Abstract
Abstract
A worn robot gripper keeps running and looks healthy until parts start going out of tolerance, so maintenance teams need a warning before that point. We compared a proprietary information monitor, which checks whether the same command still gives the same response, with standard cumulative sum (CUSUM) and exponentially weighted moving average (EWMA) control charts, on the same simulated pick-and-place machines over about 160 hours and 287,600 production cycles. The information monitor caught gripper wear on every machine at 1.2 to 1.45 times normal variation, before any bad parts appeared. CUSUM and EWMA only caught the wear at about 3 times normal variation, when nearly every part was already bad. No method raised a false alarm on healthy machines. For operations, this means gripper health can be monitored and acted upon while parts are still good, instead of reacting after scrap.