28.07.2013 Views

On the OMFIT modeling framework and the development of steady ...

On the OMFIT modeling framework and the development of steady ...

On the OMFIT modeling framework and the development of steady ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>On</strong> <strong>the</strong> <strong>OMFIT</strong> <strong>modeling</strong> <strong>framework</strong> <strong>and</strong> <strong>the</strong><br />

<strong>development</strong> <strong>of</strong> <strong>steady</strong> state scenarios for a<br />

Fusion Nuclear Science Facility<br />

November 14 - 16, 2012 University <strong>of</strong> Kyoto, Kyoto, Japan<br />

Orso Meneghini<br />

S. Smith ⋆ , H. St John ⋆ , A. Gar<strong>of</strong>alo ⋆ , L. Lao ⋆ , V. Chan ⋆<br />

Oak Ridge Associated Universities<br />

⋆ General Atomics<br />

November 16 th 2012


Integrated <strong>modeling</strong> in magnetic fusion<br />

Integrated <strong>modeling</strong>: self-consistent interaction <strong>of</strong> specialized codes<br />

to capture complex interplay <strong>of</strong> different physical processes<br />

Manual integration <strong>of</strong> different codes is a complicated endeavor<br />

Many integrated <strong>modeling</strong> s<strong>of</strong>twares have been developed<br />

(SWIM, ONETWO, CRONOS, TRANSP, CORSICA, ITM-IT, TASK, ...)<br />

Two main categories:<br />

1 Transport solver:<br />

• External components to calculate equilibrium, sources <strong>and</strong> fluxes<br />

• Tight coupling: Data exchange at memory level<br />

2 Workflow-manager:<br />

• Data “flows” through <strong>the</strong> components<br />

• Loose coupling: Data exchange through files


Conventional IM <strong>framework</strong>s define data structures, codes<br />

integration <strong>and</strong> usage for <strong>the</strong> users<br />

TOP-DOWN approach<br />

1 Overall vision guides creation <strong>of</strong> <strong>framework</strong> <strong>and</strong> its components<br />

e.g.:<br />

• Simulating plasma discharge start to finish<br />

• Couple edge <strong>and</strong> core transport<br />

• Improve physics model by high performance computing<br />

2 St<strong>and</strong>ard data structures to allow data exchange<br />

• Statefile(s), CPO, PBSD, ITER?, ...<br />

St<strong>and</strong>ard data structures are in principle an effective idea, however<br />

<strong>the</strong>y comes with additional complexity <strong>and</strong> cost:<br />

• Many subtleties need to be figured out<br />

• Agreement with all parties<br />

• Hard to convince developers to alter <strong>the</strong>ir codes<br />

There is a disconnect between this TOP-DOWN approach <strong>and</strong><br />

every-day specific needs <strong>of</strong> many <strong>the</strong>orists <strong>and</strong> experimentalist


<strong>On</strong>e Modeling Framework for Integrated Tasks (<strong>OMFIT</strong>)<br />

is a workflow-manager developed to bridge this gap<br />

Change <strong>of</strong> paradigm: BOTTOM-UP approach<br />

1 Framework provides <strong>the</strong> tools while interaction is defined by users<br />

• Allow users to define <strong>the</strong> codes coupling<br />

• Components grow in an organic fashion, based on <strong>the</strong> specific<br />

problem <strong>of</strong> its users<br />

Same difference <strong>of</strong> enciclopedia Vs. wikipedia<br />

<strong>OMFIT</strong> is a “wik-integrated-<strong>modeling</strong>” <strong>framework</strong><br />

2 Allow integration without need <strong>of</strong> st<strong>and</strong>ard data structures


<strong>On</strong>e Modeling Framework for Integrated Tasks (<strong>OMFIT</strong>)<br />

is a workflow-manager developed to bridge this gap<br />

Change <strong>of</strong> paradigm: BOTTOM-UP approach<br />

1 Framework provides <strong>the</strong> tools while interaction is defined by users<br />

• Allow users to define <strong>the</strong> codes coupling<br />

• Components grow in an organic fashion, based on <strong>the</strong> specific<br />

problem <strong>of</strong> its users<br />

Same difference <strong>of</strong> enciclopedia Vs. wikipedia<br />

<strong>OMFIT</strong> is a “wik-integrated-<strong>modeling</strong>” <strong>framework</strong><br />

