129 USNRC Research Program and Preliminary Results of PRA Modeling of Digital I&C Systems (PSAM-0329)

Nuclear power plants rely on instrumentation and control (I&C) systems for monitoring, control, and protection. The Probabilistic Risk Assessment (PRA) modeling of digital I&C systems is important to support a risk-informed approach to evaluating and selecting digital systems. However, there is a lack of an acceptable approach for modeling digital systems in PRAs. To address these issues and for the United States Nuclear Regulatory Commission (USNRC) to independently assess risk-informed digital system applications, USNRC's Office of Nuclear Regulatory Research is investigating several methods for the development of risk insights for digital systems. These methods include: a) traditional static fault tree and Markov models supported by traditional failure modes and effects analysis (FMEA) and data analysis; b) Markov models supported by advanced digital system test based methods; and c) non-traditional dynamic methods (e.g., dynamic flowgraph methodology).

This paper describes USNRC's research program and preliminary results of developing a probabilistic approach for modeling failures of digital I&C systems using traditional PRA methods (static fault tree and Markov models supported by traditional FMEA and data analysis) that can be integrated with a PRA. The research program consists of the following major tasks: (1) review the approaches on reliability modeling of digital systems that are used by non-nuclear industries, (2) obtain adequate information about the behavior of a digital system using FMEA and dependency analysis of the system so that a model of its failure behavior can be developed, (3) develop a failure rate database for digital system hardware, (4) develop and quantify a suitable reliability model for the hardware of a digital system, (5) develop and quantify methods for modeling software failures of a digital system, (6) integrate the hardware and software reliability models to quantify the reliability of a digital system, (7) integrate the combined model (both hardware and software) with the PRA, and (8) documentation of research work. The method development includes performing a case study involving a digital feedwater control system at an existing nuclear power plant. Tasks 1 and 3 are completed, and Tasks 2 and 5 are currently in progress. The major elements of this project are expected to be completed by 2008.