Three parts, inspected blind
We gave loh.ai three scanned metal parts and nothing else, then asked the questions a technician would. Here is what it found, where it told us to look, and what it got wrong.
01
Just the scan
No CAD model, drawing or specification. The detector compared each part with its own surface.
02
The same six questions
What is it, triage the scan, which surfaces matter, where to look, what might be an artefact, what to check by hand.
03
Every number checked
Each figure in an answer is checked against the detector's output before it is shown.
04
No pass or fail
Without a material, process or specification, the agent ranked what to check and gave no verdict.
Case 01
Flanged bushing
A flanged bushing, standoff or spacer, probably metal additive manufactured. It read the round flange and its two holes as a bolted mounting, and the dark speckle inside the cylinder as a bore the scanner could not see into.
- Size
- 43.3 × 43.5 × 35.7 mm
- Points measured
- 519,879
- Findings
- 143
3 critical50 moderate90 minor
Its assessment
All three critical findings are raised material hidden from every outside view, most likely on the bore wall. Gauge-check the bore first.
Check these first
- #23About 7–8 mm down the bore, at 9 o'clockA raised patch about 7 mm long on the inner wall
- #36About 11–12 mm down, at 9 o'clockA second raised patch just below #23
- #87About 16–17 mm down, at 6 o'clockA smaller raised patch deeper in the hole
Where it disagreed with the detector
#77 is graded moderate but has the largest deviation on the part (0.29 mm), so it deserves critical attention. #87 is critical only by its area.
Telling an artefact from a defect
Probe the bore with a pick, then look with raking light: real spatter casts a shadow. Blow out the bore and look again, since loose powder disappears.
Case 02
Clevis bracket
A clevis bracket, pivot lug or rod-end fitting: a round boss with a cross bore running into a flat, tapered tab. Likely a pivot in a linkage, with a pin through the bore.
- Size
- 39.1 × 20.9 × 32.5 mm
- Points measured
- 281,936
- Findings
- 94
20 critical26 moderate48 minor
Its assessment
Check whether #36 and #75 are real cavities, confirm #27 is only the part's own fillet, and look inside the bore for #76.
Check these first
- #76Inside the cross bore, lower section, about 8 mm above the baseShine a light in from below; a large rough patch on the wall
- #36Outside of the round boss, 15.7 mm above the base, right sideA pit 0.25 mm deep, the largest deviation on the part
- #27Underside, along the edge where the boss meets the tabA raised ridge running along that edge
Where it disagreed with the detector
#76 is critical by area but its average deviation is almost zero, so it may be the curved bore confusing the reference. #27 and several others sit on sharp edges and are likely edge artefacts.
Telling an artefact from a defect
Run a fingernail along the tab edge. A ridge that follows the whole corner is the part's shape confusing the software; real damage is local.
Case 03
Curved housing
A protective shell, cover or housing: a curved body, a flatter tab with countersunk fastener holes, and a grid of small through-holes read as a vent, grille or sensor window.
- Size
- 127.6 × 74.6 × 32.9 mm
- Points measured
- 2,356,804
- Findings
- 782
63 critical221 moderate498 minor
Its assessment
A large share of the critical count comes from the reference fit, not the part. The credible findings are a bump on the inner shell (#258) and a cluster of pits on the outer curve (#643, #699).
Check these first
- #258Inner surface, where the shell meets the flat tab, about mid-lengthA small, distinct bump or a stuck particle
- #643Outer surface, on the rounded end opposite the tabA small dent or pore
- #699Beside #643 on the outer curveA second pit in the same cluster
Where it disagreed with the detector
#13 and #231 are graded critical only because a poor reference fit spreads over a large area. Their average deviation is near zero and changes sign, which points to an artefact.
Telling an artefact from a defect
Look at the edges of the small holes. Rough edges that repeat on every hole are a limit of the scan; damage on one or two holes is real.
What it missed
One edge of the shell carried real metal shreds, deleted from the mesh by accident during cleaning before the test. The agent flagged that edge as abnormal and put it on its check-first list, but read it as damage around the holes rather than foreign metal. 221 of the 782 findings sit on that edge.
Try it on one of your own parts
Bring a scan to a 30-minute call.