Analysis of Disruption Scenarios and

Their Possible Mitigation in ITER

M. Sugihara1, V. Lukash2, Y. Kawano3, R. Khayrutdinov4, N. Miki3, A. Mineev5, J. Ohmori3, H. Ohwaki6, D. Humphreys7,

A. Hyatt7, V. Riccardo8, D. Whyte9, V. Zhogolev2, P. Barabaschi10, Yu. Gribov1, K. Ioki10, M. Shimada1

1 ITER IT, Naka JWS, 2 Kurchatov Institute, 3 JAERI, 4 TRINITI, 5Efremov, 6 Keio Univ., 7 General Atomics, 8 EURATOM/UKAEA, 9 Univ. Wisconsin, 10 ITER IT, Garching JWS sugiham@itergps.naka.jaeri.go.jp

Purpose and outline Design Philosophy for disruptions in ITER - In-vessel components and vacuum vessel should be

designed to withstand the expected Electro-Mechanical (EM) load by disruptions

- At the same time, to minimize the number of disruption event and its impact on machine by mitigation technique is highly desirable

1. Analysis of Disruption Scenarios and EM Load (to check the robustness of the design)

- Database analysis and physics guidelines for current quench rate and halo current

- Simulation of representative disruption scenarios by DINA code and EM load analysis

- Trade-off between eddy and halo currents for EM load with respect to current quench rate

2. Disruption mitigation by massive and moderate noble gas injection

- Assessment of current quench rate and runaway electron generation for various gas species and optimum amount of gas in ITER.

- Optimization of response time, force on gas inlet valve and mitigation success / false rate based on neural network disruption prediction system.

Representative disruption scenarios • Fast current quench : VDE, Major Disruption (MD) • Slow current quench : VDE (down & upward) Note: VDE is a rare event

- failure of control - break down of control system - failure of mitigation system

Origin of most severe EM load on each component • Blanket & divertor : Eddy current + halo current

(MD & VDE with fast current quench) • Vacuum vessel : Halo current

(VDE with slow current quench)

Current quench rate

• Minimum Dt(100%)/S are from JT-60U

• Definition of Dt ; Dtmax(100%) ≡Ip0/(dIp/dt)max Dtmax(100%)/S ≈ 0.8(60%)/0.6 =1.2 ms/m2 fi Dtmax(100%)ITER ≈1.2 ms/m2¥ 21m2 ≈25 ms

• However, Dtmax(100%) ≈ 25 ms cannot be applied to ITER. It should be always;

Dt(100%) > Dtmax(100%)

From IPB [1]: Different definitions are mixed

0.1

1

1 0

1 0 0

1000

0.1 1 1 0

ASDEX-U/80-20%[dt/S]Tore-Supra/max[dt/S]JET/100-40%[dt/S]dt/S/80-20%[C-Mod]

JT-60U/max[dt/S]TFTR/90-10%[dt/S]DIII-D/90-10%[dt/S]

dt/S

[m

s/m

2, d

I p=6

0 %

]

Jp o [MA/m2 ]

Te=2eV

Te=5eV

Te=8eV

ITER9 MA 24 MA

< ne020>=0.72 m-3

• Good measure for quench time is to use average quench rate during (0.8-0.2)Ip0 (recommendation by ITPS);

Dt(100%) = Ip0 / <(dIp/dt)80fi20>

• Presently available data (Fig.) [1] with this definition except for

DIII-D : <(dIp/dt)90fi10> JET : <(dIp/dt)100fi40> (<(dIp/dt)80fi20> is smaller than <(dIp/dt)100fi40> due to existence of runaway electron)

0

1

2

3

4

5

0 1 2 3

ASDEX-UC-MODDIII-D (90=>10)Ip0

JET (100 => 40)Ip0JT-60U

Dt /

S (m

s / m

2 )

Ip0

(MA)

