Resources/Compressed Air: The Condition Monitoring Gap in Your Plant
Operational Excellence

Compressed Air: The Condition Monitoring Gap in Your Plant

Compressed air gets managed as a utility bill instead of an asset class. Here is what to instrument, which pressure and dew point trends predict failure, and how to make leak repairs stick.

13 min read
By Monitory team

It is 3am and the packaging line is down. Operators say the filler is dropping out of sequence. The night shift mechanic finds nothing wrong with the filler. Three hours later, daylight arrives with the real answer: the desiccant dryer has been passing wet air for weeks, and the header pressure sags below the filler's minimum setpoint every time the case erector cycles.

Nobody saw it coming because nobody was watching the air system. This article covers what to instrument on a compressed air system (kW, discharge pressure and temperature, dew point, differential pressure across separators and filters, header flow, and point-of-use pressure), which trends predict failure (dew point drift, rising start counts, specific power creep), how to turn an ultrasonic leak survey into repairs that actually close, and how to report results against both reliability and energy goals.

Compressed air is usually managed as a utility bill and a compressor skid. It is actually a distribution system with four production-stopping failure modes: leaks, dryer failure, degrading air/oil separators, and short-cycling compressors. None of those live on the energy report.

The 3am Call That Was Never About the Compressor

Ownership is the first problem. The energy team tracks kWh and demand charges. Maintenance tracks the compressor skid because it has a nameplate and a PM route. The header, the dryers, the drains, and the point-of-use regulators sit between the two and belong to neither.

That gap shows up in the work order history. Search most CMMS systems for air-related failures and you find compressor PMs, a handful of motor repairs, and a recurring annual line item for a leak survey. You rarely find a failure code for "wet air downstream of dryer" or "pressure excursion at critical machine," which means air-caused downtime gets charged to whatever machine stopped.

Department of Energy operations and maintenance guidance treats program structure, planning resources, and documented practices as the foundation of maintenance performance [5]. For compressed air, that structure is usually missing entirely. The equipment is instrumented well enough to run, not well enough to diagnose.

Why Air Systems Hide Their Own Degradation

A compressed air system is engineered to absorb abuse. Leaks grow, a trim compressor picks up the load, and pressure holds. A filter fouls, the operator bumps the setpoint up, and the symptom disappears. The cost compounds quietly while the pressure gauge looks fine.

Watch for compensation behavior, because it is the clearest evidence of hidden degradation:

  • Setpoint creep: discharge pressure has been raised over the years and nobody can name the machine that required it
  • A trim unit that never trims: the second compressor runs at part load continuously instead of cycling on demand
  • Drains blowing open: timed condensate drains set to dump long and often because somebody got tired of water carryover
  • Regulators wide open at the machine: point-of-use pressure cranked to compensate for header sag
  • Nuisance filter changes: filters replaced on a calendar because differential pressure was never trended

Demand-side failures almost always present as quality or production problems first. Pneumatic actuators slow down and cycle times drift. Wet air reaches instrument lines and valve positioners. Rust appears in cylinders. Paint shops see defects that get blamed on the booth.

ISO 8573-1 exists precisely to prevent this ambiguity. It defines air purity classes for particulate, water, and oil contamination so a plant can specify the air quality its processes require. A plant that never specified a class has no reference point, so it cannot tell when its air got worse. It can only tell when a product failed.

Failure modeFirst observable signalWhat production seesTypical misdiagnosisMonitoring method
Leak load growthIdle-period header flow rising week over weekPressure sag during peak demand"The compressor is undersized"Thermal mass flow meter at header, idle baseline
Desiccant dryer degradationPressure dew point drifting upward over weeksWater in instrument lines, valve faults"Bad solenoid batch"Dew point transmitter at dryer outlet
Air/oil separator foulingDifferential pressure across separator trending upHigher energy cost, hot discharge"Normal seasonal variation"DP transmitter, trended not spot-read
Short-cycling compressorStarts per hour rising, load/unload chatterMotor and contactor failures"Motor quality problem"Start counter plus control band width
Stuck-open condensate drainContinuous flow at a drain with stable loadUnexplained flow increase overnight"Leaks got worse"Area sub-metering, drain inspection round

What to Instrument on the Supply Side

Start with the minimum viable instrumentation per compressor, and insist on kW rather than amps alone. Amps tell you the motor is energized. kW tells you what the machine is costing you at a given flow, which is the only way to compare units or track degradation.

