Design and implementation of systems for monitoring lifetime data

Abstract

We discuss the problem of monitoring data streams corresponding to a sequence of lifetime tests with censoring. The parameters of the lifetime distribution undergo abrupt changes of unknown magnitude at some unknown points in time. Situations of this type are common in the areas of ongoing reliability tests (ORT) and analysis of warranty data. Themain technical difficulty of the detection problem is related to the dynamic nature of the data acquisition process, as the information collected at consecutive points in time generally affects data for a broad range of lifetime tests in progress. We discuss issues and approaches related to monitoring such types of data and give examples related to reliability monitoring of computer components. The emphasis is on a Cusum - based approach and issues related to its implementation in systems for warranty data monitoring deployed in the IBM Personal Systems Division. © Springer-Verlag Berlin Heidelberg 2012.