Dt=Ip0/<dIp/dt>80=>20 except for DIII-D and JET

• (Dt(100%)/S)min ≈ 1.8-2 ms/m2 ==> Dt(100%) ≈ 40 ms in ITER • Linear waveform as simple choice

• Many fast quench disruptions have exponential-like waveform

- Runaway electron is associated with fast quench [2-4]

Fast quench disruption

0

1

2

3

0

2 1016

4 1016

6 1016

7.035 7.04 7.045 7.05

Ip_ExperimentIp_Linear fitIp_Exponential fitNeutron

Neutron (s-1)

I p (MA)

Time (s)

RMSE=0.157

RMSE=0.015

JT-60U 31708

Slow quench disruption

0

0.5

1

1.5

2

2.5

0

1 1016

6.065 6.07 6.075 6.08

Ip_ExperimentIp_Linear fitIp_Exponential fitNeutron

Time (s)

I p (MA)

Neutron (s-1)

JT-60U 31525

RMSE=0.031

RMSE=0.096

• Time constant of exponential waveform consistent with the database

• Exponential waveform, which passes the 80% and 20% of Ip0 for the linear waveform of the fastest current quench disruption (Dt(100%) ≈ 40ms)

fi Exponential waveform with time constant of t ≈ 18 ms in ITER

0

2

4

6

8

10

12

14

16

0 10 20 30 40 50

Ip (M

A)

Time (ms)

25 ms

25 ms

exp(-t/18)

-(15/40)t + 15

• Both linear and exponential waveforms are examined • These linear and exponential waveforms have the same

Dt(80-20%) and can be reasonable initial choice of the waveforms.

• EM load on blanket are similar for both waveforms.

Force on key (only by Ieddy)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 10 20 30 40

#17_linear_40ms#18_linear_40ms#17_exponen_18ms#18_exponen_18ms

Time after TQ (ms)

Forc

e on

key

(MN)

Design target

Force on key by eddy currentDownward VDE with fast quench

Force on flexible joint

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 10 20 30 40

#1_linear_40ms#2_linear_40ms#1_exponen_18ms#2_exponen_18ms

Forc

e on

flex

ible

join

t (M

N)Time after TQ (ms)

Design target

Force on flexible joint by eddy currentDownward VDE with fast quench

• Global feature of EM load could be checked either by

linear or exponential, but can depend on scenarios and components.

Halo current

• Toroidal peaking factor; TPF

• Halo current fraction; Ih,max/Ip0 Local max. halo current

µTPF¥Ih,max

• TPF¥Ih,max/Ip0 < 0.7 for most of the machines [1,5,6]

• TPF¥Ih,max/Ip0 depends on current quench rate (JET) [7];

fi VDE with slow current quench has the largest TPF¥Ih,max/Ip0

0

1

2

3

4

5

0 0.1 0.2 0.3 0.4 0.5 0.6

C-MOD(94-95)C-MOD(96)COMPASS-D(IPB)COMPASS-D_ELMCOMPASS-D_VDEASDEX-UMASTDIII-DJET1999-JET2002+JT60_wo_puffJT60_w_puffTPF*Ih/Ip0=0.7TPF*Ih/Ip0=0.42

TPF

Ih,max

/ Ip0

Data from IPB [1] + new data from JET, MAST and JT-60U

Specification for ITER

For VDE with slow current quench - TPF¥Ih,max/Ip0≈0.7 - Ih,max/Ip0 ≈ 0.44 - TPF ≈ 1.6 For VDE with fast current quench - TPF ≈ 1.6 (same as slow) - Ih,max/Ip0 is to be evaluated by simulation code - TPF¥Ih,max/Ip0 < 0.7 (value depends on Ih,max/Ip0)

Disruption simulation by DINA code [8]

- 2D free boundary equilibrium calculation

- Transport and current diffusion in the plasma (1D averaged on flux surface) are solved

- Circuit equations for toroidal current in PF coils, vacuum vessel (modeled by series of plates) and blanket (modeled by boxes;right lower figure)

