1 of 20

Integrated modeling �for equilibrium, scenarios, �and disruption processes in tokamaks �with DINA and NSFsim

Eduard Khairutdinov & Next Step Fusion team

October 2025

nextfusion.org

17th Integrated Modelling Expert Group Meeting (IMEG)

2 of 20

2

INTRO

NSF Simulator (NSFsim) is based on over 30 years of experience in tokamak experimentation and control

NSFsim is an advanced Grad-Shafranov 2D solver with a 1D kinetic component. It is based on the renowned DINA simulation approach and a modular architecture that has been extensively tested with numerous tokamaks

nextfusion.org

3 of 20

3

nextfusion.org

Direct calculation

Scenario calculation

No-plasma calculation

Inverse solver for ML datasets

Disruption (MD/VDE)

Equilibrium reconstruction

Integrated modelling

Online simulation platform

API interface for online calculations

ML-based controllers

NSFsim is IMAS-compatible for data storage and coupled simulations with external codes

SIMULATION SCOPE

4 of 20

4

nextfusion.org

Currents in poloidal field coils

Plasma parameters

Set currents, voltages, or both currents and voltages for different active coils as the input

Choose kinetic mode for calculation:

  • Temperatures and density fixed as profiles
  • Density fixed as profile
  • Temperatures fixed as profiles
  • Density fixed as average value

DIRECT CALCULATION

5 of 20

5

nextfusion.org

Plasma parameters evolution

ST VNS machine example

Include ion/electron heating with current drive generation or just pure ohmic heating

Choose kinetic mode for calculation:

  • Prescribed evolution for Ne, Te, Ti
  • Prescribed evolution for Ne, calculate Te, Ti

Scenario for coil’s currents

SCENARIO CALCULATION

6 of 20

6

nextfusion.org

Currents in poloidal field coils

Passive currents,

EMD signals

Validation of the 2D electromagnetic model used for other calculations

Double-check for list of ”approved” EMD signals for feedback controllers

NO-PLASMA CALCULATION

7 of 20

7

nextfusion.org

Plasma parameters

Parameters for direct calculation initialization

Calculate set of different

”start points” for ML learning

INPUTS:

  • Plasma shape
  • Plasma current
  • Coil currents constraints
  • Temperatures and density as profiles
  • Psi on axis

INVERSE SOLVER FOR ML DATASETS

8 of 20

8

nextfusion.org

Two stages:

  • Thermal quench
  • Redistribution of the current profile after mixing

Features:

  • Characteristic Ip spike
  • Calculate Halo area with poloidal/toroidal currents
  • RE current
  • Eddy currents in passive structures
  • Forces distribution in passive structures

Artificial triggering or self-consistent evolution of disruption conditions

DISRUPTION (MD/VDE)

9 of 20

9

nextfusion.org

Reconstruction based on:

  • Plasma current
  • EMD signals with ‘valid’ mask
  • Coils current

Two types of reconstruction:

  • Fixed filaments
  • Floating filaments

EQUILIBRIUM RECONSTRUCTION

10 of 20

10

INTEGRATED MODELLING

nextfusion.org

Ongoing projects on integrating codes around NSFsim for self-consistent

plasma simulations for design and control (a.k.a. flight simulator)

TRAVIS - ECRH and ECCD ray-tracing code | ASCOT5 - NBI and fast-particles

TGLF - turbulent transport | MISHKA - NN surrogate model for pedestal region

11 of 20

11

INTEGRATED SIMULATION CASE: ECCD WITH NSFSIM

nextfusion.org

TRAVIS is a multi-beam and multi-pass ray-tracing code for electron cyclotron resonance heating (ECRH). �Courtesy of IPP Max Planck.��Key features of TRAVIS:

  • TRAVIS works with an arbitrary 3D magnetic equilibrium being applicable for both stellarators and tokamaks.
  • The equations for ray tracing are taken in the weakly relativistic approach with thermal effects taken into account.
  • Absorption, current drive and emissivity are calculated in the fully relativistic approach.
  • For the calculation of ECCD, an adjoint technique with parallel momentum conservation is applied.

[N.B. Marushchenko, Y. Turkin, H. Maassberg, Computer Physics Communications 185-1, 2014]

�We use a combination of NSFsim and TRAVIS as a simulation core within our ML plasma control pipeline.

Coupling with other codes (ASCOT5, MISHKA for NBI, TGLF for turbulent transport) is being implemented.

12 of 20

12

nextfusion.org

Fusion Twin Platform https://fusiontwin.io is a cloud-based platform for running highly customizable tokamak simulations, uploading and visualizing fusion data, collaborating and sharing with others, and more. To run realistic simulations, the Platform utilizes digital replicas of tokamaks, including DIII-D, ISTTOK, SMART, and NSF NTT.

ONLINE SIMULATION PLATFORM

Platform API facilitates software-in-the-loop testing of tokamak plasma controllers. It integrates with NSFsim via MATLAB/Simulink, Python, C, and other languages, exposing all simulator inputs and outputs. The API is freely available for academic use.

https://github.com/Next-Step-Fusion/platform-api-examples

13 of 20

13

ONLINE SIMULATION PLATFORM

nextfusion.org

14 of 20

14

nextfusion.org

ONLINE SIMULATION PLATFORM

15 of 20

15

nextfusion.org

ONLINE SIMULATION PLATFORM

16 of 20

16

nextfusion.org

PUBLIC API WITH MATLAB/SIMULINK

17 of 20

17

ML-BASED CONTROLLERS

nextfusion.org

DIII-D

Dynamic control

of arbitrary plasma

shape in DIII-D tokamak.

Accuracy ~ 1 - 3 cm

Variable plasma shape

18 of 20

18

PLANS

nextfusion.org

ABOUT UPDATES, NEWS OR SMTH

19 of 20

19

SUMMARY

nextfusion.org

We welcome collaborations, joint benchmarking, and feedback from the community

And add contacts to last slide (as company and as person)

20 of 20

20

nextfusion.org

THANK YOU FOR YOUR ATTENTION