Transmission EOL: 66 s Is a Budget, Not a Robot Count
A transmission assembly is large and carries many inspection points, reaching the final station on a slat line after assembly. Manual inspection fails directly: many points, scattered positions, long time, and a higher chance of missing items toward the end. This cell builds equipment around the assembly structure and inspection list, splits points across several robots, gives each two 20 MP cameras for archiving and inspection, and uses a pneumatic or servo floor track to extend coverage. The cycle is 66 seconds.
Sixty-six seconds is a total budget, allocated per robot
The easiest mistake in this kind of cell is assuming more robots equal a faster line. Total takt is not set by robot count but by the slowest branch. The list is therefore split by time as well as by location: each robot should receive a similar number of points, a similar capture path and a similar judgment load, or one machine becomes the bottleneck. With an uneven split, the extra robots are only waiting alongside the others.
Judgment has to fit the conveyor's locating window
The slat conveyor runs continuously, and the assembly is fixed by the locating mechanism when it reaches the final station; inspection completes inside that fixed window. The window length determines when the verdict can be output. The robots move along the floor track around the located assembly and capture, and the whole inspection synchronizes with the conveyor without adding offline time. A stable unit pose inside the window is the precondition for every fixed viewpoint.
The price of parallelism is choreography
Several robots working around the same assembly overlap in motion. Without choreography of capture timing, the best case is mutual waiting and the worst is interference. Choreography covers three things: the working zone of each robot, the sequence for entering and leaving, and avoidance when several operate locally at once. This work is often underestimated, yet it decides whether the takt is actually met. The gap between the theoretical cycle and the real one usually lives here.
Dual cameras: evidence at the moment of release
Each robot carries two 20 MP cameras, one archiving and one inspecting. The value is evidence at release: if a quality dispute arises later, no on-site recheck is needed, because the archived image can be retrieved to review the decision. The inspection camera outputs the verdict in real time, the archive camera records in parallel, and neither consumes the takt.
Lighting by feature, not by station
Reflection on metal surfaces affects imaging consistency, a general difficulty in assembly inspection. The cell uses high-brightness lights for large surfaces and bar lights for holes and narrow gaps. The two areas have different imaging needs, and one light for everything guarantees that one class is judged badly. Pairing lighting configuration one to one with points is the basis of repeatable judgment.
The longer the list, the harder the trade
The cycle is 66 seconds and data is traceable, subject to site acceptance results. Custom equipment cost correlates directly with inspection item count, so compiling the list means balancing completeness against equipment complexity. Add every possible failure and the equipment inflates past economic sense; omit a key item and final inspection loses its purpose. The basis for the trade is failure consequence: items that would cause assembly or after-sales problems must stay, while pure appearance items can be sampled by batch. Overconfiguration and underconfiguration are both expensive.
Questions about takt and allocation
Why can robot count not be derived from takt?
It can, but from the slowest branch rather than total workload divided by single-machine capacity. Balanced allocation is what sets the cycle.
How do 66 seconds align with the slat line?
Inspection completes inside the fixed window after the assembly is located, robots capture in parallel within that window, and release or interception is issued at its end.
Servo track or pneumatic track?
Choose servo for dense points needing precise alignment; choose pneumatic for wider spacing and cost sensitivity. The layout decides, not a preference for accuracy.
How long should archived images be kept?
It depends on traceability requirements and the dispute cycle. Disputes often appear late, so the retention period should be agreed with quality at proposal stage.