Bayesian Networks for Dependability Analysis: an Application to Digital Control Reliability

TL;DR

This paper explores the use of Bayesian Networks as a probabilistic tool for dependability analysis, demonstrating their advantages over traditional methods through a real-world PLC reliability case study.

Contribution

It introduces Bayesian Networks as a formalism for dependability analysis and shows how they can address limitations of traditional combinatorial methods like Fault Trees.

Findings

BN effectively models dependability issues.

BN overcomes limitations of Fault Trees.

Application to PLC reliability demonstrates practical benefits.

Abstract

Bayesian Networks (BN) provide robust probabilistic methods of reasoning under uncertainty, but despite their formal grounds are strictly based on the notion of conditional dependence, not much attention has been paid so far to their use in dependability analysis. The aim of this paper is to propose BN as a suitable tool for dependability analysis, by challenging the formalism with basic issues arising in dependability tasks. We will discuss how both modeling and analysis issues can be naturally dealt with by BN. Moreover, we will show how some limitations intrinsic to combinatorial dependability methods such as Fault Trees can be overcome using BN. This will be pursued through the study of a real-world example concerning the reliability analysis of a redundant digital Programmable Logic Controller (PLC) with majority voting 2:3