Ceiling mounted cobot camera checks a safety critical connector inside 60 second takt

Contract subassembly of automotive engine cradles, run inside a third party logistics operation · Inspection · Vendor published

The job

On a horseshoe shaped assembly line in Youngstown, Ohio, Comprehensive Logistics had to confirm on every engine cradle subassembly that a steering gear wire harness connector was fully seated and locked, a fault the customer rates level 8 on the severity scale because it can cause loss of power steering. A fixed multi camera rig could not reach the tight spots repeatably, and a probe style robot also missed the repeatability target. A ceiling mounted UR10 carrying a vision camera now moves between inspection points, including underneath the assembly, photographs each connection and retracts to a home position. If a single point fails it can be reinspected on its own rather than rerunning the whole sequence, which is what keeps the cell inside takt.

The shop

100 to 500 employees, published-by-vendor as a range. Two or three shift operation running round the clock, published-by-vendor. Exact headcount: not published.

The equipment

Universal Robots UR10, ceiling mounted, carrying a vision camera; no gripper, the arm is a camera positioner. Turnkey application built by Automation Zone, an Ohio based UR certified systems integrator, with remote dial in to add or change inspection points. Conveyor assembly line with a 60 second station takt. Camera make and model, lens and lighting: not published. No light curtains or guarding were required, published-by-vendor.

The published numbers

Inspection cycle: approximately 32 seconds, inside a line cycle time of about 48 seconds within a 60 second station takt, published-by-vendor. Quality: 100 percent quality claimed, against a stated industry wide manual inspection effectiveness of about 80 percent, published-by-vendor. Payback: about 7 months for the entire turnkey application and probably less than 4 months for the robot hardware alone, published-by-vendor, with the superintendent noting the calculation excludes avoided customer quality penalties. Uptime: two years of operation across two and three shift patterns with no breakdowns, published-by-vendor. Application cost: not published. Defects actually caught: not published. Headcount change: not published. False reject rate: not published.

Where a source did not publish a figure, this entry says so rather than guessing.

What we would check before believing it applies to you

Moving one camera on an arm beats installing many fixed cameras when the inspection points are in tight or shadowed places, and this cell shows the real tradeoff plainly: it burns about 32 of a 48 second window to reach points a fixed rig could not see at all. The detail most buyers miss is the reinspect behavior, where a single failed point reruns on its own instead of restarting the sequence, and you should make your integrator commit to that in writing, because a full sequence rerun will push a cell like this straight out of takt. The 80 percent manual effectiveness figure is an industry rule of thumb quoted by the customer, not a measurement of this line, so it flatters the before case considerably. Use the 7 month application payback, not the 4 month hardware payback, since the hardware figure excludes the integrator, engineering, camera and fixture that make the cell work at all. Most importantly, the whole justification rests on avoiding customer penalties for a level 8 safety defect, which is a value driver a typical job shop does not have.

Sources