An unsupervised machine-learning checkpoint-restart algorithm using Gaussian mixtures for particle-in-cell simulations

Chen, Guangye; Chacòn, Luis; Nguyen, Truong

doi:10.1016/j.jcp.2021.110185

Cited by 10 publications

(6 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…( 12). Method ii) can be generalized by transferring the anisotropic temperature generation information from f p to f a or using a set of Maxwellian basis functions 24 .…”

Section: Pullback Transformationmentioning

confidence: 99%

Electromagnetic total-f algorithm for gyrokinetic particle-in-cell simulations of boundary plasma in XGC

Hager¹,

Ku²,

Sharma³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…( 12). Method ii) can be generalized by transferring the anisotropic temperature generation information from f p to f a or using a set of Maxwellian basis functions 24 .…”

Section: Pullback Transformationmentioning

confidence: 99%

Electromagnetic total-f algorithm for gyrokinetic particle-in-cell simulations of boundary plasma in XGC

Hager¹,

Ku²,

Sharma³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…( 50) and velocity moment perturbation from Eq. ( 52) to a high degree of accuracy for the non-uniformly distributed macro-particle profile, we compute the individual particle weights (w p ) and velocities (v 0 p ) by inverting a mass-matrix [63]. These particles are initialized with a quasi-quiet start.…”

Section: Kink (M = 1) Mode In Helical Geometrymentioning

confidence: 99%

A conservative implicit-PIC scheme for the hybrid kinetic-ion fluid-electron plasma model on curvilinear meshes

Stanier,

Chacon

2021

Preprint

View full text Add to dashboard Cite

The hybrid kinetic-ion fluid-electron plasma model is widely used to study challenging multi-scale problems in space and laboratory plasma physics. Here, a novel conservative scheme for this model employing implicit particle-in-cell techniques [1] is extended to arbitrary coordinate systems via curvilinear maps from logical to physical space. The scheme features a fully non-linear electromagnetic fomulation with a multi-rate time advance -including sub-cycling and orbit-averaging for the kinetic ions. By careful choice of compatible particle-based kinetic-ion and mesh-based fluid-electron discretizations in curvilinear coordinates, as well as particle-mesh interpolations and implicit midpoint time advance, the scheme is proven to conserve total energy for arbitrary curvilinear meshes. In the electrostatic limit, the method is also proven to conserve total momentum for arbitrary curvilinear meshes. Although momentum is not conserved for arbitrary curvilinear meshes in the electromagnetic case, it is for an important subset of Cartesian tensor-packed meshes. The scheme and its novel conservation properties are demonstrated for several challenging numerical problems using different curvilinear meshes, including a merging flux-rope simulation for a space weather application, and a helical m = 1 mode simulation for magnetic fusion energy application.

show abstract

“…Innocenti et al (2013Innocenti et al ( , 2015b and Innocenti, Tenerani & Velli (2019b), or by coupling the kinetic and fluid descriptions, e.g. Daldorff et al (2014), Ashour-Abdalla et al (2015) and Lautenbach & Grauer (2018). Second, both spacecraft observations and numerical simulations produce an increasing amount of data.…”

Section: Introductionmentioning

confidence: 99%

“…Amaya et al (2020) and Roberts et al (2020), have been used for the classification/clustering of solar wind states. Unsupervised techniques, namely Gaussian mixture models, have recently proven quite effective in PIC simulations, either for the identification of regions of interest (Dupuis et al 2020), where the particle distribution functions deviate from Maxwellian, or to encode particle information for later resampling during simulation restarts (Chen, Chacón & Nguyen 2021). In this paper, we will use an unsupervised clustering technique based on self-organizing maps (SOMs, Kohonen 1982) to cluster simulated data points obtained from a PIC simulation.…”

Section: Introductionmentioning

confidence: 99%

Unsupervised classification of fully kinetic simulations of plasmoid instability using self-organizing maps (SOMs)

2023

View full text Add to dashboard Cite

The growing amount of data produced by simulations and observations of space physics processes encourages the use of methods rooted in machine learning for data analysis and physical discovery. We apply a clustering method based on self-organizing maps to fully kinetic simulations of plasmoid instability, with the aim of assessing their suitability as a reliable analysis tool for both simulated and observed data. We obtain clusters that map well, a posteriori, to our knowledge of the process; the clusters clearly identify the inflow region, the inner plasmoid region, the separatrices and regions associated with plasmoid merging. Self-organizing map-specific analysis tools, such as feature maps and the unified distance matrix, provide us with valuable insights into both the physics at work and specific spatial regions of interest. The method appears as a promising option for the analysis of data, both from simulations and from observations, and could also potentially be used to trigger the switch to different simulation models or resolution in coupled codes for space simulations.

show abstract

An unsupervised machine-learning checkpoint-restart algorithm using Gaussian mixtures for particle-in-cell simulations

Cited by 10 publications

References 54 publications

Electromagnetic total-f algorithm for gyrokinetic particle-in-cell simulations of boundary plasma in XGC

Electromagnetic total-f algorithm for gyrokinetic particle-in-cell simulations of boundary plasma in XGC

A conservative implicit-PIC scheme for the hybrid kinetic-ion fluid-electron plasma model on curvilinear meshes

Unsupervised classification of fully kinetic simulations of plasmoid instability using self-organizing maps (SOMs)

Contact Info

Product

Resources

About