What a measured leakage-repair record has to prove
Most leak programmes end with a spreadsheet of tags and a headline saving. That is a useful start, but it does not show what changed in the compressed-air system. A plant engineer, an energy manager or an auditor reviewing the work will want to see evidence that the leaks existed, that the repairs were done properly, that delivered air fell at the same pressure and production state, and that the result was still there weeks later.
We treat a leakage project record as a chain. Each link must be supported by its own evidence, and a weak link limits what can honestly be claimed further down. If the baseline was measured at 7.5 bar (108.8 psig) and the follow-up at 6.8 bar (98.6 psig), for example, part of the apparent reduction is simply the lower pressure, not the repairs.
The record should answer these questions plainly:
- Where were the leaks, how were they found and on what date?
- Which tags were repaired, by whom, using what method, and did each one pass a retest?
- What was the measured system or branch leak flow before and after, at what pressure?
- How did the compressors respond, and what did the station draw in kW?
- Did anything else change in the same period, such as production, shift pattern or compressor dispatch?
- Was the reduction still present at a later check?
Six claims that should never be blended
The most common weakness in leak reporting is sliding from one claim to the next without the evidence for the second. The table sets out what each claim needs and what it does not, on its own, establish.
| Claim | What establishes it | What it does not establish |
|---|---|---|
| Leak located | Survey tag, location, detection method, date and surveyor | Leak flow or any saving |
| Leak repaired | Work order, repair description, parts used and a passed retest | That the repair will last |
| Leak flow reduced | Before and after flow measurement at matched pressure, or a bounded estimate with its uncertainty | An equal reduction in electrical kW |
| Compressor demand changed | Measured delivered air or loaded hours from the station | Annual kWh without knowing how the controls responded |
| Electrical energy reduced | Matched, adjusted station kWh within a defined meter boundary | That leaks alone caused it if other changes occurred |
| Cost reduced | An agreed tariff, period and energy basis supplied by the plant | A guaranteed return for any other site |
A record can be perfectly valid at the first two levels and still say nothing about energy. That is acceptable, provided the wording stays at the level the evidence reaches.
How the leaks are found, and what each method can prove
Different survey methods produce different kinds of evidence. None is wrong, but each has limits.
Ultrasonic detection
An ultrasonic detector picks up the high-frequency sound of air escaping through a small opening, even in a noisy plant. It is excellent for locating leaks at fittings, couplings, hoses, FRL units, drain valves and cylinder seals. Some instruments give a flow estimate from the sound level, but that estimate depends on distance, angle, orifice shape, pressure and background noise. Treat it as an indicator for ranking, not as a measured flow.
Flow logging and isolation tests
A flow meter on the main header or a branch, logged during a non-production period with all air-using equipment stopped, gives a direct reading of the leak flow in that boundary. Isolating branches one at a time shows where the losses sit. This is the strongest basis for a before-and-after comparison, provided pressure is matched.
Drawdown and load-time methods
Where no flow meter is fitted, leak flow can be estimated from the time a known receiver and pipe volume takes to fall between two pressures, or from the compressor’s loaded time during a shutdown with known free air delivery (FAD). Both methods are well established but carry more uncertainty: the stored volume is often guessed, and the compressor’s actual FAD may differ from its nameplate rating.
Power logging
Logging compressor input kW alongside pressure shows how the station responds to the change in demand. It is essential for any energy statement and useless on its own for locating leaks.
Why pressure must be matched before and after
A leak behaves broadly like an orifice. At the pressures used in most plants, the air escaping rises with the upstream absolute pressure, so a system run at a higher pressure leaks more through the same holes. If the before and after tests are run at different pressures, the comparison mixes two effects: the repairs and the pressure change.
The practical rule is simple. Record the pressure at the point where flow is measured, use the same measurement point both times, and either run both tests within the same pressure band or correct the result to a common pressure with the correction method stated. Always note whether readings are gauge or absolute; a mismatch between bar(g) and bar(a) is a common source of error.
Pressure also affects the energy side. As a typical figure, each extra bar of discharge pressure costs about 7% more compressor power for the same delivered air. So a project that both repairs leaks and lowers the setpoint must separate the two contributions, or report them together as one combined intervention.
The repair register and retest chain
Every tag in the survey should end in one of a small number of states: repaired and retest passed, repaired and retest failed, deferred with a reason, not repairable without a shutdown, or found not to be a leak. A record that only lists the repaired tags hides the open ones, and those still consume air.
