Air Quality Excursion Investigation Workbook

Air Quality ExcursionInvestigation Workbook

Validate the evidence, trace the contamination pathway and demonstrate recovery before closing the incident.

1. Engineering answer: an adverse result begins an investigation, not a diagnosis

An air quality excursion investigation workbook connects an unexpected compressed-air result to its sampling evidence, applicable requirement, operating events and possible contamination pathways. It helps establish what happened, where the evidence applies, which users may be affected, what correction is justified and what must be demonstrated before technical closeout. A number above a limit may be important without identifying the failed component; a questionable sample may require urgent attention without proving that the process air was contaminated.

It preserves the original result while examining both the measurement chain and the physical air system.

The central distinction is between an observed signal, a valid result, a requirement excursion, a supported cause and an authorised release. These are different milestones. Replacing a filter does not establish why a sample failed. A favourable repeat sample does not reconstruct the earlier operating period. An upstream pass does not establish the condition at a downstream connection that was never tested.

The workbook therefore produces a traceable incident record rather than a convenient pass/fail headline. Its proposed workflow and numerical examples are new explanatory authoring built on the supplied foundations. They are not recovered plant measurements, laboratory procedures or evidence that a real investigation has been completed.

2. Define the scope and keep the decision states separate

Use this resource for industrial compressed-air quality concerns involving particles, water vapour, liquid water, oil fractions and explicitly specified process contaminants. An event may begin with a laboratory report, a dew-point alarm, a contamination complaint, visible condensate, an unauthorised bypass or a treatment failure. Record the trigger in its original form. Do not upgrade an operator's description of an oily smell into a quantified total-oil result.

Breathing air, medical gases, occupational exposure and validated sterile-process decisions require their own specialist requirements. A pressure or purity calculation cannot authorise those uses.

Keep technical investigation status separate from operational status. A line may be isolated while the cause remains unknown, or restored under an authorised interim arrangement while longer-term effectiveness review remains open. Record both facts rather than forcing them into a single closed label. The proposed states below are a workbook design, not a universal certification vocabulary.

3. Protect the process without destroying the evidence

The first response follows the consequence of the concern and the site's authorised procedures. Notify the utility owner and relevant process or quality owner. Where required, they determine isolation, diversion, controlled shutdown, alternative supply or a product hold. The workbook records these decisions and their times; it does not invent shutdown thresholds or issue equipment-control commands.

Do not delay necessary protection solely to obtain a better sample. Conversely, avoid unnecessary changes that erase the as-found condition when the site can safely preserve it. Capture available trends, active compressor and dryer selections, valve indications, alarm history and the sample arrangement before routine resets or maintenance overwrite them. Where an immediate intervention is unavoidable, record what changed and which observations could not be collected.

A temporary supply must be assessed at the same relevant quality, pressure, flow and connection boundary as the service it replaces. A clean compressor connected through an unsuitable hose can leave the original exposure question unresolved. An emergency arrangement needs an owner, operating limitations, verification requirements and an expiry or review trigger.

Never disable protective devices or bypass required treatment merely to obtain a favourable test. Sampling connections, isolation and disconnection must be performed by competent personnel using the applicable pressure-system and site controls. Containment protects service or product; preservation protects the ability to explain the event. Both objectives belong in the incident record.

4. Open one incident record with stable evidence identities

Assign an incident ID that survives changes in file names, contractors and report versions. Connect it to site, air system, affected process, initial detection time, notification time and the person responsible for coordination. The well-defined starting statement should describe the concern without embedding an unproven cause: “terminal sample above the specified particle limit” is more defensible than “compressor contamination failure” before the pathway is known.

Retain the original laboratory report, raw instrument export, sample identifier, field sheet and associated photographs. Store corrections and interpretations as additional records, not replacements for the initial evidence. A new report version should identify the superseded issue and explain whether the change concerns transcription, units, method, calculation or interpretation.

Separate the incident, sample, result, operating event, corrective action and release decision into linked records. One incident can include many samples, and one sample can produce several contaminant results. A single laboratory attachment should not be counted as several independent investigations merely because it contains multiple pages.

