Sorry that my answer is going to be a bit long winded, but consider the following incident in the OSHA accident investgation database:
Incident #[url='http://arcflashforum.brainfiller.com/accidentsearch.accident_detail?id=201282837']
201282837[/url]
At approximately 4:00 p.m. on August 18, 2007, Employee #1 and a coworker were removing the 120/240-volt electrical panel at a building in Doraville, GA. The two men had arrived at work at 3:00 p.m. dressed in tank tops and jean shorts and were waiting for Georgia Power to deenergize the circuit, which was scheduled for 5:00 p.m. Before Georgia Power arrived, however, they began to remove the panel by disconnecting circuit breakers and removing the panel's supports. At approximately 4:00 p.m., the coworker was on a ladder above the panel, and Employee #1 was standing on the ground in front of it when there was an arc flash and fire. The workers were transported to Grady Memorial Hospital for treatment. Employee #1 suffered severe burns over 60 percent of his body and died on October 15, 2007. The coworker's legs were badly burned and he was sent home to recover.
Key takeaways here are:
1. This is the only case I know of where an arc flash occurred which caused a fatality. Note that I do not count any incidents involving a "burn" automatically as arc flash unlike some practitioners. The physical effect of an electrocution is NOT an arc flash and should not be treated as one and the same. The actual number of arc flash fatalities is not very big. From 2007 to 2011, there are only 11 incidents in the entire database of fatalities, though there are many more injuries than fatalities (roughly 10:1 ratio).
2. Arc flash fatalities are possible at 120/240 VAC. This contradicts another paper out there which looked at the same database from 2002 to June of 2007. It is unfortunately bad timing that the same author did not put out the report just a few months later.
3. Based on the description although we can't tell if the "tank tops" were synthetic fibers, it seems unlikely that this amount of material would be enough to cause a fatality in and of itself. I would almost try to look at it as if the two victims were naked at which point the 1.2 cal/cm^2 threshold would apply because that is the Stoll curve reference point, if no clothing were worn at all. This confirms in a roundabout way that the incident energy was more than 1.2 cal/cm^2.
4. Based on the description although we don't have a transformer size it is most likely to be a typical utility feed with enough capacity to feed multiple locations. We cannot answer the question of size/magnitude based on this incident, only the question of whether or not a fatal arc flash is possible. The rest of this discussion revolves around trying to find reasonable thresholds.
National Electrical Safety Code, 2012 edition lists "4 cal" for all single phase panelboards up to 100 A and for all 3 phase panelboards rated over 100 A for a voltage range of 50 to 250 V. EPRI has also tested wiring in the "open air" configuration at 480 V with up to 40 kA and not exceeded the 4 cal threshold value. The table entries are somewhat nonsensical since the other cases (single phase >100 A, three phase <=100 A) aren't listed but one can clearly read between the lines. Note 2 for these entries reads: "Industry testing on this equipment by two separate major utilities and a research institute has demonstrated that voltages 50 V to 250 V will not sustain arcs for more than 2 cycles, thereby limiting exposure to less than 4 cal." It then refers to reference [B1]. B1 is "208-V Arc Flash Testing: Network Protectors and Meters", EPRI.
I can clearly identify one of the two "major utilities" as PG&E and the research institute is documented as EPRI. I don't have any idea who the second "major utility" is.
The data I have supporting the 45 kA argument comes from an EFCOG presentation as follows:
http://www.efcog.org/wg/esh_es/events/ESSG_Fall_11_Meeting/presentations/10-208V%20arc%20flash%20calcs%20efcog%20fianl%2010-12.pdf Jim Philips is mentioned in this report confirming the error in 70E which purports that the cutoff is 240 V, NOT 208 V, which was later corrected. In the presentation from the PG&E data at 208 V, it clearly shows that it is possible for arcs to sustain up to around 10 cycles at 1/2" arc gaps but not at 1". This was done using essentially the "standard" IEEE 1584 test without barriers and using a 12 gauge wire "fuse". Also the lowest short circuit current tested was 18 kA, well above the available fault current for a "125 kVA or smaller" transformer at 208 V. Subsequent research has shown that the fuse wire diameter is important and that barriers to plasma flow can dramatically stabilize arcs at marginal conditions.
There is then quite a bit of data presented from a paper listed as "Effect of Insulating Barriers on Arc Flash Testing", IEEE Transactions on Industrial Applications, V 44, No. 5, September/October, 2008. This looks a lot like the Mersen data. Many things were quoted from the paper but the takeaway from it is that arcs were self-sustaining with a gap of 12.7 mm (1/2") at 4.5 kA at 208 V. When the gap was increased to 1", arcs were self-sustaining at 32 mm. This testing was done with a barrier. At 12.7 mm, the highest arc flash value at 0.1 seconds was 2.7 cal. At 32 mm, it increased to 3.2 cal. For low X/R ratios, arcs at 12.7 mm could be sustained from 4.5 kA or higher but not at any current tested at 32 mm. Thus 4.5 kA is used as a threshold value in the EFCOG presentation.
The EFCOG presentation assumed an infinite bus and calculated the short circuit current (not arcing current) based on impedance. At 3% Z, the short circuit current for a 45 kVA transformer is 4.17 kA. For a 75 kVA transformer the short circuit current does not get below 4.5 kA untl the impedance increases to 5%. At 5%, a 112.5 kVA transformer has a short circuit current of 6.25 kA.
The reported arc flash values are quite low. I don't have the current EPRI report but I've seen similar data reported in the publicly available EPRI reports that continues to state arc flash values in a similar (roughly 3 cal) range. The IAS paper referenced in EFCOG also states 3 cal. The IAS paper also gives an upper limit of 0.1 seconds for overcurrent protection but the EPRI data that is publicly available and the NESC do not put a limit on the trip time for a 120/208/240 V panelboard. The NESC standard gives a rating of 4 cal for PPE but no overcurrent protective device cutoff. This would seem to follow since even getting to a self-sustaining arc at 250 V or less is not likely although the time to extinguish is much longer than a couple cycles.
So in conclusion it appears that for 1.2 cal, there is ample evidence to support a "4.5 kA" rule. Translated into transformer sizes this would be 45 kVA, if transformer impedance is not taken into consideration. For a 4 cal rule, NESC essentially allows for an unlimited transformer size. I would hesitate a little bit on this one though. The IAS paper stated around 3 cal after 0.1 seconds of arcing time, or 6 cycles. The PG&E data shows arcing at up to 10 cycles, which exceeds 4 cal if the data can be extrapolated linearly, but very high (20 kA or larger) fault currents are necessary. If we split the middle, we could reasonably anticipate that a 4 cal rule could apply to the "125 kVA or less" prior recommendation in IEEE 1584 provided that the threshold is 4 cal. This seems to be supported by the PG&E data for 1/2" arc gaps at least as far as fault current goes. Otherwise as a third potential route and in keeping with 70E table style, a 4 cal PPE cutoff with a 0.1 second opening time (ala NFPA 70E tables) would allow for a nearly unlimited transformer size based on the IAS paper. Otherwise as it stands right now, use IEEE 1584 calculations as a guidance. It is grossly overly conservative but there's not much data to support anything else.
Thus I would argue that depending on short circuit current available, a number of different options are available.
