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.
stop criterion adaptive mesh refinement
Posted 15 nov 2012, 04:35 GMT-5 Mesh, Studies & Solvers Version 5.2 4 Replies
Please login with a confirmed email address before reporting spam
Hi,
I have a question concerning the stop criterion of the adaptive mesh refinement (adaptive solver) algorithm:
In the Reference Guide stop criterion is only described by:
“Terminate execution if it has made the requested number of refinements or if it has exceeded the maximum number of elements”.
Is there any other stop criterion implemented like: Terminate execution if the estimated error is small enough?
I have a question concerning the stop criterion of the adaptive mesh refinement (adaptive solver) algorithm:
In the Reference Guide stop criterion is only described by:
“Terminate execution if it has made the requested number of refinements or if it has exceeded the maximum number of elements”.
Is there any other stop criterion implemented like: Terminate execution if the estimated error is small enough?
4 Replies Last Post 22 dic 2015, 10:52 GMT-5
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
isnt that the the standard (usual) stop methode ?, and for me the remeshing is not a "stop", but a criteria to hold and restart with a new mesh, but based on previous values, remapped to the new mesh, or am I missing something ?
--
Good luck
Ivar
isnt that the the standard (usual) stop methode ?, and for me the remeshing is not a "stop", but a criteria to hold and restart with a new mesh, but based on previous values, remapped to the new mesh, or am I missing something ?
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi again,
yes, it is the standard stop criterion.
Nevertheless in the Comsol Reference Guide the algorithm is only described as:
"The adaptive solver performs the following iterative algorithm (Ref. 9):
1 Solve the problem on the existing mesh using the stationary or eigenvalue solver.
2 Evaluate the residual of the PDE on all mesh elements.
3 Estimate the error in the solution on all mesh elements. The computed error
estimate is really just an error indicator because the estimate involves an unknown
constant (C above).
4 Terminate execution if it has made the requested number of refinements or if it has
exceeded the maximum number of elements.
5 Refine a subset of the elements based on the sizes of the local error indicators.
6 Repeat these steps."
If there is implemented a stop criterion like mentioned above, I wonder how comsol decides that the error estimate is small enough?
Ivar, do you speak about mesh refinement in a stationary background or about the time adaption algorithm?
yes, it is the standard stop criterion.
Nevertheless in the Comsol Reference Guide the algorithm is only described as:
"The adaptive solver performs the following iterative algorithm (Ref. 9):
1 Solve the problem on the existing mesh using the stationary or eigenvalue solver.
2 Evaluate the residual of the PDE on all mesh elements.
3 Estimate the error in the solution on all mesh elements. The computed error
estimate is really just an error indicator because the estimate involves an unknown
constant (C above).
4 Terminate execution if it has made the requested number of refinements or if it has
exceeded the maximum number of elements.
5 Refine a subset of the elements based on the sizes of the local error indicators.
6 Repeat these steps."
If there is implemented a stop criterion like mentioned above, I wonder how comsol decides that the error estimate is small enough?
Ivar, do you speak about mesh refinement in a stationary background or about the time adaption algorithm?
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
unfortunately I do not have aclear view of this process, I agree it's rather important to understand, to be able to decide which number to push where, and I cannot really remember that this subject was treated in my last solver COSMOl course, but that was already some time ago, with all the new V4 features, I suspect several elements have beed adapted
--
Good luck
Ivar
unfortunately I do not have aclear view of this process, I agree it's rather important to understand, to be able to decide which number to push where, and I cannot really remember that this subject was treated in my last solver COSMOl course, but that was already some time ago, with all the new V4 features, I suspect several elements have beed adapted
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Hi,
I want to revive this thread since in the last 5.2 COMSOL I still don't see a way to stop the adaptive mesh refinement when the error drops below a given threshold.
With the addition of the "Functional" estimation as a way to drive the adaptive algorithm, now the absolute value of the error becomes meaningful to the user so being able to specify a stop threshold is extremely useful.
For example, I am running an adaptive optimization in electrostatics using a boundary integral of charge to estimate capacitance as the functional that drives the algorithm. That means that the error is actually measured in terms of total charge difference between steps. I would like to chose to stop the algorithm if the delta drops below 1fC for example.
This is a feature that is either hidden somewhere I cannot find or should be added in the next release.
Thanks,
Roberto
I want to revive this thread since in the last 5.2 COMSOL I still don't see a way to stop the adaptive mesh refinement when the error drops below a given threshold.
With the addition of the "Functional" estimation as a way to drive the adaptive algorithm, now the absolute value of the error becomes meaningful to the user so being able to specify a stop threshold is extremely useful.
For example, I am running an adaptive optimization in electrostatics using a boundary integral of charge to estimate capacitance as the functional that drives the algorithm. That means that the error is actually measured in terms of total charge difference between steps. I would like to chose to stop the algorithm if the delta drops below 1fC for example.
This is a feature that is either hidden somewhere I cannot find or should be added in the next release.
Thanks,
Roberto