Allocate responsibilities explicitly. Operations establishes the operating history; maintenance establishes the equipment work; the measurement provider addresses sample and method validity; engineering evaluates the physical pathway; the authorised quality or process owner determines the service and product decisions. One person may perform several roles where appropriate, but the record must still show which decision was made in which capacity.

5. Freeze the requirement and the exact point where it applies

Retrieve the requirement used for the original assessment. Record its source, revision, contaminant or fraction, numerical limit, units, averaging basis, point of application and operating condition. Preserve the wording alongside the measurable interpretation. A report headed “air quality” cannot be judged against an unspecified class or an informal expectation that the air should be clean.

Distinguish a warning threshold, an action threshold, an equipment protection setting and a product or utility specification. They may differ legitimately. Crossing an internal alert may require investigation before any product specification is exceeded. Remaining below an equipment alarm does not establish compliance with a stricter point-of-use requirement.

Check that the observed point matches the required point. A sample upstream of the final filter, downstream of a lubricator or taken with the normal hose removed answers a different question from a sample at the production connection. Keep useful diagnostic evidence, but do not use it to close an obligation it does not represent.

When a requirement is missing or disputed, record the unresolved decision and its owner. Characterisation can continue under an approved scope, while acceptance remains undecided. Do not select a more convenient limit after seeing the result. Any legitimate requirement revision needs its own reason, approval, effective period and assessment of earlier decisions affected by the change.

6. Review sample validity without treating doubt as a pass

A validity review asks whether the result represents the claimed quantity at the claimed point and condition. Check sample identity, collection and analysis times, method revision, instrument configuration, sample pressure, accepted volume, collection medium, handling, laboratory qualifiers and the applicable quality-control records. A technically suitable instrument cannot compensate for sampling the wrong branch.

Inspect the measurement path as well as the process path. Tubing, connectors, pressure reducers, dead volume and temporary fittings may introduce, retain or selectively remove contamination. A protective filter used only for sampling changes a particle result; removing a production hose changes the delivered-air boundary. Document these differences rather than silently correcting them with a generic factor.

Record a concern as a specific validity issue: unresolved sample-volume basis, inadequate stabilisation, disputed point identity or a failed associated control. Do not invalidate a result merely because it is inconvenient or because a repeat is lower. An invalidation needs an evidenced reason and the responsible review. Preserve the original result and its relationship to the revised disposition.

Validity and operational precaution are independent. When measurement evidence is insufficient, the process may still require containment until suitable evidence is obtained. Equally, a valid adverse result can remain unexplained mechanically. The workbook should allow these combinations rather than assuming that uncertainty means no risk or that validity proves the compressor is the cause.

7. Read the result exactly, including qualifiers and uncertainty

Retain the reported value and qualifier separately. “Below reporting limit,” “not detected,” “estimated,” “outside range” and a numeric zero are not equivalent. Record the reporting threshold, units and target fraction. A less-than value with a threshold above the specification cannot establish compliance on that statement alone. Missing oil-vapour evidence must not become zero in a total-oil calculation.

Use the decision rule agreed for the assessment. A simple comparison and an uncertainty-aware comparison may produce different dispositions near a limit. The workbook must show which rule was selected, by whom and against which requirement revision. Do not switch rules after an adverse result in order to manufacture acceptance.

This is a proposed worksheet convention, not a mandatory rule for every laboratory or contract. Missing U blocks this mode rather than silently becoming zero.

The rule does not repair an invalid sample or automatically apply to censored, viable-count or non-numeric results. Those need their method-specific treatment. Keep the arithmetic disposition separate from the final process decision, and apply comparisons before display rounding. A displayed equality at limited decimal precision may conceal a material difference in the retained result.

8. Reconstruct a chronology that distinguishes observation from notification

Place the sample collection interval, online indication, laboratory analysis, report issue, notification and protective response on one timeline. A report received on Friday may describe air collected on Tuesday. Treating Friday as the excursion start can exclude the period that the sample actually represents. Conversely, the Tuesday sample does not establish that the same condition persisted continuously until Friday.

