[Yahia Shannan]

Hello everyone,
I'm simulating a 3D offset duct in fluent 16. Realizable ke with enhanced wall treatment is incorporated. The hexahedral cells are 3.5 million. Coupled scheme and PRESTO for pressure are used as well as Pseudo Transient. I'm expecting some adverse pressure gradients and flow separation due to the physical curvature of the duct. Unfortunately, after only 13 iterations an error prompted "Divergence detected in AMG solver : temperature". I couldn't find any explanation so please help me. I'm running out of time !
Best regards

[racing_student]

Is there a special area where the divergence appears? If yes you could remesh this part and see of there's a influence

[LuckyTran]

Divergence that early means something is very wrong in the mesh or the settings. Hopefully your mesh quality is okay... Then check your boundary conditions. And then check your initial conditions.

[Leonardo.flores]

If it is transient you might have a high time step. Also, coupled might be very expensive, did you try SIMPLE? Also try lower underrelaxation factors at the beginning.

[Yahia Shannan]

Originally Posted by racing_student

Is there a special area where the divergence appears? If yes you could remesh this part and see of there's a influence

Actually the mesh quality is pretty good. Very fine near the wall in order to capture the boundary layer very well.

[Yahia Shannan]

Originally Posted by LuckyTran

Divergence that early means something is very wrong in the mesh or the settings. Hopefully your mesh quality is okay... Then check your boundary conditions. And then check your initial conditions.

Mesh was done in ICEM CFD, minimum and maximum elements quality is 0.7 and 1.00 respectively. The mesh is very fine near the wall to capture the boundary layer. Boundary conditions were set according to an experiment with pressure inlet at the inlet and mass flow inlet at the outlet !

[Yahia Shannan]

Originally Posted by Leonardo.flores

If it is transient you might have a high time step. Also, coupled might be very expensive, did you try SIMPLE? Also try lower underrelaxation factors at the beginning.

It's steady state. And since it's compressible internal flow coupled scheme was incorporated. I'm also using Pseudo transient to speed up the convergence but still temperature divergence appears !

[LuckyTran]

Originally Posted by Yahia Shannan

It's steady state. And since it's compressible internal flow coupled scheme was incorporated. I'm also using Pseudo transient to speed up the convergence but still temperature divergence appears !

Don't try to optimize anything for speed or physics. Divergence at 13 iterations means it's a setup problem. Try simple and don't use pseudo transient. Start with robust settings. If your boundary conditions are not blatantly wrong, it's probably just a poor initial guess in the initialization. For compressible flows, it's usually because of the pressure & temperature fields are initialized to values that are inconsistent with the velocity. Ease your way into the solution.

[Jiricbeng]

Generally, I ve got good experience with:
- Pressure based solver - pressure velocity coupling. In most cases it turned out better than "simple".

- Initialize for small velocity
- Use small Courant number (1 - 2)
- First order upwind, then switch to second.
- Double precision mode (I have experienced a rapid improvement in convergence several times when swtiching from single precision to double).