Ideas for making rounds

Dave, Can you post a sumpin' like it, native and / or neutral, somewhere (mcadcentral?).

First, I thought the simulation stuff was generally done (Vince Adams, "Building Better Products Through FEA",

Second, I didn't see "realistic" in your list of criteria. I say this because no welder is going around sharp corners and leaving them sharp. So, if you say there's no radiuses on those corners, the automated MIG welder begs to differ. It will, whether you acknowledge it or not, round those corners. My recommendation, deferring to reality, would be to put variable rounds on those corners ~ zero at the top and full round at the bottom (problem is thinness of impeller section vs bead size). Then the round will go easily around the the entire welded intersection. Or approximate the same with a VSS or construct a curve on the impeller at the top of the bead, an end rounded curve on the shaft at the impeller/shaft intersection and some bead profile curves at the corners, then construct a boundary suface from these curves.

David Janes

Reading your reply was a smack on the back of the head I needed. I'd gotten lost trying to visualize the shape Dave was describing and never made it to "Mechanica model". If I were shell meshing something like that I might break the sharps and blend the root with 'chamfer-ish' facet surfaces. Solid meshing and Mechanica are out of my league. Creating those root fillet (round feature) surfaces without radical boundary curvatures or singular / degenerate edges (either of which will drive the mesher nuts?) is not possible without first putting generous rounds on the vane / blade edges. Even then I'd create 'nice' boundary / trim curves and Boundary Blend or, maybe, VSS a conic arc between the hub and blade. I'll go back to my hole now, watch and see if I might learn from those that have something worthwhile to say. ;^)

FEA",

Hey wait I have that book. I haven't been thru it in a couple of years - I should skim it again.

I have done this with some success - the failures relating to the variable radius rounds themselves, or not having enough space for them. I'm actually working on more complicated ear geometry with rib features, and the rounds tend to overlap/intersect a lot. Which is fine for welding, but hell for modeling.

Just after posting this (of course) I did think of an elegant way to do this: I made a surface offset from the cylinder by the length of the weld leg, and did an intersect to make a curve. Then I offset the intersection curve of the ear and cylinder so that I traced the other leg. Then I boundary blended the curves.

Dave

I don't bother trying to fight with problematic rounds, as you say its a real time waster - just go straight to surfacing,

Sean

While surfacing can make the geometry, I usually find a whole bunch of boundary blends bring their own misbehavior to the party.

I guess I was looking for reassurance that rounds are some of the most time-intensive features in Pro/E. It seems I'm not alone.

Dave

I agree that resolving the issue with surfaces brings its own problems but I find these issues a little more transparent and the functionality more flexible to fine tune the surfaces and resolve them.

Sean

~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Sean Kerslake Dept of Design & Tech Loughborough University LE11 3TU

The last element I thought (after going over this discussion) to add was the "accuracy factor". Accuracy comes up constantly as a factor in feature failures and other strange error messages. Setting absolute accuracy to the same value for the assembly and all its components alleviates many regen errors. The same can be said for curves and surfaces as well as import surface errors: many can be eliminated with adjusting accuracy or by switching from relative to absolute accuracy. I've eliminated many goem chks (including on imported models) simply by setting absolute accuracy to a correct value.

David Janes