Page MenuHome

Mantaflow - liquid refuses to rise above level of inflow object
Confirmed, NormalPublicKNOWN ISSUE

Description

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

Blender Version
Broken: blender-2.90.0-fd8d245e6a80-windows64
Worked: 2.83

Here is what appears to be a bug in the Mantaflow liquid calculations in Blender 2.90

In the first Blender file - manta_vol_bug.blend - the inflow object is within an effector-collision object. When the simulation runs the level of the fluid refuses to rise above the Z-axis of the inflow object. This behaviour does not happen with this file in Blender 2.83

It must also be noted that there is a severe difference of the liquid inflow rate using the same simulation numbers between version 2.83 and version 2.90. That may be a problem on its own, but I cannot say if this has any influence on the Z-axis volume bug that I am seeing.

I have uploaded two video files of the rendered results of this file.

I have also uploaded a second Blender file - mantaflow volume_bug 02.blend - where the volume flow relative to the Z-axis of the inflow object is more obvious. Strangely enough, this file when loaded into Blender 2.83 will show the same incorrect behaviour.





Event Timeline

Richard Antalik (ISS) changed the task status from Needs Triage to Confirmed.Jun 22 2020, 5:58 AM

This problem is now showing up in the latest release: v2.83.1 LTS

The difference seems to be caused by rB21485e94aac1: Fluid: Refactored fluid gravity settings.
In particular, by the change to these lines:

# translate invels (world space) to grid space\n\
if using_invel_s$ID$:\n\
    x_invel_s$ID$.multConst(scaleSpeedTime_s$ID$)\n\
    y_invel_s$ID$.multConst(scaleSpeedTime_s$ID$)\n\
    z_invel_s$ID$.multConst(scaleSpeedTime_s$ID$)\n\

Unfortunately, I do not know if the old or new behavior is correct. @Sebastián Barschkis (sebbas), is inflow supposed to work when the inflow object is under water?

Just adding my opinion here:
If the Mantaflow liquid simulation is meant to reflect real world behaviour then, 'yes', an inflow object that is under the surface of a fluid should still be able to 'pump' water into the tank, vessel, domain or what have you.

Sebastián Barschkis (sebbas) changed the subtype of this task from "Report" to "Known Issue".Jul 22 2020, 4:14 PM

Yes, ideally an inflow object should act like a water tap and continuously add fluid to the scene.
Right now though, any inflow object will never add more fluid than "its own highest level". This can be seen when minimising the forces in the scene (i.e. no initial velocities and near zero gravity) - the level of fluid should not go higher than the inflow object itself.

Once the forces are higher and there is some splashing going on you will see fluid levels increasing. Fluid can expand better and this can, for example, lead to the impression of a bucket filling up. In the 2.83 scene the gravity forces were a bit stronger and so it did a better job of filling up the bucket.

Can this be behaviour be improved? Certainly! Ideally, there should be a force that counters gravity and pushes fluid particles up once an area has been filled up with fluid.
I am tagging this as a known issue for the time being.