I am new to CFX and was hoping that someone may be able to give me some pointers. I am trying to numerically model a fan which has been tested experimentally in a circular duct. I have used the exact CAD prototype and imported into CFX and then set up the numerical model to repeat the experimental set up.

The model consists of three parts. A cylinder which I have modelled as an opening to represent the inlet of air from the surroundings in the room, which is stationary. The second part is the fan cut out of the duct in which the fan is rotating and the tube it is cut out from, is counter-rotating. The final is the main outlet tube with a mass flow outlet on the outlet of the tube. You can see the set up here: http://www.dezignit.com/setup.jpg

I have input the rotational speed from the experimental results and the mass flow, calculated from the dynamic pressure measured in the experiment. I am modelling as a steady state solution and using a time step of 0.03 (~1/rotational speed).

I was having a lot of trouble with the mesh with negative parameter errors as I had too many controls on the edges and faces of the fan itself. I've stripped the mesh to be simple, but now I have very large jumps at the two interfaces (see http://www.dezignit.com/interfaces.jpg) which I think may be the problem as I can't get the solution to converge after many hours of running. The residuals don't go below 10e-3, except for the mass which dives down to 0 almost straight away.

I would really appreciate any suggestions anyone may have on what may be the problem. How do I get a better transition between the meshes of the three distinct volumes at the interfaces (I have used GGIs).

Thanks in advance for any help/suggestions anyone can give me.

Jenny

I forgot to say in my previous post that I get the error:

Volume elements with aspect ratio(s) > 50 are present in the mesh.

Which is why I am wondering about the change in tet size between the two interfaces.

Dear Jennny,

u r going in the right direction for modelling but as far as the mesh is concerned why u r not using icem? U can mesh 3 different domains with two stationary domains using hex and rotating domain with tet. Connect using Frozen rotor. Check your mesh quality before importing in Pre. Hope this will do.

try total pressure inlet.

Regards,

Pankaj.

I think you'd better use gridgen or Icem for the process of grid generation and try to generate high quality hexa grids. Unstructrured grids is apt to cause problems, and mind the skewness of your grids rather than the ratio.

hope that you can overcome the problem

Is the model non axi-symmetric or could you model a single blade passage? If you model a single passage, you could improve the mesh in each domain and have better resolution,etc. Like Pankaj said, if you try the Ptotal Inlet it should help. Using a static pressure outlet would allow you to see what mass flow rate you are getting and then you could vary the mass flow in further runs. There is a good example of this in the tutorials.

Good Luck,

Erich

Hi Erich,

Thank you for your reply. I am not sure about the blade interaction and since the model is only small (80mm diameter), I am OK with modelling the whole thing. It is a unique design and would be difficult to split into periodic boundary conditions also.

I was actually at a CFD course this week and have more understanding how the program works, and will try everyone's suggestions.

Thank you again.

Best wishes, Jenny

Hi Pankaj,

Thank you for your reply. Yes, I also wanted to use swept meshing for the two stationary domains, but unfortunately due to the imprinting of faces and insertion of straighteners in the outlet I can't sweep. I have meshed the inlet, rotor and outlet separately all with CFX-mesh and then joined together with a GGI using the Frozen rotor method.

I am looking at the mass outlet as I varied this by a factor of ten and got very quick convergence, so I will try the Pressure in and outlet as Erich suggested and see if this yields a better result.

Best wishes, Jenny

Thanks for your reply. I'm looking into refining the mesh as it's still quite coarse. I was having problems of different mesh size on both sides of the GGI between the rotating fan and the stationary outlet. I have now set these as constant so the tet elements don't have the big leap in size across the interface, but this still isn't helping. I'm having a look at other options including the boundary conditions on the inlet and outlet.

Best wishes, Jenny

I suggest you could try Upwind scheme first.