I have a part that is the result of sweeping a rectangular section about a helical path for about 1/8 of a turn. As such there are really no good planar surfaces to align to the standard three view planes. There are features located on the helical faces that need to be located. What do folk do with this kind of geometry?
I work on mostly parts of this type. Usually I don't make 2D drawings, but my last one required some. The part didn't have a flat surface, and didn't even have any planes of symmetry about any axis, but it did relate to the ground, so the ground determined the orientation of the drawing views, just so there was some concrete context. I gave X, Y dimensions, but also an aligned dimension for the overall length.
How does the guy with a CMM fixture this to check it? He could CNC a fixture directly from the part file just to get true XYZ references and properly orient the part. This part didn't mate with other parts, so there were no location bosses, screw holes, etc.
A lot of curvilinear parts don't get drawings made of them at all, or if they do it might be a few overalls, an approx weight range, C to C for screw bosses, thickness, and a few simple tolerances. Since in many ways dimensioning curvy parts is impossible, or at least unlikely to be interpretable, it's often just not done. If it comes off the tool and fits together, then it was ok. I guess most of the checking comes in the design, making sure it looks good in the CAD system, making sure the shrink rates used were apprpriate, and that the process was giving expected results.
Machine designers keep reminding starry eyed CAD vendors that 2D paper drawings probably aren't going away any time in the next 50 years, but for plastics designers, the drawings already have disappeared.
There has been a lot of talk of 3D annotationless drawings where tolerances are assigned by feature color. Size and Position are given by the electronic CAD file. This seems to make a lot of sense, because if a standard were acheived, it could be given from CAD to CAM and the CAM software could automatically assign tools, feeds/speeds, clearances, etc based on the tolerance colors.
It's already available to put a CMM type ball on a mill to check parts as they're made. A lot of the pieces of the future of manufacturing are here already.
I still have to specify where the plane is. And that depends on the surface. So how do I document the surface in the drawing? Using a tangent plane will get me a good representation of the feature, but not it's location.
This geometry is easy to create in CAD, but hard to document because all the construction stuff is gone once the part is built.
I guess we have gotten into an area that would make Gaspard Monge* roll in his grave.
If you are molding a part you have a cavity set in a mold base so there are at least a few data available from which to perform a coordinate based specification of the part. What the final part ends up being is still dependent on process parameters. So how would dimensions be specified to check for example, warpage? If the part allows it, datum pad could be added. But this is not always the case.
Well, yes, if by that you mean that the paper isn't held in such high regard universally.
All that really gives you is a pull direction. If you're the guy making the mold, you have to establish the rest.
Warpage would be measured like anything else; allowable deviation (tolerance) from the theoretically perfect electronic model. On a computer controlled CMM, select a spot on the CAD model, and it goes to that point on the physical model to measure where reality is in relation to theory.
When you're dealing with electronic instead of paper data and computer control instead of optical comparitors or calipers / gages, you could even start talking about non-planar datums.
I am not fully understanding maybe what you are trying to achieve. However, you can show planes on the drawing and dimension from these planes to points on the curve. Sort of like the old days of dimensioning a 2D cam profile, pick incremental lengths along the x-axis and project upwards to the curve, then dimension these points from a reference point. Would SW intersect curve command be on any use in your application?