CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] Meshing Issues in ICEM (

amod_kumar May 24, 2010 13:10

Meshing Issues in ICEM
Hello All,
Though I highly appreciate ICEM CFD, lately extensive usage of this software has revealed lot of weakness in this software. I hope my message reach to the development team:

1. Periodicity and Swept block: Even when periodicity is set on vertices, if an array of blocks encompassing the tow faces of periodic surfaces are marked as Swept, the priodicity gets lost. This is very troubling as you have no other option but the swept blocks to reduce the node count on parallel edges.:mad:

2. The "Pre-mesh Quality" under Blocking menu calculates Equi-angle Skewness of prism element zero. However, mesh quality under edit mesh seems to be OK. This is very time consuming.:confused:

3. The definition of ASPECT RATIO under Pre-mesh Quality and Edit Mesh Quality is on different scale. Why can't both the method reports A.R. on same scale (say either 0~1 or 1~ infinity).

4. The swept block method works well (i.e. generates good quality mesh) only when the associated faces are plane. On curved faces, they tend to make highly skewed hexa/prisms.

5. When blocking is active and user tries to set mesh count on a particular line or surface, the Edge Parameter setting on edges gets changed to some arbitrary number even when Update Parameters button has not been used. This is required in V12 where a surface mesh can be associated to free face of the swept block.

6. Many times when you use UNDO button, the colour scheme of parts under model tree gets changes to same colour.

7. Even when the geometry is a cuboid, it is impossible to generate All-Hex mesh without using Blocking technique. :mad:

With growing timeline pressure, I think it is wiser to recommend my manager to switch to other software such as ANSA or AutoGRID. I see no major development in ICEM since last 3~4 years.

PSYMN June 21, 2010 11:22

Questions, etc.
4 Attachment(s)
Sorry, I was busy traveling the week you send this in and am only now reading this thread.

1) Sorry, I don't understand, please explain some more.

2) I will investigate.

3) Fixed

4) This has had a lot of development work for 13.0, and even back for 12.1. What version are you using?

5) are you saying that changing the curve parameter so that you can generate a patch dependent surface mesh also changes the edge params even though you didn't use the "update parameters" button? If not, please try again to explain. If so, I have never seen this, but will test and will pass along a not for Testing to look out for this behavior.

6) I think this rare undo issue was fixed. If you saw it regularly, do you recall what sort of commands you were "undoing"? I can have someone look into it again just in case.

7) Which method were you referring to (that couldn't give all hexa in a box). Sure, the Cartesian methods, (such as BFCart) will generate an all hexa mesh on a cube (or even a sphere as shown) without hexa blocking.

In these first two, you see the outside and cross section of the sphere.
Attachment 3845
Attachment 3846

In this next one, you see the box around the sphere (it is even easier to get pure hex without the sphere).
Attachment 3847

In this next one, you see with some refinement at the sphere.
Attachment 3848

In this next one, you see if the sphere touches the wall of the box (solved in Fluent)

Other methods like "Hexa dominant" have also been signfiicantly improved and will work also. There is also "Multi-zone" which is automated hexa and will easily take care of it. Even Octree Tetra followed by 12 to 1 tetra to hexa conversion will give pure hexa in a box if your smoothing is turned off and your box size is a multiple of your mesh max size.

For 12.0, we resolved >480 feature requests and defects in ICEM CFD, for 12.1 (6 months later) we resolved over 180 more. Development has continued at that same pace for 13.0 and we have the lowest bug count I have seen in my 9 years at ANSYS. Looking back, I would say that there has been quite a lot of development including two totally new mesh methods (BFCart and MultiZone), as well as significant improvements to Hexa, Tetra and Prism. For instance, between 12.0 and 12.1, we improved octree tetra's memory management by 25 percent and added SMP (parallel) to a few meshing methods and smoothing. At 13.0 we have SMP (parallel) for almost all our meshing methods. Every release has included numerous improvements to CAD imports, mesh editing, output interfaces, defaults and the GUI. If you have access to the customer portal, you can download the past 4 years of update ppts.

I have never heard of AutoGrid (I will look it up later), but recently went to an ANSA 13.1 conference and noticed how hard they were working to copy ICEM CFD. Of course their new Hexa blocking tools have a long way to go before they are competitive with ours. If you do try them out, I am very interested in your opinion.

amod_kumar June 22, 2010 13:00

'[ICEM] Meshing Issues in ICEM'
1 Attachment(s)
1. Periodicity Check for extreme faces of swept block
I cannot post the actual geometry here as the project is very confidential. However, I have tried to prepare a geom in Excel and the meshing situation is attached image (SeptBlock.jpg).

