How to Return a Machine to Service and Verify It

Why this matters

The most dangerous minute on a mechanical job is not when the machine comes apart. It is the first rotation after it goes back together, when a person is standing next to an assembly that has never turned in its current configuration, holding an expectation instead of evidence. Treat that as "hit the button and see" and you are betting on every assumption you made in the last three hours at once. The way to stop betting is to stop thinking of return to service as a restart and start thinking of it as a sequence of machine states, each with a condition that must be satisfied before you leave it, and each capable of revealing defects that no other state can.

The states, and why the order is fixed

Skipping a state does not defer its defect. It relocates the defect to whoever finds it next, usually the customer, usually at the worst hour.

Step 1: Close out the isolated-and-open state properly

Before anything closes, do the physical account: every tool out, every rag out, every temporary support removed, every fastener at its specified torque, every shim and setscrew where it belongs, guards refitted rather than leaned against the wall. Turn the shaft by hand through at least one full revolution, with the machine still locked out, and feel for a catch or a rub. A hand turn is the only test in this whole procedure that costs nothing and catches a hard interference before it becomes a bearing.

Step 2: Perform the release from lockout as the standard writes it

29 CFR 1910.147(e) is not a formality and it has a specific shape. Before any lock comes off: inspect the work area and confirm nonessential items have been removed and the machine components are operationally intact; check that all employees are safely positioned or clear of the machine; and notify affected employees that the devices are coming off and the machine is about to be energized. Each person who applied a device removes their own device; nobody removes anybody else's.

If the work included electrical work inside a panel or on branch-circuit conductors, that side of the job sat under 29 CFR 1910.333(b)(2) rather than 1910.147, which excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C), and the proving-dead sequence used before that work was the live-dead-live check in NFPA 70E-2021, 120.5. Re-energization there means the panel is closed and secured first, not last.

Step 3: Energize without running, and confirm nothing moves

Restore power and stop. Confirm the machine does not move, that the controller shows no fault, and that no start command is pending from a building system, a timer, a pressure switch or a remote operator. This is the only state in which you can discover that the machine was going to start the instant power returned, and you discover it while standing clear rather than while holding a wrench.

Step 4: Bump it, and check direction against the arrow

One person at the disconnect or the hand-off-auto switch, everyone else clear of the plane of rotation and clear of the discharge, all guards fitted. Momentary start, immediate stop.

Check the direction against the rotation arrow on the housing, never against output. This matters more than it sounds. A centrifugal fan running backwards still moves air in the correct direction, at greatly reduced volume, and a centrifugal pump running backwards still develops some head. "It seems to be working" is not a direction test on either one, and a machine left running backwards after a motor or controller change will be diagnosed six months later as a capacity complaint. If a shaft-end reflective mark is the only way to see rotation, use a non-contact optical instrument aimed through an existing port from outside the plane of rotation.

While it coasts down, time it and write the number down. Coastdown time is only comparable to future coastdowns on the same machine in the same system state, since a damper or valve position changes it, so record the system state alongside the number. It is a crude measure and it is the cheapest drag baseline you will ever get.

Step 5: Run it under load, cold, and read current against expectation

Bring it to normal operating load and take running current at the starter, which is energized electrical work under 29 CFR 1910.333(b)(2), in arc-rated PPE and voltage-rated gloves after an arc-flash risk assessment, clamping one insulated conductor with the door opened no further than the reading needs.

Compare against the pre-repair figure if you have one. Interpret the direction, not the magnitude: motor current is not proportional to shaft load, because the magnetizing component is roughly constant, so a change in current tells you reliably which way the load moved and tells you its size only through that motor's own current-versus-load curve.

Step 6: Stay for the thermal state, because it is the one that costs you

This is the state people skip, and it is the state that produces the callback. Take bearing housing temperature and ambient at intervals and read the increments, not the absolute numbers.