2 Allow integration without need <strong>of</strong> st<strong>and</strong>ard data structures<br />

GREAT ATTEMPT<br />

INTEGRATED MODELING<br />

FESTIVAL 2012<br />

“IT SIMPLY FILLS YOU WITH JOY”<br />

“AN ABSOLUTE MUST SEE”<br />

“IT WILL HAVE YOU HOULING !”


<strong>OMFIT</strong> philosophy <strong>and</strong> design<br />

• Recognize <strong>and</strong> encourage reuse <strong>of</strong> existing work<br />

• Allow inter-operativity <strong>of</strong> many data formats<br />

• Easily integrate existing scripts/widgets/s<strong>of</strong>twares/...<br />

• Ease <strong>the</strong> way <strong>of</strong> working...<br />

• Interactive graphical environment<br />

• High level APIs for remote& parallel execution, GUIs, visualization<br />

• GUIs hide complexity for streamline analysis & inexperienced users<br />

• ...without limiting possibilities<br />

• All data/parameters are always accessible<br />

• Modular structure with arbitrary organization<br />

• Scripting (Python)<br />

• Facilitate <strong>modeling</strong>-experiment interaction<br />

• Integration with MDS+, PTDATA, SQL databases, etc...<br />

• Create a cooperative environment<br />

• Sharing <strong>of</strong> physics modules among users<br />

• Open-source, based on open-source s<strong>of</strong>tware<br />

• It’s all about being generic, generic <strong>and</strong> generic


1 The <strong>OMFIT</strong> <strong>framework</strong><br />

The <strong>OMFIT</strong> tree structure<br />

<strong>OMFIT</strong> interactive graphical environment<br />

Additional features: user GUIs, high level APIs, MDS+ integration,<br />

visualization<br />

2 Achievements <strong>and</strong> physics results<br />

Control-room analysis <strong>of</strong> gyro-kinetic stability<br />

MHD stability survey<br />

Magnetic flutter model for RMP-induced transport<br />

NTV model for NRMF-induced rotation<br />

Scenario <strong>development</strong> for DIII-D AT discharges<br />

3 Development <strong>of</strong> <strong>steady</strong>-state scenarios for FDF<br />

Improved LHCD system <strong>of</strong> FDF<br />

Integrated <strong>modeling</strong> in <strong>OMFIT</strong> with Eφ = 0 <strong>and</strong> β constraints<br />

Preliminary results using GCNMP/GLF23 transport model<br />

4 Conclusions <strong>and</strong> future <strong>development</strong>s


Main idea: to treat files, scripts, experiment data, texts,<br />

plots, executables, ... as a uniform collection <strong>of</strong> objects<br />

• Centralize data from different sources, independent <strong>of</strong> origin<br />

• Store everything deemed relevant, with no a-priori decision <strong>of</strong><br />

which variables are stored <strong>and</strong> how<br />

• Support <strong>of</strong> few data formats allows interaction with many codes<br />

(Does not preclude use <strong>of</strong> statefiles, etc...)<br />

MDS+<br />

NetCDF<br />

Python<br />

FORTRAN namelist<br />

<strong>OMFIT</strong><br />

Non st<strong>and</strong>ard<br />

IDLsave<br />

MATLAB<br />

Unied<br />

data access<br />

Unied<br />

data structure<br />

Unied intertask<br />

communication<br />

Unied<br />

GUI, API <strong>and</strong><br />

visualization


Data is organized in a a single, self-descriptive, hierarchical<br />

structure structure (<strong>OMFIT</strong> tree), which provides unified data<br />

