Recently, digital instrumentation and control systems have been increasingly installed for important safety functions in nuclear power plants such as the reactor protection system (RPS) and the actuation system of the engineered safety features. Since digital devices consist of not only electronic hardware but also software that can control microprocessors, the functions specific to digital equipment such as self-diagnostic functions have been becoming available. These functions were not realized with conventional electric components. On the other hand, it has been found that it is difficult to model the digital equipment reliability in probabilistic risk assessment (PRA) using conventional fault tree analysis technique. OECD/NEA CSNI Working Group of Risk Assessment (WGRisk) set up the task group DIGREL to develop the basis of reliability analysis method of the digital safety system and is now discussing about several issues including quantitative dynamic modeling. This paper shows that, taking account of the relationship among the RPS failures, demand after the initiating event, detection of the RPS fault by self-diagnostic or surveillance tests, repair of the RPS components and plant shutdown operation by the plant operators as a stochastic process, the anticipated transient without scram (ATWS) event can be modeled by the event logic fault tree and Markov state-transition diagrams assuming the hypothetical 1-out-of-2 digital RPS.