Realistic ice melting effect

[VAco]

Hi,

I have a project where I want to create a realistic ice melting effect.
I am quite new to FLIP fluids, however I read a lot and watched some masterclasses about how the solver works in Houdini. But I still struggle at a lot of areas.
 
Reference for the effect:

 

What I have done currently is played around with the viscosity. Basically I took my geometry and created points out of it, then I an SDF volume from the same geometry and passed attributes the values onto points. So, the inner most points would have the lowest value and going to the surface it would reach 0. I fitted those values from a range like 1,000,000 to 0 and then using a geometry vop in the dop context i just subtract some viscosity from the points at each time step. 

 

Here are my low resolution tests:

Particles:

Meshed:

https://www.youtube.com/watch?v=IokJPFoNi5w&feature=youtu.be
I got a somewhat decent result, however it looks more like ice cream melting than ice. It feels like the melting is not solid enough. The particles are not holding the shape as ice would.
 
That's how far I got for now. I have a few more ideas, though (haven't tried it yet):
1) Instead of using viscosity I was thinking to start with all the points disabled (not solving) and then gradually activate them. 
2) I wanted to try out using pressure, like in this example (they are using realflow):

 

http://https://vimeo.com/13599797
but have no idea how to manipulate the pressure field: 
 
I was hoping maybe somebody with more experience would put me on a better track and maybe have more ideas or suggestions on what would be the best way to achieve this result.
Also, maybe you have some suggestions on the shading part? I have some ideas, but I always tend to take the most complicated approaches
 
I am using H15.
 
Thank you in advance

 

 

[Farmfield]

Mmm, I get your idea here but personally I'd go for a viscosity attribute transfer from scattered points on the surface of the frozen part, basically using a solver node, separating the "frozen" and liquid parts, scatter points on the frozen surface, then do the attribute transfer. For added realism, I'd set up some viscosity fluctuations both static for the frozen parts as well as use some animated noise values on top of the viscosity values on the "melting-points" - no pun intended, hehe...

 

Simple solution but it should work. 

Edited February 27, 2016 by Farmfield

[6ril]

Hey,

I don't have the level to point you to the right direction technically in houdini, but I feel like your test is showing something about the time scale importance of melting ice.

Like scene scale is impacting the perception of a simulation (i.e: breaking object) time speed is giving your sim a fluid motion, as well as particles interactions, that doesn't suit ice melting.

 

maybe a low time speed and taking only a frame every n frames would help a lot the effect, and reduce the chaos of the puddle under.

 

Also scene scale seems way too big in your test, am I right?

 

the cream feeling would disappear if you manage to maintain shape for a while, somehow.

 

 

anyway, I'll follow this thread, interesting.

 

Shading.. would a glass material and somehow a growing fog volume work? But the main aspect would be the cracking propagation ! Can the breaking/voronoi tools do that easily? Combine it with the particle meshing.. I would know how to do that.

 

good luck!

 

*edit*  I'm so helpful with my somehows

Edited February 27, 2016 by 6ril

[VAco]

Mmm, I get your idea here but personally I'd go for a viscosity attribute transfer from scattered points on the surface of the frozen part, basically using a solver node, separating the "frozen" and liquid parts, scatter points on the frozen surface, then do the attribute transfer. For added realism, I'd set up some viscosity fluctuations both static for the frozen parts as well as use some animated noise values on top of the viscosity values on the "melting-points" - no pun intended, hehe...

 

Simple solution but it should work. 

 

Thanks for replying

I shall try your approach.

The biggest issue I am having is maintaining the frozen shape as ice would. I will try going higher res as well, which might help.

 

Hey,

I don't have the level to point you to the right direction technically in houdini, but I feel like your test is showing something about the time scale importance of melting ice.

Like scene scale is impacting the perception of a simulation (i.e: breaking object) time speed is giving your sim a fluid motion, as well as particles interactions, that doesn't suit ice melting.

 

maybe a low time speed and taking only a frame every n frames would help a lot the effect, and reduce the chaos of the puddle under.

 

Also scene scale seems way too big in your test, am I right?

 

the cream feeling would disappear if you manage to maintain shape for a while, somehow.

 

 

anyway, I'll follow this thread, interesting.

 

Shading.. would a glass material and somehow a growing fog volume work? But the main aspect would be the cracking propagation ! Can the breaking/voronoi tools do that easily? Combine it with the particle meshing.. I would know how to do that.

 

good luck!

 

*edit*  I'm so helpful with my somehows

 

Thanks.

 

Yeah my scale was wrong. Fixed it now. Hopefully it will help with the look of the melting.

I was planning on using a glass shader with some fog or some geometry that would scale down based on some attribute. Don't know about the cracking, I would just probably add a texture or some noise to replicate the scratches. Need to get the initial melting stuff looking good first

 

Hopefully it won't take too long.

Edited February 27, 2016 by VAco

[Farmfield]

I shall try your approach.

The biggest issue I am having is maintaining the frozen shape as ice would. I will try going higher res as well, which might help.

 

Yeah, my approach would keep the shape better, I believe, and you'll have great control using an attribute transfer from the surface in.

[Tridae]

You could use something like this 

 

It's a flip corrosion setup that eats away at a vdb. I haven't had time to look at that .hip yet however I'd imagine that you'd be able to instance that solid object as an animated object that could even have feedback dynamics on the flip sim.

 

The simplest approach would be to take that kind of setup and just gradually spawn flip particles on the surface. As they eat away at the surface you could spawn more flip particles in the cracks, thus adding to the melting effect. 

 

Like I said, haven't had time to check the .hip posted there myself yet so I might be totally off. 

 

The approach seems more stable and solid than just messing with viscosity. 

[VAco]

You could use something like this 

 

It's a flip corrosion setup that eats away at a vdb. I haven't had time to look at that .hip yet however I'd imagine that you'd be able to instance that solid object as an animated object that could even have feedback dynamics on the flip sim.

 

The simplest approach would be to take that kind of setup and just gradually spawn flip particles on the surface. As they eat away at the surface you could spawn more flip particles in the cracks, thus adding to the melting effect. 

 

Like I said, haven't had time to check the .hip posted there myself yet so I might be totally off. 

 

The approach seems more stable and solid than just messing with viscosity. 

 

Thank you for your suggestion.

 

I have tried to implement it, I was somewhat successful. I made the object collide with particles and I spawned more particles over time from the surface. Then the particles would eat way the geometry, in a controlled way.

 

Thing I noticed if I use a more complicated geometry some part start to float mid air since it is a static vdb volume (kinda fixed the issue with a vdb smooth operation).

And it is quite sensitive, if I use too many particles it will decay way too quickly and it will be very turbulent or if I use not enough particles the geometry will take for ever to melt and then the flip particles would not preserve their volume on collision to the ground, because there are not enough of them (maybe higher res is needed ?).

 

Nevertheless, a very interesting approach. Probably need more tests and playing around. I am quite certain it could work.

 

Currently I am leaning a bit more towards the viscosity approach. But shall continue doing tests and see what happens.

 

P.S. I don't have the flipbooks with me, otherwise I would post the tests.

Edited March 2, 2016 by VAco

[zhuyuxin]

Hello, I am also doing the effect of ice melting. Did you succeed in the end? My idea is the same as what you just started, but the ice melts too much, or there is no detail. If you have succeeded, can you give the document a look? Thank you. I look forward to your reply.