During this period of frenzied investment in the 70’s, two major directions for fusion research emerged: the torus or donut shaped “Tokamak” design utilized by Princeton Plasma Physics Laboratory, and MFTF-B's “magnetic mirror” based approach, with a linear vessel housing the superheated plasma bouncing the plasma off two opposing magnetic “mirrors” at either end of the chamber.

🖼️🛳️ NEW! Prints for sale: One year of ship movements

Photos: Lawrence Livermore National Laboratory.

On Friday, February 21, 1986 a group of 300 scientists, engineers, contractors and government officials gathered for a dedication ceremony at Lawrence Livermore National Laboratory. After final diagnostic tests, the "Mirror Fusion Test Facility-B" (MFTF-B) completion was celebrated, and a letter from John Herrington, Ronald Regan's Secretary of Energy was presented to program director T. Kenneth Fowler extending his congratulations on a job well done.

On the very same day after nearly a decade of development and nearly a billion dollars\* of funding, the project was shut down, the massive machine having never been turned on.

"I want all of you to know how much I regret the fact that, just as you complete this remarkable new facility, the budget pressures dictate that we must put it into standby and not operate it as you might have hoped. This is frustrating, and perhaps not the best use of our national talent and resources, but we must bring the deficit under control," wrote Herrington in the letter to Fowler.

Photos of the 400-ton "yin-yang" magnet mirrors used in the MFTF-B. Source: Lawrence Livermore National Laboratory. Photo 1, Photo 2.

I came across photos from the construction of the components of the MFTF-B on Lawrence Livermore National Laboratory’s website and was captivated by a photo from 1980 showing a strange twisting mass of metal that at the time was the largest superconducting magnet in the world.

This 350 ton magnet was encased in stainless steel built in a distinct “yin-yang” shape, and its interior was cooled to temperatures of 425º F below zero, by pumping liquid helium through the vessel. Thirty miles of copper and niobium-titanium wire were wound over the course of a year to make the magnet’s conductor. The magnet was capable of generating magnetic fields 150,000 times that of Earth’s that could contain the 500-million Kelvin degree plasma generated by the fusion device.

A photo of the MFTF-B under construction in 1978. Source: LLNL

The Race for Fusion

“According to Greek mythology, fusion energy, the fire of the Sun is a gift hard won”, goes the first line of the T. Kenneth Fowler’s 1997 book, “ The Fusion Quest”. Fowler continued, “Today, irresistibly drawn to the challenge of bringing fusion energy down to Earth from the stars, scientists tempt Zeus still.” The promise of mastering this elemental force is a clean, safe and near limitless energy source, which could literally save the planet.

While the more widely known process of nuclear fission has been producing energy at commercial scale in power plants since the 1950’s, harnessing nuclear fusion – which Fowler’s poetic description accurately captures as the same reaction taking place in the heart of the Sun – has been the subject of an international race for decades.

Only in December of 2022 did scientists at the National Ignition Facility at the Lawrence Livermore National Laboratory announce that they had achieved the first recorded fusion reaction with a net energy gain – meaning that it released more energy than was put in, which was a crucial milestone for the whole field, though it could be decades before this is put into practical use.

The energy crisis of the 1970s motivated the U.S. government to throw lots of money at alternative energy sources, and fusion was one of the big areas of interest.

According to the Department of Energy, the basic principle of nuclear fusion is the fusion of two lighter nuclei (such as the commonly used combo of deuterium and tritium) to form a heavier one (helium) which releases energy (and subatomic particles, such as neutrons). To do this, super-hot plasma is created in a vacuum to create the fusion reaction, and either lasers or powerful magnets are used to control and contain the plasma.

During this period of frenzied investment in the 70’s, two major directions for fusion research emerged: the torus or donut shaped “Tokamak” design utilized by Princeton Plasma Physics Laboratory, and MFTF-B's “magnetic mirror” based approach, with a linear vessel housing the superheated plasma bouncing the plasma off two opposing magnetic “mirrors” at either end of the chamber.

1975 photo of the Princeton Large Torus at the Princeton Plasma Physics Laboratory, which is an example of a "tokamak" fusion reactor design. Source: Princeton Plasma Physics Laboratory, Public domain, via Wikimedia Commons

Research at Lawrence Livermore National Laboratory developing the mirror based approach showed promise in smaller scale tests, which led to the decision to go all in on a large scale device. The fact that other major labs were going all in on the tokamak approach, left an opening for hedging the big fusion bet on this potential alternate path. The question of whether there was enough sound evidence to ramp up to the scale of the MFTF-B was subject to debate at the time, and the final decision seemed to partly come down to ideology and gut instinct.

In a really thorough 1987 Science magazine story on the MFTF-B Edwin Kintner, who was the associate director of the Department of Energy’s Office of Energy Research at the time said as much. Looking back on the decision to go all in on the MFTF-B, Kintner is quoted as saying “Everybody was concentrating on tokamaks. I thought it was necessary for these tokamak guys to have to look over their shoulders.”

In the same article, MFTF-B program director Fowler is quoted as saying, “You could debate the decision, but it wasn’t illogical. Building big machines is a mixture of lead times, resources, prudence and gambling.”

_Below: "Mirror Fusion Test Facility magnet system— Final design report. Sept. 3, 1980. Source: Lawrence Livermore National Laboratory."_

6653255 (1) \| DocumentCloud

V1 )0-2243

QrPI t-RI.-52955

Mirror FusionTest Facility

magnet system—

Finaldesign report

C. D. Henning

A. J.Hodges

J. H. VanSant

E. N. Dalder

R. E. Hinkle

J. A. Horvath

R. M. Scanian

D. W. Shimer

R. W. Baldi

R. E. Tatro

September 3, 1980

.lil«Mr;:i:

\\* '«

UCRL-52955

DistributionCategoryUC-20

Mirror Fusion Test Facility

magnet system—

Final design report

C. D. Henning

A. J. Hodges

J. H. VanSant

E. N. Dalder

R. E. Hinkle

J. A. Horvath

R. M. Scanlan

D. W. Shimer

R. W. Baldi

R. E. Tatro

Manuscript date: September 3,1980

LAWRENCE LIVERMORELABORATORY^

University of California\* Livermore, California • 9 4 5 5 0 ^ ^

Availablefrom:NationalTechnicalInformationService •U.S. Departmentof Commerce

5285 PortRoyal Road•Spnncficld.VA22161 •$10.00 per cop\ •(MicroficheS3.50 J

mSTHWUTKHI OF THIS MCUKEKTISLS'ILIMITEB

ACKNOWLEDGMENTS

The authors wish to acknowledge the many contributions of all those who worked on the magnet

design. Special credit is given to the following persons:

Andersen, A.

Berfcey,J.

Bulmer, R.

Chang, Y.

Cornish. D. N.

Depue, D.

Johnson. G. L.

Karpenko. V. N.

Ko/.man. T. A.

Lathrop, G.

Lietske, J.

MacDonald. J.

Martinek, R.

Mon(05'a. C.

Peterson. R.

Podesta. D.

Ross. W. N.

Shook, W. R.

Thomasscn. K. I.

Wilhcrell, C.

Zimmerman, B.

Jones. R. G.—General Dynamics, Cnnvair

O'Ncil, R. F.—GeneralDynamics'Convair

Roy, C. E.—General Dynamics.Convair

TABLE OF CONTENTS

1.MagnetDescription1

2.SuperconductorManufacture9

3.MechanicalBehavior of ConductorWinding13

4.Coil Winding25

5.Thermal Analysis41

6.Cryogenic System49

7.Power Supply System57

8.Structural Analysis67

9.StructuralFinite Element AnalysisRefinement79

10.Structural Case Fault Analysis87

11.StructuralMetallurgy97

References117

iii

Mirror Fusion Test Facility

magnet system—

Final design report

SECTION 1

MAGNET DESCRIPTION

The Mirror Fusion Test Facility (MFTF) is the

largestofthemirrorprogramexperimentsfor

magnetic fusion energyIt seeks to combine and ex

tend the near-classical plasma confinement achieved

in 2X1IB' with the most advanced neutral beam and

magnet technologies available. The product of ion

densityandconfinementtime willbeimproved

while the superconducting magnet weight will be ex

trapolated from15 tons in Baseball II" to 375 tons

inMFTF.Otherprojectparameterslistedin

Table 1 show that the MFTF will traverse much of

the distance in magnet technology towards the reac

tor regime.3

Authorized to start construction in FY 1978.

the MFTF project is close to its schedule for com

pletionin October1981\. Following a change in

geometry at the end of the preliminary design stage.

detailed design was commencedin May 1978. By-

August of1979 the design was complete and the

first coil constructed. The second coil was finished

in March1980\. Final assembly, including the case

structure, will be completed and ready for test in

early 1981. An extension of the project called M FTF-

B has been authorized and will delay project com-

p'-lion three years.

MAGNET DESIGN

Figure 1 is a computer graphics display of the

MFTF magnet with neutral-beam injection access.

The magnet is a yin-yang pair with an average ma

jor radius of 2.5 m and an average minor radius of

0.75 m. The geometricalcenters ofthe pairare

overlappedby0.7 mtoproduceanetoutside

dimension of about 8 m and a plasma length be

tween mirrors of 3.6 m. A peak magnetic field of

7.68 T occurs at the windings in the minor radius.

Because the field is a cusped magnetic well, it drops

rapidly to 4.2 T at the mirrors and 2.0 T in the cen

ter.

A current-versus-field curve is shown in Fig. 2.

withsuperconductorstabilitylimitsdetermined

fromtest coil results reported by Cornish, et al.4

The conductor exhibits cold-end recovery, and the

stability limit appears to extrapolate in accordance

with the copper magnet resistance and a modest sur

face heat flux of 0.19 \\V-cm~2

. This experimental

observationcanbe explainedbyaveragingheat

fluxes of0.4 W-cm~2

and 0.1 W-cm"2

over the open

external and restricted internal cooling surfaces', in

accordance with usual heat transfer experiments.

The MFTF conductor is the result of a two-

year development effort.-' Listed in Table 2 are the

TABLE 1. MFTF parameters.

ParameterValue

I'lasma:

Ion density x containment time, s - c m-

'i n1 2

Ion temperature, keV511

Klectron temperature. keVt

Plasma/magnetic pressure0.5

Startup beams, A. koV10(10. 20

Sustaining beams. A, keV7511\. SO

Magnet:

Maximum field. 1'7.68

Centralfield,T2.0

Mirrorratio2.1:1

Mirror-to-mirror length, m3.6

Major radius (mean), m2.5

Minor radius (mean), m0.75

Current. A5775

Turns1392

Stored energy,M.Im

Conductor current density, A -cm3729

Coil current density. A - c m\- 2

2525

Surface heat flux, W-cm0.19

Conductor length, km50

Total Height, kg341,000

-Yin-yang

magnet

FIG. 1.Computer graphic display of MFTF magnet with neutral beam injection access.

FIG. 2.MFTF magnet load line.

primaryspecificationsofbolhthecoreand

stabilizedconductor.Itconsistsofa6.5-mm-

square, copper-stabilized,niobium-titaniumcom

positewrappedinanembossedandperforated

sheathof high-purity copper. The core-to-sheath

bond had to be improved by replacing the original

90/10 Pb-Sn solder with 50/50 Pb-Sn solder for im

provedwetabililyatlower bonding temperature.

Onceinitialmanufacturingdifficultieswere

resolved,bolhqualityandproductionefficiency

were good.

A typical cross section of the coil is shownin

Fig. 3. An inner coil form of 3I6L stainless steel is

leveledwithepoxyandglassfibers.Thenfive

overlapping layers of Kapton film are installed as a

ground-plane insulation, perforatedNEMA G-11

epoxy-glass laminate is placed on top oftheKapton

for helium circulation before the 58 layers (24 turns

TARI.K 2. MIT!conductor specifications.

