Learning Forum

[m.shafiee1374]

Hello,

I'm trying to simulate an array of nine cubes using structured hexahedral mesh as shown in the pictures. In this simulation i need high accuracy for flow variables just above the cubes. Also, I'm using k-w sst turbulence model, since it has a good accuracy for both close and far from the walls. After the simulation, y+ values near wall boundaries were way above the recommended value for this turbulence model.

[Karthik Remella]

Hello I would try and stay away from a mesh with hanging nodes. This is a simple geometry - all rectangular blocks. You should be able to generate a good hex mesh without hanging nodes here. The domain might need some decomposing and you will need to perform the share topology operation at the CAD phase. Here are some videos that might help you. This is a 5 part video and I'm only sharing the 1st part.
Regarding your y+ question, after you have fixed your meshing workflow, I would decrease the first layer thickness further. You should be able to estimate, for the Reynolds number in your problem, the desired 1st layer thickness. If I'm not wrong, there are y+ calculators available online.
I hope this helps.
Karthik

[m.shafiee1374]

Hello Karthik and thanks for your response I watched the above tutorials but I have generated the mesh using gambit since I'm more familiar with it. Those hanging nodes are from different bodies and I used interface boundary condition to connect the larger domain with the smaller one, Is this a wrong thing to do?
Also for my second simulation i used the pointwise y+ calculator that gave me 0.008 m for cell height near the walls which also didn't work out. And my domain is quite large (around 250*300*100 m^3) so if i reduce the mesh size even further, I wont be able to solve it using the hardware that I have available.
But I've noticed that if I reduce the surface roughness length (from 0.03 to 0), y+ will also be reduced significantly. So maybe I'm missing something with modeling surface roughness length?

[Karthik Remella]

Hello What y+ value are you targeting? What does the 1st cell thickness of 8 mm correspond to? And which wall model are you using?
Karthik

[m.shafiee1374]

Hello This geometry is actually an array of simplified buildings for which a simulation have been carried out in a published journal using k-w sst turbulence model. Now I'm working on the same geometry and to validate my results I prefer to use the same turbulence model so that I can compare the results. However, since I realized that in order for k-w sst model to have accurate results, y+ value should be around 1, I'm also thinking about trying realizable k-e model with standard wall function so that y+ can be as high as 300.
So I guess my targeted y+ value is 300 unless my results turn out to be wrong.

[m.shafiee1374]

Hello again So I've been working on this problem and tried different domain sizes, mesh resolutions and mesh strategies in the past month all with no acceptable results (I couldn't validate my results). In my latest attempt, I reduced the number of cubes to 3 and used symmetric boundary condition on the sides of the domain to simulate the existence of the other 6 cubes. I also eliminated hanging nodes as you suggested. I'm posting a few pictures of my current mesh here:

But now my solution diverges after around 30 iterations ! Check mesh didn't fail only gave a warning about aspect ratio and my minimum orthogonal quality is around 0.5 and orthogonal skew around 0.4, I also checked my boundary conditions several times but couldn't find any issues. I'm feeling too frustrated with thisWhat is wrong here? Is it a mesh problem?

[Rob]

The jump in cell size and aspect ratio is most likely the cause of your problems. Either decompose the domain or just use poly cells. Don't use the blocking like that.