Multiphase Flow - Science topic

Yes, and it does not show good precision.

Ziyi Wu, Thank you, I know what you said. But my question is about how to simulate dynamic contact angle instead of static contact angle.

THE ENHANCEMENT OF HEAT TRANSFER BY TWO PHASE OR MULTIPLE IS MORE THAN SINGLE PHASE

Without more details, I only may speak about the phenomenological aspects that may be verified. (i) When one applies User Defined Function (UDF) this should not contradict the mass conservation of the dispersed phase (ii) the initial diameter would depend on many factors: surface tension, turbulence, injection device...References on dispersed turbulent flows are available on the following research projects: https://www.researchgate.net/project/Single-Phase-and-Multiphase-Turbulent-Flows-SMTF-in-Nature-and-Engineering-Applications https://www.researchgate.net/project/Interfacial-Transfer-Closure-for-Eulerian-Eulerian-Two-Fluid-Modeling https://www.researchgate.net/project/Turbulence-Closure-for-Eulerian-Eulerian-Two-Fluid-Models

Hello, Armin Hajighasem Kashani

First difference (discretization method): FLOW-3D implements finite differences while Ansys Fluent solves using finite volumes.

Second ("speed"): FLOW-3D is numerically "faster", due to the discretization method and because it has the possibility to simulate a single fluid (Water). Fluent resolves both fluids (Air-Water), at least with VOF.

Third (sediment transport): FLOW-3D allows coupling semi-empirical models of sediment transport to flow hydrodynamics (somewhat simplified, but better than nothing!). Until the version I have handled in Fluent (2019R2), there is no way to model sediment transport. You need to make a UDF for this purpose or try with particle models, for different grain sizes (field unknown to me!).

In my opinion, FLOW-3D is much more suitable CFD software for solving river hydraulics problems. In that aspect it seems to me that it has a great advantage over Fluent,

Good Luck,

Leonardo

More information is needed to answer this question. Basically, even in single-phase flows, the normal gradient of the pressure is almost zero only in parallel or almost-parallel vertical laminar flows. The average pressure is "permeated" by turbulence even in 2D parallel flows. When it comes to two-phase flows and within the framework of RANS modeling using Eulerian-Eulerian two-fluid models closure is needed in order to express the pressure in one phase as a function of that of the second phase. The simplest considers that the two pressures are equal in the absence of surface tension. See references within the research projects: https://www.researchgate.net/project/Single-Phase-and-Multiphase-Turbulent-Flows-SMTF-in-Nature-and-Engineering-Applications https://www.researchgate.net/project/Interfacial-Transfer-Closure-for-Eulerian-Eulerian-Two-Fluid-Modeling https://www.researchgate.net/project/Turbulence-Closure-for-Eulerian-Eulerian-Two-Fluid-Models

Noted professor Stam Nicolis , I will keep that in mind. Thanks for your time and help!

VOF is used where surface tracking among the phases is desired. In rest of the cases scholar mostly used Eulerian Model for steam water direct interaction problems.

VOF:DOI:

Eulerian Model : Numerical study of flow and direct contact condensation of entrained vapor in water jet eductor

We have tried AMR for multiphase flow with phase-field in the past. Usually, refinement is dependent on the phase-field parameter and you can control the number of refinements in a given time scale of simulations. You can refer to the following articles for more information on the implementation of AMR in the Comsol phase field.

1) Yue, P., Zhou, C., Feng, J. J., Ollivier-Gooch, C. F. and Hu, H. H., "Phase-field simulations of interfacial dynamics in viscoelastic fluids using finite elements with adaptive meshing," Journal of Computational Physics, 2006, v. 219, n. 1, pp. 47-67.

2) Zhou, C., Yue, P., Feng, J. J., Ollivier-Gooch, C. F. and Hu, H. H., "3D phase-field simulations of interfacial dynamics in Newtonian and viscoelastic fluids," Journal of Computational Physics, 2010, v. 229, n. 2, pp. 498-511.

3) Yue, P., Feng, J. J., Liu, C. and Shen, J., "A diffuse-interface method for simulating two-phase flows of complex fluids," Journal of Fluid Mechanics, 2004, v. 515, pp. 293-317.