Per compressor, capture:

1. kW from a power meter or the controller's power output 2. Discharge pressure and discharge temperature 3. Load/unload state on fixed-speed units, or VFD speed feedback on variable units 4. Running hours and start counts, both read as cumulative counters so rate can be derived 5. Differential pressure across the air/oil separator element

Separator and filter health needs trending, not spot readings. A technician who writes "DP OK" on a PM sheet has recorded nothing usable. A DP transmitter that logs every minute produces a curve, and the curve tells you when the element is loading up versus when it just got dirty from a single upset.

Dryers deserve their own instrument set. On a desiccant unit, measure pressure dew point at the dryer outlet and confirm tower switching, because a failed switching valve can hold a tower on line past saturation while the panel lights stay green. On a cycling refrigerated unit, trend refrigerant suction temperature and outlet dew point together so you can separate a refrigeration fault from a load problem.

Storage and control close the loop: receiver pressure, header flow from a thermal mass flow meter, and the control band width configured on the master controller. A narrow or collapsed control band is one of the most common causes of short-cycling, and it is a configuration problem, not a mechanical one.

Vibration earns its place on an air compressor later, not first. Condition monitoring platforms built around vibration and temperature sensing with alerting and technician feedback are well suited to rotating assets [1], and on an air system the right targets are the motor bearings and the gearbox on a rotary screw package, once bearing failure history or asset criticality justifies the coverage. If you have not yet instrumented dew point and separator DP, vibration is not your gap. Our condition monitoring guide covers how to sequence sensing by failure mode rather than by sensor availability.

The Demand Side Everybody Skips

Supply-side instrumentation tells you how hard the compressors are working. It does not tell you why. For that you need pressure at the machines that care.

Install point-of-use pressure transmitters at the three or four most pressure-sensitive assets in the plant, not just in the compressor room. Those are usually the machines with the highest minimum operating pressure and the tightest cycle timing. When one of them faults, you want a timestamped pressure trace at that machine, not a header gauge two hundred feet away.

Idle flow is the cleanest leak indicator available. Measure header flow during a non-production window with the plant down and the air system still pressurized. Everything moving at that moment is leakage, stuck drains, or demand that should not exist. Repeat the measurement monthly and the trend becomes a leak management KPI nobody can argue with.

Demand-side walkdown checklist

Capture these fields for each finding so the survey produces data, not impressions:

  • Open blow-off at conveyors or dryers: location, estimated nozzle count, whether engineered nozzles are installed, area owner
  • Cabinet cooling with compressed air: panel ID, vortex cooler or open tube, available alternative, electrical area owner
  • Air motors and vacuum generators: asset ID, duty cycle, whether a dedicated blower or pump is feasible
  • Regulator settings at critical machines: current setting, machine minimum per the OEM manual, delta
  • Condensate drains: drain ID, type (timed, zero-loss, manual), observed discharge behavior, last service date
  • Abandoned drops: location, whether the drop is valved off, parts needed to cap it

Sub-meter flow by production area if the piping allows it. A flow increase that can be assigned to an area has an owner and a supervisor who can act. A flow increase measured only at the header belongs to everybody, which means it belongs to nobody.

Reading Pressure and Dew Point Trends as Early Failure Indicators

Dew point drift is the most underused predictive signal in the plant. On a desiccant dryer, outlet pressure dew point rarely fails suddenly. It walks upward over weeks as the desiccant loses capacity or a switching valve begins leaking across. If you trend it daily and alert on rate of change rather than a fixed limit, you get a planning window instead of a water carryover event.

