Hi,

I am using the Fluent 6.1 to model the flow and heat transfer around a roughed cylinder. The diamteter of cylinder is 0.15m,roughness height is 1.35mm,and Reynolds number=2e6 .There are experimental data under those conditions. I care more about the front part of cylinder,so I use the steady state sover(segeregated, SIMPLE algorithm).

I have tried the k-e,k-w,sst,and SA turbulence model,but I found only the S-A model can give an result which has the same trend as the experimental data. The others give a much lager Nusselt Number in the staganation region. Though the S-A results are more satisfying in the trend, the difference to the experimental data is still very large. Also under the low Renolds number(2*e5),it can't give the transitional zone.

I don't know if it is because of the grid and the settings of roughness. In the user mannual, it is suggested that distance from the centroid of the first cell to the wall should be larger than the roughness height. That means the distance should be about 0.000675m in my case. I tried that only to find the covergence was difficult uless i used the very small relaxation fators. Also I think maybe it is not suitable to have a so large distance, because the distance of the second cell to the wall can't be large in order to capture the bounday layer. So the grid is not smooth(it is usually more dense near the wall as we all known).

I don't know how to deal with this problem. Anybody has this kind of experience? Please give me some suggestions, I'll appreciate any response! Please help me!Thanks!! (By the way,I have seached the fomer discussions in this form, and found some discussions on this issue,but I still can't have the idea.so ask for help here! thanks )

Maximus

With this sort of geometry (a bluff body) I wouldn't try a simulation with the Spallart-Almaras model, because it will be present quite strong adverse pressures and recirculations behind the cylinder: SA performs better with streamlined bodies and it is not your case. I would try it with RNG or, better, the Realizable models with non-eq wall functions.

If your computational resources are good enough then you may try the RSM with a grid of YPlus~1 for the best results for sure.

Hope this helps. Good luck!

Freeman

Hi,Freeman,thanks for ur response.

I have tried the Realizable models with non-eq wall functions,but I found it is not satisfying especially in the staganation zone where it gives larger results than the experiment data(Nusselt number). I don't know if it is the drawback of k-e model. Maybe it's the drawback of all the turbulence models,that the transitonal flow can't be well predicted.

In my project,I do deal with the streamlined bodies(airfoil with roughened surface).I compute the cross flow around a cylinder in order to validate the turbulence models for roughened surface.So I only compare the results of the front part of cylinder.

But I found only the SA can give a similar trend as the experimental data(only under high Reynolds number). Maybe it is because of the natural transition of the SA model in the transitional zone. but it give a lower results than the experimental data, more than 30% difference in some zone.

I have also tried the RSM,it seems the results are not as good as SA model.Maybe I need to refine the grid. It is surprising that it can give a better results under the low Reynolds number to use the the boudary layer method to deal with the roughened surface. But it can't give better results under high Reynolds number.

I am really confused. I am wonder if turbulence models can give well heat transfer predictions on roughened surfaces.

(By the way, I have used my own code to predict the heat transfer coefficients, but found it is not satisfying. I have thought maybe there is something wrong in my code. So I use Fluent for the same prediction, only to get the similar unsatisfying results.)

So expecting some suggestions!

thanks a lot!

Maximus