Learning Forum

[BenB]

Hi all,

Trying to perform a multiphase simulation, having some issues A) getting convergence and having the boundary conditions operate properly.

 

 

To add some context, I am doing a multiphase (water/air) simulation. The geometry is above. Water is the primary phase, I have a water inlet (mass flow rate) and the outlet is the whole lower section as you can see, for the fluid mixture water/air. The upper tube goes into the lower section and the wall of the upper tube are set as a wall boundary condition. I patch volume fraction air=1 at the beginning of the simulation.

I use a VOF simulation, Simple, Presto!, I run a hybrid initialization. The mesh has skewness 0.85, relatively even, fine around the walls and at angular sections, and is conformal over the different sections. The layers visible have different permeabilities. The time step is very low (E-6), 45 or so iterations per time step. Basically, I followed a lot of tutorials and tried to replicate the settings for what I was doing.

Issue 1: often have the simulation work for a few time steps, then the residuals will shoot up and it will present with a 'floating point exception'. It will often also say reverse flow at certain boundaries.

Issue 2: the results of the simulations that I have tweaked to make work, do not provide outputs I don't think are what I intended. Please see images below:

 

Left: whole cross section with water volume fraction

Right: upper left corner of the lower section. Clearly, there seems to be water coming into the outlets somehow, despite the fact there is no water coming through the inlet.

 

My general query is to ask:

1. What general features of multiphase simulations need to be understood and improved to prevent 'floating point exceptions'.

2. I know that floating point exceptions happen when the solver is trying to divide by 0 (and probably other situations too), but what does that mean in the context of CFD? I'm a non-expert and I just want to know if there's an easily pictured situation that causes this.

3. What elements of my model (mesh, geometry configuation, the solver method, initialisation, etc) could be causing the odd results pictured? (i.e. the water concentration). Is it poor mesh resolution/shape?

 

Thanks!

Ben

 

[Rob]

From the sounds of it the cells are OK, but you may not have enough of them. Also check aspect ratio as VOF doesn't like stretched cells. Don't use hybrid initialisation: you need to initialise to the conditions at t=0 

[BenB]

 

So as an update; at long simulation times the flow from the outlet gets more and more prevalent. Any tips?

[Rob]

What are you using for the backflow condition on the outlet?

[BenB]

Pressure outlet, normal to boundary, total pressure

[Rob]

And the back flow phase?

[BenB]

Aha, right. Just found that setting. Will run it over the weekend and see on Monday!

 

As a side note; I don't really believe this situation ought to be generating any backflow (from a physical point of view); I know it's about the calculated pressure gradients, so should I be trying to increase grid density close to the walls?

[DrAmine]

Backflow can occur in an iterative process as pressure field is developing even if it does not occur in reality.

[DrAmine]

Backflow can occur in an iterative process as pressure field is developing even if it does not occur in reality.