A Model That Alibre Cannot Create

[HaroldL]

When I'm not working on Too Tall Toby's practice drawings I usually scan the internet for drawings/models created in other CAD programs and recreate them with Alibre. Usually Alibre will have no problem recreating the model although sometimes with a different workflow. However, I did find one on an Inventor Tutorial web site run by Professor Constantin Stancescu in Romania that Alibre is not able to complete. Granted it does involve Surface tools such as Delete Face and Thicken but I had hoped Alibre would handle it without any problems. Well, not so - Alibre failed on the thicken feature. I followed the professor's video to understand how he created the model in Inventor. I changed the workflow that he used to get to the Delete Face feature (he used a surface Patch to create a surface, I used an extrude boss) but I don't think that is the reason for the thicken failure. I also tried using Shell but that failed to work.

Here is the drawing that I used for this model:


And here is a screen shot of how it looks after the failed thicken feature.


I ran the Check Part tool and it has a failed face and a long list of failed vertexes

I'm going to send this model and drawing in to Support so Dev can take a look at it for improvement. Meanwhile I'm going to try a couple different workflows to see if I can get Alibre to complete the model.

 

[Joseph_L]

I'm seeing a few things, the uploaded model has some sort of topology issue for sure. I'm not saying you did anything wrong, it just looks like something happened and some features are working in unexpected ways.

I think the way you use an extrusion to create a subdivided face is brilliant.

I have a few suspicions- I suspect that places where faces meet are throwing off some of the topological naming for the thicken surface, and that extra bubbly surface from the top is a continuation of a lower face. The circled portions in the images might cause a surface equivalent of a 'zero thickness error', where some faces that should be different are named the same.

I'm also observing the transition on the top of the part from the curved surface into a completely vertical one. The kernel seems to need a little help



I've uploaded three parts that represent different ways to handle this. I'm not sure if surfaces meeting at points is causing an error or not, but I dimensioned things like 49.999 instead of 50 as a precaution.

I also suspect the uploader has made a part with a tricky and unforgiving geometry for basic operations like shell and uses surfacing because that is the only way to get the part to work in that platform. If so, it's hard to get parts to work between platforms like that.

Uploads:

New Part Harold performs the top cut in multiple sections to avoid any problem with transition.

JBL Handle 3 thickens the surface and ignores the top transition issue.

JBL handle 4 gives some extra meat to the top to make the intended transition clear to the kernel, and then offsets the cut to be within fractions of a millimeter to the desired location. This allows for a much better transition.

 

[HaroldL]

@Joseph_L, thanks for joining this discussion and for the samples, the more eyes on it the better, I think.

When modeling these online examples I try to use the original dimensions, workflows, and features as much as possible. David at Support also gave me some feedback on this model with a couple of ideas to get Alibre to complete it.

I have a few suspicions- I suspect that places where faces meet are throwing off some of the topological naming for the thicken surface, and that extra bubbly surface from the top is a continuation of a lower face. The circled portions in the images might cause a surface equivalent of a 'zero thickness error', where some faces that should be different are named the same.

I'm also observing the transition on the top of the part from the curved surface into a completely vertical one. The kernel seems to need a little help

David thought there may be an ACIS limitation and that maybe this model could be sent to their Dev team for an enhancement if Alibre Dev decides to. Whatever the final decision, hopefully this will help Alibre get even better.

 

[HaroldL]

Thanks to @DavidJ and @Joseph_L for pointing out some things to check on this model. I found that Alibre, or maybe ACIS, had its biggest issue with a cut that was tangent to a fillet edge. I increased the offset dimension from 55 mm to 55.1 mm and that helped a lot.

I used a Thin Extrude Cut to trim off the model. Atom 3d will have to use a fully closed sketch for an extrude cut.



The other issue was the opening at the top that followed the cut curve. (Thanks guys for pointing out the use of Project and reminding me of Remove Face). I was able to get the cutout to within 1.5 mm of the edge of the cut before it started giving odd results with Thicken.



With those changes I was able to finally get Alibre to complete the model. It took multiple fillets to finish the top cutout edges but I was able to fillet the bottom cutout with two 1.5 mm fillets to fully round the edges.

 

[stepalibre]

David thought there may be an ACIS limitation and that maybe this model could be sent to their Dev team for an enhancement if Alibre Dev decides to. Whatever the final decision, hopefully this will help Alibre get even better.

I don't think that's a ACIS limitation (well not completely). The limitation is Alibre's feature robustness. There's always a 100 ways to create models with the right toolset. In Alibre as the topology change further down the tree and as complexity increase Alibre's feature failure rate will also increase. And the lack of feature robustness means it'll be very difficult to find and fix topology errors in those features and to even know what is wrong.

Some of these features seem to cause the most errors for me. They don't seem to be fully parametric or can resolve topology changes very well.

I'm seeing a few things, the uploaded model has some sort of topology issue for sure. I'm not saying you did anything wrong, it just looks like something happened and some features are working in unexpected ways.

I think the way you use an extrusion to create a subdivided face is brilliant.

I have a few suspicions- I suspect that places where faces meet are throwing off some of the topological naming for the thicken surface, and that extra bubbly surface from the top is a continuation of a lower face. The circled portions in the images might cause a surface equivalent of a 'zero thickness error', where some faces that should be different are named the same.

I'm also observing the transition on the top of the part from the curved surface into a completely vertical one. The kernel seems to need a little help

View attachment 42871View attachment 42873

I've uploaded three parts that represent different ways to handle this. I'm not sure if surfaces meeting at points is causing an error or not, but I dimensioned things like 49.999 instead of 50 as a precaution.

I also suspect the uploader has made a part with a tricky and unforgiving geometry for basic operations like shell and uses surfacing because that is the only way to get the part to work in that platform. If so, it's hard to get parts to work between platforms like that.

Uploads:

New Part Harold performs the top cut in multiple sections to avoid any problem with transition.

JBL Handle 3 thickens the surface and ignores the top transition issue.

JBL handle 4 gives some extra meat to the top to make the intended transition clear to the kernel, and then offsets the cut to be within fractions of a millimeter to the desired location. This allows for a much better transition.

Great work @Joseph_L! Speaking of feature robustness, when you change a dimension in a model they break. Adding equations might help fix this, but my point is that we could benefit from better feature troubleshooting tools.

 

[Ex Machina]

I think this would make a great video for your YT channel Harold!!!