Reliability Limits of Programmable Protection Systems (R&D on Reactor Design – Safety)

Programmable systems are now an established part of nuclear power Control & Instrumentation (C&I) solutions. Their design and justification are significant engineering challenges, but they offer functionality well in excess of the legacy hard wired systems they can replace. This project will explore the options for justifying the use of programmable C&I for very high reliability, and whether this could support an argument for a reactor design having only programmable control and protection systems.

Part of BEIS' Nuclear Innovation Programme. There are no general techniques that are able to demonstrate reliability for software, which differs from hardware becuase all faults are systemmatic. There are some that work in some circumstances and these should be developed urgently.

To date... Scientific demonstration of system reliability requires further research, since the traditional methods based on random failure assumptions do not apply for software. Protection systems are amongst the most safety critical known to man, and require the highest levels of assurance. Statistical testing and formal methods offer the only scientifically defensible approaches to this, and there is potential for these methods to be combined to enable high confidence. Advances in demonstration of software diversity in multi-channel ('N-version') architectures have been made but are insufficient to justify high confidence in these systems.