Abstract:
|
In mass-production environment systems for reliability monitoring typically act as search engines that scan the data on a regular basis and select data streams that indicate presence of unfavorable conditions, such as unacceptable parameters of survival characteristics of monitored part populations. Furthermore, the selected data streams are prioritized based on interests of various groups of users. In this paper we will discuss statistical issues related to survival models used in the process of monitoring, including false alarm/sensitivity trade-offs and design of alarm attributes. We illustrate the methodology based on examples from supply chain operations, including ongoing reliability tests and warranty data monitoring.
|