I'm simulating a flow trough a pipe with an elbow of D=0.1m ,L= 800mm and Rey=100000. even the residuals convergence reach 10-5 and the min quality mesh= 0.101, the max skewness=0.89848 and the number of nodes 400064. the results show a streamline as a laminar flow and a blue contour of turbulence. I'm lost to capture the turbulence phenomena in my simulation.
Dear Ben Tra,
at first I will recommend you to make some analytical calculations on the Reynolds number, so that you will know for sure, if you need to account for turbulence or not. To capture turbulence you can add a "simple" turbulence model, you should first find a proper one for your case. Here you can find more information, about one of the most used turbulence models, the kappa-epsilon model:
https://www.cfd-online.com/Wiki/Realisable_k-epsilon_model
Best regards,
Antoni Artinov
"Capturing" turbulence means nothing. If you want to solve turbulence (and in this sense capturing all features of turbulence in your problem) you must use a DNS.
Differently, you need to accept some type of modelling either in RANS or LES formulation.
Thus, what are you using as formulation? From the number of nodes I suspect you are doing something like a 2D RANS. This formulation is based on a statistical averaging and consequently the solution appears smooth as it were laminar-like. You cannot capture the unsteady fluctuations. Furthermore, the details of the BL depends also on the type of BC at the wall (resolved or modelled BC.).
Thus, I suggest to provide the details of your simulation.
thanks for all , I am simulating a flow through a circular duct (D = 100mm, L = 25D) Which contains 2 elbow 90 ° and a flow conditioner. I started the 3D simulation under fluent Ansys with the model of turbulence realizable K-epsilon and the enhanced wall function. The boundary conditions are velocity inlet (15m / s) and pressure outlet.
in response to sir Antoni Artinov , the Reynolds number calculation in done by experiment work.
What is the problem in the results? Do you have some measurements from experiments?
If you have the experiment data , you can first validate the your obtained resulats espicialy at the inlet . Next you can also validate the all zones , you can not see the turbulence in the far zones because the flow is establisshe . If you have not a best validation you must change the turbulence model .
I will agree with Fillippo that if you are solving with some approach like RANS you will not be able to see the fluctuations. you should compare your results with some of the measured data in that way you will be able to validate your results and simulation.
Dear Ben,
If you're using unsteady simulation you should check your time step size and consequently CFL number. If it's too big you may not be able to model the turbulence.
Dear Ben Tra!
As pointed out by Prof. Filippo Maria Denaro: "Capturing turbulence is nothing"
If you are solving by enabling turbulence model the solver will produce some output irrespective of your mesh or methodology (you might not get good results but for sure you will get some results). What specific quantities you are looking for? It would be helpful to understand the problem if you can post some pics of your results.
Additionally you can check the equations which you are solving in solution control. Check if the equation for turbulence model is unchecked by mistake.
Or in post-processing check some quantities indicating turbulence, these might be TKE, fluctuating velocity components. Streamlines can't be the only indicator of turbulence.
With applied e.g. 2-equation turbulence modeling in a straight pipe there is essentially no turbulence generating mechanism. So the specified turbulence intensity from the inlet cross section (e.g. 10% turbulence intensity) will only attenuate, i.e. the turbulence will dissipate over the pipe length. In a pipe with an elbow and if you are sure, that your Reynolds number is as high as specified, you probably should observe some kind of flow separation/recirculation behind the pipe elbow. Make sure that you use e.g. SST turbulence model, a fine enough grid with a near-wall mesh resolution of approx. y+
- Badges
- Science topic