Sufficient statistics process control: An empirical Bayes approach to process control

High volume, highly automated information intensive, short cycle manufacturing systems severely tax most conventional statistical process control techniques. To meet this new manufacturing domain’s control requirements, a new approach is needed. This paper presents such a process control procedure, sufficient statistics process control (SSPC). By drawing on empirical Bayes techniques, SSPC models the time sequence of the process while simultaneously reducing to a few sufficient statistics the large volume of incoming data. As a result, it provides real time, on-line quality control. The paper discusses the conceptual and mathematical foundations for SSPC. Its operation is illustrated through an example. Finally, the paper concludes with a discussion of the limitations of SSPC.