The Compressor Types and What Each One Is Suited To

Why this matters

Almost every bad air-room decision starts the same way: somebody added up the tools on the wall, got a peak number, and bought a machine that could deliver it. The machine works on day one and then either wears out early or quietly costs more every hour for fifteen years, and nobody connects the two, because the sizing question everyone asked was answered correctly.

Peak flow picks the size. The duty cycle and the shape of the part-load demand pick the family, and getting the family wrong is the expensive error, because it cannot be corrected without replacing the machine.

The two numbers that pick the family

Duty cycle is the fraction of running time the machine is actually compressing. A machine that compresses for 20 minutes out of every hour is at roughly 33 percent duty. Some families tolerate 100 percent duty indefinitely and some are built with a duty limit that the manufacturer publishes, and exceeding it is a heat and lubrication problem, not a warranty technicality.

Part-load shape is what the demand does between peaks. Two plants with the same peak can have completely different profiles: one holds a steady base load with occasional surges, the other sits near zero most of the shift and spikes hard. The first suits a machine that is efficient at a steady output near its rating. The second suits a machine that costs nothing when it has nothing to do, plus storage.

Storage is the third lever and it changes which family you need, because it converts a short spike into an average. The relation is developed in the receiver-sizing article: the stored volume in cubic feet needed to cover a demand deficit is the deficit in scfm, times the event duration in minutes, times 14.7, divided by the pressure drop you will allow in psi. Every gallon is 0.134 cubic feet.

The families

Reciprocating. A piston in a cylinder. Simple, tolerant of dirt, cheap to repair in the field, and the only common family whose part-load behaviour is genuinely proportional, because start/stop control means a machine with nothing to do draws nothing. That is its real advantage and it is usually described as a disadvantage. The cost is a duty limit: splash-lubricated single-stage units are commonly published in the range of half to three-quarters duty, while pressure-lubricated two-stage industrial units are built for continuous running. Take the duty rating off the machine's own documentation rather than a rule of thumb, because it varies more across this family than any other. Recips also make more noise and more vibration, and they deliver air with more pulsation, which is one of several reasons they are always paired with a receiver.

Rotary screw, oil-injected. Two meshing rotors with oil injected into the compression chamber to seal, lubricate and absorb heat. Continuous duty, quiet, compact, and the default for anything running most of a shift. The weakness is part load: a load/unload machine with nothing to do keeps spinning and keeps drawing, commonly a quarter to a third of full-load input depending on the control scheme and the machine, so an oversized screw bleeds continuously. Before variable speed, check which part-load scheme an existing machine is actually on, because inlet modulation without unloading is the worst behaviour in the family: published curves start well above two thirds of full-load power at zero output, and nobody in the plant usually knows it is fitted. Variable speed drive fixes most of that across the middle of the range, and it is bounded at both ends: below its minimum speed the machine goes back to cycling, and at full load it gives nothing back, because the drive's own conversion loss makes it slightly worse than fixed speed there. Its advantage lives entirely in the middle, which is why the demand profile rather than the peak decides whether it pays.

Rotary screw, oil-free. Same rotor geometry with no oil in the compression chamber, so it needs staging and intercooling to survive the heat the oil was absorbing, and it is normally two-stage above small sizes. It delivers air with no compressor-derived oil, which matters where the process cannot tolerate any. It does not deliver dry air, and it does not remove the oil already in the ambient air the intake is breathing.

Rotary vane. Sliding vanes in an eccentric rotor, oil-injected, continuous duty, mechanically simple, low speed. A steady, unglamorous fit for a constant base load in the small to mid range.

Scroll. Two interleaved spirals, oil-free at the compression element, quiet, and small. Usually installed as a bank of modules so capacity steps up and down in increments, which gives a bank good part-load behaviour by turning modules off rather than throttling one machine.

Centrifugal. A dynamic machine: it accelerates air with an impeller and converts velocity to pressure. Efficient at large flows and near-constant output, oil-free at the air path, and it has a behaviour no positive-displacement machine has. Below a certain flow it surges, so its turndown is limited and the standard protection is to blow excess air off to atmosphere, which costs full input for zero delivered air. A centrifugal wants a large steady base load underneath it and something else handling the swing.

