Transcript IRTMC at JET - ICALEPCS 2005

Real Time Measurement and Control at JET
Overview & Status
Robert Felton1, and JET EFDA Contributors
1
Euratom / UKAEA Fusion Association, Culham Science Centre, Abingdon, Oxon, OX14 3DB, UK
This work this work has been performed under the European Fusion Development Agreement.
It is funded in part by the United Kingdom Engineering and Physical Sciences Research Council and by EURATOM
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
1
JET = Joint European Tokamak
•
•
Fusion plasmas in reactor-relevant conditions
Theory - Deuterium and Tritium easiest to access
–
–
–
–
–
D + D = T 1MeV + p 3MeV
D + T = 4He 3.5 MeV + n 14 MeV
temperature 100 M oC,
density 2-3 x 1020 m-3 (1 mg m-3)
confinement > 1s
• improved confinement modes
– complex interplay of magnetic and kinetic forces
• internal and edge instabilities with pressure gradients
• short and long range forces: not “classical ideal gas”
•
Practical - Toroidal Magnetic Confinement
–
–
–
–
magnetic confinement, shape and current
power loads on vessel components
particle fuelling and exhaust
impurities from plasma-wall interaction
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
2
JET = Joint European Tokamak
•
Machine Engineering - many and varied issues
–
–
–
–
–
–
–
–
–
vessel
wall
vacuum
magnets
heating
fuelling
radiological
remote-handling
diagnostics
toroidal R 3m, r 2m, 200 m3, Inconel
CFC tiles (Beryllium and Tungsten coming)
base 10-8 mBar (cryo), plasma 10-5 mBar
32 Toroidal, 9 Poloidal, ~ kV, ~ kA
NB, 20MW, RF 30MHz 8MW, 3.7GHz 10MW
12 gas injectors + pellet; ~500 mBarl per pulse
Biological shield, Tritium compatibility
radioactive and toxic (Be) components
magnetic, thermal, optical x-ray .. visible, neutronic ...
Pulsed ~ 10s 300MJ
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
3
JET = Joint European Tokamak
•
Systems Engineering - many and varied
– machine control
• hierarchical, distributed, pulsed
• home-grown
– real-time communications
• analogue, digital signals
• data packet networks
– operations data
• 15000 points, 35000 pulses and growing
– data acquisition
• 1ns … 1s, nV .. kV,
• VME, PCI, CAMAC, PLC
– data analysis
• traditionally post pulse,
• increasingly real-time
– remote participation
• VRVS
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
4
Tokamak Measurement & Control
Hierarchical machine control
Systems (vessel, magnets, gas, auxiliary heat & fuel, diagnostics)
Independent, with common, distributed time-base (fibre-optic + local decode)
Controlled by specific Operators
Connected by ethernet (TCP/UDP/IP; > 100 systems, miles of copper/fibre)
Operations (experiments)
Parameter sets designed by Session Leader in pulse schedule
Distributed to the Systems by Level1 Supervisor infrastructure
Checked and loaded to machine by Engineer-in-Charge, and System Operators
Distributed real-time control
Systems
Real-time, calibrated outputs (avoid device dependence)
Real-time data sent to/from a Central Controller over ATM AAL5 (~ 40 systems)
Central Controller has its own Operator (PDO)
Operations
Control algorithm - conceptualised by Scientists, realised by PDO
Event driven (step NB on n=2 mode) and feed-back (3He conc, q-profile)
High level language in pulse schedule
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
5
Hierarchical Machine Control
L1 Machine Supervisors
user interface
component data
parameter data
results data
user & system logs
L2 Machine Systems
control & status
start & stop
set-up & readout
r-t signal processing
r-t physics
Level 1
Magnets
Magnets
Pulse
Gas
Gas
parameters
Level 2
NB oct4
RF
Level 3
psu
Heat
Heat
...
gas valve
results
LIDAR
control
laser
Diagnostics
Diagnostics
ECE
...
status
recorder
...
L3 Device Drivers
specific functions
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
6
Machine Operations
Check &
Load
Edit
Pulse
Schedule
Pulse
Schedule
SL
Pulse
Schedule
EIC,
Operators
Pulse
Schedule
JET
plant state
log
Pulse Schedules
reference to other pulse schedules or JET
pulses
convert physics parameters to control
parameters.
validate parameters for consistency and
safety.
non-experts use expert scenarios for
otherwise tricky situations (shape)
October 2005
The EIC and Operators validate the
parameters (JET Operating
Instructions) and load the plant.
Other users (e.g. Heating,
Diagnostics) set-up their
equipment.
The Plasma Duty Officer prepares
and loads Real-Time Control
Algorithms.
