two classes of rwms internally non- resonant with -11 ≤ n < 0 externally non- resonant...
DESCRIPTION
Intelligent shell experiments on EXTRAP T2R EXTRAP T2R group J. R. Drake, Jenny-Ann Malmberg, Per Brunsell, Dmitriy Yadikin Chalmers theory group D. Gregoratto, Y. Liu, A. Bondeson RFX R. Paccagnella, S. Ortolani, P. Martin, G. Manduchi + many others. - PowerPoint PPT PresentationTRANSCRIPT
1Alfven Laboratory Mode Control Workshop, Austin 2003
Intelligent shell experiments on EXTRAP T2R
EXTRAP T2R groupJ. R. Drake, Jenny-Ann Malmberg, Per Brunsell, Dmitriy Yadikin
Chalmers theory groupD. Gregoratto, Y. Liu, A. Bondeson
RFX R. Paccagnella, S. Ortolani, P. Martin, G. Manduchi + many others
2Alfven Laboratory Mode Control Workshop, Austin 2003
Intelligent shell experiments on EXTRAP T2R
Outline of talk
1. Intro to RFP RWMs and motivation.
2. New RWM measurements on T2R
3. Description of T2R sensor and active coil arrays
4. Theory for feedback with partial shell coverage
5. First results with intelligent shell feedback applied
6. Plans for the future
3Alfven Laboratory Mode Control Workshop, Austin 2003
Unstable modes in Extrap T2R
‘Internally resonanttearing modes’
‘Internally non-resonant’RWMs
‘Externally non-resonant’ RWMS
‘Externally resonant’
-14
-12-11
-10
+5
+4
-13
q0
q=0
q<0
qa
axis edge
Toroidal mode numbern for m=1 modes
m=0
Two classes of RWMs
Internally non-resonant with
-11 ≤ n < 0
Externally non-Resonant with
0 < n < 7
4Alfven Laboratory Mode Control Workshop, Austin 2003
EXTRAP T2R front endNote:• The blue ”shell” surface• The spacing of the 64 TF coils
5Alfven Laboratory Mode Control Workshop, Austin 2003
Suitability of EXTRAP T2R for resistive wall mode active control studies
1. -relaxation < -shell < -pulse.
2. Internally resonant modes are rotating so their b-radial perturbation is suppressed.
3. Extensive magnetic diagnostics to measure mode spectra and growth rates.
4. RWM perturbations measured at b-r / B-equilib ≈ 10-
3.
5. Both internally and externally non-resonant modes are
observed.
6. Growth rates are dependent on current density and pressure profiles.
6Alfven Laboratory Mode Control Workshop, Austin 2003
Suitability of EXTRAP T2R for resistive wall mode active control studies (continued).
1. The surface where the saddle coils are installed is relatively accessable and well-defined.
2. Plasma current levels are low (<100 kA) so the power
requirements for the amplifiers for the active saddle coils is modest. Cheap loudspeaker amplifiers can
be used.
7Alfven Laboratory Mode Control Workshop, Austin 2003
Active mode control methods studied.Collaboration Alfven Lab, Consorzio RFX
and Chalmers theory group.
1. Intelligent shell - Alfven Lab taking the lead
2. Mode analysis - RFX taking the lead
One sensor coil
One PID controller to freeze flux at zero
One active saddle coil coinciding with sensor coil
Sensor coil array
Real time mode analysis
Voltage output to an array of active saddle coils
SIMO controller
8Alfven Laboratory Mode Control Workshop, Austin 2003
Relevance of resistive wall mode active control studies done on the T2R reversed-field pinch
1. The collaboration includes Anders Bondeson’s theory group at Chalmers and the RFX theory and
experiment groups. Emphasis is on comparison of theory and experiment.
2. There are features of feedback systems common for both the tokamak and the RFP i.e. The systems are based on fields produced by arrays of active external
coils interacting with plasma modes.