Conductor parametersValue

Superconductor:

Critical current, kA at 7.5 T , 4.2 K10

Copper In superconductorJ.7:1

Number or filaments480

HUment diameter, mm0.20

Twist pitch, mm180

Conductor-resistance ratio150:1

Core size, mm6.5 X6.5

Stabilized conductor:

Maximum conductor field, T7.68

Maximum conductor current, A5775

Conductor operating temperature, K4.5

Overall: copper to superconductor ratio6.7:1

Stabilizer copper-resistance ratio220:1

Copper resistance

(at 7.68 T. 4.5 K). nit-cm46

I Iclium-cooled surface area,

cm• c m-

8.17

Required heat transfer rate, VV • c m-2

0.19

Overall size, mm12.4 X12.4

each) of conductor are wound. Afterclosing the

Kiipion ground-plane insulation aroundthe coil,

two sheets of Mylar are installed as a siip plane.

Inthesmall-radiusarea,anadditional

crushable Dacron-felt layer is applied for controlled

spacing. All other spaces are filled with G-11 blocks

and fiber-filled epoxy before the outer coil jacket is

welded in place.

A stainless-steel bladder is installed around the

encased coil so a urcthane shim can be injected be

tween the coil jacket and structure. Disks welded to

the inner structure surfacepreserveconductance

throughthe guard-vacuumspace fordifferential

pumping.Thisarrangementgreatlyrelaxesthe

helium-leakage requirements for both the coil jacket

and structure.

The coil-winding operation is shown in Fig. 4.

A tension of 600 lb is maintained on the conductor

to control the accumulation of winding tolerances

below 0.005 in. (0.12 mm) perturn.Compaction

tests on conductor stacks and computer modeling of

the winding motion confirmed that such tolerances

werecompatiblewiththeallowableconductor

strain. On initial energizing the conductor strain is

0.3%. but is reduced to 0.1% for repeated stress cy

cles.Nodegradationhasbeenobservedfor

niobium-titanium conductors at these strain levels.6

Conductor joints are made by cold welding the

central core and soldering the conductor into a cop

per tray. The joint exceeds the strength and stability

of the core and is redundant, so that some quality

control problems experienced with the cold welding

were alleviated.

Figure 5 shows the coil-winding rate. Much of

the rate improvement was associated with the ef

ficiency in joint making. Once mastered, joints were

routinelymade in less thanfour hours. Figure 6

shows the first completed coil being removed from

thewinderforplacementontheshim-bladder

assemblystandbeforethestructureisv.ciaed

around the coil.

THERMAL AND QUENCH

PROTECTION

Thermal conditions for the magnetare sum

marized in Table 3. More than 8,000 liters of liquid

helium will circulate by natural convection between

the magnet and a storage Dewar locatedon the

fourth floor of the building. The convection is the

resultof heat inputto the magnet. A computer

model of the helium loop was developed from the

Blasiusfrictionequation,Darcy'sporousmedia

equation, and a three-dimensional orificemodel.'

Estimates of effective hydraulic diameter, flow tor

tuosity, porosity, permeability, friction factor, and

effectiveorificedimensionsweremadeforthe

magnet and connection piping. The effects of two-

phase flow were included by using the Lockhart-

Martenelli correlation.8

Helium How rates were es

timated by an iterative method corresponding to an

TAHI.K 3. Thermal conditions in the magnet.

ParameterLocationValue

Helium lemperalurc. KMagnetinlet4.36

Pressure, kl'a at1.28 atmMagnet bottom130

Saturation temperature. KMagnet bottom4.52

Helium rate

(natural circulation), g/s=700

Meat load, \ \Magnet=.350

Mean quality of helium,%Magnet outlet< 5

Minimum transition

temperature, K(onduclor4.9h

-Guard vacuum space

-Copper injection bladder

-Injected urethane

filler

-316L-steel coil jacket

-Fiber filled epoxy

-Mylar-Dacron

slip plane

• Kapton ground

plane insulation

-G-11 filler block

-Conductor stabilizer

-Superconductor

core

Perforated G-11

insulation.

FIG. 3.Coil assembly detail.

assigned heal load for each model element. It was

found that, with a magnet heat load of 350 W, the

circulating helium flow is 700 g/s with less than 5%

vapor volume at ihe top of the magnet.

Quench protectionfor the magnet is accom

plishedby conventionalmeans using an external

dump resistor and a 1,000-V discharge. The magnet

time constant (see Table 4) is 69 s and an adiabatic

conductor temperature rise of 200 K is calculated

assuming a 10-s delay for the quench detection cir

cuit to sense the quench condition and activate the

circuit breakers to the power supply.

TAIil K4.Quench characteristics.

CharacteristicValue

Coil inductance, H11.0

Mutual inductance, H1.2

Peak voltage, V1000

Quench time constant, s69

Peak conductor temperature

after 100 s, K<200

Delay lime, s10

Propagation velocity, m-s~\*1.2

Quench resistor, 110.17

FIG. 4.Coil winding operation.

soliliIilii

012243648

Layer number

FIG. 5.Magnet winding rate. Rate does not include

out-of-conductortime, but does include turnaround

time and side glue blocks.

MAGNET STRUCTURE

No code or standard guidance exists for the

design of magnet structures. Instead, reference was

made to Sections III and VIII of the ASME Unfired

Pressure Vessel Code. However, blind obedience to

existing codes can result in either excessively heavy

or dangerously fracture-pronestructures. For ex

ample,paragraphUA-500oftheASMECode

recommends that 1 /4 of the tensile strength or 5/8

ofthe yieldstrengthbeusedfordesignstress.

FIG. 6.First completed coil being removed from the

winder for placement on the shim-bladder-assembly

stand.

whichever is lower. For some stainless steels like

304, the design would be limited by yield strength

andbeexcessivelyconservative,consideringthe

very high tensile strength and toughness at low tem

perature. Also. Charpy impact iests at 77 K are not

at all representative of the fracture toughness and

crack-growth properties at 4 K. so that insufficient

fractureresistance might result.Figure 7 shows a

better relationship to compare fracturetoughness

and yield strength of several stainless steels.

Ourcriteria''forthedesignoftheMFTF

magnet are summarized in Table 5. Note that the

percents of yield and tensile strength are higher than

those recommendedinUA-500for Iwo reasons:

sophisticatedelectromagnetic computercodes ac

curately resolved the forces on the magnetic struc

ture,andtheenvironmentisbenignandnon-

corrosive. Because of the tendency of materials to

embrittleatlowtemperatures,thedesignstress

dependence upon fracture mechanics at 4.2 K was

more restrictive. The plane-strainfracturetough

ness, Kic. had to be compatible with the detectible

flawsize.a.Equallyimportantwasthecrack

growthrate, da/dn. during cyclic loading condi

tions. MFTF was designed for a life of 2,000 stress

cycles corresponding to a safety factor of four dur

ing the expected10-year service life.

Whentheaforementionedgeneraldesign

criteriawereappliedtoMFTFandmaterials

properties evaluated,the specificstructure-design

criteriain Table 6 were adopted for the detailed

400SymbolMaterialCondition

400•Nitronic40Annealed

O304Annealed

i304 LAnnealed

£i304 LNAnnealed

"55\- „ „•516Annealed

J2300•516 LNAnnealed

£\••Kromarc 58Annealed

P®310SAnnealed

§°\\s>a310 SSensitized

2®\ANitronic 33Annealed

se200" o•^ V Z \_VNitronic 50Annealed

$" o•^ V Z \_ANitronic 50 KHigh-temp

n fra

a^\\\\^\_anneal

Plane-stra

I,I,I

50100150

Yield strength (ksi)

200

FIG. 7.Decline of fracture toughness with increasing yield strength.

stressanalysistobeperformedbyGeneral

Dynamics,ConvairDivision.1 0

Afinite-element

SAP IV computer analysis was performed and sup

ported by various detailed calculations. In the peak

stress region of the structure, three-dimensional ele

ments in the NASTRAN code were used to refine

the analysis, so that no calculated stress exceeded

the 80 ksi criteria.

After a review of available material properties.

304LN stainless steel was chosen for the structure.

Extrapolation ofNBS data" indicated that the yield

strength of 304LNwould equal120 ksi. if 0.14%

nitrogencouldbeadded,whilethefracture

toughness. Kjo was expected to remain about 200

ksh/nT. Production of 800,000 lb of 304LN steel was

successfully completed with the material passing all

TABU"5.Designstresscriteriaformagnetic

structure.8

TAB 1.1\*' 6. Minimum structural materials properties.

StressCriteriaProperty

Design

Design, cycles

2/3 yield strength"

\[00% yield strength\*\
\
1/2 tensile strength\
\
1/2 KK7%/™\
\
4 lifetimes\
\
'The lowest stress criterion is chosen from among those listed.\
\
Primarily tensionMMcombinedstresses.\
\
€\
\
l'rim»rily acrting.\
\
Design stress, ksi\
\
Yield stress, ksi\
\
Ultimate stress, ksi\
\
Elongation,%\
\
(harpy impact at 77 K\
\
Absorbed energy, ft-lb\
\
Lateral expansion, in.\
\
Toughness (K\|( ), ksK/In".\
\
120\
\
160\
\
20\
\
40\
\
0.030\
\
120\
\
6\
\
chemistry, microMructural, and ultrasonic inspec\
\
tions, and ihe steel was supplied for the coil struc\
\
ture fabrication shown in Fig. 8.\
\
Anextensivedevelopmentprogramwas\
\
necessaryto achieveweldpropertiesthatwould\
\
match those of the base melal. Because 316L weld\
\
metal promised to have good yield andultimate\
\
strength,itwasselectedfordevelopment.Steel\
\
toughness was known to be influencedby ferrite\
\
content1\
\
" (see Fig. 9), so very low ferritecontent\
\
was necessary. However, below 3% ferrite, miero-\
\
fissuring of this weld metalbecomes a problem.\
\
Several welding methods were attempted, but only\
\
shielded metal-arc welding produced an adequate\
\
combination of toughness, sufficient welding speed,\
\
and versatility.1\
\
" Table 7 summarizes a few of the\
\
weld and base m^tal properties which qualified the\
\
coilstructuremanufacturing.Carefulcontrolof\
\
purity, ferrite level, and welding methods achieved\
\
the demanding requirements.\
\
\\A\r•i w 4\
\
TABLE 7.MFTFstructural material properties."\
\
YieldUltimate\
\
K \| £ ,strength,strength,Reduction,\
\
ksi\\/in".ksiksiarea-Tc\
\
3I61.-15\
\
Weld183112.(1183,228.1\
\
304LN\
\
Ruse material2112111.6237.436.0\
\
a\
\
Ali measurements made at4K.\
\
175\
\
kS150-N\
\
S125-\
\
100\
\
t:75-\
\
50\
\
11\
\
1\
\
•\
\
0\
\
ring rang\
\
issures\
\
cro fissu\
\
w micro\
\
\_-\
\
WeldElectrodeBase\
\
metalcoatingmetal\
\
-A316 LTitania304 L"\
\
A316Titania304 L\
\
0316 L\
\
1\
\
Lime\
\
i\\
\
304 LN\
\
8\
\
Ferrite (%)\
\
12\
\
FIG. 8.Coil structure fabrication.FIG. 9.Effect of ferrite content on weld toughness.\
\
7-8\
\
SECTION 2\
\
SUPERCONDUCTORMANUFACTURE\
\
The MI-TI-' conductor design was developed as\
\
pariof the general magnet development work at\
\
I l.M ."" FachU.S.manufacturerproducedone\
\
hillelo\\'proton pe core material and, based on the\
\
resultsofthiswork,aspecification(MKL77-\
\
OOI37\
\
3A) was prepared. After selecting a manufac\
\
turerfhiiermagneticsCieneralCorporation),the\
\
details of ihaimanufacturer'sdesign were incor\
\
poratedintoamodifiedspecification(MEL77-\
\
00\]37

3O which was used as a basis for the produc

tion order,

•\coppt'rsheath,showninFig.10,was

soldered to the core to provide add'Jonal cooling

lor ci'ostatic stabilitv. This process was contracted

loAirco."

