Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
K-Epsilon Turbulent Mixture Model Homogeneous Flow gives the same results for different inputs
Posted 2011/05/23 6:55 GMT-4 11 Replies
Please login with a confirmed email address before reporting spam
I am using the Mixture Model for Turbulent Flow (using k-Epsilon) and I'm trying to simulate
a situation there are glass spheres dispersed in the fluid and where the slip velocity is 0. It's the option of "Homogeneous Flow".
However when I try to change the Inlet BC or the Schmidt Number in the constants,
the results are always the same.
Also, I can't get the problem to converge with PARDISO, only with MUMPS or SPOOLES solvers.
I get a overestimated pressure drop along the pipe.
I'm comparing the results to published bibliography by Shook.
I have attached the file.
Hope someone can help.
Cheers.
Attachments:
Please login with a confirmed email address before reporting spam
First the change in BCs, I am not sure of what you mean with 'results are the same', if I change the velocity at the inlet or the dispersed fraction I get different results. What may be suspicious is the fact that the dispersed phase is homogeneous over all the domain, but I think it makes soemwhat sense: you have a case of mass transport by means of a fully turbulent flow (Re is about 10e5) i.e. a very effective flow at trasporting stuff so all in all at steady state you may well have a constant homogeneous distribution of the dispersed phase especially given the fact you have a constant inlet concentration.
Second, solver. I tweaked your model a bit but definitely I could get convergence only for a MUMPS solver, I am sorry I don't have much time now ti think of it, if anything comes to my mind I will let you know.
Third pressure drop. The pressure drop I get is perfectly fine, you get something int he order of 0.07 bar for a 3m/s inlet velocity over a 4 meters (!!!) pipe. Again, this is perfectly acceptable. If I reduce the inlet velocity to 0.9 m/s I get a drop of about 0.02 bar, which is fine again. Can you give the details of your source?
Cheers
Please login with a confirmed email address before reporting spam
first of all, thank you for your answer.
What I mean is that by changing the turbulence length scale and intensity I always get the same pressure drop,
and I assume it should be different.
My source is the article by Shook : "Experiments with concentrated slurries of Particles with Densisties near that of
the carrier fluid".
I am trying to simulate the case for fine particle with 34% solids.
The author claims that at that concentration and speed, there is no formation of slurry.
The particles are all fluidized or dispersed.
And I am expecting a homogeneous flow, but the pressure drop does not match with the experiments.
Roughly, with that concentration and bulk velocity of 2 m/s, the headloss is about 0.08 (m/m) and the pipe
is 2.9 m long of fully developed flow.
I have the first section just to introduce the fully developed flow in the second one.
The pressure drop that interest me is for the 2.9 m section, as you have already perceived.
As for the solver, I get convergence with SPOOLES and MUMPS...probably a problem with my meshing??
Hope to ear from you again.
Thank you again.
Cheers.
Please login with a confirmed email address before reporting spam
First of all turbulence length and intensity are 'artificial parameters' used to find a solution for a turbulent flow, i.e. they do not exactly correspond to a physical phenomenon so if you change them you may get non-convergence but it is quite difficult you get 'different' results, that would cast an ominous light on your model.
Second, if teh author claims no slurry then I would expect to have a homogenous dispersion with no high concentration anywhere in the pipe, so I would say the results make sense and agree with literature. Do you have the details of the publication? Journal, volume, issue etc.?
Third, why do you use headloss? That is going to make things much more complicated for you, in experiments you measure pressure drop by means of pressure gauges. Unless your pipe is sloping downwards and the flow is mostly gravity driven, but then you need to model a 3D pipe and you need to include voluem forces! Have you converted the headloss m/m in bar?
Fourth, I have changed the mesh but Pardiso did not work either. With my mesh I get both delta_wall_plus and delta_wall perfectly in range so I would say that the fluidics is right here.
Cheers
Please login with a confirmed email address before reporting spam
I'm using headloss because it's used in the article.
I have converted it to Pa: so 0.08 (m/m) for a 2.9 m pipe will result in approximately 2275.6 Pa or 0.02 bar.
So there is still a difference between my simulated results and the experimental data.
The turbulence modulation should not be important for the inlet BC and the results in general?
Could you provide a little more information about the the 3D and volume forces? Why do you say that I should do 3D?
Isn't a 2D axisymmetrical enough for pipe flow?
Also, I didn't quite understand the fourth item.
Thank you again.
Cheers.
Please login with a confirmed email address before reporting spam
First of all we are talking about 0.02 against a drop of about 0.026 (with Comsol) I would say the results match perfectly especially since the study you rely upon is old and you are not sure about their instrumentation error. Therefore you do not need volume forces nor 3D sims.
Turbulence parameters are crucial to get a convergent numerical solution, but if you change them you may simply get divergence and no-solution at all. You need to read some literature about turbulence, you may begin with Comsol documentation, then the best high-level introductory book may be Frank White 'Viscous Fluid Flow'.
The same goes for the wall distance, start with the Comsol manual about the wall lift-off and then read some literature like the White book.
Cheers
Please login with a confirmed email address before reporting spam
Going to try to find some recent studies about suspension flows.
I will consult the book like you suggested.
I just have one final question: is it possible to input a custom Wall Function in COMSOL?
Thank you very much for your patience and kindness Amir.
All the best.
Rui
Please login with a confirmed email address before reporting spam
...
I just have one final question: is it possible to input a custom Wall Function in COMSOL?
...
Ay! There's the rub!
Theoretically it may be possible and that is something I haev been trying; however, I would recommend you first study a little bit more about turbulence and modelling of turbulence. To find a custom wall function is complicated to say the least for several reasons from issues with the solver itself to being sure it is physical enough etc etc etc...
Cheers
Please login with a confirmed email address before reporting spam
With "To find a custom wall function is complicated" you mean to rewrite any equation in comsol?
Could you tell me if it's possible to change the number for y+ ( in comsol it's called delta+ )?
I am studying a pipe turbulent flow but I don't know my current value of yplus, but I would also like to change the first grid point to log-law region.
Thank you
Please login with a confirmed email address before reporting spam
Not exactly. Wall functions are not just equations/functions, in a very simplified way you can think of them as a method to approximate the turbulent flow as it gets closer to the no-slip boundary wall.
y+ or delta+ cannot be defined at leisure, its value is a consequence of the fact that turbulence modelign requires some thinking and some approximating.
One good book which starts explaining a bit about y+ etc. is Frank White 'Viscosu fluid flow', the best book on boundary layer theory for turbulent flows (I think) is Joseph Shetz 'Foundations of boundary layer theory for momentum, heat and mass transfer'.
I am not sure about your last remark but you can change the thickness of the first layer in the boundary layer mesh, you can find it under Settings.
Cheers
Please login with a confirmed email address before reporting spam
I will take a look at the books you recommended.
Actually, I thought it was possible to change the y+ value in comsol. With this, I could set the first grid point in the log-law region. Now that you explained, I will try the other way.
Do you know how to get the y+ values after study?
Thanks for your patience!
Please login with a confirmed email address before reporting spam
In any 3D plot add a Surface plot or just select the already existing Surface plot, then in the expression field select the double arrows on the right and under turbulent flow select wall lift-off viscous unit and you are done.
Cheers
Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.
Suggested Content
- BLOG Modeling Approaches in Heterogeneous Catalysis
- KNOWLEDGE BASE Solving time dependent models with inconsistent initial values
- KNOWLEDGE BASE Computing Time and Space Integrals
- FORUM Differences in the results of the same model in 3D and 2D.
- KNOWLEDGE BASE Avoiding Negative Concentrations