Record timezone, clock source, timestamp resolution and any supported correction. Retain raw and corrected times. Distinguish an event time from the time it was noticed or entered into a maintenance system. A rounded operator entry may be adequate for a shift review but inadequate for determining whether a brief dryer changeover preceded a spike.

Bring compressor selections, dryer cycles, drain events, filter service, temporary-air connections, valve changes, shutdowns, ambient changes and production states into the same record where relevant. Use the trace to identify sequences worth testing, not to convert proximity in time into proof of causation.

If a time is uncertain, preserve a range or an explicit unknown. Do not align curves by shifting them until a preferred explanation appears. A useful timeline can remain partly uncertain while still identifying the next discriminating record, such as a controller export, signed isolation sheet or source-change log. Preserve disagreements rather than letting the latest narrative overwrite the original observations.

9. Define observed, possible and unknown affected periods

Separate the period directly represented by adverse evidence from the wider period that may require impact assessment. A continuously validated monitor may support a bounded event duration. An occasional grab sample usually cannot identify the exact onset or duration between samples. The last acceptable sample is evidence at its point and collection period, not proof of uninterrupted acceptable operation until the next test.

Define an outer assessment window from the available history, relevant changes and the process owner's risk decision. State what supports each endpoint and what remains uncertain. A source change after the last acceptable sample can be a useful investigation boundary, but it must not be chosen solely because it reduces the amount of production under review.

Keep separate windows for different branches or contaminants where their pathways differ. The time at which a bypass closes is not necessarily the time that every receiver and terminal branch has recovered. Transport, mixing and retained contamination can make the effective boundary different; use relevant observations rather than an assumed instant reset.

When several potential windows overlap for the same process, calculate their union before estimating duration or affected production. Do not add overlapping intervals twice. Unknown intervals remain visible. An outer-window production estimate is a screening input for disposition, not proof that every unit produced in that window is contaminated or that units outside it are automatically acceptable.

10. Map the contamination pathway before blaming a component

Build the actual air path for the affected period: intake, active compressor, aftercooler and separator, dryer and filters, receiver, distribution route, local treatment and final connection. Record bypasses, cross-connections, alternative supplies and the direction of flow where it can change. A current drawing may not represent the arrangement that existed during temporary work.

Select diagnostic points that distinguish competing explanations. A treatment-outlet result, branch result and terminal result can help identify where a change emerges, provided their methods, states and timing are comparable. A difference between non-simultaneous samples may reflect a changing source rather than contamination introduced between the points.

Keep the contamination mechanism explicit. A component can introduce contamination, fail to remove it, store it and later release it, or create a misleading indication through its sampling connection. These mechanisms lead to different investigations. A dirty-looking receiver does not by itself quantify contamination delivered to the process; a clean-looking filter does not establish that its element or seals perform as required.

The following pathway matrix is a hypothesis aid. None of the patterns is a stand-alone verdict. Use it to choose observations and preserve contradictory evidence, including cases where an upstream result is acceptable but the terminal connection remains adverse.

11. Investigate moisture with pressure and physical state intact

First identify whether the concern is water vapour, an indicated pressure dew point, entrained liquid or water accumulated at a low point. These observations are related but not interchangeable. Record the measurement and process pressures. An atmospheric dew-point value cannot be directly judged against a line-pressure requirement merely because both are expressed in degrees Celsius.

For the measurement branch of the investigation, examine leakage, line materials, dead volume, stagnant flow and condensation in the sampling path. Vaisala's sampling guidance identifies these as influences on moisture representativeness and response. Use the selected instrument's actual requirements; do not transfer a generic flow or flushing time into every test. [R1]

For the system branch, compare aftercooler condition, drainage, dryer duty, regeneration or refrigeration state, bypass history and downstream cooling. A wet final connection may involve retained condensate after an earlier upset even when the current dryer outlet appears satisfactory. A single dry reading does not demonstrate that every low point or intermittently used branch has recovered.

Keep hypotheses conditional. High inlet temperature may challenge treatment, but the selected dryer data and operating evidence determine its significance. A dew-point sensor disagreement may indicate sampling or instrument issues, yet it does not cancel independent evidence of liquid water. Retain both until the investigation explains their different locations, conditions or measurement capabilities. Required corrective work and release evidence must address the actual moisture mechanism.

