Coupled Quench + Circuit modeling for the High Luminosity LHC upgrade with the

STEAM co-simulation framework

1

Matthias Mentink1, Emmanuele Ravaioli2, Samer Yammine1, Lorenzo Bortot1, Alejandro Navarro1, Marco Prioli1, Michal Maciejewski1, Arjan Verweij1

1. CERN, Meyrin, Switzerland

2. Lawrence Berkeley National Laboratory, Berkeley, CA, USA

21/09/2017

Motivation

2

High Luminosity Large Hadron Collider Upgrade

• 1 BCHF upgrade of LHC, to be completed in 2025

• Objective: To increase luminosity of LHC, i.e. produce more focused beams leading to more collisions

• HL-LHC triplet circuit: 6 Nb3Sn quadrupoles powered by four nested power supplies, that focus /defocus the particle beam

• Quench protection: CLIQ (Coupling-loss induced quench) + quench heaters

• Quench behavior: Very complex circuit comprising six interacting magnets and multiple power supplies Requires consideration of all six quenching magnets at once Co-simulation

18 kA PS

Magnets: Q1a+Q1b Q2a Q2b Q3a+Q3b

±2 kA PS ±2 kA PS

±35 A PS

HL-LHC triplet circuit Nb3Sn quadrupole

B [T]

STEAM co-simulation: Wave-form relaxation method

3

• STEAM calculations of HL-LHC triplet circuit combines Pspice (=circuit solver) with LEDET (numerical tool for calculating quench behavior of a magnet [1])

• Implements wave-form relaxation method

• For a given time window, Pspice provides initial guess of I(t)

• Each of six LEDET models calculate resulting voltages V(t)

• Then Pspice calculates revised I(t) and the LEDET models calculate revised V(t)

• Keep iterating until convergence is achieved and then move on to next time window

For a given time window t = tstart – tend:

LEDET1LEDET2LEDET3LEDET4LEDET5LEDET6

Pspice I(t)0 V(t)1 Pspice

LEDET1LEDET2LEDET3LEDET4LEDET5LEDET6

I(t)1 V(t)2 Pspice

Keep iterating until max(I(t)i+1-I(t)i) falls below convergence threshold And then move on to next time window

I(t)2 …

[1] Ravaioli et al. – Cryog. 80, p 346 (2016)

Scenario #1: Localized quench, all magnets operating at 17.8 kA

4

• Starting conditions: 17.8 kA in all magnets

• Quench detected and validated after 15 ms

• Quench protection: CLIQ + quench heaters

• Maximum MIITs: 31.7 Tmax = 280 K

17.8 kA

Magnets: Q1a+Q1b Q2a Q2b Q3a+Q3b

0 A

35 A

HL-LHC triplet circuit, initial conditions

0 A

After quench detection + validation (15ms):Capacitive discharge over poles of magnets (CLIQ)

+600 V -600V +1000V +1000V +600 V -600 V

Calculation result

𝑀𝐼𝐼𝑇𝑠 = 10−6න0

∞

𝐼 𝑡 2𝛿𝑡

Scenario #2: Global quench (Q2b)

5

• Starting conditions: 15.8 kA in Q1a, Q1a, Q1a, Q3a, Q3b, 17.8 kA in Q2a, Q2b, global quench in Q2b

• Quench detected and validated after 10 ms

• Quench protection: CLIQ + quench heaters

• Maximum MIITs (Q2a): 32.4

17.8 kA

Magnets: Q1a+Q1b Q2a Q2b Q3a+Q3b

-2 kA

35 A

HL-LHC triplet circuit, initial conditions

-2kA A

Calculation result

𝑀𝐼𝐼𝑇𝑠 = 10−6න0

∞

𝐼 𝑡 2𝛿𝑡

Global quench in Q2b inhomogeneous current distribution between magnets

6 kA

Other HL-LHC simulations with STEAM 1/2

6

• Bortot et al., “STEAM: A Hierarchical Co-Simulation Framework for Superconduting Accelerator Magnet Circuits”, presented at Magnet Technology 25 conference, Amsterdam (29 Aug., 2017)

• Combines multiple models in a single circuit (using Pspice, Comsol 1D, Comsol 2D, LEDET)

