Robust Fault Injection For Quantitative Reliability and Risk Assessment: A Demonstration on a Benchmark IC System

摘要

The introduction of new instrumentation and control (I&C) systems into nuclear power plants has raised concerns regarding the possibility that the fielding of these systems, may introduce unanticipated failure modes that have been overlooked during design of the safety system function. To prevent such situations from arising, there is a great need for these systems to be dependable across a wide spectrum of perceived threats, faults and failures. These systems must be able to detect faults and initiate appropriate mitigation actions to ensure overall safety. In order to test the fault tolerance features of an I&C system, some form of systematic fault insertion is needed. In this paper, a new technique for robust fault injection in operational digital I&C systems is proposed. This method makes use of in-circuit emulators, which are debugging tools used by designers of embedded systems, to achieve directed fault injections. We show how this method supports data collection for the statistical estimation of specific parameters required by analytical reliability, safety, and PRA models. The advantages of an in-circuit emulator-based fault injector include improved observability and controllability, low intrusiveness, and the ability to realize multiple fault models.