4) Yue, P., Zhou, C. and Feng, J. J., "Spontaneous shrinkage of drops and mass conservation in phase-field simulations," Journal of Computational Physics, 2007, v. 223, n. 1, pp. 1-9.

5) Yue, P., Zhou, C. and Feng, J. J., "Sharp-interface limit of the Cahn–Hilliard model for moving contact lines," Journal of Fluid Mechanics, 2010, v. 645, pp. 279-294.

6) https://www.emerald.com/insight/content/doi/10.1108/HFF-01-2020-0001/full/html

I doubt you gonna get reliable results studying droplet deformation in liquid-liquid dispersed flow by means of simulations with commercial codes. The phenomena involved is too complex. Droplets break and coalesce under simultaneous population and wall effects, also depending on the interfacial tension between phases as well as on the local pressure (assuming constant temperature). For a given pair of fluids, the relative velocity between droplets and continuous phase depends on size and shape of droplets. The relative velocity also induces pairs of vortices rotating in opposite directions inside droplets (conservation of angular momentum). These may cooperate for the droplet breakage.

Even though it may be well outside your research scope, experiments are required if you want reliable data.

I guess a little craft hability is required. Build a mechanical device where small parallel glass plates (same width and thickness as those used in microscopy) are pulled along in opposite directions, similar to your Couette rheometer. Clearly the experiment will be 2D. The small gap between plates would be filled with the liquid-liquid dispersion. Some sort of rail system will be required for guiding the sliding plates. The plates could be pulled with the help of an electrical DC electrical motor with variable rpm and a system of pulleys. Make photograph/film from above. Ah, include my name on the Patent application (laugh).

Some math about your questions and useful references.

Resources and academic tools on turbulent single-phase and multiphase flows in industrial facilities and in natural environments

SINGLE PHASE AND MULTIPHASE TURBULENT FLOWS (SMTF) IN NATURE AND ENGINEERING APPLICATIONS | Jamel Chahed | 3 publications | Research Project (researchgate.net)

I refer you to two discussions which bear on the two problems of Eulerian two-fluid models closures: Turbulence and momentum interfacial transfer:

(6) Turbulence Closure for Two-Fluid RANS Modeling (researchgate.net)

(6) Interfacial Transfer Closure for Two-Fluid RANS Modeling (researchgate.net)

I refer you to two discussions which bear on the two problems of Eulerian two-fluid models closures: Turbulence and momentum interfacial transfer:

(6) Turbulence Closure for Two-Fluid RANS Modeling (researchgate.net)

(6) Interfacial Transfer Closure for Two-Fluid RANS Modeling (researchgate.net)

Look forward to exchanges on these topics

Almost a year later, but anyways:

I wrote together with someone else an article that probably suits your needs:

Preprint is on researchgate.

Giuseppe Altieri thanks very much for your answer

Yes you can, if those different fluids are separated. Just add another fluid region to your case.Then you can define individual thermophysicalProperties.

Let's say you have your heat exchanger which is set as a solid region, then you can define water as one fluid region on the outside of the HX and and Air as another flui region on the inside of the HX. The two fluids will be interacting with eachother only thermally.

In my limited understanding, i utilized a steady state model to calculate the permeability of certain structures (resembling metallic foam). The simple way that I used to adopt was to apply a known pressure difference at the inlet and outlet, and obtain volume flow rate through post processing. This would simply give you the permeability. Ofcourse, the higher the volume flow rate, the more permeable the porous medium shall be. However, under transient conditions and using variable viscosity, the problem gets much more complicated. Nonetheless, steady state isothermal condition can be a good starting point and may give you a reasonably accurate ball park value.

Go for Ansys CFX or COMSOL.

Dear Shilpa,

the source of the problem can be varied from one model to another. However, overall, coupling process of solute transport with the phase flow behavior can cause such issue. The associated factors, however, can be changed from traditional factors (i.e. grid size and time step) to non-conventional factors ( e.g. Effective concentration and the shape of rel perms). Also the solution type of the numerical scheme can be another source of such issue (e.g. fully coupled schems causes convergence issue more than decoupled scheme). See SPE-192074-MS and

Al-Ibadi, H., Stephen, K., & Mackay, E. (2020). Pulse Generation and Propagation in the Numerical Solution of Low Salinity Water Flooding. Journal of Petroleum Science and Engineering, 108151.

