The failure you want to catch is the one that has not happened yet
A tile that has already fallen is a claim. A tile that is debonded but still in place is a decision — and it is only a decision if somebody knows it is there.
That is the gap thermography fills. Debonding is invisible from the ground and, in most cases, invisible from a metre away. It is not invisible to heat.
The physics, briefly
A facade tile bonded properly to its substrate is thermally continuous with the wall behind it. Heat flows into and out of it through that bond.
Introduce a void — a pocket of air between tile and substrate, whether from adhesive failure, thermal cycling or water ingress — and you have introduced an insulator. Air conducts heat roughly two orders of magnitude worse than mortar.
The consequence is straightforward:
- Under solar gain, the debonded tile has less mass behind it to absorb heat, so it warms faster than its neighbours.
- During cooling, it has less thermal reservoir to draw on, so it cools faster too.
At either moment, that tile is at a different temperature from the sound tile beside it. The difference is often only a few tenths of a degree, sometimes less. A camera that resolves 0.05°C can see it. A person standing on the pavement cannot.
Why the survey window matters more than the camera
This is the part most discussions of thermal survey skip, and it is the part that determines whether the data is worth anything.
Thermal contrast only exists while heat is moving. On a wall that has been in shade all afternoon and has reached equilibrium, a void and sound substrate sit at the same temperature. A perfect camera returns a perfectly uniform, perfectly useless image.
So the survey has to be flown inside a window:
| Condition | Why it matters |
|---|---|
| Active heating or cooling | No thermal gradient, no signal. Post-sunset cooling is usually the cleaner of the two |
| Dry facade | Evaporating moisture cools the surface unevenly and swamps the signal |
| No rain for several hours | Wet substrate behind the tile changes conduction and produces false readings |
| No direct low-angle sun | Grazing sunlight creates surface reflections that read as thermal anomalies |
| Low wind | Wind cools the surface unevenly across the elevation |
In practice, this is why we favour scanning after sunset. Heat loss to the night sky is far more uniform across an elevation than solar gain across a facade with balconies, fins and reveals casting shadows.
What thermography cannot do
Being clear about the limits is what makes the findings trustworthy.
It cannot measure void depth. A thermal anomaly tells you something is different behind that tile. It does not tell you whether the void is 2 mm or 20 mm, and depth matters for repair specification.
It cannot distinguish every cause. Water ingress, a void, and a change in substrate material can all produce a thermal anomaly. Interpretation requires context — which is why the visible-light imagery is captured alongside and registered to the same coordinates.
It cannot sign a report. Under Taiwan's regulations, the diagnosis is made by a qualified inspector or diagnostician. Thermography is evidence they use, not a substitute for their judgement.
From thermal anomaly to defensible finding
The sequence we run is deliberately conservative:
- Capture — full-elevation thermal and visible-light imagery, geometrically registered so a thermal anomaly maps to an exact position on a real wall.
- Quantify — recover per-pixel temperature and compute ΔT against the surrounding sound area. A finding is a number, not a colour on a heat map.
- Classify — the AI reads the anomaly in context: is this consistent with hollowing, with water ingress, or with a material transition?
- Verify — for anything in the classes that can fall and injure, a qualified inspector attends and confirms on site. Automation may flag; it may not dismiss.
- Record — the finding, its coordinates, its ΔT and the verification are all retained, so next year the same patch of wall can be compared against itself.
Step 5 is the one that changes what the survey is worth. A single thermal survey tells you where the voids are today. Two surveys a year apart tell you whether they are spreading — and spreading rate is what a repair budget is actually built on.
What this looks like in a report
Every hollowing finding in our reports carries four things: a position (GPS plus elevation coordinates), a measured ΔT, paired visible and thermal images, and a severity grade. The grade feeds the building's health score, weighted at 30% — the heaviest of the five dimensions, because hollowing is the precursor to the failure mode that actually hurts people.
If your building has tiled elevations
Age and material combination matter more than absolute age. A 25-year-old tiled facade in a coastal or high-thermal-swing environment can be further along than a 40-year-old one inland.
The way to find out is to establish a baseline. Book an asset assessment or call +886-2-7733-7678.
