What a Compressed Air System Actually Costs to Run

Why this matters

Every other utility in a building arrives with a meter on it. Compressed air does not. It arrives on a wire, gets converted into pressure by a machine in a room nobody visits, and then gets used by people who have never been told it costs anything. That is why a plant will argue for twenty minutes about a light fixture and leave a blow gun hissing into an empty bench all weekend. A shop that can walk into that room and put a defensible number on what the air is costing is selling a conversation the customer has never had.

The number itself is not hard. What makes it wrong, nearly every time, is the boundary somebody drew around it.

The boundary is the air room, not the compressor motor

The run cost of compressed air is the electrical input to everything in the air room, over the period you are measuring. Not the compressor's nameplate horsepower, which is a motor rating and not a draw. Not the shaft power, which is what the motor delivers to the airend and excludes motor loss. The input.

Inside the boundary, at minimum: the compressor package motor, its cooling fans, the dryer, any separate aftercooler fan or cooling-water pump serving the air room, and the condensate management gear. If a device only exists because you make compressed air, its input is part of the cost of compressed air.

Outside the boundary: the air itself, which is free, and the purchase price of the equipment, which is a capital question and belongs in a different conversation. Mixing capital into a run-cost number is the fastest way to lose an owner who already knows what the compressor cost.

The negative space: three terms a mental estimate leaves out

Ask an owner what their air costs and you will get a figure built from the compressor's rated horsepower times the hours the plant is open. That estimate is not slightly low. It is structurally low, because it omits three terms, and all three run in the same direction.

It omits unloaded running. A load/unload rotary screw that has nothing to do does not stop. It spins, vents its sump, and keeps drawing. Manufacturers commonly publish unloaded input somewhere in the range of a quarter to a third of full-load input for oil-injected screws, but the spread across control schemes and machine sizes is wide enough that you measure it rather than assume it. Take the figure from the package's own performance data sheet or read it at the machine.

It omits the auxiliaries. A non-cycling refrigerated dryer runs whether or not air is flowing through it. Its input is small next to the compressor and it runs a great many more hours, which is exactly the shape of a term that hides.

It omits the leak load. Leakage is not a maintenance annoyance that shows up as a hiss. It is a demand, indistinguishable at the compressor from a tool, and it runs during every hour the system is pressurised including the ones nobody is in the building. The US Department of Energy's compressed air materials put leakage in a plant that has never managed it at a substantial fraction of total output, commonly cited around a fifth to a third. Rating and fixing individual leaks is its own job; for a cost number you only need the aggregate.

The unit that survives: full-load-equivalent hours

Currency figures go stale and vary by utility. The unit that travels is the full-load-equivalent hour: one hour of the package running at its full-load input. Normalise full-load input to 1.0, express every other operating state as a fraction of it, multiply by hours, and add. The result is a count you can hand to anyone, compare year over year, and convert to money once at the end using the customer's own rate.

It also forces the honest question, because a full-load-equivalent hour that produced nothing is visible in the arithmetic instead of buried in an average.

Getting the input reading without opening a panel

Most packages built in the last couple of decades will show input power, or at least load state and hours, on their own controller. Read it there first. Load and unload hour counters plus a single full-load input figure from the data sheet will get you a defensible number.

If you need a direct reading, a permanently installed meter on the compressor feeder is the answer, and installing one is licensed electrical work. Clamping a meter onto energized conductors inside a panel is energized work: it is performed only by a qualified person wearing the arc-rated protection the employer's electrical safety programme assigns for that task under NFPA 70E-2021 as the employer has adopted it, with the work permitted under 29 CFR 1910.333(b)(2) rather than 1910.147, because the hazard is electrical rather than stored mechanical energy.

Error basis before the number goes anywhere. An hour counter has a fixed-count resolution, typically one hour, which is a bound rather than a spread and stays the same size no matter how long the window is, so it shrinks as a percentage over a longer window. A clamp reading carries a percent-of-reading accuracy from its own data sheet, which shrinks with the reading. Do not mix the two into one plus-or-minus. Carry the counter error as a bound and say so.