Hasan

Dear Ziqiang Cui,

Within the classical framework, one can minimize the error between measurements and calculated values, in function of the multiphase electrical properties. There are different optimization algorithms which deal with non-linearity, with the most popular being Gauss-Newton, Simulated Annealing and Kalman filtering. Of these three, Simulated Annealing requires a large processing time and Kalman filtering may be suitable for real time (depending on other aspects).

Right click 'Setup' and press reset. This will force you to start everything after the updated mesh from scratch. Don't bother looking for the error.

Well did you got the answer ?

Don't know with fluent but with Hypermesh mesh refinement can be achieved.

These videos will help to learn 2D and 3D mesh transition techniques.

It is better to approach Eularian-Lagrangian Model in this regard. Most of the comparative research articles shows better fit with this model than any other models.

Leonardo Henao; Use the HRIC Scheme and refine the mesh in the z-direction (the wave elevation direction).

I think that the "species transport model" is the most suitable for your simulations, it's more suitable for the mixing of miscible fluids but you must be careful because you need to disable the reaction process.

For the colleagues who recommended the VOF model, I inform them that it is suitable for multiphase flows where fluids are immiscible, and for the others like mixture and Eulerian models the problem arises that you can't introduce the diffusivity coefficient particularly to validate with previous work.

Good luck.

Hi Manu Francis ,

it basically should work the same, as in single phase flow, since your are not interacting with any of the phases?

But why to you use Thread *t = Lookup_Thread(d,zone_id); and then doing a thread_loop_c(t,d) where you loop over all threads in the domain. That makes no sense at all, but it also should not really matter, if you did not activate the source term in the fluid cell zone.

Best regards

Christian

A good starting point is the book of Krüger et al. as well as the open source code OpenLB and the LBM spring school, see www.openlb.net .

Thank you very much Swapnil! I

Please mention the direction of motion of both phases i.e., water and air and tracer inlet position.

Guangpu Zhu

Hi, I was able to to it. Thanks for your reply. It was really helpful.

You also need to consider contact angle; so the corrected version is as follows based on Franken's model:

Pe = -2*B*sigma*cos (theta)/ R_max

where B depends on the pore geometry shape: 1 for cylindrical pores and 0<B<1 for non-cylindrical ones. Note that for B =1, the equation converts to well-known Young-Laplace equation.

Dears Somnath Rangrej , Purushottam More

Please write your considering sources as formulas here. Maybe I could help.

Thank you.

You can also use the C_VOF. It works for both models.

For example, you can just simply write: v1=C_VOF(cell,liq1);

Dear Zhang Chunhua

I think you are using the Shan Chen model, whereas the Free Energy and Phase Field models can eliminate spurious velocity.

Okay I made a systematic study recently with OpenFOAM about the issue that you were talking about. And it seems not surprisingly, you will over all lose mass/vol. while you track the whole mass in the system. And the reason for this is that the equation which is solved for the interface (alpha advection) is of hyperbolic type and once you try to solve it with a dissipative discritization (Upwinding) it'll make minor errors in your alpha field resulting in overall misbalance of mass/vol. in your enclosed medium. Of course this is not noticed once you try to simulate an open case. Anyways, what is important is that you cannot avoid this dissipation in the solution of alpha because it has also a good side and that is "stabilization". The good news is refining the cells at the interface would make this effect less vibrant.

BTW check the implementation of VOF in OF as well (interFoam solver) regarding the interface compression...

Sorry for the late reply, I didn't receive a notification for your last reply for some reason.

How can I set a pressure drop exactly?

Gravity is on.

For the geometry, it's a 2D rectangular tube 20 mm long and 0.7 mm wide placed vertically, I patched the whole tube with powder.

DPM is working fine but the eulerian granular model is more interesting for me.

with a bit sense of humor I would say, Omar see what you get, several conflicting views on the matter, depending on how many experts you ask as much expert views you will get. I am thankful to you putting forward this question.

I highly value the knowledge of Bernhard and John, however, I have to disagree to some degree.

Based on how I understand classical German Colloid science, here is my version -

Dispersions are inhomogeneous systems characterized by having at least one interface where molecules feel a different environment than same molecules in the bulk, there are always molecules at interface but also in bulk. Based on this, suspensions (nothing else as solids dispersed in liquid, which may settle under given circumstances or not) and emulsions (liquids dispersed in liquids) are both dispersions beside other ones, e.g. fog.

