CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   NPSH3% simulation with ANSYS CFX (https://www.cfd-online.com/Forums/cfx/88071-npsh3-simulation-ansys-cfx.html)

ragazzaccio79 May 7, 2011 11:03

NPSH3% simulation with ANSYS CFX
 
Dear all,

I am trying to simulate with CFX the NPSH head drop curve for an end suction pump working at its BEP point at maximum diameter.

I have done a steady state simulation using the Rayleigh-Plasset cavitation model. Previously, I have done all the simulations needed to find out the whole Flow-Head performance curve (always running several steady state simulations with the cavitation model disabled and using Froze Rotor algorithm).

After the real physical test of the pump, I have discovered that simulated performance flow-head curve is in line with the test (differences of 2%/4% on the whole range of flow are considered acceptable). The problem is for the real tested NPSH head drop curve at BEP flow (Knee curve). The real tested value of NPSH3% (3% of head drop) is 2.5m, while the value found with CFX is 1.15m. Considering that that flow at wich the NPSH3% has been simulated is 50m3/h (its BEP point at maximum diameter) and the pump is running at 2980RPM, the relevant SuctionSpecifSpeed is 16335 (US units), quite unrealistic and anyway not in line with the physical test.

Has anyone discovered same discrepancies between real an simuated values for NPSH3%? Could be because I am running a steady state simulation with Frozen Rotor insetad of a real transient simulation? Do you have some test to suggest?

Thank you very much for your help.

Regards,

ragazzaccio79 May 8, 2011 03:43

No one is familiar with the concept of NPSH3% or no one has any advice for me???

Thanks.

ghorrocks May 8, 2011 08:27

Your initial post was only a few hours ago and it is Sunday - But already we are being chased up for answers, alas.

I assume you have done the checks described here:
http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

If the majority of the performance curve is accurate then obviously your model is pretty good. But it seems the error is down in the region where cavitation is occurring. Doesn't this suggest the cavitation model is not perfect?

What fluid are you using? The cavitation model is probably using constants suitable for pure water.

ragazzaccio79 May 8, 2011 15:04

Thanks Glenn for your prompt reply and very sorry for the hurry.

I am using the homogeneus interphase transfer model, with water vapour and water at 25°C.

I have modeled the pump as it really is: leakages, stuffing box, volute, impeller balancin holes, ring clearances. I don't think that problem is the mesh: i have done a sensitivity analysis and mesh is so fine at the leading edge of blades that results don't change if you increase the number of elements in that area.

I believe that issues could be related to cavitation model (I have used the Rayleigh Plesset model with the setup suggested by the existing cavitation tutorial in CFX).

I am wondering if the problems come because I am running a steady state simulation, but at the moment I don't have enough resources to run a transient simulation.

That is why I am looking for other people to share their experiences about cavitation simulation for pumps and turbomachinery in general.

Thanks again and regards.

ghorrocks May 8, 2011 19:05

A long time ago I did some work on cavitation in power steering valves but that is quite different to what you are doing. I have no experience in calculating NPSH for pumps with cavitation.

My point is that cavitation models are usually tuned to pure water. Things like dissolved air, particles and dissolved stuff all affect the cavitation through things like the nucleation size, inception and recovery. So if your fluid is not pure water you may well need to re-tune the model for your fluid.

zona February 8, 2012 09:38

@ragazzaccio79
Hi, can you plaese hlep me about cavitation setup for pumps, I'l done 15-20 saimulations of pumps and later with the experiments Q-H and Q-eff. curve seemd OK (around 3-7% difference) but I have never done NPHS calculations, can you please whelp me aroun this how do I setup the calculation

ghorrocks February 8, 2012 16:49

There is a best practises guide which describes this which comes with the CFX documentation. Have you read it?

pump_passion July 22, 2012 03:30

Quote:

Originally Posted by ghorrocks (Post 306708)
A long time ago I did some work on cavitation in power steering valves but that is quite different to what you are doing. I have no experience in calculating NPSH for pumps with cavitation.

My point is that cavitation models are usually tuned to pure water. Things like dissolved air, particles and dissolved stuff all affect the cavitation through things like the nucleation size, inception and recovery. So if your fluid is not pure water you may well need to re-tune the model for your fluid.

Glenn,

sorry, could you explain how to set up a simulation with cavitation activated a considering not a pure water but a water with 23ppm of dissolved air?

Many thanks

ghorrocks July 22, 2012 08:20

This is not a straight forward task. You will need to do experiments (or access to experimental results), and tune the model to match the experiments. Alternately search the literature to see how other people have approached it.
Quote:

Originally Posted by pump_passion (Post 372837)
Glenn,

sorry, could you explain how to set up a simulation with cavitation activated a considering not a pure water but a water with 23ppm of dissolved air?

Many thanks


luigi79 October 15, 2012 09:58

hi everybody,

I'm a new member and I hope this is the right thread to post my question.

Someone can help me setting a transient CFX simulation of a pump with cavitation model? I modeled the inlet anulus duct coupled with the whole 4 blade impeller. I set a transient calculation with region of motions specified while the interface is a Transient rotor Stator, with 360°/360° specified pitch angle.
I tried to run with 1/12 omega (omega=2950rpm) for the time step and 20 loop and 15*omega as total time but my results are not real, and also convergence is not satisfactory. what is wrong? the mesh seems good, I have no problem.
What is the best practice for this calculation?

thank's

ghorrocks October 15, 2012 17:35

This FAQ covers some basics:
http://www.cfd-online.com/Wiki/Ansys...gence_criteria

The best practises guide in the documentation has lots of good hints.

"The mesh seems good, I have no problem" - what make you say that? Do you actually know that the mesh is good or are you just guessing? One of the most common problems on the forum is poor mesh quality and resolution and I cannot count the times beginner CFD people have told me "my mesh is good".

luigi79 October 16, 2012 03:25

hi ghorrocks, thank's for the quick reply!

I think that my mesh is good because I ran a transient solution of the whole pump (the mesh of inlet duct and impeller is the same) without cavitation model activated and the solution has a good agreement with the sperimental tests , while the calculation with Raileigh Plesset model turned on shows a cavitation bubble growing in only one impeller vane, and the relative position of this bubble doesn't change with the rotation of the impeller. I tried to lower the timestep till 1/40 omega and i put 10 convergence loop but I abtained the same result.

This is the reason because I guess that this problem could be related to a wrong setting of loop/timesteps/total time when using cavitation model.

Have you any idea?

thank's


All times are GMT -4. The time now is 08:59.