access <strong>and</strong> sets objects in a context<br />

ONETWO<br />

(modules)<br />

Gfile<br />

(EFIT file)<br />

…<br />

FILES<br />

(directory)<br />

CASE (str)<br />

<strong>OMFIT</strong><br />

RMAXIS (float)<br />

NW (int)<br />

PSIRZ (array)<br />

Kfile<br />

(namelist)<br />

EFIT<br />

(module)<br />

SCRIPTS<br />

(directory)<br />

runEFIT<br />

(python)<br />

GUIS<br />

(directory)<br />

PLOTS<br />

(directory)<br />

SETTINGS<br />

(namelist)<br />

• Objects are fur<strong>the</strong>r organized into modules: collections<br />

<strong>of</strong> objects used to solve elementary physics problem<br />

• Data containers (e.g. files) are interpreted <strong>and</strong> <strong>the</strong>ir<br />

data populates <strong>the</strong> tree<br />

• Objects can be directly accessed/manipulated in a<br />

unified way, independently <strong>of</strong> <strong>the</strong>ir type or origin<br />

e.g. <strong>OMFIT</strong>[EFIT’][FILES’][Gfile’][PSIRZ]


Unified data structure inherently defines a memory space<br />

where inter-tasks communication can dynamically occur<br />

ONETWO<br />

(modules)<br />

Gfile<br />

(EFIT file)<br />

…<br />

FILES<br />

(directory)<br />

CASE (str)<br />

<strong>OMFIT</strong><br />

RMAXIS (float)<br />

NW (int)<br />

PSIRZ (array)<br />

Kfile<br />

(namelist)<br />

EFIT<br />

(module)<br />

SCRIPTS<br />

(directory)<br />

runEFIT<br />

(python)<br />

myExample<br />

(python)<br />

GUIS<br />

(directory)<br />

PLOTS<br />

(directory)<br />

SETTINGS<br />

(namelist)<br />

<strong>OMFIT</strong>[’ONETWO']['SCRIPTS'][’ONETWOexec'].run()<br />

for scale in [0.9,1.0,1.1]:<br />

<strong>OMFIT</strong>[‘EFIT’]['FILES'][‘Kfile']['IN1']['PRESSR']=\<br />

<strong>OMFIT</strong>[’ONETWO’]['FILES']['statefile']['press’]*scale<br />

<strong>OMFIT</strong>['EFIT'][’SCRIPTS']['EFITexec'].run()<br />

print(<strong>OMFIT</strong>[‘EFIT’]['FILES'][‘Gfile’][‘RMAXIS’])


<strong>OMFIT</strong> data structure allows unified top-level GUI to manage<br />

tree objects, run simulations <strong>and</strong> analyze data interactively<br />

Tree browser<br />

Objects<br />

description<br />

Search<br />

Status bar<br />

Console<br />

Comm<strong>and</strong> box


High level Python APIs allow users to:<br />

Execute tasks remotely <strong>and</strong> in parallel<br />

• Remote code execution <strong>and</strong> file upload / download via SSH<br />

• Parallel execution <strong>of</strong> <strong>the</strong> same task with different input<br />

parameters, on multiple remote machines<br />

• Real-time monitoring <strong>of</strong> local / remote <strong>and</strong> serial / parallel tasks<br />

• Run codes where <strong>the</strong>y already work!<br />

→ No need to compile all codes on one system<br />

Host<br />

Server on<br />

same network<br />

ssh<br />

Firewall or<br />

login node<br />

Internet<br />

Servers<br />

directly<br />

reachable<br />

Servers<br />

behind<br />

firewall<br />

Monitor progress <strong>of</strong> parallel execution<br />

Running process Completed process


High level Python APIs allow users to:<br />

Quickly create Graphical Users Interfaces (GUIs)<br />

User GUIs speed-up routine analysis<br />

<strong>and</strong> hide any <strong>of</strong> <strong>the</strong> underlying<br />

complexity to inexperienced users<br />

• GUIs are python scripts <strong>and</strong> can<br />

be created/ modified by users<br />

<strong>the</strong>mselves<br />

• Easy! For each GUI entry need to<br />

specify:<br />

• A label<br />

• Associated <strong>OMFIT</strong> tree variable<br />

• GUIs can be nested to create<br />

arbitrarily large GUIs, while<br />

ensuring consistency<br />

ONETWO<br />

KineticEFIT<br />

EFIT<br />

PROFILES


Visualization <strong>of</strong> experimental <strong>and</strong> <strong>modeling</strong> data allows<br />

