[Benfa]

Hi,

we are trying to varifiy our CFD calculation with CFX on experimental data of a Stairmand Cyclone. Therefore we made different mesh ranging from 300.000 to 1.7 mill. cells. we start with steady state calculation using SST and after 100 iterations we switch to transient and RSM (SSG) turbulence modell. To judge convergence of the quasi-stead-state solution we monitor presure drop and mass flow at the bottom and top outlet. After geeting stable values we switch on the euler-lagrange modell using one way coupled and the Schiller-Naumann law. Different discrete sizes are used to get the efficiency curve for each particle diameter.
Problems that occur:
Pressure drop fits well to experimental data
None of the particles pass trough th top outlet. All of them just turn down.

inlet: inlet velocity
bottom and top outlet: opening, entrainment and opening presure (because of reverse flow)



We checked everything more than twice! We tryed another dimensions of the Stairmand that has experimental data too and we were not able to reproduce them too. Is there maybe something wrong with particle tracking.

Did anybody do such cyclone separation efficiency calculations and has any tips or hints how to get it done. Ansys support tells us maybe to use Fluent but CFX should be able o do this too (or what do you think?).

Thanks in advance!

[ghorrocks]

You are correct that simply switching to Fluent will not help, and that CFX can model this just fine.

Have you tried LES approaches? Some cyclone flows I have seen require LES as the core jiggles about and interacts with the turbulent structures in a fashion which cannot be modelled with RANS (including RSM models). It would not surprise me that you need resolution of the large turbulence structures to get particles to go out the top outlet - and that implies LES is required.

[Lale]

Dear benfa,
I have the same problem in fluent even by LES for a long time, please let me know how you solve the problem. Thanks in advance.

[Gert-Jan]

1) Turn on turbulent dispersion.
2) Pay attention to the inlet. Probably you used uniform inlet profile and particles distributed uniformly in space. This will not be the case in reality. It could be that in reality you have many more particles in the top part, close to the roof of the cyclone that can travel downward along the vortexfinder, towards the outlet.

3) Flow inside a cyclone can be very complex. Is the central vortex precessing? This can lead to pick up of small particles, ending up at the top outlet.


Question: You say particles end up at the bottom outlet. Did you modify the restitution coefficients of the wall? To lower values than 1?

[Lale]

Dear Gert -jan,
Many thanks for your reply.
1. I have used turbulent dispersion.
2. Please let me know about uniform inlet profile more. How I should impose it or change it.
3.There are central negative pressure and air core permanently in my simulation
4. I have not modified the restitution coeffient of the wall. How can I know its value. Whether the default is 1?

[Gert-Jan]

2) I think your inlet profile is uniform. You should change it to the individual components in line with your experiment, or include more ducting in front of your inlet. At least put you inlet far away such that it does not influence the processes in your cyclone.
3) probably your mesh is (by far) not fine enough.
4) I don't know the value. It depends on the materials you use. A value of 1 suggests there is no energy losses when particles hit the wall (action=-reaction)

[Lale]

Thank you for spending time on answer.
I believe in the effet of uniform inlet . I wonder that experimental efficincy is 30% for the sake of ligh paricle ( particle density = 1024 kg/m3) while the simulation reach to 100%. Wether the effect of uniform inlet can be how so much.

[Lale]

Dear gert-jan,
Could you please check my simulation it is possible there is a critical problem becuase of big discrepancy between experimental and simulation result

[Gert-Jan]

I already mentioned that this is a really tough simulation. The result can depend on a lot of factors. E.g., if solids load is high, you should run two-way coupling. If the core is precessing, you should run transient.

Questions:
- Did you perform a mesh dependency check?
- You apply openings as BC. What pressures do you set?
- Do you know the gas massflows through top and bottom outlet?
- Do they agree with experiments?

