[vietle]

Hi, I'm currently setting up a steady simulation for combustor (k-w SST model) with Ansys Fluent and I'm setting my mesh edge spacing 1e-6 in ICEM CFD for required y+ value (first cell thickness is 0.001 mm so y+ < 1).


The aspect ratios up to 40000 and the orthogonal quality is very low 0.003.


I've tried to increase the nodes but it didn't change much (I want to do a coarse grid though so I don't want to change the nodes), these two values change a lot only when I make the edge spacing smaller (1e-5 or 1e-4) and ratio but it will make the y+ greater than 1.

Will that have an influence on the convergence? If yes, does anyone know how to fix this?

I will attach images below.

[zacko]

Am I seeing it right, that you only have 5 elements in the vertical direction including your mentioned first element height of 0.001 mm? The expansion ratio is way too big. This will for sure cause numerical problems. You need to increase your node count. It is not only sufficient to make the first element height according to your y+ value. You want to accurately capture the boundary layer flow.

[zacko]

I haven't used ICEM in a while but I remember that it is possible to adjust the node count and keep the first element height the same as you want it to be. The documentation or maybe tutorials should help you

[vietle]

Originally Posted by zacko

Am I seeing it right, that you only have 5 elements in the vertical direction including your mentioned first element height of 0.001 mm? The expansion ratio is way too big. This will for sure cause numerical problems. You need to increase your node count. It is not only sufficient to make the first element height according to your y+ value. You want to accurately capture the boundary layer flow.

I set the nodes to 15. Maybe the expansion rate is way too big (I set the ratio in edge parameter 1.2), I'll try to change it.

P/s: I've asked someone and they said this "Minimum edge length is 10E-6, so the max edge length for a cell needs to be 10E-3 in order to get an Aspect ratio of 1000."

[shereez234]

Originally Posted by vietle

I set the nodes to 15. Maybe the expansion rate is way too big (I set the ratio in edge parameter 1.2), I'll try to change it.

P/s: I've asked someone and they said this "Minimum edge length is 10E-6, so the max edge length for a cell needs to be 10E-3 in order to get an Aspect ratio of 1000."

Just to add, the ratio you set is not the ratio you get, especially if the number of elements on a long edge is under-resolved

[vietle]

Originally Posted by shereez234

Just to add, the ratio you set is not the ratio you get, especially if the number of elements on a long edge is under-resolved

I know the ratio will change, I add the grid details for you guys go see. Don’t know why they can achieve the y+ < 1 with 50 000 nodes (maybe their AR and orthogonal bad though)

[shereez234]

Originally Posted by vietle

I know the ratio will change, I add the grid details for you guys go see. Don’t know why they can achieve the y+ < 1 with 50 000 nodes (maybe their AR and orthogonal bad though)

Y+ does not have anything to do with no of nodes. You can specify a Y+ value of 1 even with 5 nodes. However, in under resolved meshes you should specify a higher wall Y+ like 2-5. As you increase the refinement you can decrease the Y+ with icem mesh feature scale the mesh and tick on scale initial spacing as well.

Since you don't want to increase the number of nodes, then the only approach you have is to play with the distribution function like bigeometric or geometric2,1 or hyperbolic. You may also want to use absolute spacing in wall spacing instead of the default settings.

A true mesh quality is evaluated once you convert your pre-mesh into unstructured mesh and evaluate quality from there (not pre-mesh quality). You may also want to see how your mesh performs and check wall Y+ after solving in your solver and then improve your mesh and wall spacing based on your solution.

I hope this helps.

[vietle]

Originally Posted by shereez234

Y+ does not have anything to do with no of nodes. You can specify a Y+ value of 1 even with 5 nodes. However, in under resolved meshes you should specify a higher wall Y+ like 2-5. As you increase the refinement you can decrease the Y+ with icem mesh feature scale the mesh and tick on scale initial spacing as well.

Since you don't want to increase the number of nodes, then the only approach you have is to play with the distribution function like bigeometric or geometric2,1 or hyperbolic. You may also want to use absolute spacing in wall spacing instead of the default settings.

A true mesh quality is evaluated once you convert your pre-mesh into unstructured mesh and evaluate quality from there (not pre-mesh quality). You may also want to see how your mesh performs and check wall Y+ after solving in your solver and then improve your mesh and wall spacing based on your solution.

I hope this helps.

Thank you so much for very detailed answer, my mistake is trying to achieve y+ ~1 with coarse grid. I’ll do what you advise for my mesh.

Thanks!!