Infrared Scan of a Panel

Purpose

Produce a thermal survey of an energized panel that a third party can act on: every hot spot located, classified against a stated criterion, and corrected or scheduled. A loose lug is the most common electrical fire origin a shop can actually find before it burns, and it announces itself as heat long before it announces itself as a trip.

This needs an SOP rather than a camera because three things quietly invalidate a scan and none of them show up in the image: the panel can be too lightly loaded for a defect to develop visible rise, the surface being read can be shiny metal reporting far below its real temperature, and the deadfront can be on, in which case you are photographing a steel cover. A survey that ignores any of the three produces a clean report on a failing panel, which is worse than no survey because the customer now believes it was checked.

Scope

Covers a routine or complaint-driven thermal survey of energized panelboards, switchboards, disconnects and motor control equipment at 600 V or less.

Does not cover the remedial work. Once a connection is classified for repair, the de-energize, re-torque and re-scan sequence is the panel-maintenance job, and the burned-receptacle SOP owns the case where damage has already reached the conductor. Medium-voltage equipment is out of scope; the qualification, approach boundaries and PPE there are a different conversation.

Roles and handoffs

Procedure

1. Establish the energized-work basis before anything else, because the cover has to come off. A steel deadfront is effectively opaque to the wavelengths the camera reads, so a scan through a closed cover is not a scan. Removing it on energized equipment is energized work: it requires documented justification, an energized electrical work permit, a qualified person, and arc-rated PPE selected from the arc flash risk assessment, per NFPA 70E-2021, Article 130, in the edition your employer's electrical safety program has adopted. Acceptance: permit signed and PPE on, or a listed infrared window already installed. Wrong looks like a cover coming off in street clothes because it is only a photo. Stop rule: no permit and no window means no scan; reschedule and quote the window.

2. Confirm the panel is carrying enough load for a defect to show. Heat rise at a resistive joint goes roughly with the square of current, so a defect scanned at low load can look benign. Read each phase with a clamp ammeter at the main and record it against the panel or equipment rating. Acceptance: at least 40 percent of expected peak load on the phases being surveyed, which is the loading level thermographic acceptance criteria are written around. Wrong looks like a Saturday scan on a building at a fifth of its weekday draw. Stop rule: below 40 percent, either bring load on or reschedule to a loaded period, and never report the survey as complete at low load without printing the load figure next to the result.

3. Set emissivity and reflected temperature before the first image, not in the software afterward. Bare copper and aluminum bus have emissivity in the range of roughly 0.05 to 0.10 and will read tens of degrees low. Either aim at painted, insulated or oxidized surfaces, or place a known-emissivity target such as a strip of electrical tape near the joint and read that. Acceptance: the surface you quote a temperature from has a stated emissivity, and reflected temperature is set from an ambient measurement in that room. Wrong looks like a report quoting a shiny lug at 34 C when the joint is at 90 C. Stop rule: if you cannot get a valid surface, say so on that finding rather than publishing the low number.

4. Image every connection in the enclosure, not just the ones that look interesting. Work top to bottom: line and load lugs at the main, each phase of the bus, every breaker's line-side stab area and load lug, the neutral bar, and the ground bar. Take a wide shot and a close shot of each area, plus a matching visible-light photo so a reader can locate the finding later. Acceptance: full coverage documented, with each image carrying its own spot temperature and the ambient. Wrong looks like six images of a forty-two space panel. Hazard clause: keep the camera and both hands outside the restricted approach boundary and never rest a hand on the enclosure edge to steady a shot, because the shot is not worth entering the space to get.

5. Classify each finding against a stated criterion, using both comparison bases. Compare the suspect component against a similar component under similar loading, and also against ambient air. Widely used thermographic acceptance criteria, such as the NETA Maintenance Testing Specifications table your specifying engineer references, treat a similar-component difference of 1 to 3 C as a possible deficiency warranting investigation, 4 to 15 C as a probable deficiency to repair as time permits, and above 15 C as a major discrepancy to repair immediately; the ambient comparison runs on its own bands and above 40 C over ambient is likewise immediate. Acceptance: every finding carries both differences and the band it fell in. Wrong looks like a temperature with no comparison at all. Stop rule: where the two bases disagree, take the more severe band and say which basis drove it.

6. Project the rise to expected peak load before you classify, and print the correction. A measured rise taken below peak understates the defect. Scale it by the square of the current ratio: rise at peak is approximately the measured rise times the square of expected peak current divided by measured current. That square law assumes joint resistance does not change with temperature and that the enclosure sheds heat roughly in proportion to the rise, which holds well enough below about 100 C; above that, resistance climbs with temperature and the real rise runs higher, so treat the projection as a floor. Acceptance: the projected figure appears as its own line beside the raw one. Wrong looks like the raw rise classified as if the panel were at peak.