In this wraparoundtechnique, the supercon

ducting core is passed through a continuous elec

troplating process prior to applying the stabilizer.

1 he latteris preparedfroman oxygen-freehigh-

FIG. 10.MFIT superconducting core and copper

sheath.

conducting (OFHC) copper by first slitting and roll

ing a strip, followedby punching helium access

holes. On a roll-forming line, the stabilizer is wrap

ped around the core, and the conductor is then sized

by a Turk's-head. The completed conductor then

passes through the soldering furnace, a quenching

system, and a cleaning process before being wound

onto the supply drum.

The choice of a monolith conductor presents

severalmanufacturingandqualityassurance

problems not encountered with a cable or braid. In

order to achieve optimum critical current in Nb-Ti,

alarge amount of cold work is required.'5

This

cold-work requirement can be met for a cable or

braid by starting with a billet of 150- to 200-mm

diameter, because the braid strand size is typically

0.5 to 2.0 mm in diameter. The final size of the

MFTF core is 6.5 mm by 6.5 mm; consequently, a

large-diameterbilletmustbeused,andthe

manufacturingrisks are greater. In addition, the

MFTFmagnetrequiresrontinuouslengthsoi

greaterthan380 m;large-diameterbilletsare

necessary to produce these lengths in an economical

manner, and breakage during manufacture must be

minimized.

The absence of redundancy in a monolith such

astheMFTFsuperconductorrequiresthat the

quality assurance tests be especially rigorous. Con

sequently, samples from each end of each length

are checked for critical current (Ic), matrix-resis

tivityratio(R293K/RIOK).copper-to-supercon

ductor ratio, twist pitch, filament size, and filament

integrity. A basic premise for the quality assurance

of the MFTF superconductor is that sampies from

both ends of each length are sufficient to guarantee

the quality of each length. Nb-Ti alloy in the com

position range 46 ±1.5 wt% Ti, balance Nb, was

specified. The Nb-Ti alloy presented few problems

and rejections were under 1%. Defects were mainly

associatedwithsurfacequality,size,and

straightness. Many of the surface quality rejections

were discovered by the manufacturer to come from

"fretting" (self-abrasion)ofthematerialduring

transit. The solution was simply to pack such that

relative motion between the rods was minimal. In

gotandproductchemistrywerealwayswithin

specification.All copperused in the billets was

eitherphosphorous-deoxidizedoxygen-free

(PDOF)orOFHCbrandwithaguaranteed

resistivity ratio of greater than 180:1.

Conductorfabricationfollowedarelatively

standardsequence for all Nb-Tisuperconducting

composites.'6

The billets were assembled using hex

agonal copper tubes in which cylindrical Nb-Ti rods

were inserted. All elements were chemically cleaned

and stored under dry nitrogen prior to assembly.

The array was slacked to fit within a copper extru

sion can which became the shell of the conductor.

The billet was thencapped with a nose and lid.

evacuated, and sealedby electron-beamwelding.

Electron-beam welding was used to ensure repro

ducible welds.

The extrusion operation reduced the billet to a

much smaller rod and metallurgically bonded the

interfaces togethertoyield a true composite. The

presscapacityof5000tonnewas almostfully

utilizedin the extrusionof the billet. An initial

problem of tooling failure resulted in the misextru-

sion of several billets. The problem was remedied by

the design of special tooling to fully utilize the press

capacity.

The ends of the extruded rod were cropped and

the rod wascutinto two lengths. These rods were

drawn through a series of dies using several draw

benches.

After the rods werereducedto 20-mm diam.

drawingwas doneonconventionalbullblocks

speciallymodifiedtohandlethehightensile

strengths of these conductors. An intermediate heat

treatmentwasintroducedduringthedrawing

processtoobtainthe maximumsuperconductor

properties. Few problems were noted in the drawing

for sizes below 20 mm. The conductor was twisted

and Turk's-headed by a special machine designed to

rotate the pay-off spool. The final conductor sizing

to a tolerance ±0.05 mm was done with a tungsten

carbide die. A final annealing step was performed to

restore high conductivity to the copper matrix.

QUALITY ASSURANCE

The various steps in the manufactureofthe

MFTFsuperconductorwerereviewedfromthe

standpoint of quality assurance, and a comprehen

sive planwas prepared.This planspecifiedthe

documentation required for each length of conduc

tor and the method for verifying that each manufac

turingstephad beenperformed.A writtendis

crepancy report was required for any condition or

procedure which differed from the quality assurance

plan.Whenmanufacturewas complete,samples

from each end of each length were tested to see that

the specifications were met. These results, together

with any discrepancy reports filed during manufac

ture, were reported to a MaUrials Review Board

whichmadethefinaldecisiononanyout-of-

specificationmaterial.

CRITICAL-CURRENT TESTS

A total of 350- to 400-crilical-currenl measure

ments were anticipated and a test fad lit) capableof

handlingthis large numberof samples was con

structed. A superconducting solenoiu with a 25-mm

by 50-mm access produces a maximum transverse

field of 8.5 T. and a regulated12.000 A(dc) power

supply provides current to the sample. The sample

hoiderwithhelium-vapor-cooledlendscanbe

changedwhilethesolenoidismaintainedat

cryogenictemperatures.Threeofthesesample

holders were constructed so that four samples can

be tested in one day. Voltage laps spaced 25-mm

apart were attached 'oAtesample in ihe unir

orm

field region, and a criterion of 10"'Si-cm resistivity

(fortheNb-Tiarea)isusedtodefinecritical

current.

One problem encountered in testing large high-

current specimens is that the self-field generated by

the currer-i. in ihe specimen becomes significant. Fjr

theca.se ofuntwistedfilaments,calculation- in

dicatedthattheinfluenceofself-fieldonthe

measured critical current would be small, due to this

effect enhancing the field on one side of the sample

but reducing ihe field on the other side. However,

when the twist pitch (190 mm) of the filaments in

the MFTF conductor is taken into account, the self-

field acts to reduce the measured critical current by

about7%. Measurementsof criticalcurrents on

identical samples in the twisted and untwisted con

ditionverifiedthiseffect.However.ILforthe

MFTF samples were typically 15 to 20'~r above the

specificationof10.000 A at 7.5 T. so self-fieldef

fects have not created a problem in acceptance of

samples. Anotherconsiderationfortestinghigh-

current samples is the current-transfer length: both

the lengthnecessary to transfercurrent from the

sampleholderto the sample and the lengthfor

redistributionof the currentfromthe outer fila

ments to the inner filaments as the sample passes

fromthelow-fieldtothe uniformhigh-fielden

vironment;ireimportant.Insufficientareafor

current transfer from sample holder to sample can

result in heating. Insufficientlength in the uniform

field region on either side of the voltage taps can

resultin a voltage signal due to currenttransfer,

rather thandue to exceeding the criticalcurrent.

The length in the first case can be obtainedfrom

Rcfs. 16 and 17. The second case is more difficult to

calculate,so we determinedit experimentallyby

placingvoltagelapsspaced25-mmand50-mm

apart. As a result, we found that using a 610-min-

long sample with 100-mm-long solder joints al the

ends and voltage taps spaced 25-mm apart would

prevent current-transferproblems from occurring.

COPPER-TO-SUPERCONDUCTOR

RATIO

The copper-to-superconductor ratio {Cu:S.C.)

is required to be between 1.63 and 1.77. The choice

ofthisrangewasbasedonconsiderationsof

stabilil}.strength,andqualityofcold-welded

joints.l s

This parameter is determined by weighing a

100-mm length of core, dissolving the Cu matrix.

andweighingtheNb-Tifilaments.Themajor

variable affectingthe Cu:S.C. ratio (assuming the

currentratio is used in billet construction) is the

amount of cropping done al the nose and end of the

extrudedrod. Initially, a number of lengths were

foundlo be outside this specification. Ho'vever. a

master curve of ratio versus distance along the con

ductor length at final size has been developed, so

that the amount of cropping necessary can be deter

mined from the Cu:S.C. ratio al any point along the

leniith.

MATRIXRESISTIVITY

RATIO

The matrix resistivity ratio is specified to ex

ceed a value of150\. This ratio is determined by

measuring the resistance at 293 K and at 10 K: the

i0 K value is obtained by lowering the sample and

its thermometry through the temperature gradient

above liquid helium. We mount up lo six samples in

a horizontal plane and connect them in series, so

thai six measurements can be made in each experi

ment. Only one lenglh of MFTF core failed to meet

the resistivity ratio specification.A review of the

records indicatedthat this length has missed the

finalannealingstep;afterannealing,itmet

specification.

FILAMENT TWIST PITCH

The twist pitch requirement for mirror fusion

magnets such as MFTF are not as stringent as for

Tokamak magnets, since the mirror magnets are not

exposed to rapidly varying fields. The twist pitch for

theMFTFsuperconductorisdictatedbythe

proposedchargingrateandshouldbe less than

100 cm. A specificationfor twist pitch was chosen

as between 16.5 cm and 19.0 cm; this range is easily

achieved in a conductor the size of MFTF without

danger of the twisting operation causing filament

breakage.

Samples were foundwithout-of-specification

twist pitch at the early stages of conductor delivery.

This problem was traced to a method of starting the

twisting operation. The manufacturingprocedure

was changed to avoid this problem by performing

the end-cropping operation after the twisting opera-

lion.

The only olher problem involving twist pitch

resulted from a malfunction in the twisting-squaring

operation when a pin sheared in the twist machine

drive. This resulted in a 12-m length having no twist

and required that the entire length be rejected. This

experience is one case in which sampling each end of

each lenglh was inadequate to guarantee the quality

of the length; it was necessary to rely on manufac

turing quality control to note the problem and to

file a discrepancy report.

FILAMENTSIZE

ANDINTEGRITY

SinceMFTFisacryostable,steady-state

magnet, filamentsize does not play an important

role. However, fiiament uniformity and filament in

tegrityareanimportantindicationofgood

manufacturing practice. A specification of filament

size between 0.18 mm and 0.23 mm was established

for this conductor. Metallographic examination of

sample cross sections indicated that lengths from

billets prepared in the early stages of this program

contained nonuniform filaments. A typical case en

countered contains several filaments measuring 0.4

mm in diameter (the worst case identified showed

one filament measuring 0.8 mm by 0.4 mm). This

condition can arise if a billet is not packed densely

enoughandcanoccuranywherealonga given

length. This is the other instance in which we found

a problem that can remain undetected in samples

taken from each end of each length.

Subsequently,improvedbilletassembly

procedureswereintroducedandtheseextreme

rangesinfilamentsizes werereduced.This ex

perience suggests that applications needing a close

toleranceonfilament size, e.g., those requiring in

trinsic stability or low ac losses, should specify close

tolerances on billet density.

