Your Compressed Air Leak Estimate Is Wrong in Two Directions

Compressed air is the most expensive utility in a plant per unit of work delivered, and it is the only one whose losses are invisible. You cannot see a leak. You cannot see artificial demand. You cannot see a compressor burning 70% of full power while delivering nothing useful. Walking the floor tells you almost nothing, and listening tells you less over ambient noise.

The physics sets the stakes. Only 10–15% of the electrical energy put into a compressed air system reaches the tool as useful work — the rest leaves as low-grade waste heat and leakage (US DOE / Bonneville Power Administration). Air can still be the right answer for safety, for duty cycle, or for tooling you already own. It is never the cheap answer.

So the prize is real. The trouble is how it usually gets quantified.

The standard estimate carries two errors, pointing opposite ways

Almost every leak-cost pitch is built the same way: look up a leak diameter in an orifice table, multiply by hours and an electricity rate, and present the annual dollars. Both inputs to that multiplication are usually wrong.

The orifice table overstates the flow. The widely circulated table — 1/32" is 1.6 cfm, 1/8" is 26 cfm, 1/4" is 104 cfm at 100 psig — comes from Compressed Air Challenge Fact Sheet 7, and it is the theoretical orifice: a discharge coefficient of 1.0. A real sharp-edged leak flows about 62% of that. Run the physics at cd = 0.62 and the 1/8" row that reads 26 cfm computes to 16.1 scfm. The table overstates by roughly 1.6×.

The electricity rate understates the cost. The leak-cost tables that circulate with those figures are mostly built at $0.05/kWh, which is roughly half of current industrial rates.

Put them together and the two errors partly cancel. That is the trap, and it is worth being precise about why.

An estimate that lands near the truth because two mistakes offset carries no information about the plant it came from. You cannot tell a good one from a bad one, because you never knew the size of either error here. Leak geometry is not sharp-edged in any predictable way, so the 1.6× is a central tendency and nothing more. And the two inputs are not equally knowable: the electricity rate is on the client's bill and can be exact, while the flow through a leak you have not opened up cannot be.

The practical consequence is worse than the arithmetic. Quote a table at $0.05/kWh to a plant manager who knows what they pay, and every other number in your report gets discounted along with it.

Measure the whole system at once

There is a way to get the leak number without an orifice table, without a discharge coefficient, and without extrapolating from a sample. Measure the entire system's leakage in a single test.

Run it with production down, all air-operated equipment off at the machine, and the system at normal operating pressure — a weekend, a holiday, or between shifts. Under those conditions, anything still consuming air is by definition a leak.

On load/unload compressors, let the machine cycle with no production demand and record at least five complete cycles:

Leak load (% of capacity) = (T × 100) ÷ (T + t)

  T = loaded (on-load) time
  t = unloaded (off-load) time

The method is documented by CAC Fact Sheet 7 and confirmed independently by PNNL's Uniform Methods Project — useful to have in your pocket when a client challenges the number.

For any other control type, charge the system, isolate the compressor, and time the decay:

Leak flow (cfm) = (V × ΔP ÷ (T × 14.7)) × 1.25

  V  = system storage volume (ft³), receivers + piping
  ΔP = pressure drop during the test (psig), about 50% from normal
  T  = time for that drop (minutes)
  1.25 = correction for leaks falling off as pressure decays

Read the receiver nameplate for volume. Do not guess it — a guessed V puts you right back where the orifice table left you.

What the number licenses you to say

The point of the test is a decision, and the benchmarks are published rather than invented (CAC Fact Sheet 7):

Notice that the first row is a finding. A leak test that comes back at 7% has told you where not to spend, and that is worth as much as one that comes back at 25%. An assessment that can only ever discover a problem is a sales process wearing an assessment's clothes.

Then cost it on their bill, not on a table

Once you have a measured flow, the annual cost is straightforward:

$/yr = CFM × (SP ÷ 100) × H × R

  SP = specific power of the marginal compressor (kW per 100 cfm) — measure it
  H  = annual running hours
  R  = the client's marginal electricity rate, the one that actually changes
       when load changes, not the blended average

Take a single 1/8" leak at 100 psig, a compressor at 18 kW/100 cfm, running continuously, at $0.10/kWh. From the orifice table it costs $4,100/yr. Computed at a real discharge coefficient it costs $2,539/yr. Both are large, and plants usually have dozens — but only one of those numbers survives being checked.

Flow scales with orifice area, which is why a handful of large leaks almost always beats a long list of small ones for repair sequencing. Each doubling of diameter is about four times the cost.

The sequence matters more than any single number

ISO 11011:2013 splits the system into supply, transmission, and demand, and the audit order is not cosmetic. Demand-side waste inflates everything upstream of it, so sizing supply before correcting demand means buying efficiency you did not need. Plants that fix leaks first often find they can shut a compressor off entirely — a saving no supply-side project could have produced.

Measure demand first. Then size supply to the corrected demand.

The full field protocol behind this article — six tests, decision criteria, and every threshold carrying a graded source marker — is published as part of the SVEND method library, aligned to ISO 11011:2013. Each figure in it is marked as standard, program, trade, or vendor material, because knowing which kind of number you are holding is most of the skill.