Thermal Imaging for Predictive Electrical Maintenance

A bolted connection that is starting to fail does not fail quietly. It heats. Current through a resistance produces heat, and the heat rises with the square of the current, so a joint that has loosened by a quarter turn or corroded under a lug will run hotter than its neighbors under the same load, week after week, until the insulation around it carbonizes or the conductor anneals and the joint lets go.¹ That heat is visible to an infrared camera months before it is visible to anyone else. Infrared thermography is the practice of going to look for it on purpose.

For most of its history, thermography was a good idea that well-run facilities adopted and everyone else deferred. In January 2023 that changed. NFPA 70B, which had been a recommended practice since 1973, was reissued as a standard, and its language moved from "should" to "shall."² Thermography is now the one maintenance task the standard requires on every piece of electrical equipment it covers. This article explains what that requirement is, how a survey is judged, what makes a survey reliable or worthless, and how to build a program that produces engineering decisions instead of a folder of colorful pictures.

What NFPA 70B now requires

The 2023 edition of NFPA 70B organizes maintenance around an Electrical Maintenance Program and a condition assessment of each piece of equipment. Chapter 9 assigns each asset a condition of 1, 2, or 3 based on its physical condition, its criticality, and its operating environment; the worst of the three governs. Condition 1 is equipment in like-new condition with a clean, dry, secure enclosure. Condition 2 is equipment with deviations from previous results or active recommendations. Condition 3 is equipment that has missed its last two maintenance cycles, needed major component repairs in the last two cycles, or carries unresolved monitoring notifications.³

Table 9.2.2 then sets maximum maintenance intervals by task and condition, to be used where the manufacturer has not provided its own. Infrared thermography is the only task in that table listed for all equipment. Its interval is 12 months for Condition 1, 12 months for Condition 2, and 6 months for Condition 3.⁴

Section 7.4 sets the rules for how the survey is performed.⁵ Thermography is used to verify temperature differences between similar components under similar loading and between components and ambient air. All accessible and necessary covers are removed before the inspection. The temperature difference between the area of concern and its reference is documented. The inspection is performed at normal circuit loading, and where normal loading is not feasible, loading of not less than 40 percent of nominal is permitted. The circuit loading is documented and retained for future comparison.

Every one of those sentences has a consequence. The cover-removal rule puts the thermographer inside the arc flash boundary of energized equipment, which brings NFPA 70E into the room; more on that below. The loading rule means a survey performed on a Sunday with the plant idle is not a compliant survey. The documentation rule means a thermal image without a recorded load and a recorded reference temperature is not a finding; it is a photograph.

How severity is judged

NFPA 70B requires the temperature difference to be measured and recorded, but it does not publish numeric severity bands of its own. The bands the industry uses come from the ANSI/NETA Maintenance Testing Specifications, Table 100.18, "Thermographic Survey, Suggested Actions Based on Temperature Rise."⁶ The table works on two comparisons. The first is the temperature difference between similar components under similar load, for example three phase lugs on the same breaker. The second is the difference between a component and ambient air.

For similar components, a difference of 1 to 3 degrees Celsius is a possible deficiency that warrants investigation; 4 to 15 degrees indicates a probable deficiency to be repaired as time permits; more than 15 degrees is a major discrepancy to be repaired immediately. Over ambient, the bands are 1 to 10 degrees (possible deficiency), 11 to 20 degrees (probable deficiency, repair as time permits), 21 to 40 degrees (monitor until corrective measures can be accomplished), and more than 40 degrees (major discrepancy, repair immediately).⁶ The table carries a footnote that matters: temperature specifications vary by equipment type, and even within a class of equipment, such as cable, ratings differ. The bands are a screening tool, not a substitute for the component's own rating.

FM Global's Property Loss Prevention Data Sheet 5-20 uses the same bands and adds the operational translation that plant managers actually need. A probable deficiency means shutting down within months for repair. The monitoring band means shutting down within weeks. A major discrepancy means shutting down within hours.⁷ The American Bureau of Shipping, in its guidance on condition monitoring for classed vessels, is less prescriptive and says only that plus or minus one full color band from the normal operating range is usually a satisfactory envelope, while noting that limits vary by application.⁸

A competent survey report assigns each finding to one of these bands, states which comparison produced it, and records the numbers behind it.

The physics that make or break a survey

An infrared camera does not measure temperature. It measures radiant energy arriving at the detector and converts it to a temperature using assumptions the operator supplies. Get the assumptions wrong and the temperature is wrong, sometimes by tens of degrees, with no indication on the screen that anything is amiss.

The first assumption is emissivity, the ratio of a surface's radiance to that of a perfect blackbody at the same temperature. Painted enclosures and black electrical tape sit near 0.94 or 0.95, and a camera set for them reads well. Bare metal does not cooperate. Polished copper has an emissivity around 0.01, polished aluminum around 0.05, and even strongly oxidized aluminum only reaches about 0.25.⁹ Oxidized copper ranges from roughly 0.60 to 0.90 depending on how dark it has become.¹⁰ Below an emissivity of about 0.6, accurate measurement is compromised, because most of what the camera sees from a shiny lug is reflected energy from the room, not energy emitted by the lug.¹¹ A bright bus bar can appear cooler than it is, or hotter, depending on what is behind the thermographer.

