Modeling the Reliability of Complex Systems with Multiple Data Sources: A Case Study on Making Statistical Tools Accessible to Engineers

摘要

Estimating the reliability of complex systems, such as Department of Defense weapons, often involves a meta-analysis using multiple data sources, including expensive and destructive full system tests, as well as nondestructive subsystem and component-level tests. Using statistical methodology developed by the Statistical Sciences Group at Los Alamos National Laboratory, a statistical engineering process was developed for estimating and predicting future reliability of individual units and of a population of units in a stockpile. A multiphase software tool, SRFYDO, was then developed to make this process accessible and understandable to the system engineers who need to perform these analyses. In this article, we present a short overview of the method but focus on how the software was developed with the goal of assisting engineers in using a sequence of statistical tools to gain improved understanding and effectively answer a high-consequence question with their analyses.