• Initial quench development (Comsol1D model) drives circuit switches

Simplified HL-LHC triplet circuit

Comsol representation of magnetic fields, during quench

Other HL-LHC simulations with STEAM 2/2

7

• Navarro et al., “Simulation of a Quench Event in the Upgraded High-Luminosity LHC Main Dipole Circuit Including the 11 T Nb3Sn Dipole Magnets”, presented at EUCAS conference, Geneva (21 Sept., 2017)

• LHC main dipole circuit modification: Installation of more powerful 11 T dipoles, to make space for new collimators

• Simulation of the modified main dipole circuit of the LHC, combining Pspice and four LEDET models

• Complexity: More than one power convertor (somewhat similar to the triplet circuit)

• Results:

• Maximum current in the trim circuit in case of a quench: 180 A Acceptable

• Peak hotspot temperature of 326 K in the 11T dipole AcceptableComsol representation of magnetic fields, during quench

Conclusions

8

• High Luminosity Large Hadron Collider: Features complex circuits combining multiple power supplies and multiple magnets into single circuits

• Co-simulation needed to understand complex interaction between quenching magnets

• STEAM co-simulation framework has reached sufficient maturity to meet co-simulation needs

• Specific circuits currently under investigation:

• HL-LHC triplet circuit

• HL-LHC modified main dipole circuit

Magnet quench protection of the FCC-hh16 T block-type dipole magnet by means

of Quench Absorption Coils

9

Matthias Mentink1, Lorenzo Bortot1, Marco Prioli1, Michal Maciejewski1, Zhao Junjie2, Tiina Salmi2

1. CERN, 2. Tampere University of Technology

21/09/2017

EuroCirCol 16 T single aperture block coil option [1]

Motivation

10

16 T dipole for FCC-hh (Shown here: Block coil option)

• High magnetic field: 16 T, Umagnetic scales as B2 --> High stored energy

• Superconducting coil preferably compact, to efficiently use conductor --> Limited conductor volume

• High stored energy + limited conductor volume --> High energy density --> Rather high voltages to ground (>1000 V) and hotspot temperatures (>300 K) during a quench

• During a quench, how can peak voltage to ground and peak hotspot temperature be kept within reasonable limits?

[1] Lorin et al. – IEEE Trans. On Appl Supercond. 27, p. 4001405 (2017)

Possible solutions

11

How to quickly induce a homogeneous quench

• Fast detection and validation

• Quench heater optimization [6]

• Coupling Loss Induced Quench [2, 3]

How to subsequently limit Tmax and Vgnd,max

• Extraction (For instance E3presso [4])

• Inductive transfer of stored energy to external circuit

• Quench absorption coils [5]

• ICED concept [6]

[4] Nugteren - CERN internal note: 2016-23 (2016).[5] Mentink, Salmi – SuST 30, 064002 (2017) [6] Murtomaki et al. – under review (IEEE TAS 2017)

Quench absorption coils [2]

[1] Salmi et al. – IEEE TAS 27 (2017)[2] Ravaioli – PhD thesis (2015)[3] Prioli et al. – presented at FCC week 2017

Coupling Loss Induced Quench (CLIQ) system [2,3]

Outline

12

• Which computational model?

• Where to place Quench Absorption coils?

• CLIQ unit discharged over superconducting coil

• No Quench Absorption coil

• Quench Absorption coil on top / bottom

• Quench Absorption coil on sides

• CLIQ unit discharged over quench absorption coils

• Summary

Which computational model?

13

[7] Ravaioli et al. – Cryog. 80, p 346 (2016)[8] Bortot et al. – Currently under review (IEEE TAM 2017)

• Studied design: EuroCirCol 16 T Block coil, version v20ar

• Available computational models:

• Ledet [7]: numerical, very efficient

• Comsol [8]: FEM-based, real-time magnetic, electric, and thermal calculation

• Comparison: Consistent in terms of current, comparable in terms of Vgnd

• Model used for QA studies: Comsol model

EuroCirCol 16 T single aperture block coil, v20ar

-1500

1500

Vgn

d[V

]

Comsol, Vgnd,max

Ledet, Vgnd,turns

0 0.5

2000

-2000

Vgn

d[V

]

Where to place Quench Absorption coils?

