Why a Filter That Is Not Changed Costs More Than One That Is

Why this matters

A loaded filter element does not stop working. It keeps filtering, and it charges the compressor a growing toll to do it, every hour the plant is pressurised, including the hours nobody is in the building. The element itself is a low-cost consumable fitted in a task measured in minutes on a visit that was already scheduled. The restriction it becomes is billed continuously for as long as it is left in. That asymmetry is the whole article, and it is why "it is still working" is not a reason to leave an element in place.

The other half is less obvious and it catches experienced people: the element whose differential pressure never rises is not the healthy one.

Isolate before the bowl comes off

Close the isolation valves either side of the housing, vent the housing through its own bleed, and confirm zero on the housing gauge with the bleed still open before a bowl is turned, then lock and tag the isolation under 29 CFR 1910.147, which covers stored mechanical and pressure energy. A bowl still holding pressure becomes a projectile at head height when the last thread releases.

The old element and the liquid in the bowl carry oil. Wear chemical-resistant gloves and sealed eye protection to pull it, bag it, and route the disposal of oil-bearing condensate to your local sewer authority's pretreatment rules or to a licensed hauler rather than to a floor drain. Do not attempt to blow an element clean for reuse; besides not working, cleaning with compressed air is limited under 29 CFR 1910.242(b) to under 30 psi with effective chip guarding and personal protective equipment.

Where a new differential tap has to be drilled and tapped into a live header, that is the same isolation sequence on the section of pipe, not on the housing alone.

The gate, stated once

Per element, change when measured differential pressure at peak plant flow reaches the manufacturer's stated change point, OR when the manufacturer's stated calendar interval expires, whichever comes first. That is an OR. Either leg on its own is sufficient, and neither waits for the other.

Two parameters have to come off the element's own data sheet, not off this page, because they differ by media and grade. A clean element commonly starts in the 1 to 2 psi range at its rated flow. Change points are commonly set in the 2 to 3 psi range. Calendar intervals for coalescing elements are commonly 12 months. Read all three from the data sheet and write them on the housing with a paint pen, because the next person will not have the sheet.

The unit of analysis is one element, measured across its own housing. A train figure tells you the train is restricted; it does not tell you which stage to touch, and the stage that has aged worst is rarely the one anybody guesses.

Why the drop is the charge and the element is not

Pressure drop through a filter is compressor work that never reaches a tool. The compressor has to discharge higher to deliver the same pressure at the point of use, and the accepted planning rule for a lubricated rotary screw at the 100 psig class is that roughly 1 percent of specific power tracks each 2 psi of discharge pressure. Specific power here means power per unit of delivered flow, so this is a permanent efficiency charge rather than a one-off.

The comparison that decides the maintenance policy is a comparison of durations, not of parts. The element is a task on a visit. The restriction runs for every hour the system is pressurised, which on a two-shift plant is thousands of hours a year, and it grows the whole time. There is no month in which deferring the change comes out ahead once the drop has moved meaningfully above clean.

What that does not mean is "change every element every visit." An element changed early throws away service life that had no cost attached to it, because a clean element's drop is not an avoidable charge, it is the price of having filtration at all. The gate exists precisely to separate those two cases.

Measuring the drop so the number means something

Three things decide whether your differential reading is worth acting on.

Flow. Drop rises with flow, so a reading taken at 7 am on a quiet plant understates the restriction the plant actually pays for at peak. The manufacturer's clean figure and change point both belong to rated flow. Take your reading as close to peak plant flow as the site allows, and write the flow condition beside the number so the next reading can be compared to it honestly.

The instrument's basis. A gauge's accuracy is usually quoted as a percentage of full scale, which is a fixed number of psi at every reading rather than a percentage of what you are looking at. On a 0 to 200 psi gauge, a 1 percent of full scale specification is 2 psi. The quantity you are trying to resolve is 1 to 3 psi. That is the whole measurement, inside the instrument's own error.

The instrument's character. A percent-of-full-scale specification from a data sheet is a worst-case bound, not a measured spread. Bounds from two separate instruments add linearly, so differencing two 200 psi gauges gives a bound of 4 psi on a quantity of about 2 psi, which is worthless. Move one gauge between the upstream and downstream taps instead: its fixed offset is then common to both readings and cancels in the difference, leaving only its repeatability and its resolution. Better still, fit a low-span differential gauge or transducer, 0 to 15 psi class, where the same percentage of span is a much smaller number of psi.

