Electrical
Infrared thermography inspection field guide for electrical equipment
Read the heat before the connection fails: load the gear, compare the phases, judge delta-T against NETA, and write the find down so it gets fixed at the right outage.
Direct answer
Infrared thermography is the inspection of electrical equipment with a thermal camera that reads surface temperature, so a loose, corroded, or overloaded connection running hot under load shows up before it fails. It only works with current flowing, commonly at least 40 percent of rated load, and NETA criteria set the action thresholds.
Key takeaways
- Infrared thermography needs current flowing, with equipment carrying at least 40 percent of rated load before a scan is meaningful.
- NETA criteria: a rise of 4 to 15C over a similar component (11 to 20C over ambient) is a probable deficiency, repair at next outage.
- A rise over about 15C over a similar component, or over 40C over ambient, is a serious deficiency to repair immediately.
- Read connections off high-emissivity surfaces (lug, painted housing), never bare shiny metal, which reflects surrounding heat and reads false.
- Clamp the current before calling a hot phase a bad connection, because an unbalanced load makes the heavier phase run hot honestly.
Infrared thermography, and what it catches before the failure
Infrared thermography is the inspection of electrical equipment with a thermal imaging camera that reads the heat coming off a surface and turns it into a picture you can measure. A loose, corroded, or overloaded connection has higher resistance than a good one, so it runs hotter than everything around it when current flows through it. The camera sees that heat. You find the bad connection while the gear is still working, not after it has burned itself open and dropped the load.
That is the whole value of the method. It is non-contact and it is done with the equipment energized and loaded, so you catch the defect at the stage where it is cheap to fix, a clean and a re-torque, instead of the stage where it is a failed bus and an unplanned outage. The connection that is about to fail almost always tells you first by getting hot. Heat is the early warning, and infrared is how you read it.
What IR does not do is find a problem that is not making heat yet. A connection with no load on it sits at ambient no matter how loose it is. So infrared is a load-on, current-flowing inspection, and a scan of dead or lightly loaded gear is close to useless. Get that wrong and you will sign off a panel full of bad connections because none of them were working hard enough to show.
How does heat reveal a bad connection?
A bad connection makes heat through resistance heating, the I squared R loss. Power dissipated at the joint is the current squared times the resistance of that joint, so two things drive the temperature: how much current is flowing and how bad the connection is. The current-squared term is why heat climbs so fast as load comes up, and why a marginal connection that looks fine at light load lights up under a heavy one.
The catch is that the camera does not measure temperature directly. It measures radiated infrared energy and converts it to a temperature using the surface's emissivity, the fraction of energy a surface radiates compared to a perfect emitter. A painted breaker, oxidized copper, or taped insulation has high emissivity and reads close to true. Bare, shiny metal has low emissivity, radiates little, and instead reflects the temperature of whatever is around it, including the heater across the room and the thermographer.
So you read the find at the connection: the lug, the insulation, the painted housing, not off the polished bus bar. A shiny bus can read cool when it is actually hot, or read hot when it is only mirroring a heat source nearby. That reflected energy is the reflected apparent temperature, and ignoring it is how a new thermographer reports a fault that is not there or misses one that is.
Emissivity and the reflected-temperature trap
Emissivity is the setting that decides whether the camera's number is the surface's real temperature or a fiction. A surface with high emissivity, near 0.95, radiates almost all the heat it holds, and the camera reads it accurately with little correction. A surface with low emissivity, like polished copper or aluminum down around 0.05 to 0.10, radiates almost nothing of its own and the camera mostly sees reflected energy from the room. The same hot connection can read 35°C on its shiny lug and 60°C on the painted cable a foot away, and the painted reading is the true one.
The fix on the bench is to read a high-emissivity surface and read it consistently. Painted breaker housings, taped insulation, oxidized terminals, and phenolic or rubber surfaces all sit high on the scale and give an honest number. Some shops apply a small patch of high-emissivity tape or paint, or a stick-on emissivity target, to a bus or a lug they need a true number from, then read the patch. For comparison work you care about the delta between similar surfaces more than the absolute value, so reading both phases off the same kind of surface keeps the comparison honest even if the absolute emissivity is off.
Reflected apparent temperature is the partner setting. It tells the camera how hot the surroundings are so it can subtract the reflection. On a low-emissivity target near a hot source, getting reflected temperature wrong throws the reading by tens of degrees. The rule that keeps you out of trouble is simple: do not trust an absolute temperature off bare bright metal, and never base a delta-T on a reflective surface when a high-emissivity one is sitting right next to it.
How much load do you need for an IR scan?
You need current flowing, and as a working minimum the equipment should carry at least 40 percent of its rated load before the scan means anything. That figure shows up in both NFPA 70B and NETA guidance as the floor for a meaningful electrical survey. Verify the exact number against the current edition, but 40 percent is the number the trade carries in its head.
The reason is the I squared R curve. Heat at a bad connection rises with the square of the current, so at low load even a genuinely failing joint barely warms above its neighbors and slips under the threshold. Scan at 20 percent load and you can walk a switchboard full of deficiencies and call it clean. Scan the same board at 60 or 80 percent and the bad connections separate from the good ones by a wide margin.
Higher is better. If you can schedule the survey for peak load, or at least well above 40 percent, the lower-grade defects that are still early show themselves. Give the equipment time under load too, on the order of a half hour or more, so the connection reaches a steady temperature instead of being caught while it is still heating. And whatever the load is, write it down, because a delta-T with no load behind it is a number nobody can act on.
What delta-T means repair?
Delta-T is the temperature rise of a hot component over a reference, and the reference is either a similar component under similar load or the ambient air. The action depends on which reference and how big the rise. A small rise gets monitored. A larger rise gets repaired at the next outage. A large rise gets repaired now.
The widely used thresholds come from NETA's thermographic survey criteria. A rise of roughly 1 to 3°C over a similar component, or 1 to 10°C over ambient, is a possible deficiency worth investigating. A rise of about 4 to 15°C over a similar component, or 11 to 20°C over ambient, is a probable deficiency to repair at the next scheduled outage. A rise over about 15°C over a similar component, or over 40°C over ambient, is a serious deficiency to repair immediately. Confirm these against the current NETA edition, because the exact bands and the wording move between revisions.
The component-to-component comparison is the stronger of the two references, because both parts see the same conditions. Comparing to ambient is the fallback when there is nothing similar to compare against, like a single transformer or a lone disconnect. Either way, the delta-T is meaningless without the load it was taken at recorded next to it.
| Delta-T over similar component | Delta-T over ambient air | Suggested action |
|---|---|---|
| 1 to 3°C | 1 to 10°C | Possible deficiency, investigate |
| 4 to 15°C | 11 to 20°C | Probable deficiency, repair at next scheduled outage |
| Not specified in this band | 21 to 40°C | Monitor until corrective action can be scheduled |
| Over 15°C | Over 40°C | Serious deficiency, repair immediately |
Phase-to-phase comparison, the most reliable read
The cleanest find in electrical thermography is the odd phase out. On a three-phase circuit carrying a balanced load, the three phases should run at nearly the same temperature. When one lug, one pole of a breaker, or one phase of a connection runs hotter than the other two, that phase has the problem, and you found it by comparing it to its own siblings under the exact same load.
This beats comparing to ambient because the three phases share everything: the same load, the same enclosure, the same air, the same time of day. The only variable left is the connection itself. A 15°C spread between phase A and phase C on the same breaker is a high-resistance connection on the hot phase, and you do not need to know the ambient or the emissivity precisely to trust the comparison.
There is one trap. A genuinely unbalanced load makes the heavily loaded phase run hotter for an honest reason, not a defective one. So you confirm the current on each phase with a clamp meter before you call a hot phase a bad connection. If the currents are balanced and one phase is hot, it is the connection. If the hot phase is also carrying more current, the heat may just be the load, and the find is load balance, not a defect.
Field example: a hot B-phase lug on a 400 A feeder
A 480 V three-phase feeder is running about 70 percent of a 400 A rating, so each phase is near 280 A. Through the closed cover, the camera reads the three load-side lugs on the breaker: A phase at 41°C, C phase at 43°C, and B phase at 68°C, against an ambient of 30°C. The lugs are identical, under the same load, in the same enclosure, so the comparison is clean.