- Divertor is not modeled yet

- 8

- 6

- 4

- 2

0

2

4

6

8

0 2 4 6 8 1 0 1 2 1 4 1 6R (m)

Z (m

)

Physics guidelines for simulations

Representative scenarios Physics guidelines

Major Disruptions (MD)

Down/upward VDE with fast and slow Ip

quench

1. Current quench waveform and time (fast quench)

Linear 40ms and Exponential 18 ms

‹

2. Thermal quench (T.Q.) time duration

Beta drop : 1 ms [1] j flattening : ≈ 3 ms

‹

3. Surface q value at T.Q. 3 1.5 – 2 [9] 4. Beta drop during T.Q. ≈ 0.72 - 0.75 ≈ 0.75 - 0.4 5. Change of li during T.Q. 0.15 - 0.2 ‹ 6. fh≡(Ih,max/Ip0)¥TPF for VDE with slow current quench

0.7 for downward VDE with slow

quench

Calculation results

Downward VDE with fast quench

EM load on BM & divertor due to eddy and halo is expected most severe 0

5

10

15

20

- 4

- 3

- 2

- 1

0

660 680 700 720

Cur

rent

(M

A)

Time (ms)

Z (m)

I p

ZIhalopol

(a)

(b)

( c )

-5

-4

-3

-2

-1

0

1

3 4 5 6 7 8 9

Z (m

)

R (m)

(a)

(c)

(b)

VDE downward

1

2

3

4

17

15

16

18

Major Disruption

EM load on BM & div. due to eddy and halo is less severe than VDE but number of disruption is large

0

5

10

15

20

0

1

2

3

4

5

0 10 20 30 40 50 60

Curr

ent (

MA)

Time (ms)

Ihalopol

Ip

(a)

(b)

(c)

(d)

(e)

(f)(g)

Z (m)

Z

MD

0

500

200 400 600 800 1000

R[cm]

(a )

(b)

(c)

(d)

(e )

( f )

(g)

Z [c

m]

-400

M D

Support of blankets on vacuum vessel by

- Key (Fp) - Flexible joint (Fr)

Moments Mr, Mp, Mt are calculated by FEM (induced eddy current)

Force on each module Fp ‹ Mr + (Fp by halo) Fr ‹ Mp + Mt

Br

Bt

Bp

Mr(DBp)

Mp(DBr)

Mt (DBr)jr X Bt

jr

jtjp

jp X Bt

jt X Bp

jpjt

jr

(DBp)

(DBr)

(DBr)

Force on Key

- Force by eddy current is dominant but force by halo is also significant for the peak force - Nr. 1 and 18 BM are close to design target. There is some margin, but not so large.

0

0.2

0.4

0.6

0.8

1

1.2

0 10 20 30 40Time after TQ (ms)

Forc

e on

key

(MN)

Force on key of # 1 module

Design target

Total (fh=0.34)

by Eddy current

by Halo current

fh=TPF * (I

h/I

p0)

Total (fh=0.21)

fh=0.21

fh=0.34

40ms_linearDownward VDE with fast quench

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 10 20 30 40

Design target

Total

by Eddy current

by Halo current

Forc

e on

key

(MN)

Time after TQ (ms)

Force on key of # 18 module

Ihalo

(#18)=0.8*Ihalo,total

TPF = 1.6

40ms_linear

fh=TPF * (I

h/I

p0)=0.34

Downward VDE with fast quench

Force on Flexible joint

- Dominant force is by eddy and force by halo is small

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 10 20 30 40

#1_linear_40ms#2_linear_40ms#1_exponen_18ms#2_exponen_18ms

Forc

e on

flex

ible

join

t (M

N)

Time after TQ (ms)