- Survey: assign each leak a unique tag, location, photograph where permitted, detection method and severity band.
- Plan: group repairs by shutdown need, isolation, parts and permit requirements.
- Repair: record the work order, what was done (re-made joint, replaced hose, new drain valve, seal kit) and the parts used.
- Retest: check each repaired point with the same method used to find it, under pressure, and record pass or fail.
- Close: sign off the tag only after a passed retest. Failed tags go back to the repair list.
- Measure: run the post-repair flow test at matched conditions.
- Follow up: repeat the check after an agreed interval to confirm persistence.
The full register method is covered on our leak repair register page; the point here is that an energy or flow claim is only as good as the closure of the register beneath it.
Confounders that can make or break the result
Leak flow is rarely the only thing that changes between two measurement periods. These are the factors we look for before attributing any change to the repairs:
- Production volume, product mix or shift pattern changed.
- New air users were added, or old machines were removed or isolated.
- Compressor setpoints, pressure bands or sequencing were adjusted.
- A dryer, filter or drain was serviced, changing pressure drop or purge demand.
- A compressor was serviced and now delivers closer to its rated FAD.
- Ambient temperature differed markedly, for example a test in the Telangana summer at 40–45 °C against one in a cooler month, which affects compressor capacity and cooling.
- Valves to unused zones were closed, which is a valid saving but not a repair.
Where a confounder cannot be removed, it should be stated and, where possible, adjusted for. If it cannot be adjusted, the claim should step down a level, for example from “energy reduced” to “leak flow reduced in the tested boundary”.
Uncertainty and conservative reporting
Every measurement has an uncertainty: the meter’s stated accuracy, the installation effects (straight-run lengths, swirl), logging resolution, the stability of the test condition and the length of the logging period. When before and after values are close, the uncertainty can be larger than the change itself. In that case the honest statement is that no reduction could be confirmed within the measurement accuracy, not that a small saving was achieved.
Good practice is to report the measured change together with its uncertainty band, use the conservative end of the band for any downstream estimate, and state clearly which figures are measured and which are calculated. Our measurement uncertainty worksheet sets out a simple way to combine these sources.
Persistence: did the result last?
Leaks come back. Push-fit couplings loosen, hoses age, drain valves stick open and new leaks appear as maintenance work disturbs the network. A leak flow measured the day after repairs is the best case, not the long-term result.
A persistence check repeats the matched-condition flow test after a defined interval and compares it with both the baseline and the immediate post-repair value. It should also record any new leaks found since the original survey, so that a stable flow figure is not masking fresh leaks replacing repaired ones. Where the plant runs a continuing leak programme, the trend over several surveys is far more informative than any single before-and-after pair.
How to word the result
The wording should follow the evidence. These examples show the difference:
- Bounded scope: “A leak survey of the packing hall identified tagged leaks at fittings, hoses and drain points; repairs were made and each repaired point passed an ultrasonic retest.”
- Measured statement: “Non-production leak flow on the main header, measured at the same pressure band before and after repairs, fell by an amount larger than the stated measurement uncertainty.”
- Modelled statement: “Based on the measured flow change and the compressors’ measured specific power, the estimated energy effect is stated with its assumptions; it has not been confirmed by metered kWh.”
- Not acceptable without evidence: a percentage saving, a cost or kWh figure for another site, or a claim that all leaks were removed.
Any project we publish follows these rules, is anonymised unless the plant has agreed otherwise in writing, and contains no figures the plant has not approved.
How Reddy Compressors helps
We have worked with compressed-air systems since 2002 from our office at Tarnaka, Secunderabad, and our service teams cover the whole of Telangana and Andhra Pradesh, with project support elsewhere in India where feasible. As an independent multi-brand specialist we service and support ELGi, Atlas Copco, Ingersoll Rand, KAESER, Chicago Pneumatic and other makes, so the station can be assessed as a whole rather than one brand at a time.
For a leakage programme we can carry out the survey, tag and record the leaks, plan and carry out repairs that fall within our scope, retest repaired points, log flow, pressure and power where suitable instruments can be fitted, and follow the ISO 11011 assessment method for the wider audit. We do not issue conformity certificates. We also check the compressors themselves, because a machine that is not delivering its rated air, or a control setting that keeps a unit running unloaded, can hide or exaggerate the effect of the repairs. Our team of over 20 service engineers and technicians can also support the AMC and preventive maintenance that keeps leaks from returning.
