Thermal Trend Diagnostic Workbook

Thermal TrendDiagnostic Workbook

Compare heat generation, cooling conditions and measured temperature trends before selecting the cause or the repair.

1. Engineering answer: a hotter compressor is a question, not a diagnosis

A thermal trend diagnostic workbook compares compressor temperatures with the operating conditions that produced them. It brings together load, pressure, speed, local cooling-air or cooling-water temperature, cooling flow, equipment state and measurement quality. Its purpose is to distinguish an expected change in duty from a restriction in the heat-removal path, an internal equipment problem or a misleading temperature indication. The useful output is a supported investigation direction and a repeatable verification plan—not simply a red temperature tile.

A higher airend reading does not, by itself, identify a dirty cooler, defective thermostat, damaged bearing or inadequate ventilation.

Begin with the actual complaint. Did a protection trip occur, did a recorded temperature rise above an accepted baseline, did the dryer receive hotter inlet air, or did an infrared survey identify an apparent hot surface? These are different questions. They require different measurement points, comparison conditions and response owners. The workbook must retain those distinctions even when several symptoms occur together.

A diagnostic comparison never replaces the equipment's protective functions. Existing manufacturer limits and authorised site procedures govern the immediate response. A normalised temperature or favourable average cannot cancel an actual overtemperature event. Where the evidence remains incomplete, report the unresolved cause and the next discriminating observation rather than recommending a component replacement from one trace.

2. What this workbook owns—and what remains separate

It is suitable for recurring high-temperature events, performance changes after maintenance, cooler deterioration, heat-rejection complaints and unexplained differences between similar operating periods.

It does not replace an equipment manual, a pressure-system inspection, an electrical diagnostic procedure or a detailed exchanger design. It does not calculate remaining bearing life, establish lubricant condemnation limits, approve a higher trip setting or certify safe continued operation. The calculations are diagnostic tools whose applicability depends on the stated measurements and assumptions.

The three editable worksheets later in this document are designed for manual use in Word. The implementation appendix specifies how a future digital workbook should behave. No automatic Word calculations, live sensor connection or deployed dashboard are claimed. Numerical examples are fictional and are labelled accordingly; their limits and settings are not recommended operating values.

3. Define the complaint and freeze the relevant configuration

Create a campaign record with the site, compressor tag, exact model, serial number, cooling arrangement, installed auxiliaries and affected service. Record when the complaint began, the frequency of occurrence and its consequence. An alarm without a production interruption is not the same event as a protective shutdown, and a shutdown is not proof that the initiating sensor was correct.

Preserve the configuration that existed during the event. Relevant records include pressure settings, load-control mode, speed, enclosure condition, fan operation, water-system arrangement, lubricant identity, recent cooler work and any temporary ventilation. Date each change. A trace taken after the service team altered a valve or removed an obstruction must not be presented as the original condition.

Record the criterion separately from the observation. The criterion may be a manufacturer operating limit, an accepted commissioning baseline, a project investigation threshold or an unresolved requirement. Store its source, revision, measurement point and applicable state. Do not substitute an internet temperature recommendation for the exact equipment requirement.

The campaign question should be narrow enough to test: “Why did the aftercooler outlet approach increase during the same loaded duty?” is more useful than “Why is the compressor hot?” State what decision the evidence will support, which conditions cannot safely be reproduced and what remains outside scope.

4. Map the heat path and give every temperature a physical meaning

A channel called TEMP1 is not a sufficient engineering definition. Identify the physical point, medium, sensor method, unit, response and relationship to the cooling circuit. The compressor inlet, cooler cooling-air inlet, room inlet and outdoor ambient can be at different temperatures. Likewise, oil entering the cooler, oil leaving it and mixed oil returning to the element describe different functions.

Use a current schematic to show which readings form a legitimate pair. Shared coolers, mixing valves and parallel water circuits require special attention. A temperature downstream of a mixing junction cannot automatically be assigned to one upstream exchanger. Record inaccessible points as gaps; do not replace them with nearby temperatures merely because those channels exist.

