I don't check the options of COMSOL every time I run it and I didn't face such issues before. You can say that I'm noob XD but I don't know how it happened that this option was unticked. As soon as I tick it, I got very same results of simulation as on laptop but faster (11 min.) But I also realized that I didn't tick "Memory allocator" to turn on "Optimized for multicore" option. Anyway, I set 8 cores for laptop (COMSOL v5.5) and 12 cores for desktop (COMSOL v 5.6). Today, I'm not sure that this is a good approach (I would appreciate your opinion on that). A few years ago, I figured out that it results to full load of processor and shortest time of simulation. Since I'm very greedy, I set 2x number of physical cores = number of logical cores. Here we can set number of available cores. Is it worth to say that I almost got crazy searching for the cause?.įinally, today, Friday evening, I've found the problem where I'd never think to find!įile> Preferences>Multicore and Cluster computing. Anyway, results were different, I mean not time only, but convergence plot that is much more important. And be sure that I created the file on the laptop, ran simulation, copied to desktop and again ran the file. Everything the same, excepting GCRO-DR solver in version 5.6, which can be turned off. I spent 4 days to find the difference in model boundary conditions, mesh, solver configuration and so on. If something goes wrong, I use my own laptop as universal problem solver =D (COMSOL 5.5, core i5-8300H, 12 GB RAM). Then I set P2+P1 and simulation took 16 HOURS! Btw, Configuration of computer is next: COMSOL 5.6, core i7-4th gen, 64 RAM The simulation execution took about 11 seconds. Also, It is a well-known fact that for Laminar flow module the P2+P1 (quadratic for velocity and linear for pressure) discretization is precise enough. For this case, I planned to calculate velocity and pressure field via Stationary study as flow doesn't change with time. So, I build coarse mesh, set linear discretization and so on, just to get fast result and then refine it. Usually, I create very coarse model just to check if I made not some stupid mistakes in IC/BC, coz according to my experience 99% of issues are caused by incorrect IC/BC. But let's focus on Fluid mechanics module. I was going to check velocity distribution (Laminar flow) and mass transfer (Diluted species). I have very trivial geometry of microfluidic channel (3-D). A few days ago, I made a post, which was inspired by some troubles I was struggling in Comsol.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |