Chassis Studs and Locating Holes: 3D Check in 40 s
A vehicle chassis is a large flat assembly with studs and locating holes scattered across several regions, and they provide the datum for later positioning and fastening. A missing or tilted stud, or a displaced hole, becomes a mating failure at the marriage station. This cell applies inline 3D vision, judges presence and position within a 40 second cycle, and outputs a measurement report with anomaly alarms. It runs 7 x 24, subject to site acceptance results.
Error proofing and measurement want different failure modes
Measurement pursues accuracy; error proofing pursues zero escapes, preferring to block too much rather than let a bad unit through. That means trading detection rate against false-block rate and accepting continuous validation against production data. Treat error proofing as measurement and boundary samples pass; treat measurement as error proofing and the line blocks constantly. The metric here is not how precise the system is, but whether it stopped what should have been stopped.
A stud gets two tests; a hole gets two tests
A stud is confirmed by height and cross-section profile together: height decides whether it is seated, profile decides whether its form is normal, and a tilt or a low sit leaves traces in both. A locating hole is checked by centre coordinates and diameter together. Centre alone can be pulled off by oil, and diameter alone cannot show displacement. Two criteria are not complexity for its own sake; they are what holds up under dirt and reflection.
Only feature regions fit inside 40 seconds
The chassis is large, but the inspection elements are discrete studs and locating holes. The cell images only feature regions rather than scanning the full surface, which is the key to a 40 second cycle. A full scan brings a large volume of data unrelated to the verdict, slowing the cycle and loading the algorithm.
Rules catch this unit; trends catch the batch
Rules come from process standards and answer whether this unit is acceptable. Trends come from historical statistics and answer whether the batch is drifting. The classic case is a locating fixture drifting slowly, moving all hole positions together: rules alone cannot see it, trends expose it early. Used together, they avoid both false alarms and missed systemic risk.
Oil and reflection are standing conditions, not exceptions
Line oil and metal reflection degrade point-cloud quality. The imaging end controls exposure and polarization, and the algorithm end makes robust estimates for missing regions, judging only valid feature areas. These steps do not change the criteria, but they decide stability on a real line.
The report follows the chassis
The measurement report binds to the chassis serial number and travels with the vehicle as a single-unit quality file. The station sits at the quality gate before marriage, and alarms map to the serial number, so repair can locate which feature and how far it deviated without re-inspecting everything.
What the chassis team asks before sign-off
How is stud presence judged in 3D?
The point cloud around the stud is captured, contour and expected position and height are compared, and a missing feature is reported as absent. Height data also reveals tilt and low seating.
How is locating-hole position deviation measured?
Hole edge contour is extracted to compute actual centre coordinates, and deviation beyond tolerance fails the unit. Tolerance follows the process standard, and the report records the deviation value.
What does trend alarming catch?
Gradual anomalies such as a fixture drift moving hole positions together. Single units may all pass, but consecutive data shows a systematic shift that trend comparison exposes early.
Where else can this error proofing apply?
Assembly processes datumed on studs, locating holes or nuts, such as axles, subframes and battery housings, especially lines where an escape is costly and interception is required.