Recently, i am using a pipe flow case to test fluent's calculation. size of the pipe: D =23mm,L=0.1m; boundary condition:velocity inlet1.5m/s;

pressureoutlet:0pa(gauge pressure) i have tried two turbulent models(standart k-e and RSM),but neither leads to the just result(mainly the pressure drop ) corresponds to the empirical formulas(this is a typical case,you can find the relative fomulas in the literatures). i have check the mesh carefully,and also adapt the cells near the wall,but it seems nothing favors the final results .

what is wrong?? is it possible the cassical emprical correlations are wrong? or the fluent model are not exact?

thanks for any suggestion!

Are you sure that your flow is turbulent?

wether it is turbulent or laminar can be easily judged by the Re number,(>2320),surely it is turbulent.

Hm. I get Re=u*dh/vis=1.5*0.023/1.7e-5=2030 i.e. transition if disturbed and laminar if not disturbed.

If you use the length (L=0.1m) to calculate the Re you get a larger number, but I dont think that would be correct. The diameter is the correct "length".

It isn't so easily to judge if a flow is turbulent or not when you are in the transition domain (see your Re number). Increase your velocity inlet to be sure to have a turbulent flow and normally you will find the same results as the theory.

The classical empirical correlations are correct, no doubt about that. Turbulence models are never completely exact, so you always find some deviations. The question is if the deviations are large in your case. Perhaps the Reynolds is low and low Reynolds turbulent flows are quite difficult to predict.

Tom

The classical correlations may be correct, but you have to know how to use them. One has to be very careful to choose the correct correlation for the setup. And as I and HVN have pointed out the flow is in a Re area where it is difficult to determine if it is laminar or turbulent. If WZH uses the wrong correlation for the setup, strange results may be expected.

Also: the pipe length is too small to set up a developed flow. Generally the empirical relations apply for pipes that are longer than 10 times the diameter. So make your pipe at least this long and then postprocess your run for the piece of pipe after that.

the flow in that is turbulent,get the Re=u*dh/vis=1.5*0.023/1.007e-6=34260(you got the wrong viscosity?) , maybe Erwin is right,for a fully develop flow ,the pipe length should be more than 10 times of the tube diameter

ny=1.007e-6 ???

My book ("introduction to heat transfer" by Incropera & DeWitt) says that the kinematic vis. is : ny=11.44e-6 @ T=250, ny=15.89e-6 @ T=300, ny=20.92e-6 @ T=350, and I dont recall WZH mentioned a temperature so I assumed that it was in that range. Where do you find your viscosity ?

I agree that the flow isnt fully developed and that most correlations assume fully developed flow, but it is possible to compensate for this (empirically also). This, of course, adds an additional "error" to the result. But it can be estimated.

And I assume that the fluid is air, btw.

sorry, i have not made it clear that the fluid is water. and when the fluid is water, the Re number is much more than that results from air. anyway,thank you, Christian.

Oh. Water. Hm. I should have thought of that possibility. Guess that I am too accustomed to deal with air

If you decide to follow Erwins suggestion or try to compensate for the inlet conditions in your analytical calculation it would be nice to know about the result.

Good luck.

Hi, or set a fully develloped velocity (and if necessary kand eps-) profile at your inlet! You can do that by an UDF

or

by a translatoric periodic calculation (setting inlet and outlet as periodic). Unfortunately there is a bug in fluent 6.0 with that model, it gives divergence. But with fluent5.5 it works fine.

Volker

I'd try looking at the fanning friction factor at a 'point' in the pipe where the flow is fully developed. Get that agreeing with empirical then start to consider the accuracy of the developing inlet section.