Run
Pulse
EIC
ICALEPCS 2005 / RTMC at JET / R.Felton
JET
machin
e
7
The JET Real-Time Control Facility - Basic
Shape & Current
Control
Magnetics
PF Coils
Interferom Density
GAS + Pellets
NBI
plasma
ICRH
LHCD
TAE
Comms network
analogue
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
8
The JET Real-Time Control Facility - 2005
Shape & Current
Control (PPCC)
PF Coils
Magnetics
VUV impurities
GAS + Pellets
Interferom/Polarim
Vis Da, Brem, ELM
NBI
Neutron X-ray etc.
Vis H/D/T
ICRH
plasma
ECE Te (R)
Confinement
LHCD
CXS Ti (R)
q profile
TAE / EFCC
LIDAR Ne&Te(R)
Wall Load
MSE pitch (R)
Flux surfaces EQX
EQX kinetic map
Simulink code
Coil Protection
R-T Controller
R-T Signal Server
October 2005
X-ray Ti (0)
Comms network
ATM, some analogue
ICALEPCS 2005 / RTMC at JET / R.Felton
Pale blue = Diagnostic,
Sky blue = Analysis,
Red = Heating / Fuelling
/ Magnets & Power,
Yellow = PPCC (XSC),
Green = RTMC
9
Distributed Process Control (Real-Time)
•
Data
–
–
–
–
•
Level 2
Magnetics
Interferom
Connections
–
–
–
–
–
•
physics device independent
standard data sets
sizes : 4 to 400 float pt nos.
rates : 1 to 250 ms
fast, low latency < 0.15 ms
one-to-many
changes : local impact
isolation : fibre-optic
range : 1 .. 100 m
RTControl
NB
LH
q-profile
...
...
Technologies
– analogue messy
– ATM AAL5 configurable, reliable, available
– Industry standard, multi-platform, multi-vendor
•
Time - a seperate network
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
10
Diagnostics & Analysis
Earl Ferrers:
“My Lords, what kind of thermometer reads a
temperature of 140 million degrees centigrade
without melting?”
Viscount Davidson:
“My Lords, I should think a rather large one.”
from a debate on JET in the House of Lords (1987)
The JET
LIDAR
Thomson
scattering
system
October 2005
Wide range of processing techniques, and
space / time resolution
Filtering and down-sampling
Black Body Bolometer 48 chan, 2ms out
Cross-calibration factors
Electron Cyclotron Emission 96 chan. 2ms
Phase tracking of modulated signals
Far InfraRed Interferometry 15 chan. 2ms
Lock-in amplifiers (in software)
Motional Stark Effect 25 chan. 2ms
Levenberg Marquadt spectral fitting
Charge Exchange Spectr. 14 spectra, 50ms
Thomson Scattering
LIDAR laser 250ms, analysis 25 ms
Plasma magnetic boundary by Taylor expansion
“XLOC” 65 coeffs, 2ms
Finite element MHD equilibrium Grad-Shafranov
“Equinox” 500 pt mesh 25ms
Interpolation Te, Ne, q, etc on flux surfaces
“Equinox map” r/a = 0; 0.1; 1.0
ICALEPCS 2005 / RTMC at JET / R.Felton
11
Magnets, Heating & Fuelling
Physics
Inputs
Outputs
Rate
Shape & Current
Vertical Stability
Magnetics
PF currents [9]
Fast Radial Field
2 ms
0.2 ms
Gas & Pellets
Density Control
GIM[3]
Dens[3]
GIM[3]
Dens[3]
10 ms
10 ms
Pact[8]
10 ms
Neutral Beam
Preq[8]
120kV 60A 20 MW
Ion Cyclotron RF
25..50 MHz 4MW
Preq[4], dFreq[4]
Pact[4], dFact[4]
10 ms
Lower Hybrid RF
12 GHz 4MW
Preq[3]
Pact[3]
10 ms
Fact
10 ms
Alfven Eigenmode Freq
[n] refer to Groups == flexible selection of different NB PINIs, RF oscillators, antennae,
gasses, etc.
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
12
The Real-Time Controller
Preparation
•
•
Level1
User (PDO) designs and loads the algorithm
High level process block / data flow language
Diag.
Inputs
Operation
•
•
•
Level2
RTCC receives measurement data
RTCC evaluates the user algorithm
RTCC sends heat /fuel requests
Algorithm
Features
• flexible, general purpose (not low-level code)
• easy (for PDO) :
Event-triggered e.g. disruption avoidance, MHD
Feedback SISO e.g. b with NBI
•
difficult (even for PDO) :
MIMO control e.g. profiles
Vector, matrix calculations, state-space
Modular sub-routines
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
RTCC
evaluator
Heat/Fuel
Outputs
Real-time
10 ms cycle
13
The Real-Time Controller - Matlab/Simulink extension
Preparation
•
•
•
Matlab/Simulink & Level1
User designs Matlab / Simulink models
User generates C function, data and DLL files
User transfers the code and parameter files to RTMX
Operation
•
•
•
Level2
RTMX receives Diagnostic data, etc.