quick inspection <strong>and</strong> analysis<br />

Kinetic EFIT iterations<br />

M3D-C1 simulation <strong>of</strong><br />

RMP pressure perturbation<br />

• 1D/2D arrays can be<br />

quickly inspected with <strong>the</strong><br />

push <strong>of</strong> a button<br />

• Plots are objects which<br />

can be stored in <strong>the</strong><br />

<strong>OMFIT</strong> tree<br />

• Plots are interactive <strong>and</strong><br />

can be customized (à la<br />

MATLAB, though still with<br />

some limitations)<br />

• Over-plot results from<br />

different analyses / codes<br />

/ iterations / ...<br />

• More exotic plots can be<br />

scripted in Python<br />

• Plots scripts can be<br />

assigned to specific<br />

objects


MDS+ tree, PTDATA signals, SQL tables are also integrated<br />

into <strong>the</strong> <strong>OMFIT</strong> tree<br />

• Experimental data can be directly accessed <strong>and</strong> visualized<br />

• e.g. Create input for codes or compare models with experiments<br />

MDS+ traverser


Python scripting empowers users with limitless capabilities<br />

<strong>and</strong> available libraires simplify users job<br />

Example: Easy to setup non-linear parameter optimization based on<br />

cost function approach:<br />

• Change SCALEPP EXT variable in EFIT to obtain βp = 0.4<br />

def cost_f(x):<br />

<strong>OMFIT</strong>['EFIT']['FILES']['rfile']['PROFILE_EXT’]['SCALEPP_EXT']=x[0]<br />

<strong>OMFIT</strong>['EFIT']['SCRIPTS']['EFITexec'].run()<br />

cost=(<strong>OMFIT</strong>['EFIT']['FILES']['aEQDSK']['betap']/0.4 - 1)**2<br />

return cost<br />

x=[1.0]<br />

res = scipy.optimize.fmin(cost_f, x)<br />

Cost function Vs SCALEPP_EXT per iteration<br />

2<br />

4<br />

6<br />

5<br />

7<br />

8<br />

3<br />

1


Growing list <strong>of</strong> available modules<br />

Equilibrium<br />

• EFIT<br />

• KineticEFIT<br />

Transport<br />

• ONETWO<br />

• GCNMP<br />

• FASTRAN<br />

Pr<strong>of</strong>iles<br />

• GApr<strong>of</strong>iles<br />

• ZIPFIT<br />

Gyro-kinetic<br />

• TGLF<br />

• GKS<br />

<strong>OMFIT</strong><br />

MHD stability<br />

• PEST3<br />

• GATO<br />

O<strong>the</strong>rs<br />

• GENRAY<br />

• M3DC1<br />

• NTV<br />

• Magnetic<br />

flutter<br />

• Integrating codes into<br />

<strong>OMFIT</strong> automatically<br />

endows <strong>the</strong>m with new<br />

capabilities<br />

• Easy to support new codes,<br />

especially if <strong>the</strong>y use<br />

st<strong>and</strong>ard file formats like<br />

namelists or NetCDF<br />

• Need <strong>the</strong> know how <strong>of</strong> an<br />

expert to setup a modules.<br />

Users exploits this know how,<br />

especially with available<br />

users GUIs.<br />

• Modules are shared among<br />

users (now via GitHub),<br />

tools to compare/merge<br />

import/export


1 The <strong>OMFIT</strong> <strong>framework</strong><br />

The <strong>OMFIT</strong> tree structure<br />

<strong>OMFIT</strong> interactive graphical environment<br />

Additional features: user GUIs, high level APIs, MDS+ integration,<br />

visualization<br />

2 Achievements <strong>and</strong> physics results<br />

Control-room analysis <strong>of</strong> gyro-kinetic stability<br />

MHD stability survey<br />

Magnetic flutter model for RMP-induced transport<br />

NTV model for NRMF-induced rotation<br />

Scenario <strong>development</strong> for DIII-D AT discharges<br />

3 Development <strong>of</strong> <strong>steady</strong>-state scenarios for FDF<br />

Improved LHCD system <strong>of</strong> FDF<br />

Integrated <strong>modeling</strong> in <strong>OMFIT</strong> with Eφ = 0 <strong>and</strong> β constraints<br />

