The Mechanical Checks Worth Doing Before You Leave

Why this matters

A departure checklist is only as good as the machine state you can actually create before you drive off. Every check is a question asked of a machine in a particular condition, and a machine that is cold, unloaded, and running in the wrong season simply cannot answer some of the questions on a typical list. Ask them anyway and you get an answer, you write it down, and the record now says the machine passed a test it was never in a position to fail. That is worse than leaving the line blank, because the next person believes it.

The safety state this work happens in

Every mechanical check below that involves touching a drive happens with the equipment removed from automatic control and the energy-isolating device locked and tagged under 29 CFR 1910.147, with the wheel or rotor watched until it has fully stopped, because a coasting fan carries stored rotational energy long after the contactor drops. Belt tension is set and checked on a stopped, locked drive against the manufacturer's force-deflection figure. Any running check is made with guards fitted and everyone clear of the plane of rotation, and any current reading at the starter 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. On a rooftop in summer, plan the fluid and shade breaks into the visit before the heat makes the decision for you.

The callback that reformed one shop's list

A commercial rooftop unit, spring service, belts and fan-shaft bearings replaced. The shop ran a mechanical departure list of eleven items and signed every one. The tech was not rushing and was not careless; the list was genuinely worked.

Nine days later the unit was down with a thrown belt.

Reconstructing which check should have caught it

The useful question after a callback is not "who missed it." It is "which line on this list was capable of catching this, in the state the machine was in when that line was signed."

Belt tension was on the list, and it had been set correctly to the manufacturer's force-deflection specification on a locked-out drive. The tech had done the work properly. But the belts were new, and a new belt set seats into the sheave grooves during its first period of running and loses tension as it does. The manufacturer specifies a run-in re-tension after a stated period of operation, and that period is longer than any service visit.

So the departure check verified the installed tension. The tension that threw the belt was the seated tension, which did not exist yet when the box was ticked. The line on the list was not wrong. It was answering a different question from the one the failure asked, and nothing about how carefully it was performed could have changed that.

The test they applied to the rest of the list

The shop then ran every remaining line through one question: can the state I can create before leaving reveal the defect this line is looking for? Three answers are possible, and each leads somewhere different.

Answer one: yes, this state answers it now. Six of the eleven landed here.

• Direction of rotation, checked against the arrow on the housing at a momentary bump rather than against output, because a centrifugal fan or pump running backwards still produces some flow and "it seems to work" proves nothing about rotation.
• Guards, checked by mechanism rather than by presence. Which nip, which entanglement point, which opening at which distance. The guarding article in this library carries that method and it is not repeated here.
• Lubricant level and condition at rest, which is a static property and fully observable with the machine stopped.
• Hand rotation through a full revolution on the locked-out machine, feeling for interference.
• Coastdown time, recorded with the damper or valve position noted beside it, since the system state changes the number and an unlabelled coastdown cannot be compared to anything later.
• Running current, cold, read for direction of change against the pre-service figure rather than converted into a load fraction, since motor current is not proportional to shaft load.

Answer two: yes, but only if I stay. Three items landed here, and these are the ones that get quietly dropped for schedule.

• Bearing temperature trend, which needs successive readings until the increments collapse. One reading is a number, not a verdict.
• Belt behaviour under real load, which needs the machine loaded rather than idling.
• Whether anything loosens or shifts in the first period of running, which needs that period to elapse.

The honest handling of this bucket is to decide deliberately: stay for it, or move it to a scheduled return. What is not available is signing it after twenty minutes.

Answer three: no, this state cannot answer it. Two items landed here, and this is where the reform happened.

• The run-in re-tension, which by definition requires a period of operation the visit does not contain. It came off the departure list entirely and became a scheduled return visit with its own work order.
• Performance at design load in the design season. A spring visit on a cooling machine cannot verify summer behaviour. Anything on the list that implied it did was reworded to say what was actually observed and under what conditions.

Six now-checks, three stay-or-schedule, two scheduled returns. The list still has eleven lines. Two of them stopped pretending.

The line they swapped out

The coastdown entry in bucket one is a replacement. What sat in that slot before was a line reading "check for unusual noise," which had been signed on every visit for years and had never once been marked as a finding.

A check that cannot produce a "no" is not a check. There is no threshold, no reference value, and no state in which it fails, so its only real function is to make the sheet look thorough. Coastdown time took the slot because it produces a number that can disagree with the last number.

That test travels. Read your own list and mark every line that has never in living memory been marked as a failure. Some of those are genuinely rare faults worth continuing to look for. The rest are lines with no possible negative answer, and they are consuming the attention that the stay-or-schedule bucket needs.

How they confirmed the reform worked

Not with a season of statistics, which would have taken a season. On the next unit that received a new belt set, the run-in re-tension visit was scheduled, and when it was performed the tension had fallen below the manufacturer's specification. That is direct evidence, on one machine, that the departure check could not have caught what the return visit caught, and it settled the argument about whether the extra visit was worth booking.

What changes which bucket a check belongs in

A machine that runs continuously moves items out of bucket three. On a machine that will be under real load an hour after you leave, a run-in condition is reachable that afternoon rather than next month, so the scheduled return can be short and soon rather than seasonal.

Reused components move the run-in item out of the list entirely. A belt set that was not replaced is already seated, so there is nothing to re-tension for seating. The reason to re-check it is different and belongs on the normal service interval.

A cold alignment is always a prediction. Aligning a machine at rest and verifying it at rest are the same act performed twice; the machine runs hot and grows, and whether the cold offset was right is only answerable at operating temperature. The thermal-growth and alignment articles in this library own that method, and the point for a departure list is that "alignment checked" belongs in bucket two or three, never bucket one, on any machine with meaningful thermal growth.

A hot re-check is not automatically available. Where a manufacturer specifies a re-check on a warm assembly, follow their procedure with thermal protection appropriate to the surface temperature, and never on a pressurized joint, a combustion path, or anything serving a relief or protective function. If the procedure would require working on a hot pressurized joint, the correct answer is to isolate, allow it to cool and depressurize, and accept a return visit.

How to verify you got this right

• For each line on your list, name the machine state it requires. Cold and stopped, running unloaded, running loaded, thermally stable, seasonal load. Any line whose required state is not the state you can create is misfiled.
• Confirm each line can produce a negative. If it has no threshold and no reference value, it is not a check.
• Confirm the record says which question was answered. "Belt tension set to specification at installation" and "belt tension verified after run-in" are different statements and the difference is the whole article.
• Confirm bucket-three items exist as scheduled work, with a work order and a date, not as an intention. An item moved off the departure list and nowhere else has been deleted, not deferred.
• Confirm the numbers you leave behind are comparable. Coastdown with system state, temperature rise with ambient, current with the operating condition. A number without its condition cannot be compared to the next one, which means it will never detect anything.

References

• 29 CFR 1910.147, control of hazardous energy, for isolation and stored rotational energy before any drive work
• 29 CFR 1910.333(b)(2), electrically safe work practices for energized current readings; NFPA 70E-2021 for arc-flash risk assessment and PPE selection
• Manufacturer drive documentation for force-deflection tension specifications and the specified run-in re-tension period
• See related: What a Guard Is Actually Protecting Against; How to Return a Machine to Service and Verify It; How to Allow for Thermal Growth When You Align Cold; How to Verify Your Repair Actually Held Before You Leave