Compressed Air and Phantom Loads: Find Off-Hours Factory Waste

The cleanest energy measurement you will ever take is the one where nothing is being produced.

During a shift, draw is confounded by product mix, run rate and operator behaviour. Overnight and at weekends, whatever is still consuming power is consuming it for a reason you can isolate: a compressor cycling against a leak, heaters holding temperature nobody needs, drives energised on idle machines, or equipment somebody never switched off. This guide decomposes off-hours draw into five categories, uses compressed air as the worked case, ranks investigations by annualised cost, and verifies the fix against the same window.

Off-hours energy at a glance

Why off-hours is the cleanest window

Because production stops confounding the measurement.

During a shift, a machine's draw is the sum of its base load, its process load, the operator's behaviour, the product being run and the rate it is being run at. Attributing a change to any one of those requires controlling for the rest, which is why in-production energy analysis needs the mix normalisation described in energy by shift and product.

Overnight and at weekends, all of that disappears. Whatever is still drawing power is drawing it for a reason unrelated to production, and that makes the signal unusually interpretable. A step change at 18:15 every evening is a thing that gets switched on. A sawtooth continuing all night is a compressor cycling. A flat elevated line is something held at temperature.

Guidewheel's machine-level energy monitoring guide already covers finding phantom loads as one part of a broader energy method. This article takes the off-hours window as the subject in its own right, because the diagnostic discipline it supports is more specific than a general energy review, and because compressed air rewards it more than anything else in a factory.

Practical setup. Take a full non-production window, ideally a weekend from end of Friday production to Monday start-up. Change nothing during it, which is the hardest part: the instinct on seeing a large overnight draw is to switch something off immediately, and doing so destroys the baseline you need in order to prove the saving later.

Record the whole window at machine level. Site-level data will tell you the plant consumes something overnight and will not tell you what, which is the difference between an observation and an action.

With a clean window recorded, the draw separates into five categories.

Decomposing the draw

Five categories, each with a distinguishable signature.

True baseline. Flat, low, unchanging across the entire window. Safety systems, emergency lighting, network equipment, control cabinets, fire and security systems. This is the floor, and it is legitimate. Establish it before doing anything else, because everything above it is what you are investigating and treating baseline as waste wastes the investigation.

Idle equipment. Step changes that correlate with nothing. A drive left energised, a conveyor running empty, a pump nobody switched off, an extraction system left on. The signature is a discrete step up at end of production that persists, then a step down when someone eventually notices on Monday. This category is usually the easiest win in the entire plant and requires no capital, only a shutdown checklist that someone actually performs.

Leaks. A repeating cycle with no corresponding demand. On compressed air this is unmistakable: the compressor loads, builds pressure, unloads, pressure decays, and it loads again on a regular period, all night, with no tool or actuator using air. The period between load events is itself informative, because a shorter period means a larger leak.

Warm-up and thermal holding. Sustained elevated draw on ovens, furnaces, barrels, hot runners and heated tanks. Sometimes required, because bringing a thermal asset up from cold takes hours and costs more than holding it. Frequently not required, because the asset is being held through a weekend for a Monday start that does not need it. The signature is a flat elevated line with periodic small cycling as the controller maintains temperature.

Off-shift activity. Irregular, human-shaped bursts. Cleaning crews, maintenance work, a shift that ran unplanned overtime, someone testing a machine. Confirm before classifying as waste, because this category is frequently legitimate and misreading it damages the credibility of the whole exercise.

Of the five, one repays attention out of proportion to the others.

Compressed air specifically

Compressed air dominates off-hours waste in most factories for a structural reason: a leak consumes continuously, and nothing about a leak announces itself.

A mechanical fault stops production and gets attention within minutes. A leaking fitting produces no fault, no alarm and no quality problem. It simply makes the compressor work slightly harder, every hour of every day, including the 128 hours a week when nothing is being produced. Compressed air is also expensive to generate relative to the work it does, so a modest volumetric loss represents a disproportionate energy cost.