Comparison

When to pick which

• Demand is intermittent and the plant is small. Storage plus a reciprocating machine sized near the average, not the peak. The recip's duty limit is met because it stops, and it costs nothing while stopped. Size the storage against the motor's published maximum starts per hour, checked at the worst case of about half of capacity where cycling is fastest rather than at peak; the sequencing article carries the cycle-time relation.
• Demand runs most of the shift at a fairly steady level. Oil-injected screw sized close to that level, with storage covering the surges.
• Demand swings widely and continuously. Variable speed screw, sized so the normal operating band sits above its minimum speed, or two machines with one trimming.
• The process cannot accept compressor oil. An oil-free machine, plus the filtration and drying the process actually specifies. Oil-free is a statement about the compression element, not about air purity, and the purity classes belong in the purchase specification under ISO 8573-1:2010 in the edition the contract names.
• Flow is large and steady. Centrifugal for the base, something with real turndown for the swing.

Worked example: the shop that bought the peak

A three-bay service shop logs its demand for a week. Peak is 45 scfm, in bursts of about 30 seconds when two impact wrenches and a blow gun overlap. Averaged across the 8-hour day, demand is 11 scfm.

The shop was quoted a 45 scfm rotary screw on load/unload control. Run its duty: 11 / 45 = 24 percent loaded. Over 8 hours that is 1.92 hours loaded and 6.08 hours unloaded. Using the machine's own measured unloaded input of 0.30 of full-load input, and normalising full-load input to 1.0:

1.92 x 1.0 + 6.08 x 0.30 = 1.92 + 1.82 = 3.74 full-load-equivalent hours per day.

Now size to the average instead. A 20 scfm machine runs 11 / 20 = 55 percent duty, which is 4.4 hours loaded per day. At equal specific power its full-load input is 20 / 45 = 0.444 of the larger machine's, so in the larger machine's units that is 4.4 x 0.444 = 1.96 full-load-equivalent hours per day.

Same air delivered. The oversized machine costs about 1.9 times the smaller one. The condition on that ratio is equal specific power between the two machines, which favours the larger one slightly in reality because bigger airends are usually a little more efficient, so treat 1.9 as the upper end and the direction as certain.

The 20 scfm machine cannot serve the 45 scfm burst on its own, so check the storage. Deficit is 45 - 20 = 25 scfm, event is 30 seconds which is 0.5 minutes, and the shop can accept a 20 psi swing between the pressure the header actually sits at when the burst starts and the tools' minimum inlet pressure plus the measured distribution loss to that drop at the burst's flow. Not between the unload setpoint and the tool's rated pressure, which is how this band gets overstated and the vessel comes out light:

0.5 x 25 x 14.7 / 20 = 9.2 cubic feet, which is about 69 gallons.

An 80-gallon receiver covers it with margin, and it is the cheapest component in the room.

The failure mode of getting this wrong. The shop that buys the 45 scfm screw never sees a symptom. Air is always there, tools work, nothing alarms. The 1.9 multiple runs quietly for the life of the machine, and because there is no fault to diagnose, no service call ever surfaces it. That is why this is a walk-in finding rather than a callback: the only way it gets found is somebody logging duty.

What flips the recommendation. If the shop adds a process that runs continuously at 30 scfm, the average moves close to the peak, the profile flattens, and the 45 scfm screw becomes the right machine at good duty. Family choice follows the profile, so re-run it whenever the plant's work changes.

How to verify you got this right

Log load and unload hours off the controller for a full week, not a day, and include at least one non-production period. Compute duty from the loaded hours. Then compute the same duty from a stopwatch on the machine over one representative hour and compare: if the controller and the stopwatch disagree by more than the counter's one-hour resolution spread across the week, you are reading a counter that includes something you did not think it did.

Standing next to a running compressor to time it is where people get hurt. Compressor rooms commonly reach the 85 dBA eight-hour time-weighted average action level that triggers the hearing conservation requirements of 29 CFR 1910.95, so wear the protection the site's programme specifies before you open the door. Keep clear of belt guards and coupling guards and never time a machine with a guard removed. Discharge piping and the aftercooler burn on contact, so take temperatures with a non-contact instrument and keep leads clear of hot pipe and rotating parts. Nothing in this procedure requires opening a pressurised component; if a reading does, stop the machine, isolate and lock its energy under 29 CFR 1910.147, vent the sump and receiver, and confirm zero on a gauge before any fastener moves.

References

• 29 CFR 1910.95, occupational noise exposure, including the 85 dBA eight-hour time-weighted average action level
• 29 CFR 1910.147, control of hazardous energy, for isolation and stored-energy release before opening a package
• ISO 8573-1:2010, compressed air purity classes, which binds through the purchase specification or contract rather than by law
• CAGI performance data sheets and ISO 1217 acceptance-test conditions in the edition the manufacturer's data sheet names, which bind through the CAGI data-sheet programme the manufacturer participates in or through a purchase specification that calls the standard out, not by law, for comparing published capacity and specific power between machines
• See related: How to Work Out Whether a Receiver Is Sized for the Demand; What a Compressed Air System Actually Costs to Run