Wedge elements

[BioMes]

Hello everyone!

What do you think about the performance of wedge elements? I know that element choice is case-dependent but I’m looking for some general recommendations that can be followed in most cases before going into details and mesh convergence studies (like the recommendation that hex elements are normally preferred instead of tetras). I’ve read conflicting opinions - one saying that wedges are generally better than tetras but worse than hexas and one claiming that they can be even significantly worse than tetras due to inaccuracy in interpolation.

My question is mostly about second-order wedges because, from what I know, first-order ones are too stiff like linear tetras and thus should be used only as fillers.

So how would you compare second-order wedges to tetras and hexas when it comes to typical structural analysis applications?

 

[rb1957]

I would run side-by-side single element tests on both elements.

"Hoffen wir mal, dass alles gut geht !"
General Paulus, Nov 1942, outside Stalingrad after the launch of Operation Uranus.

 

[BioMes]

I was thinking about some benchmark studies but the problem is that the number of those elements used to mesh the same geometry (even a very simple rectangular cantilever beam) will always differ depending on their type. Single element tests make more sense but likely only for wedge vs hexa (a single tetra sounds just weird) and the shape will still be different, making it hard to compare the results. Also, the range of conditions that could be tested this way would be rather small - limited only to basic forms of loading.

Thus I’m asking for some general recommendations that may come from software documentation (Abaqus manual doesn’t cover this) or your experience.

 

[SWComposites]

2nd order tets and hexes are generally ok. So 2nd order wedges should also be ok for same applications.

Also depends on the specific element type, particularly for thin beam/shell bending models. See the Abaqus documentation for element performance in bending.

You could run a cantilever beam test model; start with hex mesh, then convert each hex to two wedges of same element type.

 

[BioMes]

Unfortunately, the Abaqus documentation doesn’t mention anything about the performance of second-order wedges.

I guess that comparison with tetras will be more tricky. I can just try to put 5 tetras for each hexa but does it make sense?

 

[rb1957]

single element tests are not difficult to set up.

"Hoffen wir mal, dass alles gut geht !"
General Paulus, Nov 1942, outside Stalingrad after the launch of Operation Uranus.

 

[SWComposites]

Bio - why do you care about a comparison to tets? Just use the second order wedges if you can’t mesh with second order hexes.

 

[IceBreakerSours]

Tets are awful, sure. However, if the goal is to automate workflows that involve challenging (e.g., organic/3D-printed) geometries without sacrificing accuracy, there isn't a whole lot out there that can replace those awful tets. The trade-off is to throw lots of tets, let the hardware do the number-crunching for you while you spend your precious time doing analysis and not meshing. I don't fully subscribe to this school of thought myself but I know of applications where this school of thought has a strong argument.

I would take a step back and ask myself - What is it that I really want? One or a few aesthetically pleasing meshes that are likely to give me better "performance" but point-wise predictions or an automated workflow in which I run an array of scenarios and get statistical forecasts for the quantities of interest?

*********************************************************
Are you new to this forum? If so, please read these FAQs:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083

 

[SWComposites]

4 noded 1st order tets are awful
10 noded 2nd order tets can be ok

 

[rb1957]

2nd order TET is 10 noded ... TET10

"Hoffen wir mal, dass alles gut geht !"
General Paulus, Nov 1942, outside Stalingrad after the launch of Operation Uranus.

 

[BioMes]

I ran some tests, mainly using Abaqus benchmark models for plane stress/strain elements extruded to 3D (bending plus some with stress concentration). The results are interesting. Based on those tests, I could conclude that second-order wedges fall somewhere between second-order tetras and hexas but usually much closer to hexas as they often provide results with the same or very similar accuracy. Sometimes even a bit better than hexas. Of course, the problem is that meshes with wedges have twice as many elements and more nodes (not that bad though because wedges have 15 nodes while hexas have 20 nodes).

Would you agree with this or do you think that those conclusions are inaccurate and more tests would be needed?

 

[SWComposites]

As I stated above, I agree with the results.