What's your opinion on using the terms: "Arc Flash Risk Assessment" in place of using the terms: "Arc Flash Hazard Analysis" or "Arc Flash Study"? Can these terms be used interchangeably? I don't think so, since an "Arc Flash Risk Assessment" also involves "estimating the likelihood of occurrence of injury....but who can really estimate the likelihood of occurrence and how is that done?

Comments?

This is a general question referring to how risk assessments are done for safety purposes for a whole lot more concerns than just arc flash. As you stated, RISK = hazard + likelihood. I think that we can agree that at some point we have to define what is an acceptable risk. And the way to define that is to look at comparable risks. In other words we should treat fatalities with the same care and concern regardless of the cause. And we don't want to make a mountain out of a mole hill either...we are not going to treat paper cuts with the same level of concern as a fatality. Comparable risk is defined in several standards as about 1 in a million for a fatality. In other words, we want the likelihood of a fatality to be less than 1 in a million per worker per year. This number is used in a variety of places. For instance the "one in a million" number is even used for the general populace by EPA.

Then to get to all the other types of injuries and risks, although this is somewhat tenuous, we appeal to the Heinrich Safety Triangle. Heinrich claimed that statistically that for every fatality, we have 10 serious injuries (hospitalization), and 100 minor injuries, 1,000 first aid cases, and 10,000 near misses. The numbers aren't exact but they're close enough. Heinrich's theory has been mostly disproven but it's a reasonable target. So that means we want serious injuries involving hospitalization such as arc flash to be less than 1 in 100,000 workers per year, minor injuries to be less than 1 in 10,000 workers per year, and so forth. So this gives us a good yard stick to go by. It turns out that on average we have about 1 arc flash injury per 100,000 workers per year on average according to data published by ESFI so there is definitely statistical data to back up this as a target if our goal is to at least meet industry averages.

Next we get into the heart of things...into risk assessment methodologies. There is sort of something defined in NFPA 70E in Annex F but it's basically garbage. It is sort of based on an ANSI standard for risk assessments but that standard is first off based on moving machinery and second based on repeated exposures on a routine basis such as an assembly line worker working around moving machinery, and the types of injuries that standard is meant to address are generally much less severe. So we have a huge mismatch. A far better standard is the LOPA (Layers of Protection Analysis) methodology supported by the CCPS (Center for Chemical Process Safety). Their standards are based on comparatively (hopefully) rare chemical plant accidents which are caused usually either by human error or by issues with relatively rare machine failures such as electrical equipment failures. This is a far more appropriate safety standard to follow for analyzing likelihood.

Next we have to look at basically 3 situations when it comes to likelihood. The first case is looking at how often electrical equipment actually fails and in particular fails with an arcing fault. Several large surveys have been conducted by IEEE and and others that have been summarized in IEEE 493 which is now part of the 3000 series of standards. Among other things, the arcing fault failure rates for circuit breakers are around 1 in a million to 1 in 100,000 while those for disconnects are up in the 1 in 1,000,000,000,000 (10^12) range. Similar statistics are available but these are the major ones that concern arc flash. So in the case of equipment "spontaneously" failing and blowing up, it should be pretty obvious that this is extremely rare. 70E even states this in an informational note where it states that 600 V class equipment is generally safe for normal operation. Based on IEEE 493 data I would argue that this extends beyond 600 V and includes at least all medium voltage equipment as well. The point here though is that if the chance of an arc flash is due purely to equipment failure under normal circumstances, the likelihood is so small that it meets the comparable risk standard without doing anything special. Thus the tables in 70E also state that no PPE is needed for "normal operation".

What we can't really do is quantify much about the IEEE 493 statistics as far as what lead up to arcing fault failures. For instance do th failures happen spontaneously (doubtful) or as a result of change of state of the equipment, and if it's change of state, is that by human interaction (opening or closing a breaker) or on it's own such as when a breaker trips? Unfortunately we can't get to that level of detail so we have to assume that it's all human interaction (operating a breaker manually for instance), but obviously based on the statistics it's close enough for the purpose.

So this gets us to the next case. That is when equipment is under let's call it "abnormal operation". For instance NEMA AB-4 which is the maintenance and testing standard for all UL molded case breakers states that the breaker must be visually inspected before being placed back into service. There is a simple reason for this. Breakers are designed to withstand a limited number of faults near their design limits. So if it is designed to withstand say 100 faults, was the last trip one of the 99 faults or is fault #100 in which case it is no longer serviceable and must be replaced? If it was trip #100 then obviously we've entered "abnormal operation". Similarly a simple visual inspection (without opening the cover) should suffice in capturing most of these situations. OSHA 1910.269 annexes give a list of signs to look for such as swelling, scorch marks, bubbled paint, racked, rusted, leaking fluids, etc. Remember we're talking here about normal operation vs. abnormal operation and a simple visual inspection should suffice in categorizing the differences. Just as NEMA AB-4 recommends for circuit breakers and NFPA 70E itself states for instance to observe the blades in a visible break disconnect. So by taking these extra steps we can maintain the "1 in 100,000" or better likelihood when it comes to normal operation of equipment or really any task for that matter that depends almost entirely on equipment functioning properly if we screen out actual or potentially faulty equipment.