The criterion is that successive increments are collapsing toward zero. A machine whose rise is still climbing by the same amount at the sixty-minute mark has not stabilized, no matter how comfortable the absolute number looks. Compare the settled value against the bearing manufacturer's and the lubricant's limits, both of which are machine-specific and come from their documentation rather than from a remembered figure.

Broader system performance verification - whether the machine is meeting its design duty - is a separate job with its own method, and this library has an article for it. This step is narrower: it asks whether what you just assembled is thermally settling.

Worked example: a fan after a bearing replacement

Both fan-shaft bearings replaced, belts re-tensioned to the manufacturer's force-deflection specification on the locked-out drive, guards refitted.

State: isolated and open. Hand-turned two full revolutions, no catch. Tools counted out against the tray.

Release from lockout performed per 29 CFR 1910.147(e), with the operator on that side of the building notified before anything came off.

Energized, not running. No motion, no fault, and the building system was confirmed to be holding the unit off rather than merely not having called for it yet.

Bump. Direction correct against the housing arrow. Coastdown timed and recorded with the damper position noted alongside it, as the new baseline for this machine.

Loaded, cold. Running current came in at 0.79 of full-load amps, against 0.86 recorded before the repair. The direction of that change is the useful part: less current for the same duty is consistent with drag having been removed, which is what a bearing replacement is supposed to do. Converting 0.86 and 0.79 into shaft-load fractions would need the motor's own curve, so the honest statement is "moved the right way by a meaningful amount," not a percentage of load.

Thermal. Ambient held at 24 C through the run. Bearing housing readings gave rises over ambient of 17 C at fifteen minutes, 25 C at thirty, 28 C at sixty, and 29 C at ninety. Those four readings give increments of plus 8, plus 3, plus 1: climbing steeply at first, then flattening hard. That collapsing pattern, not the 29 C itself, is what says the machine has settled, and the settled value was then checked against the bearing and lubricant limits from the manufacturer's data.

What a different trend would have meant. Had the increments read plus 8, plus 7, plus 7, the machine would still be climbing at sixty minutes and nobody leaves. On a freshly serviced bearing the first suspect for that pattern is over-greasing: excess grease churns rather than lubricating and generates heat until it purges. The response is the manufacturer's purge procedure with the drain port open, and only where that port is outside the guarded envelope and reachable without entering the plane of rotation. If it is not, stop the machine and lock it out under 29 CFR 1910.147 first. The response is never more grease.

If a protective device operates during any of this, that device has just told you something no instrument on your belt did. Establish why it opened before resetting it. Resetting or replacing a correctly operating overload, limit or temperature switch arrives at the same place as jumpering it, one step slower.

How to verify you got this right

• Confirm you can name what each state told you. If you cannot say what the bump revealed that the energized-not-running state could not, you ran the sequence without using it.
• Confirm the direction check was made against the rotation arrow, with the answer written down. This is the single most commonly assumed item in the whole procedure.
• Confirm the thermal reading is a trend and not a snapshot. One temperature is a number. Three or more with collapsing increments is a verdict.
• Confirm no protective device was reset without a cause. If one operated and the record says only "reset," the job is not finished.
• Leave the baselines behind, in the equipment record: coastdown time with system state, running current, settled temperature rise over ambient, and belt tension specification. The next person's diagnosis is only as good as the numbers you left them, and none of these can be reconstructed later.

References

• 29 CFR 1910.147, control of hazardous energy, including the release-from-lockout requirements at 1910.147(e) and the removal of each employee's own device
• 29 CFR 1910.333(b)(2), electrically safe work practices for energized measurements and for electrical work excluded from 1910.147 at (a)(1)(ii)(C); NFPA 70E-2021, 120.5, for the live-dead-live proving sequence used before that work
• Manufacturer bearing, lubricant and drive documentation for temperature limits, relubrication and purge procedures, and force-deflection tension specifications
• See related: How to Verify a System Is Operating as Designed; How to Verify Your Repair Actually Held Before You Leave; What a Guard Is Actually Protecting Against