CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Centrifugal Pump Cavitation problem or not. (http://www.cfd-online.com/Forums/cfx/94257-centrifugal-pump-cavitation-problem-not.html)

 ismael.s November 10, 2011 09:10

Centrifugal Pump Cavitation problem or not.

Hello,

I am working in a centrifugal pump simulation, but I am having differences between the numerical and experimental results.

I´ve made 4 designs of centrifugal pump on CFX, 1 is equal my simulation so it's ok, other I have difference in experimental results but it´s ok too, it´s near the real results.

But now the question, I am having problem in 2 simulations, with a big difference between Numerical and Experimental results.

First I thought that would be cavitation problems, so I've made some studies in CFX with multiphase simulation using cavitation, I´ve reduced the cavitation as much as possible, then I made a new prototype, and I still had a big difference between results.

Some numbers to you have an idea of the difference.

- Numerical results with cavitation
1953 m³/h, 42m of head, 344hp and 87.9% hydraulic performance.

- Experimental results
1953 m³/h, 21.5m of head, 324hp and 51.4% hidraulic performance.

About the simulation. I've simulated a periodic impeller and periodic vanned diffuser, then full geometry.

This flowrate is supposed to dont have much cavitation since it's 2/3 of maximum flowrate of the pumps.

The best design point is to the pump works with 2500m³/h and 32m of head, and the maximum flow at 0 head more than 3000 m³/h of flowrate, but the maximum flow in the empirical results is 2100 m³/h at 0m of head.

So what I did to try to solve this:

I've changed the turbulence model, used k-e e k-omega (sst), I've changed the saturation pressure of water at 25°C (used 3574pa and 3169pa to see if there was any difference between them), I tried to use another methods to simulate cavitation with SST and k-e , I've tried to change steady state to transient, changed the size and quality of mesh now as last resource I am trying to simulate Transient with Cavitation.

With all those simulations I've reached the same results, 1953m³/h 41-42m of head....not even near of the 21.5m of head in experimental results.

So I´ve started to think that is not my simulation, but I really dont know.

The success pump design was done with the same way, k-e periodic geometry of impeller and diffuser, then k-w with rough mesh (I was starting cavitation simulations here), then k-e with full impeller, diffuser and return chambers to analyse losses.

I thought that I could have problem in my bench test at higher flowrates, but we had a problem with another 2 pumps, 1 centrifugal with lower flowrate and other semi-axial with high flowrate too.

Maybe it's still cavitation, because the behavior of the pump in the bench test is strange, when we start the pump, it is pumping like 2800m°3/h, then it starts to fall slowly until it reach around 2100 m³/h at o head. Those 3 problematic pumps I told you about with the same problem.

1 Centrifugal high flowrate, another centrifugal lower flowrate and the semi-axial with high flowrate and lower head.

I know it's a lot of information to start, but I guess it is better to have an ideia of the whole situation.

I would appreciate any suggestions, ideias, thoughts, and if necessary I can give you more information.

Forgive me for any gramatical problems, but english it's not my main language, I am an portuguese speaker.

And here I dont know about much more people that uses CFX to turbomachinery application, I guess the company I work is one of the first.

Thank you again.

You can talk to me by email too if you want, ismael.schroer@gmail.com

 Graham81 November 10, 2011 09:36

At 2/3's the nominal flow you might be entering suction recirculation and induce cavitation like that. This highly transient phenomenon will destroy your pumps experimental performance and the effect should be captured through running transient and cavitation, but its not an easy simulation.

Can you include a picture of the relative velocity vectors projected onto a plane somewhere spanwise between hub and shroud?

 ismael.s November 10, 2011 11:38

Hello Graham,

Thank you for your quick reply, very interesting what you told me.

So here's some pictures, if you want I can upload more.

The images are from a steady state k-e cavitation simulation. I am running one Transient but it isnt ready yet.

http://i1233.photobucket.com/albums/...D/Imagem_2.jpg

http://i1233.photobucket.com/albums/...D/Imagem_1.jpg

Vectors

http://i1233.photobucket.com/albums/...FD/Imagem3.jpg
Cavitation

http://i1233.photobucket.com/albums/...FD/Imagem4.jpg
Geometry Preview

 ghorrocks November 10, 2011 17:20

http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

 Graham81 November 11, 2011 04:25

1) Although the top 2 pictures look good, I cant judge any recirculating areas from that. Please make a cross sectional plane with projected relative velocities.

2) The suction side caviation zone (is that vapor fraction >0.5 plotted?) doesnt extend up to the shroud where tip speed is highest. Do you have significantly different leading edge blade angles at hub and shroud?

3) If your experimental pump is running unstable and your simulation is predicting cavitation, how realistic is it to get a matching QH?

 ismael.s November 11, 2011 07:07

Quote:
 Originally Posted by ghorrocks (Post 331632) Have you read this FAQ? http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F
I didn't, now I've read it, thank you. It seems to follow the direction Graham told me about.

