[dengdeng]

Hi,

I am simulating a case with Fluent which is similar to an unsteady supersonic backward facing step problem. However, dynamic mesh is included in my case so the mesh deforms as time goes by. In order to obtain better solutions I change the mesh to a finer one after some iterations, and the data is interpolated to the new mesh. Then I restart the case and encountered with this problem: 'Divergence detected in AMG for Coupled: protective actions enabled!'

It seems like the temperature exceeded the limitation first and Fluent try to fix the problem by reducing the time step. Several iterations after, new error occurred:
Internal error at line 1733 in file 'dbns\rp_mstage.c' on Node 0.
Divergence detected in AMG solver
Internal error at line 1733 in file 'dbns\rp_mstage.c' on Node 1.
Divergence detected in AMG solver
Internal error at line 1733 in file 'dbns\rp_mstage.c' on Node 2.
Finally, the simulation diverged compeletely.

The high-temperature region locates immediately after the step, which increase from around 1000K to 50000K.
I tried to reduce the CFL number from 0.5 to 0.001 but the problem has not been solved.

Does anyone have any idea to solve this problem?
Thank you in advance.

[dengdeng]

More details of my case are:
1. Unsteady supersonic flow past a deforming flat plate (dynamic mesh is introduced with a UDF), during when a backward facing step problem established;
2. The flow is inviscid;
3. Both spatial and temporal discretization methods are second-ordered.

The deformation deteriorates the quality of the mesh especially of the leading edge region, which is important for my simulation, thus, I replace the deformed mesh with a newly refined mesh every certain amount of time steps. The data is interpolated to the new mesh, then the solver mentioned temperature exceeded the limitation and divergence appeared several iteration after I restarted the simulation.

First, I checked the quality of my new mesh and the boundary conditions, all of them are reasonable. Then I checked the flow property and found that the interpolated flow field seemed to be strange at the leading edge. I tried to decrease the CFL number but nothing changed.

Finally I solve the problem like this:
First increase the limitation of temperature (for example, from 5000K to 500000K). Then switch the spatial and temporal discretization methods to first-ordered, and restart the simulation. After several time steps the temperature will finally decrease and the flow field become reasonable in comparison with the interpolated one. So I changed the discretization method back to second-ordered mode, and the simulation seems to be ok until now (I do not know if it will diverge again though).

Still, I am not sure if this is a relative 'right' way to solve the problem. If anyone has any better idea?