Torque Verification on Terminations

Purpose

A loose termination is the most common origin of a burned panel this trade sees, and an over-tight one is not far behind. Both end the same way: a joint whose contact area shrinks under thermal cycling until it makes heat instead of carrying current. Undertightened, the contact never had enough area. Overtightened, the conductor is crushed and cold-flows away from the pressure until it does not.

"Snug and then a little more" is not a method: the hand cannot tell 20 from 30 in-lb, and the two failure directions feel identical at the driver. The value exists, it is printed on the equipment or in the instructions, and applying it takes a tool the shop already has to own.

The second half is the part that gets skipped. A torqued joint cannot be verified by torquing it again, because reapplying a wrench to a settled connection reads static friction rather than clamping force. The check that works happens under load, with a camera.

Scope

Covers terminating and verifying conductors at screw terminals, set-screw lugs, mechanical lugs and busbar connections in load centers, panelboards, meter mains, disconnects and transfer switches, on residential and light commercial work: finding the value, tool selection, conductor preparation, application, marking and post-energization verification.

Does not cover compression connectors and their dies, or the thermographic survey itself, which the infrared scan of a panel SOP owns and which this procedure hands its verification to. Does not cover the isolation sequence, owned by the lockout tagout for branch circuit work SOP.

Roles and responsibilities

Procedure

1. Establish and prove an electrically safe work condition before any terminal is touched. Acceptance: the circuit is isolated, locked, tagged and proved dead live-dead-live, the instrument proved on a known live source before and after the dead check, per NFPA 70E's proving sequence in the edition your program names, with work practices under 29 CFR 1910.333(b)(2) or 29 CFR 1926.417 on construction. Wrong looks like a lug torqued live because it is "just a screw." Stop rule: no proof, no tool on a terminal, and if nobody on site is a qualified person for that equipment the work goes to someone who is. Hazard: a torque wrench is a long conductive lever, and the reach to seat it puts a hand deeper into the enclosure than a screwdriver does.

2. Find the manufacturer's value before you decide what tool to fetch. Acceptance: a numeric value with its units, read from the equipment label, the wiring diagram inside the cover, or the installation instructions; where none exists, the fallback tables in UL 486A-486B, with the substitution recorded. NEC 110.14(D) requires torque values to be as indicated on the equipment or in the manufacturer's instructions and an approved means used to achieve them; it appeared in the 2017 cycle, so check which NEC edition your jurisdiction adopted before treating it as binding. Wrong looks like a value remembered from a different make of panel. Stop rule: no value from any of the three sources and the work stops for the supervising master, who names a source in writing. Hazard: none at this step, it is a document read outside the enclosure.

3. Match the tool to the value and confirm its units and its calibration. Acceptance: the value sits inside the tool's usable range rather than at either extreme, the tool reads in the same units the label uses or the conversion is written down at 12 in-lb to 1 ft-lb, and calibration is inside the interval the tool's own manufacturer states. Wrong looks like a 275 in-lb value read as 275 ft-lb, roughly twelve times the intended clamping, which crushes the conductor outright. Stop rule: a value outside the tool's range, or a tool past its calibration date, means the right tool comes from the shop. Hazard: none at this step, it is a tool check made at the van, and making it there rather than at the panel is why it works.

4. Prepare the conductor before anything is tightened. Acceptance: strip length matches the lug's strip gauge so no bare conductor shows outside the terminal and no insulation sits inside it, no strand is nicked or cut, the conductor is fully seated, and the lug is listed for the conductor material with no aluminium and copper sharing a terminal not listed AL/CU. Aluminium conductors are wire brushed and given antioxidant compound where the instructions call for it, in the order they state. Wrong looks like a ring of nicked strands where a stripper was rotated. Stop rule: a nicked strand is cut back and re-stripped, not tightened over. Hazard: brushing aluminium and burnishing an old lug liberates metal oxide dust and, on a previously burned connection, combustion residue, so it is done with eye protection and local ventilation rather than by blowing it out, and any respirator is under a written program per 29 CFR 1910.134, not a dust mask from the van.

5. Apply the torque in one continuous motion and stop at the indication. Acceptance: the tool reaches the value and releases or clicks once, on a single steady pull, and the connection is then marked with a torque-seal stripe across the screw head and the body. Wrong looks like clicking the wrench three or four times to be sure, which walks the fastener past the value each time. Stop rule: if the fastener turns without the tool ever reaching the value, the terminal or the conductor is wrong for that lug and it comes apart rather than being run down further. Hazard: a wrench that slips off a set screw travels into whatever is beside it, so the free hand stays out of the enclosure and the body is positioned so a slip carries the tool away from the busbar rather than across it.

