Walking the Map

He signs in at the front office, and most days that's all it is, especially for a tech who's been to the plant before. His being there is enough to put the contractors element on the route, though, and most of what that element asks for should already be done. The plant checked his company's safety record before hiring it. His company trained him on safe work practices and the hazards of the work and documented it. The plant made sure his company knew the hazards of the process and the parts of the emergency plan that apply, and it controls when he's in the process area.

What shows up on the day is a JHA or a pre-work authorization, the job's own list of what could hurt him.

Plenty of plants isolate the package and blow it down without touching the panel, and that isn't wrong. I was trained to pull it down first, and I still would wherever the system allows it, because every pound the compressor pulls back into the system is a pound that isn't behind the valves when he opens the separator. Some plants won't do it because one setpoint change can set off an MOC, a procedure revision and training.

That's a reasonable answer to the program they have, but it lets the maze pick the method instead of the hazard.

This plant pulls it down, so the job starts in the plant's operating procedures. With the plant's operator on board, he adjusts the setpoints the procedure calls for, including the package's low-suction cutout, and lets the machine pull the package down to the floor the procedure sets. On some systems that means adjusting other machines' setpoints too, and the procedure says which. Once the package shuts off and the discharge check closes, the separator bleeds back to suction and comes down with it.

Adjusting those settings takes the compressor outside its normal operating conditions on purpose. That's a temporary operation. IIAR 7 defines a temporary operating procedure as one that describes the "deliberate and planned operation of a piece of equipment at conditions outside of its normal operating conditions, but within its upper and lower safe limits." OSHA requires operating procedures to cover temporary operations, and adjusting compressor setpoints temporarily is one of IIAR 7's own examples.

The low-suction cutout is why the procedure needs more than steps. It's a safety system, and it's what keeps the machine from pulling the package below the floor. That floor is process safety information, and here it comes from the MDMT on the suction vessels. With the cutout adjusted, the procedure has to explain what the cutout normally protects, what happens if he goes past the limit and what keeps him from getting there. It also has to return the machine to normal. A cutout left at its pumpout setting is a safety system nobody knows is lowered.

With the package down, he closes the isolation valves and runs a hose from a vapor connection on the separator to a bucket of water to bleed off what's left. The hose comes out of the water as soon as the package reaches zero, because ammonia pulls water back in, and it never goes in at all if the package is in a vacuum.

Before he opens the separator, the ammonia sensor over the package goes into test at the detection controller so the residual ammonia doesn't trip an evacuation. IIAR 6-2025 uses this exact job, a coalescer change on a screw compressor, as its example of a nuisance alarm during maintenance.

Then he opens it. The plant's safe work practices, part of the operating procedures element, cover this part: lockout/tagout, opening process equipment and confined space entry if the plant has classified the separator as a confined space. Lockout, confined spaces, respirators and PPE also have their own OSHA standards outside the fourteen elements, and they govern this part of the job as much as anything on the map.

A coalescer change is preventive maintenance on the compressor package, and PM lives in mechanical integrity: written procedures for the work and parts that fit the service.

Once the package is back at pressure, he leak-checks every joint he opened with litmus paper. Plenty of techs reach for a sulfur stick instead, one of the conventional methods IIAR 6-2025 lists, and OSHA's 2024 interpretation letter treats lighting one as hot work. Had he used one, a hot work permit would be part of the job.

The PM procedure is also where this job most often comes up short. The OEM calls for running the machine two hours after the cover goes back on and retorquing while it's hot. When I inspect, I check the manway retainer bolts on the separator, and I've found them loose more times than I can count. I've seen one of the two retainer bolts lying on the ground with the other holding the cover by itself. Make sure the hot retorque is written into the PM procedure, and where the threads allow, add a second locking nut.

Before he can hand the system back over, he has to return the machine to normal. The setpoints and the cutout go back where they were, the sensor comes out of test and the plant's operator takes the machine back.

Two people ran this procedure, and PSM only requires one of them to be trained on it. The operator falls under the training element: trained on the operating procedures before he runs them and refreshed at least every three years. The tech who adjusted the setpoints doesn't. His company has to train him on the work practices and hazards of his job, not on your procedure. Train him on it anyway. An operating procedure is a training tool as much as a compliance document, and he's the one adjusting the cutouts.

At this plant, the coalescer change touches five elements every time: contractors, operating procedures, process safety information, mechanical integrity and training. It doesn't take much to make it seven. If the leak check is a sulfur stick, hot work is on the route. At a plant with no written procedure for the setpoint adjustments, management of change becomes the practical way to evaluate and approve them, and the plant has to stop and run one, at least once for each machine, before the work goes ahead.

That's a lot of program for a few hours of work, and the paperwork that comes with it is real: the JHA or pre-work authorization, the lockout, the line-break permit, a confined space permit if it's a permit space, the PM work order and a hot work permit if a sulfur stick comes out. As a tech, I dodged every piece of paper I could. As a consultant, I know why a line break and a lockout get written down, because those are the steps that hurt people. What keeps him safe is the thinking behind the paper, and at this plant most of that thinking happened once, before he got there. His company was vetted and trained him. The procedure went through MOC once when it was written, so adjusting the cutouts inside its limits isn't a new change. Its limits came from PSI and the operators were trained on it. That leaves him the work itself, and the procedure tells him what to adjust, how far and how to put it back.

At a plant that never did that work, the questions don't go away. They get answered by whoever is holding the wrench, instead of once, at a desk with the PSI open.

That's the difference between a map and a maze. The map is the work you do once. The maze is what every job runs into when nobody did it.

This job touches five elements, sometimes seven, and most plants manage them in different places. The procedure lives in a binder, the contractor records are on ISN or something like it, the PSI is somewhere else and the maintenance history is in SAP or a CMMS. This is where people become the glue that holds the program together. Somebody has to remember where all of it lives and how it connects, and that's how and why mistakes happen. A cutout left at its pumpout setting is one of them. However those connections are managed, by people, software or AI, they all face the same challenge: making sure the next person doing the work sees the whole route, not just the first step in front of him.

If your program already works this way, the walk is short. If it doesn't, pick a job your plant does every month and walk it. Anywhere the walk stops on "we'd need an MOC" or "he knows how," that's work that should have happened once and hasn't.

The sensor he put in test still has to come back out, and knowing it works again is the next walk.

References

• 29 CFR 1910.119. Signing in: (h)(2)(i)–(iv), (h)(3)(i)–(iv). Pulling it down: (f)(1)(i)(C), (f)(1)(ii), (f)(1)(iv), (d)(2)(i)(D), (l)(1). Opening it up: (f)(4), (h)(2)(iv), (h)(3)(iv). Buttoning it up: (j)(2), (j)(6)(iii), (k); OSHA interpretation letter, April 30, 2024. Handing it back: (g)(1)(i), (g)(2), (h)(3)(i)–(ii). The map and the maze: (f)(1)(i)(C), (l)(1).
• ANSI/IIAR 7-2025, §2.2 (temporary operating procedures) and Appendix C 3.b.i.
• ANSI/IIAR 6-2025, A.5.5.3 (leak testing after routine maintenance) and A.5.6.6 (nuisance alarms during maintenance).