Aging changes the probability of failure of components. This change in probabilities depends on environmental, functional and inherent factors. Statistical survival models represent this dependence. During some periods of a component's lifetime, it can be considered perfect, and thus reducing the search space for the diagnostic engine. At other times a component must be considered vulnerable. Depending on their probability of failure at any given time, components can be divided into two sets: vulnerable components and perfect components. This partitioning can be exploited to improve the performance of a diagnostic engine by automatically generating a system description (SD) each diagnostic session. The timely SD allows vulnerable components to fail and considers other components ideal.
Unexpected failures signal unusual operating conditions possibly
due to the failure of other subsystems. A surprise index estimates
the surprise associated with the failure of each failed component
and further tests are performed when a failure is highly surprising.