Leak Repair Persistence Dashboard

Leak RepairPersistence Dashboard

Show which compressed-air repairs remain effective, which have returned and which have not been checked.

1. Engineering answer: a repaired leak is not a permanently removed loss

A leak repair persistence dashboard follows each verified repair into later operating periods. It connects the original leak identity, repair episode, initial verification and subsequent rechecks so that maintenance teams can distinguish a repair that still passes, a confirmed recurrence and an item whose current condition is unknown. Its purpose is not to reward a high count of closed work orders. It is to show whether the physical correction remains effective under the conditions that matter.

The essential management question is: “Of the repairs that have reached their follow-up point, how many have been checked properly, how many still pass and what is happening to the rest?” A dashboard reporting 95% successful repairs without showing its eligible population and inspection coverage can conceal a substantial evidence gap. A repair that has not been revisited is not automatically failed, but neither is it proven persistent.

These are connected outcomes, not interchangeable measures. A successful coupling repair may reduce a local loss while a new leak elsewhere increases zone demand; a demand reduction may also fail to reduce electricity as expected until the compressor controls respond.

2. Page purpose and relationship to the existing engineering content

It does not replace those pages or rewrite their source histories.

They extend the supplied engineering material into a usable dashboard specification; they are not presented as formulas already contained in a manufacturer standard. All four worked examples are synthetic training cases, not records of Reddy Compressors customer projects.

3. Define the reporting boundary before counting repairs

Start with the site, compressed-air system, pressure zone, selected campaign and reporting cutoff. Record the operating service to which the dashboard applies: production air, instrument air, workshop distribution or an explicitly defined subsystem. Keep different systems separate where a common comparison would hide incompatible pressures, air-quality obligations or operating patterns.

Specify whether the population is all actionable leak findings, all reported repairs, all initially verified repairs or the subset due for a selected follow-up. Each population answers a different question. The dashboard must display its chosen population and all active filters beside the result, including excluded zones, unavailable assets and the dates that define the campaign.

A leak register does not establish survey completeness by itself. Keep the survey route and inaccessible-area record alongside it. Likewise, a persistence dashboard establishes the status of known repairs, not that every component in the plant has been examined. A campaign covering only the accessible production floor must not be described as a whole-site leak-free assessment.

4. Preserve three identities: location, incident and repair episode

Use a stable leak-location ID for the exact component or source, an incident ID for a distinct leakage event and a repair-episode ID for a completed intervention. Inspection records then reference the episode being evaluated. This structure avoids treating three photographs, two work orders and one invoice as six repaired leaks.

A location ID remains useful even if a paper tag is lost. Store the asset tag, component description, branch or machine position and dated photographs. If the entire component or assembly is replaced, retain its old identity and link the successor rather than silently rewriting the original record. A nearby leak is not necessarily the same defect.

5. Use separate operational and follow-up statuses

The operational status records where the work stands. The follow-up status describes the evidence at a particular horizon or reporting date. Keeping these separate prevents a machine repaired yesterday from changing a failed 90-day result for an earlier repair episode. Historical outcomes remain attached to the episode that produced them.

For each follow-up horizon, display PASS, CONFIRMED RECURRENCE or UNKNOWN, together with “not yet due” or “out of comparable scope” where applicable. UNKNOWN includes missing access, an unsuitable operating state, an invalid test or no recheck. It is an evidence condition, not a favourable or unfavourable assumption.

6. Initial verification starts the persistence clock

The persistence clock begins at the first accepted local verification after the repair, not the work-order creation date, invoice date or technician departure time. Until that check passes, there is no established successful repair whose persistence can be tracked. Repair-to-verification delay belongs in a separate execution metric.

Record the actual pressure near the defect, machine state, relevant valve or actuator cycle, instrument or inspection method, verifier and acceptance basis. An isolated, depressurised or bypassed connection cannot prove that an in-service leak has been corrected. For an intermittent defect, the test must cover the condition in which it appeared, not simply a convenient idle period.

Use precise outcome wording such as “no leak indication detected at the identified coupling during the stated production cycle, within the method’s detection capability.” This records an observed result without creating an unlimited guarantee of zero leakage. A photograph of a replaced component is repair evidence, but it is not a substitute for this test.

