nov. 2, 2009d. mccune1 status of transp and ptransp presented at aps-dpp 2009, nov. 2--6, 2009...
TRANSCRIPT
Nov. 2, 2009 D. McCune 1
Status of TRANSP and PTRANSP
Presented at APS-DPP 2009,
Nov. 2--6, 2009
*Supported by U.S. DOE Contract No. DE-AC02-09CH11466
Nov. 2, 2009 D. McCune 2
Abstract – GP8.00051D. McCune, R. Andre, E. Feibush, M. Gorelenkova, K. Indireshkumar, C. Ludescher-Furth, L. Randerson, PPPL*, G. Bateman, A. Kritz, Lehigh University – This poster describes the status of TRANSP and PTRANSP code development and run production operations. Production rates continue to climb as new users and tokamaks are added; statistics will be shown, including utilization of the recently added TRANSP MPI capability. New code features include greatly expanded flexibility in specification of transport models for density, temperature, and angular momentum profile prediction in PTRANSP runs. A new replay option enables retrieval of sources from TRANSP analysis runs for PTRANSP validation. TRANSP / PTRANSP free boundary MHD modeling options have been enhanced. New RF modeling capabilities, such as the ability to model ICRF with full toroidal mode spectrum, have been added. NUBEAM upgrades include improvements to deposition atomic physics and loss orbit distribution capture. Fusion Grid post-processing of TRANSP results has been improved using extraction of SWIM-SciDAC Plasma State files from TRANSP archives. Opportunities for PTRANSP Support of SciDAC and FSP will be presented.
*PPPL work performed under auspices of DOE contract DE-AC02-76CH03073
Nov. 2, 2009 D. McCune 3
TRANSP: Vision Statement
Provide a comprehensive end-to-end modeling
capability for magnetic confinement fusion energy experiments of today and
tomorrow.
Nov. 2, 2009 D. McCune 4
Traditional TRANSP: Overview
Preliminary dataAnalysis andPreparation
(largely automated)
DiagnosticHardware
Experiments (Asdex-U, C-Mod, DIII-D, ITER, JET, KSTAR, MAST, NSTX)
20-50 signals {f(t), f(x,t)}Plasma position, Shape,Temperatures, DensitiesField, Current, RF andBeam Injected Powers.
TRANSP Analysis*:Current diffusion, MHD equilibrium, fast ions, heating, current drive;
power, particle and momentum balance.
TRANSP Analysis*:Current diffusion, MHD equilibrium, fast ions, heating, current drive;
power, particle and momentum balance.
Experiment simulationOutput Database
~1000-2000 signals{f(t), f(x,t)}
Visualization
Load RelationalDatabases
Detailed (3d) time-slice physics simulations: GS2, ORBIT, M3D… *FusionGrid TRANSP
on PPPL servers
Pre- and Post-processing at the experimental site… MDS+
MDS+
Nov. 2, 2009 D. McCune 5
PPPL TRANSP Run Production*
0
1000
2000
3000
4000
5000
6000
Completed Runs
FY-2005
FY-2006
FY-2007
FY-2008
FY-2009**
Fusion Grid TRANSPFusion Grid TRANSP
*Not shown in graphic: independent run production at JET site (approximately 1000 runs per year).**59 NUBEAM/TRANSP production runs with MPI, up to 16p.
Nov. 2, 2009 D. McCune 6
PPPL TRANSP SERVICEFY-2005 – FY-2009 (Five Years)
0
1000
2000
3000
4000
5000
6000
Total Runs, by Tokamak
NSTX
MAST
DIII-D
Asdex-U
C-Mod
JET*
ITER
HL-2A
EAST
other*~4000 additional JET runs on JET production server over same period.
