How to create Brick elements? 1

[mikechy]

For simplicity, let’s consider flat plate. I seeded surfaces, then selected followings option, and successfully created Quad shells. Now I try to create solids (I expect to create Brick elements), but get Tetra elements (not even Wedge). Is it possible to do it in WF4?

http://files.engineering.com/getfile.aspx?folder=6fb7111b-00be-4ca8-8056-4a396783a492&file=setup.png

Thanks

 

[seymours2571]

Mike,

You can create bricks, but you will have to manually extrude them in independent mode. Mechanica's default solid mesher is a tetrahedral mesher.

However, the tets should give you the same results as bricks if both meshes are sufficiently refined.

Steve

 

[brep]

In Integrated mode, you can go to the AutoGEM-Settings menu and select the type of solid elements that will be autogenerated. By default it is set to only tetras, but from the pulldown you can choose tetra/wedge, or tetra/wedge/brick.

If the automesher detects "thin enough" geometry with "parallel enough" faces, it will attempt to combine tets into bricks/wedges. You might need to experiment with the element creation limits (same dialog box) to force it to create bricks .

Of course as seymours2571 points out, this exercise is somewhat moot, since Mechanica's P-tets probably will give just as good an answer, in much less total time (when you compare the total elapsed time for mesh + solve)

 

[mikechy]

Thanks for replies.

Yes, it can be done with Independent mode. I even remember how nice it was in old-old releases when you could go back and forward between these two modes (Link and Unlink). But now PTC doesn't support the Independent mode anymore, thus such way is out of consideration.

Yes, default solid mesh in Mechanica is Tetra. But 'default' means that it should be other choice and about such option (how to create not tetra, but brick) was my question.

Yes, I remember, how to setup AutoGEM (it was topic in Introduction class - probably more than 20 years ago, but I still remember and use it). The problem is - it doesn't work! I have passed probably, in which release it has stopped to work and thought, that maybe someone found out how to do it in WF4.

"This exercise is somewhat moot" - not at all, for example, if you have huge model, and it should help you to decrease a number of elements in two-three times.

Thanks anyway, others didn't try to answer at all.

 

[brep]

>>> "This exercise is somewhat moot" - not at all, for example, if you have huge model, and it should help you to decrease a number of elements in two-three times.

I agree somewhat, however my point is that the total time to set up the model *and* solve needs to be taken into account. Of course a perfect brick mesh is computationally more efficient, but if it takes many hours/days to achieve this perfection then where is the benefit (since the results *should* be the same...) This viewpoint is the result of my experiences consulting/training/supporting etc on Mechanica as well as another well known "high end" FE code.

At one customer a few years ago they proudly showed me an (admittedly) stunning hex mesh of a gearbox housing - the modal analysis (in an h-code) ran really fast. My Mechanica model took 5 times as long to run... Just as the analyst was starting to look smug, I asked them how long it took to mesh, "oh, about 2 weeks of work". In comparison I had autogemmed mine in 15 minutes.

Depending many factors (model type, size, analysis type...) your mileage may vary

 

[mikechy]

You are right: "Depending many factors ..".
1. If setup to create Wedge and Brick elements could work, it should be the same time as for tetra AutoGEM - I do not think about 'perfect' hex-mesh.
2. If you have a model with about 150K tetra elements, and you have 50+ load cases, and you have to run 5-10 design iteration - then time to run is big deal for you.

 

[seymours2571]

Mike,

PTC still does support Independent mechanica in Proe. When you have your mechanica model open inside proE just goto File menu then click "Independent Mechanica...". It will transfer your model will loads, geom, constraints, etc.

Mike your question was is it possible to create brick elements. It is...you can extrude them.

Most people here are fully aware of the fact that the larger the mesh the longer the run time. And, large meshes can severly prolong the analysis of complex design studies. However, maybe instead of worrying about the mesh, perhaps you should be looking to regions of the model where geometry simplification can be performed. Perhaps regions where the stress values are unimportant, but rounds and chamfers exists. Or, maybe you should focus on limiting the adaptive convergence to specific regions of the model, which would result in increased time savings.

