Quantitative reliability estimation of a computer-based motor protection relay using Bayesian networks

Abstract

A quantitative reliability estimation of a computer-based motor protection relay is presented. The evidence used for the estimation consists of expert judgements on the development process and estimated operational experience of the computer-based system. The framework of Bayesian modelling and Bayesian networks was applied throughout the assessment. A prior reliability estimation of the motor protection relay was built using the expert judgements on the product development process. The prior estimation was updated by taking into account the estimated operational experience for successive software versions. Prior estimations for possible reliability changes between different software versions were included to the assessment using the expert judgements on the version management of the motor protection relay. The Bayesian networks seem to provide an excellent way of performing quantitative reliability estimations of computer-based systems. Bayesian networks are particularly useful when different kinds of evidence is introduced and combined in the same assessment. The assessment method described in the paper gives informative posterior probability distributions for the failure rates of different software versions. The methodology can be used as a communicative tool between different participants debating on the reliability of a system and as a follow-up of reliability trend during the life cycle of a computer-based system. © Springer-Verlag Berlin Heidelberg 2003.