validation and analysis boston, ma usa june 1, 2017 · b.a. grierson / iaea tech fus. data / jun...
TRANSCRIPT
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
by
B.A. Grierson1, N. Logan1, S.R. Haskey1, L. Cui1, S.P. Smith2, O. Meneghini2, J. Buchanan3
1Princeton Plasma Physics Laboratory, Princeton, NJ 08540, USA2General Atomics, San Diego, CA 92121, USA3CCFE, Culham Science Centre, Abingdon, OX14 3DB, UK
Presented at the 2nd IAEA Technical Meeting on Fusion Data Processing, Validation and Analysis
Boston, MA USA
June 1, 2017
Interpretive Analysis and Predictive Discharge Modeling with TRANSP
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
OMFIT and TRANSP are Fulfilling Experimental and Core Predictive Whole Device Modeling Needs
● US DOE community workshops1 identified
need for streamlined experiment/theory
comparison— OMFIT2 framework provides such
workflows
● Motivates common set of tools across
machines for processing tokamak data,
managing code runs, visualizing results
● Community based development leverages
expert knowledge for emerging needs
✓✓
✓
1science.energy.gov link2O. Meneghini, et. al., Nucl. Fusion 55 (2015)
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Accurate Tokamak Power Balance Analysis Is the Cornerstone of Transport Interpretation and Model Validation1
● Understanding the balance of energy, particles and
momentum fluxes relies on:— High quality time-dependent profile data2
— Accurate source calculations
● TRANSP3 is a commonly used time-dependent
transport code for interpretive transport analysis and
predictive simulations— Wide use in the US and international
● OMFIT4 is streamlining data preparation, diagnostic
consistency, and interpretive → predictive workflows
1C. Holland, Phys. Plasmas 23 060901 (2016) and references therein
2N. Logan, This Afternoon, O-323http://transpweb.pppl.gov 4S.P. Smith, Tomorrow, O-43
Time derivative of i.e. X = n(x,t)
S = sources/sinks from neutral beams, RF, recycling, 3D fields
Spatial derivative of i.e. X = n(x,t)
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Data Processing and Preparation for TRANSP Varies Widely Across the Tokamak Community
● Contrasting DIII-D, JET, NSTX, C-Mod, etc…
displayed wide range of inputs, diagnostics
and mappings
— All flavors handled inside of OMFIT
● New multi-machine strategy— Tokamak independence for common tasks
such as OMFIT equilibria and profiles
— Tokamak dependent for device specific inputs
such as auto EQ, profiles, heating, ...
● At highest level alleviate reliance on existing
tokamak data
— New ability to design shots
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
DIII-D Reference Shot Provides Power Scan For Time-Dependent TRANSP Demonstration
● Power ramp-up over five
seconds
— Assesses L-H power threshold
— All profile diagnostics enabled
● Sawtoothing plasma with little
other core MHD
● Sufficient information for
assessing time-dependent
diagnostic verification
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Goal of Experimental Analysis Determines TRANSP Operating Mode and Fidelity of Heating Sources
● TRANSP input namelist is configurable to
meet wide range of needs
● Common use cases:— Survey, diagnostic checkout BEAST mode
— Power/particle/momentum balance
analysis mode
— Current diffusion for kinetic equilibrium
reconstruction
— Energetic particle physics and fast-ion
distribution function studies
Mode BEAST Analysis kEFIT EP
ZONES 20 50 100+ 50
Timestep 0.020 0.010 0.020 0.001
Particles 5k 32k 5k 128k++
Ex. Time NBI only
20min/s 3.2h/s 20min/s ++
Fastest possible without being worthless, useful for global and vol. integrated quantities
Sufficient space/time resolution for transport fluxes
Spatial resolution for edge bootstrap current, low heating for fast particle stored energy and neutrons
Short timestep for capturing rapid beam turn-on, highest MC statistics for dist. function
Many namelist switches to set
Operating modes scripted in OMFIT with drop-down menu
16 TRANSP variables7 NUBEAM variables
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Increase of Heating Package Fidelity Required for Accurate Statistics for Torque Calculations
● Global quantities weakly affected by
poor monte-carlo statistics
— Total neutron rate
— Vol. integrated quantities Wth
, Wfast
— Surf. integrated quantities INBI
● Derived profile quantities require
increased fidelity
— NBI prompt JxB torque and
momentum diffusivity particularly
sensitive to MC statistics1
Analysis
BEAST
1Mantica, et. al. ,Phys. Plasmas 17 (2010)
EP
Analysis
BEAST neutrons/sec
Torque (Nm/cm3)
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Increasing Fidelity Required for Accurate Fast-particle Distribution Function
● EP modes driven by
gradients in phase space
● Requires extremely high
monte-carlo statistics
BEAST Mode 0.1 hrs x1
unresolved
Step-wise evolution
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Increasing Fidelity Required for Accurate Fast-particle Distribution Function
● EP modes driven by
gradients in phase space
● Requires extremely high
monte-carlo statistics
Analysis Mode 1.0 hrs x8
qualitative
peaking
Smooth response
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Increasing Fidelity Required for Accurate Fast-particle Distribution Function
● EP modes driven by
gradients in phase space
● Requires extremely high
monte-carlo statistics
EP Mode High Fidelity 7.6 hrs x64
quantitative
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Data Consistency Checks Allow the Quantification of Systematic Uncertainties in the Plasma Measurements
● Plasma total stored energy and neutron rates
commonly used for data consistency checks
— Stored energy from profile analysis should
match equilibrium energy from magnetics
