Speakers
Description
Within the context of deploying future fusion power devices, the breeder blanket system remains at a very low technology readiness level (TRL) and therefore requires substantial research and development to ensure long-term robustness, reliability, and safety. Breeder blankets are highly complex, tightly coupled systems involving neutronics, thermal hydraulics, materials behaviour, and tritium generation and transport. This complexity presents significant challenges for experimental development, particularly where physical testing is costly, time-consuming, or carries inherent safety risks. As a result, advanced digital approaches are increasingly critical to support the design, operation, and qualification of such systems.
For modern complex systems such as breeder blankets, the development and deployment of a digital twin is especially valuable. A digital twin provides a dynamic virtual representation of a physical system, enabling prediction of system behaviour, exploration of design changes, and interrogation of operational scenarios with minimal risk to the underlying hardware. When underpinned by low-code, surrogate modelling approaches, digital twins can be rapidly developed, updated, and deployed, allowing researchers and engineers to integrate experimental data, simulations, and uncertainty quantification in a flexible and scalable manner.
As part of the UK Atomic Energy Authority’s LIBRTI (Lithium Breeding Tritium Innovation) programme, the University of Manchester has undertaken a proof-of-concept digital twin study using a Gas Driven Permeation System (GDPS) as an exemplar device. The GDPS was selected due to its relatively simple physical design while still providing rich, high-value data inputs and outputs that are directly relevant to fusion fuel cycle research. This makes it an ideal platform for demonstrating digital twin concepts applicable to more complex breeder blanket systems. In parallel, a new tritium transport modelling capability has been developed based on Bayesian inference techniques. This approach enables rapid prediction of tritium mobility and associated uncertainties within material specimens, even in cases where only limited microstructural information is available. The model has been validated using experimental data obtained from GDPS permeation studies.
This work presents preliminary results on a low-code digital twin architecture that leverages open-source software tools alongside NVIDIA Omniverse for system integration and visualisation. The architecture has been applied in a proof-of-concept implementation to a real fusion fuel cycle system through full GDPS digitisation. This includes integration of the Bayesian inference-based tritium transport model to capture tritium mass transfer behaviour relevant to GDPS permeation experiments. The digital twin demonstrates several key capabilities, including remote control of GDPS acquisition parameters, deployment of existing open-source surrogate models, and prediction of material properties derived from experimental permeation data.
Finally, this work is placed within the broader context of the LIBRTI programme. The extensible digital twin architecture provides a clear pathway towards full digitisation of LIBRTI activities, enabling improved decision-making for breeder blanket test module design, accelerating innovation in breeder concepts, and supporting enhanced preventative maintenance strategies. Ultimately, this approach offers a safer, more efficient framework for experimentation and development as fusion technology progresses toward deployment.
| Speaker affiliation | The University of Manchester |
|---|
