[freemankofi]

I am using CFX v12.1 to solve two-phase flows in a horizontal pipe. Both the continuous (liquid: density = 867 kg/m3) and dispersed (solid: sand with density 2650 kg/m3 and d32=0.43mm) are model as laminar.

Convergence is not my problem, however, the volume fraction (VF) results doesn't agreed with the published experimental data. The experimental data indicates that the sand settled and almost stratified flow. The delivered concentration was less 50% of the inlet VF. CFX indicates that there is an effective transport of the sand with almost 100% delivered concentration.

Since the VF or concentration profile is "wrong", I don't trust the velocity profile it is giving me, despite the pressure gradient is somehow closed to the experimental data.

I am not sure if it's coming from liquid-solid interaction model (kinetic theory), since it's is based on turbulent flows. should turbulent or laminar really matter for this model?

Any help on this will be much appreciated. Thanks.

[Michiel]

Hi,

Take a look at the papers published on this site: http://works.bepress.com/sandip_lahiri/

Do you run a transient analysis? Which multiphase model are you using?

You mention 50% of the sand is settled in the pipe. That means a stationary bed is formed at the bottom of the pipe, and this will happen over time.
A problem you have is that the CFD code doesn't include the settling behaviour. And that is the hard part I think. I think you should focus on this part before you can aspect some reliable results.

I'm not an expert on this subject, I just started working on a sand transportation model which is capable of handling high concentrations.

[freemankofi]

Michiel,

Thanks. Actually, that paper is on turbulent flow. Beside my group has tremendous experience on that subject and is relatively well understood on that regards.

We thought that, the laminar flow shouldn't be a problem for the current CFD codes. Never said that, I am more hesitance to assign it to the codes or models since all the models (and CFX or Fluent codes) gave the same results.

I think, due to the formation of the bed (stationary or moving), it's not clear if the length is long enough to allow this formation to happen. Although, the length is longer than what was used in experiment but I'm not worry about that since experiment was a pipe loop which might have effect on it. Secondly, when the formation of the bed takes place, does the flow become unstable or ? Finally, as you said the settling behaviour are not accounted for in the current codes, but is it a big of a deal?

I have started addressing the above issues and I am seeing an encouraging results (see the attached). I still also need to optimise it since my concentration is too high. I am more concern of geometric parameters and BC as opposed the model since my experience has thought me that there is no difference among them. I also realised that some previous turbulence works used inlet velocity profile, I think that's off of the chart since in experiment, such condition does not exist. One wants to see the actual flow development and specifying remove such observation.

Freeman

[Michiel]

Ok, I understand the difference in your case and the turbulent flow. Verry interesting stuff you work on. Unfortunately I'm more a practical then a theoretical person. But what you describe sounds a bit familiar to me.

"Secondly, when the formation of the bed takes place, does the flow become unstable or ? Finally, as you said the settling behaviour are not accounted for in the current codes, but is it a big of a deal?"

Well, I think it is a serious deal.... I know somebody who has a lot of field experience in the dredging industry, and worked for over 20 years in the Emirates on several big projects. For these guys it is common to pump a sand water mixture over 5 kilometers of pipe line or more. With clean sand/water there is no problem. The flow velocity is high enough the keep the small sand particles moving. But when gravel or rocks are picked up, the velocity is to low to keep the larger and havier particles moving. These large particles can form a bed, and eventually the pipe gets jammed on these beds. Even with a flow velocity of 4 m/s the large rocks are rolling on the bottom of the pipe.

In this practical story I can see some relations to the more theoretical case you are studying. And I would say that the settling behaviour of the particles can have a big influence on the entire flow field.

It seems that some of the multiphase models can handle high density mixtures as long as the flow velocity is high enough to prevent settling. When settling behaviour starts I think a more granular like flow behaviour is introduced in the moving or stationary bed. I don't know if todays CFD codes can handle granular flow? As i said I'm not verry strong on the theoretical side... Are you familiar with granular flows? If not, I have some information on this subject which I can send to you. Please send me your email if you are interested.

[freemankofi]

Michiel,
Thanks for the inside story. Perhaps, I could give you some background for my work. For the turbulence mode, due to the presence of turbulence mixing, the particles are well transported. This has been proved both theoretical and industrial, especially in Canada here with our oil sands.

The issue if we have now is for a tailing thickened, our industries have no choice than to transport it in laminar regime. The question then is what will be the suspension mechanisms for the coarse particles? To answer this question, researchers in our research council came up with a criteria which depends on pressure gradient. However, the tailing paste is non-Newtonian and not sure if this criteria will work or not. Even if it works, it's no-no for our industries since it'll cost more, such us abandoning centrifugal pump for positive displacement pumps. Therefore, there should be a better way out for transporting tailing thickened.

So, the current work is just validation to see how the current CFD codes could predict the experimentally observed flow in a Newtonian flow with high viscosity. For example, experimentally, it was observed that the delivering concentration was lower than in-stu concentration. This means that there was settling. This is more significant at certain velocities with certain pressure gradients. The outlet conc. profile ascertain to that. As a validation, can CFD code reproduce this criteria? If not, how far is the deviation? Note that, since CFD is an approximation, no experienced CFD practitioner expects exact solution considering the fact that this is a multiphase flow. One will only expect a ballpark result.

Please, note that these experimental works are not a "small University lab or something" but a world class industrial research lab with leading research engineers in hydrotransport serving our industries here.

As for granular flows, I have some information on it. However, it doesn't hurt to get your version if they are different. Please feel free to send to me via private message on this forum.

Thanks one more time.

[Michiel]

Most of the information I have on Granular flows comes from the university of Twente and can be down loaded at the following websites:

http://www2.ts.ctw.utwente.nl/kruyt/
http://www2.msm.ctw.utwente.nl/sluding/

Sounds like your work is way out of my league, but as I said very interesting. I think I will stick to the turbulent water/sand mixtures for now.