Short-cycle count works the same way. Rising starts per hour on a load/unload unit points at three candidate causes: insufficient storage, a collapsed control band, or a growing leak load that keeps the unit chasing pressure. The count itself does not diagnose, but it tells the reliability engineer which three things to check before the motor starter fails.

Specific power, expressed as kW per 100 CFM per compressor, exposes separator fouling and internal wear well before efficiency loss shows up on a utility bill that aggregates the whole plant. Track it per unit, at a consistent load point, and the degradation curve becomes visible.

Alert logic matters more than sensor count. A static high limit on filter differential pressure catches the problem at the point where you have already been paying for it for months. Threshold alarms and learned-behavior analysis working together are the standard approach in condition monitoring architectures [7], and the same logic applies here: keep a hard limit for protection, and add a trend alert for planning. Start with the simplest reliable signal and add complexity only when the simple version is being used, which is the core discipline in production measurement practice [6].

Trend Alerts Need a Named Decision, Not a Dashboard

Every air system trend alert should name the planner who reviews it and the work order type it creates. Condition monitoring workflows are designed as a path from measurement to alert to investigation to corrective action, with the technician recording what was found [3]. If your dew point alert emails a distribution list, it is not a maintenance signal. It is a notification.

The Ultrasonic Leak Survey That Actually Gets Repaired

Most leak surveys fail at the handoff, not the detection. A contractor delivers a PDF with hundreds of tagged leaks, the file lands in a shared drive, and twelve months later the next survey finds most of the same tags. The detection work was fine. The backlog conversion never happened.

Fix the protocol at the point of tagging. For every leak found with the ultrasonic detector, capture location (building, column line, asset ID), estimated leak rate, repair trade, parts required, and whether the repair needs a line shutdown or can be done live. That last field determines everything about scheduling.

Then write individual work orders, not one lump order titled "Fix air leaks." A single work order for 340 findings cannot be planned, assigned, or closed meaningfully. Individual orders with asset ID, cost center, and area owner can be grouped, batched, and reported. This is where work order automation earns its keep: the survey output becomes planner-ready records instead of a document.

Planner workflow for leak repair batches

1. Group by area and shutdown window, so all repairs requiring isolation land in the same outage 2. Batch by trade, so a pipefitter gets fittings and a controls technician gets solenoids and drains 3. Kit parts per batch from the findings list rather than per work order 4. Quantify estimated recovery per batch so the scheduler can prioritize by impact 5. Re-survey the area after repairs and recheck idle header flow before closing the batch

Verification is non-negotiable. A repair is proven by the idle-period header flow, not by a closed work order. Condition monitoring workflows treat recording the resolution of an abnormality as part of the cycle, not an afterthought [8], and the same rule applies to leak repair: the measurement that triggered the work must be repeated to close it.

Tying Findings to Both Reliability and Energy Goals

Build one dataset that answers two questions: did air-related production stoppages fall, and did specific power improve. The same instrumentation supports both, which is why air system monitoring is easier to fund than most condition monitoring programs.

On the reliability side, report air-caused downtime minutes, the count of pressure excursions below your critical machine setpoints, and MTBF on dryers and separators. On the energy side, report kW per 100 CFM per compressor, idle-period header flow, and any sustained reduction in header pressure setpoint. The asset condition and energy review workflow describes how to run that combined review on a regular cadence.

None of this is reportable without CMMS structure. Dryers, filters, receivers, and headers need to exist as separate assets under the compressor room parent, each with its own failure codes. If the dryer is a line of text in a compressor PM, dryer MTBF does not exist.

If you are still choosing where these signals will land, our comparison of predictive maintenance platforms sets out the evaluation criteria and a 30-day bake-off you can run on a single compressor room before committing the whole plant.

When the instrumentation and asset hierarchy are in place, Monitory ties condition signals from the air skid to planner-ready work orders and multi-site reporting, so a district manager can compare idle flow and air-caused downtime across plants. Note that adding sensors and gateways to a compressor room touches your OT network, and NIST SP 800-82 provides the architecture and risk guidance for that work [4]. Bring controls and IT into the design review before the first gateway is mounted.

