Advanced control and instrumentation (C&I) software performs critical safety functions, like system shutdown, that are crucial to the future of nuclear power – but for a long time it has been difficult to prove its reliability. Researchers at Bristol have helped to fill this gap with Statistical Testing that provides the evidence needed to support the licensing of this increasingly complex software and protect stations that produce almost a fifth of the UK’s energy supply.

Conquering complexity

Over the years, major accidents such as those at the Three Mile Island, Chernobyl and Fukushima nuclear power plants have demonstrated the need for independent and robust demonstration of safety required to obtain an operating licence.

The problem the industry faces is that modern nuclear plants use significantly more software-based C&I than previous generations to keep the public safe. Such C&I is highly complex, which means that traditional random failure reliability models which were originally developed for hardware cannot be used to test reliability.

The research group at Bristol led by Dr John May has helped to solve this issue by advancing Statistical Testing to provide a new approach to the reliability assurance of programmable digital systems.

The research was conducted within University’s South West Nuclear Hub and mostly funded by the C&I Nuclear Industry Forum, the UK’s leading body responsible for advancing nuclear C&I safety in the UK. Specific breakthroughs have included:

• Showing how Statistical Testing can be defined to simultaneously satisfy reliability targets and traditional test adequacy criteria.
• Enhanced understanding of the risks associated with using ‘smart devices’ with embedded software in nuclear plants, such as smart sensor transmitters and electrical protectionve devices.
• Improvements to failure diversity techniques needed to protect redundant channel architectures used to protect software-based C&I systems.
• A world-first model capable of estimating system reliability based on software executing on simulated hardware platforms, raising the possibility of massively concurrent testing. As well as test acceleration that is capable of breaking all current limits for reliability demonstrations.

Impact on UK and international nuclear infrastructure

As a result of Bristol’s research, Statistical Testing has now become a key component in the governance and regulation of critical nuclear systems.

In particular, the research has been a major factor behind the Office for Nuclear Regulation’s (ONR’s) decision in the UK to recommend the use of Statistical Testing as part of its Technical Assessment Guide for the design of complex C&I safety systems and equipment. The ONR has also incorporated the research findings into its Safety Assessment Principles (SAPs) that form the highest level of technical policy reflecting good practice in the UK nuclear sector.

EDF Energy, the energy company that runs the UK’s nuclear power infrastructure, has stated that: “Statistical Testing has given us key additional evidence needed for regulatory approval of these increasingly complex systems.”

Governments around the world are also commissioning programmes of research into Statistical Testing to inform policy, with investigations already under way by the United States Nuclear Regulatory Commission, the Korea Atomic Energy Research Institute in South Korea and The Nuclear Power Institute of China. In the future, the UK and many other countries intend to maintain a significant level of nuclear power production to provide a baseline carbon-neutral energy source and help meet net zero targets. This will require the construction of new plants, possibly including smaller and more advanced modular reactor designs as well as modifications to extend the life of existing plants. In all these cases, C&I software plays an important safety role and the research carried out at Bristol will help with validation for decades to come.