Hello,
I am looking for a tool to solve Transient Heat problems where the material have a custom internal heat source.
The Elmer docs states something like "Elmer may be used to solve transient problems...". The "may" made me wonder.
Also I see few Transient heat problems in the samples folder. Searching these forums, I find 5 matches to "Transient" and Heat, and they have all run into problems that I cannot find have been solved....
OK, so my real question is: Is Elmer a good tool for Transient Heat problems as specified above, or should I look elsewhere?
gos
Elmer and Transient Heat Problems
-
- Site Admin
- Posts: 4823
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Elmer and Transient Heat Problems
Hi gos
Basically Elmer is well suited for transient problems. The combination stabilized N-S solver & heat solver have been used for complex problems, for example in the area of silicon crystal growth. There a proper choice of timestepping is important. I have a gut feeling and many of the problems in the forum are due to people solving demanding problems of natural convection without proper time and space resolution. If you cannot resolve the physics properly with the resolution that you can computationally afford you should use RANS models. In there, Elmer is probably not the best choice as these have not been the focus of development, and the RANS models are also poorly documented. Elmer solvers work rather well in the LES regime where particularly the stabilization method "vms" (Variational Multiscale Method) might be worth trying.
-Peter
Basically Elmer is well suited for transient problems. The combination stabilized N-S solver & heat solver have been used for complex problems, for example in the area of silicon crystal growth. There a proper choice of timestepping is important. I have a gut feeling and many of the problems in the forum are due to people solving demanding problems of natural convection without proper time and space resolution. If you cannot resolve the physics properly with the resolution that you can computationally afford you should use RANS models. In there, Elmer is probably not the best choice as these have not been the focus of development, and the RANS models are also poorly documented. Elmer solvers work rather well in the LES regime where particularly the stabilization method "vms" (Variational Multiscale Method) might be worth trying.
-Peter
Re: Elmer and Transient Heat Problems
Hello Peter,
Thanks a lot for your detailed answer. I have currently modeled most of my model with a commercial tool which works great, but pricing may prohibit its use in an application I am working on. I am doing 2D analysis.
If you don't mind, I have a few questions for you:
a) Is there any simple Transient Heat examples that I might look at to get me started?
b) Is it possible to have time-varying equations in the solver (e.g. varying internal heat sources)
c) Is it possible to ask Elmer to integrate different variables?
d) Is it possible to plot temperature (or any other variable) at any point in 2D as a function of time?
Thanks
gos
Thanks a lot for your detailed answer. I have currently modeled most of my model with a commercial tool which works great, but pricing may prohibit its use in an application I am working on. I am doing 2D analysis.
If you don't mind, I have a few questions for you:
a) Is there any simple Transient Heat examples that I might look at to get me started?
b) Is it possible to have time-varying equations in the solver (e.g. varying internal heat sources)
c) Is it possible to ask Elmer to integrate different variables?
d) Is it possible to plot temperature (or any other variable) at any point in 2D as a function of time?
Thanks
gos
-
- Site Admin
- Posts: 4823
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Elmer and Transient Heat Problems
Hi
a-d) yes, You might have a look at the R-B case in the tutorials.
-Peter
a-d) yes, You might have a look at the R-B case in the tutorials.
-Peter
Re: Elmer and Transient Heat Problems
Hello Peter,
Thanks, I have looked at the "Rayleigh-Benard instability" example, but it is too complex to give me an idea of how Elmer performes timewise for pure heat flow in a solid.
a) This problem uses 2 solvers, one for "Heat Equation" and the other for "Navier-Stokes" which I do not know what is, and proably don't need for a simple Heat Transfer problem (with a time dependent heat source)?
Given a rectangular (2D) shaped material, width = 1 meter, thickness = 0.2 meter divided in a 20*20 grid, my current tool runs thrrough a Transient analysis from t=0 to t=16 hours with timestep = 5 minutes in about 20 seconds: Even this is a bit slow for my purpose!
My current tool uses a smart dynamic time-step where it analysis the error and adjusts the timestep dynamically: It may be as large as 1 hour where small changes occur, but go down to less than 5 minutes if there are large changes in the output.
b) Isn't there a very basic Transient time Heat-flow example that could get me started?
c) Has Elmer any dynamic time-step capability like the one I mentioned?
Thanks for any input.
Thanks, I have looked at the "Rayleigh-Benard instability" example, but it is too complex to give me an idea of how Elmer performes timewise for pure heat flow in a solid.
a) This problem uses 2 solvers, one for "Heat Equation" and the other for "Navier-Stokes" which I do not know what is, and proably don't need for a simple Heat Transfer problem (with a time dependent heat source)?
Given a rectangular (2D) shaped material, width = 1 meter, thickness = 0.2 meter divided in a 20*20 grid, my current tool runs thrrough a Transient analysis from t=0 to t=16 hours with timestep = 5 minutes in about 20 seconds: Even this is a bit slow for my purpose!
My current tool uses a smart dynamic time-step where it analysis the error and adjusts the timestep dynamically: It may be as large as 1 hour where small changes occur, but go down to less than 5 minutes if there are large changes in the output.
b) Isn't there a very basic Transient time Heat-flow example that could get me started?
c) Has Elmer any dynamic time-step capability like the one I mentioned?
Thanks for any input.
Re: Elmer and Transient Heat Problems
Hello Peter,
One more important question: I want to call Elmer from a non-fortran and non-c language: I can interface to DLLs:
a) Is it possible to compile Elmer (Solver ?) as a DLL and call functions in the DLL from my system?
I guess I could call the non-gui version using a system call that passes arguments to on the command line: However, I would rather not reveal some of the formulas used to end users
b) How would you suggest I go about to call Elmer from my system?
