Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Extreme, seemingly unprompted spike in simulation time.
Posted Nov 10, 2021, 8:15 a.m. EST Fluid & Heat, Studies & Solvers Version 5.6 1 Reply
Please login with a confirmed email address before reporting spam
I've been running many simulations on a specific geometry and mesh, the steps uses Darcy's law to produce a velocity field, then two time dependent steps measure short term and long term transport of diluted species, using the velocity field as a convection field. THe first solver takes steps of 5s from 0 to 200s and while the second takes timesteps from 0.00201d (roughly the end time of the first) and takes steps of 0.1d to 28d
I've been experimenting with options in the physics settings for some time now and the simulation consistently takes around 40 seconds to run. One time, I decided to create a new file with a few minor differences to the old file's geometry. The simulation took far longer than expected and I decided to cancel it. However, from there on out, the issue seemingly pervaded throughout COMSOL, I had made no alterations to the original model yet its study started taking over 15 minutes to run, I even ran a backup that was 5 days old at the time this issue started and it also took much longer . It's particularly stuck on both time dependent solvers.
Is this a known issue? Unfortunately, I am unable to share the specific geometry, though it is fairly complex and abstract, there has been no changes in it that have prompted this issue.
Mainly pointing to this article: https://uk.comsol.com/support/knowledgebase/1254; I have tried restoring the default solver, editing the time stepping settings in the solvers to fixed and strict instead of free, setting a manual tolerance of 1%, restarting COMSOL, restarting the computer and made sure not to have saved any changes to this model since too but the issue isn't helped.
Upon further inspection, it does seem to be that individual time steps are suddenly taking around 3-8 seconds to compute, is this usual for a 2D model taking about 15 MB?