Reading a Thermal Image: Generic Method

Why this matters

A thermal camera can find a failing connection, a hidden leak, or an overloaded circuit in seconds that would otherwise take an hour of disassembly to locate, but only if you read the image correctly. A tech who does not understand what the colors actually mean can just as easily chase a false positive (a warm spot from direct sunlight, not a fault) or miss a real one (a fault that reads only slightly warm because of how the camera's scale is set). The camera does not diagnose anything by itself; you do, using the image as one more piece of evidence.

Step 1: Understand what the camera is actually measuring

A thermal camera reads surface temperature by detecting infrared radiation, not "what is happening inside" a component. Keep two limits in mind before you trust any reading:

• It sees the surface, not the interior. A hot connection inside an enclosure shows up as heat on the enclosure surface only once enough heat has conducted through. A fault that has not yet built up enough heat, or one that is well insulated from the surface you can see, will not show clearly yet.
• Emissivity changes what the camera reports. Shiny or reflective surfaces (bare metal, glass, some plastics) reflect ambient thermal radiation and can read inaccurately, both too high and too low, unless you account for it. Matte, dark surfaces read closer to their true temperature. When comparing a shiny component to a dull one, do not trust a direct number comparison between them without adjusting for this.

Step 2: Set up the comparison correctly

Thermal imaging is almost always a comparison, not an absolute reading. Set it up so the comparison is valid.

1. Compare like to like. Look at the same type of component under the same load, ideally right next to a normal one doing the same job (three phases of the same circuit, several of the same fitting, several of the same bearing). A single reading with nothing to compare against tells you much less than a side-by-side.
2. Match load and duty cycle before comparing. A component reads hotter under higher load, so a warm reading on a heavily loaded item next to a cool reading on a lightly loaded one may simply reflect the load difference, not a fault. Where possible, compare components under similar or known load.
3. Let the system reach a stable operating condition before imaging, where practical. A system just powered on has not reached thermal equilibrium, and early readings can be misleading in either direction.
4. Control the environment as much as you can. Direct sunlight, nearby heat sources, recent contact with a hand or a tool, and airflow from a fan or vent all distort a reading. Note anything nearby that could explain a hot or cool spot before attributing it to the component itself.

Step 3: Read the pattern, not just the peak color

The camera's color scale is relative to whatever range you set it to, which means the same real temperature can look dramatically different depending on the scale. Do not read color alone.

• Check the actual temperature value, not just the color, for anything that looks notable. Cameras display a numeric readout at the point you select; use it rather than trusting "it looks red" alone, since a narrow scale can make a small, harmless difference look dramatic.
• Look for a differential, meaning how much hotter or cooler a point is than its expected comparison, rather than an absolute number in isolation. A localized connection running measurably hotter than the identical connections around it is a much stronger signal than any single absolute reading, and this differential approach holds even when you do not know the exact emissivity or true surface temperature.
• A tight, localized hot spot (one connection, one bearing, one specific point) usually means a discrete fault at that point: a loose or high-resistance connection, a failing bearing, a blocked or restricted flow path.
• A broad, gradual gradient rather than a sharp spot often points to something distributed, like insulation loss over a length of pipe or duct, or general airflow restriction, rather than one failed component.
• A cool spot where you expect heat, or a hot spot where you expect cool, is just as meaningful as an unexpected hot spot. A cooler-than-expected area on a normally warm supply line can indicate restricted flow or a blockage upstream.

Step 4: Confirm before you act

A thermal image is a lead, not a finished diagnosis. Before recommending a repair based on it:

1. Cross-check with a direct measurement where practical, an electrical reading, a pressure or flow reading, or a physical inspection of the flagged point.
2. Rule out an environmental explanation for anything you flagged: sunlight, a nearby heat source, recent handling, reflective interference. If any of these plausibly explain the reading, confirm with a second method before committing to a diagnosis.
3. Document the image with context, meaning the actual temperature values, the comparison point, the load condition at the time, and the environment. A thermal image without this context is much less useful later, both to you on a follow-up visit and to a customer or a manufacturer if you need to escalate a finding.

Common mistakes to avoid

• Trusting color alone without checking the actual temperature value.
• Comparing components under different loads and treating the difference as a fault.
• Imaging a system before it has reached a stable operating condition.
• Ignoring emissivity differences between shiny and dull surfaces in the same image.
• Treating a thermal finding as a confirmed diagnosis without a second, direct measurement to back it up.

References

• Manufacturer documentation for thermal imaging equipment, including emissivity and calibration guidance
• Trade-standard practice for infrared inspection of electrical and mechanical systems
• See related: Reading Rust and Corrosion Patterns
• See related: What You Can Only Catch With It Installed