What the off-hours signature tells you. Watch the compressor's load and unload cycling across the window with all production stopped. A system with no leaks will build pressure and stay there, with the compressor unloaded for long periods. A system with leaks cycles on a regular period, and the shorter that period, the greater the loss. Trend the period week to week and you have a leak-rate indicator that requires no additional instrumentation.

Where to look once the data justifies it, in rough order of frequency: quick-disconnect couplings, threaded fittings and joints, pneumatic cylinder seals, filter-regulator-lubricator units, condensate drains stuck open, and hoses on equipment that has been moved. Isolating sections of the ring main overnight and watching which isolation reduces the cycling will narrow the search substantially before anyone picks up a detector.

One adjacent win. Many plants run system pressure higher than any tool requires, often as compensation for leaks nobody fixed. Reducing set pressure after fixing leaks compounds the saving, and the off-hours data will show the compressor cycling less at the lower setting.

With categories identified, the question is which to fund.

Prioritising by cost

Rank by annualised cost, not by how dramatic the signal looks.

The calculation. For each identified load: average kilowatts above baseline, multiplied by the hours per year it persists, multiplied by your cost per kilowatt-hour. State the tariff you used and the date, because every figure downstream depends on it.

A worked example. An idle load of 12 kW persisting for the 128 non-production hours in a week is 1,536 kWh per week, or roughly 79,900 kWh per year. At a rate of 0.12 per kilowatt-hour that is approximately 9,600 per year, from one asset that nobody switched off. Substitute your own tariff before quoting any of this internally.

Why ranking by cost rather than by signal size matters. A furnace showing a large overnight draw may be entirely legitimate and unavoidable. A modest compressor cycle may be trivial to fix and run 8,760 hours a year. The second is often worth more than the first, and a ranked list built on peak draw will put them in the wrong order.

Scale context. Guidewheel's Nice House of Plastics case study reports idle draw falling from 1,500 kWh per week to consistently under 500, which Guidewheel puts at around 52,000 kWh per year or approximately $4,644 annually at that plant's rates. Useful as an indication of the order of magnitude available at a single site. Not a figure to reuse, because the tariff is theirs.

Sort the list into three buckets before presenting it.

• Procedural: fixable by changing what people do at end of shift. Zero capital, immediate, and usually the largest single block.
• Maintenance: leaks, stuck drains, failed controls. Small spend, fast payback, needs a work order.
• Capital: variable-speed drives, compressor replacement, insulation, controls upgrades. Real payback analysis required.

Presenting them mixed together invites the meeting to argue about the capital items and never action the procedural ones, which are free.

Verifying the fix

Re-measure the same window, under the same conditions, and compare like with like.

The same-window rule. If your baseline was a Friday-evening-to-Monday-morning window, your verification is a Friday-evening-to-Monday-morning window. Not a Tuesday night, not a bank holiday, not a week when the plant shut early. Off-hours draw varies with what production left behind, and comparing dissimilar windows produces savings that evaporate on inspection.

What to compare. Total kilowatt-hours across the window, the specific load you targeted, and the baseline floor. The third is the control: if the baseline moved too, something other than your fix changed and the attribution is unsafe.

Expect partial results. A leak survey typically finds and fixes most of a leak load, not all of it. A shutdown checklist works for a few weeks and then decays as people forget. Both are normal, and both are reasons to re-measure quarterly rather than declaring victory once.

Watch for regression explicitly. Procedural savings are the ones that come back. The checklist that eliminated a 12 kW idle load will stop being performed unless someone owns it and the number is visible. Put the off-hours figure on the same standing agenda as the in-production energy metrics described in energy by shift and product, and it stays fixed. Leave it to goodwill and it does not.

What this method cannot do, restated because it matters: it narrows where to look and quantifies what the looking is worth. It does not locate a leak, diagnose a component, or tell you a compressor is failing. Guidewheel is not a predictive maintenance tool and current data is not a component diagnosis.

For costing the findings, see energy cost per machine, and for the wider programme, cut factory energy costs.

Frequently asked questions

About the author