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.
Criteria to select the size of the mesh and time steps in a time dependent model.
Posted Dec 19, 2011, 3:27 p.m. EST 4 Replies
Please login with a confirmed email address before reporting spam
Hi
I'm running a simulation of a stirring tank by using the rotating machinery-laminar flow COMSOL module . The problem includes time dependency , and I am not sure about with size of mesh works better for my model.since I have two parameters such as: size of the mesh and delta time. I have been using the adimensional number:
(u)*(delta t)/(delta x)<0.5
Which means
(velocity)*(delta time)/(minimum size of the mesh)<0.5
Everytime I change the size of the mesh and the time steps in order to guarantee the criterion, I have a different answer in the velocity and pressure distribution.
Does anybody know something about this criterion? Does anybody know how to select the appropriated minimum size of the mesh and delta time for an accurate solution?
Thanks!
I'm running a simulation of a stirring tank by using the rotating machinery-laminar flow COMSOL module . The problem includes time dependency , and I am not sure about with size of mesh works better for my model.since I have two parameters such as: size of the mesh and delta time. I have been using the adimensional number:
(u)*(delta t)/(delta x)<0.5
Which means
(velocity)*(delta time)/(minimum size of the mesh)<0.5
Everytime I change the size of the mesh and the time steps in order to guarantee the criterion, I have a different answer in the velocity and pressure distribution.
Does anybody know something about this criterion? Does anybody know how to select the appropriated minimum size of the mesh and delta time for an accurate solution?
Thanks!
4 Replies Last Post Feb 18, 2012, 10:43 a.m. EST
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
my way (but I'm mostly in structural and HT) is to first estimate the mesh / time step ratios with formulas as yours, then solve and resolve for a 2x finer mesh, until I'm happy with the results, i.e. < 5-10% change when I double the mesh density.
now mostly you have only a few areas with critical mesh in a complex model, so it's not always required to double the mesh density everywhere, that is too expensive in computing time. mainly I check all my fluxes and see if I correctly estimate all gradients.
You also have the adaptive meshing in the solver settings, all depends on the model and time you have
--
Good luck
Ivar
my way (but I'm mostly in structural and HT) is to first estimate the mesh / time step ratios with formulas as yours, then solve and resolve for a 2x finer mesh, until I'm happy with the results, i.e. < 5-10% change when I double the mesh density.
now mostly you have only a few areas with critical mesh in a complex model, so it's not always required to double the mesh density everywhere, that is too expensive in computing time. mainly I check all my fluxes and see if I correctly estimate all gradients.
You also have the adaptive meshing in the solver settings, all depends on the model and time you have
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thank you so much!
Do you know the name of this adimensional number? I am looking for more information about mesh in finite elements, but I could not find this specific criterion.
Do you know the name of this adimensional number? I am looking for more information about mesh in finite elements, but I could not find this specific criterion.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
the average mesh size is "h" check te doc, its a field (changes with location)
--
Good luck
Ivar
the average mesh size is "h" check te doc, its a field (changes with location)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi. I know it is too late. But now I learnt it is called 'Courant-Friedrichs-Lewy (CFL) number.