The Demand Events That Size a System Nobody Measured

Why this matters

A compressed air system is almost never sized from the plant's actual load profile. It is sized from a nameplate total, a previous compressor's rating, or a supplier's guess, and all three of those describe an average. The events that actually decide whether the system works are short, they repeat, they overlap, and none of them appear in any record the plant keeps. So the shop that walks in with a written register of those events holds the one document nobody else on the job has, and it turns "you need more air" into three separate and much cheaper findings. This card is that register.

Building the register means standing next to machines while they cycle. Stay outside the motion envelope of anything that indexes, clamps or ejects, and run a sensing line out to where you are standing rather than defeating a guard or a light curtain to get closer. Where a temporary gauge or test port has to go into a live line, isolate that section upstream, lock and tag the isolation and the compressor disconnect under 29 CFR 1910.147, vent downstream and confirm zero on a gauge open to the line before you cut or break anything.

The three sizing drivers, which are three different numbers

The reason one flow figure cannot size a system is that three different components are sized by three different statistics of the same load.

• The compressor is sized by demand averaged over the repeat interval of the events. Not the shift average, and not the peak. Over one full cycle of a repeating event, the air the compressor puts in has to equal the air the plant takes out, or the pressure walks down cycle by cycle.
• Storage is sized by the air deficit during a single event, which is the event's flow above what the compressor delivers, multiplied by the event's duration. The storage article in this library owns the conversion from that quantity of air to a receiver size and a pressure drop, and you should use its arithmetic rather than re-deriving it.
• Distribution is sized by the coincident peak, the sum of everything that can physically fire in the same second, because pipe does not store anything worth counting and has to pass whatever flows through it at that instant.

Get these confused and you buy the wrong thing. A plant sized on the coincident peak buys a compressor several times larger than it needs and then runs it at part load, which the part-load article in this library shows is where the energy goes.

The three arithmetic operations behind those drivers, stated here so the filled-in register below introduces nothing new:

• Duty fraction equals event duration divided by the interval between events.
• Average adder equals the event's peak flow above baseline, multiplied by its duty fraction. Where the event only occurs during part of the shift, multiply again by that fraction of the shift.
• Coincident peak equals baseline demand plus the sum of the peak adders of every event in the same coincidence group.

The register: nine fields, and why each one exists

1. Event name and the machine it belongs to. Named by what the operator calls it, not by an equipment tag, because the operator is who you will ask about it next year.

2. What initiates it. Operator button, PLC sequence, timer, interlock. This field is what tells you whether the event's frequency is fixed or can change with production, and a timer-driven event is the one that will still be firing on a Sunday.

3. Peak flow above baseline, and the basis of that number. The basis is the field, not the number. See the section below.

4. Duration. Seconds. Timed, not estimated. Human estimates of a two-second event are reliably wrong by a factor of several.

5. Interval, and the worst-case burst interval. The normal interval sizes the compressor. The worst-case burst, which is what happens when an operator holds the button or a line runs a short part, sizes the storage.

6. Coincidence group. Which other events on this register can fire in the same second. Two events on the same interlock never coincide and should be recorded as such, because that is the one field that can shrink your distribution requirement rather than grow it.

7. Minimum acceptable inlet pressure at the machine, and where that number came from. The machine's own documentation, not the header setting. A machine whose real floor is 65 psig running on a plant held at 100 psig has 35 psi of headroom nobody knew about.

8. What fails when pressure sags during this event, and how the failure presents. Slow cycle, incomplete clamp, a part that scraps, a fault code, a quality defect discovered downstream. This field is what turns the register from an engineering document into a production one.

9. Where it was measured, by whom, on what date. A register with no provenance gets argued with, and a two-year-old event flow that predates a line change is worse than no entry.

The basis field is the one that decides whether the register is worth anything

Three ways to get a peak flow, in descending order of trust, and every entry says which one it used.

Measured from the receiver's pressure decay. Convert the decay slope during the event into a flow using the vessel's standard cubic feet per psi. This gives you the net deficit, the amount by which demand exceeded the compressor's output, not the event's total draw, and an entry that records it as total draw is the most common corruption of this register.

Measured with a flow meter on the branch. Best when the branch serves only the event. Record the meter's basis and whether its stated accuracy is a percent of reading or a percent of full scale, because a percent of full scale figure does not shrink when the reading does, and on a short event you are usually reading the bottom of the meter's range.

Taken from equipment documentation. Acceptable for a field you cannot isolate, and it needs a note saying so. Published consumption figures are usually stated at a reference pressure and at a duty the machine may not run, so a documented figure applied at a different pressure is an approximation and the entry should say the word.

A filled-in register: one shop, three events

A wood products shop, one 8-hour shift. Baseline demand with nothing cycling, which includes the leak population and the small tools, measures 30 scfm.

Now run the three operations.