7. Take the immediate action the severe bands require, before you leave the building. A finding in the immediate-repair band on a circuit serving life safety, or one already showing discoloration, deformed insulation or the smell of hot plastic, does not wait for a report. Acceptance: the affected circuit is de-energized, locked and tagged under 29 CFR 1910.333(b)(2) and proved dead live-dead-live per NFPA 70E-2021, 120.5, or the customer has been told in person and in writing that it remains energized against advice. Wrong looks like a severe finding emailed to a facilities inbox on a Friday. Stop rule: if you cannot reach a decision-maker, red-tag the panel, notify the building's responsible person on site, and escalate to your own lead.

8. Restore, then close the loop with a re-scan requirement. Deadfront back with every screw, torque restored on anything you disturbed, and each finding carrying a required action, an owner and a re-scan date. Acceptance: every open finding has a scheduled re-scan under comparable load. Wrong looks like a repaired lug that nobody re-imaged, which is exactly how a joint that was never actually re-torqued survives into the next season. Stop rule: no finding closes on the strength of a work order alone; it closes on a second image.

When the site does not match the assumption

The building will not permit a cover-off scan on energized equipment, common in healthcare and data spaces: quote listed infrared windows sized for the gear, scan through them next visit, and record that only windowed or accessible equipment was surveyed. A partial survey is honest; one stated as complete is not.

Load cannot reach 40 percent because the tenant space is vacant: scan if it is the only window available, but classify nothing as a pass. Record the load percentage and mark that section inconclusive.

The cover cannot come off without pulling a bus plug or unit: do not force it. Route to the lead, who chooses between an outage window and a window installation.

The record this produces

One survey record per panel. The header carries date, technician, camera and its last calibration date, building, panel identifier, and measured per-phase load with the rating it is a percentage of.

Then one block per finding: thermal image, matching visible image, the component named specifically enough to find again, spot temperature, the similar-component reference and which component it was, ambient, both differences, the emissivity used and the surface it was read from, the projected rise at peak load, the classification band, and the action with an owner and a date.

The lead reads the classification to schedule. The insurer or facility engineer reads the images and the load basis, which is why the load figure sits in the header rather than in the tech's memory. Next year's thermographer compares the same joint at a different load, which only works if this year's basis was written down.

Worked pass: 225 A three-phase panel, retail tenant, midweek afternoon

Step 1: cover-off scan under a signed energized electrical work permit, arc-rated coverall, face shield and balaclava per the arc flash risk assessment on the label. Step 2: clamp readings are A 91 A, B 96 A, C 88 A, against an expected weekday peak of 152 A per phase from the customer's own metering. B phase at 96 A is 42.7 percent of the 225 A panel rating, so the survey clears the 40 percent floor and the load figures go in the header.

Step 3: the main lugs are bare aluminum, so a strip of electrical tape goes on the insulated boot below each lug and readings come off the tape at emissivity 0.95, with reflected temperature set from a 34 C ambient in the electrical closet.

Step 4: forty-one images across the main, bus, twenty-six breakers and both bars.

Step 5 produces the finding. The B phase main lug reads 62 C. The A and C lugs, same bus, same enclosure, at 91 A and 88 A, read 41 C and 43 C, averaging 42 C. Similar-component difference is 20 C, above 15 C and therefore the immediate-repair band. Difference over the 34 C ambient is 28 C, which sits in a lower band. The more severe basis governs, and both figures are printed.

Step 6, correction printed: measured rise 20 C at 96 A; expected peak 152 A; current ratio 152 divided by 96 is 1.58; squared, 2.51; projected rise at peak is 20 times 2.51, or 50 C over the similar lugs. The raw 20 C and the projected 50 C both appear on the finding, and the projection is labelled a floor.

Step 7 takes its stop rule. The store is open and the affected phase feeds the walk-in cooler and the register circuits. The store manager cannot authorize an outage, so the finding escalates: panel red-tagged, the responsible person named on the record, the owner reached by phone that afternoon. The lug is re-torqued that evening under lockout, and step 8's re-scan the following Tuesday at 94 A on B phase reads 44 C against a 43 C average on A and C, a 1 C difference, which closes the finding.

The joint that mattered was found at 20 C of rise on a panel at 43 percent load, which is exactly the case that a low-load survey reports as clean.

References

• NFPA 70E-2021, Article 130 and 120.5, in the edition your employer's electrical safety program has adopted, for the energized-work permit, PPE selection and the live-dead-live sequence
• 29 CFR 1910.333(b)(2), the general-industry lockout and verification practice for electrical work
• NETA Maintenance Testing Specifications, thermographic survey suggested actions table, in the edition your specifying engineer references, for the classification bands used in step 5
• NFPA 70B, in the edition your authority having jurisdiction or insurer references, for electrical equipment maintenance program requirements
• See related: Burned Receptacle Response and Circuit Condemnation SOP, Panel Labeling and As-Built Handover SOP