5. Establish a trustworthy measurement chain

For every channel, record instrument identity, range, calibration or verification evidence, installation method and source of the timestamp. Distinguish an embedded sensor, a contact probe, an infrared surface measurement and a controller value derived from another signal. Their readings are not interchangeable simply because they use degrees Celsius.

Repeatable placement matters. A contact probe on a pipe surface responds to the local wall, thermal contact and surrounding air; it is not automatically the bulk-fluid temperature. An embedded probe can have installation and response limitations of its own. Preserve the method used for the baseline and identify any change before comparing a later campaign.

Check a suspect channel against an appropriate independent reference where safe and practicable. A common calibration bias can affect two instruments together, so agreement alone is not proof of accuracy. A step change after sensor replacement, a frozen value or an implausible response to a known event should trigger a measurement review before mechanical conclusions are drawn.

Separate the controller's protection channel from any diagnostic channel. The diagnostic system may identify disagreement, but it must not override, suppress or reconfigure protection. Inspection near hot surfaces, rotating equipment, pressurised connections or electrical enclosures requires the site's competent personnel and safe-access arrangements. The workbook is a record of evidence, not permission to expose hazards to obtain another reading.

6. Align clocks without erasing genuine thermal delay

Synchronise temperature, pressure, power, flow and machine-state records to one declared timeline. Preserve original timestamps and document any correction. Record whether a value is instantaneous, an interval mean, an interval maximum or an accumulated quantity. A one-minute maximum must not be plotted as though it were an instantaneous reading taken at the minute boundary.

Thermal response also has physical delay. A load change can precede a temperature response because metal, fluid, piping and sensors store heat or respond gradually. That delay is not necessarily a clock error. Do not move a temperature series until it visually aligns with power and then call the alignment a verified timestamp correction.

Keep two separate concepts: an evidence-based clock correction, and a diagnostic lag between a cause candidate and a thermal response. The first changes the analysis time reference; the second is a finding to be interpreted. Sensor response, fluid transport, mixing and control action can all contribute to an observed lag, so correlation alone cannot identify a unique mechanism.

Select acquisition resolution to capture the event and its recovery. A long averaging interval can hide an excursion; excessively frequent records can create large datasets without improving a slow sensor's response. Document the measurement capability rather than prescribing one universal sampling interval. Retain synchronisation uncertainty when the relative timing matters to the diagnosis.

7. Compare operating states before comparing temperatures

Define states using observable conditions: stopped, warming up, stable loaded operation, unloaded operation, variable-speed bands, fan cycling, water-control state and cooldown. Add pressure and production context where they materially affect duty. A “running” flag alone does not establish a comparable heat load.

For a stable comparison, identify a window in which the relevant operating and cooling conditions are sufficiently steady for the intended calculation. State how that window was selected. A low temperature immediately after start does not demonstrate adequate cooling at sustained duty, and a hotter observation during a different pressure or speed state is not automatically deterioration.

Do not remove transient records from the campaign. Keep them in the event view while excluding them, with a reason, from calculations that assume approximate steady operation. An event investigation may need precisely the startup and shutdown records that a steady-state cooler comparison excludes.

For multiple compressors, capture the concurrent fleet configuration. Another package starting can increase room heat or change cooling-water allocation. A comparison between one-machine and three-machine operation may reveal a shared utility constraint rather than an internal defect in the machine whose alarm appears first. Record the state of supporting equipment as well as the compressor being investigated.

8. Screen units, missing intervals and suspicious values

Retain original observations in a read-only source layer and perform conversions in a separate analysis view. Temperature unit is mandatory. Convert absolute Fahrenheit readings with T_C = (T_F − 32) × 5/9. Convert a Fahrenheit temperature difference with ΔT_K = ΔT_F × 5/9; do not subtract 32 from a difference. Celsius differences and kelvin differences have the same numerical magnitude.

Missing temperatures are not zero. Use explicit codes for unavailable, invalid, estimated and excluded values. A confirmed zero-degree observation can be valid. A channel that records zero during communication loss requires a different disposition. Preserve the instrument's fault and range-status information where available.