7. Choose follow-up horizons from duty and consequence

Follow-up can be scheduled by elapsed calendar time, pressurised hours, machine cycles or a defined combination. The appropriate basis depends on how the component is exposed. A moving hose used across three shifts accumulates a different challenge from a rarely pressurised standby branch. Record the basis and the reason for the chosen interval.

For illustration, an owner might use 30-, 90- and 180-day review points, but these are not universal recommended intervals. A high-consequence or repeatedly failing connection may need an earlier review; an infrequently used machine may need a functional check after its next relevant operating sequence. The plan should remain compatible with safe access and production requirements.

Set any early and late inspection tolerance explicitly before reporting. A check performed well before a target cannot establish the target outcome, and an excessively late inspection should not be relabelled as an on-time check. Store the target, permitted window, actual inspection time and achieved exposure separately. A due date is a planning requirement, not evidence of an inspection.

8. Build fixed follow-up groups and keep denominator changes visible

For a selected horizon, group repair episodes by initial verification period and the same exposure rule. Freeze the group ID, inclusion rules and reporting cutoff. Recent repairs that have not reached the target remain “not yet due”; they must not dilute the mature group’s recurrence percentage. This is especially important when repair volume rises sharply near the end of a reporting month.

For the default fixed-horizon view, an episode becomes due when its target and any allowed late window have elapsed by the reporting cutoff. A confirmed recurrence before that target remains a failure even if no further test was possible. Retirement before the target without a known recurrence is shown outside comparable scope, with its reason and count; it does not become a pass.

Within the due, in-scope population, every episode must fall into one of three categories: valid passing follow-up with no earlier confirmed recurrence in that episode; confirmed recurrence; or unresolved evidence. A later successful re-repair does not erase the original failure. When reviewing an existing report after data correction, issue a revised snapshot rather than quietly changing the original denominator.

9. Report persistence and coverage together

Let N be the due, in-scope repair episodes for one stated follow-up horizon; P the passing episodes; F those with confirmed recurrence; and U those whose result remains unknown. Every episode is counted once in this partition. Retired, not-yet-due and unresolved-identity exclusions are shown separately, with reasons.

N = P + F + U

Evaluated coverage (%) = 100 × (P + F) / N

Observed persistence among evaluated episodes (%) = 100 × P / (P + F)

Confirmed passing share of the entire due group (%) = 100 × P / N

Unknown share (%) = 100 × U / N

Always show the numerator and denominator, not just the percentage. If N is zero, the result is “not applicable: no due episodes.” If P + F is zero, observed persistence is “not evaluated,” not 0% and not 100%. Confirmed recurrence among evaluated episodes is F/(P + F); its complement describes that evaluated population only.

For transparency, a bookkeeping range from P/N to (P + U)/N shows what the passing share could be if every unknown eventually failed or passed. It is not a statistical confidence interval and does not correct inspection-selection bias. The headline should remain the observed result paired with coverage, not the most favourable end of that range.

10. Distinguish recurrence, incomplete repair and a genuinely new leak

A failed initial check is an unsuccessful or incomplete repair. A defect reappearing after an accepted initial pass is a recurrence candidate, subject to exact location, mechanism and test-condition review. A new leak at a different port or adjacent joint requires its own incident, even if the same machine is involved. These distinctions should be decided from evidence, not whichever classification improves the closure report.

Store the last valid passing check and the first later detection of recurrence. Unless a suitable continuous record establishes onset, the dashboard knows only that the recurrence was detected at the later check; it does not know the precise moment the seal failed. A day-90 finding after a day-30 pass means onset occurred somewhere within the intervening observation interval.

Retain the classification as suspected when identity or evidence is unresolved. Do not allow automatic text matching of “hose leak” to merge independent defects. A reviewer should be able to see the earlier photographs, component position, repair action and inspection conditions before accepting the recurrence link.

11. Compare operating exposure without inventing failure dates

Calendar age is useful for scheduling, but pressurised hours or cycles can improve comparisons between comparable component duties. Record how exposure was obtained: controller counters, operating logs, a documented shift schedule or an estimate. Running hours are not automatically pressurised hours, especially where branches are isolated outside production.

