CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   User-defined k or epsilon? (https://www.cfd-online.com/Forums/fluent/69115-user-defined-k-epsilon.html)

tstorm October 12, 2009 14:40

User-defined k or epsilon?
 
Has anyone written successful UDF source equations for turbulence kinetic energy (k) or turbulence dissipation rate (epsilon)?

I've tried writing mine based on the k-e equations here (http://www.cfd-online.com/Wiki/Standard_k-epsilon_model) but velocity profiles in the boundary layer don't match the Blaussius curve.

Thanks!

jalarron October 12, 2009 21:52

Currently I'm implementing a splitting technique for k-e model, so I have to add two k-e extra equations. I've done it successfully.
Are you trying to modify existing Fluent k-e equations? or you are trying to add k-e from scratch?
Cheers,
Javier

tstorm October 12, 2009 23:13

I'm trying to start from scratch. I'm making several v2-f models, which use the standard k and e equations.
Would you mind posting your k and e source code? I'd appreciate it.
Thanks!

jalarron October 13, 2009 00:47

Which viscous model are you using for your implementation from scratch of k-e? laminar model + 2 UDS equation for k-e?
About the source code, I have to check first with my supervisor, just for IP reasons. I think it shouldn't be a problem, because is a method already published and I'm just implementing it in FLUENT.
Currently I haven't tried to start from scratch but it is one option that we were thinking because patching the in-built k-e it's a bit tricky specially in the near-wall region.
Cheers,
Javier

tstorm October 13, 2009 02:11

I'm actually using the k-e model with turbulence equations turned off. I hadn't considered running it with the laminar model. I might try that tomorrow, it may solve some problems.

Just to give you some background, the v2-f model is based on the k-e model, and uses the same k and e equations, with some minor differences. The constants are calibrated differently - in particular, Ce2 is a function of turbulent quantities, and isn't constant. Also, the model is integratable to solid walls and doesn't use damping functions or wall functions. So, FLUENT's k and e equations won't work.

I hope you can share your code - I can't imagine that the k and e sources would be proprietary, but I understand if you can't share them.

Thanks!

tstorm October 13, 2009 12:35

I just remembered why I haven't used the laminar model with 4 UDSs (k, e, v2, f). I need to use a user-defined turbulent viscosity, which is made easy when you use the k-e model with turbulence equations turned off.

I'm assuming if you plan to start from scratch with k and e UDSs you'll need to store turbulent viscosity as well. Had you figured out how you were going to hook that to the solver?

jalarron October 13, 2009 22:21

I haven't tried anything with the laminar option, but I think that you could store the turbulent viscosity in a UDM variable, and updated in every iteration.
I'm sending you a document of what I did through a private message. OK?
Cheers,
Javier

tstorm October 14, 2009 12:55

You could store mu_t as a UDM, but I don't see how you'd hook it to the solver in the laminar model. The laminar model should have mu_t = 0. If you figure that out I'd love to hear how it turns out.


All times are GMT -4. The time now is 04:01.