12. Investigate particles without assuming their composition

A particle result requires its size-band definition, counting basis, sampled volume and method. Preserve whether channels are cumulative or separate bands. Adding overlapping cumulative channels can count the same population repeatedly. A change in the selected band or instrument settings can create an apparent trend even when the underlying air has not been assessed consistently.

Check the normal path for disturbed piping, corrosion products, desiccant dust, filter damage, bypassed filtration, new fittings and local hose deterioration. These are candidate mechanisms, not conclusions. Review the chronology and seek corroborating evidence where a specific material or component is implicated. A particle count alone does not identify chemical composition or distinguish every solid and liquid contribution.

Examine the sampling assembly for particles introduced during connection, unsuitable tubing, contamination retained from earlier work and collection outside the accepted instrument range. Review associated controls according to the method. A failed control may compromise a sample batch; it is not automatically a quantity to subtract until the result passes.

Where a localized hypothesis is tested, preserve the normal production configuration in the acceptance measurement. Testing through a new clean sampling hose can help isolate an assembly issue, but it does not qualify the old process hose still in service. Any temporary removal of local treatment or fittings must have an approved purpose, and the final verification must describe the configuration actually returned to use.

13. Investigate oil by fraction and by the complete delivery path

Identify what the report measured: liquid oil, aerosol, vapour or an explicitly defined total. Compare only compatible quantities and retain any fractions not assessed. An aerosol-only result is not a complete total-oil assessment when the requirement includes vapour. When fractions are combined, check that they do not overlap and that their locations, conditions and volume bases are compatible.

Potential pathways include intake contamination, the compressor and its separation system, treatment performance, an exhausted adsorption stage, a bypass, residual material in receivers or distribution, a lubricator connected to a sensitive branch, and temporary hoses. Do not select the compressor as the cause solely because the event involves oil. Oil-free compression also does not establish uncontaminated downstream pipework or final connections.

Review relevant maintenance and configuration changes. A replacement element, altered lubricant, disturbed separator, wrong connection or temporary supply can be significant, but causation needs more than a matching date. Preserve part identities, approved service records and as-found observations without inventing allowable carryover, service life or replacement criteria.

A lower later result may demonstrate improvement under the later sampled condition. It does not identify whether the initial oil arose in the air system or in handling, and it does not prove that retained contamination has been removed from every affected branch. Recovery planning should address the demonstrated pathway and the actual user boundary, with method-specific evidence for the fractions that matter.

14. Keep viable and named process contaminants within specialist scope

A viable microbiological result is a method-dependent observation. Retain sample identity, accepted volume, collection arrangement, handling history, laboratory qualifiers and the approved process criterion. Do not infer viable contamination from an ordinary particle count, or infer sterility from a report with no recovered colonies.

The investigation should ask whether the evidence supports the claimed location and exposure, whether the collection and laboratory controls were suitable and whether relevant system changes occurred. Distinguish possible contamination introduced during sampling from possible contamination present in the utility. Neither explanation should be adopted merely because it is easier to correct.

For named chemical contaminants, odours or process-specific targets, confirm that the selected method actually addresses the substance of concern. A satisfactory oil-vapour measurement is not a universal gas analysis. A treatment component installed for one target may not control a different contaminant, and a generic compressor quality certificate cannot substitute for the specified assessment.

This workbook does not prescribe culture conditions, disinfection chemistry, toxicological limits or product-release decisions. The responsible laboratory and process specialists determine the appropriate methods, handling, remediation and acceptance. Record their decisions and limitations in the same incident structure so specialist evidence remains connected to the operating chronology rather than becoming an unexplained attachment.

15. Use operating events to test a mechanism, not merely to tell a story

Create an event comparison that pairs quality evidence with relevant pressure, flow, temperature, active source, treatment state and production demand. Their observations can help explain the duty at the time of the concern, but neither a pressure dip nor a high temperature proves a particular contaminant was present.

Check whether the proposed mechanism predicts the observed sequence. If a dryer changeover is suspected, determine whether the relevant moisture response occurs consistently in that state, whether sample-system lag could alter the apparent timing and whether the same pattern occurs at a meaningful downstream point. An isolated coincidence is weaker than a reproducible, physically plausible relationship with suitable measurement evidence.

