An infrared camera does not look inside a roller bearing. It detects radiation reaching the instrument and uses its settings to estimate a surface temperature. That makes thermal imaging useful for finding differences worth investigating, but only when the compared surfaces and observation conditions support the comparison.
On an undercarriage, those conditions are rarely uniform by accident. One roller may have shiny exposed metal while another carries paint or dried mud. One side may face sunlight or a warm nearby surface. The first image may be taken soon after work and the second after several minutes of cooling. A difference on the camera can therefore arise before there is evidence of a meaningful difference between the components themselves.
Before treating the warmer image as a component problem, compare how the two images were made. Surface finish, reflected surroundings and elapsed cooling time can each change the interpretation. A repeatable difference becomes more informative once those obvious alternatives have been addressed.
Why infrared numbers can disagree before the parts do
Emissivity describes how a surface emits thermal radiation relative to an ideal emitter. It is not simply a permanent number attached to the word “steel.” Optris explains that emissivity depends on wavelength, temperature and viewing direction. The actual surface condition also matters to the measurement, especially when comparing exposed metal with a different finish.
Low-emissivity surfaces make reflected radiation particularly important. Radiation from the surroundings can reach the camera after reflecting from the part. A warm reflection can make an area appear hotter; a reflection of a colder environment can push the apparent result the other way. The displayed number then depends on more than the surface temperature you intended to compare.
Using the same emissivity setting on two different finishes does not make them equivalent. It merely applies the same numerical assumption to both. Equally, adjusting the setting until neighbouring rollers display the same temperature would erase the observation rather than validate it. The setting needs a defensible basis for the surface and instrument being used.
A moving bright patch can be an optical clue
Fluke illustrates how changing viewing position changes apparent temperatures on reflective metal. A feature that shifts with the reflected surroundings is a reason to examine the measurement conditions. It is not evidence that heat is physically travelling across the component at the same moment.
For an undercarriage inspection, any alternative view must be available from a safe position under the site's applicable controls. Do not approach moving tracks, reach through guards or move into a crush zone to test a reflection. A doubtful image can be recorded as doubtful; obtaining another angle is not more important than maintaining safe access.
The opposite observation also needs restraint. A warm-looking region that stays in one place may justify further investigation, but persistence alone does not identify a bearing fault. The camera still observes a surface, and load, cooling, surface finish and measurement setup remain part of the interpretation.
A small target changes what the camera can resolve
Distance affects the image detail available for a component. The roller surface must occupy enough of the instrument's measurement area for the intended reading. A visible crosshair on the part does not necessarily mean that all the radiation contributing to the reported spot comes from that feature. Background or neighbouring surfaces can influence an inadequately resolved target.
Focus, the selected temperature range and the measurement area also need attention. FLIR's thermographic measurement guide separates image-acquisition requirements from parameters such as emissivity, reflected apparent temperature and distance. These are instrument-specific considerations; a camera's ability to produce an attractive image is not enough to establish a defensible temperature measurement.
Keep the original radiometric file when the camera produces one, together with the visible-light context image. A coloured screenshot may show where an area looked warm, but it can discard information needed to review settings or measurement regions later. The palette itself is not a temperature standard: two images with different display spans can assign very different colours to the same numerical temperature.
Build a comparison around equivalent observations
A paired comparison starts with corresponding components and surfaces. Identify the machine, side and roller position, then select the same kind of exposed surface on the comparison component. A shiny flange on one side and a mud-covered roller body on the other are not equivalent targets simply because both belong to a roller assembly.
Record recent operation and the time between stopping and imaging. The two sides may have experienced different turning, travel or loading, and they may not cool at the same rate. “After work” is a broad description. A useful record says what work occurred and when each image was captured relative to it, within what the available records can establish.
Try to preserve the same camera, lens or mode, target region, viewing geometry and relevant settings across the pair. Record ambient conditions and notable surroundings, including sun exposure or nearby warm surfaces that could affect reflection. Where a condition differs, keep the difference visible instead of presenting the two numbers as a controlled comparison.
An apparent difference that needs a better comparison
Imagine a fictional check in which the right-side roller is photographed on a shiny exposed face just after shutdown. The corresponding left-side image is taken later, across a coated face at a more oblique angle. The right side displays a higher value. Several variables changed at once: surface, timing and geometry. The pair does not yet isolate a component-temperature difference.
The next useful action is to review those variables and obtain a more comparable observation if the applicable safe inspection arrangement permits it. Merely subtracting the two values, or comparing their colours, gives a precise result from a poorly defined comparison. A numerical difference can be calculated even when it has little diagnostic meaning.
Now suppose a later pair uses corresponding surfaces under documented similar conditions and shows the same location consistently higher. That result deserves more attention because the obvious comparison problems have been reduced. It still needs to be related to the machine's actual condition: a qualified inspection may examine rotation, leakage, damage, packing or other relevant evidence under the correct procedure.
This distinction prevents two opposite errors. A surface-related artefact should not trigger replacement of a sound component. A real repeatable anomaly should not be dismissed merely because infrared measurements can be difficult. The quality of the comparison determines how much weight the observation deserves.
Decide what the thermal observation has earned
Use the image as a screening result with an explicit scope. A defensible note identifies the component and surface, records the relevant settings and operating context, and describes the observed difference. It should avoid replacing the observation with an unsupported diagnosis such as “bearing seized” unless separate evidence has established that condition.
There is no universal temperature difference that proves every undercarriage roller has failed. Machine design, component position, load and the measurement method can change what is expected. Apply an actual manufacturer criterion or established inspection programme where one exists for the identified equipment; do not create an alarm limit from a general thermography article.
If the comparison remains unreliable because of reflectivity, restricted access or a poorly resolved target, say which problem prevents interpretation. The answer may be a different approved measurement method or a qualified inspection that does not depend on that surface reading. Do not improvise heating, surface treatments or tape placement on a working undercarriage to make the camera produce a preferred number.
For a repeatable anomaly at a track roller, retain the thermal image with a visible-light view locating that roller on the machine. The two views serve different purposes: one preserves the temperature pattern and measurement region, while the other lets a technician find the same component for the mechanical assessment.
An isolated bright patch may disappear when a reflection is understood. A repeatable temperature difference under comparable conditions leaves a different question: what is happening at that particular component? That is where the thermal record becomes valuable to the mechanical inspection—by identifying a location and a pattern that deserve an explanation.