[kaeran]

Hello everyone,

I got a big problem in convergence while simulating in Fluent.
I have almost tried all steps that has been suggested in a thread about convergence by CENTURION 2011(http://www.cfd-online.com/Forums/flu...nvergence.html). But I couldn't solve my problem. Let me explain it first.

Inorder to run a transient simulation, initial converged inputs are essential. So, I ran a steady state pressure based simulation.
Convection and conduction processes takes place in quenching. So,obviously energy and turbulent model (k-w) were switched on.
I used pressure based coupled solver with pseudo transient method to speed up the convergence.
The discretization schemes were (default),
Pressure- Standard
Momentum- second order upwind
Energy- second order upwind

I have tried Higher order term relaxation with default (0.75) relaxation factor and also 0.25 (in another simulation). The simulation ran for 5000 iterations with pseudo time step of 1e-05.

The result was not so convincing i.e., the solution was not converged.
At the end of the simulation, my conservative terms (residuals)were of the order
continuity e-03
momentum e-05 in x direction
e-04 in y direction
e-05 in z-direction respectively
energy e-17
mass imbalance e+02

These residuals keeps on oscillating and I couldn't standardise the solution.I have also cross-checked the mesh. It has decent orthogonal quality and skewness (non-conformal mesh).

Minimum orthogonal quality is 0.20
Max. Skewness is 0.85
Max aspect ratio is 45

but, these values are accepted as the geometry of the CAD model is relatively complex.

And my question here is , should I try for more iterations or should I further reduce relaxation terms?

If there are any other suggestions, it is always welcome.
And please correct me, if the steps that I mentioned here are inappropriate.

I have been struggling with this problem for more than 2-weeks. I really need some help

[bluenail]

First of all,

You don.t need to use coupled solver or other complex method for this

I recommend you to use simple solver and

Decrease the under relaxation factor for the momentum (0.15)

most variable in fluent depends on stability of momentum

and in my experience,

Using the fine and good quality mesh is better than using the stable method for poor mesh

sorry for poor English writing

[kaeran]

Thank you bluenail for your reply.

I tried the steps suggested by you but with a relaxation factor of 0.3 for momentum equation.

From my results, it seems that there are no problems with the momentum equations.
My main problem hangs around with the continuity equation. The residual of continuity equation oscillates between 10^-2 and 10^-3 . The convergence criteria is not met.

I am still trying out by varying relaxation factors and P-V coupling parameters.

[kaeran]

Hello bluenail,

I got my solution converged.
I used Pseudo time transient with a time step of 0.0001 and the higher order relaxation term was set to default value.

But I am not convinced with this steady state result because in Transient condition to get result for 60 seconds I need to run around 600000 iterations. This is unrealizable, as it would take atleast a week to solve in HPCluster.

Is there any posssible ways for me to approach the steady state convergence issue in a different manner?

[bluenail]

Originally Posted by kaeran

Hello bluenail,

I got my solution converged.
I used Pseudo time transient with a time step of 0.0001 and the higher order relaxation term was set to default value.

But I am not convinced with this steady state result because in Transient condition to get result for 60 seconds I need to run around 600000 iterations. This is unrealizable, as it would take atleast a week to solve in HPCluster.

Is there any posssible ways for me to approach the steady state convergence issue in a different manner?

I thought 600000 iteration is available in simulation

Most Converged problem really hard to find solution because there are many factor to affect simulation

So if you find your way to solve converged problem, go on

I hope you could be well

Thanks