Is there a good existing publication that covers wiring practices for enclosures (including multiple bay enclosures). I get questions from engineering and production asking about what methods of wiring are considered arc flash compliant. I know this is a very open question, and may result in a great many hours of writing and editing a document for guidelines for engineering and production.

No such thing as "arc flash compliant". There is "70E compliant" though, and for distribution equipment, "NESC compliant". First, see Chapter 2. There are about 3 pages of very general installatioon/maintenance practices that are best practices for any site. But they can be broken down into roughly 3 categories. First there are several documentation requirements that also appear in various standards and regulations such as the need for single lines. Second everything must be properly designed and maintained and the standard in the U.S. for this is the National Electrical Code (NFPA 70). The third requirement is a series of maintenance requirements.

Generally speaking, NEC covers wiring between equipment, not inside equipment. The one exception that comes to mind is pull boxes. The wiring requirements for inside boxes derive from the idea that components have to be Listed with an NRTL (Nationally Recognized Testing Lab). OSHA maintains a list of which labs are acceptable and which are not. Two prominent examples that are not on the list are CE (because it does not mandate third-party verification), and MSHA which is a part of OSHA (inter-agency rivalry). Some interesting "foreign" labs are also on the list such as CSA and TuV.

Moving away from this section if you turn to the parts of Article 130 that speak to arc flash hazards, there are rules for equipment that is properly designed, installed, and maintained. There are no rules for equipment that does not meet those requirements, and the informational notes clearly invalidate arc flash analysis if those criteria are not met. So you can either have equipment that is properly designed, installed, and maintained, or you are on your own with no guidance from 70E.

The tricky part to meeting this is of course what it means to be "properly maintained", and the two standards that are referenced (NFPA 70B and NETA MTS) are pretty vague in this regard, and don't even specify a specific maintenance frequency. You are somewhat on your own in this regard.

As far as how to properly maintain wiring and what the standards are, getting away from design criteria such as ampacity, the most clear standard is in Article 110 of NEC which requires that all wiring is done in a "neat and workmanlike manner". And this applies not only to original installations but also to repairs. For example time and again, I've seen electrical tape wrapped around cords. This is an immediate red flag that an improper repair was done. At best, this is an emergency repair. But when making the repair, it must meet Code. This means that for instance with a cord, a Listed splice such as a butt splcie must be used. The insulation must be restored back to original. The Electrician's Handbook and some wire and cable manufacturers have detailed documentation on proper splicing, and it takes about 45 minutes to do the splice properly for low voltage cable, if you have all the materials and tools together in one spot like I did at one time where I routinely had to perform repairs to 4/C #6 SOOW for a 480 V shore power system where it wasn't practical to replace the cord. The effort required was identical for any other size low voltage cable. For medium voltage cable the standard for a permanent splice is vulcanizing, a process which takes almost all day per splice. Suffice to say that for most cases the best approach for cords is either to install a junction box as a "repair" or in most cases, replace the cord. The time and materials involved usually exceeds the cost of the cable.

Here are some additional items that I hear crop up from time to time:
1. Doors open vs. closed. This is a silly requirement in 70E. The issue is when there is exposed wiring and the potential for maintenance tasks to trigger an arcing fault. Under OSHA 1910.269 in the appendix, they list this as crossing the Minimum Approach Distance with conductive tools or body parts. This is the same as Restricted Approach Boundary under 70E...both come from the MAD definition in IEEE 516. 70E offers no guidance where the equipment does not have a door. And most newer equipment is now guarded internally so even with the door open, wiring is often not "exposed". As of the 2014 edition, NEC now requires labels on all doors that limit access to exposed wiring, whether "high voltage" (601 V+) or low voltage (50-600 V), so if equipment is properly labelled, electricians should know before the doors are opened whether or not exposed wiring exists unless you go "label crazy" and put labels in locations where they don't have any meaning such as labelling the door of an entire room full of enclosed equipment with an exposed wiring sign...this mislabelling just destroys the value/meaning of the label. IEEE 1584 analysis is based on equipment with no doors (just a box with 5 sides on it). Furthermore, analysis and modelling by CIGRE shows that arc blast knocks the door off almost any size enclosure within 1-2 cycles which is faster than most breakers, so for all intents and purposes we are doing arc flash analysis today as if the door is not present in the first place. Doors are more of an argument about whether or not an arcing fault can be caused in the first place, not reductions in incident energy.
2. Arc tracking. When an electric arc occurs, it heats up the air around the arc which provides a pathway for initiation of an arc during the next half cycle. In addition if the air is hot enough, it becomes ionized and/or plasma. This heated air is propelled by the magnetic field surrounding bus bars at a speed of several hundred feet per second away from the power source until it encounters either the end of the bus bars, or until it reaches a substantial "barrier" that impedes movement of the conductive air. The "arc flash" appears at that location, NOT at the point where the arcing fault first originated. The effect is most prominent in 480 V distribution panels and vertical sections of MCC's. Arc flash analysis thus usually treats switchgear (which has substantial barriers) as individual sections. Panelboards and industrial control panels are treated as one panel for arc flash analysis purposes, and most practitioners treat MCC's with a main breaker as two enclosures, one for the main breaker and one for the rest of the vertical sections. But there is no accepted definition at this time for what would be considered an "acceptable" barrier. Thus questions about very large nipples between two panels or duct banks really have no clear guidance as to whether the equipment is "isolated" or not.
3. Equipment design with regards to arc flash. I've heard just about everything in this regard. Some planels have the breakers physically screwed to the front of a panel in industrial control panels and the concept is that it is more hazardous when it is "close" to the surface. See issue #1...this doesn't really matter at all.
4. Concerns about simply being in a room with say an MCC present. There are multiple 70E Committee statements stating that "just walking by" does not constitute an arc flash hazard. Although equipment can and does spontaneously fail, especially medium voltage equipment during periods of heavy dew (water+dust = dry banding = flashover), if properly installed and maintained, these incidents are rare. I could not find a single instance in OSHA's databases where someone was injured from "spontaneous arc flash". Every incident included some kind of human activity which triggered the event. Not that I think it's good practice but unless there is activity that is going on that can trigger an arc flash, there is no evidence to support this concern that I'm aware of.

