[casey_chen]

Hi everyone,

I'm meshing a very complicated 3D model.
In order to utilize paralle meshing to reduce meshing time, I devided the model into 34 parts, but how do I connect them after meshing?

Anyone have any idea to solve this issue?

Thanks,
Casey.

[AtoHM]

That depends. The standard way is to mesh each part separately and import them into CFX Pre (assuming you use CFX). I have no experience with Fluent but I guess it is the same there. Then you can setup domain interfaces between the mesh parts to connect them back together.



It is, however, also possible to mesh multiple bodies in one go. This would give you 1:1 mesh interfaces, which avoids the otherwise needed interpolation between the parts.

[Carlo_P]

There are at least 3 options.
First: put all bodies in one ansys mesher with share topology:
Results: very nice mesh, 1:1 interface
Disadvantage: 1:1 interface can mean lots of cells, in particolar if the mesh size is very different body to body.

Pseudo serial mesh, i.e. you have to wait that one body is meshed after the mesh of the other one can start.
The change in mesh in one body will afect all the bodies. Every time you have to remesh allthings.


put all bodies in one ansys mesher without share topology: Results: quite nice mesh, but not 1:1 interface.

Disadvantage: not 1:1. Can affect your calculation, and your calculation time.


Advantage: Parallell mesh, i.e. don't have to wait that one body is meshed after the mesh of the other one can start.
The change in a mesh should not have impact on the other bodies. Should(!). You should maintain the same mesh in the other bodies.


Put every body in different ansys mesher and then put all in cfx

Results: not nice mesh, not 1:1 interface.
Advantage: The change in a mesh don't have any impact on the other bodies. Disadvantage: not 1:1. Can affect your calculation, and your calculation time.
lot of time lots in open al the ansys meshing
perfect serial mesh. You can mesh only one body pro time.


I will suggest you the second way. The third way can be usefull only if you will change some parts of the geometry and take invariant other bodies

[casey_chen]

Hi, Carlo_P.

Thanks for your reply!

I took 2 parts out of my model and tested your suggestion.

The first one works perfectly, the caculation converged in FLUENT in the end, but the mesh was created body after body as you mentioned, it may take a long time to mesh 34 bodies, I'm testing it right now to see if the time is acceptable.

However, there appears to be a problem with the second one. When the model is transfered into ANSYS Meshing from Design Modeler, the software automatically created contact regions, and FLUENT recongnize these contact faces as interface, which is not corrected for my scenario.
So I deleted the contact region group in ANSYS Meshing, but in FLUENT the residual keeps bouncing up and down and the solution cannot converge.

Do you know how to solve this problem?
Regards.

[Carlo_P]

Uhm...why they are not interface?
They should be interface, in particolar they should be GGI.


I'm not an expert in Fluent, in particolar in interface, but seems that the behaviuor is correct.


Another aspect..try to impose on the two surfaces on the two different bodies similar cell dimensions, using surface refinements.
Maybe not exactly the same, but very similar. No more than 10 times.

[casey_chen]

Originally Posted by Carlo_P

Uhm...why they are not interface?
They should be interface, in particolar they should be GGI.


I'm not an expert in Fluent, in particolar in interface, but seems that the behaviuor is correct.


Another aspect..try to impose on the two surfaces on the two different bodies similar cell dimensions, using surface refinements.
Maybe not exactly the same, but very similar. No more than 10 times.

Hi, carlo_p!

I’m still working on the first solution you proposed. I put all 34 parts in DM and used the Form new part function to enable shared topology. But when meshing, the Ansys Meshing software always get stuck at:

Part: Part
Status: Meshing faces (64644/69246)

I left it for 12 hours and it didn’t change
Do you happen to know the solution?

Really appreciate your help!

Regards,
Casey

[Carlo_P]

This error appear when it can't mesh that face.
Normally it exit with an error. Did you received any error?


Normally, it appears when the surface has some error.


You can try to clean the goemetry, removing small faces/not correct edges/etc.. in spaceClaim. Sorry, but I don't know how DM works.


Otherwise, you can try to use a virtual topology.


Otherwise you can increase the value in defauturing (I'm not sure how it is called)


Otherwise you can decrease the dimension of the surface cells


Otherwise, you can mesh without share topology.


In order to understand better, you can try to mesh only the surface. In this way, you can try to understand where the problem is.


Hope it helps...

[casey_chen]

Originally Posted by Carlo_P

This error appear when it can't mesh that face.
Normally it exit with an error. Did you received any error?


Normally, it appears when the surface has some error.


You can try to clean the goemetry, removing small faces/not correct edges/etc.. in spaceClaim. Sorry, but I don't know how DM works.


Otherwise, you can try to use a virtual topology.


Otherwise you can increase the value in defauturing (I'm not sure how it is called)


Otherwise you can decrease the dimension of the surface cells


Otherwise, you can mesh without share topology.


In order to understand better, you can try to mesh only the surface. In this way, you can try to understand where the problem is.


Hope it helps...

Thanks for your quick reply!

I’m really stuck here. I tried mesh with and without shared topology both.

When meshing without shared topology, the meshing completed successfully, it took around 15min to create 68million tetra mesh. But while caculating, the solution cannot converge, still trying to figure out what when wrong.

I even tried mesh the original model as one part, but the software crashed in the process, guess the geometry is way too complicated.

Guess I’ll just keep trying both method, and hopefully it will work out in the end.

Thanks again for your help!

Regards,
Casey.

[Carlo_P]

I even tried mesh the original model as one part, but the software crashed in the process, guess the geometry is way too complicated.

Try to semplify it or clean it. Maybe you don't need all tthe parts or you can simplify it.

When meshing without shared topology, the meshing completed successfully, it took around 15min to create 68million tetra mesh. But while caculating, the solution cannot converge, still trying to figure out what when wrong.

Try to have quite the same dimension in the surface around the interface. You should use the same superficial size. Wuold be better if you can monitor different values on each body, (velocity/temperature/pressure) to understand which body goes in divergence first.





But, as always, try to semplify it. Starts with less solids and try to understand where it crashed.