Preliminary results using GCNMP/GLF23 transport model<br />

4 Conclusions <strong>and</strong> future <strong>development</strong>s


<strong>OMFIT</strong> used for control-room MSE+kinetic constrained EFITs<br />

<strong>and</strong> analysis <strong>of</strong> gyro-kinetic stability<br />

Experiments aimed at probing critical gradient <strong>and</strong> stiffness <strong>of</strong> <strong>the</strong><br />

electron temperature pr<strong>of</strong>ile by localized electron cyclotron heating<br />

• Integration with DIII-D pr<strong>of</strong>ile fitting tools, ECH <strong>and</strong> NBI from<br />

experiments (via existing IDL routines)<br />

• <strong>OMFIT</strong> provided quick feedback to guide experiments TGLF<br />

simulations to identify modes structures as a function <strong>of</strong><br />

wavenumber <strong>and</strong> radii<br />

ETG<br />

40 TGLF local linear stability simulations, run 10 at <strong>the</strong> time in parallel S. Smith APS 2012<br />

ITG


Survey <strong>of</strong> linear MHD stability at increased βn<br />

Pressure scanned<br />

by scaling <strong>of</strong> P ′ <strong>and</strong><br />

evaluated MHD<br />

stability for different<br />

toroidal mode<br />

numbers n <strong>and</strong> wall<br />

distances<br />

(conformal)<br />

220 GATO simulations run 20 at <strong>the</strong> time in parallel on 3 machines<br />

N =3.4 N =3.9 N =4.4 N=4.9 N=5.4


<strong>OMFIT</strong> as <strong>the</strong> <strong>framework</strong> for testing magnetic flutter model<br />

for RMP-induced transport<br />

Magnetic flutter model reproduces fairly<br />

well Te at top <strong>of</strong> pedestal<br />

Workflow:<br />

• M3D-C1 calculates spectrum <strong>of</strong><br />

perturbations<br />

• Experimentally measured “I” <strong>and</strong> “C”<br />

coil currents, linearly combine<br />

perturbations<br />

• Magnetic flutter model implemented<br />

in <strong>OMFIT</strong> predicts χe based on δB<br />

• ONETWO transport code calculates<br />

pedestal Te<br />

P. Raum APS 2012; S. Smith ITPA 2012


<strong>OMFIT</strong> as <strong>the</strong> <strong>framework</strong> for testing NTV model for<br />

NRMF-induced rotation<br />

Calculated torque from new <strong>framework</strong> is<br />

consistent with measured NRMF torque<br />

Similar workflow to magnetic flutter study<br />

A.J. McCubbin APS 2012


Validation <strong>of</strong> <strong>the</strong> NTV model with <strong>the</strong> DIII-D experiment<br />

NTV calculated on a grid <strong>of</strong> C-coil n=1 amplitude <strong>and</strong> phases<br />

• Linear combinations <strong>of</strong> intrinsic error <strong>and</strong> C-coil field<br />

