i have been looking through arc flash last 4 months, and i am kinda trying to figure out what is missing in arc flash calculations? in my mind, DC arc flash, single phase arc flash, etc. can you give me a few name in your mind? what is the technology gap for industry or engineer need to fill out? lets discuss and see where we lead them.

DC is pretty well covered. It's a bit young but still there. Some other areas:

1. Quantify the effect of phase dividers.
2. Quantify the effect of enclosure size.
3. Quantify "off axis" positions relative to enclosures.
4. Quantify arc blast which at this point is pretty much an unknown.
5. Quantify the effect outside of the "IEEE 1584" valid range (above 15 kV, below 480 V, above and below the kA cutoff, etc.)
6. Better quantify the lower cutoff.
7. Quantify the effect of fuses vs. circuit breakers (especially the current limiting effect).
8. Quantify the influence of doors, walls, etc.
9. Quantify the influence of other effects such as plasma, and shrapnel. It is well known that outside of a certain range of voltages and currents, plasma starts to be a much larger influence.
10. Quantify the effect of configuration...how vertical or horizontal bus bars or other structures effect the arc flash effects. Some test work has already shown that it tends to project the arc flash outwards and shoot out jets of much more damaging plasma if the bus bars are horizontal and aimed out of the enclosure.
11. Better measurements on arc blast designs. Right now arc resistant designs are limited to medium voltage and are performance based designs only.
12. Can we arrest an arc flash fast enough to clear it early and never get to the point where there is a risk...another arc resistant approach instead of with blast doors?
13. Is it possible to scale up the concept of AFCI's to detect arcing faults for real at industrial scales and trip based on current/voltage measurements only rather than using the existing fiber optic and/or Rogowski coil/fast dV/dt designs.
14. What is the failure rate of current design circuit breakers and can we improve on these? Can we do probabilistic risk assessments of arc flash and do away with the PPE when the likelihood is so low that there is effectively too improbable of a risk to concern ourselves with PPE. Are certain designs inherently safer than others? For instance is bolted, metal enclosed gear safer than draw out and is the reliability high enough to make it decidedly better?
15. How does PPE work with respect to plasma, shrapnel, or concussion (arc blast) forces? Is it possible to manipulate the PPE and provide better protection to different effects of an arc flash?
16. Is it possible to increase the gaps (say via solidly insulated gear) to the point where it is flat out no longer possible to even get an arc flash with some designs, thus making it arc proof?

These are just some ideas...it should give you an idea at least of just how little is truly known.

I would like to add:
17. ATPV concept: is "50% probability that sufficient heat transfer through the tested specimen is predicted to cause the onset of a second degree skin burn injury..." safe enough?

Hey, I can actually answer that one now partially. One of the issues (which adds yet another question...) is projecting injury/safety data beyond Alicia Stoll's data. For instance, where is the first degree burn line? I've done some simple geometrical projections (and posted them here) that suggest that the third degree/fatal line is so close to the 2nd degree line that realistically it need not be considered. More interestingly,

Alicia Stoll's curve is that...a CURVE. It is higher when the incident energy exposure rates are shorter, and vice versa. Remember the magic "1.2 cal/cm^2" number? This is Alicia's number at 1 second. It decreases at 2 seconds, and followup data at shorter intervals shows that it increases. Just as with electrical components, human skin can tolerate higher thermal impacts at shorter time intervals and vice versa.

Second, there are multiple hazards in an arcing fault incident beyond the two that are already mentioned...thermal and arc blast. Another major effect is plasma ejection. It turns out that porous PPE such as the types currently made from treated cotton and aramid fibers offer very little protection against plasma while simple rainwear stops it dead at far lower ATPV ratings. Thus the reason for the earlier questions. Leaving this aside...

Third, right now the test ratings are based on the CLOTH, not the CLOTHING. The cloth is tested as a single panel facing perpendicular to the arcing fault. No undergarments are considered (which offer considerable protection in tests that have been done), and the effect of off-angle arcing faults is not considered (which would reduce the incident energy by the cosine of the angle). So in reality, you are a lot better protected from the thermal effects than stated.

Finally, take a look at the CURVE generated by the ATPV rating (based on the ASTM procedure). Several pieces of cloth are tested and a range of incident energy values are tested simultaneously. The actual test data is purely pass/fail. The curve that is fitted to it is a sigmoid. Roughyl 0.5-1 cal/cm^2 out on most of the data that I've looked at, there are ZERO failures. Thus it seems to suggest that right at the threshold of ATPV there is indeed a 50% chance. But contrary to what the sigmoid (which is material dependent) may suggest, failure probability drops very quickly to zero as the incident energy decreases. If we were to superimpose various phase angles on the incident energy calculation and take into account the impact of assymetrical fault current, the probability of injury would become precipitously low because some of my own calculations suggest that about 89% of the time, a reduction of 0.5-1 cal/cm^2 would be covered by simple probability of the phase angle being favorable at the time of the fault.

