How to Measure Pressure Drop Across a Compressed Air System

Why this matters

A total system drop is not an answer. "You are losing 18 psi" tells an owner what they are paying and nothing about what to fix, and the two most common responses to it, repipe the main or buy a bigger compressor, are both wrong more often than they are right. The deliverable from this job is an allocation: a line for every segment, a measured share against a planned share, and one segment with an obvious variance. What follows builds that sheet field by field, because the sheet is the procedure.

Before any port is opened or any coupler is pulled

Fitting a test gauge opens a pressurised path. Close the upstream isolation valve, open the vent, confirm the section gauge reads zero with the vent still open, and lock and tag that isolation under 29 CFR 1910.147, which covers stored mechanical and pressure energy, before a wrench touches the fitting. A line opened while still pressurised ejects the fitting and whips the hose, which is a struck-by injury.

Taking the last reading means working at a tool that is in use. Never uncouple a pressurised hose: close the drop shutoff, bleed the hose through the tool's own trigger with the tool pointed into a clear area and hands and face out of the discharge path, and confirm it is dead before the coupler is pulled. Air driven into a break in the skin can enter the bloodstream, which is the same hazard behind 29 CFR 1910.242(b) limiting compressed air used for cleaning to less than 30 psi with effective chip guarding and personal protective equipment, so never clear a port or a bench with a blow gun. Air rooms and shop floors commonly reach the 85 dBA eight-hour time-weighted average action level, at which point hearing protection and the rest of 29 CFR 1910.95 apply.

Field 1: the operating state, written once at the top

Every reading on the sheet belongs to one named state, and the state goes on the page before any number does: the date, the clock time, what was running, and the flow if you can get it.

The state has to be the peak, because that is the condition the discharge setpoint has to cover. Friction loss in pipe scales with roughly the square of volumetric flow, a relationship derived for fully turbulent flow with the friction factor treated as roughly constant, so halving the flow quarters the loss and a survey taken over lunch reports a system with no problem. A filter element's differential does not necessarily follow the same exponent, since it depends on the media and its loading, so an off-peak survey understates every segment but by different amounts, which corrupts the allocation as well as the total.

If the plant's peak is unpredictable, the honest options are to stay until it happens or to fit recording gauges and let them catch it. Neither of those is measuring at 10 am and multiplying.

Field 2: the station list, defined by position

List the stations from the compressor to the worst tool, and define each by where it is rather than by what is next to it. The gap between two stations is a segment, and every segment you fail to bracket disappears into a neighbour.

A workable station set for most plants: compressor discharge; receiver outlet; downstream of filtration and the dryer; the header at the far end of the main; the branch takeoff serving the worst tool; downstream of that drop's shutoff and filter-regulator-lubricator; and the tool inlet at the end of the hose and coupler. Six segments, and any of them can be split further if you already suspect it.

Use existing gauge ports and test ports first. Every port you add is a joint you opened, and the whole procedure above applies to each one.

Field 3: the instrument row, filled in before the readings

Write down which gauge, its range, and its accuracy statement copied off the instrument rather than remembered. Then write down what kind of error that statement describes, because the answer changes what your subtractions mean.

Use one gauge and move it. A gauge's error at a given point on its dial is dominated by a systematic offset, and a systematic offset that is a fixed number of psi is common to both readings and cancels completely in the difference. That is the single strongest reason to move one instrument rather than to hang two.

A percent-of-reading systematic term does not cancel completely. It leaves that same percentage of the difference. Read 105 and 95 on a gauge carrying a 1 percent-of-reading offset and the errors are 1.05 and 0.95 psi, so 0.10 psi survives into a 10 psi difference, which is 1 percent of the difference. Small, and not zero, and worth stating so nobody quotes the cancellation as absolute.

Two gauges do not cancel anything. Their offsets are unrelated. If all you know about each is a worst-case bound from its data sheet, worst-case bounds add linearly and the result is reported as a bound, never as a plus-or-minus. Two 0 to 160 psi gauges each specified at 1 percent of full scale are a bound of 1.6 psi each, so the bound on any difference taken between them is 3.2 psi. Against the 1 psi segment on the sheet below that is not a measurement at all, and against the 9 psi segment it is still a 36 percent bound.

