The pursuit of nuclear fusion as a clean, sustainable energy source represents one of the most challenging scientific and engineering goals of our time. Fusion promises nearly limitless energy without carbon emissions or long-living radioactive waste.
However, achieving practical fusion energy requires overcoming significant challenges. These come from the heat generated by the fusion process, the radiation produced, the progressive damage to materials used in fusion devices and other engineering hurdles. Fusion systems operate under extreme physical conditions, generating data at scales that surpass the ability of humans to analyse.
Nuclear fusion is the form of energy that powers the Sun. Existing nuclear energy relies on a process called fission, where a heavy chemical element is split to produce lighter ones. Fusion works by combining two light elements to make a heavier one.
While physicists are able to initiate and sustain fusion for variable periods of time, getting more energy out of the process than the energy supplied to power the fusion device has been a challenge. This has so far prevented the commercialisation of this hugely promising energy source.
Artificial intelligence (AI) is emerging as a powerful and essential tool for managing the inherent challenges in fusion research. It holds promise for handling the complex data and convoluted relationships between different aspects of the fusion process. This not only enhances our understanding of fusion but also accelerates the development of new reactor designs.
By addressing these hurdles, AI offers the potential to significantly compress timelines for the development of fusion devices, paving the way for the commercialisation of this form of energy.
AI is reshaping fusion research across academic, government and commercial sectors, driving innovation and progress toward a sustainable energy future. For example, it can play a transformative role in addressing the challenges of developing materials for fusion reactors, which must withstand extreme thermal and neutron environments while maintaining structural integrity and functionality.
By connecting datasets from different experiments, simulations and manufacturing processes, AI-driven models can generate reliable predictions and insights that can be acted on. A form of AI called machine learning can significantly accelerate the evaluation and optimisation of materials that could be used in fusion devices.
These include the doughnut-shaped vessels called tokamaks used in magnetic confinement fusion (where magnetic coils are used to guide and control hot plasma – a state of matter – allowing fusion reactions to occur). The superheated plasma can damage the materials used in the interior walls of the tokamak, as well as irradiating them (making them radioactive).
Machine learning involves the use of algorithms (a set of mathematical rules) that can learn from data and apply those lessons to unseen problems. Insights from this form of AI are critical for guiding the selection and validation of materials capable of enduring the harsh conditions within fusion devices. AI allows scientists to develop detailed simulations that enable the rapid evaluation of materials performance and their configurations within a fusion device. This helps ensure long-term reliability and cost efficiency.
AI tools can help narrow the range of candidate materials for testing, characterise them based on their properties and perform real-time monitoring of those installed in fusion reactors. These capabilities enable the rapid screening and development of radiation-tolerant materials, reducing reliance on traditional, time-intensive approaches.
Controlling plasma
AI also offers a way to better control the plasma in fusion reactors. As discussed, a key challenge in magnetic confinement fusion is to shape and maintain the high-temperature plasma within the fusion device, often a tokamak vessel.
However, the plasmas in these machines are inherently unstable. For example, a control system needs to coordinate the tokamak’s many magnets, adjust their voltage thousands of times per second to ensure the plasma never touches the walls of the vessel. This could lead to the loss of heat and potentially damage the materials inside the tokamak.
Researchers from the UK-based company Google DeepMind have used a form of AI called deep reinforcement learning to keep the plasma steady and be used to accurately sculpt it into different shapes. This allows scientists to understand how the plasma reacts under different conditions.
Meanwhile, a team at Princeton University in the US also used deep reinforcement learning to forecast disturbances in fusion plasma known as “tearing mode instabilities”, up to 300 milliseconds before they appear. Tearing instabilities are a leading form of disruption that can occur, stopping the fusion process. They happen when the magnetic field lines within a plasma break and create an opportunity for that plasma to escape the control system in a fusion device.
My own collaboration with the UK Atomic Energy Authority (UKAEA) addresses critical challenges in materials performance and structural integrity by integrating a variety of techniques, including machine learning models, for evaluating what’s known as the residual stress of materials. Residual stress is a measure of performance that’s locked into materials during manufacturing or operation. It can significantly affect the reliability and safety of fusion reactor components under extreme conditions.
Read More: AI is bad for the environment, and the problem is bigger than energy consumption
A key outcome of this collaboration is the development of a way of working that integrates data from experiments with a machine learning-powered predictive model to evaluate residual stress in fusion joints and components.
This framework has been validated through collaborations with leading institutions, including the National Physical Laboratory and UKAEA’s materials research facility. These advancements provide efficient and accurate assessments of materials performance and have redefined the evaluation of residual stress, unlocking new possibilities for assessing the structural integrity of components used in fusion devices.
This research directly supports the European Demonstration Power Plant (EU-DEMO) and the Spherical Tokamak for Energy Production (STEP) project, which aim to deliver a demonstration fusion power plant and prototype fusion power plant, respectively, to scale. Their success depends on ensuring the structural integrity of critical components under extreme conditions.
By using many AI-based approaches in a coordinated way, researchers can ensure that fusion systems are physically robust and economically viable, accelerating the path to commercialisation. AI can be used to develop simulations of fusion devices that integrate insights from plasma physics, materials science, engineering and other aspects of the process. By simulating fusion systems within these virtual environments, researchers can optimise reactor design and operational strategies.
- is an Associate Professor (Reader), University of Surrey
- This article first appeared in The Conversation
