Arc Flash Study Requirements: What OSHA and NFPA 70E Actually Mandate

No OSHA regulation orders an "arc flash study" by name. The obligation that sends facilities shopping for one is assembled from three places: the OSH Act's General Duty Clause, the electrical safe-work rules in OSHA 29 CFR 1910.333, and NFPA 70E — the consensus standard OSHA cites when it explains what protecting workers from arc flash looks like. What NFPA 70E actually requires is an arc-flash risk assessment, and it accepts two methods for the PPE side of that assessment: an incident-energy analysis or the PPE category tables. Either route complies within its limits. "Study" is an industry habit, not a regulatory term — and that distinction decides what a small facility actually has to buy.

Last reviewed: July 21, 2026.

The requirement stack, document by document

Document	What it says	What it demands of an employer
OSH Act §5(a)(1), General Duty Clause	The workplace must be "free from recognized hazards that are causing or are likely to cause death or serious physical harm"	Arc flash qualifies: OSHA's interpretation letters treat industry consensus standards such as NFPA 70E as evidence that a hazard is recognized and that protection against it is necessary. An unassessed hazard is citable even though OSHA never adopted 70E as a rule
OSHA 29 CFR 1910.333	Live parts must be deenergized before work unless the employer demonstrates that deenergizing is infeasible or introduces greater hazards	The default is an electrically safe work condition. Whatever energized work remains needs documented safe work practices and protective equipment — which forces the hazard to be quantified
NFPA 70E, Article 130	A shock and arc-flash risk assessment before energized work: likelihood, severity, protective measures, and the label content of §130.5(H)	The assessment must be documented, must produce an incident energy or PPE category per piece of equipment, and must be reviewed on a cycle
NEC 110.16	Installation-side arc-flash warning markings on equipment	Drives label content on the installation side — see NEC 110.16

OSHA's 2008 interpretation letter on verifying deenergized circuits closes the loop: workers exposed to arc-flash hazards must be protected, and the letter points to a flash hazard analysis or NFPA 70E's task-based tables as the accepted ways to size that protection. That is the whole legal architecture. No clause in any of it says "commission a study."

Two methods for the PPE decision — either one complies

NFPA 70E gives an employer two routes, and they are alternatives, not steps:

1. Incident-energy analysis. Calculate the incident energy at the working distance for each piece of equipment — IEEE 1584-2018 is the calculation method in practice, with a published scope of three-phase AC systems from 208 V to 15 kV — then select arc-rated PPE whose rating meets or exceeds the result.
2. PPE category method. Look the task and equipment up in the 70E category tables and apply the assigned category — valid only while the equipment's fault current and clearing time sit inside the table row's stated limits.

One method per piece of equipment. A calculated 9.5 cal/cm² result stays an incident-energy result; it is never relabeled "Category 3" — the arc flash PPE category page covers that line in detail. And the tables are not the shortcut they appear to be: exceed a row's fault-current or clearing-time parameter and the table method is void for that equipment, which pushes you back to the calculation.

Neither route requires a facility-wide engagement. A per-panel incident-energy analysis with sound inputs is as compliant, for that panel, as the same number arriving inside a bound engineering report. What the requirement cares about is that the value exists, is documented, and matches the equipment as it stands today.

The five-year review

An assessment is not permanent. NFPA 70E caps the review interval at five years, and any change that moves the inputs — a service upgrade, a protective-device swap, a utility fault-current revision — restarts the work for the affected equipment immediately. The five-year rule and its change triggers have their own page; the short version is that a compliant assessment carries a date, and under NEC 2026 that date is printed on the label for any inspector to read.

What a full study deliverable contains

When you do commission the engineering version, "study" has a reasonably standard scope:

Component	What it is
Field data collection	Nameplates, conductor runs, breaker and fuse models with settings — verified against the one-line diagram, or reconstructing it
Short-circuit analysis	Available fault current computed at every bus from utility data and transformer impedances
Protective-device coordination	Time-current curves checked so upstream and downstream devices trip in the intended order
Incident-energy calculations	IEEE 1584-2018 energy and arc-flash boundary per bus and equipment location
Labels	Printed per NFPA 70E §130.5(H), carrying the assessment date
Report and recommendations	The model, its assumptions, results, and energy-reduction options: settings changes, maintenance-mode switching, device swaps

EC Magazine's guidance on performing studies under IEEE 1584-2018 treats field data collection as its own project phase, and it is the phase your own records shrink or inflate. A facility that can hand over a current one-line diagram buys a smaller study than one where the engineer starts by tracing conduit.

When the full study is the right call — and when it is not

Hire an engineering firm when any of these is true:

• A customer, insurer, or AHJ requires a PE-stamped report. Software output is not a stamp.
• The system is multi-source or network-fed — on-site generation, large motor contribution, closed-transition transfer — where per-panel treatment misses the interactions.
• Anything operates above 15 kV or on DC; both sit outside IEEE 1584-2018's published scope.
• You need protective-device coordination work, not just label values.
• The documentation is gone. If nobody can say what feeds a panel, rebuilding the one-line is engineering fieldwork, not calculation.

The opposite case is common and much cheaper: a small facility on a simple radial system — one utility service, one transformer, panels fed in a tree with no closed ties — where every panel sits inside the 1584-2018 window and the two hard inputs are obtainable: available fault current (how to get it) and protective-device clearing time. For that facility, a per-panel incident-energy analysis produces the documented assessment the requirement actually asks for. One scope line to respect: our engine accepts clearing times up to 2 seconds and refuses longer entries rather than extrapolating — a device that has not cleared a fault in 2 seconds is an engineering problem, not a data-entry problem.

For what the engineering engagement costs and the per-panel arithmetic behind quotes, see arc flash study cost — pricing lives on that page so the figures stay in one place.

A screening path that respects both cases

A defensible sequence for a small facility: run each panel through the free arc flash calculator — IEEE 1584-2018 at both arcing-current branches, every intermediate value shown — and read the results as triage. Panels with modest energies and clean inputs can be labeled straight from the calculation; a label pack turns those results into print-ready §130.5(H) labels with the assessment date on them. Panels that hit scope walls — long clearing times, missing fault-current data, energies that point at engineering controls — are exactly the ones worth paying an engineer to model. When the answer is a firm, arc flash study companies covers how to scope the request and compare quotes; the practical move is collecting two or three written scopes and comparing what each firm counts as a deliverable.

What the calculator path does not include, stated plainly: nobody stamps anything, nobody walks your site, and the inputs are yours to get right.