[swtbkim]

Hi all!


I made a boundary layer using t-rex, and imported it in fluent
Thing is that t-rex is highly skew grid, maybe the highest one.
Then boundary layer can be deformed I guess.


What happens when I smooth the grid that is made by t-rex?
Is it ok to repair the grid in Fluent which is made by t-rex??


Thanks in advance!

[cnsidero]

Originally Posted by swtbkim

Hi all!


I made a boundary layer using t-rex, and imported it in fluent
Thing is that t-rex is highly skew grid, maybe the highest one.
Then boundary layer can be deformed I guess.


What happens when I smooth the grid that is made by t-rex?
Is it ok to repair the grid in Fluent which is made by t-rex??


Thanks in advance!

Did you combine the anisotropic tets when you exported the mesh? In the File>Export>CAE... panel, there is a check box that says "Combine anisotropic tetrahedra". If not, you should enable this option.

[swtbkim]

Thank you for your suggestion, Chris.
But I have already done, of course.
I don't have any problem about it. I'm just curious about theoretical background in mesh smoothing.

T-rex tets become prism layer if I check the option.
Skewness of prism layer is not high?

How about other criterion for mesh quality e.g. aspect ratio?

[cnsidero]

Ok, good. You didn't specify so I was just checking. Sounds like you need some T-Rex parameters suggestions for improving it's quality. As it sounds like you've discovered, often its not the T-Rex (or their combined prism) cells themselves that are poor quality but the iso-tet cells in the vicinity of T-Rex extrusion front. Here's a few quick tips.

The smoothing parameter has influence on the transition on the local cell height. The instances where I modified this parameter and seen an appreciable effect is where I specified different initial cells on adjacent domains. The transition from from one to the other can be made more or less gradual. So it's likely that this parameter won't help you in general.

Without seeing your grid at all, here's some basic suggestions that I usually start with:

Advanced:
Collision Buffer = 2
Skew Criteria:
Max Angle: 160
Centroid: 0.7

Try those and see if that helps. Beyond this, you'll likely need to post a picture of the mesh, highlighting the areas you believe to bad, to be able to give you more specific advice.

[swtbkim]

Originally Posted by cnsidero

Ok, good. You didn't specify so I was just checking. Sounds like you need some T-Rex parameters suggestions for improving it's quality. As it sounds like you've discovered, often its not the T-Rex (or their combined prism) cells themselves that are poor quality but the iso-tet cells in the vicinity of T-Rex extrusion front. Here's a few quick tips.

The smoothing parameter has influence on the transition on the local cell height. The instances where I modified this parameter and seen an appreciable effect is where I specified different initial cells on adjacent domains. The transition from from one to the other can be made more or less gradual. So it's likely that this parameter won't help you in general.

Without seeing your grid at all, here's some basic suggestions that I usually start with:

Advanced:
Collision Buffer = 2
Skew Criteria:
Max Angle: 160
Centroid: 0.7

Try those and see if that helps. Beyond this, you'll likely need to post a picture of the mesh, highlighting the areas you believe to bad, to be able to give you more specific advice.

Thank you very much for your considerate reply, Chris!

1. What does collision buffer mean?
2. Centroid and equiangle skewness is not bad, but equivolume skewness is main problem in my case. what does this mean?
3. How can I attach a image file?? URL address is needed if I click 'insert image'. Is there a way to attach without URL address?

Thank you!

[cnsidero]

1. What does collision buffer mean?

Generally speaking, collision buffer controls the size of the gap between colliding layers. A higher value means more room for iso tets to be inserted - which generally results in better quality iso tets - but at the same time prevents the advancement of layers. It's a bit of a trade off. I find the default (0.5) a bit too low and prefer a value of (2.0). You'll find Pointwise people recommending a similar value in recent T-Rex related webinars and videos.

More specifically it is a normalized distance the extrusion uses to detect layer collisions. The distance is computed by adding the current layer height to the current layer height * collision buffer. E.g. for collision buffer of 2.0, the distance will be S = ds_i * (1 + 2.0) = 3 * ds_i, where ds_i is the height of the i'th layer.

2. Centroid and equiangle skewness is not bad, but equivolume skewness is main problem in my case. what does this mean?

IMHO, equivolume skewness is not a very useful metric because it is a measure of how "equilateral" or "isotropic" a cell is, yet for boundary layer techniques, we are intentionally creating non-equilateral or anisotropic cells. So in the case of a T-Rex mesh, it's a misleading metric. The two I typically use for T-Rex meshes are maximum included angle (< 165) and centroid skewness (< 0.7).

3. How can I attach a image file?? URL address is needed if I click 'insert image'. Is there a way to attach without URL address?

There is a "Manage Attachments" functionality when creating a post from which you can upload images.

[swtbkim]

I'm making a grid for CFD simulation near the car and there are two bad parts.

1. Harsh geometry between tire and road.
Road and tire are nearly tangent to each other, so the equivolume skewness is up to 0.9997
In fact, this has nothing to do with t-rex, and I decomposed the geometry and managed to decrease it below 0.95

2. three orthogonal domains which have their own t-rex tets.
The equivolume skewness is up to 0.999 near the intersection point of the three domains.
I changed the dimension of the connectors, and it was OK anyway.



Collision buffer may be very useful. I'm gonna try right now!
It is hard and inconvenient to decompose all the bad part
It would be perfect if collision buffer work

[cnsidero]

Originally Posted by swtbkim

1. Harsh geometry between tire and road.
Road and tire are nearly tangent to each other, so the equivolume skewness is up to 0.9997
In fact, this has nothing to do with t-rex, and I decomposed the geometry and managed to decrease it below 0.95

Yes, I have encountered this a well with automobiles. I used a similar approach as you - created a fillet around the tire/ground intersection line.

You may also considering refining the grid spacing around the base of the tire.

2. three orthogonal domains which have their own t-rex tets.
The equivolume skewness is up to 0.999 near the intersection point of the three domains.
I changed the dimension of the connectors, and it was OK anyway.

Again, the equivolume skewness is not a helpful indicator. I would suggest looking at maximum included angle and centroid skewness. I try to stay under 165 for max angle and 0.7 for centroid.

[swtbkim]

1. Where is the base of the tire? I'm not native in English, so I don't know where it exactly is.

2. I understood what you suggested.. But Fluent keeps warning that the skewness is larger than 0.98
Is it OK if calculation can be started even though skewness is larger than 0.98?
I didn't try, but the calculation can be even started?

Thanks a lot!