3D T-Rex Meshing
 

Hi,

I am trying to create a mesh for a conjugate heat transfer study on a high pressure rotor blade. The geometry is 3D and is quite complex. I have managed to mesh all the fluid parts structured and now I am on to the solid domain. Due to the complexity, I am now trying to finish it using tetrahedrons.

Pointwise has a T-Rex function which is very useful, (combines 5 cells to make a hexahedron) but I am having difficulty in trying to get it to work. It seems to stop a lot of cells from growing and therefore leaving many high angled pyramid cells (more than 10% over 175 degrees, this excludes all pyramid cells generated). I am willing to accept cells up to 175 degrees, but 179 degrees is certainly not right. My belief is that it is caused by the high aspect ratio structured cells within the boundary layer. However, I can't compromise on this due to the low y+ value required for the study.

Does anybody have an experience with using T-Rex on Pointwise and willing to help me to rectify this issue. Ideally, one maximum full layer would suffice, but I am not getting any after many painstaking iterations.

note. I am using a maximum angle of 175 under skew criteria as well as a collision buffer of 2.

I have a lot of experience with T-Rex. There's many things that could be causing the large maximum included angles. You'll have to be more specific and post some pictures of what your geometry looks like and where the problem areas are (ideally take a screenshot while in examine).

The more details you can provide, the better I'll be able to help you.

Hi cnsidero,

Thanks for the quick reply. Unfortunately, I am unable to send the geometry over though. I seem to be having problems occurring near the corners where 2 boundary layer meshes (structured) meet.

Attachment 32364

I have one question about 3D T-Rex too.
Before using it on the full the full aircraft that I want, I thought I should experiment a little bit with a small wing (figure).
There are two structured (right-handed) domains on the surface of the wing and two unstructured at the wingtip with their normals pointing into the block.
I set up the T-Rex to have say max 50 layers, 0 full, I enable the push attribute and set the collision buffer to 2 and the maximum angle to 160. I set all the domains of the wing to be walls with Δs of 1e-5 and I initialize the block. It starts solving, I can see the number of cells, full/max layers etc, but after a while I get the error message "one or more entities could not be initialized".

What am I doing wrong again?

PS
I should probably add that I just noticed that I get this message with the number of max layers is reached. Whatever value I specify for max layers I get the error when the number is reached.
I also tried the same with unstructured domains on the wing; the result was the same.

This isn't a T-Rex issue. There's something wrong with your block assembly. It's usually one of two things: a) a domain in the block is protruding through another domain in the block b) a domain in the block has a corrupted triangulation, i.e. folded cells.

The way to diagnose it is to copy a coordinate location from the message window and set the rotation axis to this location (View, Set Rotation Point ...) to orient yourself to the problematic region. Its usually immediately obvious what the problem is once you do this.

If it's a) you'll have to correct the guilty domain or b) usually re-initializing does the trick.

I was hoping it to be b) (I thought that since the database entities were assembled in a single model the chances for the first would be minimum) so I re-initialized the domains as you said and in the meantime I was going through the tutorials again...
It turns out that I hadn't set up the symmetry plane as "match". When I re-initialized and set the correct BCs the block was solved. T-Rex is innocent, I knew I was the one to blame :p

I think I'll give it a try on the full configuration. The surface mesh is hybrid. Regions where I could have a decent structured domain I did, but other regions are unstructured (on the lower surface of the wing for instance where I used the 2D T-Rex for the leading edge). Is there anything I should pay particular attention to? Any tips or something I should keep in mind?

Chris, I can't thank you enough for your help! I'd really be a lost cause otherwise...

It's nice when the fixes are simple. My own experiences have also been that usually the error is between the chair and keyboard ;-)

For T-Rex:

- keep the transition in surface mesh cells sizes smooth
- avoid using anisotropic surface mesh strategies in concave corners (e.g. wing-root/fuselage junction).
- in narrow gaps or other regions where adjacent extrusion fronts will collide, adjust the surface mesh resolution such that when the layers collide the T-Rex cells (aniso tets) are as close to isotropic as possible, i.e. don't let layers collide deep in the boundary layer.

The general pattern is the quality of the volume mesh is a strong function of the surface mesh. Check out this webinar my former colleague, Travis, and I did when I worked at Pointwise:

http://www.pointwise.com/webinar/scripting/

T-Rex hadn't been released within Pointwise at that time but it includes the best (IHMO) surface meshing strategy for T-Rex use with aircraft.

Happy to help. Keeps my skills sharp too.

The tutorials workbook, the user's manual and Pointwise's (CFDMeshing) and Travis' videos are always the first to read/watch before I ask for help ;) The videos are very enlightening! Then it becomes a matter of how much time I can spend looking for the answer on my own.

I've already taken care of the first pointl; the spacings are about the same from one domain to the other. In concave regions I'll select the domains and assign them to "adjacent" in the BCs menu unless this is a very bad idea. I'll watch the video again and I'll try to solve the block following your tips. I'll let you know of the result.

I think you're misunderstanding the Adjacent BC. That's used when you want to extrude from the domain of an existing extruded block. The Adjacent BC will automatically set the first cell height be taking information from the adjacent block and hence it's name.

Oh I got it. I did the same when I was creating the unstructured domains but I already had another structured or unstructured domain adjacent to the one I was assembling.
How do we treat concave regions then? Leave the BC to "off" so that the cells in that region are created as a function of the growing cells in the surroundings?

No, leave the BC type as Wall. T-Rex can automatically smooth the extrusion normals in concave corners, in a similar spirit to the structured hyperbolic extrusion. As I mentioned, the key to getting good quality cells is to start with a smooth, uniform surface mesh on the domains that make up the concave corner.