Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
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.
several problems
Posted May 24, 2012, 1:54 p.m. EDT Structural Mechanics Version 4.2a 2 Replies
Please login with a confirmed email address before reporting spam
hi,
We bought the floating license and just installed on several computers a few weeks ago. As a newer, I am confused about the following questions. Could you help me to solve it?
1: I run a CAD model in a 32 bit operation system laptop, but I even can't do the mesh, for the comsol tells me " out of memory in meshing". But I can do the mesh in another
64 bit operation system laptop. Do you have any suggestion for me to run the comsol in the 32 bit operation system laptop?
2. To save the memory, which solver will you recommend? Mumps, Paroiso and Spooles. How to determine the relationship between memory and accuracy?
3 Error: the relative residual is greater than the relative tolerance. Can you give me some suggestions to solve this problem?
4.I mesh all parts separately and there is no warning or errors, however, the plot shows there are some edge didn't mesh well. If I select the a courser mesh method and to compute the model, the relative residual will be larger than the relative tolerance.
thanks
have a great day
Best!
We bought the floating license and just installed on several computers a few weeks ago. As a newer, I am confused about the following questions. Could you help me to solve it?
1: I run a CAD model in a 32 bit operation system laptop, but I even can't do the mesh, for the comsol tells me " out of memory in meshing". But I can do the mesh in another
64 bit operation system laptop. Do you have any suggestion for me to run the comsol in the 32 bit operation system laptop?
2. To save the memory, which solver will you recommend? Mumps, Paroiso and Spooles. How to determine the relationship between memory and accuracy?
3 Error: the relative residual is greater than the relative tolerance. Can you give me some suggestions to solve this problem?
4.I mesh all parts separately and there is no warning or errors, however, the plot shows there are some edge didn't mesh well. If I select the a courser mesh method and to compute the model, the relative residual will be larger than the relative tolerance.
thanks
have a great day
Best!
2 Replies Last Post May 24, 2012, 9:15 p.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
indeed there are many things to look after. Most FEM moels require easily several Gb of RAM (if you do not spend some time simplifying it.) With a 32 bit OS your are limited to 2Gb of adressable RAM - the OS itself which tend to take a lot (often > 50% of available). with y 64 bit OS you can address about as much you can reasonably buy today so you are far less limited.
On the other hand, setting up a 64Gb model will take long to solve, and if by some simple sutides, using symmetry etc you can reduce it by 32x its worth to think over it and learn how to simplify your models.
Often direct solvers uses more RAM as iterative ones, but on the other side direct solvers often are quicker or just manage to converge when iteratives start running round in loops. here it depends a lot on the model you have. I mostly trust COMSOL, in >80% of the cases I do not touch my solver settings from the default ones.
Ther relative residual error often comes from a too coarse mesh, the result has gradients with steep slopes that are poorly resolved by the mesh, this can be very local, so often 80% of the model can be meshed coarse, while only the right 20% needs special attention. But this is very mdel dependent, CFD and turbulent flow, as well as RF models require very high mesh densities, and are often solved on clusters, at least for most reasonable enginnering models.
These are things one must learn, by doing and training.
My main avice start simple, it will mostly run quickly, then gradually compelxify until your PC gets saturated, or you get bored waiting then think of ways to simplify your model
--
Good luck
Ivar
indeed there are many things to look after. Most FEM moels require easily several Gb of RAM (if you do not spend some time simplifying it.) With a 32 bit OS your are limited to 2Gb of adressable RAM - the OS itself which tend to take a lot (often > 50% of available). with y 64 bit OS you can address about as much you can reasonably buy today so you are far less limited.
On the other hand, setting up a 64Gb model will take long to solve, and if by some simple sutides, using symmetry etc you can reduce it by 32x its worth to think over it and learn how to simplify your models.
Often direct solvers uses more RAM as iterative ones, but on the other side direct solvers often are quicker or just manage to converge when iteratives start running round in loops. here it depends a lot on the model you have. I mostly trust COMSOL, in >80% of the cases I do not touch my solver settings from the default ones.
Ther relative residual error often comes from a too coarse mesh, the result has gradients with steep slopes that are poorly resolved by the mesh, this can be very local, so often 80% of the model can be meshed coarse, while only the right 20% needs special attention. But this is very mdel dependent, CFD and turbulent flow, as well as RF models require very high mesh densities, and are often solved on clusters, at least for most reasonable enginnering models.
These are things one must learn, by doing and training.
My main avice start simple, it will mostly run quickly, then gradually compelxify until your PC gets saturated, or you get bored waiting then think of ways to simplify your model
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
thanks a lot, your suggestions are so great!