Filament integrity was not anticipated to be a

problem for the Nb-Ti core, sir.ix ;he filament size

(0.2 mm) is rather large and the amountof cold

work is moderate for Nb-Ti. However, routine ex

amination after removal of the matrix revealed slip

lines on the filaments and occasional breaks. The

densityofbreaks, i.e., approximately one in a 2-cm

length, suggestec\* that broken filaments should not

affect the short sample critical current, and the ex

perimental measurements bore this out. The other

concern was that broken filaments might af'fecl the

strength of the core, so a series of tensile measure

ments were made at 4.2 K. Again, there was no dif

ference between samples with and without broken

filaments.Consequently,lengthswithup to1%

broken filaments in a 2-cm length have been ac

cepted.

Analysis of the problem indicates that filament

breakage occurs in the final sizing steps; filament

breakage occurs only in the outer filaments where

the deformation due to converting from a round to

a square cross section is greatest. This problem can

most probably be eliminated by the addition of an

annealing step prior to final sizing or by changing

the filament distribution. However, these corrective

actions are not being attempted at this time, since

thepropertiesoftheMFTFcorestillexceed

specification and the manufacturing changes would

resultinconsiderabledelaysinconductor

manufacuture.

A benefit of this quality assurance program has

beensignificantimprovementsinmanufacturing

techniques: (1) better billet assembly procedures. (2)

less relianceonoutside rod drawing capabilities. (3)

controlled rod-cropping procedures, and (4) careful

monitoringofthe manufacturing steps to ensure

that each step is performed to all specifications.

SECTION 3

MECHANICAL BEHAVIOR OF CONDUCTOR WINDING

INTRODUCTION

The following is ;:immary of analytical and

experimental studies into the mechanical behavior

of the Mirror Fusion Test Facility (MFTF) magnet

superconductor pack. The superconductor pack is

here defined as the 24-turn by 58-layer coil with its

interturnandinterlayerinsulationandany fillei

material between the superconductors and the coil

jacket. The relative shapes and sizes of the MFTF

conductor,structuralcase,andplasmafanare

shown in Fig. 11. Mechanical behavior is defined as

the stress and strain that the superconductor pack

components experience during magnet winding and

normal 2-T central field operation, us well as any

anticipated conductor motion that impacts the in

ternal design of the coil.

Mechanical properties of the coil components

at 4 K are reviewed. This is followed by a descrip

tion of the electromagnetic loads that the coil will

experience, theirmethodof calculation,andthe

redistribution of these loads due to structural com

pliance. Investigations of the possible buckling ef

fects of superconductor pretension on the jacketed

coil are also discussed.

Superconductor stress and strain analysis, per

formed by using a selected range of coil mechanical

properties, are summarized, as are the possible ef

fects of static and cyclical strain on the niobium-

titanium superconductor and its copper stabilizer.

Finally, an analytical prediction of superconductor

motion is detailed, including a diagram of the ex

pectedsuperconductorpackdisplacementwith

respect to the coil jacket.

MECHANICAL PROPERTIES

Thefirststepinmodelingthemechanical

behavior of the MFTF coil was to determine the

physical properties of its components at 4 K. The

MFTF superconductor pack may be thought of as a

modified orthotopic composite that, can carry only

compressive loads in two directions and only tensile

loads in the third direction. In addition, the proper

tiesineachdirectionarenonlinearandexhibit

hysteresis in their stress-strain responses.

The MFTF superconductor core is a 0.25-in.-

squarecooper-stabilizedniobium-titaniumcom

posite consisting of 480 superconducting filaments

in a copper matrix with a copper-to-superconductor

ratio of 1.7.An embossed and perforated copper

wrapis soldered around the core to provide heat

transfer and mechanical support to the core. The

final assembled conductor is 0.490-in. square.\*"'

TensiletestsoftheMFTFsuperconductor

were performedin liquid helium lo establish both

thestress-strainresponsecurvesandthetensile

failure loads."" Strain-gauged samples were used to

obtain the stress-strain curves up to an elongation

of 1% using foil gauges designed for use in liquid

helium.2 3

Figure 12 shows the stress-strain response

of the wrappedsuperconductoratliquidhelium

temperature. Unload/reload cycles always followed

hysteresis loops while loadings beyond the previous

peak always followed the envelope curve until being

unloaded.

ThebareMFTFsuperconductorcorewas

found to fail in tension at about 8200 lb. while the

wrappedsuperconductorassemblywas pulledlo

beyond 10,600 lb without tensile failure. The addi

tion of a cold weld joint to the superconductor core

reduced the failure load to 7100 lb, showing very lit

tle sensitivity to the number or length of cold welder

strokes. The technique chosen for MFTF joints was

five cold welder strokes at 1/2 in. per stroke. An

nealing of the cold weld due to soldering of the cop

per stabilizer further reduced its strength to 5700 lb.

Thecold-weldtensiletestsshowedalarge

spread in tensile failure values. The design require

ment of high reliability placed on the magnet system

meant that adequate joint strength must be guaran

teed. This matter was resolved by the addition of a

monolithiccopper joint-reinforcementbar.Joint

assembly verification tests showed that the copper-

reinforcedcold-weldjointscanwithstandthree

timesthemaximumanticipatedaxialloadof

3000 lb and are stronger than the parent supercon

ductor even when no cold weld is applied, thereby

providing 100% joint strength redundancy.

The coi! pack is made up of the above super

conductor in 58 layers of 24 turns each.'Interturn

200-ir

200

FIG. 11.Relative size and shape of MFTF conductor, structural case, and plasma fan.

insulationis composed of 0.045-in.-thickNEMA

G-l 1 buttons glued to a woven string.2 6

Inlerlayer

insulation is 0.0625-in.-thickslotted NEMA G-l I

sheets.2 7

The compressive response of the superconduc

tor coil pack was measured at liquid nitrogen and

roomtemperatureusinga coil packmock-up."8

This series of tests revealed a linear compressive

modulus of 2.0 X 106

psi in the interturn (button)

direction and 3.0 X I06

psi in the interlayer (slotted

sheet) direction. Both response curves exhibited an

early soft region followed by a stiff linear modulus.

Figure13 shows compression test danin the in

terlayerdirectionwhichillustratesthisphenom

enon. The initial nonlinearcoil packbehavior is

associatedwiththeunevensurfacesof the pack

components. If ideally "flat"mating surfaces are

assumed, the early softness can be interpreted as a

0.005-in. gap per layer of superconductor.

Accumulated stack height measurements per

formed on the coil winder showed actual winding

gaps to be less than 0.005 in. per layer per turn at all

monitoredstationsonthecoil.2 9

Coilwinding

procedures such as tensioning at 600 lb. clamping,

and coil-height measurements place an emphasis on

obtainingatightwinding.Thisreducesthe

Test terminated

at 10,600 lb

20J=

0.10.20.30.40.50.60.70.8

Axial strain (%>

FIG. 12.MPfF superconductor tensiletestdone in liquid helium.

magnitudeofcoilpackmotionduringmagnet

energizing and de-energizing.

A plenum region is provided between the top of

the 58thlayer and the inside surface of the coil

jacket for the collection of helium bubbles outside

ofthesuperconductoritself.Theplenumfiller

material is composed of laminated sheets of slotted

NEMAG-ll. The laminated assembly was com

pression tested and found to be structurally sound

to beyond 4000 psi.— Compressive strengthwas

needed only during coil closure since the bubble

plenum region experiences very small compressive

loads during magnet operation.

The outside face of the 58-layer coil pack bears

againsta slip plane on the large radius of both

magnets as shown in Fig. 14. The slip plane, con

sisting of two sheets of 0.007-in.-thick Mylar, allows

relative motion between the superconductor pack

and the surrounding coil jacket to occur without

damaging the Kaplon or NEMA G-llinsulation.

Analysis oftheslip-planerequirementsfounda

coefficientoffrictionof0.9orlesstobe

3000

IIIII

2500

E = 3 X 1 06

p s i ^-

2000-

1500-

1000 Over 95% of total strainI-

500occurs below 500 psi pressureJ-500A-

I\_j\_\_—Ii l l

00.20.40.60.81.01.2

Conductor pack compressive strain (%)

FIG. 13.Compaction of MFTF conductor pack in-

terlayer (slotted G-ll) direction at LN temperature.

Slotted G-11

Superconductor

G-11 buttons

KIG. 14.Bearing face of MFTF superconductor pack.

satisfactory.'0

Laboratorytestsonaslip-plane

mockup showed a friction coefficient of 0.17 for the

Mylar-on-Mylar configuration.3 1

'3 2

Tests were also performedto verily the com

pressive strength of the filled epoxy used to grout

the completed coil prior to applying the jacket.

This investigation included a thermal shock test us

ing an embedded steel washer. No damaging effects

due to differential thermal contraction were found.

The compressive load, at which the soft copper

wrap on the superconductor begins to be perma

nently indented by intertum insulation, was found

by testing to be 4500 psi.3 4

The coining load in the

interlayer direction is 5400 psi due to the larger ef

fectivecontactareaoftheNEMAG-llsheets.

These compare favorably with the anticipated 3400-

psi interturn and interlayer compactionpressures

exerted during magnet operation at full field.3 5

ELECTROMAGNETICLOADING

The MFTF yin-yang magnet geometry and the

direction of current flow in each coil is illustrated in

Fig. 15. As thediagramsuggests, thelarge and

small radii of the two coils behave somewhat like

segmentsofsolenoids. Thepeakmagnetic field

value of 7.68 T occurson(he inside surfacecf the

coilpackinthe smallradiusregion, as seen in

Fig.16\. The similarityto a solenoiddiminishes,

however, when the transition fromlarge to small

radius isconsidered (asshown in Fig. 17) where the

transitionregiondisplays highgradients of elec

tromagnetic forces.

Most of the electromagnetic load calculations

for the MFTF yin-yang magnet were performed us

ing the Electromagnetic Fields. Forces and Induc

tance(Fl-T-T)computerprogram.3 6

"3 8

FI-FIis

capable of modeling an arbitrary system of coifs

made from circular arc and/or straight segments of

rectangular cross-section conductors, and was used

extensively in the analysis of the MFTF magnet.

The internal pressure exerted on the side vails

of the coil jacket by the superconductors has been

(2 out of page)

FIG. 15.MFTF magnet current direction.

FTG. 16.MFTF magnet field distribution—minor

radius symmetry plane.

1111

Present design

Reduced aperture design

1!1

Transition -^.\ S

11!1

Transition -^.\ S

-4

1!1

Transition -^.\ S

-41111111i

-2

-1

0.51.01.52.02.53.03.5

Distance (s) along coil centerline (m)

4.0

FK». !7.MFTF magnet. Average magnetic pressure—EFFI calculation prior to redistribution.

calculated using an HFFI computer program model

and the load redistribution extracted from integral

coil/structure finite element modeis.The analysis

models built by GeneralDynamics/Convairused

sixrodelementstorepresentthecoilpack.

Transversecoilpackmemberswithappropriate

packmoduluspropertiestransferredtheapplied

loads to the structuralcase.The finite element

model is illustrated in Fig. 18.

Another way to illustrate the EFFl-generated

loads in Fig. 17 is to use a coil sketch with applied

load vectors, as Fig. 19 shows. These loads do not

account Tor the effects of the structural case.3 9

Due

to the compliance of the supporting case and the

motionof the coil pack with respect to the coil

jacket, the electromagnetic pressure exerted on the

surrounding material is lessened in the small radius

region.4 0

This effectcanbe seenbycomparing

Fig. 20 with Fig. 19.

The superconductor and copper bus lead-outs

penetrate the coil jacket and structural case at the

helium vapor exit pipe in the manner shown earlier

in Fig. 15 and with the applied loads from Fig. 21.

The lead supportstructure has been designedto

withstandthe maximum expectedelectromagnetic

loads. ' The geometry selected for the iead-out path

allows coil pack motion to occur without jeopardiz

ing thelead-out assembly, as dealt with in more

detail later in this report.

