Skip to content

GitLab

  • Projects
  • Groups
  • Snippets
  • Help
    • Loading...
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
    • Switch to GitLab Next
  • Sign in / Register
Q
QuaLiKiz
  • Project overview
    • Project overview
    • Details
    • Activity
    • Releases
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributors
    • Graph
    • Compare
    • Locked Files
  • Issues 28
    • Issues 28
    • List
    • Boards
    • Labels
    • Service Desk
    • Milestones
    • Iterations
  • Merge Requests 0
    • Merge Requests 0
  • Requirements
    • Requirements
    • List
  • CI / CD
    • CI / CD
    • Pipelines
    • Jobs
    • Schedules
    • Test Cases
  • Operations
    • Operations
    • Incidents
    • Environments
  • Packages & Registries
    • Packages & Registries
    • Container Registry
  • Analytics
    • Analytics
    • CI / CD
    • Code Review
    • Insights
    • Issue
    • Repository
    • Value Stream
  • Wiki
    • Wiki
  • Snippets
    • Snippets
  • Members
    • Members
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • QuaLiKiz-group
  • QuaLiKiz
  • Wiki
    • Qlknn
  • QLKNN overview

Last edited by Karel van de Plassche Feb 13, 2021
Page history

QLKNN overview

QuaLiKiz Neural Network regression

A neural network regression allows to emulate the results of first-principle-based turbulent transport predictions. Model output databases are compiled with inputs spanning a portion of the experimentally relevant parameter space. The neural network then learns the mapping from code inputs to outputs. The speed of the quasilinear models allows for the production of sufficiently large training sets for this application.

QuaLiKiz was used to create a large database of 3.10^8 flux calculations using 1.3 MCPUh on HPC resources (Edison@NERSC). Embedding known physical constraints in the training of the networks is essential for the surrogate model to perform well in transport predictions. As such, we show the importance of choosing the right cost function and more fundamentally, choosing which target variables the networks have to be trained on. Custom figures of merit and visualisation tools were developed to aid with neural network accuracy verification, such as the data slicer which you can use to visualize QuaLiKiz results over the 9D database. Ongoing work is underway for development of higher input dimension (13D) QuaLiKiz-neural-networks, where due to the curse of dimensionality, the relevant input subspace is constrained by an under-construction multi-machine profile database.

There are multiple flavours of QLKNN to try:

  • "baseline": general for large-aspect ratio tokamaks: QLKNN-hyper
  • "WIP advanced physics": general for large-aspect ratio tokamaks: QLKNN-HornNet
  • "WIP jet specific": better QuaLiKiz approximation for JET only: QLKNN-jetexp
  • "Deprecated no physics": Trained without QLKNN methodology, for reference only: QLKNN-fullflux

Journal publications

  • NEW Fast modeling of turbulent transport in fusion plasmas using neural networks: van de Plassche PoP 2020
  • Real-time-capable prediction of temperature and density profiles in a tokamak using RAPTOR and a first-principle-based transport model: Felici NF 2018
  • 5D proof of principle demonstration: Citrin NF letter 2015

Presentations

  • NEW Fast surrogate modelling of turbulent transport in fusion plasmas with physics-informed neural networks AAPPS invited talk pdf
  • Towards a high input dimension database for Neural Network regression based on JET measurements: Ho poster EPS2017
  • 9D database and 10D Neural Network regression including ExB shear in post-processing: van de Plassche poster TTF 2019

Master theses

  • Inclusion of Physics Constraints in Neural Network Surrogate Models for Fusion Simulation MSc P. Horn
  • Realtime capable turbulent transport modelling using neural network: van de Plassche master thesis
Clone repository
  • Background of QuaLiKiz
  • Developer best practices
  • Home
  • Input and output variables
  • Interfacing JETTO with standalone version of QuaLiKiz
  • Numerics of QuaLiKiz
  • Physics of QuaLiKiz
  • Published integrated simulations using QuaLiKiz
  • QLKNN
    • QLKNN hyper
      • QLKNN hyper overview
      • QLKNN hyper testing
      • QLKNN hyper usage
    • QLKNN overview
  • QuaLiKiz regression
  • Running with Pythontools in standalone
  • Scripting with Pythontools
View All Pages