RTMX sends control requests to RTCC
RTCC relays the Heat/Fuel requests
Simulink
model
Features
•
•
•
•
Flexible
EFDA users work on control problem at home lab
Use Matlab / Simulink function libraries (discrete time)
Responsibilities
–
–
PDO still loads and runs RTMX and RTCC
Protection stays with Local Managers
Diag.
Inputs
RTMX
processor
RTCC
evaluator
Heat/Fuel
Outputs
Real-time
10 ms cycle
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
14
Control Design
r(t) or R(z)
reference
E(z)
error
C
Ufb(z)
feedback
P
y(t) or Y(z)
sensor
uff(t) or Uff(z)
operating point
u(t) or U(z)
actuator
System Identification
To obtain signals Actuator : u(t) e.g. PNB and Sensor y(t) e.g. bN
use theoretical models TRANSP, JETTO, ASTRA, CRONOS, GS2, …
or use experimental data.
Model the process P as a differential equation for y(t) resulting from u(t).
use State-Space or Laplace transforms : Y(z) = GP(z) . U(z)
Control Design
Design a controller C which achieves a desired reference signal r(t) by driving the
actuator u(t) using feedback of the measured signal y(t) within constraints (e.g.
error, settling time)
Check the controller C by simulation, using the process model P
U(z) = GC(z) . E(z)
October 2005
E(z) = R(z) - Y(z)
ICALEPCS 2005 / RTMC at JET / R.Felton
15
RT System Engineering
•
RT systems have been developed to satisfy JET Scientific Programme
– they work in parallel with existing measurement and control systems
– they integrate with existing system infrastructures
•
Even so, diversity and sustainability not always balanced
– Common Application Frameworks - HTTP protocol
• 1 VxWorks, 2 Windows - healthy competition - should have prize-giving !
– Common Platforms
• VME + PowerPC + VxWorks & PCI + PC + Windows - future ?
– Association-supplied Diagnostics “In-kind procurement”
•
•
Windows + Linux diverse interfaces, long-term support of internals ?
RT systems will evolve further
– Need to improve functional partitioning, and data distribution
– Model-based system engineering not yet established at JET way to go!
Diagnostic
Diagnostic
October 2005
Analysis
Analysis
Control
Control
ICALEPCS 2005 / RTMC at JET / R.Felton
Actuator
Actuator
16
Work In Progress (JET’s EP programme)
•
Magnets / Shape and Current Control
eXtreme Shape Control
Coil Protection System
•
Plasma Ops, CREATE, ENEA, CEA
Power Supplies
Heating and Fuelling
xxLM upgrade to PowerPC and ATM
RF frequency control, LH position control
•
CODAS
CODAS
Diagnostics
Bolometer, MSE, X-ray
Expts, CODAS
visible cameras, video distribution, hot spots Expts, CODAS
•
Analysis
Matlab / Simulink
Equinox and Polarimetry, MSE
Disruption Prediction
L-mode / H-mode
•
Plasma Ops
Plasma Ops, CEA, U.Nice
Plasma Ops, U.Naples, ENEA
Plasma Ops, & Murari
Databases & Communications
extend ATM network,
October 2005
Plasma Ops, CODAS
ICALEPCS 2005 / RTMC at JET / R.Felton
17
Long term To Do (JET’s EP2 programme)
•
Magnets / Shape and Current Control
– Vertical Stabilsation upgrade project ~ many Associations, ~ MEu !
– Error Field Correction Coils control ?
•
Heating and Fuelling
– ELM info for ITER-like antenna ?
– Pellet synch
•
Diagnostics & Analysis
–
–
–
–
•
EP2: Be / W Diagnostics, Neutron and Gamma Cameras
Alven Eigenmodes ?
RT magnetics analysis to speed up PostPulse Analysis
“Integrated” Analysis ancient (map onto flux) and modern (pattern recog.)
Databases & Communications & Computers
– try EPICS, MDSplus
– evaluate new network technology Is there an Integrated Services Data
Network (control, status, events, audio, video, time)?
– evaluate new computer technology PCIexpress, CELL
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
18
Summary
•
Real-time Diagnostics
– simplified operation and analysis :: reliable quick-look
– real-time processing will be “designed in” to many new Diagnostics
– limited by lines of sight, field of view, calibration dependencies
•
Real-time Magnets, Heating & Fuelling
– improving modelling and control algorithms for shape and stability
– improving power output and control
•
Real-time Experiment Control
– SISO and MIMO demonstrated; more sophisticated tools needed
•
Real-time Communications
– ATM ok - fast enough for most applications, flexible, reliable
•
Science Requirements (JET programme in support of ITER) can
best be satisified by Real-Time Measurement and Control
– Scientific Task Forces explore Plasma and Fusion Physics, and physicsbased control concepts - either simple or complex
– Real-time systems are the means to practically demonstrate the concepts.
October 2005
ICALEPCS 2005 / RTMC at JET / R.Felton
19