CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Model geometry (

oduor June 11, 2002 07:27

Model geometry
Dear all, I am modelling the hydrodynamics of a Rushton turbine stirred mixing vessel (diameter, d=4m) with CFX4.4 on a Windows NT platform. My fluid is water only, at ambient conditions.

-I use the Luo's (1972) steady (or quasi steady state) to model the impeller, ie the inner block moves with the impeller while the outer block is stationery with the baffles.

-I take the model units in Build to be meters and calculate the Re no. as such. Need less to say that if I increase the the size of the vessel (by increasing the units) the Re no increases. - If the the dimensions are less than 1 unit ( I call it 1 m)I get good convergence (less than 10^-7), however, when the diameter is more than 1m, the best convergence I can get is upto 10^-3. Put in other words, there is good convergence for my Re<10^4.

-Why does this happen? - Does this have something to do with the imperical constant of the k-eps ?

Thanks, for taking your time to read this.


Kuba June 11, 2002 08:28

Re: Model geometry
Hi, I guess that you are running both simulations with the same impeller speed so that you get different Re. Are you sure that you are in the turbulence regime with the samll vessel. With silding mesh you need a large number of revolutions. How many have you done so far? Also try small time steps and larger number of iterations per time step.

Kuba b.

oduor June 11, 2002 08:44

Re: Model geometry

oduor June 11, 2002 09:00

Re: Model geometry
Hi Kuba, Yes. With same speed, different geometry.

Of interest is the fact that in the Laminar regime (Re<10^3) K-eps turbulence model model predicts the flows as obtained experimentally.

Note that impeller Re for turbulence regime is given differntly by different authors, as for me, I take it that at impeller Re >10^4 I am in the turbulent regime.

I try to avoid laminarization of the turbulence flow by operating at not too high Re no. I feel 100 RPM with 0.378M impeller diameter should give me a turbulence flow. Please correct me if Iam wrong!!

thanks oduor

Kuba June 11, 2002 11:36

Re: Model geometry
Yes with these condiitons if D=0.378m you are in the turbulent flow Re is around 2.3*10+4.

I repeat my questions,

How many revolutons of the impeller did you simulate and with what time steps?

For your larger geometry did you incerase the mesh?

Kuba b.

oduor June 11, 2002 12:51

Re: Model geometry
Hi Kuba, Sorry, I do not get your question quite well, I thought you meant RPM. At what stage (whre?) do you set the number of revolutions? I am still, new in this field!.

I just defined the RPM and I used rotating coordinates for my "steady" state modelling, quite similar to the 'application example in CFX4'.

thanks oduor

Kuba June 11, 2002 13:03

Re: Model geometry

Lats get this sorted. Are you using 'Sliding mesh' approach to simulate your impeleler or the MFR 'Multiple Frame of reference'. From your first post You described the method of 'sliding mesh' where you set a timme step to model the impeller revolution.

In the sliding mesh you set the RPM, the number of revolutions of your impeller, the angle of your time step and the number of interations per time step.

To check this just tell me how many iterations have you set in your commnad file. If it is more than 100 iterations then you are most likely using the MFR method and the number of revolutions is less important.

P.S. (Luo 1972)??? I have not heard of this paper. But there are papers by Luo in 1993 and 1994 describing the 'sliding mesh' approach and the MFR approach. COuld you give me the title of the paper?

oduor June 12, 2002 08:00

Re: Model geometry
Kubo, You are right, I am using MFR, sorry I did not mention this earlier, explicitly. Sliding mesh is transient, whilist Luo 1994 (by mistake I wrote 1972) assumes a steady flow for what is actually a transient flow. That is the approach i use.

Again, on iteration, it is true i use more than 100, infact always more than 1300 for a reasonable result.

Infact I tried the transient flow but i got very wild divergence, and I abandoned it for I while. I will soon go back to it but we shall addres the problems i encountered with transient flow in a later posting.

Thanks for your petience, Kuba.


Kuba June 12, 2002 09:27

Re: Model geometry

I am not an expert on MFR (I use sliding mesh) but what we found is that when using MFR you need to run the simulation for at least 10,000 iterations. The due to teh geomety modelled you might find that the solution will take much longer to get to lower residuals.

I would suggest to run the simulation for more iterations.


oduor June 12, 2002 11:52

Re: Model geometry
Hi Kuba,

Thanks, Kuba, for taking your time to respond to all my quetions.

Is there any anybody reading this, and is working with MFR, please help.


All times are GMT -4. The time now is 04:24.