Statistical Trend Analysis Methodology for Rare Failures in Changing Technical Systems

摘要

A statistical trend analysis methodology based upon the evaluation of actuarial data on the temporal spacings between failure events, within a certain category is presented. It applies to the middle ground between the frequent failures in a population of a large number of identical units and the extremely rare events for which no actuarial information is available. It analyzes sparse data assembled while a technology, a system, or a component continue to change. The method accounts fully for failures during the technology evolution (with high rates if they were prevalent at early times) but also allows for system improvements to properly affect the estimates of the most recent failure rates, if such improvements are demonstrated in the actuarial data on failure spacings.