Does your arc flash study need to be redone?

The question often arrives in the same form: someone opens an arc flash study, finds “IEEE 1584-2002” in the methodology section, and wants to know whether the whole thing has to be thrown out. A superseded edition does not by itself make a study wrong. What IEEE Std. 1584-2018 did change is which inputs the model asks for, and many 2002-era studies do not record them. In my experience, the decision often turns on that silence more than on the edition number itself.

What changed, briefly

The full technical account of the model changes (the test program behind the 2018 edition and how the model changed shape as a result) is written up at length on a separate site: IEEE 1584-2018 vs 2002: what actually changed↗(opens in a new tab, external site). This page is about the decision rather than the model, so what follows is only the part that bears on it.

Three consequences of IEEE Std. 1584-2018 matter to anyone holding a study built on IEEE Std. 1584-2002:

• Electrode configuration became an explicit input. How the conductors are arranged (their orientation, and whether the arc terminates in open air, inside a box, or against an insulating barrier) changes how energy is directed toward a worker. The 2018 model asks which arrangement applies. The 2002 model separated open air from enclosed equipment but did not ask about orientation, so many studies built on it never recorded the answer, and it may not be recoverable from the document.
• Enclosure geometry entered the calculation. The physical size of the box is treated as affecting the result rather than as incidental. A study that never recorded enclosure dimensions may need a trip back to the field before it can be run under the newer model.
• The applicability envelope moved. Each edition states the range of voltages, conductor gaps, and fault currents its model is valid within, and those ranges are not the same. Equipment that sat comfortably inside the 2002 model's range may sit near or outside the 2018 model's, or the reverse. Read the ranges in your own copy of each guide and compare them against what your study actually modeled.

The 2018 edition also handles the variation in arcing current differently. Both editions recognize that the arcing current is not a single deterministic value and that a lower-than-expected arcing current can hold an upstream protective device in a slower part of its curve, producing a longer clearing time and therefore more incident energy. In my experience this step is often skipped in studies of every vintage, so it is a good one to look for early.

Why a 2002 study is not automatically wrong

Two things are often conflated: whether a study used a superseded model, and whether its results are unsafe. They are not the same question. A 2002-model study of a facility whose equipment sits squarely inside that model's range, whose protective devices have not been re-set, whose transformer has not been replaced, and whose assumptions are all written down is a document that can be examined and checked. A 2018-model study that never states its electrode configuration is not made more reliable by its edition alone.

In practice, the newer edition changed what a study has to record to run its model. It made explicit several things that used to be assumptions, and an assumption that has been made explicit is one a reviewer can check. That is why the edition question often turns into a documentation question.

Four things the decision often turns on

When this question comes up in practice, it usually comes down to four points. None of them requires recalculating anything. The first three can be answered from the study document you already have. The fourth also needs the current state of the plant.

1. Does the study state its edition and model at all? If not, the rest of this list is hard to evaluate, and that is the finding.
2. Is the equipment inside the stated applicability limits? A result produced outside the range the model is valid within should not be relied on without further justification, and some studies do not flag when they have gone outside it.
3. Was the reduced-arcing-current case run? If the study reports one incident energy value per bus with no second case and no narrative explaining its absence, the governing value may not be the one on the label.
4. Does the study still describe the system you have? The edition matters little if a transformer was swapped, a tie was added, or a trip unit was re-set since the study was issued. A model that no longer matches the plant needs attention under either edition.

Where the periodic review fits

In NFPA 70E-2018, 130.5(G) addresses two triggers for revisiting an incident energy analysis: a system change that could affect its results, and the passage of 5 years. As of November 2026 the current edition is NFPA 70E-2027, so read the edition that applies to you for the current interval and wording.

The practical point is that a system change, not the calendar, is often what makes a study stale. Facilities change continuously and quietly. The periodic review is simply the moment somebody finally has to look, which is why the edition question and the five-year question tend to arrive in the same email. If you are preparing for one, the edition is one of several things to check, and often not the most urgent.

What to do with an old study

Start by reading it. A full restudy can be a large expense, and it is sometimes the right answer. A cheaper first step is to establish what the existing document supports: which model, which inputs, which assumptions are stated, which are silent, and whether the plant still matches. With that evidence in hand, the redo-or-not conversation with your Engineer of Record, who makes the call, is easier to have.