The pop-up or coloured-band indicator built into many housings is a service reminder, not an instrument. It has no stated accuracy, and a stuck one sits comfortably mid-scale forever.

Case one: the element that goes

A coalescing element on a plant's main train. Data sheet: clean 1.3 psi at rated flow, change point 3.0 psi, calendar interval 12 months. In service 7 months. Measured with a single gauge moved between taps at peak plant flow: 3.4 psi.

Run the gate. The drop leg has triggered, at 3.4 psi against a 3.0 psi change point. The calendar leg has not, at 7 of 12 months. Under an OR, one leg is enough. Change it now.

The excess over clean is 3.4 minus 1.3, which is 2.1 psi of accumulated restriction. At roughly 1 percent of specific power per 2 psi for this class of machine, that is a bit over 1 percent of the compressor's power going into the element rather than into the plant, and it has been running that way for some part of the last 7 months and would run that way for the next 5 if left. On a plant pressurised something like 6,000 hours a year, the charge is levied across all 6,000 of them, against a change task measured in minutes.

Note what the gate did not need: an opinion about whether the element "looked bad." Nobody opened it before deciding.

Case two: the identical element that stays

Same element part, same data sheet, fitted to the train on a standby compressor that runs perhaps one week in four. In service 7 months. Measured at peak flow while that machine is carrying the plant: 1.5 psi.

Run the same gate. Drop leg: 1.5 against a 3.0 change point, not triggered. Calendar leg: 7 of 12 months, not triggered. Neither leg fires. Leave it in, write 1.5 psi and the date on the housing, re-measure next visit.

The excess over clean is 0.2 psi, about a tenth of a percent of specific power, which is inside the noise of the measurement and is not worth a truck roll. This element will reach its 12-month calendar leg long before it reaches its drop leg, and that is the correct outcome: a lightly loaded element gets replaced on time rather than on condition, because time is what is aging it.

The two cases are the same part, the same gate and the same technician, resolving opposite ways on measured evidence. That is what a gate is for. A shop running "change every element annually" would have left case one restricted for five more months. A shop running "change on drop only" would leave case two in for years.

The failure the gate's drop leg cannot see

A particulate afterfilter downstream of a desiccant dryer, clean figure 1.0 psi, calendar interval 12 months, in service 14 months. Measured at peak flow: 0.6 psi.

The calendar leg had already fired at 12 months and had been deferred by somebody reading the low drop as evidence of a healthy element. That is the error, and it is a reasonable-sounding one. Work the direction out from the mechanism instead. An element that is catching material accumulates it, and accumulation raises resistance, so a loaded element reads higher than clean and a heavily loaded one reads much higher. There is no mechanism by which an element in service reads meaningfully lower than it did clean while still filtering. A falling or stubbornly flat reading therefore means air is getting past the media rather than through it: a failed seal, a collapsed end cap, a bypass path around the element in the housing.

Cut open, this one had a failed seal. Downstream of it, desiccant fines had been travelling into the header for some part of those 14 months, which is exactly the load the afterfilter exists to stop and exactly the material that scores a valve seat.

That is why the calendar leg is in the gate at all, and why it is an OR. The drop leg detects loading. Nothing about the drop leg detects bypass, and bypass is the failure that does downstream damage rather than merely costing power. A gate written as an AND would have required both legs to fire and would have left this element in indefinitely, because the leg that would have caught it was the one that never fired.

Two practical consequences. When a measured drop comes in below the recorded clean figure for that element, treat it as a finding, not a pass, and pull the element to inspect the seal. And when you record a clean baseline on a new element, record it, because a reading that has fallen is only recognisable against a number somebody wrote down.

References

• 29 CFR 1910.147 for isolating and verifying depressurisation of a filter housing before an element is changed, and 29 CFR 1910.242(b), which limits compressed air used for cleaning to under 30 psi with effective chip guarding and personal protective equipment
• Local sewer authority pretreatment rules for disposal of oil-bearing compressed air condensate
• Filter manufacturer data sheets for clean pressure drop, change point and calendar interval at stated rated flow; gauge manufacturer specifications for accuracy basis and full-scale span
• Department of Energy compressed air system guidance for the relationship between discharge pressure and compressor specific power
• See related: What Each Stage of Filtration Removes; What Pressure Drop Through Distribution Actually Costs