Mantaflow not calculating volume correctly for liquids [due to Adaptive Time Steps?] #77510

New Issue

Ignatz · 2020-06-06T14:36:27+02:00

Ignatz commented

2020-06-06 14:36:27 +02:00

System Information
Operating system: Windows 7 Pro 64-bit
Graphics card: Nvidia GeForce GTX 1060 6gb

Blender Version
Broken: version 2.83.0

Created Mantaflow liquid simulation.

Inflow object flowing into small tub (collision object) and from there into a larger tub below (collision object)

Small tub fills in something like 70 frames and then overflows into larger tub below which fills up in another 100 frames.

This cannot be. The larger tub is something like six to eight times the volume of the smaller tub and would not fill up this fast.

In my opinion, Mantaflow is calculating the flow so that there is an equal Z-axis fill rate, but this is totally incorrect with regard to volume.

mantaflow volume bug.blend

mantaflow volume bug.mp4

**System Information** Operating system: Windows 7 Pro 64-bit Graphics card: Nvidia GeForce GTX 1060 6gb **Blender Version** Broken: version 2.83.0 Created Mantaflow liquid simulation. Inflow object flowing into small tub (collision object) and from there into a larger tub below (collision object) Small tub fills in something like 70 frames and then overflows into larger tub below which fills up in another 100 frames. This cannot be. The larger tub is something like six to eight times the volume of the smaller tub and would not fill up this fast. In my opinion, Mantaflow is calculating the flow so that there is an equal Z-axis fill rate, but this is totally incorrect with regard to volume. [mantaflow volume bug.blend](https://archive.blender.org/developer/F8586860/mantaflow_volume_bug.blend) [mantaflow volume bug.mp4](https://archive.blender.org/developer/F8586861/mantaflow_volume_bug.mp4)

Ignatz commented

2020-06-06 14:36:27 +02:00

Added subscriber: @ignatz

Ignatz commented

2020-06-07 11:16:51 +02:00

As I continue to think about this problem it suddenly struck me that Mantaflow has evidently multiplied the number of liquid particles.

Witness the relatively small particle stream from the inflow object... relative to the huge amount of flow when the small box overflows.

As I continue to think about this problem it suddenly struck me that Mantaflow has evidently multiplied the number of liquid particles. Witness the relatively small particle stream from the inflow object... relative to the huge amount of flow when the small box overflows.

Philipp Oeser commented

2020-06-11 15:45:35 +02:00

Added subscribers: @sebbas, @JacquesLucke, @lichtwerk

Philipp Oeser commented

2020-06-11 15:45:35 +02:00

Changed status from 'Needs Triage' to: 'Confirmed'

Philipp Oeser commented

2020-06-11 15:45:35 +02:00

This is indeed unexpected.
I assume this has to do with Use Adaptive Time Steps (turning that off yields more predictable results)?

CC @sebbas
CC @JacquesLucke

I would assume this is a bug, will let others decide though...

This is indeed unexpected. I assume this has to do with Use `Adaptive Time Steps` (turning that off yields more predictable results)? CC @sebbas CC @JacquesLucke I would assume this is a bug, will let others decide though...

Philipp Oeser changed title from ~~Mantaflow not calculating volume correctly for liquids~~ to Mantaflow not calculating volume correctly for liquids [due to Adaptive Time Steps?]

2020-06-11 15:46:06 +02:00

Ignatz commented

2020-06-11 16:35:46 +02:00

Philipp, you are correct. Turning off 'Adaptive Time Steps' does yield a result that is more correct relative to the volumes of the containers versus the fluid flow.

It would appear that the 'Adaptive Time Steps' option is allowing an unexpected multiplication of the fluid particles.

The multiplication of fluid particles could be interesting in some situations, but definitely not when the user is trying to obtain results that more closely match the real world.

I am assuming that the 'adaptive time steps' has some other benefit for the calculation of the simulation, but I would make particle multiplication a separate option on its own.

Philipp, you are correct. Turning off **'Adaptive Time Steps'** does yield a result that is more correct relative to the volumes of the containers versus the fluid flow. It would appear that the 'Adaptive Time Steps' option is allowing an unexpected multiplication of the fluid particles. The multiplication of fluid particles could be interesting in some situations, but definitely not when the user is trying to obtain results that more closely match the real world. I am assuming that the 'adaptive time steps' has some other benefit for the calculation of the simulation, but I would make particle multiplication a separate option on its own.

Jacques Lucke commented

2020-06-11 16:53:54 +02:00

It seems to produce fairly different results in 2.90 compared to 2.83, can you confirm that? Also it does not leak in the same way in 2.90 for me.

Philipp Oeser commented

2020-06-11 17:07:50 +02:00

In #77510#951634, @JacquesLucke wrote:
It seems to produce fairly different results in 2.90 compared to 2.83, can you confirm that? Also it does not leak in the same way in 2.90 for me.

Yep, sorry was only testing 2.83, 2.90 has totally different results, but nonetheless the discrepancy remains.
In 2.90, I get ~260 frames to fill up the top box and ~360 frames to fill up the bottom box [which is unexpected]

> In #77510#951634, @JacquesLucke wrote: > It seems to produce fairly different results in 2.90 compared to 2.83, can you confirm that? Also it does not leak in the same way in 2.90 for me. Yep, sorry was only testing 2.83, 2.90 has totally different results, but nonetheless the discrepancy remains. In 2.90, I get ~260 frames to fill up the top box and ~360 frames to fill up the bottom box [which is unexpected]

Jacques Lucke commented

2020-06-17 12:52:51 +02:00

I cannot explain this difference either. @sebbas, do you know what is causing the simulations to be that different?

Jacques Lucke commented

2020-08-14 15:46:55 +02:00

I think I misunderstood the problem last time.

In #77510#951647, @lichtwerk wrote:
In 2.90, I get ~260 frames to fill up the top box and ~360 frames to fill up the bottom box [which is unexpected]

I can reproduce it. It seems quite wrong, but might be a hard to fix problem, not sure. I'll set this as a bug for now. Maybe @sebbas has to reclassify this as a known issue at some point.

I think I misunderstood the problem last time. > In #77510#951647, @lichtwerk wrote: > In 2.90, I get ~260 frames to fill up the top box and ~360 frames to fill up the bottom box [which is unexpected] I can reproduce it. It seems quite wrong, but might be a hard to fix problem, not sure. I'll set this as a bug for now. Maybe @sebbas has to reclassify this as a known issue at some point.

Dalai Felinto commented

2022-05-09 11:02:12 +02:00

Mantaflow (fluid simulator) has no active developer at the moment. Moving those bugs as known issues so at least other users can find their issues already reported.

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Mantaflow not calculating volume correctly for liquids [due to Adaptive Time Steps?] #77510