Design target

Force on flexible joint by eddy currentDownward VDE with fast quench

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 10 20 30 40

#17_linear_40ms#18_linear_40ms#17_exponen_18ms#18_exponen_18ms

Design target

Force on flexible joint by eddy current

Forc

e on

flex

ible

join

t (M

N)Time after TQ (ms)

Downward VDE with fast quench

Force by eddy current can be reduced significantly with increase of current quench time by factor of 1.5-2

Force on key due to eddy : Linear 40 vs 80 ms (Downward VDE)

0

0.2

0.4

0.6

0.8

1

1.2

0 20 40 60 80

#1_linear_40ms

#1_linear_80ms#2_linear_80ms

Time after TQ (ms)

Forc

e on

key

(MN)

Force on key by eddy current

Design target

#2_linear_40ms

Downward VDE with fast quench

Reduction of eddy current Is significant but increase of halo current is very small

0

1

2

3

4

5

6

7

0

0.5

1

1.5

2

10 100 1000

Force (a.u.)

Current quench time (ms)

I h,m

ax (M

A)

Ih,max

Force on BM byeddy + halo

Force on VV by halo

VDE_DW

• This feature can be a good basis for the optimum current

quench time for disruption mitigation

Possible disruption mitigation in ITER -Massive [10] or moderate [11] noble gas injection • Choice of Neon as an optimum injection gas species

• tL/R is evaluated by the coupled time dependent equations [12]: - Impurity rate eq. - Plasma power balance eq. (radiation & joule power) - Plasma circuit eq. - Avalanche & Dreicer R.E. eq.

• Current quench time can be longer by factor of 1.5-2 than that of unmitigated disruption (18ms L/R time) or argon case for nNeon £ (3-5)¥1021 m-3

• R.E. is not generated for nNeon ≥(0.5-1) ¥1021 m-3

0

20

40

60

80

0.1 1 10 100

No runawayRunaway

Neon

Carbon

Argon

nz (1020 m-3)

t L/R (m

s)

ITERL / R time

Species: Neon Amount:nNeon≈(1-2)¥1021 m-3

Response time Dt , Force on gas inlet valve & Success rate - Assumption : neutral gas pressure

†

Pn ª P p ≡ 0.29 MPa for penetration to plasma center

- Required time Dt for neutral gas pressure to reach required value

†

P p can be evaluated by solving the following flow eqs. (gas flow is critical at the valve: Mach=1)

†

Vn

kT0

dPn

dt=

Am

k ( 2k + 1

)k +1k -1 P0r0 (for Pn<P*)

†

Vn

kT0

dPn

dt=

Am

2kk - 1

P0r0( Pn

P0)

2k (1- ( Pn

P0)

k -1k )

(for Pn > P*)

0 0.2 0.4 0.6 0.8 1Mas

s flo

w ra

te Q

m /

A (k

g/m

2 s)P / P

0

Critical flow

Critical pressure P*/P0=0.528

M=1

Solution

- Dt decreases with increase of reservoir gas pressure P0 fi quick response is achieved by increasing reservoir gas pressure P0

- On the other hand, force on the gas inlet valve

†

F0 = P0 A increases to achieve faster response (decreasing Dt )

0

2 105

4 105

6 105

8 105

0 5 10 15 20Pr

essu

re (P

a)Time (ms)

Required pressurefor penetration

P0 = 0.35 MPa

0.4

0.5

0.6

0.81.01.4

Dt for 0.35 MPa

Success rate of mitigation can be increased with decreasing Dt (with increasing force on gas inlet valve F0 )

- Employ neural network system by Yoshino as example [13]

In the case of massive injection

With increasing Dt, missed rate increases gradually, but for Dt≈15ms, missed rate can be still very low (3-4%) for LM, DL, high li (Ip ramp-down) disruptions. Force on the gas inlet valve reduces significantly (≈ 300kg; needs design study).