Descriptive detected-event rate = confirmed recurrence events / documented exposure × selected scale

Use a clearly labelled scale such as detected recurrences per 1,000 pressurised hours, and do not mix hours and cycles. For a first-recurrence analysis, exposure belongs only to time when the episode is under observation and at risk; stop at the recorded first recurrence, removal or observation endpoint. Intermittent inspections leave the true onset interval unresolved, so the displayed rate is a descriptive detected-event rate, not an exact physical failure hazard.

A component family with short observation or few failures should remain a limited-evidence comparison. Inspection intensity also matters: a closely monitored area can appear worse simply because more defects are discovered. Show exposure coverage, inspection frequency and population size before ranking contractors, zones or component types.

12. Maintain comparable pressure, state and detection capability

Record the same relevant physical source and operating condition at each visit. Pressure changes can alter leakage indication and flow; machine motion can expose a defect not present at rest. Noise, instrument settings, distance and access affect the usefulness of an acoustic check. Different inspection conditions need explanation before a trend is interpreted as deterioration or successful repair.

Do not convert an acoustic reading directly into a universal volumetric flow. Where the instrument supports estimation, retain the documented method, reference pressure, distance requirements and uncertainty or confidence description. A qualitative “no indication” result is valid for a qualitative inspection but must not be silently converted into a precise zero-flow value.

A pressure-normalisation model is an additional engineering calculation, not a default correction available for every leak. Its applicability depends on geometry, flow regime, pressure basis and method.

13. Keep the physical repair result separate from flow persistence

Count-based persistence answers how many checked repairs still pass. Flow-based persistence asks how much of a previously established air-demand reduction remains. These can differ greatly: many small repairs can persist while one large repaired source returns. A separate loss-weighted view can expose that difference, provided all included flow values share compatible conditions and a defensible estimation or measurement method.

Initial reduction D0 = sum(pre-repair flow − immediately post-repair flow)

Current retained reduction Dt = sum(pre-repair flow − current comparable flow)

Retained reduction (%) = 100 × Dt / D0, only when D0 > 0

Use exactly the same included component set for D0 and Dt. Exclude missing-current-measurement items from both terms of a paired comparison and disclose the excluded original flow share. Do not give those missing items zero current leakage. A qualitative pass cannot populate a quantitative field without a justified bound or method.

Values above 100% or below zero may be physically possible if a source improves further or becomes worse than its original condition. Preserve the calculated value and investigate; do not silently clip it to make a tidy graphic. Additional improvements, changed duty and measurement error must be considered before calling the difference persistence.

14. Reconnect local results with the zone and station

A zone-flow trend sees the combined effect of repaired sources, new leaks, legitimate demand, pressure changes and operating schedules. It cannot by itself allocate every change to a particular repair campaign.

After a major repair campaign, check whether the compressor fleet has actually reduced input power, loaded time or unnecessary running. Supply controls can cause a flow reduction to produce a different energy response from an average specific-power estimate.

Do not count the same benefit in both a leak-repair dashboard and a sequencing project total. The portfolio result should reconcile to the defined measurement boundary. A stable 90-day local repair cohort and rising zone demand can both be true when new defects appear elsewhere; the correct response is to locate the new demand, not to rewrite the earlier repair history.

15. Prioritise recurring mechanisms and overdue evidence

Display confirmed recurring locations by component family, exposure, consequence, repeat repair count and current estimated loss where supported. A common description alone is not proof of a common root cause.

The work queue should separate “repair needed” from “verification overdue” and “follow-up overdue.” Each has a different next action. A verification visit may resolve an evidence gap without further repair, whereas a confirmed recurrence needs a corrective work decision. Keep owner, dependency, due date and latest evidence beside the priority, rather than presenting an unexplained red score.

Safety, pressure integrity and process-critical functions outrank a purely financial ranking. No dashboard task authorises tightening pressurised connections, defeating protective devices or isolating an essential air user. The site’s competent personnel govern the repair method, isolation, safe access and return to service. The dashboard records those approvals; it does not replace them.