• 121 points, yet < 10 min thanks to parallelization<br />

• NTV model predicts peak torque which is not necessarily<br />

coincident with empirical experimental observation<br />

<strong>OMFIT</strong><br />

M3D-C1<br />

C coil<br />

ZIPFIT<br />

ONETWO<br />

Kinetic<br />

EFIT<br />

EFIT<br />

M3D-C1<br />

Error elds<br />

Linear<br />

Combinations<br />

Parallelized<br />

Flux-surface<br />

Average<br />

NTV<br />

Calculations<br />

NTV<br />

Contour Map<br />

Integrated NTV torque relative to empirical optimum<br />

NTV<br />

optimum<br />

Empirical<br />

optimum<br />

C. Paz-Soldan


Workflow to optimize DIII-D AT discharge<br />

(e.g. maximize βn) was implemented in <strong>OMFIT</strong><br />

xneo<br />

EXB<br />

EFIT<br />

Magnetics + MSE<br />

Prole tting<br />

ZIPFITproles<br />

Script-based proles tting<br />

GAproles<br />

Interactive proles tting<br />

EFIT<br />

Scale kinetic P constraint<br />

257x257 , ε=1E-6<br />

Evaluate maximum achievable βn<br />

Kinetic constraint <strong>of</strong> EFIT equilibrium<br />

ONETWO<br />

Analysis mode<br />

(NFREYA, TORAY)<br />

EFIT<br />

Magnetics + MSE + Kinetic<br />

GATO<br />

Evaluate MHD stability<br />

realistic wall, n


1 The <strong>OMFIT</strong> <strong>framework</strong><br />

The <strong>OMFIT</strong> tree structure<br />

<strong>OMFIT</strong> interactive graphical environment<br />

Additional features: user GUIs, high level APIs, MDS+ integration,<br />

visualization<br />

2 Achievements <strong>and</strong> physics results<br />

Control-room analysis <strong>of</strong> gyro-kinetic stability<br />

MHD stability survey<br />

Magnetic flutter model for RMP-induced transport<br />

NTV model for NRMF-induced rotation<br />

Scenario <strong>development</strong> for DIII-D AT discharges<br />

3 Development <strong>of</strong> <strong>steady</strong>-state scenarios for FDF<br />

Improved LHCD system <strong>of</strong> FDF<br />

Integrated <strong>modeling</strong> in <strong>OMFIT</strong> with Eφ = 0 <strong>and</strong> β constraints<br />

Preliminary results using GCNMP/GLF23 transport model<br />

4 Conclusions <strong>and</strong> future <strong>development</strong>s


Previous work found self-consistent <strong>steady</strong>-state Q=3<br />

scenario for FDF tokamak<br />

Solution found by taking<br />

advantage <strong>of</strong> disparate<br />

transport <strong>and</strong> current<br />

diffusion timescales<br />

Baseline EC+LH scenario<br />

has desirable AT features<br />

Poor LHCD performance<br />

lead to considering ECH<br />

only solution<br />

A. Gar<strong>of</strong>alo, APS 2012<br />

FDF Modeling Yields Full Non-Inductive Current Drive<br />

<strong>and</strong> High Fusion Power with Combination <strong>of</strong> EC <strong>and</strong> LH<br />

No NBI<br />

1.5-D<br />

<strong>On</strong>eTwo+GLF23+EFIT<br />

simulation<br />

Most<br />

challenging<br />

scenario, but<br />

ECH/ECCD<br />

only<br />

ECH/ECCD<br />

+ LHCD<br />

• Simplifies Tritium<br />

containment<br />

• Increases area for Tritium<br />

breeding<br />

• Negative-ion NBI technology<br />

uncertain<br />

Too much ECCD power<br />

required for <strong>steady</strong>-state<br />

baseline scenario (β N ~3.7)<br />

Steady-state baseline equilibrium:<br />

Q~3, Pfus ~230 MW,<br />

PEC =55 MW, PLH =25 MW,<br />

H98y2 ~1.2, ƒBS ~70%<br />

Steady-State Equilibrium for Baseline FDF<br />

Mode Has Desirable AT Features<br />

V. Chan, APS 2011


<strong>OMFIT</strong> used to find new solution based on<br />

improved LHCD system<br />

Existing LHCD system suffers <strong>of</strong> wave<br />

accessibility problems<br />

Waves are trapped at <strong>the</strong> plasma edge<br />

between <strong>the</strong> slow <strong>and</strong> fast cut<strong>of</strong>f layers<br />

• Due to high density <strong>and</strong> steep density<br />

gradient (H mode)<br />

• Poor control <strong>of</strong> <strong>the</strong> location <strong>of</strong> <strong>the</strong> power<br />

deposition pr<strong>of</strong>ile (chaotic solution?)<br />

<strong>OMFIT</strong> was used to asses possible solution:<br />

1 Increase n <br />

2 Change launching location<br />

Z<br />

2.0<br />

1.5<br />

1.0<br />

0.5<br />

0.0<br />

0.5<br />

1.0<br />

<strong>and</strong> find a new <strong>steady</strong> state equilibrium 1.5<br />

