CFD Online Discussion Forums

CFD Online Discussion Forums (
-   STAR-CD (
-   -   Fully developed flow (

crevoise December 16, 2011 04:40

Fully developed flow
2 Attachment(s)

I am computing a flow trough a pipe of diameter 50mm, and 700mm length. The mesh is around 300000 cells, with a finer mesh near the wall.
- On the inlet, my boundaries are the following: velocity: 25m/s and T=573K, turbulence intensity and length: 0.04 and 0.004.
- On the outlet, I apply a pressure boundary (piezo, zero grad for temperature)
- The walls are adiabatic, no slip.
- I am using K/epsi Hight Reynolds number scheme
- initial conditions: v:25m/s, T=573K

In steady state computation, I am reaching the convergence criteria (0.0001) around 260 iterations.
However, the velocity profile is not good, and I have no clue why(see residuals and velocity profile in attached).
If you have some idea about which mistake I have done, I will be glad to learn it. I guess the solution should be obvious for an experienced cfd, but...

Thanks a lot

k_k December 16, 2011 05:14

if your objective is to compute the fully developed turbulent velocity profile, use periodic boundaries on both ends of the pipe and specify the mass flow rate, turbulence conditions, pressure gradient across the pipe as inputs. you should get fully developed velocity profile on both ends of the pipe when convergence is reached.

crevoise December 16, 2011 06:12

Thanks for your reply and the tip. I will try it.

However, my aim is to get a fully developed flow before starting some injection and reactions. Thus, my scalars will not have to be equal on the two boundaries faces, which is problematic in the idea of cyclic boundaries.
I can still first compute the fully developed flow with cyclic boundaries; and for the injection, restart the computation from the developed field, with different boundary conditions.

In the idea, I wanted to avoid a restart, so then, do not use a cyclic boundary. Do you think it is possible, or is it just a loss of time from my side to fight with this?

Thank you.

crevoise December 20, 2011 10:50

Hello again,

As advised k_k, I am running the simulation with cyclic boundaries.
On the inlet cyclic boundary, I am setting an inlet flow of 0.03kg/s, with a bulk temperature of 573.15K. On the inlet, I let the cyclic boundary with an 'all match' option.
My aim is to get a fully developed flow, with 573.15K and 25m/s, to match a case made on a 1D model (on the 1D model, the starting flow is already a fully developed flow, while StarCD gives a plug flow on the inlet)

So far, I am getting the fully developed flow, but with an average inlet temperature of 585K and an average inlet velocity of 26m/s.
My question is about the temperature: while setting a bulk temperature to 573.15K, I would have expected to get a 573.15K on my inlet as a result of the computation. So how to explain the 585K?

Thank you for your light on this point.

All times are GMT -4. The time now is 18:05.