Hello Everyone

I'm exploring better ways to diagnose behaviour in robot/cobot cells during commissioning and would really value hearing from integrators, control engineers, project managers, and commissioning leads.

A few questions I'm trying to understand from people who have worked on real deployments:

1. Which types of cells have you found to be the most difficult to commission? (e.g. welding, machine tending, palletizing, multi-robot cells, etc.)
2. When a commissioning project runs late, what are the most common causes? (Is it usually a mismatch between the design model and the physical installation, or something else?)
3. When something goes wrong in a new cell, how do teams figure out what physically happened? (What tools or data do you rely on? How long does that process typically take?)
4. How often do engineers need to return onsite after commissioning to troubleshoot issues?
5. If you could rewind and replay the spatial history of the cell — seeing exactly where robots, people, and objects moved over time — would that help diagnose problems?

Any brief examples, war stories or lessons learned would be incredibly helpful.

Thanks in advance!

 

Multi-robot cells are usually the toughest. Every interaction creates another possible failure mode. Welding cells can also be difficult. Heat, fixtures, and part variation all affect repeatability.

Most schedule slips come from integration issues. Mechanical tolerances change. Sensors need repositioning. Safety circuits require validation. PLC and robot handshakes often need refinement. The digital model rarely matches the real installation perfectly.

Troubleshooting starts with controller logs, PLC diagnostics, HMI alarms, and network diagnostics. Video recordings also help. The hardest part is rebuilding the exact sequence of events. That can take minutes or several days.
Post-commissioning visits are common. Production always exposes cases nobody predicted during FAT or SAT.
One lesson I've learned is to check the basics early. Wiring faults, loose terminals, and connector problems can look like software bugs. Understanding common wiring harness connector types has saved me time more than once.
A synchronized replay of the cell would be valuable. Seeing robot motion, machine states, operator movement, and part locations together would make root-cause analysis much faster. Intermittent faults would also become much easier to explain.