[rcp]

Hello, this is my first thread. I hope this post did not break any rule here.

I have read some publications and discussion on forum about the recommendation to extend the inlet and outlet boundary, and by doing so we can:

1. force the convergence.
2. avoid recirculation which can lead to reverse flow (mostly at the outlet).
3. ensure the flow is fully developed as it enters and exits the interfaces.

My question is, if I apply velocity flow rate at inlet and static pressure at outlet of centrifugal pump, do I still need to create extended tube at both the inlet and outlet?

Also, what do we expect in the results (head, flow behavior) if we just, lets say, ignore the recommendation for the sake of saving computational time? Does every case need this extended outlet/inlet?

Thank you very much for your attention.

[Gert-Jan]

If you want to calculate the performance as if the pump is on a test rig, then you can ignore most of the inlet and outlet and evaluate the massflow average total pressures on both.
If you want to test it in real life situations, then I don't know. Mostly, problems occur at maldistribution of the velocity profile at the inlet. Then you start at a position where you are sure about the conditions or are convinced that it has no influence.

[karachun]

Why not determine this by yourself?
You can run simulation without additional inlet/outlet extensions, then extend them into 5/10/20/40 diameters and compare results. if results are close then you don’t need to extensions, otherwise you should use them.
There are many different variants: sometimes result is sensitive to extended inlet/outlet (wrong friction results, bad convergence) and sometimes – not.
In general devices with high friction resistance compared to resistance of shape (like Venturi tube) are sensitive to entrance length, to obtain accurate results they must be in fully developed region in simulation. Devices with high shape resistance (like valves) are insensitive to velocity distribution at inlet.
Extended tubes can not be longer than real pipelines connected to device you model.
https://en.wikipedia.org/wiki/Entrance_length

[xudong]

1/7 power law, calc re then vs layer, simple