6. Do not re-torque a settled connection to check it. Acceptance: the record shows the value applied once, at a stated date, with the tool ID, and verification deferred to step 7 rather than done with the wrench. Reapplying a wrench measures the static friction of a joint that has already relaxed, reads higher than the clamping force actually present, and drives an already-correct joint past its value. Wrong looks like a maintenance visit where every lug gets "checked" and half are quietly over-tightened. Stop rule: re-torque only where the manufacturer's instructions explicitly permit it and say to what value; otherwise a suspect joint comes apart and is re-terminated. Hazard: opening a settled joint releases strands under spring tension, so support the conductor as the set screw backs off rather than letting it whip.

7. Energize, load the circuit, and verify the joint thermally rather than mechanically. Acceptance: with the equipment carrying load, each terminated lug is compared against the identical neighbouring lugs at comparable current and sits within the same band the shop's thermographic procedure sets, with the spread in degrees written down. Wrong looks like signing off on the click alone, which says a value was reached and nothing about whether the joint is carrying. Stop rule: a lug outside its neighbours' band goes back to step 4 for disassembly and inspection, not to step 5 for more torque. Hazard: this step puts energy back with the cover off for the scan, so it runs under the arc flash SOP's boundary and PPE, taken from outside the boundary through an inspection window where one exists, and every protective device on the affected circuits is tripped and reset on its test button before the space is handed back.

The record this produces

Per termination, on the ticket: equipment ID, terminal by phase or position, conductor size and material, the torque value with its units, the source of that value (label, instructions or UL 486A-486B), the tool ID, its calibration date, the date applied, and the marking method used. Then one follow-up line from step 7: the load present, the terminated lug's temperature and the temperatures of the identical neighbours it was compared against.

The next technician reads the marking and the record and can tell whether a joint has been done and to what. The supervising master reads the source column, because a rising count of UL 486A-486B substitutions means the shop is buying equipment with unreadable labels. The estimator reads the step 7 lines on a panel that keeps producing findings, because that panel wants replacing rather than re-terminating.

One subpanel feeder, run

A 100 A subpanel fed from a 200 A load center, feeder being re-terminated after a nuisance-trip investigation found discolouration at one lug.

Step 1: feeder breaker opened, locked, tagged. Meter proved on a known live receptacle at 121.8 V, dead check at the subpanel lugs reads 0.0 V, meter re-proved at 121.6 V. Both bracketing readings on the ticket.

Step 2: the subpanel's main lug label gives 275 in-lb for 1/0 through 4/0 AWG. Value and units written down as printed, in inch-pounds. The jurisdiction is on an NEC cycle later than 2017, confirmed at permit, so 110.14(D) applies.

Step 3 fails. The torque screwdriver on the truck reaches only 50 in-lb, well short of 275 in-lb, and the only other tool aboard is a 20 to 150 ft-lb wrench. Converting, 275 in-lb divided by 12 is 22.9 ft-lb, which sits at the very bottom of that wrench's range, where a click-type tool is least accurate. Both fail the acceptance for different reasons: one cannot reach the value, the other is at its extreme. Stop rule fires, nothing is tightened by feel in the meantime, and a 40 to 200 in-lb wrench comes from the shop with its calibration date inside the manufacturer's interval.

Step 4: conductors are 2/0 AWG aluminium. Strip length matched to the lug's strip gauge, no bare conductor outside and no insulation inside. One strand nicked on the phase B conductor from the original strip, so that conductor is cut back and re-stripped. Wire brushed and antioxidant applied in the order the panel's instructions give, eye protection on, and the enclosure not blown out afterwards.

Step 5: 275 in-lb to each of the three lugs, one continuous pull each, one click each, torque-seal stripe across each screw head and lug body.

Step 6: nothing re-torqued. The record reads applied once, on this date, with tool ID and calibration date.

Step 7: subpanel re-energized, load restored, and after roughly an hour the three lugs read 96 F, 94 F and 95 F at comparable current. The spread is 96 minus 94, which is 2 F, and all three sit in the same band. Passes. The AFCI on the branch circuit that had been nuisance tripping is tripped and reset on its test button before the work is called finished.

Checking the run against step 3's acceptance: the acceptance names range, units and calibration, and the failing step prints a value for all three, including the conversion at 12 in-lb to 1 ft-lb that turned 275 in-lb into 22.9 ft-lb and showed the second tool was at its extreme rather than merely inconvenient.

References

• NEC 110.14(D), tightening torque values as indicated on the equipment or in the manufacturer's instructions, with an approved means used to achieve them. It entered the 2017 cycle, so confirm which edition your jurisdiction has adopted and pull that one.
• UL 486A-486B, wire connectors, for the fallback torque tables where no equipment or instruction value exists.
• 29 CFR 1910.333(b)(2) for general-industry electrical work practices and 29 CFR 1926.417 on construction. 1910.147 excludes electrical utilization work at its own (a)(1)(ii)(C).
• 29 CFR 1910.134 for any respirator use, including the written program, medical evaluation and fit test.
• The torque tool manufacturer's documentation for calibration interval and usable range, which vary by tool.
• See related: the infrared scan of a panel SOP, which owns the thermographic comparison; the lockout tagout for branch circuit work SOP, which owns step 1.