Princeton-led team develops AI for fusion plasma monitoring
A new AI software tool for monitoring and controlling the plasma inside nuclear fuel systems has been developed by an international collaboration of scientists from Princeton University, Princeton Plasma Physics Laboratory (PPPL), Chung-Ang University, Columbia University, and Seoul National University. The software, which the researchers call Diag2Diag, is described in the paper, “Multimodal super-resolution: discovering hidden physics and its application to fusion plasmas,” published in Nature Communications.
Diagnostics: The name Diag2Diag refers to the diagnostic techniques for analyzing plasma, according to the researchers. Sensors in a fusion system take measurements of plasma density, temperature, and other characteristics every fraction of a second. Thompson scattering is an example of a diagnostic technique that is used in tokamaks. However, such diagnostic measurements may not happen often enough to detect sudden changes in the rapidly evolving plasma that could make the plasma unstable and cause power production to be unreliable. However, with the AI developed by the Princeton-led team, the “missing” data can be filled in.
Jalalvand
Synthetic data: Princeton research scholar Azarakhsh Jalalvand, the lead author of the paper, explains, “We have found a way to take the data from a bunch of sensors in a system and generate a synthetic version of the data for a different kind of sensor in that system.” The synthetic data is more detailed than data from actual sensors, yet it aligns well with “real world” data, thereby increasing the robustness of plasma monitoring and control, according to the researchers. This diagnostic technique could also reduce the complexity and cost of fusion systems. Jalalvand said that the Diag2Diag AI enhances the detail of data and can recover data from failing or degraded sensors, ensuring reliability of the system.
Giving diagnostics a boost: Getting as much data as possible about the ever-changing plasma conditions is important to keep future fusion devices functioning as a reliable source of electricity. As noted by Jalalvand, “Fusion devices today are all experimental laboratory machines, so if something happens to a sensor, the worst thing that can happen is that we lose time before we can restart the experiment. But if we are thinking about fusion as a source of energy, it needs to work 24/7, without interruption.”
The Diag2Diag AI is a way “of giving your diagnostics a boost without spending hardware money,” said Egemen Kolemen, the principal investigator of the research and a professor who is jointly appointed at PPPL and Princeton University.
The pedestal: An important but difficult part of fusion plasma to monitor is the edge of the plasma, known as the pedestal, where powerful bursts of energy called edge-localized modes pose risks to the fusion reactor’s inner walls. Careful monitoring of the pedestal with the assistance of the Diag2Diag AI tool should allow scientists to enhance plasma performance by getting the most energy out of the fusion reaction in the most efficient way, according to the researchers.
Other applications: The Diag2Diag developers said that they are pursuing plans to expand the scope of their AI software. They note that Diag2Diag could be applied to other areas of fusion diagnostics besides plasma monitoring, and it is also broadly applicable to other fields in which diagnostic data are missing or limited, such as in spacecraft and robotic surgery.