Term colloid is unfortunately historically not clearly defined, something between characterized by slow diffusion and microscopically not resolved. Importantly properties of colloid dispersions are interface determinated. For more confusion there are also colloidal solutions (no dispersed bulk phase). If colloidal dispersions get flocculated, they are still interface determinated opposite to possible result of coalescence or Ostwald ripening, when dispersed phase grows beyond colloidal size.

In this sense nanofluids are colloidal dispersions and of course colloidal suspensions same time. And they settle with slow or slower velocity. When settling and diffusion velocity get to the same order you will get a so call 'isobaric' equilibrium. Even solutions (e.g. salt solutions) or mixtures of gases exhibit enrichment of higher density components towards the bottom, even at earth gravity.

I would have a look at the newly introduced expression language capabilities in ANSYS Fluent 2019 R1 and R2. I could imagine, that using expressions, you do not have to write a UDF anymore? In an expression, the defined wall flux could be just multiplied with an expression value, which is always 1 for the primary phase of your multiphase flow and always zero for all other phases. I have not yet tried this by myself in ANSYS Fluent, since usage of expressions in this solver is rather new, but that is how I would implement it in e.g. ANSYS CFX using CCL language. New expressions in ANSYS Fluent should allow to do that at some point as well.

Regards,

Dr. Th. Frank.

I would suggest trying to define wall slip in the wall boundary condition and see where that leads you.

Simply plug and play the (ijes) or parasolid model into the ansys workbanch and then try to resolve it

The Knudsen number is defined via the mean free path l_fp. In liquids the mean free path hardly can be introduced. Though liquids differ greatly from the solids, the marked molecule in liquid actually is moving under the impact of the self-consistent-field caused by the other molecules. Introducing the concept of the quasi-particles we can introduce the mean free path for those quasi-particles. The last is of the order of the mean distance between the molecules. Knudsen number as Kn_p=l_fp/D, where D is the nanoparticle diameter formally can be introduced, but how can it be used for the continuous model justification? The Knudsen number arises as a dimensionless parameter in the kinetic equation. Small Knudsen numbers correspond to the continuous model, even taking into account, that for the dense gas, all the more for the liquids, much more complicated kinetic equation compared to the Boltzmann equation should be used. That parameter has no practical use without considering of any type of the kinetic equation. For the nanoparticles Kn_p may not be small, so I see no application field for such parameter, specially taking into consideration the long-range interaction of the particles via the liquid.

Hi Weijia Lu,

Great point. Machine learning uses algorithms to parse data, learn from that data, and make informed decisions based on what it has learned. Deep learning structures algorithms in layers to create an “artificial neural network” that can learn and make intelligent decisions on its own.

Best,

Moh.

Well,

q=kA (P1- P2)/ µL---------------------- (Equation ) where

q = Volume flow rate

k = Permeability

A = Core area perpendicular to flow direction

P1 = Upstream pressure at location 1

P2 = Downstream pressure at location 2

µ = Dynamic viscosity of flowing fluid

L = Distance over which pressure drop (P1- P2) occurs.

Hi Ankita

Interfacial tension is the work (J or N.m) that is required to increase the surface area of an interface between two phases, expressed per unit surface area (m2). The units of interfacial tension is therefore J/m2, or N/m.

Given two immiscible liquids in a thin tube, capillary pressure is the difference in pressure accross the boundary between the two phases. It results from the interfacial tension between the two phases.

All the best

Johan

Thanks a lot Ricardo

Its calculated based on effective mixture properties.

I think you should check mesh independency for your solution, changing the solver, will result in change in the mesh dependency criteria.

hi

used for non Newtonian flow

You can check below mentioned article:

Article A LBM-DEM solver for fast discrete particle simulation of pa...

