1,000 Tray Points at 0.2 mm in 85 Seconds
An EV battery tray usually measures more than 1.8 m x 1.5 m x 0.3 m and carries over a thousand measurement points. Any hole that is off can cause interference at vehicle build. Sampling cannot cover that kind of point-level risk, and the tray is a welded and assembled structure whose hole pattern and mounting faces must be fully confirmed before the pack goes into the car. The requirement is therefore not a higher sampling rate but a full check at a producible cycle.
Sampling does not fail on the part, it fails on coverage
Sample inspection takes a proportion, and the one hole it did not take is exactly the one that is out. That is not a probability problem but a mismatch between the coverage method and the risk. Raising the sample rate from one in ten to one in two still leaves half the points unseen, and the cost per part climbs with it.
Seven geometric relations, not one dimension
The station evaluates hole position, flatness, profile, symmetry, concentricity and linear distance, plus missing and wrong parts. The first six are geometric evaluation; the seventh extends measurement to assembly completeness, which is what separates this from an ordinary dimensional gauge. Each of the seven has its own fit and its own tolerance, and they all reference one coordinate frame.
Parallel capture puts a thousand points near the cost of one
One camera cannot cover a 1.8 m part and keep accuracy at the same time. Several cameras cover zones, shoot simultaneously, compute in parallel and measure jointly. Parallelism is the whole point: shooting 1,000 points in series could never fit a line cycle, while zoned parallel capture makes the total time the same order as a single point. That is what makes a full check inside 85 seconds possible, subject to site acceptance results.
The second number is repeatability at 0.025 mm. A full check is not only about measuring correctly once, but about being comparable across parts and shifts. Without repeatability the data cannot go into a quality file, and the check is just a series of one-off measurements.
Calibration is the foundation of the chain
The chain runs calibration, imaging, feature extraction, dimension calculation and output. Calibration establishes the relationship between the cameras and the workpiece frame, and every camera's result is brought into that one frame before cross-zone hole relationships can be evaluated. Imaging triggers a retake if any camera fails to capture. Feature extraction locates more than 1,000 points, dimension calculation evaluates the seven relations, and output produces the report plus release or block instructions.
How it joins the line, and where it stops
Trays arrive on the line conveyor, are measured inline and released or blocked without leaving the conveyor, so the measurement interleaves with transport and occupies no extra station time.
The limits should be stated plainly. Accuracy is in the 0.2 mm class, suited to tray dimensions and assembly features, not to a metrology room's CMM work. Joint measurement depends on calibration stability and needs periodic review, and incoming attitude or surface changes have to be assessed for their effect on feature extraction.
Before a measurement quote is compared
How do 1,000 points fit in 85 seconds?
By zoned parallelism rather than faster single-camera cycle. Cameras shoot and compute at once, and the total time is the same order as a single point. Serial capture of 1,000 points could not meet the line.
Is 0.2 mm enough?
It depends on the question. Hole positions, planes, profiles and assembly relations are covered by 0.2 mm with 0.025 mm repeatability; micron-level form tolerances belong to a CMM.
How is alignment across cameras assured?
Calibration brings every camera into one workpiece frame. Its accuracy decides whether a cross-zone hole relationship is trustworthy, so it is reviewed on a schedule rather than once.