COIL WINDING TIGHTNESS

As Fig. 14 shows, the MFTF coil pack is made

upofalternatelayersofsuperconductorand

NEMAG-11 insulation. The tightness of the wound

coil influences the superconductor pack motion and

pressure loadredistribution.The degree of pack

tightness is quantitativelydescribedas the stack-

height buildup per layer of conductorabove the

sumoftheindividualcomponentheights.This

phenomenon is the result of surface features of the

superconductorandNEMAG-llwhichprevent

continuous surface-to-surfacecontact.

When fictitious, perfectly flat surfaces are con

sidered in analyses, the early reduced compressive

stiffness shown in Fig. 13 is interpreted as an initial

per-layer '"gap." As a result, reference to the word

Coil 2

Complete model — plate elements

FIG. 18.MFTF—refined model.

Rod elements modeling

conductorpack

1910 psi

3100psi

3100 psi

FIG. 19.

tral Held.

MFTF electromagnetic loads at 2-1 cen-

"gap" in the context of coil winding has come to

representiiis analog of a surface contact phenom

enon.

Severalmethods were employedtomaintain

the tightness of the MFTF coil pack. The supercon

ductor was continuously held at a tension of 600 lb

duringcoil winding.Small-radiusclamps helped

retain superconductor pretension, and side clamps

closed interlurn gaps on the large radius of the coils.

The clamping, tensioning, and compaction measur

ingtechniquesaredescribedintheMFTFcoil

winding specificaton. The actual winding tightness,

achieved with the tensioning and clamping scheme

describedabove,wasfoundfromstackheight

measurements lo be from 0 lo 5 mils per conductor.

An intermediate effect of this pretensioning is

the dislorlion or possible buckling of the jacketed

coilduringtransfertothestructuralcasesub

assembly. An early analysis indicated that a poten

tial problem existed and that furtherinvestigation

was needed.Interference between the jacketed coil

Inside surface of stainless steel jacket/coil form

Outside surface of superconductor pack

Note: Displacement exaggerated

for clarity

Ref. data from GCD-LLL-79-001

appendices A, B & C

FIG. 20.MFTF average redistributed magnetic pressure (psi).

and the structural case due to coil distortion would

resultin assembly delays andpossihly the design

and fabricationof" additional hardware.

A comprehensive buckling analysis was laier

performedb\GeneralDynamics/Convairwhich

took into account a conservative 50f

.\> relaxation of

the windingpretension.This analysis predicted

that the jacketej coil would not buckle when the

coil-formadapterstrongbackwasremoved.The

analysesalsopredicteda displacementofabout

0.9 in. at the large-radius symmetry plane which is

withintheallowablerangeforstructuralcase

assembly.

These, as well as earlier analyses, were very

conservative and highly dependent on an accurate

prediction of superconductor pretensile load reten

tion. A program of strain measurement was under

taken to experimentally determine actual winding

load on the coil form.Permanentstrain gauges,

meant to be read during magnet operation, were

readduringcoilwinding.Additionaltemporary

coil-form strain gauges were applied to complement

the permanentinstallations.4 4

All of these gauges

were also read during coil demounting and transfer.

Preliminary results from these efforts confirmed the

conservative nature of the analytical assumptions.

SUPERCONDUCTOR

STRESS AND STRAIN

Theanalyticalpredictionofsuperconductor

stress and strain in the MFTF magnet evolved from

early hand analyses to sophisticated finite element

analyses and parametric studies. Hand calculations

were performed using simplifying assumptions such

as no relative motion of conductors with respect to

other conductors and the supporting case,4 5

or of

uniform stress or solenoid-like behavior.When a

relationshipbetweensuperconductorwinding

tightness and strainwas shown,4 4

tests were ini

tiated to determine the actual stress-strain response

•vCPN

A l l values in 16 lb/in

Coil1 (west or+Zcoil)

FIG. 21.Klectromagneric forces on MF'I'F superconductor leads.

of the wound coil pack for inclusion in analytical

models.

The stress analysis of the MFTF magnet struc

turalcase'"\*1 6

ledtoimprovedmethodso\\~

predicting the effects of case deflection on supercon

ductor strain.4

This early work, however, did not

address the effects of winding tightness and coil-

packmotionwhichtendtoredistributestresses

across the pack, as Fig.22shows.

The effects of winding tightness on conductor

stressandstrainwereinvestigatedbyGeneral

Dynamics Convair.4

' A parametric study was con

ducted to measure the degree of coupling between

coil-packmodulusandconductorstress.Using

three differentsimulatedcoil-packmoduli to ad

dress winding gaps of 0. 5. and 10 mils per conduc

tor, it was found that although the coil motion in

creased with larger gap si/e. superconductor stress

Superconductor stress (ksi)

I — i — i — I — I — I

05 10 15 20 25

H(;. 22.MITK superconductor stress distribution.

tended to redistribute itseliwith a slight reduction

in the peak from a value of 22 ksi with no gaps to

about 19 ksi with 10-mil gaps. Figure 22 graphically

illustratesthechangeinsuperconductorstress

distribution.

The analyses have indicated a new conductor

stress of about 22 ksi when the magnet is at its max

imum operating current of 5775 A. Due to non

linear superconductor behavior and hysteresis, the

superconductorstrain can reach a peak of 0.25\*5

during the first energizing. This number is reached

by neglecting any radial support of the coil pack in

ihe small radius. However, of this 0.25%. 0.15% is

nonreversiblepermanentset and 0.10% is clastic

strain with little variation in the hysteresis loop for

subsequent loadunload cycles.

The wrappedsuperconductorandreinforced

joint assemblies were successfullypulled in liquid

heJium to about three times the expected axial load

levels. Of greater concern was the possible reduction

in the critical-current value of the superconductor

and degradation of the residual resistivity ratio of

thecopperstabilizerduetostaticandcyclical

straining.

A comparisono^MFTF operating strains with

currentliteratureonstraineffectsrevealedno

measurable degradation of the critical-current value

for a peak strain of 0.25% or for repeated strainings

to the same level, even to the point of fatigue failure

ofthe specimen.A 3%. reductioninthe copper

stabilizer residual resistivity ratio was predicted for

asinglestrainof0.25%.However,nofurther

degradation could be found for subsequent cyclical

straining lo 0.10r

?.4 s

COIL PACKMOTION

Superconductormotion was considered when

designing the internal components of the coil pack

suchaslead-outs, jointandrampdesigns,slip

planes, helium-bubble plenum details, and various

fillers and supports. The motion of the MFTF coil

pack with respect to the jacket and structural case

w;»s predicted by postprocessing finite element com

puteroutputprovidedbyGeneralDynamics/

Convair.\- 2 , 4 7

The diagram in Fig. 23 depicts the coil

pack with exaggerated displacements to emphasize

the directions of motion. Displacementvalue will

vary due to actual as-wound conditions.

The slip plane previously discussed allows the

magnet to safely accommodate several limes the coil

motion that is anticipated. An analysis of the design

requirements of the slip plane depicted in Fig. 14

showed that a coefficientof frictionof 0.9 or less

would prevent damage lo the slotted NFMA G-l 1

inlerlayer insulation. This was because95r

iof the

conductor motion occurs before 40% of the peak

magnetic field is reached and when electromagnetic

loads are relatively small. Tests of several slip-plane

designs showed two sheets of 0.007-in.-thick Mylar

to be more than adequate, with a coefficient of fric

tion of 0.17 or less at cryogenic temperatures.

Conductor stress and motion analyses revealed

that a reduction in normal load on the slip plane in

the small radius and a more favorable mechanical

environmentforconductorlead-outsandcoil

diagnosticswouldoccurif radialmotioninthe

small radius was allowed. In the small radius region.

GeneralDynamics/Convairanalysesindicateda

tendency of the conductor pack lo self-supportor

lightly bear on the jacket wall if radial motion was

not heavily restrained. As a result, the radial inside

surfaceof the coil jacket in the small radius was

designed lo include a layer of highly compliant felt

paddinginadditionto theMylar-on-Mylarslip

plane. This reduced or eliminated damaging radial

loads in the end regions where internalanomalies

0.30Inside surface of stainless steel jacket/coil form

-Outside surface of superconductor pack

Note: displacement exaggerated

for clarity

Ref. Data from GDC-LLL-79-001

Appendices A, B & C

0.02

0.20

FIG. 23.MFTF average superconductor pack motion (inches).

such as lead-outs and permanent clamps must be

installed.

Predicted coil motion also impacted the design

of penetrations in the NEMA G-l I sheets where the

superconductor must ramp up from one layer to the

next. The possibility of a short circuit due to relative

motion of adjacentlayers was eliminated by sur

rounding the conductor penetrations with NEMA

G-11 blocks. This arrangement safely allows over

3/8 in. of relative interlayer motion, 30 to 40 times

the amount that is predicted by idealized analyses.

Large excursions could occur if layers were to seize

rather than slip. Even though this is highly unlikely.

the conservative penetration design allows for these

movements.

Severalotherdetailsofeoiiinternaldesign

were influenced by predicted coil motion. Supercon

ductor strain-gauge leads were routed and strain-

relieved to allow for motion of the gauged region.

Interturnbutton insulation was glued to a woven

string and slots in the interlayer sheets were made

too small for the butlo/s to pass through, in order

to mechanically trap the buttons if they are scoured

offofthesuperconductor.Coillead-outswere

positionedto' allowcoilpackmotiontooccur

without mechanically loading the lead assemblies.

2.?

Glue blocks on the outer edge of the inlerlaycr in

sulation provide J flat machined surface for slip-

plane motion.

The MFTKcoil pack mechanical behavior may

best be summarized by two statements: Much effort

was pul into winding tight coils to minimize the coil

pack motion that was predicted by analyses. Given

the directions and relative magnitudes of the expec

ted motion, coil internal details were designed to

allow for several times the anticipatedloads and

motionwithoutcompromisingmagnetperfor

mance.

23-24

SECTION 4

COIL WINDING

I N T R O D U C T I O NTABLE 8. C-coil dimensions.

This section describes the basic criteria used in

the winding of the MFTFyin-yang magnets. It also

covers the equipment, special tooling, and materials

necessary for the winding and enclosure of the coil

in a 0.5-in.-thick stainless-steel jacket.

COIL GEOMETRY

The yin-yang magnet consists of two C-shaped

coils that are enclosed in a structural case and at

tached roughly in the shape of a ball. Figure 24

depicts the geometry of one coil and Table 8 lists

coil dimensions.

FIG. 24.MFTF coil geometry outline.

Value

ParameterMetersInches

Major radius2.51)98.45

Minor radius0.7529.53

Cross section

HeightU.<)H38.5H

Width(1.3915.35

The coil is wound in pancake construction with

superconducting conductor. There are 58 layers of

24 turns giving a total of 1,392 conductor turns.

Thetotallengthof superconductorused in

each coil is 25,000 m (82,000 ft). A grand total of

50,000 m (164,000 ft or 31 mi) of superconductor is

required to wind the pair of coils.

COIL WINDING

EQUIPMENT

Before the MFTF coils could be wound it was

necessary to design and develop the equipment re

quired. The initial concepts provided by A. R. Har

vey were based on experience winding the baseball

coils and other similar coils. These concepts were

pursued by R. C. Ling, R. E. Hinkle, and EG&G

designers. The winding machine, reel support, con

ductor spools, button dispenser, and cold-welding

process were the first to be developed.

MFTF Coil Winding Machine

The coil winder design was started in August

1976andthe initialdesignspecification49

was

releasedinSeptember1976\. Constructedby

Teledyne Readco, York, PA, the winder in Fig. 25

was delivered to LLNL in September 1977.

