Discussion Forum

Posted: 1 decade ago Sep 8, 2010, 1:09 a.m. EDT

Hi that is somewhat model dependent, read throigh the doc, and try on very simple models until you catch the principles -- Good luck Ivar

Posted: 1 decade ago Oct 25, 2010, 12:18 p.m. EDT

What documents do you mean? I am trying to use an adaptive mesh as well and cant seem to find much help for it. Thank you :) RM

Posted: 1 decade ago Oct 26, 2010, 4:00 a.m. EDT

Good question. I could hardly find anything to use "functional" option of adaptive mesh and finally gave up on that. I want to have mesh refinement only around certain boundaries and not on all the subdomain, but so far no luck. -- Comsol 4.0a Ubuntu 10.04.1

Posted: 1 decade ago Oct 26, 2010, 6:26 a.m. EDT

Hi I agree: local mesh refinement would be interesting, didnt work well in 3.5, and I havent tried it correctly in V4 yet -- Good luck Ivar

Posted: 1 decade ago Oct 26, 2010, 8:07 a.m. EDT

Hi, There are many options here: The default mesh gives a fairly uniform mesh if the geometry is "smooth," but it does adapt to the local curvature. The global mesh parameters for the "resolution of curvature" and "maximum element growth rate" affect the mesh density close to small details in the geometry and how quickly the mesh size grows away from those small areas. You can also specify local mesh sizes in domains, along boundaries, or at points to create a finer (or coarser) mesh in parts of the geometry. You do this by adding Size nodes to the Mesh branch. In the Settings windows for the Size nodes you can select the geometric entity level (boundary, for example) and then the boundaries, for example, where the local settings apply. You can also create special boundary layer meshes, which are suitable for resolving boundary layers in fluid flow. They typically consists of narrow layers of quadrilateral elements along the boundary. Finally, adaptive mesh refinement is available as a solver option for stationary and eigenvalue solvers. The mesh is then adapted based on an error estimate during the solution stage. Best regards, Magnus Ringh, COMSOL

Posted: 1 decade ago Oct 26, 2010, 10:10 a.m. EDT

Hi Magnus I fully agree, my models though are often of type compliant structures (ST) with aspect ratios 10'000:1 or worse. Your mesher, with some precautions is doing very well. Nevertheless, before, on another programme I could use their refinement meshing efficiently, while they had a mode that really refined where the stress gradient was high (my issues are often stress concentration points). With COMSOL's L2 mode, I notice it refines, but very globally, so I quickly run out of RAM, while my hotspots remain not that nicely resolved. This means that I must find another norm that concentrates better on the stress gradient to mainly refine therearound. I didnt manage in 3.5 and havent tried really in V4. I have nobody "paying" me for searching this, and as I have many of these issues, and my clients wants results by yesterday, it's tricky for me/us out here to test and try ;) Hope within a year or two to manage to get some "cheap" students well trained to give me a hand -- Have fun Comsoling Ivar

Posted: 1 decade ago Oct 26, 2010, 10:21 a.m. EDT

[QUOTE] Hi Magnus I fully agree, my models though are often of type compliant structures (ST) with aspect ratios 10'000:1 or worse. Your mesher, with some precautions is doing very well. Nevertheless, before, on another programme I could use their refinement meshing efficiently, while they had a mode that really refined where the stress gradient was high (my issues are often stress concentration points). With COMSOL's L2 mode, I notice it refines, but very globally, so I quickly run out of RAM, while my hotspots remain not that nicely resolved. This means that I must find another norm that concentrates better on the stress gradient to mainly refine therearound. I didnt manage in 3.5 and havent tried really in V4. I have nobody "paying" me for searching this, and as I have many of these issues, and my clients wants results by yesterday, it's tricky for me/us out here to test and try ;) Hope within a year or two to manage to get some "cheap" students well trained to give me a hand -- Have fun Comsoling Ivar [/QUOTE] Exactly. Moreover, with single physics let's say it is more or less OK. But, what if you have multiphysics... say fluid flow and heat transfer. For flow it is ok to have more or less fine mesh (usually streamline diffusion takes care of that), but with heat you need 10-100 times more denser mesh - in my case around the bubble boundary. Since I have also complex fluid formation around the bubble you can not mesh it very systematically. Hence, Adaptive mesh refinement. -- Comsol 4.0a Ubuntu 10.04.1

Posted: 1 decade ago Nov 12, 2010, 3:22 p.m. EST

Dear Magnus, I understand that meshing subdomains is possible, but I think the question is: Is there an adaptive mesh available for time dependent solvers? Say we have a rigid sphere moving within the domain. Is it possible to re-mesh at every time step? The COMSOL help documentation seems to indicate that this is possible, but there is nothing that actually states how to do this. Thanks, Todd

Posted: 1 decade ago Apr 7, 2011, 6:14 a.m. EDT

Hey, I had a problem with an acoustic mesh that was too fine at a table (boundary) for the acoustic simulations that I was doing. My solution can be found on http://comsol.gertdesitter.eu/Home/interactive-meshing Maybe that tutorial can help.

Posted: 1 decade ago May 23, 2011, 1:24 p.m. EDT

yeah, I have the same problem. Is it possible to use adaptive mesh in each time step ?????

Posted: 1 decade ago May 23, 2011, 2:00 p.m. EDT

See http://www.comsol.com/products/4.2/ Adaptive meshing for transient simulations is available starting at version 4.2.