Worked example: one week on one package

A single oil-injected rotary screw serves a fabrication shop. Metered over 7 days, which is 168 hours:

• 96 hours loaded
• 54 hours unloaded
• 18 hours off

Unloaded input, read at the machine, is 0.30 of full-load input. Normalising full-load input to 1.0:

• Loaded: 96 x 1.0 = 96.0 full-load-equivalent hours
• Unloaded: 54 x 0.30 = 16.2 full-load-equivalent hours
• Compressor total: 112.2 full-load-equivalent hours

So 16.2 of the compressor's 112.2 full-load-equivalent hours for the week, 14.4 percent of the compressor's own input, went into a machine spinning with its inlet closed.

Now the leak load. During a 60-minute window on Sunday with production down and every tool valved off at the drop, the compressor loaded for 12 minutes of the 60. That 20 percent duty cycle is the fraction of full output the system was consuming with nobody using it. It is an upper bound on leakage, not a leak rate, because anything left switched on that you did not find is counted in it, and the timing resolution of a stopwatch on 12 minutes is a bound of a few percent of reading on top.

Apply that bound to the PRESSURISED hours rather than the loaded ones, because a leak flows during every hour the system holds pressure including the ones nobody is in the building: 0.20 x 150.0 = 30.0 full-load-equivalent hours, 26.7 percent of the compressor's 112.2. Applying it to loaded hours alone is the common mistake and it understates the leak term by a third.

Add the dryer. A non-cycling refrigerated unit measured at 0.06 of the compressor's full-load input, running all 168 hours, is 10.1 full-load-equivalent hours. The air room total is 122.3 full-load-equivalent hours.

The two non-productive terms sum to 16.2 + 30.0 = 46.2 full-load-equivalent hours. Against the compressor's own 112.2 that is 41.2 percent. Against the air room's 122.3 it is 37.8 percent. Both are true and they answer different questions, which is why the base goes in the sentence every time.

What this does not license you to conclude. It does not say the shop can cut 29 percent of its air-room input. The unloaded hours are a control-strategy question and some of them are unavoidable. The leak figure is a bound that includes unattended demand. And the dryer's 10.1 hours are not waste at all; they are the price of not putting water into the tools. What the number does license is the ranking: the largest single controllable term here is the leak load, and it is larger than the entire dryer.

The failure mode. A shop that skips the Sunday window and reports only the loaded hours will conclude the compressor is well matched to the plant, because a 57 percent load duty over a working week looks healthy. The leak term alone is three times the dryer's, and it is running in hours the load duty never shows. The average hides it; only the no-production window exposes it.

How to verify you got this right

Run the arithmetic backwards. Take your full-load-equivalent hours, multiply by the package's rated full-load input, and compare against the utility's demand and consumption for the same week if the air room is separately fed or if the plant's baseline changes visibly when the compressor is down. If your figure exceeds the whole plant's off-shift baseline, the boundary is wrong or the unloaded fraction is.

Then re-run the no-production window on a different day. A leak population does not change materially in a week. If the two duty cycles differ by more than the timing resolution, something was left on during one of them, and the higher reading is the one with the unfound tool in it.

Before you do any of this at the machine. The air room is a hot, loud, stored-energy environment. Where sound levels put an employee at or above the 85 dBA eight-hour time-weighted average action level, hearing protection and the rest of the programme required by 29 CFR 1910.95 apply, and a compressor room commonly gets there. Discharge piping and the aftercooler run hot enough to burn on contact, so route your hands and your leads clear of them and let a shutdown package cool before touching bare pipe. Do not open, drain or disconnect anything to take a reading: every measurement described here is taken from a display, a counter, or outside a closed enclosure.

And if the conversation turns to the blow guns you saw on the way in, the limit is not a preference. Compressed air used for cleaning is restricted under 29 CFR 1910.242(b) to a reduced pressure of less than 30 psi and then only with effective chip guarding and personal protective equipment, because air driven into a break in the skin can enter the bloodstream.

References

• 29 CFR 1910.242(b), compressed air used for cleaning
• 29 CFR 1910.95, occupational noise exposure, including the 85 dBA eight-hour time-weighted average action level
• 29 CFR 1910.333(b)(2), electrical safety-related work practices for energized work, and NFPA 70E-2021 in the edition and form the employer has adopted in its electrical safety programme
• US Department of Energy, Improving Compressed Air System Performance: A Sourcebook for Industry, for leakage as a share of plant output
• See related: Why Most of the Energy Leaves as Heat; What a Compressor Is Actually Doing to the Air