Protection of the Hi Lumi CCT Corrector Magnets 1 M Mentink J van Nugteren G Kirby March 8 th 2017 Motivation Quench protection of Hi Lumi CCT corrector coils 2 versions 05 meter ID: 541212
Download Presentation The PPT/PDF document "Quench" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Quench Protection of the Hi-Lumi CCT Corrector Magnets
1
M. Mentink, J. van
Nugteren
, G. Kirby
March 8
th
2017Slide2
Motivation
Quench protection of Hi-
Lumi CCT corrector coils2 versions: 0.5 meter model magnet and 2.2 meter prototype magnetQuench behavior is intrinsically 3D, both longitudinal quench propagation and transverse heat exchange play a roleQuench behavior determines appropriate level of extraction, which affects insulation requirements 3D modeling needed to understand quench behavior
2Slide3
Previous work by
Juho
RustiPrevious evaluation of quench behavior by Juho Rysti [1]3d thermal model in Comsol (FEM solver)Heat balance equation solved for a model with a few turns, and resulting quench velocities are applied to entire magnet geometryResults:Short magnet is intrinsically safe: No extraction Thotspot = 190 KLong magnet requires extraction: R
dump
= 1.4
Ω
, tswitchdelay = 12 ms Thotspot = 190 K
3
1. Kirby et al., IEEE
Trans. On Appl.
Supercond
. 27, p.
4002805
(2017)Slide4
Overview
New 3d model: modeling assumptions
Thermal heat transferMaterial propertiesQuench BackModeling resultsComparison with Juho’s resultsGeneral overview of temperature and voltage resultsQuench Protection parameter variationsSummary
4Slide5
3d model assumptions: Thermal transfer
Matlab
model with heat transfer in all three directionsTransverse heat transfer within turn:1 node per domain, insulation and strands are separate domains with separate temperature-dependent properties105 transverse nodes = 20 strand nodes + 4 x 20 strand insulation nodes + 3 former element nodes + 2 former insulation nodesEffective thermal thickness of E-glass strand insulation: 180 µm, calculated with ComsolLongitudinal heat transfer (from perspective of strand): 50 elements per turn, longitudinal quench propagation velocity follows directly from heat balance equationTurn-to-turn heat transfer (includes strand insulation to strand insulation)Helium cooling: T < 32 K: 1 W/cm2, T > 32 K: film boiling regimeNumerical aspectsBrute force solving (no adaptive mesh)Typical run-time: 40 mins for 320k nodes, scales linearlyInvestigation of longitudinal node density vs. Thotspot indicates convergence
5
4mm
15 mm
0.2 mm
polyimide
6.6 mm
0.2 mm
polyimide
6.6 mm
LHe
, 1.9 K
LHe
, 1.9 K
Transverse heat transfer Slide6
Material properties
Material properties
Formers: Aluminum-6000 series, RRR = 2.9Former insulation: Polyimide (0.2 mm thick)Strand insulation: Fiberglass (E-glass) + epoxy resin (Cv = Cv,G10, k = kG10 x 1.6 [2]) NbTi strand type: MB outer layer strand, D = 0.825 mm, Cu : nonCu = 1.95, in-field RRR = 62Magnetic field per group of strands varies between 3 and 1.8 T per turn when current is at 435 AIc(3T, 1.9 K) = 1130 A Operating current of 435 A is at 38% of short sample, and at 55% of load lineCurrent sharing temperature:
T
cs
= 5.7 K,
3.8 K marginBoth temperature and field dependence of NbTi are considered in quench protection calculation
6
2.
Imbasciati
et al., FNL TD-03-019, FNL / Univ. of Milan / INFN
(2014)
3. Data provided by A. Verweij
, also see IEEE Trans. on Appl. Supercond. 16,p
.
1192 (2006)
NbTi
MB
outer layer strandSlide7
Quench back
Electrical and inductive calculations (simultaneously with thermal calculations)
Approach: Numerical solution (where M and R are matrices and I is a vector):Quench back: Assumes homogeneous current flow through former cylinders, inductances based on assumed saddle coil current flow Likely somewhat pessimistic with regards to quench backInductances are corrected for presence of iron, but linear behavior assumed (i.e. inductances are independent of current)
7
Quench back current flowSlide8
Comparison with
Juho’s
resultsComparison of resultsWith quench back suppressed, new model is slightly more pessimistic (190-200 K) vs. 220 KDifferent underlying assumptions (bronze former vs. aluminum former, 5 strands per slot vs 10 strands per slot, cable vs. individual strands, etc.) may explain this 20-30 K differenceQuench back makes a big difference: 220 K without QB, 80 K with QB!
8
R
dump
= 1.4
Ω
,
t
delay = 12 msJ. Rysti’s resultsResults of the new modelSlide9
Typical temperature and voltage distributions
9
Typical temperature and voltage
distributions along the strands
Quench originates in middle of strand bundle on inner former, in middle of magnet
Calculation result: Inhomogeneous quench back
First strands on outer former quench (due to lack of direct helium cooling), followed by strands on inner former
Strands on inner former
Strands on outer former
Position along length of strand
x
[m]
Temperature [K]
Formers
Thick outer cylinder
0
100
8 12 16 20 24 28
50
Position along length of strand
x
[m]
Voltages [V]
Peak imbalance = Peak voltage between neighboring strands ≈150 V
-350
0
0 220 440
2.2 meter long magnet,
R
dump
= 1.0
Ω
,
t
switch,delay
= 25
msSlide10
Overview of results
10
Overview of results for “standard” quench protection setup
“
standard”
quench protection setup:
1.0
Ω dump resistor 25 ms
delay between quench origination and opening of switch
With aluminum formers, quench back, helium cooling
Results:
Peak Thotspot = 140 KPeak voltage to ground: 435 V (= 1.0 Ω x 435 A)Peak voltage between neighboring strands: 150 VOpening of relay after 25 msStrands on outer former normal due to quench back
Normal zone formation on outer former
Normal zone formation on inner former
2.2 meter long magnet,
R
dump
= 1.0
Ω
,
t
switch,delay
= 25
ms
Opening of relay after 25
msSlide11
Quench Protection parameter variations
11
Quench protection parameter variations
Higher dump resistor, faster switch opening
lower
T
hotspotQuench back much lower Thotspot
Operating current increases
T
hotspot
increases
Iop = 435 ARdump = 1.0 Ω, tswitchdelay = 25 msSlide12
Summary
12
New three-dimensional model
Three dimensions: Longitudinal (from strand perspective) and both transverse directions
Considers thermal exchange between strands, formers, neighboring turns, neighboring formers, etc.
Quench back included (in somewhat rudimentary manner, likely somewhat pessimistic)
ObservationsComparison to Rysti’s Comsol model: Generally consistent results if quench back is suppressed
Quench back strongly reduces the peak hotspot temperature (typical reduction: more than 100
K
)
Peak voltage to ground dominated by dump resistor (
Vground,max = Iop x Rdump)As rule of thumb, peak voltage between neighboring strands typically equal to 30 to 40% of peak voltage to groundResults & variationsRdump = 1.0 Ω, tswitchdelay = 25 ms Thotspot = 140 K with quench back, 270 K without quench backHotspot temperature decreases with increasing Rdump
and decreasing tswitchdelay
.
For instance:
R
dump
= 1.4
Ω
,
t
switchdelay
=
12
ms
T
hotspot
= 80 K (with quench back)
Hotspot temperature increases with operating current For instance: Rdump = 1.0
Ω, tswitch,delay = 25 ms,
Iop = 735 A Thotspot
= 280 K (with quench back)