Fluids

Gaz separation/three phase environment

• javat33489
Subscriber

Hi all. I want to simulate a three phase environment: oil, gas and water. Water and gas 0.1 kg/s, gas 100 kg/s. There is an entrance and an exit. The separator is modeled. The goal is to separate the gas so that it passes through the flow meter. And also measure the gas pressure at the top:

1. I set the multiphase environment to VOLUME OF FLUID. I did not set surface tension for material pairs

2. I created three materials GAS and OIL with constants and took water from the FLUENT base

3. At the input, I set the consumption for each material

4. I left 0 at the input and output of PRESSURE GAUGE

5. I use the Ke turbulence model on base settings

6. Hybrid initialization and first start on PISO

7. I use gravity and transient analysis. Iteration time 300/1e-6 sec

The task is not solved, this is the schedule, why? Tell me your advice and how best to do it? I did according to the tutorial.

• Rob
Ansys Employee

If the three phases are initially mixed and then separate you don't want to start with VOF. Depending on the volume fractions have a look at DPM and Eulerian/Mixture Model.

• javat33489
Subscriber

I thought about it. Okay, I'll take a look. What about set costs? Is it possible to do as I did all in one INLET

Can I get more information and advice from you? How would you count? Can you write a short sequence?

• Rob
Ansys Employee

That depends on which model you use: VOF needs a unique inlet for each phase. Read through how the models work (Theory Guide) and add some more information about volume fractions etc.  Staff are limited in how how we can say, you need to do the work and we can (mostly) point out things you've missed, got wrong and (in rare cases!) done correctly. ;)

• javat33489
Subscriber

I understood about VOF I thought about it too. What else is wrong with my calculation?

• Rob
Ansys Employee

Hard to say until you let me know volume fractions etc.

• javat33489
Subscriber

Liquid:

50% water (1 kg/s)

50% oil (1 kg/s)

gas 100 kg/s

98% - gas

1% water

1% oil

Am I right?

• Rob
Ansys Employee

That'll depend on the gas density. Given the volume fractions why are you trying to use VOF?  Do you expect to be modelling every droplet with 5-10 cells across it's diameter?

• javat33489
Subscriber

Let's start over.

KEpsilon turbulence model

Energy included

Three media oil gas and water

Mass flow:

Water 1,44 kg/s

Oil 1,44 kg/s

Gas 0,1 kg/s

1. Decided to use MIXTURE.

I have attached some important screenshots:

There is no convergence, for some reason the graph behaves like this. What could be the problem?

2. Next, I decided to try the Euler model.

I have attached some important screenshots:

Convergence behaves like this, every 20 iterations there is a jump, is this normal?

• Rob
Ansys Employee

As I can't see if the Mixture model is transient it's hard to judge.

For the transient solver there is a spike in residual at the end of the time step as time is updated and the solver begins to converge the new step. However, if you're seeing the spike at the same rate as the max iterations you need to check each time step is converged. If it's not the results will need to be treated with care as they may not be correct.

• javat33489
Subscriber

What else can be done? I will also try to solve stationary analysis.

• Rob
Ansys Employee

I think you also need to review the DPM model if the liquid fraction is that low. With multiphase models the hard part is understanding how the device works to then choose the best model.

• javat33489
Subscriber

I had an error. Water + oil 90% and gas only 10%. I counted everything, but there are always bursts every iteration.

• Rob
Ansys Employee

Can you show some images of the geometry & mesh. How well mixed is the oil & water? Ie do they need to be considered as separate phases or can you assume it's one liquid?

• javat33489
Subscriber

Water, oil and gas enter the product in one stream, so I indicated all of them in one inlet.

when I corrected that oil + water is 90% and gas is 10%, the picture became better, but contyiunity always grows

the geometry is a few pipes with risers in which the gas and water are separated, i will attach these pictures later and the mesh too

• javat33489
Subscriber

Maybe it's all about the network. First, I made the grid in the WORKBEnch prepost. Now I have made a grid in the fluent itself. And this is what happened with the MIXTURE model:

This is the stationary mode. But it takes me a very long time to solve on my PC. After the solution is I will write here and attach screenshots.

• Rob
Ansys Employee

Mesh does make a significant difference if the mesh is rubbish compared to good: ie a poor mesh may work in single phase but not with multiphase models. Similarly, Fluent Meshing uses poly cells and is often able to give a better mesh as the built in algorithms tend to override poor sizing choices! Check the cell count, you want to be running 250k-1M cells per core (or node, I've lost track of the multi-cpu terminology) and also check you've got sufficient RAM.

• javat33489
Subscriber

I am using 5M cells.

I haven't completed my task yet. As soon as I count, I will send screenshots of the results and we will discuss them.

• javat33489
Subscriber

That’s what I did. The gas separated perfectly. What about convergence? This is fine? Energy surged strongly

• Rob
Ansys Employee

Looks reasonable, but you need to fully understand what's going on. Depending on how much the pressure changes you may not need to worry about such high pressures and just adjust the density.

• javat33489
Subscriber

Thank you! My concern is that gas remains in the upper part of the lower pipes.. Why?

The mixture is fed from the bottom, then the gas goes up and the liquid goes down.

• Rob
Ansys Employee

How long has the model run for? It's common to get gas stuck in systems, without a cross section how have you checked it's a sizeable hold up?

• javat33489
Subscriber

I didn't cut. Only visually assessed the overall picture. If that's okay, then fine. Thank you.

• Rob
Ansys Employee

I'd have a proper look at the results: an incomplete picture can lead to incorrect conclusions.

• javat33489
Subscriber

I can't post it here. Maybe I'll send you a picture in a personal?

• Rob
Ansys Employee

Everything has to be public, it's the only way Legal would let staff work on here to avoid export law complications.