Obi Official Forum
Suggestion / Idea Adaptive fluid simulation - Printable Version

+- Obi Official Forum (https://obi.virtualmethodstudio.com/forum)
+-- Forum: Obi Users Category (https://obi.virtualmethodstudio.com/forum/forum-1.html)
+--- Forum: Obi Fluid (https://obi.virtualmethodstudio.com/forum/forum-3.html)
+--- Thread: Suggestion / Idea Adaptive fluid simulation (/thread-2909.html)



Adaptive fluid simulation - mimarilker - 04-05-2021

Hi,

I really would like particles increasing in number around obstacles. This will really increase performance since the areas far away from obstacles will include significantly fewer particles. Moreover, this will also increase accuracy drastically since the interaction of fluids with obstacles will be solved with much more particles. Is it possible with the latest version? Can ve manually code it? Please see this 2 minutes paper video


RE: Adaptive fluid simulation - josemendez - 04-05-2021

(04-05-2021, 07:52 AM)mimarilker Wrote: Hi,

I really would like particles increasing in number around obstacles. This will really increase performance since the areas far away from obstacles will include significantly fewer particles. Moreover, this will also increase accuracy drastically since the interaction of fluids with obstacles will be solved with much more particles. Is it possible with the latest version? Can ve manually code it? Please see this 2 minutes paper video

This is not possible to do in realtime.

Obi is a fully lagrangian, particle-based simulator. The technique shown in that video is a particle-in-cell / fluid-implicit particle method, where the bulk of fluid simulation (pressure projection) is performed using a grid, then advection is done using particles.

For grid adaptivity, they generate a constrained delaunay tetrahedralization of the simulation domain every few frames. This step alone is orders of magnitude more expensive than what Obi does for the entire simulation, as tetrahedralizing a moderately large volume can take several seconds per frame using the fastest known method (isosurface stuffing, in this case: https://people.eecs.berkeley.edu/~jrs/papers/stuffing.pdf).

So this method can be an advantage for huge, extremely detailed fluid simulations that take minutes or hours per frame to compute, but it's not applicable to realtime fluid simulation where everything has to happen in a few milliseconds.