Deep Rising Fx: Potential fluid simulator

darkChief

darkChief

Member
Hey,

I would consider this vaporware for now because development isn't that far along yet. I noticed some people were interested in an intergrated Lighwave fluid solution mainly for things like water. So I gave it a shot, and decided to share early results with the community.

This is a simple test, everything is multithreaded, and it should easily scale to millions of particles in the future. This isn't a grid based solution, it is an open boundary solution.

This is a proprietary solution, based on recent research, so only the developers ability will be the limitation :D

 
toeknee

toeknee

Member
Hey darkChief, looks interesting. So if I understand you correctly this will be simular to Houdini's flip fluids?
 
Oldcode

Oldcode

Member
Looks interesting but also needs a lot of work. The particles are way too big to do any realistic looking simulation for a movie. Also, would there be a way to do different viscosity, stickiness, and other properties? It will be interesting to see.
 
darkChief

darkChief

Member
Hi,

Not FLIP (that is a grid based solution, that uses particles for surface tracking). This is pure SPH. I know SPH has had a bad reputation in VFX but that is changing. Recent research has caught up with methods like FLIP fluids recently, and the research has improved. Real Flow 2015 introduced the Dyverso solver which is a SPH solution, this replaced its old particle solution. Don’t let the particle size distract you, because the size of particles in SPH are infinite. Because of that SPH can be hard to configure, so this solution may only have two resolutions. The smaller the particles, the smaller the time steps, because the displacements will be that small. This isn’t a limitation, it just means you need a different work flow. One work flow is to scale the size of your scene, surface the particles and scale down once it is baked. The scene in the video is small, in Lightwave units it only 5x3x3 meters. And surface reconstruction for SPH is now mature, you can get thin sheets of fluids comparable to FLIP. There’s no domain to worry about like FLIP, fluids can go anywhere.

A dam break simulation is hard to pull off without viscosity ;) There’s already two types of viscosity in this solver, a standard one, and a proprietary one that allows bigger viscosity values. Also featured in the video is vorticity confinement, which is configurable.

SPH has many extensions like grid based methods, like viscosity etc.

If you wanted to flood a city, the particle size in the video is perfect :D
 
samurai_x

samurai_x

New member
Make it a reality for lw.


Kickstarter it
 
Oldcode

Oldcode

Member
darkChief said:
Hi,

Not FLIP (that is a grid based solution, that uses particles for surface tracking). This is pure SPH. I know SPH has had a bad reputation in VFX but that is changing. Recent research has caught up with methods like FLIP fluids recently, and the research has improved. Real Flow 2015 introduced the Dyverso solver which is a SPH solution, this replaced its old particle solution. Don’t let the particle size distract you, because the size of particles in SPH are infinite. Because of that SPH can be hard to configure, so this solution may only have two resolutions. The smaller the particles, the smaller the time steps, because the displacements will be that small. This isn’t a limitation, it just means you need a different work flow. One work flow is to scale the size of your scene, surface the particles and scale down once it is baked. The scene in the video is small, in Lightwave units it only 5x3x3 meters. And surface reconstruction for SPH is now mature, you can get thin sheets of fluids comparable to FLIP. There’s no domain to worry about like FLIP, fluids can go anywhere.

A dam break simulation is hard to pull off without viscosity ;) There’s already two types of viscosity in this solver, a standard one, and a proprietary one that allows bigger viscosity values. Also featured in the video is vorticity confinement, which is configurable.

SPH has many extensions like grid based methods, like viscosity etc.

If you wanted to flood a city, the particle size in the video is perfect :D
Click to expand...


Cool. I use Real Flow with Lightwave now. It works, but definitely has some quirks. One thing I've noticed is there is a fair amount of difference between SPH and both versions of Dyerso. Despite it taking much longer to simulate, I often find I'm still using the old SPH, partly because I'm familiar with how to do the setting to get the results I want.

Do you think you're SPH engine will be able to handle water on small to medium scale? For example, I'm working on a project now where I have a monster that rises up out of a swamp. I want a fairly viscous fluid (mud and slime) to stick to the creature as it rises with but with some dripping off.

Thanks!
 
darkChief

darkChief

Member
Thanks, for the encouragement. Will do my best to finish this.

- - - Updated - - -

Oldcode said:
Cool. I use Real Flow with Lightwave now. It works, but definitely has some quirks. One thing I've noticed is there is a fair amount of difference between SPH and both versions of Dyerso. Despite it taking much longer to simulate, I often find I'm still using the old SPH, partly because I'm familiar with how to do the setting to get the results I want.

Do you think you're SPH engine will be able to handle water on small to medium scale? For example, I'm working on a project now where I have a monster that rises up out of a swamp. I want a fairly viscous fluid (mud and slime) to stick to the creature as it rises with but with some dripping off.

Thanks!
Click to expand...

Theoretically and mathematically everything from a glass of wine to oceans should be possible. Since theory doesn’t always translate to implementation, I can't say exactly what the full potential of this solver is yet, it is too early to make any claims. That requires actual experimentation to see the limits and realities. So look out for practical demonstrations of simulation scenarios in the future ;)

Viscosity is mainly friction, there's also adhesion to consider. I'm also interested in implementing highly viscous fluids later (fluids that don't completely lose shape). That could be a completely different solver.
 
Last edited:
Oldcode

Oldcode

Member
Very cool. The only thing that catches my eye is the particles are all aligned before it hits the object. There's a function in Real Flow that allows a certain amount of random positioning of the particles, both vertical and horizontal, as they come out of the emitter.

Got anything like that in the pipeline?

Thanks! :D
 
You must log in or register to reply here.
Top Bottom