[Lale]

Thanks alot.
the Solid concentration is less than 4% and transient has appllied in two phase and DPM.
1. I wonder when the number of mesh is 400000 or less the efficiency is 90% while when the number of grid is around 800000 the efficincy is 100% so the diffrence between simulation and experiment increase with increasing the mesh number.
2. I set 0 barg in overflow and underflow and back pressure 1 for air entering in both of them.
3. I do not know the experimental gas massflow becuase it mixs with water. But i know the simulation data.
The result of liquid massflow is agree with experiment
In addition the pressure and axial velocity profils are in good agreement with literature.
I need your help really

[Gert-Jan]

What Solid concentration? 4% Mass or 4% volume?
But Im lost anyway. You mentioned Solids, air and liquid......... Please explain exactly what you are doing. Is it solid-gas separation? Liquid-gas separation? Or solid-liquid?

[Lale]

There is a hydrocyclone so seperates solid from liquid. There is an air core in center of hydrocyclone becuase of opening underflow and overflow to atmospher and negative pressure core in center of hydrocycloe cuase air intake.
I mentioned 4% becuase I want to say there is DPM condition (low concentration). The exact concentration is 0.05 % mass.
thank you

[Gert-Jan]

I reread the whole thread:
- It looks like you picked up an ancient query from 2013, expecting people to remember what they did 6 years ago. Why not start a complete new query?
- Moreover, the old query starts with a Stairmand cyclone, which is very misleading since it sets my mind automatically to gas-solid separators. But you have 3 phases...........That complicates things even further.
- Moreover, looks like you are working in Fluent, while this is the CFX-forum.

Why not give all information you have from the start? In this way it is all a bit a waste of my precious time, don't you think?

[Lale]

Dear Gert,
I am so sorry to cause misunderstanding. I am amature to use cfd-online. I have used a hydrocylone with 15 mm diameter to separate light particle ( density= 1024 kg/m3) from water. The rsm has been applied for water . Vof for air core and dpm for particles. BC: inlet velociy for inlet and pressure outlet for both underflow and overflow. In dpm approach the reflect BC for wall and scape for underflow , overflow and inlet. Really, Ido not know which data is important to mention here.
Please feel free to ask any other information.
Thank you for your patience to answer.

[Lale]

Dear Gert,
There is a change in efficiency during the solving time, firstly the solid exit from overflow with low solid mass flow,then the solid in underflow observe and then increase, eventually the amount of mass in underlow is 9 times to overflow .when I watch the particle animation it is true process but how the collective efficiency can be gained?

[ghorrocks]

I think your last post is asking how can you improve the separation efficiency of your cyclone. This is a cyclone design question, not a CFD question. So have you looked in cyclone design textbooks and literature?

There are design recommendations for cyclones to optimise their efficiency. Look in the cyclone design textbooks and literature to find the recommendations.

[Lale]

Dear gorroks,
Thank you for your reply. Indeed, in the last question I want to know how I gain collective efficiency in simulation becuase it changes during computation. In my project at the bigining the efficiency is quit low and then it increases but I want to know the collective efficiency rather than instntanouse.

[Gert-Jan]

- I still don't know if you use CFX or Fluent. I think fluent, not?
- You mentioned the efficiency increases during computation. Do you mean during steady state or during transient?
- As Glenn mentioned, on this forum, please raise questions related to CFD, not to the working of your configuration. I don't know if there is a magic switch that can increase or decrease your efficiency. It is a three phase hydrocyclone, which is a very specific configuration. I think you are one the specialists here on CFD-online.

[Lale]

All of my questions were about CFD and In the last question I want to know how reach a reliable report in fluent in collective efficiency rather than grade efficiency. However, I do not know why it seems about concept of hydocyclone performance. Any way I tried steady and transient and both of them have the same result.

[Gert-Jan]

I don't feel confident enough to give you advice on a Fluent calculation. Not my cup of tea. So, please go to the fluent forum. Possibly you'll get a reliable answer there.