Repeatability is the other kind. If you take a segment three times and the spread is genuinely random rather than a drift, independent spreads combine in quadrature, so two equal spreads multiply by the square root of two rather than by two. Do not fold a bound and a spread into one figure.

Field 4: the readings, each tied back to the state

Because there is one gauge and seven stations, you cannot take them simultaneously. Re-establish the same state for each reading and record the state's own witness alongside the number: the header pressure at the receiver, read the same way each time, or the machines that were running. If the witness moves, the reading belongs to a different state and it does not go on this sheet.

If a segment's two ends sit at materially different elevations, correct for the static column before subtracting; the correction and the two-tap technique for a single component belong to their own article, and this survey applies them at each pair.

Field 5: the allowance column, filled in before the variance

Write the planned share for each segment next to the measured one. Where a total distribution loss allowance is specified as a percentage of operating pressure, take it from the piping system manufacturer's design data or the design standard your specification invokes, in the edition it names, and split it across segments. For a filter or dryer, the allowance is the element's change-out differential from the manufacturer's data, not its clean rating, because the change-out figure is the one your setpoint has to carry.

Without this column the sheet reports what is happening and cannot say what is wrong.

The filled sheet

One gauge, moved between existing ports, at the afternoon peak with the plasma table and both cells running. Discharge setpoint 105 psig.

Station readings behind those segments: 105, 104, 95, 92, 90, 89, 87 psig. The total checks two ways, as 105 - 87 = 18 and as 1 + 9 + 3 + 2 + 1 + 2 = 18.

Read the variance column, not the measured one. The largest measured segment and the largest variance happen to be the same line here, but they need not be: the hose and coupler measured 2 psi and are exactly on allowance, so there is nothing to do there even though 2 psi is real pressure. The drop came in 2 psi under its allowance, which is a genuine outlier and the sheet reports it as one rather than smoothing it: that FRL is either newer or lighter duty than the allowance assumed, and it is the one segment on the page that is doing better than planned.

The finding. Filtration and dryer carry 9 of the 18 psi, half the total, against a 4 psi allowance. That is an element differential question and it is answered with a service, not with a repipe.

The second reading. Elements changed, same state re-established, treatment measures 4 psi. Total falls from 18 to 13 psi and the setpoint can come down by 5 psi, which is realised only when someone actually turns it down. What is left is a main and a branch each carrying 1 psi more than allowance, which is 2 psi of structural excess in a 13 psi system and not worth opening a wall for.

The failure mode. Run the same survey at half the peak flow. The pipe segments fall to roughly a quarter of their peak values by the square-law scaling, and the treatment segment falls too but by an exponent between one and two depending on the element's media and loading, so somewhere between a half and a quarter of 9 psi. At the quarter end that is about 2.25 psi against a 4 psi allowance, and the survey clears the one segment that is actually the problem. Every error in an off-peak survey runs in the same direction, toward a clean sheet.

How to verify you got this right

Check the sheet against itself first. The segment differences must sum to the first station minus the last, and if they do not, a station was read in a different state or a segment boundary was crossed twice.

Then check the total against the setpoint. Your first station sits at the discharge setpoint, so the last station plus the total has to return to it: 87 plus 18 is 105. Then compare that last station against what the worst tool actually needs at its inlet, because the gap between those two is your whole remaining margin. If the tool needs more than the last station delivers, the setpoint is already too low for the plant as piped, and if your segments do not account for the difference between discharge and tool, there is a loss between two stations you did not bracket. The usual culprit is a partly closed isolation valve nobody knew about.

Then re-take one segment cold, on a different day at the same state, before quoting the variance to anyone. A segment that repeats within its own error character is a finding. A segment that does not repeat was never a segment; it was a state that moved while you were walking.

References

• 29 CFR 1910.147, control of hazardous energy, for isolating and verifying depressurisation before fitting a test port or breaking a coupler
• 29 CFR 1910.242(b), compressed air used for cleaning, and 29 CFR 1910.95, occupational noise exposure, including the 85 dBA eight-hour time-weighted average action level
• Instrument data sheet for the gauge's accuracy statement and its basis; piping system manufacturer design data and the project's specified design standard, in the edition it names, for segment allowances; filter and dryer manufacturer data for change-out differential
• See related: How to Read a Pressure Drop Across a Component; What Pressure Drop Through Distribution Actually Costs; Why Pipe Size Decides More Than Compressor Size