Test based reliability quantification method for a safety critical software using finite test sets

Abstract

Software is currently used within nuclear power plants (NPPs) to digitalize many instrumentation and control (I&C) systems. To guarantee the safety of the NPP, the reliability of the software must be properly quantified. In this study, we propose a novel method for software reliability quantification. The method identifies and arranges possible internal states of the software that can occur in actual use. Based on a specific internal state, possible input sets (combination of single values of each input variable) are applied sequentially. In this process, the assigned range of each variable, correlation between variables, characteristics of analog-to-digital converter (ADC), and plant dynamics are considered to identify the possible states of each variable. The effectiveness of the proposed method is demonstrated via a case study for a trip logic in a reactor protection system (RPS). Compared with existing test-based methods, the proposed method can shorten test execution time and eliminate uncertainties derived from random sampling of input values from the operation profile. Moreover, this method can provide a number of test sets required for an exhaustive testing.