[dileeps]

Hi,

I am trying to simulate a centrifugal pump to generate head capacity curve and compare it with experimental results. My simulation results of head values much differ from the test data.

Pump specs:

Duty point discharge - 2 litres per second

Head- 21 m

Speed-2800 rpm

Impeller dia - 141 mm

Cfd gives much higher values of head in higher flows.

This pump is supposed to give 0 m head at around 3.5 lps flow, but instead gives about 19 metres of head.

I used SST & K-E models, used wall roughness, but not much advantage. Hope anyone may help me resolve this issue.

[ghorrocks]

Have you read the FAQ: https://www.cfd-online.com/Wiki/Ansy..._inaccurate.3F

[dileeps]

Originally Posted by ghorrocks

Have you read the FAQ: https://www.cfd-online.com/Wiki/Ansy..._inaccurate.3F

Hi, thanks a lot for your help . I am not experienced enough to understand everything mentioned in the article. But still i got many areas to look for correction from the list.
Any way, here are some findings of my own from a rookie perspective. Kindly let me know if any clues are revealed.
- The pressure rise within the impeller itself seems very high (from streamlines). Normally the pressure rise should decrease as the mass flow rate increases. But in my case, head capacity curve remains almost flat for a range of flow rates.
- I have tried several pump models, and this trend is visible in most of the models, but in varying degrees. But i have got satisfactory results in few models with the same setup.
- I have set monitors for head in pre, and usually monitor its variations as the iterations progresses. My convergence criteria is rms 10e^-4, and most of the times the head value corresponding to the iterations near rms 10e^-3 appears to be matching to the required value. After that head starts rising to a higher value. Most of the times solver stops with maximum set no of iterations (1000 in my case), and not by residual convergence.
- In the CFX Post, I tried to map y plus contour on blades. The maximum value comes around 300. I had used inflation with 10 layers with maximum thickness of 2 mm (impeller eye dia is 37 mm)


Hope these may help in resolving the problem.

Thank you in advance for your kind help.

[ghorrocks]

Another reference to read: Have a look at the best practices guides in the CFX documentation. A few of them discuss rotating machinery and pump performance curves.

Some comments:
* Check your boundary conditions, material properties are correct. They can cause misleading performance curves.
* You might need to run your pump faster or slower to get the shape of the performance curve you expect.
* The most important things to check for simulation accuracy is convergence, mesh size, time step size (if transient) and boundary proximity. You need to do sensitivity checks for these parameters.