EFFECTIVE MULTICRITERIA REDUNDANCY ALLOCATION VIA MODEL-BASED SAFETY ANALYSIS

Abstract The design of safety critical systems that meet dependability (e.g. safety and reliability) criteria is a hard combinatorial problem and there is a need for tools that make use of systematic optimisation algorithms to provide technological assistance. This paper describes a method that combines a Pareto-based multi-objective genetic algorithm with an automated safety analysis process. The aim is to automatically evolve an initial design, which does not meet dependability requirements, to produce a selection of trade-off designs that fulfil such requirements with minimal costs. For the purposes of demonstration, the general problem of dependability optimisation has been constrained to one of optimal redundancy allocation and the approach has been applied to a benchmark problem with results that represent improvement to those reported in earlier works.