Machine learning applications to computational plasma physics and reduced-order plasma modeling: a perspective
Farbod Faraji, Maryam Reza
Abstract
Abstract Machine learning (ML) offers a diverse array of tools and architectures for transforming data from simulations and experiments into explainable science, thereby augmenting domain knowledge. ML-enhanced numerical modeling has the potential to revolutionize scientific computing for complex engineering systems, enabling detailed analyses of technologies’ operation and facilitating automation in optimization and control. While ML applications have grown significantly in various scientific fields—particularly in fluid mechanics, where ML has demonstrated immense potential in computational fluid dynamics (CFD)—numerical plasma physics research has yet to see similar advancements. However, the strong parallels between fluid mechanics and plasma physics provide an opportunity to transfer ML advances in fluid flow modeling to computational plasma physics. This Perspective outlines a roadmap for such progress. It begins by introducing fundamental aspects of ML, including the main categories of algorithms and the types of problems they can address. For each problem type, we highlight specific examples of ML applications in CFD, reviewing several notable efforts. We also review recent ML applications in plasma physics and discuss promising future directions and pathways for development in plasma modeling across different problem types. Finally, we elaborate on key challenges and requirements that must be addressed to unlock ML’s full potential in computational plasma physics, including the development of cost-effective, high-fidelity simulation tools for extensive data generation.