A study on quantitative V and V of safety-critical software

Recently practical needs have required quantitative features for the software reliability for Probabilistic Safety Assessment which is one of the important methods being used in assessing the overall safety of nuclear power plant. But the conventional assessment methods of software reliability could not provide enough information for PSA of NPP, therefore current assessments of a digital system which includes safety-critical software usually exclude the software part or use arbitrary values. This paper describes a Bayesian Belief Networks based method that models the rule-based qualitative software assessment method for a practical use and can produce quantitative results for PSA. The framework was constructed by utilizing BBN that can combine the qualitative and quantitative evidence relevant to the reliability of safety-critical software and can infer a conclusion in a formal and a quantitative way. The case study was performed by applying the method for assessing the quality of software requirement specification of safety-critical software that will be embedded in reactor protection system