2. The "Pre-mesh Quality" under Blocking: My suggestion is to give an option in pre-mesh quality check to select type of elements for which quality parameters are to be checked.

3. ---

4. I an anxiuosly waiting for the final released version V12. I used "dev version" of ICEM V12 where new features being added are really good.:)

5. I will have to go to office, prepare some test geometry and send screen shots. I will do it in the next post.

6. No. It does not occur regularly. But the last time it happened, it put me in lot of trouble as there were around 50 parts in my geometry!

7. I was refering to "hexa dominant". In fact, we do not have license for BFcart and the user manual also does not highlight the capability of this mesh correctly. The example shown there will not meet the quality guidelines prepared internally. I would suggest you include the figures you shared here in help manual (sphere example). This would have provoked me to test this feature and talk to my manager in case the extra license needs to be procured.

AutoGrid: My colleague is using it and I found him quite satisfied. I have not personally used them. Ditto to ANSA. One of my ex-colleague was a big fan of ANSA and got me interested. However, my experiments with Ansys Classic, Hypermesh, GAMBIT and ICEM CFD will rate ICEM CFD above the remaining 3.

If you ask me to explain the biggest problem I have been facing since last 3 years: "It is very difficult to transition from FINE mesh to COARSE mesh in 3D (swept block is the only option)" and it has let me down many times."
In 2D, it is fantastic.

Thank you Simon for sharing the sphere example.


PSYMN June 22, 2010 14:42

1 Attachment(s)
1) I will create an example and test, but off the top of my head... The periodicity check in ICEM CFD compares all the elements of one part with all the elements of the other. Often users will have periodic nodes, but perhaps not periodic elements. In other words, all the nodes on one side are perfectly matched on the other, but perhaps the line elements on side A are in part "CURVES", while those on side B are in "PER2". The periodic check would flag this as a problem. I will create a defect report to see if we can sort this out.

4) Well no wonder you are having issues if you are using a pre-release of 12.0. The pre-release is usually after we have added a bunch of features, but before we sort out all the bugs. We may have followed that pre-release with 6 months of defect reduction before the actual release. Since then, we have released 12.0 and 12.1. I am using pre-release 3 for 13.0. Is that what you meant? The final release of 13.0 is due out by the end of 2010.

6) If we can repeat a failure, we can usually fix it. however, if we can't repeat it, it is hard to know if we have fixed it. Knowing what commands you "undid" would guide testing to look for the defect.

7) With 12.0 and newer, BFCart works with an ICEM CFD Hexa key. I just took a quick look at the 12.1 help and there are 9 BFCart images to illustrate the affect of various settings.

I also just went and tried Hex Dominant on a box. My box was 10 10 10. I set the size to 10 and hit compute (no other settings). I got a perfect hex mesh. (see below). I tried this in 12.1 and 13.0. I have not got 12.0 installed on my machine any more. See the image...
Attachment 3866

It may be that the past few years of development will take care of your swept block needs. MultiZone (automatic hexa and swept blocks) has had a ton of development work (hundreds of hours) over the past year. Other ways to transition include CGRID topologies or even hanging node refinement (if you are using Fluent) We could probably discuss all that on a new thread though.

Back to ANSA, yes it is pretty good for shell meshing. I would say it was similar to Hyper-mesh for someone doing a body-in-white type application. But it is far behind in terms of hexa blocking. And it only added prism in the past couple years. I am sure they are working hard, but up until relatively recently, it was just considered a surface mesher (prep for TGrid). What was your ex-colleague using it for?

amod_kumar June 23, 2010 07:57

Icem cfd 11.0.1
2 Attachment(s)
4) I am a bit old then. I am still using version 11.0.1. May be it is time to check with my IT for the latest release! If I understand correctly, the latest stable release of ICEM is V12.1?

5) The block edge mesh setting before free mesh generation and after it is attached. It has got reset to "2" which is default no. of nodes.

7) In version 11.0.1, help manual does not cover BFCart (at least the keyword seach with key "BFCart" yields nothing).

You are right! ANSA was used for surface meshing, a prep for TGrid.


All times are GMT -4. The time now is 04:05.