I was watching a video about the Open Source Ecology large 3-D printer and the problems they’re having with bushing misalignment. Basically the problem is that they have this big 3-D-printed block with spaces in it for bronze bushings, so that it will ride smoothly and with low friction on a pair of steel rods, but it doesn’t.
The problem is diagnosed to be misalignment: the bushings are 12-micron tolerance, so a 12-micron deviation in the shape of the cavity they fit into is enough to get them out of tolerance, and possibly kick the two bushings for a single rod out of parallel enough that the rod can’t slide through the easily.
It occurred to me that this is another case of needing to worry about not only geometry but the derivative of geometry with respect to force, which is to say, compliance. If there’s enough space around the ends of the bearings for them to rotate a little bit in their seats to comply with the rod, this wouldn’t be a problem, as long as the solid ring around the middle of the bearing can support the necessary load. (The idea of leaving some play to avoid binding is mentioned around minute 39 of the OSE video, but I don’t really understand if they’re talking about leaving play in the same place I’m talking about leaving play here.)
More generally you can leave space in the “solid” plastic around the cavity which allows the cavity as a whole to rotate but not translate, using established flexure designs to provide selective compliance (whether using FACT or another design approach). Even just using a softer plastic would diminish the binding problem, but that might create undesired compliance in other degrees of freedom; spaces for compliance, like a 3-D version of the Snijlab living hinge, can permit large compliances in selected degrees of freedom with minimal compromise on strength and the stiffnesses in other degrees of freedom. Leaving such spaces can be done even with conventional molding and subtractive manufacturing processes, but it’s much easier with 3-D printing or digital 2-D cutting processes.
Another aspect of the problem is that they’re building a gantry with lots of prismatic joints (built, in turn, out of cylindrical joints), which pose a lot of problems like binding under side loads, and revolute joints would have been a better choice.