Look for alternative explanations. A peak-demand period can change treatment loading, flow paths and sensor response simultaneously. A filter service can coincide with a new sampling assembly. A temporary compressor can be connected through an old contaminated hose. Treat these as competing or interacting mechanisms rather than choosing one convenient label.

Preserve events that contradict the leading explanation. A recurrence when the suspected machine is not connected is important evidence. So is a failure restricted to one sample assembly while comparable process-path evidence remains acceptable. The conclusion should explain the observations that do not fit, or retain them as unresolved limitations that affect confidence and follow-up.

16. Build a hypothesis register with evidence for and against each cause

For every material hypothesis, record the proposed source, transport path, affected point and failure mechanism. Add supporting observations, contradictory observations, missing evidence, the proposed discriminating check and its owner. A label such as “dryer problem” is too broad to direct a useful investigation; a hypothesis about an actual bypass path or switching malfunction is more testable.

Separate a direct observation from an interpretation. A valve indication showing closed is an observation about an indication; it is not always proof of physical isolation. A replaced element is evidence of work; it is not proof that the previous element caused the excursion. The register should make those distinctions easy to follow.

Choose tests that reduce uncertainty without compromising service or sampling integrity. Where safe and authorised, a suitable paired-point measurement, configuration inspection or instrument reference check may distinguish alternatives. Do not deliberately expose sensitive production to suspected contamination merely to reproduce the complaint.

Classify conclusions honestly: supported primary cause, contributing condition, sampling artefact supported by evidence, or unresolved. Not every incident ends with one exclusive cause. Preserve the evidence needed to revisit the conclusion if the event returns.

17. Plan repeat testing around the unanswered question

A repeat test should state its purpose before the result is known. It may investigate the original sample, assess the current utility condition, locate a pathway or verify a corrective action. These purposes are not interchangeable. Reanalysis of retained material, where the method permits it, addresses a different question from collecting new air after maintenance.

Specify what result would support or contradict the hypothesis and what would remain unresolved. Select the extent of repeat testing from the actual event and procedure, not from a universal number of samples or a desire to obtain three consecutive passes.

Retain every valid result and each justified validity disposition. Do not average independent failing and passing occasions simply to hide an excursion, or discard an earlier result because a later sample is favourable. Where averaging is part of the approved method or requirement, retain that basis and the individual observations.

A new sample usually describes a new collection period. State which earlier questions it cannot answer. Successful post-repair evidence can support recovery while the earlier product-impact review remains open. The repeat-test record should therefore connect to a specific investigation or release obligation, rather than allowing an undifferentiated “retest passed” statement to close the whole incident.

18. Use calculations that preserve the measurement basis

Simple calculations help reconcile evidence only when their inputs are valid for the stated purpose. For sample flow q in litres per minute and accepted collection time t in minutes, V = q × t / 1,000 gives cubic metres on the same flow basis. For changing verified intervals, use V = Σ(qᵢ × Δtᵢ)/1,000. Exclude pauses from collected volume; do not treat a missing flow record as measured zero or silently count purge time.

Where the method supports count-to-volume calculation, C = N/V gives number concentration for the named particle band. Retain raw count, accepted volume, instrument adjustments and report qualifier. Changing a reference basis does not correct particle losses or contamination introduced in the sampling path. If the instrument already reports reference volume, do not convert it again.

For two compatible, non-overlapping measured fractions A and B, C_total = A + B. Under a suitable uncertainty model, u_total² = u_A² + u_B² + 2ρu_Au_B; expanded uncertainty is U = k × u_total. Record correlation and the coverage basis. These uncertainty principles follow NIST guidance; the worked values are fictional. An incomplete fraction set or invalid sample remains incomplete regardless of the calculation. [R2, R3]

Keep interval-duration metrics separate from sample counts. Report data coverage and observed exceedance duration using valid represented intervals. Do not infer continuous performance from sparse samples or call an integrated dew-point difference a contamination dose. The formulas below are diagnostic bookkeeping, not a substitute for a method-specific exposure or product-risk model.

