Predictive Aircraft Maintenance APAC met at Marina Bay Sands on 21–22 September while MRO Asia-Pacific filled Suntec next door. The week’s language was familiar, digital twins, AI forecasts, health monitoring, smarter inspections. Safran talked product health. Veryon talked defect clustering. The combination vendors keep selling is a drone with a sensor that sees past ordinary light, a model of the same tail from last heavy check, and software that flags what changed. That is inspection physics, not a task-card app. Keep it separate from who signs the CRS.
Conventional walkaround and stand work still live on access, light, clean paint and a trained eye. Hyperspectral cameras look at wavelengths people do not. A twin can subtract last year’s surface from this morning’s pass. The useful question is whether the stack can tell paint fade from a skin crack, dirt from corrosion, a normal composite weave from delamination, a warm hydraulic line from a dying bearing, and a real defect from a dirty lens or a bad mesh. Until those five separations are published against known damage, not against a booth video, the output is a hint. Hints send a human with a tap hammer. They do not close an AD.
Accountability is the certification problem the brochure skips. If the model misses a crack, the chain is long like technician who accepted the pass, airline that wrote the procedure, drone operator, sensor house, software vendor. EASA and the FAA will have to say which file is advisory and which file may enter the approved maintenance data. Until that line is drawn, a drone that heard a crack is still a photographer. The engineer who walks the same bay with a torch remains the person who can be named when the next owner opens the record.