How Many Robots for Powertrain EOL? Start With the Point Map
Diesel engine, gasoline engine and transmission assemblies reach end of line with a long checklist attached: bolts, poly-V belt, nameplate, certificate of conformity, fan, filter. Each has to be confirmed present and correct rather than missing or wrong. The unit is heavy, the points sit on several faces, and an operator circling it to work through the list is slow and fatigue-prone. This project put a slat conveyor and robot-mounted vision at that gate: about 70 seconds per unit and 99 percent correct judgments, with a full-frame photograph archived afterwards.
A long list on a scattered point map has a low ceiling
The manual problem is not visibility, it is the phrase one by one. Several dozen items mean looking down, looking up and recording, hundreds of times a shift. Attention decays, and the decay is when a miss happens. The underside adds pressure: uneven, close to the floor, and the easiest part to skip because checking it means kneeling.
Stable conveyance comes before any robot view
These assemblies are heavy with uneven undersides, so ordinary conveying lets the part drift inside the inspection window. A slat conveyor carries it in and out without that movement. It looks like logistics, but it is the premise of the vision design: camera positions are fixed, and once attitude drifts the imaging conditions change and every threshold has to be loosened. The machine reaches 11.23 m x 5.04 m x 3.14 m, so floor space and traffic routes belong in the proposal.
The robot count is a consequence of the point map
The reusable lesson here is that robot count follows point distribution, not a standard configuration. Points grouped on one side can be covered by one robot; points spread over wide faces need several sharing the load. Two criteria decide it: whether one machine's envelope reaches every point, and how much of that envelope is left as travel time once coverage is achieved.
The same logic transfers to other large assemblies. Draw the point map before discussing robots; reverse the order and you get a system that keeps moving while some points were never imaged. The longer the checklist, the earlier a project should lean toward multiple machines rather than buying one and adding later.
What is photographed and what is kept
The sequence has four steps: barcode read on entry to bind the inspection task; robots with 20-megapixel cameras shooting preset views while oversized high-brightness light and high-brightness bar light keep a large part evenly lit; item-by-item judgment of wrong and missing installation; and a full-frame archive shot once every point is done.
The archive shot is often treated as a formality, yet it is the primary evidence. The item verdict says which component failed; the full frame says what the unit looked like. Only together do they support a review.
Where 70 seconds and 99 percent hold up
The cycle of about 70 seconds and 99 percent correct judgment are subject to site acceptance results. They rest on three things at once: stable conveyance, enough light, and a fixed checklist. Let one slip and the real miss and false-call rates have to be re-measured from site records instead of quoted from project data.
Doubts before the point map is drawn
When is one robot enough?
When points are concentrated on one side and a single envelope reaches all of them. The criterion is the point map, not the size of the assembly.
Why a slat conveyor?
Heavy units with uneven undersides need a carrier that holds attitude inside the inspection window. With stable attitude, fixed views mean something.
Is the 11.23 m station mandatory?
It is this project's machine size, set by the assembly and its point distribution. Another workpiece needs a fresh calculation.