— Neutron rate should match classical prediction
in absence of EP modes
● Further metrics for resistive current evolution
— Internal inductance li
— Surface loop voltage (plasma resistivity)
— MSE pitch-angle evolution
Impurity dilution reduces Wth through ni
ne peaking and QNImpurity dilution reduces nD
Peaked near axis
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Common Experimental Uncertainty is the Plasma Impurity Content
● Electron density, temperature profiles
typically well-constrained— Thomson + interf, ECE
● Single-ion density profile may be
available through charge-exchange— Survey spectroscopy may indicate
many other low-Z impurities
— Visible bremsstrahlung may indicate
Zeff
above single-impurity
● Motivates systematic Zeff
variation to assess
data consistency
— TRANSP SCAN provided by OMFIT
Total Stored Energy
Total Neutron Rate
⨉0.6
⨉1.4
⨉0.6
⨉1.4
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Neutron Rate and Stored Energy Exhibit Distinct Responses to Variation in Z
eff
● Increasing Zeff
directly depletes the
thermal deuterium density— Target for beam-plasma neutrons
— Absolute neutron calibration ~20%
● Increasing Zeff
reduces nD
more than it
increases nZ
— Response of Wtot
weaker
than neutrons
— Absolute equilibrium
stored energy ~ few %
● Confidence depends on particular machine
● Bounds Zeff for further analysis (i.e. GK)
Both metrics indicate overall Zeff > only carbon
Derived Zeff Derived Zeff
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Resistive Current Evolution Provides MSE Pitch Angles for Assessing Neoclassical Resistive Current Diffusion
Measured
Simulation
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Resistive Current Evolution Provides MSE Pitch Angles for Assessing Neoclassical Resistive Current Diffusion
Measured
Simulation
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Routine Predictive Simulations Maximize the Scientific Utilization of Scarce Resources; Experimental Runtime
While interpretive analysis seeks to accurately
quantify power flows and transport coefficients
Core transport simulations seek to
validate transport models
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Predictive Simulation Capabilities Now Being Routinely Utilized Through Standardized Workflow and Best Practices
● Core transport simulations replace the
fluid variables with profiles derived from
transport models
— Local transport models provide flux given
gradient
— Profile is defined by integral of local
gradients
● Enabling a predictive TGLF1 simulation in
TRANSP with PT_SOLVER requires setting
at least 50 namelist variables
Start with experimental profiles
→ Choose transport model→ Choose transport channels→ Set radial boundary condition→ Set transition time to predictive
Many namelist switches to set
Operating modes scripted in OMFIT with drop-down menus guiding switches and logic checking
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Standardized OMFIT Visualizations and Post-Processing Provides Profile Evolution and Validation Metrics
● TRANSP run interpretively in
analysis mode
● TRANSP run repeated with
TGLF to predict Te, Ti
— 15 hrs wall time
— 8 NUBEAM,128 TGLF CPUs
● How did TGLF do,
quantitatively?
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Standardized OMFIT Visualizations and Post-Processing Provides Profile Evolution and Validation Metrics
● TRANSP run interpretively in
analysis mode
● TRANSP run repeated with
TGLF to predict Te, Ti
— 15 hrs wall time
— 8 NUBEAM,128 TGLF CPUs
● How did TGLF do,
quantitatively?
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Standardized OMFIT Visualizations and Post-Processing Provides Profile Evolution and Validation Metrics
● TRANSP run interpretively in
analysis mode
● TRANSP run repeated with
TGLF to predict Te, Ti
— 15 hrs wall time
— 8 NUBEAM,128 TGLF CPUs
● How did TGLF do,
quantitatively?
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Time-Dependent Transport Model Validation Metrics Provided by OMFIT for List of TRANSP Runs
● Historically validation metrics1,2
produced for single timeslice
● Time-dependent simulations provide
metrics as EQ, heating are varied
CORE
PEDESTAL1ITER Physics Basis T&C Nucl. Fusion 39 (1999)2Kinsey, J. et. al., Phys. Plasmas 15 (2008)
GLF23 TGLF
Expt
GLF23
TGLF
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Predictive Transport Simulations with TGLF Provide Time-Evolving Turbulence Characteristics
● Validation metrics quantify
accuracy and utility of model
● Insight into the nature of the
turbulence provided by linear
modes and flux spectra
— Multi-dimensional space for ⍵
and ᶕ in (⍴,t,k)
● Ultimately, knowledge of
instability informs control
strategy
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
Predictive Transport Simulations with TGLF Provide Time-Evolving Turbulence Characteristics
● Validation metrics quantify
accuracy and utility of model
● Insight into the nature of the
turbulence provided by linear
modes and flux spectra
— Multi-dimensional space for ⍵
and ᶕ in (⍴,t,k)
● Ultimately, knowledge of
instability informs control
strategy
Growth rate increases in time with heating power
Early
Late
Ion
Electron
B.A. Grierson / IAEA Tech Fus. Data / Jun 2017
OMFIT and TRANSP are Fulfilling Experimental and Core Predictive Whole Device Modeling Needs
● Community workshops1 identified need for
streamlined experiment/theory comparison
— OMFIT2 framework provides such workflows
● Achieved common set of tools across
machines for processing tokamak data,
managing code runs, visualizing results
● Community based development is
leveraging expert knowledge for emerging
needs for the future
✓✓
✓
1science.energy.gov link2O. Meneghini, et. al., Nucl. Fusion 55 (2015)