves. You could use your own points. Compute F1 (distance to first closest point) by measuring distance between @P and result of nearpoint(1, @P) function using your points as a second input. It will create "cellular noise" perfectly matching the Voronoi Fracture setup. You can create the points with any structure of your choice.