The coil winder was designed with maximum

versatility to allow for futurechanges in the coil

geometry. It is capable of winding a two-axis coil

with a major radius from 60 in. (1.5 m) to 130 in.

(3.3 m) with 360° ofrotation. The minor radius can

vary from12 in. (0.3 m) to 60 in. (1.5 m).

The design characteristics50

of the coil winder

are given in Table 9.

FIG. 25.MFTF coil binding machine.

Keel Support

Thereelsupport(Fig. 26) wasdesignedby

EG&GdesignersunderthedirectionofR.C.

Ling5

'andfabricatedby HopperManufacturing,

Bakersfield, CA.

The reel support provides two basic functions.

One is to support the 11-ft reel of superconductor

during winding. The other is to provide constant

tension for the conductor between 50 and 1,000 lb.

A counter-balanced weight system is coupled to a

torque motor with a center sending device that ap

plies more or less torque to the reel to keep it bal

anced with the weights. This system was tested with

a dynamometer and strain gauges and found to be

accurate within ±50 lb. The entire upper assembly

that supports the reel is floated on air bearings to

provide a friction-free motion, and is capable of be

ing elevated and pivoted to keep the conductor in an

ideal winding position.

Other design characteristics\*2

forthe reej sup

port are given in '. jbJe 10.

Button Dispenser

The button dispenser was initially designed and

developed by the LLNL coil shop to apply the inter-

turn insulation buttons to the MFTF test coil. It

was redesigned by EG&G for MFTF winding and

later modified by the winding technicians.

The dispenser (Fig. 27) was designed to apply

the interlurn insulation (buttons) to the conductor's

sidecontinuouslywhiletheconductorisbeing

wound onto the coil form. The basic principle is to

apply one drop of Loctite #414 super glue to each

button, rotate the button, and hold it tightly against

the conductor for a second while the glue sets. Loc

ate #414 was selected after extensive testing.5

-1

Cold Welder

Figure28showsthecold-weldingprocess

adapted for the MFTF conductor after use on the

MFTF test coil.5 4

Purchased from the Heinlz divi

sion of the Kelsey Hayes Co., Phtladephia, PA, the

TABLE 9. MKTF magnet winder characteristics.TABI.K 10.Reel support characteristics.

Parameter

Design load, lb172 X 103

Design torque, in.-lb6 X I 06

Operatingtorque:

No load, in.-lb5.2 X I 06

Full load, in.-lb5.1 X I 06

Rotation:

C\\\and C O V , both

azimuth and elevation axesContinuous

Drive speed:

Arimuth axis, max., rnm0.75

Elevation axis, rpm0.20

Operating temperature.°K32toIltl

Design lire, yr5

Suniial seismic load, g0.25

Power: 4811 V, 3 phase, SO Hz,15(1 1W:

Elevation drive. A150

A/imulh drive. A100

Dimensions:

Height:

No load. Tt23.17

With toad, rt30

Diameter:

Service platform, ft32

Base, ft21

Weight:

No load, lb350 X 10-'

With load, lb522 X 103

welder hydraulically preK

ses the Iwo ends of con

ductor together, causing a cold flow and molecular

bonding of the copper. The Nb-Ti strands do not

bond,butdointermesh.Theresulting jointis

stronger than copper, but not as strong as the com

posite.

TOOLING

Auxiliarytoolingencompassedthreemajor

areas, adaptors for mourning the coil to the winder,

clamping systems to maintain conductor tightness,

andthemiscellaneoustoolingrequiredforcoil

finishing. Where possible, initial designs and con

cepts were fabricated and tried during the construc

tion of the MFTF test coil.

Adaptors

The need for two adaptors was apparent from

the onset of the design: one for holding the coil form

Design load, lb

Design tension, !b

Vcrliclc travel, ft

Vcrlicle speed, in. • min

Pivot (ravel, deg

Pivot speed, deg- min

Reel drive: bidirectional

Operating temperature, °l"

Operating humidity,%

Operating tire, yr

Seismic loud, g

Dimensions:

J.englh,ft

width, rt

Height, ft

Weight: without reel, Ih

Reel sizt-:

o.d.,ft

\ \ idth, ft

Hub diameter, ft

Keel weight: with

9,000 ft of superconductor, lb

12,000

600

•1.5

3.6

±7.2

2.4

Continuous-duty

ac torque motor

32 to HO

12 to 95

0.25

If)

1(1

21.000

11,17

2.33

Hl.5

11.500

while winding and one for mounting the coil-form

adaptor to the coil winder. Figure 25 shows the cuil-

form-to-coil-winder adaptor in place on the winder.

This was designed atLLNLin conjunctionwith

Teledyne Readco5 5

during the winder construction.

The coil-mountingadaptorwas designedby

EG&G5 6

and constructedby F-'MC Corporation,

San Jose, CA. along with the coil form. The design

was to supporta deadweight of 60 tons and a

winding torque of 5,000 fl-lb with a safety factor of

2.5. It had to be easily removable so that minimum

strain would be placed on the finished coil during

disassembly.

Clamping Systems

The need for quick-aciing movable clamps for

winding, and permanently fixed clamps for holding

a layer after winding, was evident from experience

with the baseball coil and the MFTF lest coil.

The clamps were reducedto two types, end

clamps?7

andside clumps.58

-''9

Four end clamps,

shown in Fig. 29, are mounted on each end of the

Soil winding mjwhihe

FIG. 26.MFfF reel support.

FIG. 27.MFTF winding button dispenser.FIG. 28.Cold-welding process.

FIG. 29.End winding clamps.

coil-mounlingadapter.Eachclampispneu

matically operated and capable of being swung out

of the way. Each exerts an axial downward ihrust of

1,500 to 2,500 lb.6 0

There are five sets of side clamps (see Fig. 30)

on each side of the large radius of the coil form.

They use a screw thread with a quick acting spring

detenttoproducefastlateralmovementand

telescope for verticle movement. The side clamps

are designed lo exert a lateral force of 500 lb6 1

holdeachturninpositionduringthewinding.

Upon completion of the layer, the clamps are reset

to hold the first turn of the next layer. To hold the

previous layer in place fixed clamps were required.

ThetowerclampsshowninFig. 31were

designed and constructed by EG&G to support the

sides of the windings. They are stackable I-in.-thick

plates capable of exerting 500 lb of force in two

places by the use of set screws.

PermanentendclampsinFig. 32were

designed6

" :o replace the end winding clamps. They

are 1/2-in.-thick. 316L stainless-steel plates, held in

place by I/2-in.-diambolts on the outside and by

3/8-in.-diam bolls on the inside. The inside bolls are

pan of the weld-backing bar. The entire system of

bolts is torqued lo 40 ft-lb each to replace the ac

cumulated 6.000 lb of end-clamping force.

Tooling tables were designed and constructed

by EG&G to provide a stable support for the tower

clamps and side clamps. This can be seen in Fig. 30.

Miscellaneous tooling covers items that have

beendevelopedbythewindingtechniciansand

LLNL design team. The major item is the routing

fixture (Fig.^}).it is constructed from two sets of

linear ball bushings and shafts that arc mounted to

the coil form to support an air-operatedrouter.6

The prime purposeofthis toolis to machine a

smooth surface on the side of the NEMA G-l I filler

blocks. This machined surface is further sanded to

become a smooth load-bearing surface that trans

mits the electromagnet forces to the slip plane and

subsequently to the structural steel case.

' ^

i^pTffll

1^0\\

FIG.30.Side winding clamps.

INSIXATION

The insulation materials have been successfully

used in previous coils. No new or unusual materials

have been needed to meet the MFTF magnet in

sulating requirements in TableIt.

TABLE 11. MFTF magnet voltages,

Magm-I areaVoh age

(irvundplant-. \ nk- J

l.a>tT-lo-layi-r,V(dc)

Turn-tn-t urn, \\'(dc)

1000

17.3

0.7

The biggest insulationproblem is to prevent

arcingingaseousheliumaspredictedbythe

Paschen curve in Tig. 34 for helium at 20°C and

Iatm.

Intertum Insulation (Buttons)

Thedesignoftheinterturninsulationwas

taken from the Baseball II coil with improvements.

The insulation must be capable of bending in two

directions to conform to the coil geometry: it must

also leave space Tor the liquid helium to circulate.

The button approach that was developedfor the

Baseball II coil depended upon the glue bond to

keep the buttons spaced along tile conductor. The

improved version (Fig. 35) added a Dacron string

attached to the buttons for spacing.

FIG. 31.Tower clamps.

FIG. 32.Permanent end clamps.

The button is an octagonally shaped 0.040-in.-

thickpiece of' iVf:MAC7-Itu ith a groove in The cen

ter. Since the breakdownvoltage6 4

on theN E M A

G-l 1 is 700 V • m i l - ' at room temperature the inter-

turn insulationrequirements were easily met.

BecauseofpoorvenderperformanceL L N L

was forcedtounderiakethe jobofdesigningthe

machinetoproduce the buttonsattherate ofat

least 1.500 ft/day.R. Leber ( M F I f D ) was responsi

ble for the success of this task. The M F T Fwinding

techniciansnowproduce the interturn buttonson

three machines in accordance with the manufactur

ing process as outlined:

Sheet Material.Sheet material of N E M A G-l I

is procured byL L N Land surfaced tomaintain a

uniform 0.040 in. with a 0.002-in. Harness.

Strips.The sheets are sent to a second vendor

to be sheared intostrips 0.437-in. wide. Agroove

FIG. 33.Router guide assembly.

Pressure spacing product (bar-mm)

IlI1IL

1 0\- 61 0\- s1 0\- 41 0\- 31 0\- 2to-1

Density spacing product (gm-cm"3

-mm)

FIG. 34.Paschen curve for helium (20°C).

1\. Dim. t o l . ± 0.015 except as noted.

2\. Cement items 1 to 2 with L L L approved adhesive,

3\. Mat\*) for item 1 tohavesurface sanded to

thickness tol. shown. Edge finish to be clean

sheared or punched.

Spaced 0.725 on centers

\\/TS/T\rr\

\\ //I^-Laminate sheet, /+0

W0.045-rn.thick^'o ^ O " \*

/\t-Woven tape, 0.048 X 0.00? in-7\- \* \-—fbot.ofc

0.437'N

— 0.437

H -1 — 0 . 4 3AH dimensions typical

and in inches

fbot. of groove}

±0.003

Laminate sheet^ \- ^=o=o=o=-=-\\-<\s-\\ryr>

CT^WW

Apply cement' Edge jFinished

Iand tapeInotchiSpacer strip

Ili

Suggested method of production

FIG. 35.MX coil intertani spice strip.

0.050-in. wide and 0.025-in. deep is milled in the ex

act center of each strip, and the strips are sent lo

LLNL.

Square Punching.The strips are inspected for

thickness and groove depth and then run through

the machine in Fig. 36 to punch a square hole in an

evenly spaced pattern.

Applying the String.The punched strips arc fed

intothestringmachine(Fig. 37)whichfeeds

Dacron string into the milled groove and applies a

single drop of Loclite ^MgluL0

-1

to the joint. This

newly formed ribbon of punched strips is wound

onto a plastic spool.

Button Punching.The square-punched ribbon is

run througha second punching operationwhich

removestheremainderofthesquareholeand

chamfers the edges of the remaining center piece.

What remains is an octagonal-shaped button glued

lothestring.Thisisrespooledontoa400-ft-

capacity spool with a ribbon of Mylar between each

layer of buttonsto preventthemfromtwisting.

Figure 38 shows the button punch with some loose

buttons being refastened to the string.