3. Role of field errors can be studied.
9Alfven Laboratory Mode Control Workshop, Austin 2003
2 4 6 8 10
time (ms)
red: n = -2 Theoretically stable.-exp /-theory = negative Small initial amplitude.
green: n = -8Theoretically unstable. -exp /-theory = 1.3 Large initial amplitude.
blue : n = -10 Theoretically unstable. -exp /-theory = 1.5 Small initial amplitude.
Observed Growth rates (w) for three RWMS.
-9
-10
-11
Ip
80
60
40
20
0
logebn
10Alfven Laboratory Mode Control Workshop, Austin 2003
2 4 6 8 10
time (ms)
n = -10(internally non res)
green: ”High ” equilibriumLower growth ratew= 1.4
blue :”Low ” equilibriumHigher growth ratew= 4.1
Observed Growth rates for n = -10 for two equilibria.
-9
-10
-11
Ip
80
60
40
20
0
logebn
11Alfven Laboratory Mode Control Workshop, Austin 2003
2 4 6 8 10
time (ms)
n = +5
(externally non res)
green: ”High ” equilibriumHigher growth ratew= 1.9
blue :”Low ”StableNo growth
Observed Growth rates for n = +5 for two equilibria.
-9
-10
-11
Ip
80
60
40
20
0
logebn
12Alfven Laboratory Mode Control Workshop, Austin 2003
n = -8• Unstable (Th & Exp)• Large ”initial”
amplitude• Mode phase is repro-
ducible in the lab frame
Observed phase of RWMs in fixed lab frame
n = -10• Unstable (Th & Exp)• Small ”initial”
amplitude• Mode phase is
random in the lab frame
• Slow rotation +
Five shots overlaid in each panel
2π8
13Alfven Laboratory Mode Control Workshop, Austin 2003
n = -2• Theor stable - Exper
unstable• Small ”initial”
amplitude• Mode phase is repro-
ducible in the lab frame
Observed phase of RWMs in fixed lab frame
n = +5• Unstable (Th & Exp)• Small ”initial”
amplitude• Mode phase varies in
the lab frame• Sometimes slow
rotation
Five shots overlaid in each panel
2π10
14Alfven Laboratory Mode Control Workshop, Austin 2003
Low
• Some shot-to-shot variation.
• Amplitudes higher than the high case below.
Raw data m = 1 B-radial perturbation (inboard - outboard) at 8 ms into discharge
Five shots overlaid in each panel
0 100˚ 200˚ 300˚ Toroidal angle
High
• No shot-to-shot variation (n = 8 domi-nated).
• Amplitudes lower than the low case above.
15Alfven Laboratory Mode Control Workshop, Austin 2003
Summary of new experimental observations concerning RWM instabilities
1. For many theoretically unstable modes, the experimentally observed growth rates are fairly well described by theory including a dependence on equilibrium profiles.
2. Some theoretically stable modes are observed to be unstable (i.e. n = -2).
3. Concerning the role of field errors:
• Modes that have a high initial amplitude during the transient discharge start-up (i.e. n = -8), have
a fixed phase in the lab frame.
• The theoretically stable n = -2 mode has a fixed phase in the lab frame.
16Alfven Laboratory Mode Control Workshop, Austin 2003
Feedback experiments underway on T2R1. Sensor coil array is in place in interspace between
vacuum vessel and shell.• 4 (poloidal) x 64 (toroidal)• saddle coils ”outboard-top-inboard-bottom”.
2. Active coils in place outside shell at eight toroidal positions.• coils are ”1/32” wide (i.e. double the width of
a sensor coil).• saddle coils ”outboard-top-inboard-bottom”.
3. ”m = 1” connected • Both sensor coils and saddle coils are series
connected (i.e. ”top & bottom” and ”inboard & outboard”).
