Machine learning (ML) is rapidly growing in popularity as new tools for applying these techniques to plasma control are developed. You’ve probably seen numerous media mentions of those projects.
ML is also incredibly helpful in data analysis and modeling. Just as our doctors might use ML to read our x-rays, today’s fusion researchers can use these tools to diagnose the presence of plasma instabilities. The DIII-D research highlight from last week shows the work of Alan Kaptanoglu and colleagues, who trained neural network-based models to identify Alfven eigenmodes.
For those of us who are fans of electromagnetic waves, the tokamak is a wonderland in its production of a huge zoo of Alfvenic modes. Identifying those modes is a time-consuming, and often manual, process. As noted in the publication, these types of models can be evolved into control applications, but they also help to improve our physics understanding of the waves.
Alan A. Kaptanoglu et al 2022 Nucl. Fusion 62 106014, https://iopscience.iop.org/article/10.1088/1741-4326/ac8a03