REAL-TIME EMPIRICAL BAYES MANUFACTURING PROCESS MONITORING FOR CENSORED-DATA

Because of the current work in manufacturing electronics which links h igh-speed test sets to host computers, it is now possible to collect a nd analyse measurements seconds after a product is tested. However, so me of the actual measurements may often not be collected. This paper d escribes a technique to monitor right-censored, left-censored or both right- and left-censored manufacturing data. Also, the technique is ap plicable to 'interval' censored data; data that is censored when it oc curs in an interval, with uncensored observations above and below the interval. This'interval' censoring can occur when automatic test equip ment cannot transmit data and test at the same time. To increase produ ction volume through the test area, measurements are often collected o nly when a test fails. The method described is designed to monitor a m anufacturing process in real-time with the goal of identifying process changes as they occur. An empirical Bayes model of the process respon se parameters is developed for the situation when censoring is present . Estimates for the model are developed, a strategy for implementing t he model is given and an example is presented which illustrates its us e to monitor a process.