Ray trajectories


Solution 1: Increase n makes waves accessible <strong>and</strong><br />

reduce <strong>the</strong> amount <strong>of</strong> refraction<br />

Ray trajectories<br />

Problems:<br />

• Low current drive efficiency <br />

∝ 1/n • Power deposited at<br />

<strong>the</strong> very edge<br />

<br />

Te(keV ) ≥ 50/n 2 <br />

Z<br />

2.0<br />

1.5<br />

1.0<br />

0.5<br />

0.0<br />

• Lower power h<strong>and</strong>ling<br />

• Coupling is more difficult (high n waves are 0.5<br />

more evanescent below cut<strong>of</strong>f density)<br />

1.0<br />

1.5


Solution 2: Inner wall launching location is promising<br />

but may be an engineering challenge Ray trajectories<br />

• n tends to upshift<br />

→ Higher n at higher ne<br />

• Density gradients are less steep<br />

→ Less refraction<br />

• Higher B<br />

→ Less stringent accessibility criterion<br />

• No trapped electrons<br />

→ Higher η CD<br />

• Robust single-pass absorption<br />

• Less power lost in <strong>the</strong> SOL<br />

• Reliable control <strong>of</strong> LHCD pr<strong>of</strong>ile<br />

How to launch <strong>the</strong>se waves?<br />

Z<br />

2.0<br />

1.5<br />

1.0<br />

0.5<br />

0.0<br />

0.5<br />

1.0<br />

1.5


Solution 2: Poloidal-splitter antenna to launch LH waves<br />

from <strong>the</strong> high field side<br />

Same antenna type used in Alcator C-Mod (low field side), <strong>and</strong><br />

proposed for <strong>the</strong> Vulcan Tokamak (high field side)<br />

• Quiescent plasma is good for coupling<br />

• Central stack as toroidally Y.A. Podpaly continuous et al. / Fusion limiter Engineering <strong>and</strong> Design 87 (2012) 215– 223<br />

Y. Podpaly, FED 2012<br />

g. 12. Vulcan LHCD concept illustration, with LH waveguide <strong>and</strong> launcher shown<br />

Plasma<br />

Dierential phase<br />

Output waveguides<br />

Common phase<br />

A<br />

D<br />

B<br />

C<br />

Fig. 13. Estimated Phase shifter power losses in avera<br />

klystron to plasma.


Solution 2: Sensitivity study to inner wall poloidal launch<br />

location <strong>and</strong> launched n<br />

150 GENRAY simulations run overnight in <strong>OMFIT</strong> to search for:<br />

• Highest current drive efficiency<br />

• Deepest wave penetration<br />

“Reliable optimum” found over a broad region in parameter space


Optimized inner wall launch shows superior characteristics<br />

compared to outer wall launch


Optimized inner wall launch shows superior characteristics<br />

compared to outer wall launch


Previously: Eφ = 0 <strong>and</strong> β = βtarget imposed by<br />

adjusting LH <strong>and</strong> ECH powers by trial <strong>and</strong> error<br />

Separation <strong>of</strong> transport <strong>and</strong> current evolution timescales<br />

<strong>On</strong>eTwo/GLF23<br />

~1 sec. evolution<br />

j, T e , T i ,<br />

Vary heating to<br />

match target beta<br />

Free-boundary<br />

EFIT<br />

Core<br />

density<br />

pr<strong>of</strong>ile<br />

shape<br />

similar to pressure pr<strong>of</strong>ile<br />

Free-boundary<br />

EFIT<br />

<strong>On</strong>eTwo/GLF23<br />

E / r = 0<br />

j<br />

Vary current drive<br />

to obtain Ef =0<br />

User-intensive process: painful, lengthy <strong>and</strong> prone to mistakes


<strong>OMFIT</strong> driven simulation with simultaneous constraint <strong>of</strong><br />

