Smoke Simulation with moving object is breaking

[robosherpa]

While creating smoke simulation with a moving object such as a sphere or a cuboid breaks the simulation.
Here is an example of the breaking simulation:

object-interaction-moving-object-sphere-broken-1.mp4

object-interaction-scenario-phiflow-can-break-at-times.mp4

It would be helpful to know if there is a plan or a patch to fix this issue.
Please let me know.

Thank you for your kind attention.

In [2]: import phi

In [3]: phi.__version__
Out[3]: '3.0.0'

In [4]: import torch

In [5]: torch.__version__
Out[5]: '2.6.0+cu126'

[p-holl]

Could you try with Solve(rank_deficiency=0) in make_incompressible().

[robosherpa]

Thanks @holl- . It seems more stable than before with Solve(rank_deficiency=0) in make_incompressible().

[robosherpa]

Still getting the same issues after trying the proposed solution.

object-interaction-moving-object-continues-to-break.mp4

object-interaction-moving-object-continues-to-break-with-velocity-field.mp4

Stationary objects seem stable however:

object-interaction-moving-object-continues-to-break-stationary-is-stable.mp4

[robosherpa]

@holl- Commenting for your kind attention. I would appreciate your feedback or recommendation. Thanks.

[Damian-Marszalski]

This issue still appears to persist, but in my case, in complex 3D plume simulations.

I've traced the problem in the develop branch to the following commit:

77f39b7

Prior to this change, everything runs without instability.
Although, in my case, as suggested previously, adding rank_deficiency=0 to Solve() in make_incompressible() fixes the problem.

As a temporary workaround, reverting to the previous commit cd7100b may work:

pip install git+https://github.com/tum-pbs/PhiFlow@cd7100ba0c63d7976903f8e39f6a300ecfd6ef3e

plume-explosions.mp4

[bqcom]

[like] Bikram Adhikari reacted to your message:

…

________________________________ From: Damian Marszalski ***@***.***> Sent: Tuesday, June 24, 2025 1:46:33 PM To: tum-pbs/PhiFlow ***@***.***> Cc: Bikram Adhikari ***@***.***>; Comment ***@***.***> Subject: Re: [tum-pbs/PhiFlow] Smoke Simulation with moving object is breaking (Issue #195) WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros. [https://avatars.githubusercontent.com/u/85636510?s=20&v=4]Damian-Marszalski left a comment (tum-pbs/PhiFlow#195)<#195 (comment)> This issue still appears to persist, but in my case, in complex 3D plume simulations. I've traced the problem in the develop branch to the following commit: 77f39b7<77f39b7> Prior to this change, everything runs without instability. As a temporary workaround, reverting to the previous commit cd7100b<cd7100b> works: pip install ***@***.*** https://github.com/user-attachments/assets/9e9c4d14-efc4-4b54-a950-35ed889d17ca — Reply to this email directly, view it on GitHub<#195 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BOSK47TGPV3UTN6QXSR5VFL3FFJDTAVCNFSM6AAAAABY4NHX2GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTAMBQGU3DMMZZHA>. You are receiving this because you commented.Message ID: ***@***.***>