Hooke's law

[michaelb-01]

I am trying to create a basic particle system with spring-mass constraints based on Hooke's law with the aim of creating some granular simulations. I am still learning Houdini so i'm just trying to assess the options to achieve this..

 

Could I just use a popWrangle to update forces and velocities inside a popnet? Or would it be better to use gas microsolvers?

 

The ultimate goal is to achieve this!:

 

 

baby steps first though..

[Mandrake0]

seriously at the moment i have not a clue how to do this. :-/

 

but maybe this information's can help: 

 

Forum Post Snow:

http://forums.odforce.net/topic/18902-inspiring-snow/

(PDF at the last Post)

 

Video Disney:

https://www.youtube.com/watch?v=4HpaB-kHMBo

[djwarder]

I've always thought a POPVOP would be best, and then calc inter-particle forces.  Also, could use point clouds to speed up nearest neighbour detection.  However, any specifics I've yet to have the time to look into!!

[ikarus]

The disney snow simulation used a FLIP-based system of storing velocity and strain on the particles and calculating the forces and such on grids.  That gives you easy access to gradient and divergence calculations involved in applying stress+strain based forces.

[michaelb-01]

The Disney example is great but I want stay away from the PIC and FLIP methods and keep it with a particle-particle (lagrangian) method...

 

The most basic spring-mass system is described here: http://paulbourke.net/miscellaneous/particle/

 

But how would I go about implementing this in houdini? Pop wrangle?

 

Cheers

[kleer001]

I'll be you dollars to donuts the Disney way is faster than the springs. Then again the Disney process for snow can take 5-15 minutes a frame.

 

And yes, coding it in a wrangle will be the fastest way to implement either.

[ikarus]

You could certainly build a spring-mass system in geometry dop / pop wrangle, but I couldn't see you getting very high point counts in this manner as spring systems need a good deal of substeps to be stable,  which when combined with pc-lookups on dense point data, will become quite slow.

Guess it depends on what resolution you're looking to achieve.

 

The other thing is I don't think you will get the effect you want with spring constraints alone, if you're going for the snow clump effect.

[michaelb-01]

Thanks for the replies.

 

The Disney way should definitely be faster since its approximating velocities on a grid, but for that same reason it isnt as good for granular simulations since you are smoothing velocities.. Unless you have a very high resolution grid I guess. 

 

Yea efficiency with a lagrangian method is definitely a concern.. which does suggest a PIC/FLIP or MPM method would be best.. I'll have to do some tests and find the best compromise for speed and aesthetics! 

 

I'm mostly looking into this for granular (dry sand) effects so I wouldnt want the clumping. 

 

Another approach (!) is what Pavel did here (http://forums.odforce.net/topic/20333-sand-blowing-over-sand-dunes/page-2), using low res rbd objects with friction to drive the velocities of a high res particle simulation. Lots of methods I guess as always, now down to some testing!

 

Thanks for the advice. If anyone does any tests on this or sees something relevant i'd love to see it!