Frequently asked questions about compressed air condition monitoring

What is the single most valuable compressed air measurement?

Idle-period header flow. It requires one flow meter and a non-production window, it quantifies leak load directly, and it gives you a repeatable number to trend monthly.

Do I need a dew point transmitter if I already have a dryer alarm?

Yes. Most dryer panel alarms indicate a component fault, not air quality at the outlet. A transmitter gives you the drift trend that precedes the fault.

Should compressed air assets be in the CMMS separately from the compressor?

Yes. Dryers, filters, receivers, and headers each have distinct failure modes and service intervals. Without separate asset records you cannot assign failure codes or calculate MTBF on the components that actually stop production.

Where does vibration monitoring fit on an air compressor?

On the motor bearings and gearbox of rotary screw packages, after dew point and separator differential pressure are covered. Vibration and temperature sensing with alerting and technician feedback is well established for rotating equipment [1], but it does not detect the air quality and leak failure modes that cause most air-related stoppages.

How do I keep leak repairs from recurring annually?

Convert every tag into an individual work order with an area owner, and verify with a re-survey plus idle flow check. A survey that produces a document instead of a backlog will repeat itself every year.

How do I justify compressed air instrumentation to finance?

Quote the two numbers the instruments produce: idle-period header flow in CFM, which converts directly to wasted compressor kW, and air-caused downtime minutes charged to the machines that stopped. Our ROI of predictive maintenance guide covers how to structure those cost categories and the payback calculation.

Start With One Header and One Weekend

The first action takes about one shift to set up: install or commission a thermal mass flow meter on the main header and record flow during the next non-production window. Write down the number. That is your leak and artificial demand baseline, and it costs you one meter and one weekend.

The second action is two transmitters on your base-load compressor: pressure dew point at the dryer outlet and differential pressure across the air/oil separator. Those two signals cover the failure modes most likely to stop production and most likely to be missed by existing PMs.

The metric to track this week is idle-period header flow in CFM, reviewed weekly with a named owner. Pair it with one governance step: assign an air system owner who reports monthly on leaks closed, current header setpoint, and air-caused downtime minutes. Without that owner, the header stays in the gap between the energy team and the compressor PM route.

The 3am stoppage at the top of this article was predictable. Dew point had been drifting for weeks and idle flow had been climbing for months. Both were measurable with instruments that cost less than one night of lost packaging output. The compressor was never the problem. The absence of two transmitters and one owner was.

References

[1] AWS, "What is Amazon Monitron? - Amazon Monitron", AWS documentation. https://docs.aws.amazon.com/Monitron/latest/user-guide/what-is-monitron.html

[2] NIST, "AI Risk Management Framework | NIST", 2023. https://www.nist.gov/itl/ai-risk-management-framework

[3] AWS, "The Amazon Monitron workflow - Amazon Monitron", AWS documentation. https://docs.aws.amazon.com/Monitron/latest/user-guide/deployed-workflow.html

[4] NIST, "Guide to Operational Technology (OT) Security | CSRC", 2023. https://csrc.nist.gov/pubs/sp/800/82/r3/final

[5] U.S. Department of Energy, "Operations and Maintenance in Federal Facilities | Department of Energy", DOE guidance. https://www.energy.gov/cmei/femp/operations-and-maintenance-federal-facilities

[6] Google, "Rules of Machine Learning: | Google for Developers", Google developer guidance. https://developers.google.com/machine-learning/guides/rules-of-ml?hl=en

[7] AWS, "How Amazon Monitron works - Amazon Monitron", AWS documentation. https://docs.aws.amazon.com/Monitron/latest/user-guide/how-monitron-works.html

[8] AWS, "Understanding sensor measurements and monitoring machine abnormalities - Amazon Monitron", AWS documentation. https://docs.aws.amazon.com/Monitron/latest/user-guide/anom-monitoring-chapter.html

Ready to put this into practice?

See how Monitory helps manufacturing teams implement these strategies.