Duty fractions. Event A: 20 divided by 40 is 50 percent while the heavy part runs, and that run is 2 of the 8 shift hours, so 12.5 percent of the shift. Event B: 8 seconds six times an hour is 48 seconds out of 3,600, which is 1.33 percent. Event C: 0.5 divided by 10 is 5 percent, all shift.

Average adders. A: 60 times 0.125 is 7.5 scfm. B: 25 times 0.0133 is 0.33 scfm. C: 40 times 0.05 is 2 scfm. Shift average demand is 30 plus 7.5 plus 0.33 plus 2, which is 39.8, call it 40 scfm.

Coincident peak. All three are in one coincidence group, so 30 plus 60 plus 40 plus 25 is 155 scfm. That is nearly four times the shift average of 40 scfm, and it is the number the distribution has to pass.

The compressor's real requirement. Not 40 and not 155. During the heavy part run, average demand over one full 40-second cycle of Event A is: baseline 30, plus A's average across its own interval of 30 scfm, plus C's 2 and B's 0.33. That totals 62.3 scfm. A compressor delivering 62 scfm holds the header through the run. Check it: during A's 20-second draw the plant pulls 30 plus 60 plus 2 plus 0.33, which is 92.3 scfm, against 62.3 supplied, a deficit of 30 scfm for 20 seconds, which is 10 standard cubic feet out of storage. During the 20-second recovery the plant pulls 32.3 scfm against 62.3 supplied, a surplus of 30 scfm for 20 seconds, which is 10 standard cubic feet back in. It balances, which is what "sized on the interval average" means.

Storage's real requirement. 10 standard cubic feet against Event A on its own. But field 6 puts B and C in the same coincidence group, so the sizing case is A with B inside it: add 25 scfm for 8 seconds, which is 3.3, and C's 0.3, for about 13.6. Size against that, and against the worst-case burst rather than the normal interval, which is what field 5 is for. Run that through the storage article's conversion against whatever pressure band the 90 psig clamp minimum leaves you and you get a modest vessel, sited near the clamp table rather than in the compressor room.

So this plant's three findings are: a compressor at roughly 62 scfm rather than the 155 someone would have quoted from the peak, a small local receiver at the clamp table, and distribution capable of 155 scfm at acceptable loss. The one number that appears nowhere in that answer is the shift average of 40 scfm, which is the only number the plant had before you arrived.

What would change this: break the coincidence. If the collector's pulse timer can be interlocked to skip a pulse while the clamp bank is drawing, the coincident peak drops from 155 to 115 scfm and the distribution requirement drops with it, at the cost of a slightly dirtier filter. That trade is only visible because field 6 exists.

The failure mode of skipping the register: the plant runs the clamp table on a header held at 100 psig, sags to 88 during the event, scraps parts at glue-up, and the maintenance log records "low air pressure" with no flow, no duration and no coincidence. Two years later a larger compressor gets installed, the coincident peak still exceeds what the pipe can pass, the sag persists, and the plant now runs an oversized machine at part load. Every part of that outcome was decided by information that was never written down.

Where the register lives and when it goes stale

Keep it with the system drawing, not in a maintenance work order, and put a review date on it. Three things invalidate an entry and each should trigger a re-measure: a change to the machine or its tooling, a change to the production rate that alters an interval, and any change to the header pressure setting, because a documented consumption figure taken at one pressure does not hold at another.

Read what Event B is before measuring it. A bench station blowing off parts is compressed air used for cleaning, so 29 CFR 1910.242(b) applies: dead-end static pressure under 30 psi, effective chip guarding, personal protective equipment. A 60 psig nozzle minimum is a finding for the register, not a demand to accept. Timing at a blow-off station also puts you next to airborne dust. Stand upwind of the discharge and out of the plume; where the material being blown is respirable, wood dust and metal fines included, do the timing from outside the cloud or wear respiratory protection selected under a program meeting 29 CFR 1910.134, because eye protection and gloves do nothing for an inhalation route. Where the dust is concrete, stone or masonry bearing, that is respirable crystalline silica and the exposure is governed by 29 CFR 1910.1053 in general industry, which means the blow-off itself is the wrong tool and the entry should say so.

References

• 29 CFR 1910.147, control of hazardous energy, general industry: the basis for the isolate, lock, tag and verified-zero sequence before installing a temporary gauge or test port in a pressurised line.
• 29 CFR 1910.134, respiratory protection, general industry: the basis for the respiratory control required when timing events in an airborne dust plume.
• 29 CFR 1910.1053, respirable crystalline silica, general industry: the basis for treating blow-off of concrete, stone or masonry dust as a regulated exposure rather than a housekeeping choice.
• See related: What Storage Buys You That Horsepower Cannot; Why Part Load Behaviour Decides the Energy Bill; What to Record About a Compressed Air System So It Can Be Compared Later.