working on my first few arc flash studies. I've been reading IEEE 1584 and NFPA 70E. My question is how far must arc flash study and labeling extend? Down to distribution equipment - panelboards, disconnects, MCCs, switchboards. For instance, in a machine shop with CNC equipment, are all machine electrical control enclosures required to be labeled?

Employees must be able to determine the proper PPE to be used for the task they are performing.

In the case of the OP, most of my customers are placing labels on each piece of production equipment, like CNC's.
The majority of these customer use some AF analysis methodology that does not require the modeling of each piece of equipment, such as an energy boundary.

It's sort of a tricky subject. NEC is pretty clear that anything requiring "frequent" energized maintenance needs a label. NFPA 70E just describes the content of the label and IEEE 1584 goes into how to estimate incident energy but not the likelihood which is covered by other standards. NEC doesn't define "frequent". It seems like the general consensus is more than once a year. The practical side is whether or not adding more labels is worth the "pester factor" (how often someone gets involved in providing data on unmarked panels). If it's not very frequent, labels might not be needed.

The second issue is whether or not you can "generically" label panels. By this I mean if you have some kind of "understood" plant rule such as that all unmarked panels are assumed to be say 1.2 cal/cm2 or 4 cal/cm2. With that baseline, all panels don't need a label but clearly this kind of rule is a Code violation.

Also of consideration is the location...whether or not the label will require frequent updating, and NEC also kind of clearly references distribution points like panelboards but not equipment accesses such as motor peckerheads. Presumably that's because the lockout point is upstream of a local panelboard or other types of on-machine boxes.

There is also a lot of debate about the lowest level require to have labeling. Currently there is a note in 1584 about 240V and 125kVA which is getting revised in the next revision. There are also not provisions to calculate incident energy for single phase systems with any known accuracy.

To answer your question - 480V/3 phase, 240V/3phase - yes. 208/3 phase, single phase - ??? generic label possibly. Machine motor starter cabinet with 480V 3 phase - yes, no 3 phase????.

At least in one operation we considered using the "generic" NEC label, the one that basically just says "Warning! Arc flash hazard. Wear appropriate PPE." Just last week when I was installing a piece of equipment it actually came with the "generic" label supplied in the box. At another operation we talked about simply using a very generic "<1.2 cal/cm2" label for equipment that we didn't need to analyze or it could also be "<4 cal/cm2" for distribution (utility) equipment. These are all NEC (enforced Code) compliant but totally non-70E compliant but since 70E is voluntary as long as you're consistent about how you label things, especially if it's in some sort of document somewhere, and if it includes procedures for electricians on how to handle the generic label situations, then I don't see any reason that OSHA would take issue with it.