CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   AMG solver error (http://www.cfd-online.com/Forums/fluent/28702-amg-solver-error.html)

Matthew Brannock July 9, 2001 23:25

AMG solver error
 
Hi,

I've been getting this AMG solver error when attempting to iterate my solution:

Error: Internal error at line 743 in file 'amgif.c'. divergence detected in AMG solve

In my CFD model I'm using a UDS transport equation to simulate the settling of solids in a sedimentation column. Concentration is the scalar term so density is inside the derivitives (the DEFINE_UNSTEADY_UDS has been included accordingly for the transient term). Perhaps I should be using mass fractions instead?

I know that there have been several questions along the same lines with regards to this error. I've taken all precautions to stop the divergence, such as very low relaxation factors and time step with lower order discretisation models and in 2ddp but I still get this error.

I know that the likely reason for the problem is the setup of my UDS UDF but if there are any other suggestions they would be appreciated.

As a side note, when compiling (with visualC++ v5.0 and for fluent 5.5) the UDF I got this warning:

warning LNK4064: conflicting subsystem; image may not run

Would this be a possible cause?

cheers,

Matthew.


Dark Pain July 13, 2001 18:32

Re: AMG solver error
 
Assuming you are using the segregated solver, this error occurs when you have a non-diagonally dominant matrix and the Gauss-Seidel smoother in AMG diverges.

Low relaxation factors may help, but the real solution is to make sure your UDF's are not creating a non- diagonally dominant system.


All times are GMT -4. The time now is 14:43.