19. Reconcile potential process exposure without claiming proven contamination

Identify which users were connected to each potentially affected air path and during which operating windows. Record whether the air directly contacted product, operated an isolated actuator or served another defined function. Similar room location alone does not establish the same exposure route. The process owner should confirm the actual interface and relevant production records.

Use batch identifiers, line states, timestamps and traceability records to identify the population requiring review. Where only a production rate is available, any time-based estimate must disclose its assumptions. Uniform production over a shift is an arithmetic assumption, not a measurement of the exact number of affected units. Whole batches may need review even when a screening estimate covers only part of their production interval.

Separate potentially exposed production, proven unacceptable production and production released through an authorised assessment. A utility excursion does not automatically prove product contamination; a favourable finished-product result does not automatically resolve the utility fault. Each decision uses the applicable evidence and process requirements.

Do not calculate contaminant mass delivered to product merely by multiplying one grab-sample concentration by total station air volume. That would require representative concentration over time, a compatible volume basis, the actual branch and transfer pathway, and an appropriate fate or exposure model. Without those inputs, retain a bounded screening assessment and identify the further evidence needed.

20. Separate correction, recovery, product disposition and effectiveness

A corrective-action record should identify the verified condition being addressed, the approved work, responsible person, completion evidence and the verification that will test its effect. Replacing several parts at once may restore service, but it limits attribution. Describe the combined intervention rather than selecting one part as the proven cause without evidence.

Remediation must be appropriate to materials, equipment and process requirements. This workbook does not supply universal cleaning, flushing or disinfection instructions. Document the applicable procedure, configuration changes, removed component evidence and relevant as-left condition. Preserve any effect on pressure, safety, treatment duty or other users.

Define recovery at the required points and operating states. An isolated compressor sample cannot release a downstream network whose contamination pathway remains unresolved. Include event-sensitive conditions where relevant, such as restart, changeover or peak duty, without inventing one test duration for every system. Record which obligations are met, which remain open and what interim controls apply.

Close product disposition separately through the authorised quality or process owner. Then plan effectiveness review against the observed failure mechanism and recurrence risk. An immediate pass supports a current condition, not indefinite persistence. Where the primary cause remains unresolved, the closeout must retain that limitation, monitoring responsibilities and a clear trigger for reopening the incident.

21. Worked case A — incomplete oil evidence followed by recovery evidence

Fictional case; all values are illustrative and are not recommended purity limits. An approved requirement in this example is total oil no greater than 0.010 mg/m³ at a terminal connection. The laboratory reports liquid-plus-aerosol oil of 0.006 mg/m³ and oil vapour below 0.008 mg/m³. The two fractions are stated to be non-overlapping and on a compatible basis.

Replacing the vapour result with zero would create an unjustified total of 0.006 mg/m³. Treating the stated threshold only as a screening upper bound gives a possible range from 0.006 to below 0.014 mg/m³ before measurement uncertainty. That range crosses the requirement. It is not a confidence interval or a complete acceptance result; it demonstrates why the original report cannot establish compliance by zero substitution. The laboratory's qualifier interpretation remains controlling.

After a separately documented intervention, new representative samples yield 0.004 mg/m³ liquid-plus-aerosol and 0.003 mg/m³ vapour. The example assumes complete fraction coverage and suitable valid methods. Standard uncertainties are 0.0006 and 0.0008 mg/m³, with independence justified for this synthetic calculation and k = 2 selected in the stated plan.

The later result supports the sampled post-intervention condition only. It does not establish the earlier onset, explain the original source or release earlier production. The closeout record keeps the original evidence, new sample IDs, intervention details and remaining exposure decision linked. A different correlation assumption would change uncertainty and must be assessed rather than hidden in a default. The exact decision rule is part of this fictional plan, not a general instruction to every laboratory.

22. Worked case B — a sample pause changes particle concentration

Fictional case; the particle band and acceptance limit belong only to this example. A terminal sample record shows 25 minutes of elapsed collection activity at a reported 2 L/min on the agreed volume basis. The event log documents a five-minute instrument pause, and confirms that no sample flow or particle collection occurred during that pause. The valid active collection therefore lasts 20 minutes, not 25.