16. Worked example A: why a high percentage can hide weak coverage

Synthetic case: 120 repair episodes were initially verified in the selected campaign. At the reporting cutoff, 20 have not reached the chosen target and 10 were retired before that target without known recurrence. The remaining 90 are due and in comparable scope. Of these, 54 have a valid passing follow-up, 6 have a confirmed recurrence and 30 have unresolved follow-up evidence.

The correct headline is “90.0% observed persistence among 60 evaluated episodes; 66.7% evaluated coverage of 90 due episodes.” Reporting simply “90% of repairs remain effective” would extend the finding to an unexamined population. The immediate action is to resolve the 30 unknowns, prioritising consequence and access, while investigating the six confirmed recurring defects.

17. Worked example B: repair count and retained flow tell different stories

Synthetic case: all four repairs below passed the agreed initial acceptance test, which permits the listed measured residuals. All flow values are comparable reference m³/min at the stated test conditions. At the selected follow-up, A and C still pass; B and D have confirmed recurrence. No current measurement is missing.

D0 = 1.40 − 0.07 = 1.33 reference m³/min

Dt = 1.40 − 0.35 = 1.05 reference m³/min; retained reduction = 78.95%

Count persistence is 2/4 = 50.0%, while 78.95% of the initial flow reduction remains. Both figures are correct because the repairs differ in flow contribution. The 0.28 reference m³/min increase since initial repair identifies lost benefit in this selected group, not the total leakage of the plant. Electricity and rupee savings still require a separate measured or justified system response.

18. Worked example C: an uncertain recurrence date changes benefit estimates

Synthetic case: an initially verified repair is rechecked and passing at day 30, then recurrence is first detected at day 90. The failure was not continuously monitored. Assume, for this scenario only, that the repair saves 0.80 reference m³/min until recurrence, then loses the full benefit. The component is pressurised for 16 hours every day. These simplifying assumptions are explicit, not inferred field facts.

Supported scenario range for effective exposure: 30 × 16 to 90 × 16 = 480–1,440 h

Avoided air volume scenario: 0.80 × 60 × exposure = 23,040–69,120 reference m³

Suppose a separately supported marginal system response for this scenario is 5.0 kW per reference m³/min. The corresponding power reduction is 4.0 kW and the energy scenario is 1,920–5,760 kWh. An illustrative energy price of ₹8/kWh gives ₹15,360–₹46,080. Neither the price nor the response factor is a current quotation or a universal compressor value.

The wide range is the consequence of an unknown onset date. It must not be presented as 90 days of verified continuous savings, and the local inspections do not independently validate the electricity calculation. Partial degradation, changed pressure or nonuniform hours would require a different model. The dashboard should retain the interval rather than manufacture a precise failure timestamp.

19. Worked example D: new leaks can mask successful earlier repairs

Synthetic case: comparable residual zone demand is 4.0 reference m³/min before a repair campaign and 2.0 immediately afterwards. At day 90 it is 2.8. The original campaign’s local records indicate that a 0.20 reference m³/min loss has returned. A separately identified new defect contributes 0.60, with all other legitimate demand and test conditions assumed unchanged for this example.

Initial zone reduction = 4.0 − 2.0 = 2.0 reference m³/min

Day-90 zone reduction = 4.0 − 2.8 = 1.2 reference m³/min = 60.0% retained

Original campaign reduction retained = 2.0 − 0.20 = 1.80 reference m³/min = 90.0%

The original repairs retain 90% of their initial reduction, but the zone retains only 60% because a new 0.60 loss has appeared. The new finding belongs in a new incident, linked to the affected asset rather than blamed automatically on the campaign contractor. The explanatory bridge is 2.0 + 0.20 + 0.60 = 2.8. In real data, this bridge may also contain an unresolved residual and measurement uncertainty; it should not be forced to balance by inventing a leak.

20. Editable repair-episode and recheck worksheet

Use the following two linked worksheets as a manual starting point. Add rows for actual records; leave absent evidence blank and explicitly classified as unknown. The Word tables are editable records, not live calculations. Initial and later inspections may use the same record layout, but their purposes and acceptance conditions must be distinguished.

21. Editable reporting snapshot and evidence-quality record

A management snapshot should be reproducible. Record the dataset version, selected filters and cutoff before entering totals. The count reconciliation is as important as the headline: confirm that the initial population is accounted for through due, not-yet-due, scope-changed and unresolved-identity categories. Explain any restatement from the previous issue.

Apply evidence flags at the individual record and result level: measured, method-supported estimate, qualitative inspection, assumed exposure, unsuitable conditions, missing evidence or unresolved identity. A high-quality photograph does not make an unmeasured flow estimate high quality. Equally, a precise meter cannot establish that two observations refer to the same component.

22. Troubleshooting dashboard signals without jumping to conclusions

When persistence falls, first check that the population, horizon and inspection rules have not changed. Then examine the recurring episodes and actual repair methods. A change from short-horizon to long-horizon reporting can lower apparent success even if the underlying programme is unchanged. A new survey team or more capable instrument can also reveal defects previously missed.

Do not resolve a disappointing indicator by lowering acceptance criteria retrospectively. Criteria can be revised for a justified technical reason, but the revised method must have its own version. Prior reports should remain available on their original basis, with a separate restated comparison where necessary.

23. Practical review cycle and sustained ownership

At each reporting cycle, reconcile new repairs, initial verifications, due follow-ups, failed checks and scope changes. Confirm that overdue critical items have an owner and an executable next step. Review the oldest unverified repairs separately from the oldest known open defects; their barriers and risks are not the same.

For recurring component families, record the proposed permanent correction and the evidence expected to demonstrate improvement. A better support arrangement, an appropriate coupling or a revised operating practice should be connected to the relevant incidents. Repeated replacement of the same component without investigating the mechanism can keep the repair count high while persistence remains poor.

The supplementary DOE/CAC guidance supports maintaining a master leak list, documenting repairs, examining recurring problems, checking the baseline and repeating the programme. It does not adopt generic leakage percentages or savings promises as site results. [R1]

Request a persistence review with the existing leak register, repair and recheck records, operating schedules, pressure information and any available zone-flow or station-energy data. Identify inaccessible areas and the reporting question. The useful outcome is a scoped evidence and action plan, not a guaranteed saving inferred from the number of tags.

24. Questions from maintenance, energy and plant management

What is leak repair persistence?

It is the demonstrated continued effectiveness of an initially verified repair over a stated operating exposure or follow-up period. The result belongs to an identified repair episode and the conditions of the recheck, not an unlimited leak-free guarantee.

Does a completed work order prove that a repair persists?

No. It records work performed. An initial service-condition check establishes local closure; later suitable inspections supply persistence evidence. Keep all three milestones visible.

Why must the dashboard show inspection coverage?

A pass percentage calculated from inspected repairs may omit many unknowns. Showing the evaluated count divided by the due population prevents a small or selectively inspected subset from being presented as the complete campaign.

Should an inaccessible repair count as a failure?

Not automatically. It remains unknown when its current condition cannot be established. Keep its access restriction, consequence, owner and next inspection opportunity visible instead of assuming success or failure.

Is recurrence the same as a failed initial verification?

No. Recurrence follows an accepted initial pass. A leak still present at the initial check indicates incomplete or ineffective repair. Preserve the distinction so repair quality and longer-term durability can be improved separately.

How often should repaired leaks be checked?

Use a site-defined schedule based on consequence, recurrence, component duty, pressurised exposure and access. Calendar or cycle examples in this page are illustrative, not universal requirements.

Can ultrasonic readings be added to calculate saved air?

Only where a supported quantification method produces compatible flow estimates and the conditions and limitations are retained. A qualitative acoustic pass cannot be converted automatically to a precise zero-flow value.

Can local repairs persist while total leakage increases?

Yes. New defects or changed operating conditions can increase zone demand while earlier repairs remain effective. Separate existing-episode recurrence from new incidents, then reconcile the system trend.

Can the dashboard calculate verified rupee savings?

Not from repair counts alone. It needs a supported energy response, valid operating periods and the relevant financial assumptions. Use the separate energy and economics tools, keeping scenarios distinct from measured outcomes.

What happens when a recurring leak is repaired again?

Create a new repair episode linked to the same location and incident history. Reverify the new work, but do not change the earlier episode’s failed persistence result into a pass. Historical failure remains useful evidence.

IMPLEMENTATION APPENDIX — NOT PUBLIC PAGE COPY