4. Present active coil array covers 25% of surface.
17Alfven Laboratory Mode Control Workshop, Austin 2003
61 62 63 64 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16
110-degree toroidal sector of T2R
toroidal direction
Polo
idal d
irecti
on
The frame of reference consists of 64 toroidal sectors numbered 1 to 64
337.5˚ 0˚ 28˚ 90˚
top
inboard
outboard
bottom
B-r sensor coils 4(poloidal) x 64 (toroidal) positions
diagnostic port sector outer shell weld shell gaps
18Alfven Laboratory Mode Control Workshop, Austin 2003
61 62 63 64 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16
110-degree toroidal sector of T2R
toroidal direction
Polo
idal d
irecti
on
The saddle coils for active feedback are twice the width of the sensor coils
337.5˚ 0˚ 28˚ 90˚
top
inboard
outboard
bottom
B-r sensor coils 4 (poloidal) x 64 (toroidal) positions
19Alfven Laboratory Mode Control Workshop, Austin 2003
61 62 63 64 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16
110-degree toroidal sector of T2R
toroidal direction
Polo
idal d
irecti
on
337.5˚ 0˚ 28˚ 90˚
top
inboard
outboard
bottom
Cartoon of an n = 8 mode
20Alfven Laboratory Mode Control Workshop, Austin 2003
Theory for both intelligent shell control and mode control has been done. Assumptions for partial coverage feedback in the T2R RFP
1. Used T2R geometry and penetration times.But assume smooth resistive shell!
2. Assume B perturbation Fourier component information
corresponding to 4 x 32 sensor coils.
3. Use 4 x 8 active coils corresponding to actual coil geometry and partial coverage (side band
harmonics).
4. Consider only m = 1 nonresonant RW modes (i.e. zero
for resonant modes and higher m modes.
5. Examine both intelligent shell case and mode analysis/control case.
21Alfven Laboratory Mode Control Workshop, Austin 2003
Theory for partial coverage feedback in an RFP
References:
1. Feedback control of resistive wall modes in RFPsPaccagnella, Gregoratto and BondesonNuc Fusion 42 (2002) pg 1102
2. Output feedback with 4 x 32 sensors and 4 x 8 coilsGregoratto, Paccagnella, Liu and BondesonManuscript
22Alfven Laboratory Mode Control Workshop, Austin 2003
Features of the feedback theory
1. Assume Fourier component bn are known for themodes.
2. Eight active coil toroidal positions allows 8 control voltages Vn (n = -3,-2,-1,0,+1,+2,+3,+4)
3. Feedback law determines the 8 control voltages.
4. All the modes of interest are potentially ”controlled” (i.e. stabilised, destabilised, reduced growth rate, increased growth rate)
5. For intelligent shell case the gains in the feedback law
are equal and positive (i.e. negative feedback).
6. For mode control case gains in the feedback law are different and are optimised (can be positive
feedback).
23Alfven Laboratory Mode Control Workshop, Austin 2003
Block diagram for the control of a single RWM
RFX figure
24Alfven Laboratory Mode Control Workshop, Austin 2003
Experiments with anaog controlled intelligent shell have started
ControllerInput is m = 1 connected sensor coil pair
AmplifierOutput to m = 1 connected saddle coil pair
B-radial frozen at zero with feedback
B-radial grows without feedback
plasma current
B-radial
Active coil current
Vacuum vessel
Shell
25Alfven Laboratory Mode Control Workshop, Austin 2003
Intelligent shell experiments
Unfortunately not all the controllers were ready as of last week. The first experiments have been done with 12 analog controllers.
This means that 6 of the 8 toroidal positions can be controlled.
First test:
Intelligent shell with 6 toroidal positions active which is equivalent to about 18% coverage).
26Alfven Laboratory Mode Control Workshop, Austin 2003
Comparison of experiment with 6 toroidal positions and theory for 8 toroidal positions.
Summary of feedback theory for targeted moden = -8
Without feedback, the n = -8 mode is unstable and has a large initial amplitude.
The n = -8 mode should be stable for the intelligent shell controller at 8 toroidal positions.
