S. Madireddy, C. Akçay, S. E. Kruger, T. Bechtel Amara, X. Sun, J. McClenaghan, J. Koo, A. Samaddar, Y. Liu, P. Balaprakash, L. L. Lao
{"title":"EFIT-Prime:用于 DIII-D 平衡重建的概率和物理约束降阶神经网络模型","authors":"S. Madireddy, C. Akçay, S. E. Kruger, T. Bechtel Amara, X. Sun, J. McClenaghan, J. Koo, A. Samaddar, Y. Liu, P. Balaprakash, L. L. Lao","doi":"10.1063/5.0213609","DOIUrl":null,"url":null,"abstract":"We introduce EFIT-Prime, a novel machine learning surrogate model for EFIT (Equilibrium FIT) that integrates probabilistic and physics-informed methodologies to overcome typical limitations associated with deterministic and ad hoc neural network architectures. EFIT-Prime utilizes a neural architecture search-based deep ensemble for robust uncertainty quantification, providing scalable and efficient neural architectures that comprehensively quantify both data and model uncertainties. Physically informed by the Grad–Shafranov equation, EFIT-Prime applies a constraint on the current density Jtor and a smoothness constraint on the first derivative of the poloidal flux, ensuring physically plausible solutions. Furthermore, the spatial location of the diagnostics is explicitly incorporated in the inputs to account for their spatial correlation. Extensive evaluations demonstrate EFIT-Prime's accuracy and robustness across diverse scenarios, most notably showing good generalization on negative-triangularity discharges that were excluded from training. Timing studies indicate an ensemble inference time of 15 ms for predicting a new equilibrium, offering the possibility of plasma control in real-time, if the model is optimized for speed.","PeriodicalId":20175,"journal":{"name":"Physics of Plasmas","volume":null,"pages":null},"PeriodicalIF":2.0000,"publicationDate":"2024-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"EFIT-Prime: Probabilistic and physics-constrained reduced-order neural network model for equilibrium reconstruction in DIII-D\",\"authors\":\"S. Madireddy, C. Akçay, S. E. Kruger, T. Bechtel Amara, X. Sun, J. McClenaghan, J. Koo, A. Samaddar, Y. Liu, P. Balaprakash, L. L. Lao\",\"doi\":\"10.1063/5.0213609\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"We introduce EFIT-Prime, a novel machine learning surrogate model for EFIT (Equilibrium FIT) that integrates probabilistic and physics-informed methodologies to overcome typical limitations associated with deterministic and ad hoc neural network architectures. EFIT-Prime utilizes a neural architecture search-based deep ensemble for robust uncertainty quantification, providing scalable and efficient neural architectures that comprehensively quantify both data and model uncertainties. Physically informed by the Grad–Shafranov equation, EFIT-Prime applies a constraint on the current density Jtor and a smoothness constraint on the first derivative of the poloidal flux, ensuring physically plausible solutions. Furthermore, the spatial location of the diagnostics is explicitly incorporated in the inputs to account for their spatial correlation. Extensive evaluations demonstrate EFIT-Prime's accuracy and robustness across diverse scenarios, most notably showing good generalization on negative-triangularity discharges that were excluded from training. Timing studies indicate an ensemble inference time of 15 ms for predicting a new equilibrium, offering the possibility of plasma control in real-time, if the model is optimized for speed.\",\"PeriodicalId\":20175,\"journal\":{\"name\":\"Physics of Plasmas\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":2.0000,\"publicationDate\":\"2024-09-06\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Physics of Plasmas\",\"FirstCategoryId\":\"101\",\"ListUrlMain\":\"https://doi.org/10.1063/5.0213609\",\"RegionNum\":3,\"RegionCategory\":\"物理与天体物理\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"PHYSICS, FLUIDS & PLASMAS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Physics of Plasmas","FirstCategoryId":"101","ListUrlMain":"https://doi.org/10.1063/5.0213609","RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"PHYSICS, FLUIDS & PLASMAS","Score":null,"Total":0}
EFIT-Prime: Probabilistic and physics-constrained reduced-order neural network model for equilibrium reconstruction in DIII-D
We introduce EFIT-Prime, a novel machine learning surrogate model for EFIT (Equilibrium FIT) that integrates probabilistic and physics-informed methodologies to overcome typical limitations associated with deterministic and ad hoc neural network architectures. EFIT-Prime utilizes a neural architecture search-based deep ensemble for robust uncertainty quantification, providing scalable and efficient neural architectures that comprehensively quantify both data and model uncertainties. Physically informed by the Grad–Shafranov equation, EFIT-Prime applies a constraint on the current density Jtor and a smoothness constraint on the first derivative of the poloidal flux, ensuring physically plausible solutions. Furthermore, the spatial location of the diagnostics is explicitly incorporated in the inputs to account for their spatial correlation. Extensive evaluations demonstrate EFIT-Prime's accuracy and robustness across diverse scenarios, most notably showing good generalization on negative-triangularity discharges that were excluded from training. Timing studies indicate an ensemble inference time of 15 ms for predicting a new equilibrium, offering the possibility of plasma control in real-time, if the model is optimized for speed.
期刊介绍:
Physics of Plasmas (PoP), published by AIP Publishing in cooperation with the APS Division of Plasma Physics, is committed to the publication of original research in all areas of experimental and theoretical plasma physics. PoP publishes comprehensive and in-depth review manuscripts covering important areas of study and Special Topics highlighting new and cutting-edge developments in plasma physics. Every year a special issue publishes the invited and review papers from the most recent meeting of the APS Division of Plasma Physics. PoP covers a broad range of important research in this dynamic field, including:
-Basic plasma phenomena, waves, instabilities
-Nonlinear phenomena, turbulence, transport
-Magnetically confined plasmas, heating, confinement
-Inertially confined plasmas, high-energy density plasma science, warm dense matter
-Ionospheric, solar-system, and astrophysical plasmas
-Lasers, particle beams, accelerators, radiation generation
-Radiation emission, absorption, and transport
-Low-temperature plasmas, plasma applications, plasma sources, sheaths
-Dusty plasmas