Dear Elmer users
This post concerns some developments almost one year back that are still not at a very mature state. I haven't had time to pursue this work so even though the material is preliminary I thought that I would share the developments with the forum in the hope of getting some feedback. If you're interested in the standard and more mature capabilities of ElmerSolver you may skip this message.
There is a new approach to assembly and solution in ElmerSolver which builds to the following principles:
- the matrix structure is defined using the real assembly loops and a list matrix structure
- only the element-level assembly must be provided by the user. The assembly loop and solution procedure are hidden
Some of the benefits of the system are that it becomes much easier to:
- write new solvers for PDEs
- write fully coupled routines even though they would posses different domains
- add constraints
- introduce robust nonlinear solvers
Some drawbacks are:
- new assembly routines must be written (or copy-pasted)
- solver specific tailoring may be more difficult
- parallelization much more difficult
- not supported by ElmerGUI
These new capabilities demonstrated by the attached 9 small examples (some info in the Readme.txt). There are examples related to strong coupling, constraints and stepsize controlled nonlinear solvers. The equation is simple Poisson. In some cases nonlinearity is introduced by a source term. Some of these could even also be solved with the more traditional approach but perhaps not with the same flexibility and modularity.
All feedback is appreciated. The syntax has not yet been fixed and major changes are still possible. None of this functionality is yet in the tests nor are they documented anywhere. So the attached package is all the information there is.
Best regards,
Peter
Developments in strong coupling and constraints in Elmer
-
- Site Admin
- Posts: 4812
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Developments in strong coupling and constraints in Elmer
- Attachments
-
- CoupledSolver.tar.gz
- (5.73 KiB) Downloaded 495 times
-
- Posts: 5
- Joined: 19 Jan 2012, 14:56
- Antispam: Yes
Re: Developments in strong coupling and constraints in Elmer
Peter,
I am impressed about the project itself. For now i used ELMER for helmholtz (steady state) , mechanical eigen value calculations, and stiffness (bending calculations). I am working in the field of multiphysics, like coupling acoustic and mechanical and the coupling to electro domain by using of piezo(s) (special case). Therefore new solvers with coupling have a lot of interest for me and will be in the future possible more important. Also for me is calculations with more then 600.000 nodes normal. I see with umfpack out of memory errors (8Gbyte internal ram), but my knowledge is to small to get into detail of the solver itself.
Vliegenthart
I am impressed about the project itself. For now i used ELMER for helmholtz (steady state) , mechanical eigen value calculations, and stiffness (bending calculations). I am working in the field of multiphysics, like coupling acoustic and mechanical and the coupling to electro domain by using of piezo(s) (special case). Therefore new solvers with coupling have a lot of interest for me and will be in the future possible more important. Also for me is calculations with more then 600.000 nodes normal. I see with umfpack out of memory errors (8Gbyte internal ram), but my knowledge is to small to get into detail of the solver itself.
Vliegenthart
-
- Posts: 36
- Joined: 22 Aug 2012, 13:43
- Antispam: Yes
Re: Developments in strong coupling and constraints in Elmer
It is a very good new!!! We need more coupled physic! for example I work, for my phD thesis, with COMSOL Multiphysic for the coupling of Mechanic+Piezo+Electric circuit. Is there a reasonably possibility to view in a short time something similar with Elmer? Thank you for your work!!