GGI interface problem (CFX-5.7.1)
 

I have problems modelling a propeller with the sliding mesh (transient rotor/stator) capability in CFX-5.7.1. When viewing the results of the simulation, there seem to be an inconsistency between the velocities and the turbulence quantities (k-e) across the domain interfaces. A big jump from large velocities/turbulence quantities inside the propeller domain to much smaller values in the tank domain. When doing transient computations I have no problem reaching convergence of 1E-5 (RMS) within 15 iterations in each timestep (t = 0.1 sec. and the propeller turns with 35 revolutions per minute).

A part from transient simulations, I have also tried a steady-state frozen rotor computation, and after 1000 iterations with an auto time step of approx. 0.02 sec. I still get the inconsistency/big jump across the interface.

I should probably say that I am modelling the complete propeller geometry (360 degrees), and I have enclosed the impeller in a cylinder with larger diameter than the propeller.

Has anyone in the forum experienced the same kind of inconsistencies across GGI interfaces?

Regards Jesper

Re: GGI interface problem (CFX-5.7.1)
 

Hi Jesper,

Are you viewing the Velocity variable or Velocity in STN Frame? Note that "Velocity" is the relative velocity. Inside the rotating domain, it is relative to the domain rotation.

Regards, Robin

Re: GGI interface problem (CFX-5.7.1)
 

Hi Jesper,

Yes I already had some inconsistencies across GGI interfaces, especially for the derivative of the quantities. But I was running my simulation in laminar flow at low Reynolds number. What's your conditions? Regards,

Thomas

Re: GGI interface problem (CFX-5.7.1)
 

Hi Robin, thank you for the response. When viewing the velocities in STN Frame, the difference is in fact small. I had previously read about the difference in Velocity and Velocity in STN Frame in the documentation, but forgot about it my search for the inconsistencies in turbulence quantities. So, sorry for that...

The main problem is with the turbulence quantities (k and epsilon). I am curious about why there could be a rather big jump across the interface when I look upon these two quantities. When plotting a contour plot of e.g. k intersecting both the impeller domain and the tank domain, I get rather large discontinuities in the contours when crossing the interface. Making the same plot with Velocity in STN Frame, makes continuous contours across the interface. Is there some obvious reason for this, that I have somehow missed? I first suspected bad convergence, but I actually have quite good residual plots. And also the grid has quite equal size distribution across the interface.

To Thomas: I have turbulent conditions. Reynolds number in the tank (a circular channel) is approx. 2E6, and the impeller Reynolds number is of the same magnitude.

Regards Jesper