[cfdbooks]

I try to create cut plane for nfaced cells, but it displays only the outline of the plane not the full plane. So I can not create contour on it. How can I get a full cut plane for nfaced cells?Thanks in advance.

[Marina G]

Hi,

There can be a number of reasons why this is happening. Please follow this checklist and verify each point, in order to understand where is the issue.

1. When you create the clip plane, you need to have the 3D fluid domain part selected. Make sure that this is indeed the part you are selecting, and not, for example, a part that is only the boundary of the fluid domain
2. What kind of elements are composing this fluid domain part? Are they elements that are recognized as 3D, or 2D? To verify this: select the part you want to use as parent part for the clip, go into Query -> Show Information -> Part. This will tell you which element types you have.
3. Try to change the part representation before creating the clip. Select the part and click the "part element settings" icon at the top-right. Set, for example, 3D border, 2D full as your representation.
4. After you've created the clip part, check its representation to make sure you are visualizing the elements correctly - same steps as in 3.

If none of these explains what you are seeing, please let me know what you find in point 2) and the exact version of EnSight that you are using.

Thanks. Best Regards,

Marina Galvagni
CEI Software Support Engineer

[cfdbooks]

Dear Marina,
Thanks for your help. When I query the mesh information, it shows the element type is nfaced. In fact, when I make a cut plane for a single nfaced element, I get an outline of the clip. When I make a cut plane for nfaced elements which are bounded by a cube, I get the points cloud of the clip. There is no use to change part representation.When I switch clip domain type from "Intersect" to "Crinkly", every thing seems reasonable. My Ensight version is 10.1.2(b).

[cfdbooks]

In fact, if I want to make a cut plane for a cube, "hex8" format gets the real cut plane and "nfaced" format gets only outline of the cut plane.

[Marina G]

Hi,

It sounds like your model contains nfaced elements that EnSight can not correctly handle.
In attachment, a simple model with a cube composed by nfaced elements. When I clip it using the plane tool on EnSight 10.1.2(b), it works correctly - I get a 2D part composed by Tria3 elements. Please load this model and create a clip. Does it work correctly for you as well?

If yes, then the problem is in the way EnSight handles the specific elements you have. In order to be able to fix this, we'll need to give a look at your model. Would it be possible for you to send us your dataset - even just the part that creates a problem is fine -? You can use the drag&drop window at the link:

http://www.ceisoftware.com/support/

Thank you.
Bests,

Marina Galvagni
CEI Software Support Engineer

[cfdbooks]

I have sent you the dataset using drag&drop windows at the link(ensight.case and ensight.geo). The vertices on the different nfaced element faces which have the same coordinates don't share the same vertex index. Maybe this is the problem.

[Marina G]

Hi,

Thank you for your data. It appears that your model part is composed by N-sided elements. N-sided are 2D elements, not 3D. If they were 3D, then they would be called N-faced.
If you do a Clip with the plane tool and Domain: Inside, you cut the model in half. In this way, you can see that your model is not composed by 3D parts, but it is a shell - only a surface. So, it is to be expected that when you clip it you obtain a line - indeed, the clipped part is composed by only Bar2 elements, that are line elements.

Does this make sense?

Bests,

Marina Galvagni
CEI Software Support Engineer

[cfdbooks]

Thanks! You are right. I will upload the nfaced files(more than 500MB,nfaced.geo and nfaced.case).Please check volume part.

[cfdbooks]

The files have been uploaded to your support web. I uploaded the files twice. The first time was failed. So please choose the larger file. Thanks.

[Marina G]

Hi,

Thank you for the dataset. Now it is clear what is happening.
Your model is composed by elements with a bad connectivity. In particular, the elements are formed by multiple faces that are not linked together; instead of being a single volume element, each element is a collection of not-connected faces. Moreover, the elements are not connected to each other: each one of them is a separate "volume" part.

