Discussion Forum

October 19, 2009 2:22pm UTC

Undefined value found

Hi,

i can solve my semiconductor model with a relatively coarse mesh.
However the found solution is not good enough. It shows some
differences to a reference (PC1D) semiconductor model.

That is why i want to refine the mesh.


But if i refine the mesh i get following error:

---------------------------------------------------------------------
Error: 7248
Undefined value found.

Details:
Exception:
com.femlab.jni.FlNativeException: Undefined value found Solvermodel
Messages:
Undefined value found Solvermodel

Stack trace:
at solvermodel.cpp, row 1837, (Undefined value found in the equation residual vector.
There are 85 degrees of freedom giving NaN/Inf in the vector for the variable phii
at coordinates: (0.000493076,-5.8675e-005), (0.00049406,-5.80051e-005), (0.000495044,-5.73351e-005), (0.000494026,-5.58557e-005), (0.000493041,-5.65256e-005), ...
)
at com.femlab.solver.FlSolver.femStatic(Native Method)
at com.femlab.solver.FemStatic.run(Unknown Source)
at com.femlab.server.FlRunner.run(Unknown Source)
at com.femlab.util.i.run(Unknown Source)
at com.femlab.util.aa.run(Unknown Source)
----------------------------------------------------------------------------

The model is quite complex and has 5 coupled application modes.
I use the Parametric segregated Solver.

I think if there would be a problem with my initialisation values or boundary contitions,
i would not be able to solve the model even with a coarse mesh?

I found out, that the problem especially occurs, if i refine the mesh in a region with large gradients.

The Comsol solution is represented by green circles and the reference solution is shown as line in the attached figure.

Has anyone an idea how the finer mesh causes "undefined value"-problems and how to overcome them?

Thank you,

Stefan

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 12, 2010 7:06am UTC in response to Stefan Eidelloth

Re: Undefined value found

Hi

My first reaction is that you are hitting a numerical limitation (you case is a typical examples of the type I send myself to COMSOL support).

As you say, normally you should have a denser mesh where your gradients are steep

Cannot comment further
Good luck
Ivar

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 17, 2010 2:12pm UTC in response to Ivar Kjelberg

Re: Undefined value found

Hello Mr. Ivar
Can you give me some advice with the following errors.
Error:
Time 1.386578133869988:
Error in Jacobian computation:
Undefined_value_found_Solvermodel
Undefined value found in the stiffness matrix.
Error:
Time 5.125383036305885:
Nonlinear solver did not converge.
Last time step is not converged.
I'm looking forward to your reply.
Thank you.
N.L

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 18, 2010 8:20am UTC in response to Ivar Kjelberg

Re: Undefined value found

Hello Mr. Ivar KJELBERG
Thank you for your advice.
I sent a private message to Mr. Ivar.
Please help me.
Sincerely,
N.L
nhhieubk@gmail.com

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

April 13, 2010 5:17pm UTC in response to nurul hasan

Re: Undefined value found

Hi,
I did not further work on that specific issue
and i did not try the "restart option",
but please keep us updated if you do.

Currently I am working on a model with a
conductive boundarie instead of a resolved emitter.
This way I can avoid high gradients.

Best regards,
Stefan

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

April 16, 2010 1:34pm UTC in response to Stefan Eidelloth

Re: Undefined value found

I had the same problem.

The reason is possibly that initial value is zero in the first calculation step. I changed the initial value from zero to a expression not Zero.

It worked.

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

November 4, 2010 3:15pm UTC in response to Aaron Quastel

Re: Undefined value found, error 7248

I am facing a similar problem now. I am trying to couple heat and mass transfer in a transient problem. For mass transfer, there is drastic difference in concentrations between subdomains. In one of the boundaries, sublimation is assigned. Please find the attached file. I set the 'Initial value' in the Solver Manager window to 'Initial value expression' and it did not work. I tried to do using stationary solver and store the values. However that also didn't work.

Thank you so much for your time.

Roopesh

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

January 25, 2012 1:39pm UTC in response to Gerard Santillan

Re: Undefined value found

Hi Gerard,

it is long time ago I worked on that issue. You might try to have an initial value expression with non-zero slope (if you did not already do so). And you might want to have a look at our CoBoGUI:

http://www.isfh.de/institut_solarforschung/software.php?_l=1

It uses a model where diffused regions are simplified to special boundary conditions.

Good luck,

Stefan

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse