Emissivity is the property that turns a thermal camera from a radiance sensor into a
thermometer. The instrument measures radiation; converting that reading into a temperature
requires knowing how efficiently the surface radiates compared with an ideal emitter. Get the
value wrong and every absolute temperature in the record is wrong with it, in a consistent
and entirely plausible-looking way.
For welding this is harder than for most applications, because the surface does not hold
still. Bright rolled steel, mill scale, an oxidised heat-affected zone, spatter and a fresh
weld bead have visibly different emissivities, and a single seam can present several of them
within a few centimetres. Temperature changes it too, and so does viewing angle.
That is why serious thermal weld monitoring is built to be robust to it rather than to assume
it away. Two approaches do the work. Relative features — gradients, cooling slopes, the
shape and symmetry of the thermal signature, run-to-run repeatability — survive an uncertain
emissivity because the error largely cancels. Calibrated absolute measurement is used
where a real number is required, as when an interpass maximum has to be evidenced, and it
depends on establishing the value for that surface and recording that this was done.
The practical rule follows from the physics: absolute temperature claims need a documented
emissivity and a documented surface state, while anomaly detection and process-drift signals
do not. A supplier quoting absolute weld temperatures without saying which value was used,
and on what surface, is quoting a number with an unstated assumption inside it.