The second assumption is the reflected apparent temperature, the temperature of whatever the target is reflecting into the lens.¹² It must be estimated and entered into the camera, and where an infrared window is used, the window's transmission value must be entered as well, because every window absorbs and reflects some of the energy passing through it.¹⁰ The practical fix for a low-emissivity target is old and reliable: apply a small patch of electrical tape to the surface, set the camera to 0.95, set the background correction to the reflected temperature, and read the tape.¹¹ The result is good to roughly two degrees or two percent.

The third assumption is that the target fills enough of the detector to be measured. Every camera has a spot size ratio, the distance-to-target ratio at which a single measurement point is valid. A 50:1 ratio means a one-inch target can be measured from no more than 50 inches away.¹³ Manufacturers also caution that a target must cover at least three by three pixels on the detector for the reading to be trusted, and that digital zoom does not help; it enlarges the pixels you already have.¹⁴ A thermographer standing across a switchgear room reading a quarter-inch terminal screw is producing a number, not a measurement.

The fourth is load. Because heating rises with the square of the current, a joint at 40 percent load produces roughly one sixth of the heat it produces at full load. NFPA 70B's 40 percent floor is a minimum, not a target; practitioners recommend surveying above 80 percent of rated load where possible and allowing the equipment to run at load for about half an hour before imaging so temperatures stabilize.¹ A survey scheduled when the plant is quiet will pass equipment that fails at shift change.

Safety: NFPA 70B and NFPA 70E in the same room

NFPA 70B says covers shall be removed. NFPA 70E, the workplace electrical safety standard, says that opening hinged doors or removing bolted covers to expose energized conductors is a task for which an arc flash incident is likely, on any equipment, in any condition.¹⁵ Performing thermography from outside the restricted approach boundary, without opening anything, is listed as a task for which an arc flash is not likely. The difference between those two rows of Table 130.5(C) is the difference between a technician in street clothes and a technician in arc-rated PPE selected by an incident energy analysis.

OSHA permits energized inspection under 29 CFR 1910.333 where de-energizing is infeasible, and testing that can only be performed with the circuit energized is the regulation's own example of infeasibility.¹⁶ Thermography is by definition such a test; there is nothing to see on a dead bus. But the permission comes with the requirement that only qualified persons do the work and that they use the precautionary techniques and protective equipment the hazard demands.

Infrared windows resolve the tension for equipment that will be surveyed repeatedly. A window is a fixed aperture with an infrared-transmitting medium, installed in an enclosure door or wall so that the equipment can be imaged with the enclosure closed and its integrity intact; UL 50V is the outline of investigation that covers them.¹⁷ With the door closed and the equipment in a normal operating condition, the thermographer is reading a panel meter, in 70E's framing, rather than exposing energized parts. NFPA 70B supports the approach directly: Section 6.8 requires that where a recognized hazard presents increased risk during maintenance, a study be conducted to develop design options that reduce it.⁵ Windows are the design option for thermography. They do not eliminate judgment; each window has a field of view, and the survey plan has to confirm that every connection of interest can actually be seen through the windows installed.¹⁸

Who is qualified to do this

The camera is the cheap part. The standard does not name a certification, but it requires that testing personnel be qualified to operate the equipment used and qualified to perform the procedure on the specific equipment, and that any instrument providing a measurement carry calibration traceable to a national standard.¹⁹

Two qualification frameworks are in general use. ISO 18436-7 defines three categories of thermographer for condition monitoring. Category I performs surveys using established procedures, after a minimum of 32 hours of training and 12 months of experience. Category II selects the technique, applies the theory, interprets results, and recommends corrective action, after 64 hours and 24 months. Category III develops the program, establishes procedures and severity criteria, and supervises the others, after 96 hours and 48 months.²⁰ ASNT's SNT-TC-1A, the employer-based scheme used across nondestructive testing, defines Levels I, II, and III along the same lines: Level I performs specific examinations to written instructions; Level II sets up equipment and interprets results against codes and specifications; Level III establishes techniques and qualifies the others.²¹

The distinction that matters to a facility owner is between someone who can operate a camera and someone who can defend a finding. A Category I or Level I thermographer can collect good images to a procedure someone else wrote. Deciding what the images mean, what the severity is, and what the repair should be is Category II work, and building the program is Category III.

