Set medium-filter replacement criteria by comparing the current pressure drop against the clean pressure drop recorded for that specific bank, at the same measurement boundary and the same duty airflow, and act when the reading reaches whichever limit comes first: the filter manufacturer’s applicable operating limit, the fan’s usable pressure allowance, or the point at which the process loses the airflow it needs. Pressure drop is the pressure difference across the filter face, and it rises as dust loads the media and narrows the air paths. The criterion only holds if the comparison stays like for like, so the reading boundary, the airflow condition and the filter identity must be fixed before a trigger value means anything.

Define the pressure boundary and clean baseline

A pressure reading has no value as a replacement criterion until you state exactly what it spans. A tapping pair across one medium-filter bank, a reading across a single filtration stage, and a reading across the whole air handling unit each describe a different restriction: the AHU figure folds in coils, dampers, silencers and the final filters, while the bank reading isolates the elements you intend to change. If the clean baseline was taken across the bank and a later maintenance check is read across the unit, the two numbers are not comparable and the trend built from them is meaningless. Fix the tapping points, record them in the maintenance instruction, and keep the same gauge or transmitter arrangement for the life of the bank.

The second half of the boundary is the airflow condition. Pressure loss through a filter varies with the volume of air passing it, so a baseline captured during commissioning balancing at one fan speed does not describe the same bank running at a different duty. Define the normal duty airflow for the system, state it with the baseline, and note the operating mode it belongs to if the unit runs at more than one setting.

With the boundary and the duty fixed, record the clean pressure drop of the bank as installed, using the actual elements and the actual rack arrangement. A catalog clean-resistance figure describes a filter model tested at a stated face velocity in a test rig; the installed bank may use a different element depth, a different number of cells, blanking plates, or a frame geometry that adds its own loss. The baseline you want is the one the maintenance team can compare against for the next several years, which means it belongs to this project and this bank, not to one published value assumed to cover every filter size, airflow and measurement arrangement.

Filter identity completes the record. Note the model, class, dimensions, depth, number of elements per bank and the rack layout alongside the clean reading, because a replacement element of a different depth or media area will shift the clean baseline and quietly invalidate the trigger. When the replacement-filter configuration is selected, it should be matched to the installed rack and the operating duty rather than to the nominal size alone.

Set the action point from filter and system constraints

The replacement action point is set by whichever of three constraints is reached first, not by the filter’s published final resistance alone. The filter manufacturer’s applicable operating limit protects the element itself: beyond it, media deflection, frame distortion or seal strain become the concern. That limit tells you how much pressure the filter can tolerate; it says nothing about whether the rest of the system can still deliver its duty at that resistance.

The fan’s usable pressure allowance is the second constraint. Every system is selected with some margin between the clean-condition total pressure and the fan’s capability at the design flow, and the medium-filter bank is one claimant on that margin along with the final filters, coil fouling and duct balance. If the fan is near the top of its curve, or already running at high speed to hold room pressures, the medium filters have to be changed well before they reach their own operating limit. Read this margin from the commissioning records and the fan curve for the installed impeller and drive, not from the design intent on paper.

The third constraint is the airflow the process actually needs. In a cleanroom served by constant-speed or limited-turndown plant, rising filter resistance reduces volume, which shows up as falling air change rates, weakening room pressure differentials, or a terminal HEPA stage no longer receiving the supply it was balanced for. A unit with variable-speed control may hold flow while the filter loads, in which case the loss appears as rising fan speed and power rather than falling flow; the constraint is then how much speed headroom remains. Either way, the airflow requirement can force action before any pressure limit is approached.

Once the limiting constraint is identified, set a warning point below the action point so that filters, access permits and labor can be arranged before the system is forced into an unplanned change. Where the change affects a classified space, the decision to return the room to service after the work belongs to the project’s approved procedure and its designated approver, not to the pressure reading alone. ISO 14644-4:2022 treats cleanroom design, construction, start-up and verification as a connected process, which is the reason a filter changeout in a classified facility is scoped with its re-verification rather than as an isolated mechanical task. Defining the trigger and the scope of the change also defines what the changeout will cost to plan for.

Interpret rising pressure at comparable airflow

A rising pressure drop only indicates filter loading when the airflow through the bank is comparable to the airflow at the baseline. Because pressure loss through a filter increases with the volume of air passing it, a system that has been rebalanced upward, had a damper opened, switched from a reduced night setting to full occupancy duty, or had its fan speed raised will read higher across the same bank with no change in the media at all. Interpret any rise against the operating point it was taken at before attributing it to dust.