InterIaver Insulation

The interlayer insulation (Fig. 30) is fabricated

in 36-in.-wide by 48-in.-long sheets 1/16-in. thick.

LLNLdrawingNo. AAA78-110166depictsthe

geometryoftheperforations.Nominallythe

material is punched with a 3/ i 6-in. wide X 1.5-in.

long slot spaced 3/16-in. apart.

FIG. 36.Square-punching machine.

These sheets are sand blasted atLLNL and

then cut lo lit the coil form in such a manner that

the slots arc 45° off the vertical centerline of the

conductor.

Giound Plane Insulation

The material selected for the ground plane in

sulation is Kaplon. It was selected because of its ex

cellentbreakdownvoltagesof3,600V-mil\- 1

roomtemperature,whichincreasesto10,800

V-niiHat -I95°C.(1

-S

Table 12 shows the mechan

ical properties of Kapton.

The Kaptonis applied to the coil form in 2-

and4-in.-widestrips(seeFig. 39). Eachstrip is

bonded with 3M =714 glue and placed so it overlaps

the previous one by 50%. A second layer is placed

over the first in the same manner with the over

lapping being 50%. This process is continuedfor

five layers to produce a type of baffled pathway that

is at least 2-in. long for protection from arcing to

ground.

Kapton exhibits an additional characteristic of

beingabletowithstand400°Cwhichgivesad

ditionalprotectionfromdamageduringclosure

welding.

Plenum Chamber

When the magnet is operating the liquid helium

coolant has bubbles of gas for which a paih of es

cape must be provided. To accommodate the gas

bubbles, 3 in. of space was left between the last Wiyer

of conductor and the coil jacket. It is this space that

is referred lo as the plenum chamber.

The 3-in. space has to be filled with a porous

material that is capableo(transmitting loads from

the conductor pack lo the jacket. The material se

lectedwastheinierlayerinsulatingmaterial

previousl}described. The orientationol'the per

forationswas studiedlo determineanoptimum

path for the bubbles. The selected scheme was to

bond four sheets together with the perforations in

line. Thesethenwerebondedloloinsheets of

similarconstructionexceptthattheperforations

were rotated 45-90°. This scheme of four sheets one

way and then four the other was continued until the

entire space was filled.

Slip Plane

The conductormotionanalysis-- shows that

the conductoris going to compress until all the

winding gaps are closed and then move out towards

MS"-''

KIG. 37.String-application machine.

the 1/2-in. jacket. To accommodate this motion a

slip plane has beenprovidedto reduce ihe coef

ficient of friction and provide a sacrificial layer of

material if local deformations are excessive.

Sampleswereproducedthatduplicatedthe

edge construction of the NEMA G-11 glue blocks

and tested inaspecial fixture at both room and liq

uid nitrogen temperature. They were placed against

sample slip planes and loadedwith a controlled

force while a measured force was applied to start the

materials slipping. Table 13 is a summary of the lest

results.

TheMylar-un-Mylarsystemwasselected

because it did not exhibit stick-slip behavior and

avoidedthe uncertainties of the addition cf the

moly-disulfide power into the system.

Glue Blacks and Filler Blocks

The material selected to fill the inside of the

coil from the conductor to the slip plane is NEMA

G-11.

GlueBlocks.The glueblocks(Fig. 33) are

fabricated with l/2-in.-wide X l/32-in.-thick X 40-

in.-longstripsofNEMAG-11. Eachstripis

bonded, on a curved form, to the next strip using

Epoxy 815 and versimide hardncr. These strips are

built up to approximately 0.45-in. thick.

The inside edges of the completed blocks are

chamferedapproximately1/8-in. wide by 1/8-in.

deep, and grooves are cut every inch to provide

cooling along the surface of the conductor.

Once a glue block is bonded to the interlayer

insulation it will provide 5001b-in.-1

of support, in

shear, to hold the layer tight against the coil form.

Once these permanent side damps are installed they

canholdthe conductorin place and the Tower

clamps are removed.

Filler Blocks.The filler blocks are of a similar

design except they are machined to lit the outside of

theconductorpackatthesmallradius.They

provide a solid filler between the conductor and the

jacket.

Felt Filler.To provide for the conductor pack

motion previously discussed, it is necessary to allow

FIG. 38.Button-making machine.

FIG, 39,Coil form No. 1 ground plane installation.

TABLE 12.Properties of Kapton-Iype H.

Typical values, 1 mil film

Property\- I 9 5 ° C25°C200°CTest method

Physical

Ultimate tensile strnegth, psi\*35.IHMI25,00017,000ASTMD-882-64T

Yield point, psi at 3%\*10,0006,0011ASTM D-H82-64T

Stress to produce 5% elongation, psi\*13,0008,50(1ASTM D-HH2-64T

Ultimate elongation,%\*270•MlASTMD-H82-64T

Tensile modulus, psi"510,000430,000260,000ASTMD-882-64T

Impact strength. Kg'cm-mil6Pneumatic inpact test

Folding endurance, cycles0

Tear strength-propagating, g-mil

10.IMHIASTMD-2I76-63TFolding endurance, cycles0

Tear strength-propagating, g-mil8ASTMD-1022-6IT

Tear strength-initial, g - m i l-

'510ASTMIM004-61

Tear strength-initial, I b - i n-

'IIIIIIASTMU-1004-61

Hunting test, psr75ASTMU-774-63T

Density, g-cm-

-1.42ASTM1)-1505-63T

Kinetic coefficient of friction

(film-to-frlm)0.42ASTMD-I8M-63

Refractive index^I.7HKncyclopedic dictionary

Thermal

Melting point

Zero-strength temperature. °C

Cut-through temperature, °C

Coefficient of thermal expansion, in./in./°C

Coefficient of thermal conductivity,

Cal'cm-cm'S-°C

Flammability

Heat scalable

Specific heat, cal-gm"'\*°C

Test condition

None

81520 psi load

for 5 s

Hot harb

4351 mil

2 to 5 mil

Weighted probe o

heated filmb

2.0 Xl ( T5

-14 to +38°CASTMD-696-44

3.72 X\] 0 ~4

25°CModel TC-KI00

3.8» xI 0 ~4

75°Ctwin heatmeter

4.26 XI 0 "4

200o

4.51 XI 0 "4

300°C

Self extinguishing when flame is removed

0.26140°CDifferential calori

275°C300°C400°C

Shrinkage, %0.30.53.030 minASTM D-1204

Heat aging (in air)Syrl y r3 mo12 hCirculating

air oven

Time to reach

1% elongation

\\* M D'Elmemtorf

"Dupont'Graves

Mull:n

SBecke line

TABLE 13. Coefficient of static friction test sum

mary.

RoomLN

Materialtemperaturetemperature

C - l lcinG-1!0.3 lo 0.350 J to 0.35

Mylar/moly/Mylar11.13 to 0.160.13to0.16a

Mylar-on-Mylar0.13 to 0.160.13 to 0.16

('oeflieientofslidingFrictiondroppedto0.99Kithmoly

cuatinj!.

the conductor at the small radius to move radially

outward. This motion is permitted by pJaeini! 3/8

in. of felt between the Kaplon and the filler blocks.

The felt will compress allowing the conductorlo

displace about 0.120 i n "

Filler Material

Upon completion of the coil winding the coil is

built up with the glue blocks, slip plane,Kapton

and NEMA G-llsheets. The jacket, I /2-in.-lhick

3I6L stainless steel, is fitted to the exterior of the

coil pack. Nominally there is l/4-in. of clearance.

Thisvolumeis filled with Epon 815 with chopped

fibers and versimide hardener. Figure 40 shows the

lop filler on the end of the first coil. When this filler

is cured the jacket is welded into place. The filler

FIG. 40.External joint.

providesa lightloadtransmittingmedialoIhe

jacket from the conductor pack.

A washer test was made using the Epon 815

and glass-fiber compoundA large steel washer was

imbedded into a sample of the material and then

thermalcycled10 limesinliquidnitrogen.No

cracking was observed from expansion and contrac

tion. Compressiontests were made in theLLNL

Test Laboratory using the same composition. The

results were10,000 psi to failure in compression,

corresponding to a safety factor of4compared to

Ihe 2500-psi maximum load predicted by analysis."6

WINDING TECHNIQUES

Prior to the commencement of winding the first

MFTF coil a practice coil form was built and 10,000

flofl/2-in.-squarecopperwire waswoundto

debug the winding equipmentand tooling. From

thiseffortadetailedwindingprocedure2

-''was

prepared which is a living document that is main

tained by the engineer in charge of winding.

The firstMFTF coil winding was staried in

February 1979 and completed six months later. Sep

tember 1979. This included unwinding the first six

layers, delays due to slivers in the conductor wrap

ping process, and several weeks of delay due to con

ductor shortages. Winding rates of four layers per

week were achieved during ihe last pari of winding

when the material logistics were corrected.

CONDUCTOR JOINING

Joining two pieces of Nb-Tisuperconductor

together has been a subject of extensive study over

thepastyears.Severalmethodswere testedin

cludingsoftsoldering,silver soldering, andcold

welding.As mentionedearlier,the coldwelding

process was Ihe most satisfactory one. It was used

successfullyfor the MFTF test coil althoughthe

joint is no; as strong as the parent material. It was

decidedtoincreasethecold-weldstrengthand

redundancy by adding additional stabilizing copper

to the sides of the core.6 7

The result of this change

was a machined joint tray into which the core is

soldered after it has been cold welded. Figure 40

shows an external joint completed in the first coil.

The joint is made by stripping back and remov

ing a section of the copper stabilizer, and then cold

welding54

the ends of the core together. The cold-

welded joint is then soldered into the copper joint

tray th:it provides the mechanic;!! strength as well as

the electrical path to assure lhal there is continuity

even if the cold weld were to break. Extensive tensile

testing has been done-^ to verify the design concept.

The joined conductor is stronger than the parent

material, even when no cold weld was made in the

core.

A detailedoperatingprocedure''1

'as well as

detailed check lists have been preparedfor joints.

The finished coil has a total of 5N joint\*-. Twenty-six

are internal joints (against the cod form) and 26 are

external joints as .shown in Fig. 40.

CURRENT LEADS

Two current leads an.\* requited lo hook the coil

to the external power supply. Both leads have been

designed'1

''to meet the worst case conditions; i.e.,

when theI'uiicurrent is carriedii,gaseous helium, if

the superconductor goes normal.

Interna)J.tad

The lead that connects the end of the llrst layer

and travel1

\- \\erlicafly across all ^layers at the cen

ter of the small radius and along the top of layer 58

and finally out the helium exhaust port is called the

internallea (J.Thisleadisconstructedofthree

massive pieces of copper or bus bars.

K\\ferny)Lvad

TIK\* external lead is theonelhal carnes the end

of the lastturnof the lastlayer on throughthe

helium exhau-4 port. It loo is constructed of a heavy

pieceof copperbus bar.

The joining of the conductor to the busr. is

done in a similar manner as the joint tray except

there is no cold weld.

The copper bus bars have cooling channelsfor

free heliumflowand a piece of Nb-Ti supercon

ducting core is soldered into a grove along the sides.

The superconductoris continuous except for two

places where joining the buses made it impractical.

The joining of the copper bus bars was done using

Handyand Harman "Easy How" silver solder. It

has a melting temperature of 1160°F. The supercon

ductor was soft solderedinthegroveswitha50-50

soft solder after the silver-solder joints were made.

Figure 41 shows the current-leadinstallation

forcoilNo. I.LLNLdrawingsAAA79-I07723,

AAA79-107724andAAA79-I0774?describethe

bus bars in detail. Drawing AAA79-107727 shows

the assembly of the lead buses.

FIG. 41.Current lead installation for coil No. I.