What a survey actually finds

The classic finding is the high-resistance connection: a loose or corroded lug, a bolted bus joint that has relaxed, a breaker line-side terminal that was never torqued. ASTM E1934, the standard guide for examining electrical and mechanical equipment with thermography, lists loose or deteriorated connections, short circuits, overloads, and load imbalances as the causes of electrical thermal exceptions.²² Insurers have quantified the stakes. Hartford Steam Boiler's loss data puts the failure rate of electrical equipment outside a scheduled preventive maintenance program at three times the rate of equipment inside one, and holds that more than two thirds of electrical system failures are preventable by routine maintenance.²³

Not every finding is hot. A blown fuse or an open pole shows cold against its neighbors, and a phase carrying no current on a three-phase load is as much a defect as one carrying too much.²⁴ Transformers and motors have their own arithmetic: an operating temperature 10 degrees Celsius above rating roughly halves insulation life, and every additional 10 degrees halves it again.⁸ Motor bearings, which thermography sees well, account for about half of motor failures in the IEEE 493 survey data, with windings accounting for another quarter.²⁵

New installations deserve a survey of their own after they have run under load. ABS notes the reason: thermal cycling loosens fasteners, and a connection that was correct at commissioning can be loose by the end of the first season.⁸

Shipboard and waterfront electrical systems

Thermography is at its most valuable, and its most difficult, aboard a vessel. ABS lists it as an applicable condition monitoring technique for circuit breakers, motors, electrical systems, transformers, and turbine generators, and it requires that in a diagnostic role it be performed only by trained, certified, and experienced personnel.⁸ Under the ABS Guide for Surveys Based on Machinery Reliability and Maintenance Techniques, equipment covered by an approved condition monitoring plan can earn credit toward the Special Continuous Survey of Machinery, which is to say a thermography program can reduce the intrusive survey burden the class society would otherwise impose.²⁶

The environment fights the physics at every step. ABS names the degrading factors: humidity from rain or condensed steam, extraneous radiation from surrounding hot objects, insulation or barriers between camera and target, and distance.⁸ A 2025 systematic review of infrared thermography in maritime systems adds salt aerosols, surface contamination, metallic reflections, restricted access, rapidly changing ventilation, and the elevated background temperatures and dense layouts of machinery spaces, and finds that existing ISO and IEC standards do not fully account for shipboard conditions.²⁷ The same review reports that only about 10 percent of the maritime industry applies condition-based maintenance at all.

Two practical rules follow. First, the survey has to be timed to the load profile of the vessel, which is nothing like a building's: a generator switchboard in port on shore power is not the same equipment as the same switchboard underway with propulsion auxiliaries running. ABS's baseline is normal steady-state load with surrounding equipment operating normally.⁸ Second, infrared windows carry more weight aboard ship than ashore, because opening a switchboard in a seaway to take a picture is a poor bargain. Marine approvals for windows exist from ABS, DNV, and Lloyd's.¹⁸

What insurers expect

FM Global's data sheet 5-20 calls for infrared surveys on all electrical equipment annually, an infrared check of low-voltage breaker connections, and an infrared check of automatic transfer switches in both positions.⁷ The Hartford Steam Boiler standard for electrical preventive maintenance calls for surveys of switchgear, distribution panels, cable and bus connections, motor control centers, and starters, during peak demand where possible and at not less than 40 percent of rated load, by qualified technicians only.²³ A facility whose survey cadence, load conditions, and thermographer qualifications fall short of those expectations should assume its insurer will notice after a loss, if not before.

Building the program

A thermography program that produces decisions has six parts.

Baseline. The first survey of every asset is the reference against which every later survey is read. It should be performed at representative load, with emissivity, reflected temperature, distance, and load recorded for each image, so that the next thermographer, who may not be the same person, can reproduce the conditions.

Interval. Assign each asset a condition under NFPA 70B Chapter 9 and set the interval from Table 9.2.2 or from the manufacturer's recommendation where one exists. Condition 3 equipment goes to six months. Criticality alone can put an asset in Condition 3: if its failure could endanger personnel, it belongs there regardless of how clean the enclosure looks.³

Access. Decide, asset by asset, whether covers come off under an energized work permit and arc-rated PPE, or whether infrared windows go in. For equipment surveyed every six or twelve months for the life of the plant, windows pay for themselves in the first PPE cycle they eliminate.

Report. NFPA 70B's annex on documentation lists what a report should carry: date, time, and thermographer qualification; equipment identification; camera settings including emissivity and reflected temperature; ambient temperature and humidity; circuit loading; thermal and visible images; the temperature difference for each finding and its reference; and a severity classification with a recommended action.²⁸ A report without the load and the settings cannot be trended and cannot be defended.

Close the loop. Every finding gets a repair order, and every repair gets re-imaged under load. A finding that was corrected and never re-surveyed is an open question.

Trend. The value of the program compounds. A connection that reads 4 degrees over its neighbors this year and 9 degrees next year is telling you something that neither survey alone would.

MD Engineering performs infrared surveys of shipboard, waterfront, and industrial electrical systems, reports findings against the NETA Table 100.18 bands with the load and camera parameters recorded for every image, and carries findings through to repair and re-survey. If your last survey report cannot tell you what the load was, what emissivity was used, and what the temperature difference was for each finding, you do not have a survey. You have pictures. NFPA 70B now asks for the survey.