What Artificial Demand Is and Why Nobody Sees It

Why this matters

A shop that has done a leak survey, fixed what it found, and reported a number to the owner has usually reported half the problem. The other half has no leaks, no faults, no alarms and no failed parts. It is air being consumed by equipment that is working perfectly, doing work nobody needs done, purely because the header is being held above the pressure the application requires. It has a name, artificial demand, and it is invisible for a structural reason: there is nothing in the building you can point at and say "that is it."

The reason to learn it is commercial as much as technical. Artificial demand is the term that explains why a plant's air consumption is higher than the sum of everything it can account for, and putting a number on it is something almost nobody has ever shown that customer.

The definition, and the word that carries it

Artificial demand is the additional mass of air consumed by the plant's own equipment because header pressure is higher than the application needs. Not because anything is broken. Two mechanisms produce it, and both track header pressure directly:

• An open path to atmosphere at a working device. A blow-off nozzle, an air motor exhaust, an open pick-and-place jet, a venturi. Any opening venting to atmosphere from above roughly 13 psig is choked, meaning air leaves the throat at the local speed of sound and the mass flowing out is proportional to absolute upstream pressure. Feed the same nozzle at a higher header and it passes proportionally more air while doing exactly the same job.
• A volume filled and then dumped. A cylinder, an air-over-oil circuit, a bag or a pad fills to whatever pressure it is fed and exhausts that whole charge each cycle. The mass in a fixed volume is proportional to absolute pressure, so a cylinder fed at 105 psig when its load is overcome at 70 psig dumps roughly half again the mass on every stroke, forever, and the stroke is no better for it.

The word doing the work in that definition is needs. Not "is set to", not "is rated for". The application's need is set by the load it has to overcome and the cycle time it has to hit, and it is almost always below what the plant is feeding it.

Four things artificial demand is not

It is not a leak. The physics of a choked opening is identical, and that is exactly what makes the two get confused, but the accounting is completely different. A leak flows whenever the system is pressurised, including nights, weekends and shutdowns. Artificial demand flows only when the application is running. And critically, a leak is closed by repairing a fitting; artificial demand is closed by setting a pressure. Fixing one does nothing for the other.

It is not production demand. Production demand does work: it moves the load, it makes the part, it hits the cycle time. Artificial demand is the increment above that, and it buys nothing. A cylinder whose load is overcome at 70 psig moves no faster on 105 psig once the load is off the seat; the extra mass is spent accelerating a piston that was already fast enough and then vented.

It is not a measurement artefact. This matters because the first reaction to a flow balance that will not close is to distrust the instrument. Artificial demand is real mass leaving the building.

It is not something a bigger compressor fixes. A larger machine holds the header up more successfully against the same demand, which raises average pressure, which increases artificial demand. That is the one intervention on the list that makes it worse.

And it is not what a leak survey finds

This is the exclusion worth sitting with, because it is the one that leaves money on the table in shops that have already done the work.

An ultrasonic leak survey is walked with production shut down, which is exactly the condition that makes leaks audible and everything else silent. Every device generating artificial demand is off during that survey. So a survey that finds fourteen leaks totalling a quarter of compressor capacity has honestly reported the leak term and structurally cannot have seen the other one. A shop that reports the survey number as the whole opportunity, and then reduces the header and finds consumption fell further than the leak physics can explain, has just learned this the expensive way, in front of the customer.

Why nothing in the plant points at it

There is no fault code. No regulator alarms because it is set high; a regulator set wide open is doing precisely what it was told. No machine runs badly. No part fails. The maintenance system has nothing to log.

The only place it appears is in the compressor's loaded hours, and loaded hours are a number almost nobody reads, on a controller almost nobody opens, in a room almost nobody enters except to change a filter.

Measuring it: the pressure-step method

You cannot measure artificial demand directly, because you cannot instrument a thing that has no device. You measure it as a residual: change the header pressure, watch how much total consumption changes, subtract the part the leak physics already predicts, and whatever is left is artificial demand.

That needs three inputs, and the first two come from the duty-cycle method covered in the leak-quantification article: total output as a percentage of compressor capacity during a representative production period, and leak load as a percentage of capacity measured at zero production. The third is the machine's own capacity at each of the two discharge pressures, read from the manufacturer's capacity data, because a positive-displacement compressor delivers modestly more flow at a lower discharge and the two measurements are otherwise being expressed against two different denominators.

