How to assess an exhaust gas temperature data provider’s accuracy

Selecting an exhaust gas temperature data provider requires more than comparing sensor specifications or dashboard features. The meaningful question is whether the delivered measurement system remains accurate across load ranges, transient conditions, fuel types, and harsh operating environments—and whether its data can support defensible decisions on emissions, efficiency, maintenance, and operating risk.

For engines, gas turbines, boilers, marine propulsion systems, and standby power assets, exhaust gas temperature (EGT) is rarely a standalone value. It is used to identify cylinder imbalance, turbine hot spots, combustion instability, fouling, injector deterioration, aftertreatment risk, and deviations from expected heat rate. A provider whose equipment performs well in a laboratory but produces unstable or poorly traceable data in service can create false alarms, conceal genuine faults, and undermine trend analysis.

Accuracy assessment should therefore begin with the complete measurement chain rather than the temperature probe alone.

Define what “accurate” must mean in the intended application

A quoted sensor accuracy often refers only to a new sensing element under specified test conditions. It does not automatically include thermowell effects, extension cable errors, cold-junction compensation, transmitter conversion, data acquisition resolution, filtering, software scaling, or installation-related heat losses. In an operating exhaust system, those elements can collectively matter more than the tolerance stated on a datasheet.

The first task is to write an application-specific measurement requirement. This should identify:

• the expected temperature range during start-up, normal operation, overload, regeneration, and shutdown;
• the maximum allowable error at the operating temperatures that actually matter;
• required response time for load steps, trips, cylinder diagnostics, or turbine protection;
• acceptable drift between maintenance intervals;
• the use of the data: protection, control, condition monitoring, emissions support, performance optimization, or reporting;
• environmental exposure, including vibration, pressure pulsation, soot, salt, oil ash, corrosive condensates, and electromagnetic interference.

A provider may be suitable for long-term average stack monitoring but unsuitable for detecting a fast cylinder-to-cylinder deviation on a medium-speed marine engine. Likewise, a system intended to support gas-turbine exhaust spread calculations needs consistent channel-to-channel behavior, not merely acceptable absolute accuracy on each individual channel.

The distinction between absolute accuracy and relative accuracy is often overlooked. For combustion tuning or cylinder balancing, a repeatable and closely matched sensor array can be more valuable than one isolated sensor with an attractive absolute specification. For compliance-adjacent records, warranty claims, or performance acceptance tests, traceable absolute accuracy becomes more important.

Request an uncertainty budget, not a single accuracy claim

A credible exhaust gas temperature data provider should be able to explain how it derives total measurement uncertainty at the point of use. The answer should not be limited to “Class 1 thermocouple” or “±0.5% full scale.”

For thermocouple systems, the uncertainty budget normally includes the sensing-element tolerance, transmitter or input-module accuracy, reference-junction compensation, cable and connector effects, analog-to-digital conversion, calibration uncertainty, mounting effects, and long-term drift. If the provider supplies a digital sensor or smart transmitter, the same logic applies even when some functions are integrated into one device.

Thermocouple wire and sensor tolerances may be specified under standards such as IEC 60584-1 or ASTM E230/E230M. Those classifications are useful starting points, but they do not prove installed system performance. A Class 1 thermocouple can still deliver misleading data if the wrong compensating cable is installed, junction temperature is poorly managed, sensor immersion is inadequate, or the probe is exposed to radiative heating from an adjacent hot surface.

The provider should state whether its published figure is a maximum permissible error, a typical value, or an expanded uncertainty, and should identify the confidence level where uncertainty is quoted. These terms are not interchangeable. A low “typical” error has limited decision value if worst-case variation, calibration conditions, and ageing effects are not disclosed.

For multi-point systems, ask for a channel consistency specification. In a 16-cylinder engine or a gas turbine exhaust frame, the important question may be whether two channels can be compared reliably at the same temperature, not whether each channel individually meets a broad tolerance band.

Examine the sensing technology against the gas chemistry and duty cycle

Most EGT applications use thermocouples because they withstand high temperatures and have practical response times. However, sheath material, junction construction, insulation, diameter, and installation geometry all affect durability and measurement quality.

Mineral-insulated metal-sheathed probes are common, but the best material depends on temperature, sulfur content, chlorides, water vapor, pressure, and thermal cycling. A probe that survives clean natural-gas service may not have the same life in heavy-fuel marine duty, waste-derived fuels, biomass exhaust, or corrosive industrial processes.

Fuel transition is making this assessment more demanding. Hydrogen-capable engines and turbines may experience different combustion temperatures, water-vapor levels, flashback control strategies, and transient behavior. Ammonia and ammonia-blended fuel projects introduce further concerns around unburned ammonia, oxidation chemistry, materials compatibility, and the interpretation of temperature trends during changing operating modes. A provider does not need to claim universal fuel expertise, but it should demonstrate relevant material selection, test history, and failure analysis for the proposed service.

Probe response time is another frequent source of mismatch. A thick, heavily protected probe may be mechanically robust but too slow for the diagnostic purpose. A fine exposed junction can respond quickly but may be unsuitable for high-velocity, particulate-laden exhaust. The correct choice is not the fastest or most durable probe in isolation; it is the probe whose response and life match the operational consequence of a delayed or failed measurement.

Verify installation assumptions before comparing providers

Many alleged sensor failures originate in the installation design. EGT readings are affected by probe immersion depth, gas velocity, pipe diameter, flow profile, bends, manifold geometry, shielding, local recirculation, and the thermal conduction path into the mounting boss. On reciprocating engines, placement differences between cylinders can produce apparent imbalance even when combustion is stable. On turbines, a non-representative radial or circumferential location can distort exhaust temperature spread.

Providers should supply clear guidance for insertion length, mounting orientation, weld bosses or compression fittings, allowable vibration loading, cable routing, bend radius, and sealing arrangements. Where the application is safety- or availability-critical, the assessment should include drawings and a review of actual nozzle or manifold geometry rather than approval based on a generic installation note.

A useful question is whether the provider can explain the likely measurement bias created by a proposed installation. A supplier that only confirms mechanical fit is not necessarily validating the measurement.

Test performance under transients and ageing, not only at calibration points

Factory calibration establishes a baseline. It does not prove that a sensor assembly will remain stable after thousands of thermal cycles, exposure to vibration, or contamination by ash and condensate. Exhaust service commonly produces drift through oxidation, contamination, junction degradation, insulation leakage, mechanical fatigue, and changes in thermal contact.

Evaluation should include evidence from comparable duty cycles: operating temperature, fuel, engine or turbine type, exhaust pressure, vibration severity, and expected service interval. Generic operating-hour claims are weak unless the provider can define the conditions behind them.

For critical assets, a practical acceptance approach is to install a limited pilot set alongside a reference method or independently calibrated comparison probes. The objective is not to expect perfect agreement between probes at every instant; two probes can legitimately see slightly different local gas conditions. Instead, assess bias, repeatability, response during defined load changes, channel stability, and drift after a meaningful operating period.

Trend behavior is especially important. A system that is consistently offset by a known and manageable amount may still support maintenance diagnostics. A system with intermittent dropouts, unexplained step changes, or unequal drift across channels is much harder to use safely.

Calibration traceability must extend beyond a certificate

A calibration certificate is valuable only when it identifies the instrument, calibration points, method, uncertainty, environmental conditions where relevant, and traceability path. Calibration performed by a laboratory accredited to ISO/IEC 17025 for the relevant temperature range generally provides stronger assurance than an undocumented in-house check, though accreditation alone does not validate the installed application.

Ask whether calibration is performed on the complete assembly or only on the sensing element. Ask how replacement probes are matched to existing channels, how transmitter configuration is controlled, and whether field verification can distinguish sensor drift from wiring or input-module faults.

For assets operating under formal quality systems, the provider should also support serial-number traceability, revision control, and clear records for repair, recalibration, and configuration changes. ISO 9001 certification can indicate a structured quality-management system, but it should not be treated as evidence of measurement accuracy by itself.

Assess data integrity as part of measurement accuracy

Modern EGT systems increasingly feed historians, engine management systems, alarm platforms, remote diagnostics, and AI-based maintenance tools. A sound sensor can still become operationally unreliable if timestamps drift, values are rounded, units are converted incorrectly, bad-quality flags are lost, or filtering hides fast events.

The provider should document sampling frequency, scan rate, input isolation, data resolution, synchronization, filtering logic, alarm deadbands, missing-data treatment, and communications behavior during network interruption. For a protection-related use case, confirm the end-to-end latency from physical temperature change to alarm or control action.

Cybersecurity and access control are also relevant where data passes through gateways or cloud platforms. The issue is not only external intrusion; uncontrolled software updates, user changes to scaling, or undocumented dashboard calculations can compromise the integrity of maintenance decisions.

Use standards as a framework, not a substitute for validation

Relevant standards depend on the asset and purpose. Thermocouple construction and tolerance standards help establish component quality. ISO/IEC 17025 supports confidence in calibration competence. Engine and turbine performance or emissions frameworks may define broader test and documentation practices, while marine installations may also be influenced by class requirements, flag-state obligations, and IMO emissions rules.

Yet EGT measurement is often a supporting parameter rather than a single universally regulated measurement. Compliance with one component standard does not prove suitability for a particular exhaust manifold, turbine outlet, catalyst inlet, or stack condition. The decisive evidence remains the provider’s documented uncertainty, installation engineering, validation method, and service history in comparable conditions.

Make the decision on lifecycle evidence

The lowest sensor price can be misleading when access requires an outage, a vessel call, scaffolding, or partial disassembly. A provider should be assessed on replacement lead times, regional technical support, spare-part interchangeability, repair policy, obsolescence management, and the ability to preserve historical comparability after hardware revisions.

Before award, require a concise technical dossier: complete uncertainty budget, calibration scope and traceability, materials specification, response-time evidence, installation drawings, environmental limits, data architecture, fault diagnostics, relevant reference cases, and a proposed site acceptance test. Any unsupported accuracy statement should be treated as a commercial claim rather than an engineering result.

The most dependable exhaust gas temperature data provider is not necessarily the one offering the narrowest catalogue tolerance. It is the one able to demonstrate that its measurement chain will remain interpretable, traceable, and stable in the actual exhaust environment—long enough to improve operating decisions rather than merely populate a dashboard.