Work the numbers two ways. Over the coolest similar phase, the delta-T is 68 minus 41, or 27°C, which lands well into the immediate-repair band over a similar component. Over ambient it is 68 minus 30, or 38°C, which is close to the over-ambient immediate threshold as well. Both references point the same direction, and they agree, which is the cross-check you want before you call something serious.
Before writing it up, confirm the load is balanced with a clamp meter. If A, B, and C all read near 280 A and only B is hot, the heat is a high-resistance connection at the B-phase lug, not the load. The action is repair at the next available outage if it can be scheduled quickly, or sooner if the feeder is critical, with a re-scan after the lug is cleaned and re-torqued.
| Reading | Value |
|---|---|
| System and load | 480 V three-phase, about 70 percent of 400 A |
| A phase lug | 41°C |
| B phase lug | 68°C |
| C phase lug | 43°C |
| Ambient air | 30°C |
| Delta-T over coolest phase | 27°C (immediate repair band) |
| Delta-T over ambient | 38°C (near immediate threshold) |
Severity, priority, and what gets fixed when
The delta-T sorts the find into a priority, and the priority is what an owner actually plans around. A small rise is a watch item that rides until the next routine survey. A moderate rise is a planned repair scheduled into the next outage, because forcing an outage early for it costs more than it saves. A large rise is a now problem, where the cost of an unplanned failure outweighs the cost of an emergency outage.
Two things bend the priority off the raw number. Criticality is the first: the same delta-T on a feeder to a chiller in July or a connection feeding a process that cannot stop justifies acting faster than it would on a spare circuit. The load at the time of the scan is the second: a serious-looking rise found at 45 percent load is worse than it looks, because the same defect at full load runs much hotter, and a borderline rise found at light load deserves a re-scan at higher load before it is dismissed.
Write the priority as an action and a timeframe, not just a color. Repair immediately, repair at the next scheduled outage, or monitor and re-scan next cycle tells the planner what to do. A find tagged red with no recommended action and no load behind it is a number that gets argued about instead of fixed.
What to scan on an electrical survey
You scan the places where current changes hands, because that is where resistance and heat collect. Bolted lug terminations, crimp and split-bolt splices, breaker line and load connections, fuse clips and fuse bodies, disconnect blades, motor terminal boxes, contactor and starter terminals, transformer connections and bushings, and the joints in busway, especially the bolted joints and the plug-in stab connections where a busway plug clips onto the run.
The body of a conductor rarely fails. The connection is where it happens, so that is where you spend the time. On busway, the joints and the plug-in points are the suspects, and a hot busway joint under load is the same I squared R story as a hot lug, just carrying a lot more current and a lot more consequence when it lets go.
The choice every survey faces is open the cover or scan through a window. Opening a dead-front to expose energized connections puts a person inside the arc-flash boundary, which is the most dangerous moment of the whole job. A permanently installed infrared window lets you read the same connections with the cover closed. Where there is no window, opening the panel is a qualified-person task under full arc-flash PPE, and that decision is not the thermographer's alone.
IR windows and arc-flash safety
An infrared window is a permanently installed port, made of an IR-transparent crystal or polymer, that lets you scan energized connections without opening the enclosure. That matters because opening an energized panel to scan it is the step that exposes a worker to an arc flash. The window removes the exposure: you read the connection through the closed cover and never cross into the arc-flash boundary.
NFPA 70E treats this directly. Scanning through an installed window, with doors and covers closed, is recognized as a task with a much lower likelihood of an arc-flash event than opening the equipment, and the same equipment-condition logic that governs energized work applies. Confirm the current 70E edition for the exact task table and approach-boundary language before you write a procedure around it.
If there is no window and the cover has to come off, this becomes energized work. A qualified person, the arc-flash PPE for the incident energy at that equipment, and the established work permit and boundaries all apply, and the call to open it belongs with the people who own the electrical safety program, not with the thermographer wanting a cleaner shot. The grounding and bonding of the gear should already be sound before anyone works it energized, which is its own discipline covered in the grounding electrode system and bonding guide.
The camera and the qualified thermographer
A bad operator makes bad data with a good camera, so the person matters more than the tool. The recognized path is thermographer certification under ASNT's SNT-TC-1A framework: Level I to scan and collect data to a procedure, Level II to set up the analysis, interpret the results, and judge severity, Level III to write the procedures. Level I commonly runs around 32 classroom hours plus documented field experience before certification. Confirm the current requirements, because the training hours and experience logs are set by the employer's written practice under SNT-TC-1A, with ISO 18436 as the parallel path where that route is used.