Don't know if that answers your question though because I pointed out one reason why the ATPV rating may be meaningless (plasma), and several more indicating that for most realistic scenarios, employees are far better protected than currently predicted even according to current knowledge.

Incident energy in cal/cm² already has time of exposure built into it. A given arc will produce twice the IE in cal/cm² at two seconds than it will at one second. How does this relate to Alicia's number?

If the exposure is 2 cycles, the onset of a second degree burn will be higher than 1.2 cal/cm^2.
If the exposure is 60 cycles (1 sec), the onset of a second degree burn will be 1.2 cal/cm^2.
If the exposure is 120 cycles (1 sec), the onset of a second degree burn will less than 1.2 cal/cm^2.
That's what I understand from Paul's post.

That is it exactly. Incident energy calculation assumes that exposure is linear. IE, the probability of a second degree burn depends on the amount of thermal energy AND the length of exposure...the chance of injury is nonlinear.

http://www.esps.ca/data/1/rec_docs/102_Oberon_WP_Understanding_the_Stoll_Curve.pdf

There has been research to extend the Stoll curve down below 1 second. Notice however that the result is probably not what we'd like it to be. At 2 seconds, the threshold is 1.46 cal/cm^2. At 1 second, it's the familiar 1.2 cal/cm^2. As the time interval gets shorter however as I recall (can't find the data right now) it does tend to go back up after a point. Essentially for very long arcing intervals, we are getting some cooling effect. At shorter intervals it does not have enough time to absorb the energy.

Fortunately, this problem is only a problem for "PPE 0". For tested ATPV ratings, this issue goes away. The clothing is tested in the ASTM standard against the Stoll CURVE. If it deviates above the curve at any time during the test, the clothing "fails". Thus the ATPV rating as far as thermal energy is concerned "passes" regardless of the nonlinearity of the curve.

Hi, Paul. Looking for a test summary, I found that ATPV is 8,4 cal/cm2. This value results from sigmoid curve for 50% probability of burn. For 5% probability of burn, the incident energy is 7,6 cal/cm2. My question is why accept 50% probability of burn if 5% probability of burn, on my concept, is more safe? In this case, the clothing used on HRC 2 could be used only on HRC 1.

My question is why accept 50% probability of burn if 5% probability of burn, on my concept, is more safe? In this case, the clothing used on HRC 2 could be used only on HRC 1.

Taking that to it's logical conclusion says why wear anything less than a 100 cal/cm^2 flash suit?

I have no interest in letting random luck slip through either, but getting fixated on this ONE part of the picture leads to erroneous conclusions. The Stoll curve is an average. The incident energy curve fit (IEEE 1584 calculation) is an average. All the way along we are averaging and curve fitting ourselves to death. In the end it would seem like everything is a 50% crap shoot and that no consideration is being given to looking at say the variance (sigma) and taking that into account. It's a valid concern and one that I've considered in the past. However, there are reasons why the "50% threshold" is not really 50%.

The ASTM test is not reflective of reality. The reality is that the result of the ASTM test is a 50% probability of failure for the cloth sample placed perpendicular to a relatively low energy arcing fault.

Carefully examine the raw test data. It is fit to a sigmoid because that's what the procedure calls for. The reason that a sigmoid is used is frankly because the math is easy. It's the same reason that gaussians are used as an approximation to beta distributions even when the beta distribution is the actual statistically correct model. Reading anything into the printed % failure rates of the report EXCEPT the 50% cutoff generates invalid results as you can plainly see with a graph showing both the curve fit and the actual test data. Look closely and find the first test that actually failed. Is this anywhere near 7.6 cal/cm^2 or somewhat higher? This is why the sigmoid is not a good fit for looking at <50% cases to begin with...because the math surrounding it does not model the actual data well. In looking at more than a dozen similar curves, I have seen no evidence of a "long tail" where the probability never quite reaches zero, which is what the sigmoid math suggests happens. The cutoff is around 0.5 to 1 cal/cm^2 below the ATPV. Below that point, no samples fail, but the sigmoidal curve suggests a failure rate in the neighborhood of around 5-15% depending on the report. Given that roughly 30 pieces of cloth are destructively tested, there are enough samples that statistically at least 1 or 2 will show up in this range once in a while but none ever do.