Thanks Paul.

Much of this I knew. I was really looking for guidelines on what does not exist in a well-defined form yet. This seems to be a recurring theme in many of my questions such as previous questions like 'What constitutes a suitable barrier' and 'How large an opening is acceptable'

Being a manufacturer, we are of course just as interested in the wiring standards for the inside of the enclosure. We do have wiring standards. Of course they were developed before the coming of the great arc-flash focus. We want to insure everything we build is compliant with UL, CE, OSHA - basically all European and North American standards.

Sorry...missed it and when it said production and maintenance I was thinking of an end user, not a vendor. As you are probably aware the closest thing to a "generic" UL requirement is 809A for industrial control panels. Manufacturers are slowly getting on board with other directions though. There are multiple "arc resistant" standards, although a large portion of "arc resistant" gear is in fact not actually "arc resistant" as such. An example is low voltage MCC's that are tested following the ANSI C37 protocols for medium voltage switchgear. There is no such thing as ANSI-compliant "arc resistant" gear because the relevant (UL) standards don't cover this. It comes from either misuse of the ANSI standards or else use of the IEC standards.

A huge body of research including enough details to write code and/or do calculations on arc resistant gear was recently released by CIGRE. As an end user I found it fascinating and lacking at the same time. There are two issues with this body of research. The first is that essentially the most basic model considers a heat source producing a relatively constant amount of heat and the pressure rise due to that heat source. The model then arbitrarily uses a cutoff (Pmax) where beyond this value, the enclosure ruptures. The first issue is that although a lot of testing and modelling was done, they failed to be able to provide a way to calculate Pmax so this value must be measured through testing. Similar issues exist in explosion proof enclosures though so I can't really fault them for this. Second issue is that the arc input energy was modelled using the Lee model which is not a great start in the first place, and then using somewhat arbitrary constants to correct the results. All published test work used artificial heat sources instead of voltage and current as inputs. It stands to reason that substituting a better model (say IEEE 1584 differentiated with respect to time) would result in a better fit to the data but there's not enough published data to go down this path.

What I got out of it without using the model for its intended purpose (as an end user I don't particularly care about modelling arc resistant gear) is that the enclosure can be treated almost like a sealed box if the total surface area of openings is less than 10%. Otherwise, the full model applies.

Second item that was discussed in that work is alternative atmospheres such as SF6 and what a dramatic difference it can make. And the third interesting item is using "absorbent" materials that can significantly retard arc flash. Rockwell Automation has done some proprietary research on this subject as well and it almost looks like all they are talking about is what color the enclosure is painted but they aren't sharing details. With respect to paint, I can easily understand this. The emissivity of aluminum is down around less than 0.1 while it is around 0.65 for rusty steel, around 0.80 for painted, and close to 0.9 for black painted steel. Thus we can significantly alter the absorpotion or reflection of heat of the enclosure merely by altering the material and/or color. A black colored absorber with a high heat capacity within the enclosure is likely to significantly draw heat away from the arc flash and retard the time before it ruptures.

