[surajp92]

I am trying to model ducted Ka 4 70 Propeller in open water. Instead of modeling the whole propeller I am modeling only quarter domain with one blade and quarter of the nozzle. I am using below physical model:

Constant density
Steady State
SST K-w turbulence model
Gamma transitional model

I am using velocity inlet and pressure outlet boundary condition. Also I am using periodic rotational boundary condition for two sides of the domain. I have two domain rotating and stationary. Rotating domain is surrounding only propeller and stationary domain is outside rotating domain. And I am applying moving reference frame for rotating domain and stationary domain has lab reference frame. I also created the internal interface at the contact surface.

I am having some trouble with the convergence of residuals to very small value. After few iterations all residuals remains almost constant. My velocity component has converged up to 10-4, Continuity up to 10-3 and k-e up to 10-2. I think K-e residuals attain constant value and after that rest of the residual are attaining constant value. So how can I get k-e residual to get converged to smaller value? I changed the turbulent intensity at inlet and outlet to 0.001.

Also, my thrust of the propeller, duct and propeller torque becomes constant once the residuals attain constant value.

Thanking you.

[me3840]

With your current physics and mesh resolution, that's the amount of error you can reduce. If you want to change it you will have to change the mesh or resolve the physics better.

[surajp92]

I tried using Transitional Gamma model also for capturing the transitional effects. But the same thing happened and residuals converged only up to 10-3 to 10-5. Can you recommend some changes I should make to make my residuals converged to lower value?

I am thinking of using unsteady analysis with sliding mesh but I think for this I would have to model full domain and not just quarter domain.

thank you.

[me3840]

Honestly I think the quarter domain and steady-state are the two biggest problems. Those are pretty poor assumptions for a flow and geometry like this. Steady state with the full domain would be better.