The test is performed with a piece of cloth placed perpendicular to the arc. As you move off axis or even if the cloth is not pointed perpendicular to the arc, the thickness "increases" by the cosine of the angle. Real people are not flat objects. The data only applies directly in the center of the chest area assuming that it is centered directly in front of the arcing fault. If nothing else, this significantly reduces the area that could potentially be affected. Given that somewhere around 10-20% burns of 2nd degree or higher in the face/chest area is fatal, the fatality rate (which is what the ATPV is all about in the end) is drastically less than 50% even at the ATPV.

The test is for the cloth only. If the worker is doing as little as wearing an undershirt, that adds significant additional protection, even if the undershirt is not in itself arc rated because it provides additional thermal insulation as long as it is non-melting. This would seem to suggest that a simple cotton T-shirt increases the ATPV by1.2 cal/cm^2 but this vastly under-rates the protection afforded. In actual tests that have been performed with clothing systems (multiple layers), the results are far greater than just simply adding their ATPV ratings. See this web site for actual test data using clothing systems that has been published:
http://www.arcwear.com/arctest/arctest.php

Final reason is that arc flash incident energy is modelled as worst case. Even something as simple as the phase angle when the arc first initiates (if it does) dramatically decreases the assymetrical fault current and thus can give you the decrease of around 1 cal/cm^2 that you are looking for from a purely statistical point of view with no consideration given to the PPE at all.

I will give a much softer reason. I have not seen any published reports yet where a worker was injured by an arc flash while following the work methods published in 70E. Despite the obvious mathematical discrepancies, it seems to work.

If you are still not convinced, why not just adjust your cutoffs? Ignore the artificial 4/8/25/40 cutoffs that 70E uses. Use your chosen cutoff (say 5%) from the clothing manufacturer's test data. In fact as a simplification it appears in most cases that you can just subtract "1" from the ATPV rating and get to the 5-10% range that you are referring to, and to the "100% protection" that I was suggesting. I arrived at this conclusion by looking at about a dozen ATPV reports. In every case, the cutoff seems to be around 0.5-1 cal/cm^2 before there are simply no recorded clothing failures. In your example, you'd wear "H/RC 2" clothing for anything up to 7.6 cal/cm^2, not stopping at 4 cal/cm^2. I'll bet if your plant is anything like mine (I've run this kind of analysis before), only about 3-5% of your buses fall in the "underprotected hole". That's IF you don't apply any credit for any of the other reasons (wearing underclothing, off-axis, phase angle probabilities).


In summary though, I offer the major reason for not just increasing the PPE without darned good reason for doing so. Increasing PPE requirements increases the hazards of heat related injuries (heat exhaustion or heat stroke), dehydration, decreases visibility (dramatically), and decreases manual dexterity. In short, increasing the PPE to anything more than the minimum required to protect against a potential arc flash hazard also causes a greater hazard in and of itself.

I'm all for doing a full risk assessment...we want to be able to perform a job in the safest manner possible. I do not believe in decreasing/eliminating PPE where it is clearly required (where there is a signifant likelihood of an arc flash). Going overkill with the PPE especially thermally protective PPE does not result in the safest approach...quite the opposite happens.

Hi, Paul. Thanks very much for your time. My interest is not to increasing the PPE as I agree with you about the other hazards related. I am only trying to understand the value "50%". I read that in Europe there is a moviment to decrease that value (increase the safety) but unfortunately I could not find that material.
Again, thanks very much.

In the end, we have a test. It produces a result. Utilizing the result we have developed safety standards. The ATPV is just that...a result from an ASTM test. You may as well forget about the "50% passing" phrase and just treat it as a rating system with no inherent meaning whatsoever. So far I do not know of any cases where PPE was worn where the IEEE 1584 standard was used to calculate an appropriate ATPV and the person being protected was burned. I'm reasonably certain that the moment that one "slips through", we're all going to hear about it pretty quickly. IEEE 1584 standard has been around for over 10 years now. Right now I would conjecture that in most cases, we are either adequately protecting people or overdoing it. As with most others, I'm not really willing to decrease the standard without further research and details on cases where overprotection appears to be the case, but I'm also very interested in any case where current standards have failed. If the failure rate is truly 50%, I would be highly surprised if we didn't already see a published case by now which seems to suggest that overprotection is a distinct possibility. The evidence to date seems to suggest that using the ATPV rating and IEEE 1584 is at least "good enough".