Hi Kevin, yeah unfortunately fillet problems tend to be very difficult to fix.

The whole calculation of a fillet has quite a few different pieces in it, it involves calculating offsets of surfaces, intersections of surfaces, extensions, construction of fillet surfaces along rail paths...

Any single one of those is actually a pretty complicated operation, and fillet has layers of many of these parts.

In your particular original case, it looks like the part that is breaking down is the calculation of the intersection of the fillet with the side surfaces for that middle piece. The steps that Burr shows above cause those side pieces to be replaced by plane primitives, those can get some special case assistance in intersection calculations that seems to help out in this case.

It is good to have examples of things that don't work right, but for filleting most of the time I will be gathering them up and sending them over to the people who make the geometry library that I'm using so that they can examine them and try to tune things up. That will tend to be a rather long and drawn out procedure though.

In some cases I may be able to do something about it sooner than that, like Pilou posted one recently that I was able to fix up right then actually. But that will tend to be rather rare with fillet problems however.

- Michael