Before stepping the pressure down, three exclusions, each of which is a hazard the test itself creates:

• Never reduce supply pressure to a circuit that holds a clamp, a brake, a counterbalance, a fixture or a suspended axis without the machine builder's stated minimum pressure in hand and the load blocked or lowered to a mechanical rest first. Lowering header pressure releases holding force everywhere at once, and the machine gives no warning that it was relying on it.
• Exclude any pneumatic safety interlock, damper or protective function from the test, and confirm with the equipment's documentation which circuits those are before you touch a setpoint.
• Changing a compressor setpoint means working inside an enclosure with energised conductors. De-energise the starter, lock and tag it, and prove the circuit dead with a meter before working inside, per 29 CFR 1910.333(b)(2), which is the electrical standard here; 1910.147 expressly excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C). On a construction site the electrical counterpart is 29 CFR 1926.417.

Worked example: what the residual actually came to

A plant runs its header at 105 psig. Measurements, all by the duty-cycle method:

• Total output during a representative production hour: 78 percent of capacity at 105 psig
• Leak load at zero production, same pressure band: 25 percent of capacity at 105 psig, which sits inside the fifth-to-a-third range typical of an unmanaged system

The header is stepped down to 95 psig, held for a full week so the plant runs its normal mix of work, and total output is re-measured during a comparable production hour: 68 percent of capacity at 95 psig.

The naive reading is a 10-point fall. That reading is wrong, and it is wrong in the flattering direction, because the two percentages are against two different capacities. The manufacturer's capacity data for this machine gives about 3 percent more flow at 95 psig than at 105 psig. Putting both figures on the 105 psig capacity basis: 68 percent of a capacity 3 percent larger is 70 percent of the original basis. So the honest reduction is 8.0 points, not 10.

Now subtract the part that is not artificial demand. Leaks are choked, so their mass flow scales with absolute pressure. Going from 105 psig to 95 psig takes the absolute from 119.7 psia to 109.7 psia, a ratio of 0.916, so a 25-point leak load becomes 22.9 points. That is a predicted fall of 2.1 points, and it is a computed systematic term, not a measurement uncertainty.

Residual: 8.0 minus 2.1, which is 5.9 points of capacity. Against the original 78 points of total output, that is 7.6 percent of everything the plant was consuming, and over the week at 95 psig the shop's cycle times and reject rate were unchanged, so none of those 5.9 points was doing work.

Two things this example does not prove, and saying so is part of the method. The two production hours are samples, so if the week's work mix differed from the baseline week the comparison is contaminated, and the fix is to repeat it rather than to argue about it. And the residual is the total of every over-fed device in the plant, not a location; finding where it sits means walking the machines with the manufacturer's minimum pressure for each and comparing it against what the regulator is actually set to.

What flips the recommendation

The whole method assumes the header pressure can come down. Sometimes it cannot, and the tell is specific: one machine, usually one, whose documented minimum pressure genuinely sits near the current header. In that case lowering the plant to serve the rest of the shop and giving that one machine a dedicated arrangement, whether a point-of-use booster or its own supply, is the move; holding the entire plant at one machine's pressure is how a header ends up where it is in the first place.

The other flip is a plant whose measured leak load is already very low. Where leaks are a small term, the residual method's subtraction is small too, and nearly the whole observed reduction is artificial demand. That is a good outcome, not a suspicious one, and it means the opportunity is at the regulators rather than at the fittings.

References

• 29 CFR 1910.333(b)(2), the electrical standard governing work inside an energised enclosure such as a compressor control panel, with 1910.147(a)(1)(ii)(C) as the carve-out that sends this work there; 29 CFR 1926.417 is the construction counterpart
• Compressor manufacturer's capacity-versus-discharge-pressure data, needed to put two duty-cycle measurements on the same denominator
• Machine builder's documented minimum operating pressure for each pneumatic circuit, particularly any circuit providing holding or clamping force
• See related: How to Put a Number on What Leaks Are Costing; Why Raising System Pressure to Fix a Symptom Costs More Than the Symptom