About the physics, I suppose it's ok, people that work for ESSS (Ansys vendor in south america) helped me start the simulations. As I told you before I made one pump reach the results, other near and I have 3 with problems (remembered now to tell you, 2 of these 3 have very high flowrate, higher than the ones approved).

The pumps are supposed to be very similar, the only differences are the type of impellers. 2 Radials (Pumps working fine), 2 mixed flow (high specific speed) and one semi-axial impeller (higher specific speed).

About being turbulent, I made Reynolds calculations and 2 problematic pumps are not laminar but aren't turbulent either (Re= 3500 and 3900). The other 2 (radial pumps) have Re = 4975 and 5260, both are turbulent. 1 I didnt calculate yet.

I will upload few more images to show you the pump design and what are being predicted in simulation.

Thank you for the help until now.

Ismael

 ismael.s November 11, 2011 08:35

Quote:
 Originally Posted by Graham81 (Post 331685) 1) Although the top 2 pictures look good, I cant judge any recirculating areas from that. Please make a cross sectional plane with projected relative velocities. 2) The suction side caviation zone (is that vapor fraction >0.5 plotted?) doesnt extend up to the shroud where tip speed is highest. Do you have significantly different leading edge blade angles at hub and shroud? 3) If your experimental pump is running unstable and your simulation is predicting cavitation, how realistic is it to get a matching QH?
Hello Graham.

1) Something like this?

http://i1233.photobucket.com/albums/...D/Imagem_3.jpg

2)Yes, it is the water vapour volume fraction 0.5 and the LE angles are different. I made the design in the Ansys Blademodeler, I cant tell you now which are the angles exactly because I lost this data from ansys, somehow the project was misplaced.

3) 3) If your experimental pump is running unstable and your simulation is predicting cavitation, how realistic is it to get a matching QH? I didnt understand what you mean here, please, can you explain other way?

Well I am posting 3 more images to show you how is the pump in question.

What's is my periodic simulation using

And here I have one simulation with full geometry, 360° and return chambers, as much completed as I could. Steady state simulation, same design, 2500m³/h and 32m of head (the results), no cavitation.

http://i1233.photobucket.com/albums/...D/Imagem_4.jpg

Thank you.

 Graham81 November 11, 2011 08:54

Hey Ismael,

1) Yes that is what I meant. There are no recirculating zones apparantly.
2) So the pump is cavitating heavily in your problematic simulation.
3) What I mean is: if the pump on the testfloor is running across its curve (2800-2100 m3/hr at 0 head), it will be very difficult to get a simulation to match your experimental results. One result is that you predict it to have large cavitation pockets under these conditions that could be the root cause of the unstable operation.

I must say I dont have much experience with these small (close-coupled 25-30 kW mixed flow end suction?) machines.

 ismael.s November 11, 2011 09:11

Well, agreed with that, that's why I am trying to validate the fowrate around 1950m³/h, since the experimental test shows more stable.

And the pump is 294,2kW - 400hp.

The cavitation / transient simulation is still running, I guess it will run until wednesday, he will have a big bank holiday here.

I will still be checking and talking to you about the case.

Ah, about running 2800-2100m³/h at 0 head, it starts with 2800m³/h and start to decrease and ends around 2100m³/h so it becomes 'stable' flowrate.

The phenomenom I tried to explain is that it starts high then decrease until 2100 and stop at 2100 at 0 head. When we closes the valve to increase the pressure, the flow decrease normally as it should.

 ismael.s November 16, 2011 08:44

Hello, giving you feedback about the simulation.

I've finished the cavitation transient simulation with stage interface, the results are close to reality now.

1953m³/h - 5m head - 228cv (Numerical)

http://i1233.photobucket.com/albums/.../Imagem1-1.jpg

http://i1233.photobucket.com/albums/.../Imagem1-2.jpg

http://i1233.photobucket.com/albums/.../Imagem1-3.jpg

http://i1233.photobucket.com/albums/.../Imagem1-4.jpg

Now I will run it transient, with cavitation and transient stator/rotor interface, to try to identify what's happening inside the pump since I couldn't identify well with stage interface.

Oh, you can compare the results with the other images in this post since is the same geometry and mesh.

Thank you.

 ismael.s November 21, 2011 09:55

About the transient simulation, I was checking and the total time wasn't enough to solve the problem, so I added more time to the simulation and I got 42m of head again now, so the transient simulation didnt solve my problem.

Does anyone have any other suggestion of what could I do?

Thank you.

 zona February 8, 2012 10:28

Hi, I just saw this thread
can you send me your 3D files and ansys setup files so that I can look at them
I've made not too much simulations and they all came OK (+/- 5-10%) from experiment or expected values

 ismael.s February 27, 2012 07:15

Hello, thank you for repplying. I was on vacation travelling, I saw your repply but I didn't answer earlier because I didn't have access to the files until now.