The camera spec that matters is detector resolution, thermal sensitivity, and focus. More detector pixels, 320 by 240 at a minimum and 640 by 480 for fine work, put more measured points on a small lug at a distance. Thermal sensitivity, the NETD, decides whether a small rise is even visible. And focus is not optional: an out-of-focus thermal image does not just look soft, it reports the wrong temperature, because the energy from the hot spot smears across cooler pixels and the peak reads low.
Before the first image, the thermographer sets emissivity and reflected apparent temperature for the surfaces being read, and confirms the camera's temperature range covers what is expected. Those settings are the difference between a number you can put in a report and a number that is off by tens of degrees.
Survey conditions and steady-state load
A delta-T is only as good as the conditions it was taken in, so the survey records the conditions, not just the temperatures. The equipment should be at a steady-state load, meaning it has carried its load long enough to reach a stable temperature, not just switched on. A connection caught mid-warmup reads lower than its real steady value.
Ambient temperature gets recorded because the over-ambient criteria depend on it directly. Outdoors, wind is the quiet enemy: even a light breeze convects heat off a hot connection and pulls the apparent temperature down, so a real defect can read as a minor one or vanish. Note the wind, and treat outdoor over-ambient numbers on a windy day as understated. Sun load does the opposite, heating surfaces that carry no fault at all, so account for solar gain on outdoor and rooftop gear.
The one thing that ties it all together is the load at the time of the scan. Record the percent of rated load and the measured current per phase. Without it, the delta-T cannot be compared to a later survey, cannot be judged against the 40 percent floor, and cannot be defended when someone asks whether the find was real or just a lightly loaded panel.
What goes in the report
A thermographic report that an owner can act on pairs a thermal image with a normal visible-light photo of the same component, so the repair crew knows exactly which lug to open. Each find carries the location, the equipment and component identified, the load at the time of the scan, the measured temperatures and the delta-T, the reference used (similar component or ambient), the severity, and the recommended action with a priority.
The visible photo is not a formality. A thermogram alone shows a hot blob with no context, and the crew sent to fix it cannot tell phase A from phase C, or one identical breaker from the next in the row. The paired visible image, with the hot spot marked, is what makes the find repairable by someone who was not standing there when it was shot.
The other half of a good report is the baseline. The first survey of a piece of gear becomes the reference the next one is compared against, so the same component scanned the same way a year later shows whether it is stable, improving after a repair, or trending hot. Without a repeatable baseline, every survey starts from zero and the trend, which is often the real signal, is lost.
Beyond electrical: mechanical, envelope, and roofs
The same camera and the same physics find heat anywhere a problem makes it, so a thermographer rarely scans only electrical gear. On the mechanical side, a failing bearing runs hot, a misaligned coupling or an overloaded motor shows up as heat, and so do a slipping belt, a blocked cooling passage, or an overheated gearbox. The method is the same: find the component running hotter than its healthy twin or its own history.
On the building side, infrared reads missing or wet insulation, air leakage, and moisture, because wet and dry materials hold and release heat differently. A roof moisture survey uses exactly this: trapped water under the membrane stores the day's heat and releases it slowly after sunset, so a scan in the right evening window shows the wet areas glowing against the dry roof. That is its own discipline, with its own timing and confirmation by core sample or capacitance meter, and it overlaps with electronic leak detection on low-slope roofs.
The point for an electrical tech is that the instrument is multi-use, and the discipline of load, emissivity, reflected temperature, and a measured reference carries across every one of those applications. The targets change. The method does not.
How often should you run an IR survey?
The common cadence is an annual infrared survey of the electrical equipment, with more frequent scans on gear that is in worse condition or more critical to the operation. NFPA 70B, the recommended practice for electrical equipment maintenance, moved in its 2023 edition toward treating periodic infrared inspection as a standard part of the program rather than an option, and where it is adopted or referenced by an authority or an insurer, that expectation has teeth. Confirm the interval against the current 70B edition and its equipment-condition tables, because the frequency scales with condition.