This means the trend record has to carry its accompanying operating data. A pressure figure alone is not a trend; a pressure figure with the airflow, fan speed or control setpoint at the moment of reading is. Where the unit uses variable-speed control to hold a fixed supply volume, the airflow stays constant and the pressure rise does reflect loading, but the fan speed and power become the useful secondary evidence that the margin is being consumed. Where the unit runs at fixed speed, both pressure and flow move together, and the pair has to be read as one operating point.

Having separated a more demanding flow duty from genuine resistance, the shape of the trend is informative. Gradual, steady rise over months is ordinary dust loading and lets you forecast when the action point will be reached. An abrupt step with no change in duty suggests a sudden event, such as construction dust, a prefilter stage that has failed or been removed, or a filter displaced in its frame. The exception to patient interpretation is independent evidence: visible damage to media, frame or gasket, or any other replacement criterion the project has defined, calls for its own action regardless of what the pressure trend shows.

Investigate unexpected pressure falls and damage

A pressure drop that falls while the system stays in service is a signal to investigate, never a reason to extend the change interval. Loaded media does not clean itself, so a lower reading means that either less air is passing the filter, the instrument is no longer measuring what it was, or some of the air is bypassing the media.

Check the airflow first, because a reduced volume lowers the pressure across the bank even as the filter continues to load. Falling flow can come from a closed or slipped damper, a fan belt or drive fault, a speed-control change, or heavy fouling further along the air path. The second check is the measurement itself: blocked, kinked or disconnected sensing tubes, a leaking tapping, a drifted or zeroed-out gauge, and a transmitter that has lost power or calibration all read low without anything happening to the air.

The third possibility is a new bypass path, and it carries the most consequence for the space downstream. A torn element, a filter unseated in its frame, a failed gasket, a loose clamping mechanism or a gap at a blanking plate lets air past the media rather than through it, which reduces resistance while reducing filtration at the same time. On a medium-filter bank, the practical effect is that the final filters downstream receive dirtier air than designed and load faster. Each of these three findings calls for a different correction — restore the duty, repair the instrument, or replace and reseat the element — so the investigation has to reach a cause before any decision about the change interval is made.

Write the replacement instruction and action record

The maintenance instruction turns the criterion into something a technician can execute without reinterpreting it, and it needs five things on the page: the normal measurement condition, the warning point, the action point, the independent damage criteria, and the person or role who owns the action.

The normal measurement condition names the tapping points, the operating mode and the duty airflow the reading must be taken at, together with the clean baseline for that bank and the filter identity it was recorded with. The warning point is the value that triggers procurement and scheduling; the action point is the limiting condition agreed for the installation, whether that is the filter’s operating limit, the fan’s usable allowance or the airflow floor the process needs. The independent damage criteria stand apart from pressure entirely: visible media, frame or gasket damage, water damage, or evidence of bypass mean replacement whatever the gauge shows. Naming an action owner keeps the trigger from being recorded and then left unassigned.

The wording matters more than the format. An instruction phrased as trend at normal duty, investigate unexpected falls, and replace at the agreed limiting condition tells staff how to use pressure evidence without inviting them to invent a universal threshold or a fixed calendar interval that the installed bank never justified. Each reading should be logged with its date, the airflow or fan setting it was taken at, and any observation about the unit’s operating mode, so the record supports a trend rather than a sequence of unrelated numbers.

Before approaching a supplier about replacement elements or about setting the criterion for a new bank, have the installed filter model, class and dimensions, the rack arrangement and element count, the clean pressure drop recorded at a stated duty airflow, the tapping arrangement used for that reading, the fan curve and commissioning balance data, and the airflow the served space requires. With those in hand, the action point can be set against the constraint that actually binds the installation, and the replacement filters can be matched to the rack they have to seal into.

Frequently Asked Questions

Q: Can we copy the replacement setting from another medium-filter bank?
A: Only after confirming that it applies to the installed bank and its duty. Compare the filter identity, rack arrangement, airflow and pressure measurement boundary, then check the applicable filter limit and system constraints. A setting based on a whole air-handling unit cannot be assumed to apply to a reading taken across one filter stage.

Q: Does a pressure rise after an airflow change mean the filters need replacing?
A: The rise alone does not establish filter loading. Interpret it alongside the changed airflow before deciding on replacement, and check whether the current operating point still meets the required airflow within the filter and fan constraints. Keep airflow conditions with the pressure record so later comparisons remain meaningful.

Q: Can a lower pressure reading justify postponing a planned replacement?
A: Not without checking why the reading fell. Verify airflow, instrument connections and possible filter or seal damage before extending the interval. Reduced airflow or a bypass path can produce a lower reading without showing that the filters are in better condition.

Q: Should we wait for the pressure action point if a filter or seal is visibly damaged?
A: No. Treat visible damage as an independent maintenance decision rather than letting a pressure reading postpone action. The replacement instruction should state the applicable damage criteria and who is responsible for acting on them, alongside the pressure-based warning and action points.