Modeling of Complex Redundancy in Technical Systems with Bayesian Networks

Abstract

Redundancy is a common approach to improve system reliability, availability and safety in technical systems. It is achieved by adding functionally equivalent elements that enable the system to remain operational even though one or more of those elements fail. This paper begins with an overview on the various terminologies and methods for redundancy concepts that can be modeled sufficiently using established reliability analysis methods. However, these approaches yield very complex system models, which limits their applicability. In current research, Bayesian Networks
(BNs), especially Dynamic Bayesian Networks (DBNs) have been successfully used for reliability analysis because of their benefits in modeling complex systems and in representing multi-state variables. However, these approaches lack appropriate methods to model all commonly used redundancy concepts. To overcome this limitation, three different modeling approaches based on BNs and DBNs are described in this paper. Addressing those approaches, the benefits and limitations of BNs and DBNs for modeling reliability of redundant technical systems are discussed and evaluated.