Eφ = 0 <strong>and</strong> β = βtarget imposed by Newton method<br />

ONETWO/GCNMP<br />

Evolve Te,Ti<br />

~1 sec, xed sources<br />

Find δβn/δPECH <strong>and</strong> change<br />

ECH power to obtain target βn<br />

PECH + δPECH<br />

ONETWO/GCNMP<br />

Evolve Te,Ti<br />

~1 sec, xed sources<br />

PECH<br />

ONETWO/GCNMP<br />

Evolve Te,Ti<br />

~1 sec, xed sources<br />

TORAY<br />

GENRAY<br />

statele.txt<br />

ONETWO<br />

Make self-consistent<br />

statele for GCNMP<br />

statele.nc inone<br />

eqdsk<br />

TORAY<br />

GENRAY<br />

EFIT<br />

Find equilibrium<br />

satisfying force balance<br />

inone statele.nc<br />

PLH<br />

ONETWO<br />

Evolve J<br />

δEφ/δρ = 0<br />

PLH + δPLH<br />

ONETWO<br />

Evolve J<br />

δEφ/δρ = 0<br />

Find δEφ/δPLH <strong>and</strong> change<br />

LH power to obtain Eφ = 0<br />

ONETWO<br />

Evolve J<br />

δEφ/δρ = 0<br />

eqdsk*


1st step <strong>of</strong> current evolution<br />

Eφ = 0 condition is well imposed with 20 MW <strong>of</strong> LH power<br />

Initial<br />

current evolution step 1


After 1s <strong>of</strong> GCNMP/GLF23 Te Ti evolution<br />

pr<strong>of</strong>iles develop fine structures <strong>and</strong> steep gradients<br />

0.1 s<br />

...<br />

1.0 s<br />

-6<br />

t=10 s<br />

• Strongly affects boostrap current → current drive & equilibrium<br />

• Not observed in original ONETWO/GLF23 simulations<br />

• Need to underst<strong>and</strong> why, it’s ongoing work


1 The <strong>OMFIT</strong> <strong>framework</strong><br />

The <strong>OMFIT</strong> tree structure<br />

<strong>OMFIT</strong> interactive graphical environment<br />

Additional features: user GUIs, high level APIs, MDS+ integration,<br />

visualization<br />

2 Achievements <strong>and</strong> physics results<br />

Control-room analysis <strong>of</strong> gyro-kinetic stability<br />

MHD stability survey<br />

Magnetic flutter model for RMP-induced transport<br />

NTV model for NRMF-induced rotation<br />

Scenario <strong>development</strong> for DIII-D AT discharges<br />

3 Development <strong>of</strong> <strong>steady</strong>-state scenarios for FDF<br />

Improved LHCD system <strong>of</strong> FDF<br />

Integrated <strong>modeling</strong> in <strong>OMFIT</strong> with Eφ = 0 <strong>and</strong> β constraints<br />

Preliminary results using GCNMP/GLF23 transport model<br />

4 Conclusions <strong>and</strong> future <strong>development</strong>s


Conclusions <strong>and</strong> future <strong>development</strong>s<br />

Newly developed <strong>OMFIT</strong> <strong>framework</strong> is a workflow manager which<br />

based on a bottom-up approach<br />

• Framework provides <strong>the</strong> tools while interaction is defined by users<br />

• Allow integration without <strong>the</strong> definition <strong>of</strong> st<strong>and</strong>ards<br />

→ tree structure<br />

<strong>OMFIT</strong> is being used for a wide range <strong>of</strong> tasks:<br />

• Kinetic EFITs & control room analysis: support for experimentalist<br />

• Validation <strong>of</strong> NTV/magnetic flutter <strong>the</strong>ories<br />

• Preliminary scenario <strong>development</strong> for DIII-D <strong>and</strong> FDF<br />

Future work:<br />

• Add more data-types <strong>and</strong> modules based on needs<br />

• Transition from ONETWO to GCNMP with sources calculated<br />

within <strong>OMFIT</strong><br />

• Tutorials, online help, support<br />

• Work closely with experimental DIII-D groups

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!