Insurers drive a lot of this. Property carriers such as FM Global commonly require a periodic IR survey of electrical equipment as a condition of coverage, and a missed or failed survey shows up at renewal. The interval and the qualified-thermographer requirement are usually spelled out in the policy or the loss-prevention recommendations.
There is also the acceptance scan. New gear gets an infrared survey once it is energized and under load during commissioning, which catches the connections that were never torqued right before the building is occupied. That belongs in the commissioning sequence, and on a data center or mission-critical job it is part of the power QA covered in the data center electrical commissioning guide, where the plant is proven under load before the critical load arrives.
After the find: repair and re-scan
Finding the hot connection is half the job. The other half is fixing it correctly and proving the fix. The repair is energized work run in reverse: de-energize, lock out, verify dead, then open the connection, inspect it, clean the contact surfaces of oxide and contamination, and re-make it. A connection that ran hot from being loose gets re-torqued to the manufacturer's value, not to feel, because over-torque damages the joint as surely as under-torque, and the torque discipline on a bolted connection is its own skill worth getting right.
Cleaning matters as much as torque. A connection that overheated has often already oxidized or pitted the contact faces, and re-torquing a corroded joint to spec still leaves a high-resistance connection. Bad contact surfaces get cleaned or the lug gets replaced, and aluminum connections get the right antioxidant treatment where the manufacturer calls for it.
Then you re-scan. Put the gear back under load, let it reach steady state, and shoot the connection again. The delta-T should be gone. A re-scan that still shows the hot spot means the repair did not take, usually a contact surface that was never cleaned or a deeper problem in the device. The re-scan closing the loop is what turns a find into a documented, verified fix instead of a work order someone signed.
What to document
Every find becomes a record, and the record is what survives the technician leaving and the next survey happening. Capture enough that someone a year out can locate the component, understand the severity, reproduce the comparison, and confirm whether it was fixed.
| Field to record | Why it matters |
|---|---|
| Equipment and component | Locates the exact lug or device for the repair crew |
| Load percent and current per phase | A delta-T is meaningless without the load behind it |
| Reference temperature and type | Similar component or ambient, the basis for the delta-T |
| Delta-T | The measured rise that drives the severity |
| Severity and priority | Maps the find to monitor, next outage, or immediate |
| Recommended action | Tells the crew what to do, not just what is wrong |
| Thermal plus visible image | Makes the find repairable by someone who was not there |
Common mistakes
- Scanning at no load or below 40 percent, so real defects never warm enough to show.
- Ignoring emissivity and reflected temperature, then reporting a shiny bus that is reflecting, not hot.
- Comparing dissimilar components instead of phase-to-phase or to a true twin.
- Not recording the load, which leaves the delta-T impossible to judge or compare later.
- Calling a hot phase a bad connection without clamping the current to rule out an unbalanced load.
- Opening an energized panel without the arc-flash PPE and the qualified-person controls.
- Shooting out of focus, which reads the hot spot low and hides a real defect.
- Skipping the re-scan after repair, so nobody knows whether the fix actually took.
Field checklist
Want this checklist to run itself on every job — with photo proof and a signed record crews can hand the customer? That's FieldOS.
Standards and references
NETA, the InterNational Electrical Testing Association, is where the thermographic survey criteria live. NETA's maintenance testing specifications include the temperature-difference table that sorts a find into investigate, repair at the next outage, or repair now, by delta-T over a similar component or over ambient. Cite the current NETA edition, because the bands and the wording change between revisions, and NETA acceptance testing covers the commissioning scan as well.
NFPA 70B, the recommended practice for electrical equipment maintenance, frames the periodic infrared program and the inspection intervals, and its 2023 edition strengthened the expectation that infrared inspection is part of a maintenance program. NFPA 70E governs the safety of doing the survey on energized equipment: the arc-flash boundary, the PPE, and the recognition of closed-cover IR-window scanning as lower-risk. NFPA 70, the NEC, is the installation code behind the gear being scanned.
Thermographer qualification runs through ASNT's SNT-TC-1A, the employer's written practice for Level I, II, and III, with ISO 18436 as the parallel condition-monitoring route and ISO 18434 covering thermography in that framework. Insurers such as FM Global and the equipment manufacturers impose their own survey intervals and torque values, and where they do, those control. Confirm every edition, section, and threshold against the documents the jurisdiction and the project have actually adopted before citing them on a report.