What to send us
- A simple layout or list of the compressed-air network and main branches.
- Compressor makes, models, ratings and nameplate photographs.
- Normal operating pressure and the setpoints on each compressor.
- Production hours and any non-production window when a leak test could run.
- Any previous leak survey, register or meter data you already hold.
- What you want the record to support: maintenance closure, energy reporting or both.
Frequently asked questions
How do I prove a compressed-air leak repair actually saved air?
Measure leak flow in the same boundary before and after the repairs, at the same pressure and with production stopped or matched. A flow meter on the header or branch gives the strongest evidence; drawdown or loaded-time methods are acceptable with a stated uncertainty. Then check that no other change, such as a new setpoint or isolated zone, explains the difference. A list of repaired tags alone shows work done, not air saved.
Can an ultrasonic leak detector measure leak flow in cfm?
Some ultrasonic instruments estimate flow from the sound level, and those estimates are useful for ranking leaks and planning repairs. They are affected by distance, angle, orifice shape, pressure and background noise, so they are indicators rather than measurements. For a flow or energy claim, use a flow meter, a drawdown test or the compressor loaded-time method at matched pressure, and treat the ultrasonic figures as supporting detail.
Why does system pressure matter when comparing leak tests?
Leaks behave like orifices, so the air lost through the same holes rises as the upstream pressure rises. If the baseline test runs at a higher pressure than the follow-up test, part of the apparent reduction comes from the pressure change rather than the repairs. Use the same measurement point, keep both tests in the same pressure band or correct to a common pressure, and record whether readings are gauge or absolute.
Does reducing leaks always reduce compressor electricity use?
Not always by the same proportion. The power saved depends on how the compressors respond: a variable-speed unit trims down smoothly, while a fixed-speed machine may simply spend more time unloaded, still drawing a share of its power. Poor sequencing can absorb the saving altogether. That is why an electricity claim needs station kW or kWh measured within a defined boundary, not just a lower leak flow.
How long should we wait before checking that leak repairs have lasted?
There is no single correct interval; it depends on how quickly the network changes and how the plant runs. Many sites choose a follow-up after several weeks of normal production and then a regular survey cycle. What matters is that the persistence test uses the same method, boundary and pressure as the baseline, and that new leaks found since the first survey are recorded separately from repaired ones.
Can we publish leak repair results without naming the company?
Yes. A project can be described by industry, region and system size without naming the plant, provided the details are not enough to identify it and the owner agrees to publication. We publish only anonymised, factual records the plant has approved, with no prices, staff names or commercially sensitive production details. Figures are shown only where the measurement evidence behind them has been reviewed.
Related pages
Calculators and worked examples
These tools give indicative figures for screening, comparison and for preparing an enquiry. They run in your browser and send nothing. No prices or tariffs are assumed: enter your own to see costs. Efficiencies and other figures shown are example values: replace them with your own. Results are not a performance guarantee or a design certificate; final selection is confirmed by a service engineer against the machine data sheet and the site.
System leakage test: Load/unload timing, residual flow or pressure decay
Turn a non-production leakage test into a leakage flow, a share of compressor delivery, leak power and annual energy. Choose the method you used; only that method’s inputs are read. Optionally enter the result after repairs, or the measured station power before and after, to state the reduction.
The worked example below shows the method with fixed numbers. Turn on JavaScript to change the inputs.