Check duplicate records using channel identity, source record and timestamp, not timestamp alone across every instrument. Review flatlines, abrupt offsets, unit changes and time reversals. An outlier should remain visible until its measurement and operating context have been assessed. High-temperature events must not disappear because an automatic smoothing rule considered them inconvenient.

For paired calculations, use the common valid observation window. Do not subtract a one-hour average cooling-water inlet from a ten-minute maximum outlet and call the difference a measured approach. Where a gap intersects the event, record which metrics become uncertain and whether the missing interval prevents a conclusion about the peak or duration.

9. Calculate temperature rise, cooler approach and available margin

Temperature differences become useful when the two points are named. Use the following definitions only for the stated boundaries; different equipment may require another documented comparison.

ΔT_component = T_component − T_local_reference

A_cooler = T_hot_stream_out − T_cooling_medium_in

ΔT_cooling_water = T_water_out − T_water_in

R_inlet = T_cooler_air_in − T_room_fresh_air_reference

M_high = T_applicable_high_limit − T_measured

A positive R_inlet indicates that the measured cooler inlet is warmer than the selected fresh-air reference. It does not quantify the percentage of recirculated air: other local heat sources and measurement positions can contribute. The cooler approach is likewise an observed temperature difference, not a direct measurement of fouling thickness, efficiency or remaining cooler life.

Only calculate M_high when the limit applies to the same physical point, operating state and measurement convention. A positive margin means the value lies below that stated upper limit; it is not a declaration that the entire compressor is safe. A lower-temperature criterion, when relevant, requires its own limit and sign convention.

Do not report temperature deterioration as a percentage of an arbitrary Celsius reading. A rise from 40°C to 50°C is 10 K; describing it as a 25% increase in absolute thermal state would depend on the chosen scale. Use temperature differences, state-matched residuals or an explicitly defined dimensionless engineering indicator.

10. Use time-weighted summaries and declare excursion conventions

For observations representing valid intervals, the time-weighted mean is T_mean = Σ(T_i × Δt_i) / ΣΔt_i. A row-count mean is suitable only when every accepted observation represents equal time. Sum the supported intervals, not the number of records, to determine coverage.

Coverage = 100 × valid represented duration / scheduled analysis duration.

Show scheduled, valid, estimated and unresolved durations separately. A valid-duration mean must not be labelled as the complete shift mean when part of the shift is missing. A maximum is the maximum observed under the instrument and logging method, not proof of the true instantaneous maximum during an unobserved interval.

Define the excursion rule before counting events. For the examples in this workbook, a strict upper exceedance means T > L; equality is not an exceedance. A project may require another convention, including persistence time or hysteresis. Store that convention as an explicit input, not an undocumented software default.

For interval-representative values, observed time above L is ΣΔt_i over accepted intervals satisfying the rule. A descriptive excess-temperature exposure is X = Σmax(T_i − L, 0) × Δt_i, expressed in K·min when time is in minutes. X is an event-summary measure, not a validated damage, lubricant-life or remaining-life model.

Where values are instantaneous samples, interval membership and threshold crossing require a declared method. A supported linear interpolation can estimate a crossing between adjacent good samples, but not across a trip, state change or long gap without justification. Keep observed and estimated event duration separate. Missing data must not join two episodes into one “continuous” event or prove that temperature remained below the limit.

11. Build a state-matched baseline without adjusting away the problem

A useful baseline represents accepted operation at known configuration, load, pressure and cooling conditions. Preserve its dataset, date, point definitions and acceptance basis. It need not be the coldest recorded period. Selecting a favourable cold-start trace would make normal loaded operation appear defective.

Begin with direct matched-state comparisons. For example, compare aftercooler approach at similar compressed-air duty, inlet-air condition, cooling-air flow and fan state. If the raw outlet temperature rises while the approach remains similar, a warmer cooling sink may explain part of the change. That is a diagnostic interpretation requiring the other conditions to be checked—not a universal correction rule.

