Dependent failure—a multiple-state stochastic description

Abstract

This paper describes the relation between objective variability and subjective uncertainty and the influence of both features in creating the effect of dependent failure, i.e. the enhanced occurrence frequency for multiple failure events in redundancy structures, known from operational experience. As an extension of the state-of-knowledge correlation model to the case of a non-homogeneous population with several possibilities of failure behaviour, a unified approach is developed which identifies and models both subjective uncertainty and variability of the failure behaviour in a stochastically consistent way. Both types of uncertainty are represented by different aspects of the new model and hence kept separate.