Bladeless Fan : Polyhedral Mesh : Courant Convergence Challenges
 

Hello everyone,

Have modeled a blade-less fan in 18.2 with polyhedral mesh.

Original mesh was tetra, but the mesh count was way higher. Hence converted to Polyhedral in Fluent to reduce the count and get the simulation moving forward.

Have setup the model with k-e scalable wall functions; Pressure based solver; Coupled Scheme for Pressure-Velocity coupling.

At the start of the solution, the Courant number was 200.

400 iterations were run before any change was made in the courant number. Each courant number change was 1/2 of the earlier one.

E.g from 200 courant number to 100, then 50 and then 25.

Residuals do come down with the reduction of Courant number till 50, but the moment when it is changed to 25, they go way up.

Have been trying to converge the model for the last 24 hours, but still it is not converged.

For the record, the mesh size is less than 5 million elements, with the skewness ratio of 0.85.

Wondering if there is any "range" of Courant number with which one can play with to get to the required convergence ?

If manipulating the courant number does not help, what other methods do we have to help in convergence of residuals?

Googling talks about pseudo time step (manual), automatic time scale factors, etc. which is only leading to confusion.

Can the experts in here guide with respect to achieve the convergence of the model ?

Awaiting advice

This is a steady case and you are talking about the Flow Courant number setting in the COUPLED P-V scheme yes?

What is wrong with a courant number of 200? If it doesn't diverge on 200, I would keep it there. If residuals are not going down the way you like, then that hints that something else is a problem. Turning down the courant number just to watch residuals go down is fruitless.

Because of the way it works you can crank the Courant number really low all the way down to 1. But you do this only if you have trouble keeping your case stable.

Implicit under-relaxation and the courant number setting is anyway already a pseudo time-step. You don't need to branch out and confuse yourself anymore

Thanks a ton for the guidance.

On Courant number at 200, Steady solution does not even converge to default convergence levels. Surprisingly, it does when the Courant number is gradually reduced :confused:

Another play currently being explored is to reduce the relaxation factors in an orderly fashion to get the data converged. Not sure how far can it be done, but will try that option.

In coupled P-V scheme, Fluent does give the choice of adjusting Courant number. Not really positive about its impact in Steady state explicit option (still a student), but it surprisingly works !

Thanks a lot