Where sufficient representative data exist, a reviewed model can estimate expected temperature as a function of independent operating variables. Define the residual e_i = T_measured,i − T_predicted,i. Record the model's training range, variables, validation and uncertainty. Do not fit a model solely to the abnormal period and then declare the abnormal temperature normal because the model learned it.

A model that includes a failed cooling-flow condition as an ordinary predictor may explain the reading without proving acceptable operation. Keep the operational question visible. Statistical adjustment must not hide a violated equipment limit, a deteriorating utility or a configuration change. If the conditions lie outside the baseline's supported range, mark the comparison as not comparable or model-limited rather than manufacturing a precise corrected result.

12. Diagnose air-cooled packages and shared-room effects

Separate the temperature of air entering the cooler from the temperature measured elsewhere in the room. Map the path by which fresh air reaches the package and warm air leaves it. Inspect relevant cooler condition, air-path restrictions, enclosure state, fan operation and recent ducting changes through competent site personnel.

Supplementary manufacturer guidance also identifies high ambient temperature, restricted ventilation, hot-air recirculation and maintenance condition as possible contributors to overheating. This supports the investigation categories; it does not provide a universal operating-temperature limit or demonstrate which cause exists at a particular plant. [R1]

A fan-running indication is not a measurement of airflow through the heat exchanger. Record speed or operating state where available, blocked screens, damaged shrouding, duct or damper changes and the applicable equipment data. Additional ducting can change resistance; a larger extraction fan cannot be assumed to correct every airflow problem.

Compare the implicated unit alone and under the relevant concurrent fleet states when such observation is safe and authorised. A neighbouring exhaust can heat the local inlet, while a shared outlet restriction can affect several packages. The machine showing the first alarm may simply have the least remaining margin.

Do not use open enclosure panels as a default permanent remedy. Preserve the designed airflow and guarding requirements. Any temporary mitigation needs an owner, approved operating boundary, expiry and permanent corrective plan. Diagnostic observation should not become an uncontrolled change to the package's cooling arrangement.

13. Diagnose water-cooled packages with flow as well as temperature

Record water inlet temperature, water outlet temperature, actual flow and the boundaries of the circuit. A pump-running lamp or an open-valve command does not prove that the required flow reaches the exchanger. Shared headers, strainers, control valves, bypasses and parallel paths can change the delivered cooling condition.

For a defined single-phase liquid stream under approximately steady conditions, the sensible heat-transfer estimate is:

Q_thermal,kW = ρ × q_v × c_p × (T_out − T_in) / 3600

Here ρ is liquid density in kg/m³, q_v is volumetric flow in m³/h, c_p is specific heat in kJ/(kg·K), and the temperature difference is in K. Use properties appropriate to the actual fluid, mixture and temperature. The factor 3600 converts the hourly mass-flow basis to seconds.

A larger water-temperature rise does not automatically mean more heat is being removed. Flow can fall enough that Q_thermal decreases even while ΔT increases. Conversely, lower ΔT can result from higher water flow rather than a loss of duty. Preserve flow and both temperatures instead of retaining only their difference.

Do not substitute package motor kW for heat measured in one water circuit. Energy can leave through other coolers, surfaces, discharged air or auxiliaries, and transient heat storage can matter. Where condensation or another phase change occurs in the hot stream, a simple dry-air sensible-heat estimate will not capture its complete duty. The measured single-phase water-side calculation may still be useful, but its measurement boundary must be explicit.

Before interpreting a changed heat-duty estimate, check load, pressure, bypass state, common utility conditions and instrument quality. Fouling is one hypothesis, not the only explanation. Cleaning methods, water chemistry, allowable flow and pressure limits remain equipment-specific; this workbook does not prescribe chemical cleaning or valve adjustments.

14. Interpret lubricant, mixing and compression-stage trends

In an oil-injected compressor, temperature observations belong to the actual lubricant and thermal-control arrangement. Record the lubricant identity, service history, relevant level or condition evidence and the points used for hot, cooled and mixed temperatures. Follow the equipment's approved checking procedure; an unsupported reading at the wrong state can mislead the diagnosis.

