Learning Forum

[Manish Patel]

I am trying to do a saturated pool boiling simulation over a finned surface.

I did initial simulation with coarce mesh and I got some initial results without much problem. I am using VOF method with lee model for evaporation-condensation. 

Now, when I refine mesh and run the simulation with same setup, the solution diverges. I can not understand what is wrong here. I tried choosing first order as well as second order schemes. Also, I reduced time step from 0.001 s to 0.00000001 s. But, still the solution diverges at initial time steps itself. What can be the possible reason? 
 
Coarse mesh

Fine Mesh

Residuals:

[Rob]

Where is the free surface? Where is the heat coming from? 

Looking at the model, the mesh doesn't look to be that fine, and the jump in cell size is excessive for VOF. 

[Manish Patel]

The pressure outlet is at the top. Those protruded rectangular fins are the heated walls. The side walls are adiabatic.  

As I said, the coarse mesh did work without error and got some results with visible boiling. Coarcer mesh also has jump in cell size and works good.

But, as soon as I replaced the mesh with finer one (2 times finer), the simulation diverges. 

The mesh chek for coarse gives following output in Fluent:

Minimum Orthogonal Quality =  1.00000e+00 cell 2886 on zone 23 (ID: 6469 on partition: 2) at location ( 5.21808e-03,  1.42462e-01)

Maximum Aspect Ratio =  5.03856e+00 cell 2924 on zone 23 (ID: 4861 on partition: 1) at location (-5.21808e-03,  1.44615e-01)

The mesh check for finer mesh gives the following output in Fluent:

Minimum Orthogonal Quality =  1.00000e+00 cell 3908 on zone 23 (ID: 23410 on partition: 2) at location ( 3.87500e-03,  1.49462e-01)

Maximum Aspect Ratio =  6.06954e+00 cell 4647 on zone 23 (ID: 19736 on partition: 4) at location (-1.05750e-02,  2.09605e-03)

The zoomed-out and zoomed-in snapshots of the coarse and fine meshes are attached here. 

Coarce:

Fine:

[Rob]

It's hex so the cell checks ought to pass. What's not covered is whether the mesh can capture the gradients. As you refine near the fins you're going to be adding more heat into the adjacent cells (per m3), and as you stretch the cells the VOF gradient calculation is going to become very stiff (hard to converge). 

[Manish Patel]

Okay, so I have found out the problem. The Ansys fluent automatically splits the interior into two parts: 1. interior-part-surface-body 2. part-surface-body. 
So, I used combine > merge zones tool to merge those two. 

So, when I run simulation without merging those two zones, the simulation runs smoothly. 

When I merge those two zones, the simulation diverges. 

Wihout merging, the mesh looks weird as some cells are missing from interior. 

Comparision of the with merging and without merging looks like this:

What should I do now? 

[Rob]

The missing line will be a separate interior face, look for something labelled interior-fluid-fluid1 or similar. Unless you also have a wall and wall:shadow inwhich case there's an unexpected wall in your domain. 

[Manish Patel]

I checked and found nothing labelled as "interior-fluid-fluid1" or similar. It only shows two parts in interior: 1. 1. interior-part-surface-body 2. part-surface-body. 

When I merge them, solution diverges at first time step itself. 

Still could not find any solution. 

Though, I was trying to simulate without merging those two parts, simulation runs smoothly, but I suspect the results are not reliable. 

Please help. 

[Rob]

Can you post some results from before the solver fails? I can't diagnose without seeing anything. 

[Manish Patel]

There are no results with "not merged" mesh simulation. Since the solution fails at the initial 5-10 time step itself. 

[Rob]

So you do have results. I don't expect them to be converged, but they might highlight some issues for you to work on.