[Shomaz ul Haq]

Dear all,

Hi. Hope are well. I wanted to ask on what basis are the number of nodes selected for a particular length in Edge Params in Pre-Mesh Params? I am using an O-grid for circular double pipe geometry and the length of my pipe is 1 m. In that case what number of nodes should I set for the Edge of "length of pipe". Would be grateful. Thanks.

[Shomaz ul Haq]

Is there anyone who can help me? Would be grateful. Thanks.

[bluebase]

Hi Shomaz,

there is no recommendation for such kind of problem. You have to test out how many cells are enough to achieve your mesh independent solution.
For some special cases even one cell lengthwise could be enough, like laminar couette flows.
You might need the following information http://www.grc.nasa.gov/WWW/wind/val.../spatconv.html

With regards,
Sebastian

[Far]

I would recommend this

1. No of nodes should be enough to represent the geometry well. You can start with 20 -40 nodes on full circle

2. Make mesh Independence study, but generally speaking flow normal to geometry is more important than the flow along the stream-wise direction, unless you are modelling transition from laminar to turbulent flow.

[Shomaz ul Haq]

Thank you bluebase. You are absolutely right. I shall check the link too and will test different number of nodes. Thank you Far. Far you said I should start with 20-40 nodes on "full circle". What does that mean? Is that along the length of the pipe or along any other edge? Also why generally flow normal to geometry is more important than flow along stream wise? Also modeling laminar to turbulent flow is important to me. Would be grateful for answers.

[Shomaz ul Haq]

Dear all. Hope all are well. Waiting for reply. Would be grateful for help. Thanks.

[bluebase]

Hi Shomaz,

i think Far means the circumferencial number of nodes. So, 20-40 elements per 360 degree.
And the second point, why "flow" is more important normal to the geometry... Well it's good to know some basics on discretization. Let's imangine an example.
Think of a parabolic function, plain old f(x)=x^2. Now, take a ruler and try to draw this function only with straight edges. Start with 2 lines only. Then three, four and so on... Do you get the idea? the more lines you can use, the more the functions actually look like the original parabolic function. But just two lines are not enough.
The analogy to cfd is, you usually linearize the equations which describing the flow. Those are typically non-linear. So, to have a good representation of the flow you put many "straight lines" where the flow changes rapidly. Or with other words, put many elements where gradients are very high.
Where do you expect high gradients in your flow? Where does the flow/velocity/pressure changes rapidly?
Search for keywords like "boundary layer flow" in your text books or google.
Then you'll probably understand why you should put more cells normal to a wall than you probably require downstream (for laminar flow). Though, in turbulent flows you probably also need many more elements downstream than a laminar flow would require.

With regards,
Sebastian

[Shomaz ul Haq]

Thanks a lot Sebastian. Now I know how to select number of nodes along the radial edge of an O grid but how can I select number of nodes for circumference? Is it done using the same Edge Params in Pre-Mesh Params? Would be grateful. Thanks.

[Far]

Originally Posted by Shomaz ul Haq

Thanks a lot Sebastian. Now I know how to select number of nodes along the radial edge of an O grid but how can I select number of nodes for circumference? Is it done using the same Edge Params in Pre-Mesh Params? Would be grateful. Thanks.

yes, use edge mesh parameters