Here is how you can see this. Select part 4 (your volume part). With the calculator, you can calculate the function EleMetric(face count), that returns the number of faces per element. If you then calculate its average (with the calculator function StatMoment(average)), you get a value of 6.3814. From the Query -> Show Information -> Part, you get that this part has ~ 1.5 million elements. If you assumed that each element is a quad (because the number of faces per element is ~ 6), then each element should have 8 nodes. Let's assume that each element stands alone: this would mean that each element has independent nodes, as no nodes are shared between different elements (if they were, the two elements would be connected). This case would give you the highest number of nodes you can have, as there are no shared nodes between elements. In this case, you should have ~ 1.5 mill X 8 = 12 million nodes on the part. Instead, according to the Query -> show information -> Part, you have 19 million nodes! Your part has definitely too many nodes. This means that not only the connectivity is wrong, giving you elements that are not connected to each others, but also that some elements have multiple nodes on the same vertices for the same face.

Ok, let's now see one element more in detail to see better what is happening. Extract from your part 4 the element with element ID = 1, using the Edit -> Subset feature (in attachment, snapshot of its edit window). This gives you a new part that is only one element. Visualize its nodes (in attachment, snapshot of how to do it). Now, you get that the element is clearly a Hex, but has 24 nodes instead of 8 (you can verify this with the Query -> Show information -> Part). This means that at each node you are actually assigning multiple node IDs. When this happens, the faces are not connected, but are considered disconnected. If you want two faces to be connected, they need to share the node ID, not only to be at the same (x,y,z) position. Indeed, if you save a .case gold file that contains only this element and look at its geometry file, you get the following connectivity:


4 1 2 3

8 5 6 7

12 9 10 11

16 13 14 15

20 17 18 19

24 21 22 23


So, you have 6 faces (6 rows), each one of which has different nodes! The result is that they are considered disconnected. You need to have repeating node IDs between faces to make them share the same node, and therefore be connected.

I hope this clarifies what is happening - the problem is on the connectivity side. Please let me know if you have more questions.

Would it be possible for you to tell us from which solver this data come from? If this is a commercial software, we'll need to contact their devs group to make sure their exporter routines in EnSight Case gold are correct.

Thank you.
Best Regards,

Marina Galvagni
CEI Software Support Engineer

[cfdbooks]

Hi,Marina,
Thanks for your warm-hearted help. I know what you mean. The source code of cartesian grid generator was written by myself. It is cumbersome to define the vertices globally. The output file format is similar to NASA's cart3d which defines the grid vertices locally.

[cfdbooks]

Although the cut plane is not displayed correctly, the flowfield seems reasonable(on wall nsided surface and on crinkly cut plane ).

[Marina G]

Hi,

I'm glad that the results you are getting are correct, despite the format not being exactly what EnSight expects. Please note that because of this difference in the format, EnSight needs more memory to load the data (as the number of nodes is higher than what is actually needed), and this can lead to some slow down / running out of memory. If your model is not very large, this effect could be negligible, but for large models you could have some problems / significant slow-downs (in which case you might want to consider changing your exporter to fit the .case gold format).

Let us know if you have any additional question or problems.
Best Regards,

Marina Galvagni
CEI Software Support Engineer

[cfdbooks]

Thanks again for your help!

[cfdbooks]

Hi,Maria,
I still have one question. If the dataset I sent you is exported to Ensight as hex, the clip plane can be created correctly. If it is exported as 'nfaced',the clip plane can not be created correctly. Maybe there is something Ensight can do to improve the clipping method for these special 'nfaced' elements . Of course, changing data format to fit Ensight is the best way. But sometimes it is very difficult.

[Marina G]

Hi,

Here is one possible thing you can do. There is an environmental variable that you can set before opening EnSight, that will tell the reader to build a hashing table for the node positions. This will help the reader recognize when the nodes are at the same (x,y,z) position and merge them. Note that this has a number of side effects, mainly:
- EnSight will become really slow at loading the data. This is because it needs to build the hash table, which is a slow process. Moreover, the time required for building it grows more than linearly with the number of nodes.
- The hashing table is build using the (x,y,z) in single floating precision. This means that it is still possible that two nodes in the same position are actually not recognized as coincident (numerical errors). So, you can still have holes in the clips.

So, as you said, the best solution is to change your dataset to be "real" .case gold format. Use this environmental variable as a back-up when you really can't change the dataset.
The env. variable is ENSIGHT_DUPLICATE_NODE_CHECK and needs to be set to 1. You'll need to have installed EnSight version 10.1.2(b) or newer to have it recognized.

Let us know if you have questions.
Bests,

Marina Galvagni
CEI Software Support Engineer

[cfdbooks]

I will modify the format to Ensight. Thanks.