18003 Total Runs
Nov. 2, 2009 D. McCune 7
FY-2009 PPPL TRANSP TeamName TRANSP PTRANSP SciDAC Other
R. Andre 70% 30%
Marina G. 100%
Kumar 65% 35%
McCune 25% 35% 20% 20%
Feibush 20% 15% 65%
Ludescher 80% 10% 10%
Randerson 60% 40%
Total FTE 4.2 0.75 0.70 1.35Color code: Physics, Visualization, Engineering/ Operational Support
Nov. 2, 2009 D. McCune 8
TRANSP Staffing Level• Staff levels still insufficient:
– Cannot promptly meet short term code development requests of production users;
– Long term development projects have lagged.
• Main reason: production system support.– Trouble-shooting 100s of crashed runs.– Detailed answers to 1000s of user queries.
• New hire in FY-2009: Marina Gorelenkova.– Financed by experimental projects: big help.
Nov. 2, 2009 D. McCune 9
A Cautionary Note for FSP• Production support very labor intensive!
• Immediate user needs take precedence.– FTEs for long term development = Total FTEs
less FTEs for support at best.– It’s worse than this, due to nature of support:
• Repeated interruptions;• Numerous short term, urgent, unplanned tasks;• User’s research dependent on speedy completion;• Highly disruptive to long term planning and effort.
• [P]TRANSP code development suffers.
Nov. 2, 2009 D. McCune 10
Separate Production Support and Code Development Roles?
• Might work for a sufficiently large team.
• But, support personnel need knowledge similar to that of code authors.
• Trouble shooting crashed runs = response to user “beta testing” of research code.
• Integral part of “Validation & Verification”.
• Code developer involvement is important.
Nov. 2, 2009 D. McCune 11
A Day in the Life: Oct. 20, 2009 (zero new lines of code)
• Detailed email exchange:– JET physicist query,
momentum balance terms. – TRANSP extracted Plasma
States for GYRO @ GA.– Ideas for NUBEAM
simulation of mode induced diffusion of fast ions.
– FMCFM module design for PTRANSP, FSP, etc.
– NUBEAM development planning (ADAS commit).
– SWIM progress report for FY-2009 for DOE.
• More emails:– NUBEAM failure in MST-
RFP simulations.– Diagnosis of TRANSP
output system failure, 160GB ITER file (output time step too small).
– NSTX physicist query, fast ion trapping definition.
– Help sought, debug DIII-D “Coppi-Tang” predictions.
– Cursory review, two co-authored papers.
– Join FSP “frameworks” email list.
Nov. 2, 2009 D. McCune 12
New for Users This Year• Fusion Grid NUBEAM, MPI up to 16p:
– ADAS atomic physics: state excitation effects on neutral beam deposition.
– New current shielding option: Y.R. Lin-Liu, F.L. Hinton, Phys. Plasmas 4 (1997) 4179.
– New outputs @ times of interest OUTTIM(…):• Both GC & FLR computed slowing down ion fb(…);• Detailed deposition distribution function data;• Distribution data on lost orbits.• All available using GET_FBM.
• Same NUBEAM in SciDACs: CSWIM, FACETs– MPI Scaling to ~1024p, 1-4000 ptcls/cpu.
Nov. 2, 2009 D. McCune 13
New for Users This Year• NUBEAM_Fgen – see C. Stark, JP8.0060:
– Replay NUBEAM calculation over short time window; produce data file for GET_FBM.
– ElVis (Fusion Grid) User Interface;– Start from any existing NBI TRANSP run, any time of
interest; replay using Plasma States…– Choose particle statistics (NPTCLS);– Choose number of CPUs up to 16;– SULI summer project.
• PTRANSP Free Boundary MHD equilibrium– Details below…
Nov. 2, 2009 D. McCune 14
New for Users This Year• Plasma State based heating and current
drive source replay:– TRANSP rerun using sources from prior run.– Sample application (PTRANSP): test transport
model variations, holding sources fixed.
• “Steerable” namelist quantities.• Two production versions: tshare & pshare:
– Latest code in tshare; tested version in pshare.
• Prompt input data error checking– At job queuing time; save queue waits.
Nov. 2, 2009 D. McCune 15
Technical Progress (Code Not Ready for Users)
• New RF-related Modules:– MPI-TORIC, GENRAY, CQL3D.– See GP8.00052 (next door).– TORIC Toroidal Mode Spectrum is available.
