I have a 2D TFM that is about 0.2m * 0.2m, with 100*60 mesh and max 10 facets per cell. I also have a simple chemical reaction for the model. To simulate the first 2 seconds of my model, I expected it to take maximum 2 hours. However, when I ran it, MFiX GUI indicated that it would take more than a day. I need to simulate the system for 12 minutes, which will take about a year…
I’m very new to CFD so I’m not sure whether this is actually the amount of the time needed to do the simulation, or I did something wrong in my model. Is there a way to roughly estimate how long a model takes to simulate? I thought that if every cell is computed 1,000 times for each time step (1e-3 seconds), then for 6,000 cell we need to numerically compute 6 millions Navier-Stokes equations. I guessed that an average computer can compute N-S equations 1,000 times each second, so it would take less than 2 hours. Am I wrong?
Also I attached my mfx file and user_rates.f in case anyone would like to help me take a look. 2d_jelly_roll.mfx (23.4 KB) usr_rates.f (2.4 KB)
Jack - I haven’t had a chance to run your example yet, but often the initial time estimate is inaccurate, and after a while it converges to a more reasonable value. Off the top of my head, one year sounds excessive. If you let it run for a bit, does that number decrease?
I’ll take a closer look when I’m back at my desk next week.
Thank you Charles! I reduced the number of cells (from 6,000 to 1,500) and the run time is lower, but it still takes quite long. Here are the updated files. td_jelly_roll.mfx (25.9 KB) usr_rates.f (2.4 KB)
I’m also running it on a cluster now using this command
build_mfixsolver --smp
env OMP_NUM_THREADS=16 ./mfixsolver -f td_jelly_roll.mfx
It doesn’t seem like the run time was improved much. I ran it for about 13 hours on the cluster, and it seems to takes more than a month to run 720 seconds of simulation based on the log file.
I think I solved the run time problem by running the DMP solver in 16 processors. It takes only 3 days to run the simulation now. However, could you help me take a look at my reaction rate coefficient problem? Thanks a lot!