Dear Chinmoy,

Using Re as such is NOT wise as we fundamentally have a different physical domain for single and multi-phase fluid flows. In case of a single phase fluid flow, we have a solid-fluid boundary, where the concept of "NO-SLIP" boundary condition can be applied comfortably,

whereas,

in case of multi-phase fluid flow,

what we have is a fluid-fluid (i.e., an immiscible liquid-liquid boundary) interface at the fluid boundary. Here, we cannot apply the conventional "No-Slip" boundary condition and hence, we may need to apply a "free-slip" boundary condition as the fluid interface is NOT rigid and it is deformable and this deformability depends on the intensity of the interfacial tension. And it requires a knowledge of Laplace number, which is the ratio between the fluid pressure and the Laplace pressure (Sigma/L). Determining the value of L (the capillary length) remains a real challenge.

Good Luck.

Tristan Davenne is entirely right. Essentially what you mentioned about your current solution points in the same direction. It seems to me, like ANSYS support has recommended to you to switch from a disperse multiphase flow formuölation to a VOF-like (VOF = volume-of-fluid) formulation. VOF exists in the ANSYS software in two flavours - homogeneous and inhomogeneous. The homogeneous model with just 1 set of momentum equations follows the assumption, that both phases move with the same velocity field. In that case no assumptions or knowledge about interfacial momentum transfer is needed. ANSYS has implemented an inhomogeneous VOF model as well. In that case from the local ratio in volume fractions of both phases an inherent drag term is deduced, but in that case you don not have to specify it yourself. It assumes, that separatign interfaces between the two phases are rather macroscopic and resolved by the simulation like typically done in free surface flows.

Regards,

Thomas Frank.

Dear Professor Suresh,

Thank you for your reply.

Much appreciated.

If I understand correctly, it is an elongated tube that can serve as a waveguide. The electromagnetic waves can not penetrate the wall of the pipe, even at very high power. You have to couple the energy on one side of the wet diaphragm and measure the electromagnetic energy on the other side with a similar probe. Suitable coupling elements can be found here:

Thanks Ricardo Vicente de Paula Rezende .

What would you say about the 'Bulk mass flow rate' available for Outlet boundary condition.

For that, no volume fraction information is asked in CFX-Pre.

Should I conclude that, if I use that option: the liquid+vapor=bulk; it would mean that it is necessarily kept constant (0.3kg/s both for inlet and outlet)?

If so, is it Physically possible as well?

P.S.: my fluids are liquid water and water vapor (wall boiling case).

Dear Sir,

For multiphase flow inside porous media, you can use the mathematical formulation of TPMM along FVM instead of VOF. For further information, please see my articles in my page and also you can contact with me at any time.

best regards

I fully agree with previous answers. But your question is not clear enough.

the distribution of fluids (previous history) and wettability (if the viscosities are similiar) are the controlling factors for the pressure needed to flow. involving capillary pressure muddles the issue.

For example, if your water saturation is 40 % and all the other variables are the same. Not all the pores have 40%! some will have 100% and others 20%. Therefore, the pressure drop will be different for different pore size distributions, mainly if they are not unimodal.

Also, during imbibition and drainage at the same saturation you will have different fluid distributions and as a consequence different pressures.

by P. Vorobieff, C. A. Brebbia

Publisher: WIT Press / Computational Mechanics (April 18, 2018)

Efstathios Michaelides, Clayton T. Crowe, John D. Schwarzkopf

September 1, 2016 by CRC Press

Thanks a lot for your kind references.

Deewakar

As long as you have low mixing, a sufficiently high water column and volume (to provide buffer against liquid/gas phase contact) and a membrane/module which can inject the gas (smaller pores work better; "pore-free" silicone is also effective), and you can control the gas input pressure/flow, you should be able to inject high amounts before the liquid phase becomes fully saturated and bubbles start to form. The Henry's constant and mass transfer characteristics (which would severely limit the amount of H2 you could inject (i.e. 1.6mg/L) if the liquid phase is open to atmosphere) are then not limiting in this case. In a similar vein, if you have high H2 partial pressures in the gas phase and your reactor is well-sealed, you can dissolve more as well.

Not a VOF expert here, but for what concerns the Immersed Boundary (IB) there are several possible routes to fix the conservation problem, depending from what method you actually use.

Actually, I expect you to be working in a finite volume (FV) context, otherwise you have additional sources of non conservation, besides IB itself.

The general idea to fix the conservation problem of IB with a FV method is to track the mass flux on your boundary (but it can be also an energy flux, etc.). This is trivial in a FV context, as you know all the cell faces that make up your immersed boundary and you apply the boundary flux there.