A thermostatic or bypass arrangement can change how much fluid passes through a cooler. Mixed-return temperature therefore cannot automatically be used as the cooler outlet. A changed reading may reflect control behaviour, a changed duty, a circuit restriction or a sensor problem. Preserve the schematic and valve-state evidence before selecting a cause.

Do not assume that colder operation is always preferable. The equipment's required thermal operating range remains relevant at low load and during repeated short runs. Any requirement must come from the identified machine and its actual conditions.

For multistage compressors, retain each stage's inlet condition, outlet temperature and pressure basis. A downstream stage can run hotter because its inlet conditions changed, even if its own mechanical condition did not. Compare the stage and intercooler path rather than diagnosing the final discharge sensor in isolation.

Where temperature change coincides with vibration, wear indications, lubricant findings or a performance change, link those records. The combination can strengthen a mechanical hypothesis, but neither a hot casing nor a calculated residual alone proves bearing damage or justifies an overhaul.

15. Use motor, drive and infrared evidence with the correct boundary

Motor housing, bearing housing, winding sensor and drive-panel air temperatures describe different locations and mechanisms. Match the method and operating state when comparing periods or similar machines. A drive fault log and panel-cooling condition may be more discriminating than the general compressor-room average. Electrical investigation belongs to qualified personnel under the site's safe-work process.

Infrared measurements need their own method record. FLIR explains that low-emissivity surfaces can reflect surrounding thermal radiation and give misleading apparent temperatures. Surface condition, emissivity and reflected-temperature treatment therefore matter; a polished metallic surface cannot be evaluated by assuming every bright region is internally hot. [R2]

Retain the radiometric file where available, the visible reference image, target location, surface condition, camera settings, viewing geometry and operating context. A screenshot with a colour palette alone is insufficient for quantitative comparison. The workbook should label the reading as a surface measurement rather than silently treating it as the temperature of internal fluid or a winding.

A surface treatment or reference target used for measurement must be compatible with the component and applied only under an approved safe method. Do not touch, tape or access live or hot equipment solely to improve an image. Where the method cannot support a reliable quantitative value, use the image as qualified screening evidence and define the confirmation needed.

16. Turn patterns into competing hypotheses

Use the following matrix as an investigation aid. The patterns are not automatic fault codes. More than one condition can be present, and every conclusion should identify the evidence that distinguishes it from credible alternatives.

A diagnostic record should say “consistent with reduced cooling flow; flow confirmation required” when that is all the evidence supports. Reserve “confirmed cause” for a mechanism supported by appropriate investigation and verification. Do not convert a pattern-matching score into a repair instruction without review.

17. Quantify uncertainty and keep it separate from acceptance tolerance

Temperature subtraction inherits uncertainty from both channels. For D = T_1 − T_2, first-order propagation gives u_D² = u_1² + u_2² − 2cov(T_1,T_2). With defensibly independent inputs, the covariance term is zero. Expanded uncertainty is U_D = k × u_D for the stated coverage factor and interpretation. NIST's propagation framework supports this treatment; it does not supply the instrument uncertainties for a particular campaign. [R3]

Convert an instrument certificate's expanded uncertainty to standard uncertainty using its stated factor before combining terms. Do not insert accuracy limits, resolution and standard deviations into one formula as though they were identical quantities. Calibration, installation, spatial variation and response effects need an appropriate budget without counting the same contribution twice.

When comparing two approach values, define J = A_after − A_before. Its uncertainty depends on both approaches and their covariance. Shared sensors or calibration effects can create correlation; independence must not be assumed solely because the readings were taken on different days. Unknown material correlation should remain a limitation or a sensitivity assessment.

An interval overlapping zero means the chosen uncertainty treatment does not resolve the sign of a small change at that stated coverage. It does not prove there was no change. Likewise, a diagnostic uncertainty interval is not automatically the contract's conformity decision rule.

