Software reliability growth model of reactor protection system based on ratio analysis of critical faults

The application of digital instrumentation and control systems in Nuclear Power Plants (NPPs) provides a series of advantages, but it also raises challenges in the reliability analysis of safety-critical systems in the NPPs. Software testing is one of the most significant processes to assure software reliability, and the safety-critical systems of NPPs are sensitive to the severity of software faults, especially the critical faults that infect the system function greatly. Previous software reliability models related to the analysis of fault severity were mostly based on the assumption of different severities. In this paper, the fault severity data collected during the software test process were used for modeling the test process with a Software Reliability Growth Model based on a non-homogeneous Poisson process. The mean value function was derived by considering the ratios of critical and non-critical faults and was named as “Ratio of Critical-Faults model” (RCF model). The fault data collected while developing the safety-critical system were used to validate this model. According to the analysis, RCF model had fitting abilities similar to that of the Goel-Okumoto model and Inflection S-shaped model whereas the prediction effect of the RCF model was better than that of these two models, especially when little data were collected, which could be used to determine the release time of the software.