The final category is cases where the task involves manipulating electrical components in a way in which whether or not an arc flash could occur depends partly or almost entirely on human performance...not making mistakes. This would include for instance blindly fishing cables into an energized enclosure or landing connections on or near live bus while it is energized, or racking breakers or MCC buckets on and off live bus although this can be argued is more about the equipment design than the task. In every case described it is not a matter of mechanical action of the equipment...it's when human interaction comes into play. So we have to ask ourselves can we achieve a level of human performance of 1 in 100,000 or lower mistakes per worker per year? Again we don't really have to. There have been tons of studies done on the subject particularly for military, aircraft control, and nuclear power plant applications. The conclusion is that in unstressed conditions (no emergencies, healthy individuals not having trouble at home or lack of sleep or sick or any other risk factors) with very repetitive and simple tasks AT BEST human error rates get to about 1 in 1,000 while with tasks where there is a lack of training, or the task is not clear or easy, or it's an emergency situation, or any of the risk factors are present for the people doing the task, the likelihood of an error goes up to about 50%. To get to that 1 in 1,000 case requires a very simple, routine task that is drilled on over and over again with military precision and training. It just doesn't happen by accident. And yet even at that level of error rates, it is still two orders of magnitude higher than our comprable or acceptable error rate of 1 in 100,000. So the result is that if the task involves human performance either partly or entirely to avoid causing an arc flash, there is simply no way to achieve that level of performance. Thus an arc flash risk exists if the hazard is great enough.

However even without this level of rigor in analysis, and granted it's pretty simple since we have a binary decision to make (does human error play a significant role in the task safety or not), but even failing that there are already standards in place for doing the likelihood part of the analysis. The tables in 70E provide a pretty large task list. There is also a table of tasks in Annex A of 30 CFR 1910.269 (OSHA 1910.269) that is a lot shorter and easier to apply, and the one I recommend. So rather than reinventing the wheel, one can simply adopt one of these lists. Or if you prefer something more akin to a peer reviewed paper Roberts has addressed this issue in exactly the same way in a couple IEEE papers.

Hello

I totally agree with your comments regarding using the terms interchangably. The problem goes deeper however and that is we use the term "Risk Assessment" so fluidly, we anticipate that workers particularly QEW have an understanding of what a risk assessment is and what it entails. This is not the case. I cannot speak for the United States however here in Canada there needs to be a great deal of educating on risk assessment. It is not something taught in apprenticeship school, not exactly a topic around the coffee truck or break times. Ourselves as knowledgeable in electrical safety understand the theory of "likelihood of occurance", avoidance, severity, etc. but most electricians do not. I think we need to do a whole lot of educating and awareness on this subject and stop worring so much about changing terms used in NFPA 70E and CSA Z462.

I will put it this way. The risk assessment standards (ANSI, CCPS LOPA, etc.) and mind you NFPA 70E and CSA Z462 are NOT on that list of risk assessment standards recognize two things. First they recognize that any one person still has a significant chance of letting things slip through the cracks. It is simply not possible for any one person to be effective at doing risk assessments. They might miss things and they will taint the results with their own biases. It should be pretty obvious to almost everyone how bias creeps in to where electricians would like to not do something in terms of safety if it severely inconveniences them and likewise safety generalists tend to go to the extremes when it comes to likeilhood and typically bring an operation to its knees if allowed to push policies without any checks on what they are doing. So the standards all recommend a team effort consisting of a mix of subject matter experts (electricians, production workers, safety generalists, etc.) working together on risk assessments in order to both avoid the error of omission and biases from the system. This isn't a pencil whipping form that can be done in under 5 minutes though many operations are implementing and calling the so-called "pre task plans" risk assessments when they are anything but and have not been proven to provide any improvement in safety at all.

That being said, I can also confidently say that I've been trained in performing risk assessments and leading teams in performing them. Non-engineers are actually really, really good at these things because they tend to have a very practical and hands-on view of the situation. Given the chance and good coaching they definitely keep the others in line and also the process reinforces to them what the very real risks are and why they are doing what they are doing...it's a great training tool for all involved.

There are two kinds of risk assessment standards. The tables in NFPA 70E/CSA Z462 are a prescriptive standard. They give very specific rules. The "engineering" approach is a performance based standard...you can implement it any way you want. The performance safety standards are a three step process and this might sound a whole lot like the tables in 70E/Z462 for good reason. First evaluate the hazard of the task with NO protective measures whatsoever. What is the worst that can happen? There can be multiple hazards such as cut with a wire, a fall, shock, or arc flash. If the injury is nonexistent or minor, then move on. If not, apply all the protective measures and evaluate the likelihood. If it modifies the hazard this modification might also be needed. Compare this to the acceptable risk standard (1 in 100,000 per year for arc flash for instance). If it is acceptable then the process is complete. Move on to the next task. If not, think up other ways to protect against it. If there are multiple competing approaches such as changing the protective device settings and PPE, use the "hierarchy of controls" to select among them...PPE is always the last resort.

In practice working as a coach one of the big problems I had is that I can't keep a stable team when it comes to production and maintenance workers. Between shifts, different skills, and the fact that doing risk assessments at a chemical plant is boring, often emotionally intense, and burns people out pretty fast, they get rotated. The result is that I spend about 5-10 minutes out of a 2-3 hour session doing introductory training on doing risk assessments quite often. But what I also found is that tradesmen pick this stuff up VERY fast. After doing just one introductory task, they get it and we move through the process very quickly.