@Max Weber , thanks for posting this, this is a really interesting conundrum.
I think the fixture plate idea is OK / fine.
Depending on how many you have to do.
I'd be tempted to
1. Switch the order of operations (to reduce chatter / ringing of unsupported thinner features.).
2. Reduce / eliminate one set up (potentially) . in this case your setup labelled "2".
View attachment 292269
^^^ Click to blow up.
I'd be tempted to change the order of operations as you have it labelled to,
4, 1, 3,
So 4 to 1 would be like a conventional two sided flip, then finish up on 3.
IF I could devise a way to mill the sloped flat surface (marked "number" B :-) ) and maybe use a form tool or other milling strategy on the radius / feature marked A and A' … then one
might be able to eliminate your setup 2 altogether. ? [Haven't had my Weetabix yet … so who knows .].
These days you see 'Peeps" 3D-printing (low res cheap resin) representations of partially machined parts (in various process stages / ops) to figure this stuff out, maybe even like potato block painting, (if children still do that or have an app for that), imprint them into slabs of plasticine / playdough to get a visual, mental and tactile firm "Grip" on fixture design (maybe). Heck I bet this part could be made of LEGO and imprinted into playdough ;-) at various depths and orientations to get a rough handle on fixture design.
I like Milland's idea of the 4th axis indexers that have a plunger you can hit with the end of the tool.
I think the fixture plate idea is a good exercise - especially for repeat parts.
______________________________________________________________________________________
-
Further brain fart - It could be that you could design a fixture such that
some of the profiles and fixture elements can be superposed on top of each other or "nested" in place. That way better part precisions and accuracies could be maintained over the machine (
yielding higher localized repeatabilties),rather than moving in process parts from one end of the machine's table to the other. I.e. keep each part in process to finished part as much as you can within the machine's local coordinate system for that part, (so you'd be rotating each in process part mostly in place, so your 4, 1 and 3 fixture profiles are on-top of each other and possibly at different depths; (maybe ? - looks like you might have to rotate Op 3, 90 degrees in the horizontal plane relative to the long axis of the other setups after the first "Flip".). ~ 5 axis machining
claims better part accuracy as the part is mostly machined to completion in the "sweet spot" of a machine with minimal operator intervention. However, on a three axis mill without scales the local precisions and accuracies over a 4" range should be pretty good versus moving parts op 10 to 20 to 30 in 16" ranges each time to the furthest extents of the table (as per the original scheme illustrated by Max Weber/ OP) - and also possibly at the same time eliminate a lot of probing cycles, almost zero point clamping / fixturing idea.
-
Post brain fart - IF you can eliminate one set up and nest the features and profiles for fixturing each Op then theoretically
you could get 4 completed parts (at a time) for opening the door of the machine three or four times and flipping parts in place. Rather having work still in process to be completed, start and end of the "Run" , 4 completed parts full cycle full ops being a smidge more efficient perhaps. That would more than compensate for time lost using a bull nose/or ball end mill on that sloped feature / contouring + different strategy on the Tricky external radii of feature A to eliminate an additional setup (2). In some 5 axis scenarios you open the door twice to top and tail one part at a time, versus your 3 axis mill scenario with a fixture plate where you have roughly one completed part for each door opening of the machine (4 parts at a time for 4 door openings , as it were). ~ 5 axis being one form of enhanced automation , but also having a lot of parts on the table of a 3 axis mill being flipped by a robot or cobot would be another approach to automation. In some instances might be nice to walk away from the machine for longer periods to work on something else maybe ?