At this point in time there are 4 basic ways to attack arc flash as a general rule:
1. Reduce the arcing fault time. This is the realm of relay manufacturers. I haven't seen an "AFCI" for 480 V yet but I have no idea why it can't be done. Right now we have light detectors, di/dt detectors, differential relaying zone selective relaying (allows increased use of instantaneous tripping without consequences to coordination), and standard relaying as our means of detection/tripping.
2. Arc elimination. This consists of either high resistance grounding which reduces the likelihood of arcing in most cases, or various devices that suffocate the arc by providing a lower impedance path temporarily in conjunction with a main breaker.
3. Reducing the current. This is a two edged sword though. Current reduction can also mean increases in opening time and hence increased incident energy if not coupled with changes to trip functions. This causes odd effects such as switching to a current limiting fuse can actually increase rather than decrease incident energy.
4. Various schemes such as some items in the new IEEE MCC guide document that result in a reduction of hazards during certain tasks. For instance racking out buckets is normally pretty hazardous. Changing the mechanism to incorporate a disconnect or shorter stab arms or adding "closed door" racking in combination with arc resistant gear eliminates the hazard altogether from this particular task. Various means of verifying voltage with the door closed also significantly reduce the hazard during testing for absence of voltage. And finally as I already mentioned, eliminating any and all exposed wiring with the doors open is a big step forward. In combination with upcoming IEC rules for test meter tip designs, there will no longer be any exposed wiring during testing functions.

In the medium voltage world, the IEEE 386 "elbow connectors" are finding more and more uses outside of their original intended use for underground equipment. What you start with is a molded rubber connector that can be load break or non-load break with standardized fittings that extends shielding continuity all the way through the connector to both ends so that there are no exposures to any of the conductors. The connector itself is designed for burial so it is waterproof and dust/dirt proof. With no external exposures and shielding all the way around, arc flash is virtually eliminated. A similar design exists with SF6, rubber, or solid epoxy insulated switchgear. The gear is frequently sealed for life and maintenance schedules are very low (lubrication/inspection 3-4 times over the entire lifespan of the equipment). This is a far cry from drawout gear where every racking operation is questionable in terms of safety and the gear requires extensive annual PM's that were generally not being followed. As an example just last month I had a Westinghouse breaker where the gear was stripped out so when I went to rack it out, the breaker just stopped halfway in/out. Even if we cut off power and accessed it from the back, the entire racking mechanism was built into the breaker and if we had not managed to coax it out, there was no practical way to remove the breaker short of disassembling the entire switchgear or going in with demolition equipment like torches and cutting the breaker apart. This is as opposed to the UL rated equipment such as switchboards and panelboards where everything is bolted so it can't be maintained "online" and must out of necessity have much longer PM schedules. Today the designs have converged. Switchgear manufacturers have standardized on a small number of breakers that are used in both drawout and bolted designs for cost reasons, so we now have very long maintenance schedules even on drawout switchgear, but the breakers are not the issue. It's the drawout mechanism itself. As an example the ABB AMVAC breakers that have been out for close to 20 years have proven themselves in use. They were originally advertised as "zero maintenance". Far from it...they require some maintenance, but not very much. The big problems I've seen with them are where the low voltage signal connector doesn't engage properly, or the cell fingers don't engage properly, or the cell arms misalign for one reason or another, leaving it in various states of not quite truly closed/open or engaged/disengaged. ABB themselves report that 80% of arc flashes in drawout gear is in the drawout mechanism, not the breaker. This all points to one clear conclusion: perhaps it is time to retire drawout gear. Panelboards are a lot less expensive and much smaller. And gas insulated gear is the ultimate in reliability and small size, but also can't be removed without tearing the equipment apart. Adding spares is relatively inexpensive for both types of equipment. A final option that provides the ability to take breakers on/offline without a drawout mechanism is to use the elbow connector to tie breakers together. I've actually seen equipment meant for underground vaults built this way but there's no reason to restrict use to underground only.

Taking this all together I believe the trend is going to be towards defining what specific tasks are required for maintaining equipment followed by what the hazards of those tasks are, and designing equipment around those hazards so that they no longer exist in the future. Medium voltage gear is becoming more standardized at least from a single manufacturer and becoming more of a modularized affair with bolt-together sections that are stacked to produce a particular function. Similar trends are occurring in 600 V class gear. Many overload relays are now microprocessor driven electronic relays and the inputs/outputs are designed to bolt directly onto the lugs of the contactor without any jumper wires. It would not surprise me to see the same trend in breaker or fuse clip marriages to the contactors. There are already some small (10 HP) starters on the market from Rockwell Automation that don't even have enclosures anymore. The entire "MCC" is now a set of tightly integrated modules with no user serviceable parts and in some cases, even the cables to the motors are premolded assemblies, and even field disconnects are available in the same system. Most of the maintenance tasks that would have an arc flash potential are elminated as a result. If this design extends upwards into higher currents, although arc flash is still a very real threat, it could become a very limited and specialized problem.

Depending on what sort of equipment you are producing, NFPA 79 May also be worth looking at as a guide. It's orgin was as a standard for Metal working and Plastics working machinery, but is now "Industrial Machinery'and it largely just echos the NEC.

Paul, Thanks for the responses. I think you mean "UL 508A, Standard for Safety, Industrial Control Panels" instead of 809A?

Correct, 508A. Sometimes remembering all the different standard numbers is not easy.