- DLL
- Command Line (any support for encrypting formulas passed along on the command line?)
gos
One more important question: I want to call Elmer from a non-fortran and non-c language: I can interface to DLLs:
a) Is it possible to compile Elmer (Solver ?) as a DLL and call functions in the DLL from my system?
I guess I could call the non-gui version using a system call that passes arguments to on the command line: However, I would rather not reveal some of the formulas used to end users
b) How would you suggest I go about to call Elmer from my system?
- DLL
- Command Line (any support for encrypting formulas passed along on the command line?)
gos
-
- Site Admin
- Posts: 4823
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Elmer and Transient Heat Problems
Hi gos
If you do not have coupling to fluid flow I would not expect any significant difficulties with the heat solver. Of course Elmer might not be as fast a tool as one that has been tailored for the purpose. In Elmer things can be very generic and this has some cost. Still you should be able to optimize the performance by selecting the linear system solvers and if possible, assembling the stiffness matrix only once.
In time integration you should try to higher order schemes for accuracy. There is also adaptive timestepping (see test case TimeAdapt). I don't have too much experience with that. You can also give the timestep as an external function.
Calling Elmer through the file system will of course be possible but you pay a heavy price in I/O. I know that there has been some use of Elmer via API but I'm not too familiar with it. I would think that its difficult to pass information through the current API, it's more a handle for the control of the simulation. The encryption is not something that we've considered. It sort of goes againts the spirit of open source
-Peter
If you do not have coupling to fluid flow I would not expect any significant difficulties with the heat solver. Of course Elmer might not be as fast a tool as one that has been tailored for the purpose. In Elmer things can be very generic and this has some cost. Still you should be able to optimize the performance by selecting the linear system solvers and if possible, assembling the stiffness matrix only once.
In time integration you should try to higher order schemes for accuracy. There is also adaptive timestepping (see test case TimeAdapt). I don't have too much experience with that. You can also give the timestep as an external function.
Calling Elmer through the file system will of course be possible but you pay a heavy price in I/O. I know that there has been some use of Elmer via API but I'm not too familiar with it. I would think that its difficult to pass information through the current API, it's more a handle for the control of the simulation. The encryption is not something that we've considered. It sort of goes againts the spirit of open source
-Peter
Re: Elmer and Transient Heat Problems
Hello Peter,
Thanks yet again. I still do not understand how I can use Elmer from my App though.
Can Elmer be compiled into a DLL?
If it can, are you saying that the current API calls are not easy to use?
The FAQ states:
Q: Can I use parts of Elmer in my own code?
A: Yes, with the condition that if you publish the code it must be published under the same license i.e. GPL.
Then, how are people generally using Elmer if they integrate Elmer with their own C or Fortran code?
They certainly must be using the Elmer API in some way. But if they do, it should be possible to put a C wrapper around the Elmer code that exports the API functions I need to use?
Confused...
gos
Thanks yet again. I still do not understand how I can use Elmer from my App though.
Can Elmer be compiled into a DLL?
If it can, are you saying that the current API calls are not easy to use?
The FAQ states:
Q: Can I use parts of Elmer in my own code?
A: Yes, with the condition that if you publish the code it must be published under the same license i.e. GPL.
Then, how are people generally using Elmer if they integrate Elmer with their own C or Fortran code?
They certainly must be using the Elmer API in some way. But if they do, it should be possible to put a C wrapper around the Elmer code that exports the API functions I need to use?
Confused...
gos
-
- Site Admin
- Posts: 4823
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Elmer and Transient Heat Problems
Hi gos
The viral effect (copyleft) of GPL codes is there intentionally. Everybody is free to use and develope the code as they see best but they cannot distribute it under a different license. The are more liberal licensing schemes such as BSD or Apache. We have discussed on using LGPL or at least adding a linking exception to the dynamic solvers. Still, ElmerSolver is primarily a stand-alone software with great flexibility in the modularity of the simulation. We have well defined APIs when Elmer is the master but not vice versa. The answer in FAQ does not mean that there would be loads of projects utilizing Elmer through an API, it was more about interpreting the licensing scheme in concrete terms. There are pure FE libraries, such as LibMesh++ (under LGPL) or Deal II, if you want to include the simulation within your own software.
About the licensing options for commercial applications there is some discussion in:
http://programmers.stackexchange.com/qu ... of-process
-Peter
The viral effect (copyleft) of GPL codes is there intentionally. Everybody is free to use and develope the code as they see best but they cannot distribute it under a different license. The are more liberal licensing schemes such as BSD or Apache. We have discussed on using LGPL or at least adding a linking exception to the dynamic solvers. Still, ElmerSolver is primarily a stand-alone software with great flexibility in the modularity of the simulation. We have well defined APIs when Elmer is the master but not vice versa. The answer in FAQ does not mean that there would be loads of projects utilizing Elmer through an API, it was more about interpreting the licensing scheme in concrete terms. There are pure FE libraries, such as LibMesh++ (under LGPL) or Deal II, if you want to include the simulation within your own software.
About the licensing options for commercial applications there is some discussion in:
http://programmers.stackexchange.com/qu ... of-process
-Peter
Re: Elmer and Transient Heat Problems
Hello Peter,
Ah, so I have misunderstood the GPL license: I cannot use Elmer within a piece of commercial software... right?
Yes, I have seen both Deal II and LibMesh: Deal II is not free, and LibMesh has very limited docs if I am not wrong...
gos
Ah, so I have misunderstood the GPL license: I cannot use Elmer within a piece of commercial software... right?
Yes, I have seen both Deal II and LibMesh: Deal II is not free, and LibMesh has very limited docs if I am not wrong...
gos