Interface contracts beat pretty HMI screens

Most of the downtime I see on high-speed packaging lines is not fancy PLC bugs. It is mismatched assumptions between machines: the case packer expects the starwheel to be home before the conveyor jogs, the OEM skid drops its permit on a minor fault, or CIP steals plant air mid-cycle. The fix is not a nicer HMI. It is a clear interface contract at the system level.

What has worked for us is a one-page interface control document per handoff. List signal names, who owns start/stop, permissives, fault latching, timing tolerances, and what each device does on comms loss. Then prove it with a stopwatch, a wire break, and a network pull, not just at FAT. We also standardized a small set of interlock patterns and fault classes so techs know what to expect.

Since doing this, nuisance stops and startup finger-pointing have dropped, and brownfield swaps are cleaner because the boundaries are explicit. For plants mixing legacy PLCs with new EtherNet/IP gear, how are you formalizing interface behavior? If you use an ICD, what single clause has saved you the most downtime?