• PTRANSP– Density prediction: tested, some aspects OK, still
needs work.– Control Upgrades (being implemented):
• Choice of solver algorithm• Choice of summed transport models • Axial, Confinement & Edge plasma regions…
1 1 2 2 ...c M c M
Nov. 2, 2009 D. McCune 16
Details: ADAS in NUBEAM
Figure from “ADAS in Fafner and TRANSP/ NUBEAM” – Michael Kraus et al, October 2009 ADAS Workshop.
Asdex-U 17847n(0) ~ 1.2e20 m**-3NBI source #1
• ADAS atomic physics – correction for atomic state excitation – now available in NUBEAM for H-isotope neutral beams• Significant change from ground state for high target densities (shown).• Improvement over legacy excitation model Janev, C.D. Boley, D.E. Post, Nucl. Fusion Vol. 29, No. 12, pp.
2125-2140 (1989)
Collaboration: PPPL & IPP/Garching
Nov. 2, 2009 D. McCune 17
Details: Isolver Free Boundary Solver• Isolver is a free boundary equilibrium solver originally written by J. Menard in IDL
and rewritten in Fortran for use in TRANSP/PTRANSP.• Development of free boundary capabilities proceeding in stages:
– EFIT analysis: Completed• Isolver analyzes psi(R,Z) from EFIT time slices used as input to TRANSP to find flux surface
representation, separatrix, X points, …
– Least squares coil currents (original Isolver mode): Completed• Isolver determines the poloidal field coil currents which will best match a prescribed
boundary.
– Measured coil currents: 70% Complete• Isolver uses the measured poloidal field coil currents along with a feedback circuit to keep
the plasma centered in the vessel.
– Passive structures: 30% Complete• The measured poloidal field coil currents are used along with a circuit equation to include the
currents induced in the nearby passive structures.
– Flux Diffusion: To Be Done• The poloidal field diffusion equation, which is used to evolve the plasma q, is coupled to the
equilibrium solver.• The poloidal field coil currents are driven through a circuit equation to best match the
measured coil currents.
– Plasma current prediction: To Be Done• The plasma current is predicted from the circuit equation.• The drive voltage of the ohmic coils is chosen to match a prescribed plasma current
waveform.
Nov. 2, 2009 D. McCune 18
Isolver: EFIT Analysis
• EFIT psi(R,Z,t) slice data is input to TRANSP/PTRANSP
– LEVGEO=8 mode in TRANSP/PTRANSP for prescribed equilibrium data.
– New NLPSIRZ_FLUX=.TRUE. mode uses psi(R,Z,t) input to evaluate flux surfaces.
• Isolver determines flux surface based on location of X points and limiter surface.
– Flux surface representation returned to TRANSP/PTRANSP is displaced in from separatrix to avoid singularities.
– Separate analysis is performed with boundary at separatrix to evaluate enclosed flux and self inductance Li.
• Available for any tokamak with EFIT and limiter data.
• For application to D3D, see Budny et al. APS poster JP8.00102, Tuesday afternoon session.
Nov. 2, 2009 D. McCune 19
Isolver: Least Squares Mode• Fit coil currents to best match prescribed
boundary.• Coil currents can be adjusted freely, fixed to
measured data or constrained by measured data.• Coil geometry currently available for NSTX and
D3D.– D3D example in Budny, et al. JP8.00102 poster.
• ITER coil geometry awaits TRANSP/PTRANSP output pipeline upgrade to accommodate large amount of psi(R,Z,t) data.
Note: Ohmic coil current KK_OH left unconstrained and became nonzero only during the part of the shot where Isolver had difficulty finding a solution.
NSTX
Nov. 2, 2009 D. McCune 20
Isolver: Measured Coil Currents Mode• Measured poloidal coil currents drive Isolver coil currents.• Feedback circuits position the plasma.
– Stabilizes plasma against vertical and horizontal motion.– Feedback circuit currents are added to measured coil
currents.– Prescribed boundary is used to identify feedback points for
centering the plasma.