The n = -8 should also be stable for the mode controller with active coils (1/32 wide) at 8 toroidal positions.
27Alfven Laboratory Mode Control Workshop, Austin 2003
Partial intelligent shell. 6 toroidal positions - inboard/outboard+top/bottom (18% coverage)Green is without feedbackBlue is with feedback
With FB
n = -8
The initial amplitude is lower.
The growth rate is not changed.
The phase is unchanged.
Without FB
phase
28Alfven Laboratory Mode Control Workshop, Austin 2003
n = -2 (impossible case)
The harmonics ”controlled” are n = -10, -2, +6, +14.
The n = -2 mode is theoretically stable but experimentally unstable.
The n = -10 and +6 modes are unstable both in theory and experiment. The n = +14 is stable in theory and experiment.
Feedback with partial coverage of 8 toroidal positions cannot stabilise all these modes.
For the intelligent shell controller, the n = -10 growth rate can be decreased but the n = -2 and n = +6 have higher growth rates and the n = 14 is destabilised!
The mode controller is better. However not all three unstable modes can be stabilised.
29Alfven Laboratory Mode Control Workshop, Austin 2003
Partial intelligent shell. 6 toroidal positions - inboard/outboard+top/bottom (18% coverage)Green is without feedbackBlue is with feedback
With FB n = -10
The initial amplitude is not changed.
The growth rate is not changed.
The phase is changed.
Without FB
phase
30Alfven Laboratory Mode Control Workshop, Austin 2003
Partial intelligent shell. 6 toroidal positions - inboard/outboard+top/bottom (18% coverage)Green is without feedbackBlue is with feedback
With FB
n = -2
The initial amplitude is slightly lower.
The growth rate is slightly increased (in agreeement with theory).
The phase is not changed.
Without FB
phase
31Alfven Laboratory Mode Control Workshop, Austin 2003
Partial intelligent shell. 6 toroidal positions - inboard/outboard+top/bottom (18% coverage)Green is without feedbackBlue is with feedback
With FB
n = +6
The initial amplitude is not changed.
The growth rate is slightly decreased (not in agreeement with theory).
The phase is not changed.
Without FB
phase
32Alfven Laboratory Mode Control Workshop, Austin 2003
Test with 8 toroidal positions, but only inboard/outboard saddle coils activated.
Unexpected result:
The n = +6 mode is stabilised.
The other modes are only slightly changed.
33Alfven Laboratory Mode Control Workshop, Austin 2003
Partial intelligent shell. 8 toroidal positions - inboard/outboard ( 12% coverage)Green is without feedbackBlue is with feedback
With FB
The n = +6 mode is stabilised.
b n = +6b n = -10
The n=-10 mode is slightly lower
Without FB
phase
34Alfven Laboratory Mode Control Workshop, Austin 2003
Partial intelligent shell. 8 toroidal positions - inboard/outboard ( 12% coverage)Green is without feedbackBlue is with feedback
With FB
The n = -8 mode is unchanged..
b n = -8 b n = -10
The n=+14 mode has a higher amplitude
Without FB
phase
35Alfven Laboratory Mode Control Workshop, Austin 2003
Partial intelligent shell. 6 toroidal positions - inboard/outboard+top/bottom (18% coverage)Green is without feedbackBlue is with feedback
With FB
b-radial pertur-bation late in pulse.
Intelligent shell controllers at 6 toroidal positions indicated by vertical dashed line.
Without FB
36Alfven Laboratory Mode Control Workshop, Austin 2003
Future experiments
With the present 8 toroidal position set up we will continue the studies and compare experiment with theory.
Both the intelligent shell controller and the RFX mode controller will be used (and compared with the analog IS).
For these studies destabilisation is as interesting as stabilisation since the goal is to verify that the theory models the relevant physics.
Study the field error effects.• initial amplitude.• destabilisation of a stable RWM.• phase.
Add more active coils. Next step is 50% coverage.
Use the flexibility of the RFX controller for mode rotation.