14

QA coils on top / bottom

QA coils on the sides

Schematic representation (not true to scale)

QA coils on top / bottom:

+ Advantage: Easier to

implement from a design perspective

X Disadvantages: Lower coupling factor (~60%), less homogeneous magnetic field distribution over superconducting coil

QA coils on side:

+ Advantages: Higher

coupling factor (~80%), field more homogeneously distributed over superconducting coil

X Disadvantages: Harder to implement from design perspective, secondary coils have to sustain large stresses

CLIQ unit discharged over superconducting coils (1/2)

15

• Magnet initially powered at 12040 A (105% nominal current)

• During regular operation, blocking diodes prevent induced current in QA coils

• CLIQ quenches the magnet --> High dI/dt, diode threshold voltage exceeded --> Inductive transfer of stored energy from primary circuit to secondary circuit

• Presence of quench absorption coils --> Significant reduction in MIITs

QA coils Blocking diodesPower supply,

no energy extraction

CLIQ unit: 40 mF, V0 = 1200 V

CLIQ lead resistance: 20 mΩ

Supercond. coils

M

No QA coils Top/bottom QA coils Side QA coils

CLIQ unit discharged over superconducting coils (2/2)

16

• Assumed detection + validation + CLIQ discharge time: 21 ms

• QA coils on top / bottom: Inductive coupling of ramping QA coil is concentrated on a single layer --> Increased Vgnd,max

• QA coils on side: Vgnd,max decreased by nearly a factor two

• Presence of QA coils: Hotspot temperature reduction by up to 100 K

Scenario Volumeratio,

primary to QA

UQA,disp

/ Ustored

[%]Peak

Vgnd [V]

MIITs [MA2s],

with detection

Thotspot, adiab.

[K]

No QA coils ∞ 0 1470 17.2 370

QA coils on top / bottom

3.5 21 1560 15.4 300

QA coils on side

3.9 33 790 14.5 270

QA coils on sideQA coils on top/bottom

CLIQ unit discharged over QA coils (1/2)

17

CLIQ discharge over QA coils (similar to [2,9-12])

• CLIQ unit electrically insulated from primary circuit

• CLIQ discharge --> Net flux per half turn equal to zero --> No current oscillations in primary, negligible impact on Vgnd

QA coils Blocking diodes

Power supply, no energy extraction

CLIQ lead resistance: 20 mΩ

Superconducting coils

M

CLIQ unit: 40 mF, V0 = 1200 V

Initial CLIQ current discharge: No net inductive voltage on

sup. coilsEventual current discharge CLIQ discharge over QA coils

[9] Agustsson et al. - Proc. PAC2013[10] Bromberg et al. – PSFC/ JA-11-26 (2011)

[11] Bromberg et al. – IEEE TAS 22 (2012)[12] Bromberg et al. – US 7701677 (2006)

CLIQ unit discharged over QA coils (2/2)

18

External CLIQ:

• Slightly less efficient in inducing normal zone for same CLIQ parameters --> Hotspot temperature 10 K higher

• But also: Homogeneous distribution of inter-filament losses over superconducting coil--> Lowest peak voltage to ground

Scenario with side QA coils

Volume ratio, superconducting coils to QA coils

UQA,disp / Ustored [%]

Peak Vgnd [V] MIITs [MA2s] with

detection

Thotspot, adiab. [K]

CLIQ discharge into superconducting coil

3.9 33 790 14.5 270

CLIQ discharge into QA coils

3.9 33 730 14.6 280

CLIQ discharge over QA coils

CLIQ discharge over superconducting coils

Summary

19

‘Quench Absorption Coil’ concept for 16 T accelerator dipole

• Disadvantages / uncertainties

• More complexity, with extra copper coils and diodes

• Mechanical implications currently not yet understood

• Advantages

• Significant enhancement in quench behavior, without increasing size of superconducting coils

• 100 K lower hotspot temperature, 50% lower peak voltage to ground

• Allows for external CLIQ discharge

• CLIQ device electrically insulated from primary circuit

• No current oscillations and no CLIQ-induced voltage over the primary circuit

• About as efficient as CLIQ discharge in superconducting coils