Horizontal feedback point
Vertical feedback point
Horizontal feedbackcurrent source
Vertical feedbackcurrent source
DIII-D
Nov. 2, 2009 D. McCune 21
Details: Updatable Namelist Variables
Time NBI ECH Total80-130 34 40 74
130-300 17 30 47300-400 17 15 32400-500 0 4 4
Namelist variables designated with tilda (“~”) in TRANSP’s port.spec are “updatable”. For example: beam and RF powers. Hypothetical application:
[main namelist] … … nantech=1 ! 1 gyrotron nbeam=2 ! 2 beams powech=0.0 pinja=2*0.0
~update_time = 80.0 powech=40.0e6 pinja=17.0,17.0
~update_time = 130.0 powech=30.0 pinja=0.0,17.0 … … [etc.]
Nov. 2, 2009 D. McCune 22
Details: Running MPI NUBEAM
• In Namelist, set:– NBI_PSERVE = 1
• Command line job setup as before– tr_start <runid> [tshare]– It will ask you how many CPUs (Np)?– Note: Np > 16 likely to result in long queue
wait.
• In the near future:– NTORIC_PSERVE = 1 will be available.
Nov. 2, 2009 D. McCune 23
ElVis/Java Fusion Grid Services
• Available Services*:– Nubeam_Fgen, get
fast ion distrib. fcn.– Get_fbm, view fast ion
distrib. fcn.– Cstate, view Plasma
State file.– Rplot, view TRANSP
run results.– ElFresco, compute
reflectometry
• Recent Changes:– Nubeam_Fgen,
Cstate, and Get_fbm are new.
– Rplot:• MDS+ run selector
panel;• Script editor;• Arrow Key command
line recall.
– Heavy use of ElFresco service.*partial list
Nov. 2, 2009 D. McCune 24
ElVis/Rplot
Nov. 2, 2009 D. McCune 25
I-phone Has an App for that…
TRANSP Fusion Grid Users can access these services…
Nov. 2, 2009 D. McCune 26
Future Plans• Continue Run Production & Support.• Develop RF Modules with MPI.• Further Develop MPI NUBEAM.• Develop Free Boundary MHD Equilibrium.• Develop GUIs– ElVis/Java Services.• PTRANSP / FSP / SciDACs:
– PTRANSP winding down, much work needed.– Level of effort going forward is unclear.– Note: ITER task agreements need PTRANSP.
Nov. 2, 2009 D. McCune 27
1d Transport Solver Compatible with PTRANSP, FSP, SciDACs…
Control Object(s):• Choice of Solver• Choice of Transport Models• Options
P0: Plasma State @t0• Past time step geometry• Past time step profiles
P1: Plasma State @t1• new geometry• new input profiles
• missing profiles => equations to solve
• new boundary conditions• sources (heating, fueling).
Portable S
olver M
odule FM
CF
M
Collection of transport models:• NCLASS• GLF23• MMM95• MMM08• TGLF• etc., etc.
New profiles
Current Status: PTRANSP generates P0 and P1
Question: Shall we build the rest of this?
Nov. 2, 2009 D. McCune 28
Opportunity to Take Advantage of TRANSP Legacy
Plasma State Time Series from SciDAC/FSP SimulationIncludes Machine Description
TRANSP Namelist Time Dependent Equilibrium & Profiles & Sources
TRANSP Analysis
TRANSP Output Dataset
Access to Physics Modules that start with TRANSP data, including many first principles codes.
Question: Shall we build the rest of this?
Status: Replay Mode demonstrated transfer of sources from Plasma State time series.
Nov. 2, 2009 D. McCune 29
Summary• Continued heavy research production use
of TRANSP and PTRANSP.• Continued Active Development Program
– But some developments are “behind schedule”.
• PTRANSP funding & work coming to end.– Some goals not met, but, continuation of
PTRANSP activity could help FSP & SciDACS.– ITER task agreements depend on PTRANSP.
• The current TRANSP staff is stretched thin, mainly due to production support.