CW7 - improper mesh around node n

[macPT]

That's the message I receive from processor when it tries to run a solid mesh study.

The mesh is created with no problem reported and, visually, I don't find any part of the mesh with a suspicious appearence. All the surfaces are meshed.

So I should know where is node n to be able to check what is going on. But, how to do that? How can I locate node n? Do you know any tools to analyse mesh or read CW files?

Regards

 

[JNR]

I'll ask our CosmosWorks/CosmosM dudes and report back Monday.

I was - and he did. So at least I didn't get coal.....
OK, OK, It's a reference to my holiday sig. "Be naughty - Save Santa a trip..."

 

[agbaca]

macPT,

It’s been awhile since I ran CW 7 but if I recall there was no way to “probe” your model at the node level. The newer versions of CW are slightly more user friendly when trying to debug your mesh or “probing” your model for nodal stresses. Beware that CW will give you some type of mesh unless if it fails miserably. A good rule of thumb is to never exceed an aspect ratio of 5 for elements sides. I suspect somewhere in your model you have an element (probably many) that is exceeding this ratio---somewhere probably near a hole or transition from thin to thick section. If you are versed in the “geostar” module (and have it) you can export your model (.geo) to geostar and interrogate the nodes by using the node plot command. If you’re not versed in geostar, I would recommend that you re-mesh your model using the “Apply Control” option (in the mesh menu). Pick the critical areas that I mentioned above and this should take care of your problem.

Good Luck,

Allen G. Baca

P.S. I’ve also seen your error message when meshing assemblies that have a slight interference (.00001) between parts. (Due to import tolerances or sloppy modeling)

 

[macPT]

JNR - thanks. I´ll wait for your help

agbaca - we can probe the model in the post processor. before running the case, there is no probe tool available (as far as I know).

I have tried to play with mesh configuration, element size and jacobian chack, but the result was the same in all meshes. I think I have an error in the model (it is an assembly so there's maybe some gap or interference that is causing problems - as in your PS) but still I can't find it. If I knew where the node is, I could focuss my efforts correcting the error. Otherwyse its like shooting a gun in the dark.

But the idea to export mesh to another FEA it's maybe a good idea. I will give it a try.

Thanks

 

[macPT]

I've found a way to solve the problem!

I've created a micron sized gap between some components (a gap lesser than the value from wich CW would not consider contact)and voilá! it worked! SW should have a small interference somewhere that I couldn't figure out.

From the deformation and stress distribution, this gap didn't have any impact on the results. So I manage to fool CW well.

Even well folled, the fact is that CW is very poor in tools for checking and debbuging. I remember the old days of Algor, where many of these tools were provided (like manualy modify the stiffness matrix!).

JNR, I am still wayting for your report.

Thanks

 

[JNR]

Sorry it took a while - we work a 9/80 arrangement. 9 hours M-Th then 8 alternate F and day off on the other F's (and it's not the same Friday for everyone!!!).

Anyway here is the answer from our Cosmos guru.

1. True - you cannot show the node numbers in CW. You need full-up CosmosM for that. (We have both.)

2. Try: Preferences - Mesh - Jacobian Check - At Nodes.

3. If the problem persists use the Sparse Solver in stead of FFE New or FFE Old.

I was - and he did. So at least I didn't get coal.....
OK, OK, It's a reference to my holiday sig. "Be naughty - Save Santa a trip..."

 

[macPT]

Thank you for your tips JNR.

I think my problem is solved for now, but I will try to reproduce the error and see if those tips fix the error and have equivalent results.

Regards