Screening Fixtures for 1 to 2 mm Deviation With a Tablet
Large welded structures, frames, tooling and fixtures in heavy equipment and machinery plants are awkward to move. Fixtures serve for years, and deformation, knocks and assembly changes accumulate. Local wear and loosened fasteners are often only millimetres deep, hard to notice, and they gradually affect assembly quality. There are many fixtures spread over a wide area, while metrology capacity sits in one fixed zone, so sending everything for measurement is neither physically practical nor affordable.
Fixture condition drifts slowly
The traditional answer is to move the workpiece to a fixed inspection area for a preliminary screen, which costs handling and occupies precision metrology hours. For a fixture in long service, a single result matters less than the trend of its condition, and a trend only exists if the screen is light enough to repeat on a cycle.
One to two millimetres is a boundary, not an accuracy claim
The scheme is tablet AR comparison: an inspector takes the tablet to the shop floor or the warehouse, the camera recognises and tracks the workpiece, and CAD is superimposed live. Deviations at roughly the 1 to 2 mm level can be seen. That number exists to draw a line: anything that reaches the millimetre level is doubtful, everything else passes. The verdict has only two outcomes, release or refer, which is simple enough to execute quickly on site.
Anchor on the main structure, work from large faces down
Fixtures have rich structure and good stiffness, so tracking is stable and the overlay has a reliable base. Working region by region, the main structure is the datum, large faces come before local details, and the walk path is fixed. Observation distance and angle decide how visible a deviation is: a scan from standing height at a distance misses what a close pass catches, so each region has to be approached.
Clean the surface first, then register
Oil, rust and old paint on a production floor degrade feature recognition. The fix is simple: clean the feature areas before registering so the overlay can be trusted. That step is done by the inspector on site, and it decides whether the overlaid picture is usable as evidence.
Doubtful parts are photographed at the deviation location and handed over with the metrology request, so the second check has a specific target. Passed parts are released where they stand. Screening records are archived by asset number to track condition over time, and repeated deviation patterns become a checklist that is updated as experience grows and used to focus the next screen.
Tiered inspection moves the bottleneck
On-site screening and metrology sit in series, and screening throughput is far higher than metrology, so overall inspection speed is set by the fast step rather than the slow one. Handling drops, impact risk drops with it, and metrology hours concentrate on doubtful parts. Buyers should first weigh the tolerance the part allows against metrology load: where sub-millimetre tolerance judgment is required, go straight to metrology and use AR only as a pre-filter. Figures are subject to site acceptance results.
Questions from the floor
Is 1 to 2 mm good enough?
It depends on the purpose. It filters obvious anomalies and releases normal parts; it does not issue a tolerance verdict. Precision judgment belongs to metrology.
What is the handoff for doubtful parts?
The deviation location is recorded and photographed, transferred with the metrology request, and metrology confirms whether it is out of tolerance.
When should AR not be the only method?
When the tolerance is below a millimetre or the check involves precision geometry, the 1 to 2 mm capability cannot decide it and metrology should be scheduled directly.