Retain equipment limits and observed excursions separately. Do not add uncertainty to a trip threshold to obtain a more favourable operating margin. Acceptance decisions must use the agreed applicable rule. For heat-duty calculations, uncertainty in flow, fluid properties and temperature difference also matters; a very small ΔT can dominate the relative uncertainty of the result.

18. Worked example A: separating a warmer inlet from a larger approach

All figures in this example are fictional. Three steady observation windows are assumed to have comparable compressed-air duty, pressure, fan state and measurement configuration. The outlet sensor measures compressed air leaving an aftercooler; the reference measures cooling air entering that cooler.

In the second row, the absolute outlet has risen by 8 K, while the defined approach is unchanged. Under the stated comparison assumptions, the data do not show an additional approach deterioration. They direct attention to the hotter air entering the cooler and to whether the resulting outlet remains acceptable for the equipment and downstream treatment.

In the third row, 8 K of the 14 K outlet increase corresponds arithmetically to the cooling-inlet increase, and a further 6 K appears in the approach. This decomposition describes the observations; it is not a proof that 6 K was caused by fouling. Cooling airflow, bypass state, load matching, sensor behaviour and exchanger condition still need review.

Suppose an illustrative applicable outlet limit is 55°C. The measured margins are 15 K, 7 K and 1 K respectively. The unchanged approach in the second row does not restore the original absolute margin. A healthy-looking normalised indicator cannot erase the consequence of a hotter sink. Conversely, the 1 K nominal margin in the third row is not sufficient for an unrestricted acceptance decision without the relevant operating and uncertainty requirements.

19. Worked example B: larger water ΔT, lower heat transfer

This fictional comparison concerns one single-phase cooling-water circuit. Assume approximate steady operation, density 1,000 kg/m³ and specific heat 4.18 kJ/(kg·K) solely for the example. These rounded values are not a property specification for glycol, contaminated water or another fluid. Instrument readings are assumed valid and aligned.

Reference duty = 1,000 × 6.0 × 4.18 × 6 / 3600 = 41.80 kW. Later duty = 1,000 × 3.0 × 4.18 × 9 / 3600 = 31.35 kW. The measured water-temperature rise is 50% larger, but the calculated sensible heat transfer is 25% lower because flow has halved.

The example demonstrates why water ΔT alone is an incomplete diagnostic. It does not establish that the compressor generated exactly the same heat in both runs, or that the missing 10.45 kW accumulated entirely inside the machine. Other heat paths, transient storage and changed load could contribute. Confirm the full operating context before attributing the difference to reduced water delivery.

If this circuit alone sustained 41.80 kW for two hours under the assumptions, the transferred thermal energy would be 83.60 kWh. That is thermal energy crossing the measured boundary, not automatically electricity saved or useful recovered heat. A recovery claim would also require evidence that the heat was delivered to a useful demand and displaced another energy input.

20. Worked example C: time weighting, excursions and an unobserved interval

The fictional table contains interval-representative temperatures—not isolated spot readings. Each accepted value represents its full stated interval under the example's measurement convention. A 5-minute interval is unresolved and is not interpolated. The diagnostic upper threshold of 85°C is illustrative and uses the strict T > 85°C rule.

Scheduled duration is 40 minutes and valid represented duration is 35 minutes. Coverage is 35/40 × 100 = 87.5%. The time-weighted valid mean is (80 × 10 + 88 × 5 + 84 × 20)/35 = 83.4286°C, displayed as 83.43°C. The simple mean of the three available values is 84.00°C and does not represent their unequal durations.

Observed time above the threshold is 5 minutes, or 14.2857% of valid duration. It is also 12.5% of the scheduled window, but that figure must be labelled as observed exceedance divided by scheduled time—not the complete-window exceedance fraction. Without additional evidence, total exceedance could occupy between 5 and 10 minutes of the scheduled window under the interval convention, or 12.5% to 25%.

Observed excess-temperature exposure is (88 − 85) × 5 = 15 K·min. No finite upper value for the missing interval's excess exposure can be inferred without a justified bound on its temperature. The maximum observed interval value is 88°C; the missing interval prevents confirmation of the overall maximum. These limitations must remain visible even though most of the campaign is complete.

