Evaluation of Nuclear Safety-Critical Software Reliability of DCS

This paper tries to choose Bayesian Belief Networks（BBN） method to build the evaluation models. The stage evaluation models are constructed based on NUREG-0800 BTP7-14. It finished the software real-time evaluation in stages using 13 first-level indexes, 74 second-level indexes, 326 third-level indexes.HUGIN software is chosen as the Bayesian Networks Analysis tool. Predictive inference and sensitivity analysis are given to the testing object. After testing, the degree of compliance with standards at different stages of the life cycle can be obtained. The software reliability is 98.84% by using the evaluation model, thus, the reliability of the failure-free software operation can be calculated from the quantitative aspect. After predictive inference, sensitivity analysis, the weaknesses in software development process can be found from the qualitative aspect. It can be a base for evaluating the reliability and safety of the DCS.