Dyverso viscous and rigid bodies interaction issues

[Orlando Esponda]

Hello,

First post and actually very new to Realflow (second day using the demo) so please bear with me if this has been asked before. I actually found another post about it but at the end the OP decided change his Dyverso sim for another technique, so it doesn't help with my issue.

So, I'm trying to simulate a Dyverso viscous fluid spinning and a couple of objects must fall into the liquid... They are supposed to float on the liquid but I can't get them to do it so. But that's not the main issue: whenever an object touches the fluid particles, it starts spinning, and keeps accelerating. At some point these spinning objects are thrown away from the fluid.

I'm not sure if I'm doing something wrong (because of my lack of knowledge) or if these kind of interactions are not supposed to be done this way (Dyverso + rigid bodies).

It's a very simple setup: One object that acts as container for the fluid, one emitter, 6 small objects that fall on the fluid, and some basic daemons (gravity, drag, vortex, noise). https://drive.google.com/file/d/1JNb02q ... sp=sharing

I'm trying with Dyverso to take advantage of the GPU processing speed, so I really hope this issue can be solved without the need to switch to Hybrido or the legacy SPH particles. In other words, I really hope it's user error.


Here's a video showing what I'm trying to describe: https://drive.google.com/file/d/1XinDw1 ... sp=sharing


Lots of thanks for any help,
Orlando.

[Thomas Schlick]

Hello Orlando,

No, you don't do anything wrong, but what you're reporting here is not unusual with Dyverso and rigid bodies. Unfortunately this issue is unsolved and I'm not sure what causes it. What I know is that it mostly occurs with small objects with small mass values.

So, one method is to increase the objects' masses, but then you might see unnatural splashes and fluid-object interactions. My first idea is to go to the "Simulation Options" menu (right-click on the small triangle next to the "Simulate" button and choose "Options..."), open the "Caronte" tab and set "Quality" to something like 75+. Under the "Dyverso" tab look for min/max substeps and start with to something like 2/4 or or 4/8. Increase these values if necessary, but don't make them too big, because they affect the fluid as well.

Another idea is to let Dyverso "Rigid" particles drive your objects. Here's a short explanation how to use this solver: http://support.nextlimit.com/display/rf ... imulations

Finally, it's possible to increase scene scale, but I don't think this will really help here, because you need a really big scale to compensate for the small masses and this means that you'll also need way more fluid particles.

[Orlando Esponda]

Thank you Thomas,

I tried increasing the substeps as well as the Caronte quality, but it didn't help. So I'm trying with Dyverso rigids, but I can't find a way to make it work. In a simple scene (following the example in your link) it works, the objects are affected by gravity and they bounce off the ground plane. In my scene I can't make them move at all. They stay in the place where the particles were generated. Gravity is not affecting them at all. Do you have any idea on what could cause this?

If the fluid touches this rigids, they move as expected. Is just the forces that are not affecting them at all.

Again, simple scene (see attachment)

[Orlando Esponda]

So, after some more experimentation I found out why Dyverso rigid bodies were not moving. Turns out it has to do with a combination of fps and substeps. Not only Dyverso Rigid substeps, but also (in my case) Viscous substeps. I'm trying to simulate a slow motion fluid, so the scene is set to 240 fps. With this framerate it is possible to simulate the viscous fluid with higher settings than the default ones (let's say 8 min/max)m but if I add some Dyverso rigid bodies to the scene they are static. At some point I set the min/max substeps to 6 or so, and the objects were not influenced but the forces, but they were interacting with the fluid (if the fluid "touches" the object, it starts moving). The problem were the daemons. So I reset the substeps for rigids and the problem was still there. I then reset the substeps for the viscous fluids and voila!, the rigid bodies were affected by the daemons as expected.

I think this is a bug as the behavior makes no sense at all, but please let me know if I'm doing something wrong or if it's a known issue.

I recorded a video with a simple scene showing what I described before: high fps and higher than default rigid substeps will cause the rigid bodies not to move at all, or move partially:

https://drive.google.com/file/d/1S58NZu ... sp=sharing


Again, thanks for any help
Orlando.

[Thomas Schlick]

Thanks for the additional information. There are a few things that really puzzle me and I'd like to dig a little deeper here. Is it it possible to get the project file you've been using in the video where you experiment with fps and substeps? It's the easiest and fastest way, and you can PM me a link to the scene. I only need the FLW file and the objects. Any simulation data is not required.

Thanks,
Thomas

[Orlando Esponda]

Thanks for the additional information. There are a few things that really puzzle me and I'd like to dig a little deeper here. Is it it possible to get the project file you've been using in the video where you experiment with fps and substeps? It's the easiest and fastest way, and you can PM me a link to the scene. I only need the FLW file and the objects. Any simulation data is not required.

Thanks,
Thomas

Sent, please check your inbox.

I'm using the demo version 10.1.2.0162 under Linux (Fedora 27). Maybe irrelevant, but maybe this particular build is not working as expected... who knows, I just thought it's worth mentioning it.

Orlando.

[Thomas Schlick]

Thanks for the file. I'll take a look and let you know what I've found out.

By the way, there's no difference between the demo and the full version.

[Thomas Schlick]

I've tested your scene and I was able to make the static objects move in conjunction with a DyViscous fluid. In my tests with 240 fps and 2/2 substeps I've used the following settings:

DY_Domain01 (rigid) > Resolution > 75
Particle Skinner > Search Distance > 0.25 | Max Control Particles = 2 | Outside Control Particles = Yes

I didn't test all different parameter combinations, but with the settings above it worked. Maybe you find other values. As said, this made the objects move and collide with the viscous fluid, but the interaction turned out poor and the objects didn't sink in the fluid - even with rigid domain's Density set to 20000 they remained on top of the fluid's surface. With 24 fps and the above settings the objects dip into the fluid though, and with density set to 50000 they even go to the fluid's bottom. But, with a rigid density of 250.000 I've made the objects dive in the fluid at 240 fps.

So, it's not a so much a problem of substeps, but Particle Skinner settings.

A few words about substeps:
All materials are controlled by the Dyverso framework, and if you simulate multiple materials Dyverso will take the highest substeps values for the simulation of all materials. Let's say Viscous = 4/4, and Rigid = 2/2 then the entire simulation will be performed with 4/4 substeps. You also have to be careful with substeps in Dyverso, because they also influence a fluid's behaviour, as it becomes more rigid. If you need more substeps you should decrease iterations.

...set to 240 fps. With this framerate it is possible to simulate the viscous fluid with higher settings than the default ones

This is something I cannot confirm. I've simulated viscous fluids with 50 and 75 substeps at various fps rates and it always worked.

Then I've seen that your scene was pretty slow and I've also found out why. The "Bacio" objects' Display > Isosurface option was turned on. Once I've disabled this option everything was much faster.

[Orlando Esponda]

Thanks Thomas,

I will keep testing with different settings and will report back if I find something.

[duke3d]

Hi
recently I was working on the same simulation. I fount that any PBD liquid (viscous as well) works very good with Particle Skinner. I used Dyverso Rigid for the objects that fall into liquid (In my case it was ice cube falling into water and nuts falling into chocolate). The result was very good

I've also tried SPH liquid and granular, didn't work with Particle Skinner.