Worked example
Inputs
- Test method: Load/unload timing (fixed-speed compressor)
- Compressor actual delivered air, FAD (m³/min): 10
- Load/unload: loaded time per cycle, averaged (s): 30
- Load/unload: unloaded time per cycle, averaged (s): 70
- Residual: measured flow with production stopped (m³/min): 5
- Residual: verified legitimate continuous users (m³/min): 1.2
- Decay: system volume, receivers plus pipework (m³): 5
- Decay: start pressure (bar g): 7
- Decay: end pressure (bar g): 6
- Decay: time for the pressure fall (s): 60
- Atmospheric pressure (bar a): 1.013
- System pressure during the test (bar g): 7.5
- Normal operating pressure to restate the result at (bar g): 7.5
- Leakage after repairs, same method and pressure (m³/min; 0 if not yet measured): 0
- Station specific power (kW per m³/min): 6.5
- Pressurised hours per year: 6000
- Optional: station power in the idle test before repairs (kW; 0 to skip): 0
- Optional: station power in the idle test after repairs (kW): 0
- Idle (non-production) hours per year for the measured power: 2000
Result
- Loaded share of the cycle: 30% T ÷ (T + t) = 30 ÷ 100 s
- Estimated leakage at the test pressure: 3 m³/min free air = 50 L/s = 105.9 cfm (load/unload)
- Leakage as a share of compressor delivery: 30% of 10 m³/min
- Leak power at the operating pressure: 19.5 kW at 6.5 kW per m³/min
- Leak energy per year: 1,17,000 kWh over 6,000 pressurised hours
- Leak energy cost per year: Enter your electricity tariff (cost per kWh) above
Method. Load/unload timing (fixed-speed load/unload compressor, production stopped, timed over several cycles): leakage ≈ FAD × T ÷ (T + t), with T the loaded and t the unloaded time per cycle. Residual flow (flow meter, production stopped): unexplained residual = measured residual flow − verified legitimate users (dryer purge, instrument air, drains); it is the figure to investigate, not automatically all leakage. Pressure decay (compressors stopped and isolated, no users): leakage in free air ≈ V × (p_start − p_end) ÷ (p_atm × time), with V the receiver plus pipework volume in m³, pressures in bar and time in minutes; the pressure difference is the same in gauge or absolute terms. Leak flow through an orifice rises roughly in proportion to upstream absolute pressure, so a result taken at one pressure is restated at the normal operating pressure by × (p_op + p_atm) ÷ (p_test + p_atm). Leak power = leakage × station specific power (kW per m³/min); annual energy = power × pressurised hours. The after-repair figure must come from the same method at the same pressure. The measured-power option gives (kW before − kW after) × idle hours, which is a measured saving, not an estimate. Money is shown only when you enter your own tariff.
What to take from it. Record the method, the pressure and the plant state with every result, so the next test is comparable. A lower leak flow saves energy only if the compressors then unload, slow down or stop.
Technical content reviewed by D. Srinivas Reddy, Managing Partner and service engineer, Reddy Compressors. Last updated 17 October 2026. How our engineering content is reviewed.
Leak repair persistence
For one cohort of verified repairs that has reached its follow-up date, report inspection coverage, persistence and recurrence separately, with the evidence gap and the air still being lost.
The worked example below shows the method with fixed numbers. Turn on JavaScript to change the inputs.
Worked example
Inputs
- Number of eligible repairs (follow-up date passed): 120
- Number of repairs re-checked: 90
- Number of repairs still passing: 72
- Number of repairs leaking again (recurred): 18
- Number of recurrences in the most frequent component group: 11
- Estimated flow of recurred leaks (m³/min): 0.9
- Estimated flow of new leaks found nearby (m³/min): 0
- Specific power (kW per m³/min): 7
- Running hours per year: 6000
Result
- Inspection coverage: 75 % (90 of 120)
- Persistence rate: 80 % of re-checked repairs still pass
- Recurrence rate: 20 % of re-checked repairs leak again
- Other outcomes among re-checked: 0 repairs (neither passing nor recurred)
- Unknown status: 30 eligible repairs not yet re-checked
- Persistence range over all eligible: 60–85 % (all unknowns failed – all unknowns passed)
- Most frequent component group: 61.1 % of recurrences (11 of 18)
- Residual leak flow: 0.9 m³/min (32 cfm)
- Compressor power for residual leaks: 6.3 kW
- Energy per year: 37,800 kWh
- Cost per year: Enter your electricity tariff (cost per kWh) above
Method. Coverage = re-checked ÷ eligible. Persistence = still passing ÷ re-checked; recurrence = leaking again ÷ re-checked (inspected items only). Unknown = eligible − re-checked. The persistence range assumes every unknown repair failed (low) or passed (high), each over all eligible repairs. Residual flow = estimated flow of recurred leaks + new leaks nearby; power = flow × specific power. Other outcomes = re-checked − passing − recurred (for example component removed). Default: the page example (120 eligible, 90 re-checked, 72 passing, 18 recurred, 11 of them quick couplings, 0.9 m³/min).
What to take from it. Report coverage beside persistence. A high persistence figure from a low coverage describes a small, possibly unrepresentative sample.