 

[mikechy]

1. PTC doesn't support Independent mode anymore. You can use it in your own risk. Support means to give you the help when you need.
2. When you go in Independent mode, any changes which you'll make, can't be retrieve in Pro/E. So, for design optimization it has no real value.
3. "Large model" means large geometry (component, sub-assembly, assembly). All simplification have been done already - it's common approach for any FEA.
4. Single-pass adaptive isn't good idea because we should use Mechanica result to calculate fatigue life, and in this case: Garbage in - Garbage out.

 

[seymours2571]

1. "Yes, it can be done with Independent mode. I even remember how nice it was in old-old releases when you could go back and forward between these two modes (Link and Unlink). But now PTC doesn't support the Independent mode anymore, thus such way is out of consideration."

PTC's supposed lack of support for the independent mode GUI doesn't preclude its ability to create brick elements. What help would you need on this anyway from tech support?


2. The "unlinking" of the geometry in the Independent mode or more accurately the export of the proe geometry to Mechanica has always been the case. Your statement of the obvious and well known is odd. Plus, independent mechanica can perform design optimizations. It just "depends" on your simulation model requirements.

3. First, I don't even know where you are coming from on this statement. Secondly, "large model" doesn't mean large geometry in numerical simulation (i.e. FEA, CFD, etc) world. "Large model" means a great number of elements or cells. You can have large geometry and also have a low element count. You can have very small geometry, yet huge element counts as in SIF calculations or fatigue crack propagation simulations.

4. What are you even talking about here? Just because you are performing fatigue calculations doesn't eliminate the SPA method of convergence (as if SPA method was the method I was referring to - you could reduce the MPA max polynomial order number if the max order numbers occur in regions of known singularities or non-important geometry). If the stress in the region of concern has converged within your required tolerance then it is converged. Your mentioning of the GIGO principle doesn't even apply to your sentence. GIGO applies to incorrect modeling (i.e. material properties, constraint methods, loading values), GIGO doesn't apply to convergence. You are just throwing around terminology like candy.

Honestly, I don't even know why you posted a question. You apparently already have all the answers. You also obviously assume you are the all knowing power of all things FEA. You should have just posted a question to yourself.

"Thanks anyway, others didn't try to answer at all. " - maybe they thought your post was juvenile and not worth responding to.

Good luck

 

[mikechy]

I've asked straight question: Does someone has experience to create the Wedge/Brick elements in WF4/Integrated Mode. I did not bring any another subjects - I only tried to answer on the suggesting.
>> mechanica can perform design optimizations
We do reliability based design optimization. Mechanica (Fatigue Advisor module) can't do it, at least for our purposes.
>> ... statement. Secondly, 'large model'
It's not the statement. It's an explanation what we have: huge application and a lot of elements.
>> ... SPA method
From our experience, distinction in stresses between SPA and MPA can be up to 15%. Using such data for fatigue calculation could give uncertain result. In this case it's incorrect simulation of fatigue. I've called it GIGO.

Thanks for Good luck.

 

[brep]

OK, just to try and get this back to the original question...

The fact remains it *is* possible to use AutoGEM to (automatically) create wedges or bricks in Integrated Mode WF4. In the top level menu of Mechanica (first get there using Applications-Mechanica) you'll find the AutoGEM menu. Choose "Settings" and you'll get the dialog like the WF4 screenshot I've attached to this post.

Another recommendation is is to tick the checkbox for "Display AutoGEM prompts".

This latter option helps by ignoring the thickness check that otherwise would have stopped wedges/bricks from being formed.

Sometimes it helps to create volume regions to isolate chunkier/curvier areas where only tets will suffice.

Anyway, for those still reading here's a bit of extra info:
1. to address some situations where SPA does not give as good of results at local maxes (rather than just at overall max) Wildfire 5 has much more control over local stress errors (i.e. you can highlight points/edges/surfaces/volumes/components and give them a local stress error target)
2. We are planning in WF6 to offer much more powerful brick meshing in a couple of ways. Firstly with mapped volume regions, and secondly with 2.5D extrusions (useful for glue/gaskets etc)

So you can see that there is still plenty of cool stuff being added to Mechanica. I'll be interested to see how people like the plasticity option in WF5... there's more to come in that area too

 

[mikechy]

brep,
Thanks for good news about WF6. It's long way to wait, but it is good to know.
I can confirm that WF5 is better, and I like many new things in it.
You mentioned "We are planning ...". Are you from PTC? If it's so, I can give you details on Bricks problem...