Once you know the amount of your leak (and hopefully it is just proportional to the interpolation error on the boundary) you can cancel that leak by applying a global correction.

One possible way is to apply the correction to all the cells of your domain as a volume source of mass (or energy, etc.). This is done, for example, here:

but I don't like it as it acts as an instantaneous propagation of a local information (boundary) the the whole domain without a specific physical meaning.

A better approach, the one I currently use in my code, is to apply the correction as an additional boundary flux on the same boundary where the leak is created. This is not different from correcting the outflow mass flow in an incompresible flow in order to preserve the divergence free condition.

A possible implementation is here:

Note that these two implementations in the papers above are typical of unstructured settings, where the "solid" part of the grid is not actually present in the computation. But similar treatments have been developed in the more common setting where the whole cartesian block grid is part of the computation.

I don't know if VOF in these settings would require additional attention.

https://www.cfd-online.com/Forums/cfx/90063-ansys-cfx-solver-exited-return-code-1-a.html

http://www.cfd-online.com/Forums/ansys-meshing/79815-proper-way-name-boundaries-2d-model-use-cfx.html

https://www.researchgate.net/post/Problem_with_CFX_solver

https://studentcommunity.ansys.com/thread/cfx-solver-error/

Following!

Krzysztof,

I believe there are alternatives which can be safer.

Regarding liquid-liquid emulsions, (Matthias Kraume ) has tested different oils when he was looking for Toluene alternatives.

See attached file. There are different oils & their physical properties. I hope it will help you.

Best regards.

You shoud add function ovject to your controlDict in order to read forces etc.

From the forces probably you can calc what u need.

Franco

Hi

You can simply understand it by " Surface area to volume ratio" (SAV) concept. From dimensional analysis, SAV=1/L. so when a channel characteristic length is 1mm, SAV=1000 , that is surface effect are 1000 times higher than volume effect. Gravity is a body force which acts on the volume so its effect compare to lets say friction is 1/1000 smaller in a minichannel with characteristic length 1 mm. For heat transfer surface area is important so mini or microchannels are better heat exchangers.

Dear Mr. Pandey,

let me try to help you out and clarify your doubts in a simple way.

A gas is constituted by many many molecules, each one having its own velocity. However, while studying a gas flowing in a given domain, we cannot follow each molecule so we introduce the concept of particle.

A particle is constituted by a bunch of many molecules (say millions) in the neighbourhood of a point, which however has dimensions much less than the domain to be studied. So the studied domain is composed of a very large number of particles. This is called the continuum assumption which is not valid, e.g. in rarefied flows.

More precisely, the particle must have a volume that is much larger than the cube of the gas mean free path and this volume must be much smaller than the volume of the studied domain. For air at STP, a cube having its side equal to a mean free path contains about 10,000 molecules.

The velocity of a particle of a single component gas is generally defined as the vectorial average of the several molecules contained in it. Of course, in a multi-component gas, a weighted vectorial average has to be performed. So, the velocity of a particle can be regarded as the velocity of the center of mass of the particle itself. Besides, the velocity of a particle depends on the adopted reference frame.

But we have to remember that each molecule has its own velocity so, we may define the peculiar velocity of a molecule as its actual velocity minus the velocity of the particle it belongs to. The peculiar velocity of a molecule does not depend on the adopted reference frame because it is always measured with respect to the center of mass of the particle.

In this simple model, the internal energy of the particle is represented by the average kinetic energy of all the molecules contained in the particle evaluated with the peculiar velocities.

It is like if we split the total kinetic energy of the molecules in two parts, a macroscopic one (the particle kinetic energy) and a microscopic one (the particle internal energy).

Going to your example, in the gas expansion, the particle velocity increases ond so does the kinetic energy, while the internal energy (and so the gas temperature) decreases.

I hope to have been clear enough even not using equations.

Best regards and good luck for your work

Carlomagno

Are you having back flow issue in defining the outlet pressure in FLUENT? You have to be selective in inlet velocity also to avoid the back flow.

Dear Mohammed,

I'a agree with James Garry. In my thesis preparation, i tested the position of temperature sensor in cylidrical tube. The accuracy of results depends on the position of the sensor and the flow regime which must have a high Reynolds number value.

I f you need more details about this, i'am ready to help you.

With my best regards!

Dr. Adel OUESLATI