The accepted count is 12 particles in the specified non-overlapping size band. No additional dilution or instrument correction applies in this simplified case. The actual accepted volume is 2 × 20 / 1,000 = 0.040 m³, and concentration is 12 / 0.040 = 300 particles/m³. Incorrectly using the full elapsed time would yield 0.050 m³ and 240 particles/m³.

With an illustrative simple-comparison upper limit of 250 particles/m³, the correct arithmetic lies above the limit; the incorrect volume would appear below it. This is not a rounding issue. It changes the interpretation because the denominator includes time during which nothing was sampled. If the instrument's own validated totaliser already reports 0.040 m³, retain that evidence and do not apply the pause deduction a second time.

Suppose the laboratory also flags an associated sampling-control concern. The workbook retains the 300 particles/m³ calculation but marks its evidential use under review. It does not automatically subtract the control, label the process acceptable or erase the adverse observation. Appropriate protective decisions remain with the process owner while a valid follow-up plan addresses both sample integrity and the required terminal boundary. This case demonstrates a volume correction; it does not identify the physical particle source.

23. Worked case C — observed moisture duration and missing evidence

Fictional case; assume the stated logger intervals have been reviewed as adequately representing their assigned classifications. For a 60-minute observation window, an approved moisture requirement is pressure dew point no higher than −40 °C at the specified line pressure. The record contains 30 valid minutes at −45 °C, 10 valid minutes at −30 °C, 5 valid minutes at −25 °C and 15 missing minutes. Pressure basis and channel validity are established for the observed intervals.

Valid represented duration is 45 minutes, so coverage is 45/60 × 100 = 75%. Observed exceedance duration is 15 minutes. It represents 15/45 × 100 = 33.33% of valid time and 15/60 × 100 = 25% of the entire window. These percentages answer different questions and should both retain their denominators.

If the 15 missing minutes were all non-exceeding, the classification-based exceedance share of the complete window would be 25%; if all were exceeding, it would be 50%. These are missing-data bounds under the stated interval classification, not statistical confidence limits. They do not prove what occurred in the gap or capture fluctuations that the measurement method could not resolve.

The excess integral is a trend descriptor only. Dew point is not a linear measure of water concentration, and 175 °C·min is not a water mass or product-contamination dose. The incident timeline must retain the missing interval and the actual air-use states. A later dry reading cannot establish that the gap was dry. For sparse spot readings rather than adequately represented intervals, these duration calculations would require a separate defensible time-assignment model or remain unavailable.

24. Worked case D — overlapping assessment windows and production review

Fictional case; the times are screening assumptions supported by the synthetic incident record, not actual customer observations. Two potentially relevant pathway events produce assessment windows of 08:10–08:50 and 08:30–09:00 for the same production line. The durations are 40 and 30 minutes, but the overlap from 08:30–08:50 is 20 minutes. The union is therefore 40 + 30 − 20 = 50 minutes, covering 08:10–09:00.

Lot A ran from 08:00–08:30 and produced 600 units; Lot B ran from 08:30–09:00 and produced 900 units. For an explicitly assumed uniform production rate within each lot, the time-overlap estimate is 20/30 × 600 = 400 units from Lot A and 30/30 × 900 = 900 units from Lot B. The combined estimate is 1,300 potentially relevant units, not 1,500 and not an overlap-double-counted total.

That arithmetic does not establish which individual units had relevant contact with the air, whether the contaminant reached product or whether any unit is unacceptable. The actual traceability and process-contact records govern the assessment. If partial-lot traceability is unavailable, the quality owner may need to review both complete lots; the workbook must not automatically release the remaining 200 units of Lot A.

Suppose treatment is restored at 09:10 and the first suitable terminal recovery evidence is collected later. Keep those times separate. Equipment restoration is not automatically the end of possible downstream exposure, and a favourable sample does not retrospectively narrow the earlier window without evidence. The report should state the current utility status, the original assessment windows, the basis of the production estimate and the distinct authorised lot decisions.

