Is there a efficient way to reduce computation time in structures connected with long pipes? Trying to model a system with large chambers connected by 300 m of 0.25 m pipe with periodic pump ports.
Thanks, Jon
Hello Jon,
Yes and no... well, the particles have to do the random walk through the long thin pipes, which will always take a long time.
You could, maybe, use periodic boundary conditions instead of modeling the whole periods of long pipe/pump assemblies, but if I understand your description correctly, that wouldn't work either because your system is not infinite (but the large pipes are only connecting large tanks).
You could, however do a trick, but it's pretty advanced and prone to errors. I demonstrate it below. Say you have a system with two large volumes connected by a long pipe:
You can isolate the pipe+pump assembly and launch particles from one, then the other side of the pipe, while keeping the pumping speeds:
On this screenshot (right click+open image for full size) you can see that ~98% of particles come back to the first volume, 1.4% gets pumped and 0.73% goes through.
You should repeat the same launching particles from the other side if the pipe is not symmetric.
Then with this precalculated 97.9% return, 1.4% pumping and 0.7% crossing rate you can simplify your system:
(I moved the cubes closer but it's not necessary, they will be connected by "teleports").
Instead of the long pipe and pump, I use a succession of 3 circular facets at the opening (0.001cm away, so that particles don't leak between them):
1) First facet with 98% opacity (will turn back 98% of the particles, just like the long tube)
2) Second facet that teleports to the next cube, with opacity=0.36 (36% of the remaining 2% of particles will be teleported to the next cube, which equals 0.73% just like in the original system)
3) Third facet with sticking=1 (the rest, or 1.4% will be pumped)
Then you do the same on the other cube.
There are a few tricks when setting up the teleports (the facets have to be two-sided and facing the same direction), as described here. I've also added a reflective collar around the three circular facets to prevent particles leaking out between them.
Here are the three stages:
original system.zip
pipe analysis.zip
simplified equivalent.zip
The result is very fast (no pipe at all to traverse) but setting it up is labor-intensive. In the end, you have to find a good balance between your time (which is expensive) and computation time (which is cheap) :) If you have the same long pipe (300m and 0.25 diam), then one precalculation is enough. Good luck!