Divergence detected in AMG solver: pressure coupling

  • 162 Views
  • Last Post 20 February 2019
Tayyaba posted this 18 February 2019

Hi

I am doing 2 way FSI simulation of  a flap.Case is 3D and flow is laminar .I started simulation with pressure based transient flow using time step 1e-4. Coupled scheme is used with second order implicit transient formulation.

Case worked well until 1800 coupling steps then suddenly I got the error

"Divergence detected in AMG solver, pressure coupling"

I want to know the reason for this type of error, if any one could guide me please. Is there any problem in solver or discretization scheme?

Waiting for a kind response.

 

Kind regards

Tayyaba

 

Order By: Standard | Newest | Votes
abenhadj posted this 18 February 2019

Might be due to many reasons: you need to debug! Just check the results before that point and store cell residuals so that you find the regions which might resulted to divergence. Check if the displacement sent to Fluent from Mechanical are pretty large and here you need to assess for more stability (implicit update might be helpful here)

Best regards,

Amine

Tayyaba posted this 18 February 2019

Thank you very much for your kind reply.

Implicit update means change in scheme?

 

Regards

Tayyaba

abenhadj posted this 18 February 2019

Check the documentation. In Transient run do not reduce the URF to smaller value (best to be one not lower then 0.8). You need to monitor the force on the FSI walls in Fluent to monitor its evolution of the number of steps.

You have in Fluent the Solution Stabilization: Dynamic Mesh>System Coupling boundary (Solver Options). 

And for general Dynamic Mesh Motion: Implicit Update

Best regards,

Amine

Tayyaba posted this 18 February 2019

Thanks once again.

But I was working with URF as 0.1 in data transfer 'Transient structural' as a participant and for 'Fluent flow' as a participant I used  URF as 1.

And yes I have already used this solution stabilization option in some other simulation but still the question is which method to choose?Either volume based or coefficient based? and the right value of scale factor.

Also does 'double precision' effect this type of error? as I did not make it on.

 

Regards

Tayyaba

abenhadj posted this 18 February 2019

Coefficient based has shown to be more robust and not mesh dependent. So go for it.

Best regards,

Amine

Tayyaba posted this 20 February 2019

Thank you very much for your reply.

Is it possible to run the existing case with these changes? or should I have to run the case from beginning which is quite hard (:

as I got the error almost at the middle of solution.

 

Regards

Tayyaba

abenhadj posted this 20 February 2019

Best to run from scratch as usual.

Best regards,

Amine

Close