I am trying to measure the reattachment length in Adverse pressure gradient flows over a forward facing step. Which RANS model predicts reattachment length formed at the step consistent with experimental results? I tried but the results are not consistent. I used diverging channel to create APG at step.
Velocity inlet - 0.8 m/s
Pressure outlet - 0 gauge pressure
Once I used the k-epsilon model with other wall laws different from the classical log law and was able to capture with a quite good agreement the reattachment lenght for a classic experiment conducted by Buice and Eaton (1985). Which turbulence models you have used so far?
I have used 5 RANS models
K-epsilon- standard, realizable, RNG,
K-omega
and T-SST (4 equation)
With increasing APG reattachment length is increasing with K epsilon RNG. which is supposed to decrease
others are not consistent
Could you please explain,,, what do you mean by other wall laws?
I have used enhanced wall treatment y+=1
The classic k-epsilon model is what we call a "high Reynolds model", wheres the k-omega for example is a "low Reynolds model" the terminology here is not necessarily linked with the Reynolds number of the flow, but with its ability in predicting in a consistent way the near wall behavior, which is related with lower values of y+ (which is a kind of Reynolds number based on the distance from the wall and on the friction velocity).
The k-epsilon model is not efficient in capturing the near wall behavior unless extremely refined meshes are used in the near wall region. To avoid this issue people don't simulate the near wall behavior and simply use a wall law to model this near wall behavior. The most used wall law is the classic logarithm law, which neglects pressure gradients inside the turbulent boundary layer. In your case since the boundary layer dettachment is induced by a sudden expansion I believe you were not suposed to expect any problems by using the classic log law.
Have you tried varying your y+ value with the classical k-epsilon model?
I am using 3d model and I once tried using RNG K epsilon, non equilibrium wall functions, y+=30 , but the results were not good. My mesh is very fine near the step 0.3 mm. I cannot go more finer as the number of cells increases and it is beyond computational capability. I did not use 2D model for my case study as turbulence is a 3d phenomenon. Do you recommend 2D?
What wall treatment and model do you recommend for APG flows? As per manual for APG flows non equilibrium wall function is recommended. What y+ value you recommend? My Re= 4800. step height = 6 mm. Free stream velocity = 0.8 m/s. Working fluid is water.
Using y+=1 my first layer height is 0.016 mm.
Well, although turbulence is indeed a 3D phenomenon with multiple lengh scales, the use of a RANS based simulation already filters important informations about the flow. In my understanding when you use a Reynolds Average Navier-Stokes simulation the idea is to obtain mean fields in accordance with those observed in experiments. But always mean fields. We know that vorticity dynamics is intrinsically connected with the production and dissipation of turbulent kinectic energy. For a 2D model, vortex stretching is always null for example. Even though, when you are trying to obtain the average behavior of a given fluid flow, sometimes you are willing to loose some details of the flow in order to obtain primary information with low computational cost.
Velocity profiles, boundary layer dettachment and reattachment, the lenght of a recirculating zone, temperature profiles, and other important informations may be obtained with 2D simulations using simply the classical k-epsilon model. The important thing about this model is how you treat the near wall behavior. Calibrating the values of y+ and paying good attention to the mesh refinement near the wall will lead you to a consistent behavior of the mean fields.
I've worked for 5 years with numerical modeling of turbulent flows using only 2D geometries. I've never worked with a commercial code. Here we developed a Fortran code based on Finite Elements that was basically based on the k-epsilon model with some interesting non-commercial wall laws. I've simulated a great variety of flows over different geometries using 2D meshes and k-epsilon.
I would suggest you to start simulating your flow with a 2D fine mesh using the classical k-epsilon model with your non equilibrium wall function, varying the value of y+ in the range 1
Thank you Mohammed Kamel, Dr. Rafael for your comments, I have few questions:
1. I want to run a simulation with experiments done in free surface water tunnel. Which model you think can predict good reattachment length.
2. I am using turbulent length scale 6mm and turbulent intensity of 5%. If the reattachment length is not close which parameter should i adjust.
3. I am using 0 pressure gauge as outlet boundary condition. However experiment was conducted in close recirculating water tunnel. Is it correct? or should i use outflow as my outlet boundary condition?
Dear Hassan, sorry for the delay. I did not see your question here. We will keep on talking in private.
- Badges
- Science topic