Decision Making with Competing Objectives
Nuclear power generates around 15% of the UK’s energy. It is a vital supplier of stable, low-carbon electricity. At the heart of EDF's Advanced Gas-cooled Reactors (AGRs), thousands of interlocking graphite bricks ensure core stability and regulate the nuclear chain reaction.
As these bricks degrade over time, understanding their condition is essential. But inspections require shutting down the reactor, interrupting power generation and incurring significant costs. EDF must combine rigorous safety requirements with making the most efficient use of targeted operational data.
We developed CrackSmith, a bespoke software tool that uses Bayesian statistical modelling to predict graphite degradation. This is now embedded in EDF’s safety case development, helping to ensure safe, continuous reactor operation. Our work supports the extension of reactor lifetimes, contributing to Britain’s transition to clean energy by 2030.
Understanding risk with ageing core components
The Challenge
The condition of the graphite bricks is critical to how long an AGR can safely operate. As these bricks age, small cracks form. These are not a safety concern at first, but if the cracking progresses too far, it could compromise the reactor’s structural integrity or its ability to shut down in an emergency. Limits on acceptable cracking are a vital element of the safety case and dictate the reactor’s lifetime.
To meet regulatory standards, EDF must demonstrate to the Office for Nuclear Regulation (ONR) that each reactor remains safe under even extreme scenarios, such as earthquakes or major electrical faults. These safety cases rely not just on inspection data, but also on robust predictions of how degradation will evolve.
Although sophisticated imaging allows internal inspection of bricks, each inspection requires a reactor shutdown – an expensive disruption to a key source of low-carbon electricity. With data limited, EDF need reliable models to assess which bricks are most at risk and how cracks may develop over the coming months and years of operation, ensuring safe operation while meeting strict regulatory requirements.
UTILISING EXPLAINABLE AI TO UNCOVER VULNERABILITY DRIVERS
The Solution
Smith Institute developed CrackSmith, a predictive model that estimates the number and severity of graphite brick cracks. By combining limited core observation data with knowledge of the physical behaviours of the graphite bricks, the model constructs both current and possible future reactor states.
At the heart of CrackSmith is a Bayesian statistical framework, which enables predictions to be continuously refined as new data becomes available. It quantifies uncertainty in crack progression down to the individual brick while integrating expert physical understanding. This avoids the overly conservative conclusions of a purely statistical model and prevents premature reactor decommissioning.
CrackSmith uses advanced high-efficiency sampling techniques to generate and analyse vast numbers of possible reactor scenarios. From this, EDF can clearly demonstrate risk and uncertainty to the regulator, optimise the timing of inspections, and reach informed conclusions about reactor life extension.
Enabling continued safe operation of Britain's nuclear fleet
The Impact
CrackSmith plays an integral role in EDF’s safety cases for all eight AGRs at Hartlepool, Heysham and Torness, informing the conclusions that underpin the regulatory submissions that demonstrate the reactor’s continued safe operation. As well as the main forecasts, we deliver conditional “what if” forecasts that help EDF answer key questions about safe operating windows relative to allowable crack limits, supporting confident, data-driven decisions.
This capability enables the continued operation of Britain’s AGR fleet, a critical contributor to Britain’s clean energy supply. As outlined in the Clean Power 2030 report, nuclear power is essential to achieving a low-carbon, reliable electricity system in the years ahead. By supporting safe extension of reactor lifetimes, our work helps ensure that nuclear power remains a stable pillar of Britain’s energy mix for years to come.
CUSTOMER TESTIMONIALS
Smith Institute have shown unique capability in two key areas; they have been able to generate a robust bespoke approach to a difficult problem in a short timescale but equally important they have been able to explain their approach in a very clear, transparent and understandable set of reports. We had already assessed that Smith Institute were a very highly competent organisation to deliver the mathematical/modelling work we required. Their ability to clearly explain to key stakeholders their work transformed the task from a ‘very good piece of analysis’ to an ‘outstanding success’
