If you go the blend route, another thing you can do is adjust the bulge factor, if you reduce it the blend will become more flattened and get close to your original shape, here is an example where I typed in a bulge factor of 0.5 for each of the 4 blends:



The central surface is still not completely flat like in your original, but it is getting pretty close to it, it is less bendy than the previous one. And then it still has the property that it is smooth with the other pieces where they touch, which makes the fillet work.

Let me know if you would like a more step-by-step run-through for how this one was constructed.

- Michael

Thank you very much, Michael, it's all pretty clear. I did guess it was a problem of the filleter.

Anyway, even if there are still a few tricky shapes the filleter can't cope very well with, I find it MUCH smarter than Rhino's built-in filleter (even version 4) ;)
You're doing a remarkable job!

Thanks again, and all the best

Daniele

Ok, here different way
fillet two half cylinders, draw line between corner points, extract fillet edge, make a new profile.


use sweep-command


use boolean-diff

and still more simply, loft-command

Yes, these types of situations are pretty annoying to solve. Rhino users use a method where they build a pipe across the edge they want to fillet, then use that to trim the main object and remove the pieces which creates a gap for the fillet. You can then blend the upper and lower surfaces but that might not always work. You can also create a fillet profile and use the upper and lower edges as rails. But because these rails have sharp corners there might be spaces etc.

It's kind of frustrating, but that's Nurbs for you. :P

I've noticed that many Rhino users (that use it for precise modeling work) also have a copy of SolidWorks or Pro/E that they use for these kind of filleting jobs in the end of the process. Those are geared more for engineers and doing this kind of thing in a brute force method seems to be their forte, while they aren't intuitive for modeling.

So obviously not a problem specific to MoI, as even surface modellers that cost an arm and a leg like Alias StudioTools have the same problems (from what I was able to test with the PLE version).

Hi Gent, yup basically this is a common problem. Fillets are a very difficult area of calculation, particularly in corner areas where more than two edges are coming together.

The fancy solid modelers do solve this in a kind of "brute force" method like you described - they have a lot of special case code that recognizes different specific configurations and does the right thing to handle each one on a case by case basis. They've spent a huge amount of time and effort adding more and more special cases to make them work better.

The geometry library that MoI uses just has not had as much time spent on all those different special cases.

Some general advice is to try and smooth out the areas to be filleted, because sharp corners are more difficult, and also to try and have pieces set up accurately. Things should be snapped together well, any kind of small mis-alignments cause further problems. Doing those kinds of procedures basically makes the filleters job easier and will make it more likely to work.

- Michael