Abstract Details
Abstracts
Author: Diego del-Castillo-Negrete
Requested Type: Poster
Submitted: 2026-03-20 22:41:15
Co-authors: M. Yang, G. Zhang, D. Zarzoso, B. Clavier
Contact Info:
University of Texas at Austin
2515 Speed
Austin, Texas 78712
US
Abstract Text:
We report recent progress on generative artificial intelligence (AI) methods to accelerate turbulence and kinetic computations in plasmas of interest to controlled nuclear fusion. For turbulence, we present the GAIT (Generative Artificial Intelligence Turbulence) framework based on the coupling of convolutional variational autoencoders (VAEs), that encode precomputed turbulence data into a reduced latent space, and recurrent neural networks and decoders that generate new turbulence states [1,2]. We also present results on PreVAE-Turb, a surrogate modeling framework that leverages pre-trained VAEs from the Stable Diffusion image generation model for efficient spatial compression of turbulence fields. The pre-trained VAE is fine-tuned on turbulence data and combined with convolutional long short-term memory networks to learn temporal dynamics in latent space, enabling autoregressive prediction of turbulent field evolution. We present application to Hasegawa-Mima fluid turbulence and gyrokinetic turbulence computed using the GENE code. For kinetic problems, we report progress on the use of AI methods to accelerate particle-based computations. The AI methods include Normalizing Flows [3] and Diffusion Models [4]. Convergence analysis, along with numerical test experiments are provided to demonstrate the effectiveness of the proposed methods. Applications include transport in 3D chaotic flows, and runaway electrons in magnetically confined fusion plasmas.
[1] B. Clavier, D. Zarzoso, D. del-Castillo-Negrete and E. Frenod, Phys. Rev. E Letters 111, L013202 (2025).
[2] B. Clavier, D. Zarzoso, D. del-Castillo-Negrete and E. Frenod, Physics of Plasmas 32 063905 (2025).
[3] M. Yang, P. Wang, D. del-Castillo-Negrete, Y. Cao and G. Zhang; SIAM journal of Scientific Computing, 46, (4) C508-C533 (2024).
[4] M. Yang, Y. Liu, D. del-Castillo-Negrete, Y. Cao and G. Zhang, Journal of Computational Physics, Vol 544, 114434 (2026).
Characterization: 4.0
Comments: