Termination of T-rex layers against adjacent domain

Hi, 

Thank you very much for sending your question. 

You have set the shown bounding domain to a match T-Rex boundary condition; this is the correct way to go. However, could you please clarify if you have check on the Push Attributes option inside the T-Rex panel for the block? This will modify the bounding domain(s) to ensure a nice match between the T-Rex volume surface and volume grids. 

1. Select the block
2. Grid, T-Rex
3. Set the desired number of T-Rex layers
4. Make sure the Push Attributes option is checked on
5. Solve tab
6. Initialize the block

For additional information on this feature, please check the Fidelity Pointwise User Manual: 

1. Help, User Manual
2. Grid, Solve
3. Unstructured Blocks
4. T-Rex Tab
5. Search for Push Attributes (You can use Ctrl+f for search box)

Please let me know if this helps.

Best regards,

Dear Claudio,

Thank you for answer. I am aware of the push attributes option and it works to some degree: it improves the mesh of the block. However, it does not answer my question fully. This is because now there is a discontinuity in the number of inflation layers between this block and the adjacent block, see the image below. Therefore, I would like to not use the push attributes option and have a smooth transition in the number of inflation layers from block one to block two (as in the bottom picture of my original question). I hope this makes clear what I try to achieve.

Hi, 

Thank you very much for the additional comment, this clarifies things a bit. 

The behavior you see in the image is caused by the requested number of layers on the left and right block to likely be different and to both blocks probably being initialize independently (i.e., one at a time). I would recommend to reduce the requested number of layers on the right block to match what was generated on the left block. Unfortunately, it is a bit tricky to get a full understanding of the entire volume grid and its set up from a few images. 

That being said, please note that the recommended approach to obtain a smooth transition between the two block is to set the interface domain to a Match T-Rex boundary condition for both blocks, to check on the Push Attributes option for both blocks, and lastly to initialize both blocks at the same time. Would you mind to please give this approach a try?

Looking forward to your additional comments. 

Best regards,

The reason I split the mesh in two blocks here is because I wanted to have two different number of layers, so setting both blocks would remove issue, but also does not fulfill my requirement. The reason I set the left block to a lower number of layers is because the automatic termination of the T-rex algorithm (triggered by the collision prediction) creates highly skewed cells. Therefore I reduced the number of layers such that it creates only full layers. However for the right block, I would like to have the  larger number of layers again such that the transition to the isotropic cells is smoother.

I tried initialising both blocks at the same time (by selecting both blocks and hitting solve), but in fact this still solves both blocks after each other and I end up with same result as before (see image below).

So what I am looking for is a nice transition from X T-rex layers to X+5 T-rex layers. In other meshing software such as Ansys meshing or Fluent meshing this can be done so I expected something similar to be possible with Pointwise.

Hi, 

Thank you very much for the clarification on what your end goal is and how you are approaching this problem. Unfortunately, it wasn't clear to me that you actually wanted that difference in number of T-Rex layers across the interface; now it is. 

As you noted, when you enter the unstructured solver with both blocks, Fidelity Pointwise will initialize one block at-a-time; this is the expected behavior. My recommendation for using this procedure was considering setting the Max. Layers parameter to something large enough so the T-Rex solver can grow as many layers as possible and optimize the transition region (my apologies; perhaps I should have mentioned this explicitly). Note that, if each block is set a different number of layers, entering the T-Rex solver with both blocks will likely give a similar result as when initializing one block-at-a-time. 

So, we are not sure at this point what is causing the jump in T-Rex layers that you are experiencing (original post). Is there any chance you could share your Fidelity Pointwise file with us to take a closer inspection? Perhaps at this point it would be useful to move from this forum into a regular support ticket in the Cadence Online Support platform. Would that be possible?

Until this issue can be investigated thoroughly, there are some potential workarounds that you could try:

Option 1 (easy change): You mentioned that the reason why you need two different number of T-Rex layers on both blocks is due to skewed cells triggered by the collision prediction. Have you tried increasing the Collision Buffer parameter to give the iso cells in the region a bit more space? 

Option 2 (easy change): You can use the set up that leads to the third image you shared (i.e. right block looks great and there is a discontinuity on the left block) and set the Iso Seed Layers parameter on the left block to 4 or 5 and then re-initialize it. Note that this will affect the entire left block and not just the region close to the interface domain. 

Option 3 (more involved): In your original setup (where the T-Rex jump happens), you could split the left block (Edit, Split) at exactly the X number of T-Rex layers where T-Rex stops. The objective here is to split the interface domain (unfortunately there is no way to just split the domain at this time; the entire block needs to be split). Once the interface domain is split, please re-initialize the right block setting the "lower" interface domain to Match and the "upper" interface domain to Off and make sure the Push Attributes option is turned off. If this work, you can re-join the left block (Edit, Join)

Looking forward to your additional comments. 

Best regards,