0

0.02

0.04

0.06

0.08

0.1

0

0.2

0.4

0.6

0.8

1

0 5 10 15 20

Mis

sed

Rate

Force on gas inlet valve (ton)

Dt (ms)

Force

Neutral gas column10 cmf, L=4 m

Ip ramp-down

Density limitLocked mode

In the case of mild gas puff

Response is very slow, i.e., thermal quench occurs ≥ 100 ms after puff valve open;

Bakhtiari, NF 42 (2002) 1197

0

0.2

0.4

0.6

0.8

1

0 20 40 60 80 100 120

Density limitIp ramp-downLocked mode

Mis

sed

Rate

Dt (ms)

beta limit

More than 40% of disruption will be missed, while force on the gas inlet valve is not an issue.

ITER value for massive and mild injection

Massive injection

Mild injection

Dt (ms) 15 100 Missed rate (%) 3-4 40 False rate (%) 2 2 Force on gas

inlet valve (kg) 300 Not an issue

• In this system, by optimizing the alarm level, the false rate can be reduced significantly;

with Alarm level: 0.98 => False rate ≈ 2 %

• Increasing alarm level can further reduce the missed rate but the false rate significantly increases.

Massive injection

Mild injection

Success shots 970-960 600 (≈ 700) Missed shots 30-40 400 (≈ 300)

Total shots: 104

Disruptive shots : 103

False shots 200 200 (≥ 500) ( ) ; rough estimation for alarm level=0.99

•Key point to enhance the effectiveness of mitigation: Increase of success rate for disruptions close to beta limit and ITB [13,14].

- Success rate cannot be increased without increasing false rate (algorithm by DIII-D)

DIII-D; Wroblewski et al., NF 37 (1997) 725

Conclusions

• Physics guideline for the current quench time (Dt=40 ms linear , time constant of t=18 ms exponential waveform) and halo current (TPF¥Ih,max/Ip0 < 0.7) are derived from disruption database.

• Representative disruption scenarios are analyzed by the

DINA code and EM load on the blanket modules and vacuum vessel are analyzed by FEM code. There is some margin, but the margin is not so large, which indicates the importance of mitigation.

• Increasing current quench time by a factor of (1.5-2) can

decrease the EM load due to eddy currents significantly, at the expense of small increase of the EM load due to halo currents.

• Such mitigation can be achieved either by massively or mildly injecting (1-2)x1021 m-3 of Neon gas without runaway electron generation.

• Success rate of mitigation can be increased with

increasing the pressure of gas reservoir (decreasing the response time) at the expense of increasing the force on gas inlet valve.

• Coupled with a neural network disruption prediction system, it is found that the success rate can be >95% for massive injection with moderate force of gas inlet valve (≈ 300 kg). However, only ≈ 60% of disruptions can be mitigated successfully for mild injection method due to its longer response time.

References [1] ITER Physics Basis, Nucl. Fusion 39 (1999) 2137. [2] Sugihara, M., et al., J. Plasma Fusion Science 79 (2003) 706. [3] Riccardo, V., et al., submitted to Plasma Phys. Control. Fusion. [4] Sugihara, M., et al., 30th EPS, St. Petersburg (2003), P-2.139. [5] Riccardo, V., Plasma Phys. Control. Fusion 45 (2003) A269. [6] Neyatani, Y., et al Nucl. Fusion 39 (1999) 559. [7] Riccardo, V., Plasma Phys. Control. Fusion 46 (2004) 925. [8] Khayrutdinov, R.R. et al, J. Comp. Physics 109 (1993) 193. [9] Sugihara, M., et al., Plasma Phys. Control. Fusion 46 (2004) 1581. [10] Whyte, D., J. Nucl. Material. 313-316 (2003) 1239. [11] Bakhtiari, M., et al., Nucl. Fusion 42 (2002) 1197. [12] Mineev, A., et al., ITER Task Report (to be published). [13] Yoshino, R., Nucl. Fusion 43 (2003) 1771. [14] Wroblewski, D., et al., Nucl. Fusion 37 (1997) 725.