Search

A research team at Princeton University has succeeded in significantly improving the quality of disruption predictions for a plasma in a tokamak fusion device through deep learning.

The team led by plasma physicist William Tang had access to the database of the Joint European Torus (JET) tokamak to demonstrate the potential of deep learning and neural networks in the analysis of large amounts of plasma disruption data. They fed more than a half-petabyte of information into their fusion recurrent neural nets (FRNN) deep learning software.

Through GPU-accelerated computing—i.e., the use of a graphic processing unit together with a CPU—disruption prediction improved dramatically in accuracy and speed. The Princeton team also demonstrated the scalability to over thousands of even more advanced GPUs.

According to Tang, this development holds a huge potential for accelerating scientific discovery through deep learning for fusion research.

Read the full article here.

The Massachusetts Institute of Technology (MIT) has just released plans for a number of initiatives that will broaden its engagement in fusion research.

MIT has been a long-standing player in fusion research, receiving support primarily from the US Department of Energy that has included funding in three major experiments at MIT's Plasma Fusion and Science Center (PFSC). The last of these—the Alcator C-Mod tokamak—ended operations in 2016 after 23 years and approximately 33,000 plasma shots.

Now, in a renewed approach, researchers at MIT will be working with a newly formed company, Commonwealth Fusion Systems, to combine the experience of a well-established lab with the more nimble and longer-term financing perspectives from the private sector.

Commonwealth Fusion Systems is an independent, for-profit company created by former MIT staff and students for the accelerated commercialization of fusion energy.

Joint teams will be working to develop a new class of high temperature superconducting magnets, followed by the conception and construction of a compact 100 MW fusion experiment called SPARC, and finally a larger demonstration power plant. 

Commonwealth Fusion Systems has obtained $50 million in funding from the Italian energy company Eni, which is also a founding member of the MIT Energy Initiative (MITEI). Eni has also announced funding to the order of $2 million per year for PSFC's newly created Laboratory for Innovation in Fusion Technologies, which will focus on cutting-edge solutions for some of the challenges on the road to a steady-state fusion reactor.

Read more about the new initiatives on the MIT website.