21. Worked example D: an apparent improvement smaller than its uncertainty

This fictional case compares the same defined cooler approach before and after work. Before work, hot-stream outlet is 44.0°C and cooling-medium inlet is 32.0°C, giving 12.0 K. After work, the corresponding values are 42.8°C and 31.0°C, giving 11.8 K. The observed approach change J is −0.2 K.

Assume, for this example only, that each of the four temperature observations has standard uncertainty 0.4 K and that all four contributions are independent. Each approach then has u_A = √(0.4² + 0.4²) = 0.5657 K. The change between the two approaches has u_J = √(0.5657² + 0.5657²) = 0.8 K.

With the explicitly selected coverage factor k = 2, U_J = 1.6 K. The reported comparison is therefore −0.2 K ± 1.6 K under that uncertainty treatment, giving an interval from −1.8 K to +1.4 K. The interval includes zero. The appropriate conclusion is that this comparison does not resolve an approach improvement—not that the work had no effect and not that it achieved a statistically established improvement.

The assumed independence is important. Reusing sensors can retain correlated calibration effects, while changed installation conditions can introduce new contributions. A real investigation must use its own uncertainty budget. A narrower uncertainty interval may become possible with better paired measurements and controlled conditions, but it must not be obtained by deleting inconvenient contributions.

The absolute outlet also decreased by 1.2 K, while the cooling inlet decreased by 1.0 K. That observation is useful, but it cannot be attributed entirely to the maintenance action. Record the work completed separately from the evidence needed to verify its thermal effect. Never convert a nominal 0.2 K change into a savings or remaining-life claim.

22. Editable worksheet A: campaign, points and comparison basis

Complete this record before selecting the analysis window. Replace the entry prompts with project information; do not treat the prompts as measured values. Keep one campaign identity and issue revision, with linked records for each compressor or exchanger under investigation.

Keep the source register alongside the campaign rather than embedding confidential customer data into public website examples. A campaign may be complete as a definition while measurement is still pending. Use a clear status so the record cannot be mistaken for an executed investigation.

23. Editable worksheet B: observation and event register

Create one record per channel observation or defined interval. Use stable IDs that survive sorting and export. Large time series can remain in their original structured format; the Word sheet can hold the selected diagnostic windows and references to the complete data.

The entry must identify whether the temperature is a fluid, surface or internal-sensor measurement. Record raw values even when corrected values are used for analysis. Keep a correction log for unit conversion, clock adjustment, channel reassignment or confirmed sensor offset.

An event note should describe what was observed before naming a cause. “Water flow decreased after the valve-state change” and “a valve fault caused overheating” are different statements. The second requires more evidence. For every calculated indicator, retain the exact observation IDs, time window, equation, assumptions and rounding convention.

24. Editable worksheet C: hypothesis, action and verification

Use this record to make the investigation actionable without overstating certainty. A completed inspection or repair does not automatically close the thermal finding. Closure requires the agreed evidence under a relevant operating condition.

For a temporary measure, add an expiry, permitted operating conditions and the permanent-action reference. Do not convert a temporary fan, restricted load or altered schedule into a permanent engineering solution by leaving the record open indefinitely.

The final diagnostic statement should identify what is known, what remains uncertain and whether the action addressed the original complaint. When multiple changes were made together, attribute the observed result to the combined intervention unless the separate effects were measured or otherwise defensibly established.

25. Verify the thermal result and preserve a useful operating record

Repeat the relevant measurements after work using the same or reconciled points, methods and operating context. Compare absolute temperatures, defined approaches, cooling-flow evidence and event behaviour. Check the original service consequence: an improved average is insufficient if the machine still trips during the same production sequence.

Verification should cover the limiting credible duty that the agreed scope requires. A short quiet-shift run cannot prove hot-weather or multi-compressor performance. Where those conditions cannot be observed, state the limit of the verification and the agreed follow-up. Do not erase the original baseline; issue a new accepted reference only when the change and its evidence justify it.

