Unlock Fusion’s core secrets in Ai-Enhanced simulation

Creating and maintaining fusion reactions – essentially replicating star-like states on Earth is extremely difficult, and Dr. Nathan Howard, a leading research scientist at the MIT Plasma Science and Fusion Center (PSFC), has revealed that , I think that is the most important thing. The fascinating scientific challenges of our time.

“Both the overall promise of fusion as a science and clean energy source is really interesting. It motivated me to come to graduate school (MIT) and work at PSFC,” he says.

Howard is a member of the PSFC’s Magnetic Fusion Experiment Integrated Modeling (MFE-IM) group. MFE-IM Group leaders Pablo Rodriguez Fernandez, Howard and his team use simulation and machine learning to predict how plasmas will work on fusion devices. MFE-IM and Howard’s research aims to predict the performance of a particular technology or configuration before being manipulated in a real fusion environment, allowing for smarter design choices. To ensure their accuracy, these models are continuously validated using data from previous experiments, and the simulations are actually grounded.

Published in Nuclear Fusion, Howard is a paper entitled “Performance and Turbulence of Italian Combustion Plasmas with Nonlinear Gylocytetic Profile Prediction and Turbulence,” Howard is a swirl structure present in a plasma called turbulence. Explain how you used high-resolution simulations and see it. The world’s largest experimental fusion device currently under construction in southern France works as expected when switched on.

He also demonstrates how different operating setups produce roughly the same amount of energy output, but the low energy inputs can have a positive impact on the efficiency of typical fusion devices.

The biggest and best ever built

Forty years ago, the US and six other member countries came together to build Iter (Latin for “The Way”), a fusion device that runs a 500 megawatt fusion capacity and a plasma that can be generated 10 times. More energy than absorbed from external heating.

The plasma setup designed to achieve these goals – the most ambitious of fusion experiments – is called the ITER baseline scenario, and achieves this plasma as fusion science and plasma physics progresses. The methods are refined using increasingly powerful simulations such as modeling. The framework used by Howard.

In his work to validate the baseline scenario, Howard used CGYRO, a computer code developed by Howard’s collaborators at General Atomic. CGYRO applies complex plasma physics models to a defined set of fusion operating conditions. Although time-intensive, CGYRO generates highly detailed simulations of how plasmas work at different locations within the fusion device.

A comprehensive CGYRO simulation was performed via the Portals Framework, a collection of tools originally developed at MIT by Rodriguez Fernandez.

“The portal runs fidelity (CGYRO) and uses machine learning to build a quick model called “surrogate,” which can mimic the outcome of more complex executions, but much faster,” says Rodriguez. Fernandez explains. “Only high fidelity modeling tools like portals can glimpse the plasma core before they are formed. This prediction-first approach allows for more efficient plasma creation in devices such as Iter. .”

After the first pass, the surrogate’s accuracy is checked for high fidelity executions, and if the delegate does not produce results along the CGYRO results, the portal will be run again to make the CGYRO results better I improved the surrogate to imitate it.

“The good thing is that once you have built a well-trained (surrogate) model, you can use it to predict different conditions.

Once they are fully trained, surrogates are used to explore how different combinations of inputs affect the predicted performance of ITER and how it achieved the baseline scenario. It was done. In particular, running the surrogate took only a few minutes and was able to provide a boost in conjunction with CGYRO, producing detailed results more quickly.

“I just stopped by to see what my condition is.”

Howard’s work with CGYRO, Portal and Surrogates examined specific combinations of operating conditions that were predicted to achieve the baseline scenario. These conditions include the magnetic field used, the methods used to control the shape of the plasma, external heating applied, and many other variables. Using 14 iterations of CGYRO, Howard was able to confirm that the current baseline scenario configuration can achieve 10 times the output of the input to the plasma.

Regarding the results, Howard said, “The modelling we’ve carried out is probably the highest fidelity possible at this point, and almost certainly the highest fidelity that has been published.”

14 iterations of CGYRO included confirmation of plasma performance and a running portal to build a surrogate model of input parameters, so we connected the surrogate to CGYRO to operate more efficiently . The additional iteration of CGYRO simply explored an alternative scenario where ITER predicted that it could generate roughly the same amount of energy at about half the input power.

The surrogate enhancement CGYRO model revealed that the temperature of the plasma core, and therefore the fusion reaction, is not undue affected by the lesser effects of the power input. The lower the power input, the more efficient operation is. Howard’s results are a reminder that there may be other ways to improve Iter’s performance. They have not been discovered yet.

“The fact that the results of this modeling can be used to influence the planning of experiments like Iter is exciting. For many years, this has been the goal of our research, and now we have actually made it. I’ve said I’ll do it – it’s an amazing arc and it’s really fulfilling.”

This story has been republished courtesy of MIT News (web.mit.edu/newsoffice/), a popular site that covers news about MIT research, innovation and education.