Units, terms, and conversions
Infrared thermography goes by a few names and the readings come with their own vocabulary, so the same idea reads differently across a camera menu, a report, and a standard.
It is called IR thermography, thermal imaging, or an infrared survey. Temperatures are in degrees C or degrees F, the rise is the delta-T, and the camera's sensitivity is its NETD. The two settings that make a reading honest are emissivity and reflected apparent temperature, and the safe way to read energized gear is through an IR window.
- Delta-T
- The temperature rise of a hot component over a reference, either a similar component under similar load or the ambient air
- Emissivity
- The fraction of energy a surface radiates compared to a perfect emitter, 0 to 1; bare metal is low, painted or oxidized surfaces are high
- Reflected apparent temperature
- The temperature of the surroundings that reflects off the target and into the camera, which corrupts the reading on low-emissivity metal
- IR window
- A permanently installed infrared-transparent port that lets you scan an energized enclosure without opening it
- NETD
- Noise-equivalent temperature difference, the camera's thermal sensitivity, which sets the smallest rise it can resolve
- Thermographer level
- ASNT SNT-TC-1A qualification tiers, Level I to scan, Level II to interpret and judge severity, Level III to write procedures
FAQ
What does an infrared inspection find?
An infrared inspection finds electrical components running hotter than they should, almost always at the connections: loose or corroded lugs, failing breaker poles, overloaded conductors, and bad busway joints. The heat comes from resistance at the joint under load. It also catches load imbalance across phases and, on mechanical gear, hot bearings and motors.
How much load do you need for an IR scan?
You need current flowing, with a common working minimum of 40 percent of rated load before the scan is meaningful. Heat at a bad connection rises with the square of the current, so a lightly loaded scan hides real defects. Higher load is better, and you record the actual load and current with every reading.
What delta-T means repair?
Under the widely used NETA criteria, a rise of about 4 to 15°C over a similar component, or 11 to 20°C over ambient, is a probable deficiency to repair at the next outage. A rise over about 15°C over a similar component, or over 40°C over ambient, means repair immediately. Confirm the current edition.
What is an IR window?
An IR window is a permanently installed port made of an infrared-transparent crystal or polymer that lets a thermographer scan energized connections without opening the enclosure. Because the cover stays closed, the worker stays outside the arc-flash boundary, which removes the most dangerous step of the survey. NFPA 70E recognizes closed-cover scanning as lower risk.
Why does a shiny bus bar read the wrong temperature?
Bare, shiny metal has low emissivity, so it radiates little of its own heat and instead reflects the temperature of its surroundings into the camera. A shiny bus can read cool when it is hot, or hot when it is only mirroring a nearby heat source. Read the connection, the lug, or a painted surface, not the polished metal.
How often should electrical equipment be scanned with infrared?
A common cadence is an annual infrared survey, with more frequent scans on critical or poor-condition gear. NFPA 70B frames the periodic program and its 2023 edition strengthened it, while property insurers such as FM Global often require a survey for coverage. New gear gets an acceptance scan under load at commissioning. Confirm the interval.
What do I do after infrared finds a hot connection?
De-energize, lock out, and verify dead, then open the connection, clean the contact surfaces, and re-make it, torquing a bolted joint to the manufacturer's value rather than by feel. Replace a pitted or corroded lug instead of re-torquing it. Then put the gear back under load and re-scan to confirm the delta-T is gone.
Is phase-to-phase comparison better than comparing to ambient?
Yes. Comparing the three phases of one circuit under the same load isolates the connection, because the phases share the same load, enclosure, and air, so the odd hot phase is the find. Comparing to ambient is the fallback for a single component with no twin. Clamp the current first to rule out an unbalanced load.
Can I do an infrared survey without opening the panel?
Yes, if the equipment has installed IR windows you can scan the energized connections through the closed cover and stay outside the arc-flash boundary. Without windows, reading the connections means opening the panel, which is energized work requiring a qualified person and arc-flash PPE. The decision to open belongs with the electrical safety program.
People also ask
Codes cited in this guide
This guide is written and reviewed against the published standards below. Always confirm the current adopted edition with the authority having jurisdiction.