Keep thermal performance, pressure integrity, air quality and electrical condition as separate acceptance subjects. A cooler that transfers heat effectively can still have an integrity defect. Lower aftercooler outlet temperature does not certify downstream particle or oil condition. A favourable thermal trend also does not establish electricity savings without compatible energy and operating measurements.

Preserve the work record, raw and processed data, calculations, source criteria and review disposition. A maintenance team should be able to understand the next alarm without reconstructing the campaign from email history.

A concise project conclusion can read: “At the stated loaded duty, the cooler approach returned to the accepted comparison range after the recorded intervention. The water-flow cause is supported by the linked measurements. High-ambient operation was not observed and remains a follow-up condition.” This is example wording, not a completed customer result.

26. Frequently asked engineering questions

Why can a compressor run hotter without being mechanically damaged?

The cooling-air or cooling-water inlet may be warmer, load or pressure may have changed, cooling flow may be restricted, or a sensor may be giving a misleading reading. Compare the operating and cooling context before deciding that the airend or bearing is defective. A real thermal concern still needs the appropriate protective response even when its cause is not yet known.

Is subtracting ambient temperature enough to compare two periods?

Not by itself. The reference must be the relevant local cooling condition, and load, pressure, flow, fan or valve state and measurement method must be comparable. The resulting difference is a diagnostic indicator. It does not automatically normalise every thermal effect or demonstrate that an absolute equipment limit remains satisfied.

What exactly is cooler approach temperature?

In this workbook it is hot-stream outlet temperature minus the relevant cooling-medium inlet temperature. Both points must be named. For an air-cooled aftercooler, that means compressed air leaving the aftercooler and cooling air entering it. A different equipment method may define another approach; preserve that definition instead of comparing unlike values under one label.

Does a higher cooling-water outlet temperature mean the cooler is working better?

No. Evaluate inlet temperature, outlet temperature and measured flow together. The water can become hotter because less water carries the duty, while total heat removal may fall. The single-phase heat equation is useful when its assumptions and fluid properties apply, but a calculated duty still requires operating-context review before a cause is assigned.

Can thermal imaging replace installed temperature sensors?

It can provide complementary surface evidence, but it does not automatically measure internal fluid, winding or bearing temperature. Surface condition, reflected radiation, camera setup and safe access affect interpretation. Keep the measurement method visible and use appropriate confirmation where the diagnostic decision requires quantitative confidence.

How often should temperature be recorded?

Choose recording resolution and observation duration to capture the event, sensor response and relevant operating cycle. A short excursion requires a capable measurement chain; a slow drift requires a sufficiently representative history. There is no single interval that fits every compressor, cooling circuit and diagnostic question.

Can the workbook predict exactly when a high-temperature trip will occur?

No. A measured rate of rise describes the selected interval, not a guaranteed future trajectory. Controls, thermal storage, changing duty and cooling conditions can alter the response. Do not extrapolate a short slope into an assured time-to-trip or use that extrapolation as permission to continue operation.

Can missing temperature data be replaced with a straight line?

Only as an explicitly labelled estimate when the method is appropriate for the limited purpose. Do not interpolate through a trip, unknown state change or long gap merely to complete a graph. Preserve the original gap, show the estimated share and identify whether the missing interval prevents a conclusion about peak temperature or event duration.

Does a lower post-service temperature prove the repair succeeded?

Only when the comparison addresses the intended result under appropriate conditions. Cooler inlet temperature, load, pressure, fluid flow and measurement method may have changed. Retain the work record separately from the evidence of restored performance. A completed maintenance activity is not the same as a verified thermal outcome.

What information should be supplied for a thermal review?

Provide equipment and cooler identities, the complaint and alarm history, time-stamped temperatures, load and pressure records, cooling-air or water conditions, flow evidence, recent maintenance and the applicable equipment limits. Include the circuit or airflow arrangement and identify missing information. Reddy Compressors can use this brief to scope an evidence-led investigation without inventing the missing operating data.