25. Three editable investigation worksheets

The following tables are reusable Word worksheets. Replace the bracketed prompts, duplicate rows as required and retain a versioned copy for each incident. They contain no automatic calculations or signatures. Blank means unresolved, not zero, not acceptable and not not-applicable. Record an explicit reason when a field does not apply. Keep sensitive source files and customer information in the approved private evidence location.

Worksheet A — incident, requirement and containment definition

Complete Worksheet A before allowing a headline result to drive the entire investigation. The incident scope can expand as evidence develops, but retain the original scope and explain the change. A narrower final scope requires supporting route or time evidence; it should not arise merely because one later sample is acceptable.

Worksheet B — sample and event evidence register

One row should have one identifiable record type. Link multiple contaminant results to the same sample rather than duplicating its collection event. Preserve original timestamps and any correction references. Where a sample is invalid for one use but useful for another, name both uses and retain the explanation. Do not collapse the evidence register into a list containing only favourable results.

Worksheet C — hypothesis, correction and verification closeout

The final summary should answer four separate questions: is the current utility condition demonstrated, what happened to potentially affected production, how well is the cause supported, and what remains due? Do not hide an unresolved cause inside a completed work order. Do not close the whole incident because three of four obligations are complete. The authorised closeout should reference the version of each supporting record.

26. Engineering questions and answers

1. Does one failing sample prove that the compressor is the cause?

No. The result must first be linked to a valid method, point, condition and requirement. A supported excursion may originate in generation, treatment, distribution, a final connection or the sampling process. The workbook preserves the adverse evidence while testing these pathways instead of converting the sample location into a cause statement.

2. Does an invalid sample mean that the process air was acceptable?

No. It means the sample cannot support the specified conclusion in its present form. The actual air condition may remain unknown. Document the validity defect, preserve the record and arrange appropriate evidence while the responsible process owner determines whether protective measures remain necessary.

3. Can a successful retest replace the original adverse result?

Not merely because it is favourable. New sampling usually represents a later condition, potentially after the system or sample assembly changed. Preserve the original and any evidenced invalidation. State whether the repeat addresses current recovery, an analytical question or an unresolved pathway; it does not automatically answer all three.

4. Can an upstream pass release a downstream user?

Only evidence applicable to the defined requirement can close that obligation. An upstream result can help locate contamination, but it does not include downstream receivers, pipework, treatment, hoses or connections. Any accepted representation of another point needs an explicit technical basis, not an assumption that all points share the same air.

5. Is a less-than result the same as zero?

No. Retain the qualifier and its threshold. A result below a reporting threshold that exceeds the requirement may remain inconclusive. Likewise, one unmeasured oil fraction cannot be replaced by zero to create a favourable total. The laboratory's stated interpretation and the agreed decision rule govern the evidence use.

6. Does a dryer-panel dew-point value describe every terminal point?

No. It describes the indication at its own measurement boundary and conditions. Confirm sensor and sample pressure, method and representativeness. Distribution conditions and retained liquid may differ downstream. Keep the panel trend as useful operating context without promoting it into an independent terminal acceptance test.

7. How many passing samples are required before return to service?

There is no universal count in this workbook. The authorised recovery plan should specify points, contaminants, operating states, methods and the extent of evidence needed for the particular concern. The plan may need to address a switching or restart event that several convenient quiet-period samples would miss.

8. Can production between the last pass and first failure be called contaminated?

Not from those two snapshots alone. That interval may be a justified outer assessment window, but exact onset and product contact remain separate questions. Preserve the uncertainty and use actual operating and traceability records. Potentially affected, proven unacceptable and authorised-for-release populations must remain distinct.

9. Does replacing the suspected component complete the investigation?

No. Replacement is action evidence. Recovery must be demonstrated at the relevant boundary, prior production needs its own disposition, and the cause should be supported or explicitly unresolved. Follow-up should test whether the corrective action remains effective under conditions relevant to the observed mechanism.

10. What information makes an enquiry immediately useful?

Provide the original result and qualifier, sample plan and point identity, requirement revision, event chronology, treatment and distribution arrangement, recent work and the protective actions already taken. Identify the decision needed next. Reddy Compressors can use that defined brief to scope the required engineering and measurement support without claiming that the supplied records already prove a cause.