[linlin.geng]

when I use CFX to simulate the cavitation with a homogeneous mixture model, I found the following problem:
I use the following two CEL languages to define the turbulent viscosity of k-eplison model,
1. u_t= rho_L*ke*ke/ed;
2. u_t= rho_M*ke*ke/ed;
where rho_L is the liquid density, rho_M is the density of liquid and vapor mixture.

by comparing with the results obtained with original k-eplison model, I found that the first CEL language provides the same result, and the second one has a relatively big difference.

Does this mean that the turbulent viscosity was calculated with the density of single primary phase?

Thanks for your attention, and thanks for every help and discussion.

[ghorrocks]

The documentation should hopefully define this. Have you checked?

Note that combining turbulence models with multiphase (and especially cavitation multiphase) is beyond the validated range of turbulence models. You will need to be vary careful with results and cannot assume that it will work accurately.

[linlin.geng]

Originally Posted by ghorrocks

The documentation should hopefully define this. Have you checked?

Note that combining turbulence models with multiphase (and especially cavitation multiphase) is beyond the validated range of turbulence models. You will need to be vary careful with results and cannot assume that it will work accurately.

thanks for your answer, here is the statement from CFX Theory help: In homogeneous multiphase flow, bulk turbulence equations are solved which are the same as the single phase equations, except that the mixture density and the mixture viscosity are used.

in my opinion, the mixture viscosity means the mixture dynamic viscosity which is defined as the summation of separate equations of the phase quantities. and for the turbulent viscosity, I did not find any helpful information in Help document.

[ghorrocks]

Your quote from the documentation is what I expected, and it seems to contradict your first post where you said the function where you used the liquid density was more accurate than your function where you used the mixture density.

Have you compared your two simulations with one using the default k-e turbulence equation where it calculates the turbulent viscosity itself?

[linlin.geng]

Originally Posted by ghorrocks

Your quote from the documentation is what I expected, and it seems to contradict your first post where you said the function where you used the liquid density was more accurate than your function where you used the mixture density.

Have you compared your two simulations with one using the default k-e turbulence equation where it calculates the turbulent viscosity itself?

yes, as you mentioned, I have compared the results obtained from that two equations to the default turbulence model.
for K-e model, these three configurations give almost same results. then I code the SST model for another validation. The function 1 (using liquid density) give same result as the default one, but function 2 (using mixture density) provide a totally different result than the default model.

I am confused now.

[ghorrocks]

Now I am confused as well. Your first post said your two k-e based models gave different results, but now you say the k-e results are the same but the SST models are different? Now I have no idea what is going on.

[linlin.geng]

Originally Posted by ghorrocks

Now I am confused as well. Your first post said your two k-e based models gave different results, but now you say the k-e results are the same but the SST models are different? Now I have no idea what is going on.

PLease see the attached Doc file, Please comment on it. Thank you very much

[ghorrocks]

Your conclusion appears to contradict the CFX documentation, so I strongly doubt you are correct.

Have you done a verification and validation on this model before you compare the results? If your simulation is not accurate then you cannot draw any meaningful conclusions from it. So have you done mesh size and convergence tolerance sensitivity checks?

[linlin.geng]

Originally Posted by ghorrocks

Your conclusion appears to contradict the CFX documentation, so I strongly doubt you are correct.

Have you done a verification and validation on this model before you compare the results? If your simulation is not accurate then you cannot draw any meaningful conclusions from it. So have you done mesh size and convergence tolerance sensitivity checks?

Yes, i have done the mesh independence validation, and the convergence is very well, the RMS residual is below to the 10-6. I promise that these three configurations presented in Doc have the same numerical setting and all get a well converged result.

[ghorrocks]

Looks like you need to do some hunting around to find out what is going on. A suggestion is to check the turbulent viscosity is behaving as expected.

I would write your two functions to a variable and write that to a results file, and compare it to a default k-e model and the turbulent viscosity variable included in the results by default. Then you can compare the turbulent viscosities directly.

[linlin.geng]

Originally Posted by ghorrocks

Looks like you need to do some hunting around to find out what is going on. A suggestion is to check the turbulent viscosity is behaving as expected.

I would write your two functions to a variable and write that to a results file, and compare it to a default k-e model and the turbulent viscosity variable included in the results by default. Then you can compare the turbulent viscosities directly.

I am looking forward to your answer. in addition, i have to mention that the viscosity i comapred in Doc file is the variable of water. eddy viscosity computed by three different configurations.
By the way, I found another point. There are two variables: water. eddy viscosity, vapor.eddy viscosity, if i use default setting, the ratio of water.eddy voscosity to vapor.eddy viscosity is 1000, which does make sense. however, when we use CEL language to replace the default turbulence model, the ratio is equal to 1, namely, water eddy viscosity= vapor. eddy viscosity. This is unexpected.

besides, the same investigation was on the SST model, and you can found the convergence situation was totally different. (left: default SST model, middle: SST model defined by CEL with liquid density, right: SST model defined by CEL with mixture density).

i am really confused with the results and the interal handling of CFX.

Thanks for your attention.

[ghorrocks]

If the vapour turbulent viscosity equals the water turbulent viscosity in your CEL functions then something is wrong here. They should be different by a factor of about 1000, just like the default setting. So you need to do some hunting to find why your CEL function is not working as expected - have a look at all the input variables of both functions.

[ErikaG]

Hi!

I also have the same problem.
I find a factor =1 between the turbulent viscosity of the vapor and liquid phase. Did you get any new results? Can you share the CEL expression you used for the Eddy Viscosity in Turbulence modelling?

Thank you in advance!