The leads are brought out through the helium

exhaustport. Analysis"41

shows a magnetic attrac

tion bejiveen the leads. They are separated by 2.5-

in.-thick NHMAG-M dividers lhal have been glued

together. The divider is locked into place in the ex

haust lube by means of welded guides. The leads are

boiled lo the divider in such a fashion that the bolts

do not go through and are separated by a minimum

2-in. space.

COIL CLOSURE

Figure 42 shows a typical cross section of the

coil at the large radius. The closure is done sequen

tially starting at the small radius ends and progress

ing towards the center.

The general sequence of events in closing the

coil are as follows:

a.install all the NEMA G-l 1 glue blocks,

b.rout the glue block surface,

c.install the weld backing bars,

d.install the slip plane,

c.install the plenum chamber,

f.complete the ground-plane Kaplon insula

tion,

g.fillthe remaining volume to withinl/4-in.

of the jacket with solid NEMA G-llmaterial.

hfillthe remaining1/4-in. with fiber-filled

Epon 815,

i.install the jacket section and tack weld in

place, and

j .continue the process working from the coil

ends to the center.

3 1 6 LCrescoil

jacket

Fiber-filledF.pon815

1/16N E M AG-11

Kapton

G-11 perforatedplenum

chamber

\- G - 1 1interturninsulation

-Interlayerinsulation

-G-11filler

\- Kanton

KIG. 42.Large radius—typical cross section.

INSTRUMENTATION

Strain gauges have been installed on the coil

formpriortowindingaspartofihcGeneral

DynamicsInstrumentationPlan.7 0

Additional

strain gauges were installed on conductors at layer 2

and26 tomonitortheconductormotionsand

strains.4 7

Other strain gauges were added to the under

side of the coil form to monitor the coil form strains

duringwinding. Voltagelaps7 1

were installedas

parto\[Ihc coil closure to monitor the voltages at\
\
variouslayers of thecoilduringoperation.Ad\
\
ditional voltaic laps have been added lo the current\
\
leads for similar reasons.\
\
QUALITY ASSURANCE\
\
ThebasicQualit}Assuranceprogramfor\
\
winding is described in the Magnet System Quality\
\
Assurance Pian (M-078-06-01). This plan describes\
\
i!.e type of records lo mai.'Uain as well as respon\
\
sibilities oT individuals.\
\
The detailed Qualit}Assurancerequirements\
\
for the winding of the coils arc given as p:irl of the\
\
lesl of the Ml-TF Coil Winding Procedure (MKL-\
\
78-001432). This defines the responsibilities as well\
\
as ihe forms used lo report the measurements and\
\
inspections made during the winding operation.\
\
Several inspections are routinely made al Ihe\
\
completion of winding a la\\er:\
\
a.visual inspectionfor debris, flatness, and\
\
any protruding insulation;\
\
b.electricalresistance andHi-Pol measure\
\
ments to detect any shorting or debris that may ex\
\
ist:\
\
c.winding gap measurements at four points,\
\
measuring ihc stack height and width; and\
\
d.joint inspection for cleanliner-s.\
\
All anomalies are corrected before continuing\
\
f\
\
o wind the next layer.\
\
Check lists are used for each joint to assure that\
\
no detail is overlooked in the process. Each step of\
\
the procedure is signed off by the operator, and the\
\
final li.-a is suied by the shiftsupervisor.\
\
Operatingproceduresarespecifiedforeach\
\
oper 'lion. These are referenced in the coil winding\
\
procedure and all are available on the winding plat\
\
form .\
\
Shiftlog books and photographic records arc\
\
maintained lo re ord problems and progress. Each\
\
shift supervisor completes a new page each shift.\
\
All of the data is reviewed daily by the engineer\
\
in chargeof winding. Anydiscrepanciesare re\
\
cordedt)iiaNonconformanceReportwhichis\
\
reviewed hy a Materials Review Board that decides\
\
uponthe corrective actionordispositionofthe\
\
discrepancy.\
\
40\
\
SECTION 5\
\
THERMAL ANALYSIS\
\
The cooldown and warmup thermal analyses of\
\
theLawrenceLivermoreNationalLaboratory\
\
(LLNL)MirrorFusionTestFacility(MFTF)\
\
magnet system included investigationof a broad\
\
rangeofflowratesandsupplytemperature\
\
schedules. At the onset of these studies there were\
\
several objectives or design goals: (1) to cool down\
\
and warm up the magnet system within three to five\
\
days; (2) to yield acceptable levels of thermally in\
\
ducedstressesresultingfromtransverseand\
\
longitudinaltemperaturedifferentials;and (3) to\
\
yield acceptable stress levels with or without flow\
\
imbalances in separate sections of the magnet. All\
\
of these analysis objectives were met.\
\
Details of the initial studies are contained in\
\
Ref. 72. As the design evolved and the interplay be\
\
tweenthermodynamicandstructuralanalysis\
\
becamebetterunderstood,aniterationof the\
\
detailed cooldown and warmup analyses was per\
\
formed.Thisanalysisparticularlyemphasized\
\
longitudinaltemperaturevariationsand is con\
\
tained in Ref. 73. This document is a condensation\
\
andintegration of the results of those two prior\
\
studies.\
\
COOLDOWN AND WARMUP\
\
BASIC ASSUMPTIONS\
\
Cooldown and warmup of the MFTF magnet\
\
is achieved by means of helium through-flow in two\
\
separate, parallel-flowpassages: one through the\
\
conductor bundle, and the other through the guard\
\
vacuum.A massflowfractionfor eachof the\
\
parallel flows was apportioned on the basis of the\
\
relative mass to be cooled or warmed, 19% of the\
\
total flow to the conductor, and 81% to the guard\
\
vacuum. The latter flow split was established early\
\
in the program and maintained throughout. Subse\
\
quentincorporationofadditionalmassto the\
\
magnet case in the form of stiffeners did not alter\
\
this recommendation.\
\
The magnet assembly was modeled numerically\
\
(Fig. 43) in terms of successive arrays of block-type\
\
Conductor\
\
helium, 19%\
\
Large model © ,\
\
644 nodes\
\
Impose boundary\
\
temps TBfrom small\
\
models\
\
Typical section nodes\
\
Large model © ,\
\
661 nodes\
\
\- Intercoil, guard vacuum helium, 81%\
\
Plan as viewed from below magnet\
\
FIG. 43.Cooldown and warmup analysis large models (1) and (2) address transverse thermal gradients.\
\
41\
\
nodes(i.e.,rectangularparallelepipeds)repre\
\
sentingelementalmassesofconductorregion,\
\
potting, jacket, case, coil-extension structure, and\
\
intercoilstructure.Theblock-typenodeswere\
\
providedwithconvectivelinkagestoadjacent\
\
helium-flownodes,whichwereordereddirec-\
\
tionally to appropriately model the helium through-\
\
flow. Convective linkages between flow nodes and\
\
surfacesofconductornodeswerebasedona\
\
laminar flow Nusselt number of 4.0, aneffective\
\
hydraulic diam of 0.131 cm, and a surface area per\
\
conductor of 8.17 cm2\
\
/cm. Each conductor region\
\
node contains 348 conductors.\
\
Guard-vacuum-spaceheliumflowincludes\
\
flow between the jacket and the case, and between\
\
theguard-vacuumbaffleandthecase.Flow\
\
passages ofl/8-in. depth in the jacket cavity and\
\
1/4-in. depth in the baffled cavity cover 50% of the\
\
case inner surface, and a laminar Nusselt number of\
\
4.0 was again assumed. Helium flow passes through\
\
the inlercoil and coil extension structures prior to\
\
entering the guard-vacuum space. For the intercoil\
\
andcoilextensionsurfaces,aturbulent-flow\
\
natural-convection Nusselt number was computed:\
\
NN l l= 0.13(GrPr)1 / 3\
\
.(1)\
\
Theflowratesemployedinthecooldownand\
\
warmupanalyses,upto340 g/s,representeda\
\
preliminary value supplied by LLNL. Supply tem\
\
peraturesfor flow schedules were assumed to be\
\
controlledon the basis of a measured case tem\
\
perature located at the lower minor radius. Local\
\
flow rates contacting each conveclively cooled or\
\
heated node were apportioned based on ratios of\
\
local-to-lotal cross-sectional flow area.\
\
Effectivepropertiesofthe conductorregion\
\
nudes were computed by an independent analysis,\
\
in which mass-weighted average specific heats were\
\
generatedasfunctionsoftemperature,and\
\
series/parallelconductivelinkageswere resolved\
\
into directional thermal conductivities, also as func\
\
tions of temperature. Appropriate heat and conduc\
\
tivity data for 316 SS. 304 SS, and NEMAG-ll\
\
fiberglass epoxy were obtained from the exit near\
\
thetopmostlocationofthecoil jacket.These\
\
several flow paths and splits were simulated in both\
\
large and small analysis models. Figure 44 show? a\
\
worst-caseflow-rateschedulewhere340 g/sof\
\
heliumentersthemagnetduringthetimethat\
\
return temperature is 300 K and varies linearly to\
\
150 g/swithreturntemperaturedecreasingto\
\
100 Kandremainsat150 g/sforreturntem\
\
peratures decreasing to 4.5 K. The accompanying\
\
inlettemperatureprofileconsistedoftwo steps:\
\
80 K for return temperature >100 K and 4.5 K for\
\
return temperature < 100 K.\
\
Because the external case sliffeners of Large\
\
Model 1 do not have a direct convective linkage to\
\
the cooldown/warmuphelium, it was anticipated\
\
that they would sustain the most severe transverse\
\
temperaturedifferentials.LargeModel 1 was\
\
thereforerunwiththeseverecooldownflow\
\
schedule of340 g/sand136 g/s at300 K return\
\
temperatureto assess the effectsof flosv rate on\
\
transverse differentials. Large Model 2 was run only\
\
with the cooldown flow schedule with 136 g/s initial\
\
flow rate. The small model was run with these as\
\
well as several other cooldown and warmup flow\
\
schedules including constant flow rates.\
\
ANALYSIS RESULTS\
\
Figure 45 shows resulting transversedifferen\
\
tials and cooldown duration as a functionof the\
\
flowratewith300 Kreturntemperature.This\
\
severe flowrateandinlet temperature profileis\
\
responsible for the large temperature differentials.\
\
Structural analysis of this maximum differential be\
\
tween the stiffencr. which is not directly cooled, and\
\
therestofthemagnetwhichis"wetted"by\
\
cooldown helium gas, shows that the temperature\
\
differentialup to180 g/s yields acceptable stress\
\
levels. This corresponds to a minimum cooldown\
\
duration of 44 h.\
\
Figure46showsthedetailedtemperature\
\
history data during cooldown with an initial flow\
\
rateof136 g/sandtwo-stepinlettemperature\
\
profiles. These data are for the cross section midway\
\
in the majorradius section of the magnet where\
\
maximum transverse gradients occur. The "wetted"\
\
portions of the magnet exhibit temperatures in a\
\
narrow band during cooldown. The stiffener, which\
\
is not directly cooled, remains warmer than the case\
\
temperatures.In this case, the stiffeneris about\
\
95 K warmer (max) than the rest of the magnet and\
\
it has been determined that this condition is accept\
\
able by structural analyses.\
\
In contrast.Fig. 47 shows similar transverse\
\
gradienttemperaturehistoriesforthetwo-\
\
temperature step cooldown with maximum initial\
\
42\
\
Intercoil,\
\
node 2 - ,\
\
Coil jacket, and case nodes,\
\
X44\
\
Return\
\
temp (K)\
\
300\
\
100\
\
4.5\
\
Maximum\
\
He (g/s)\
\
340\
\
150\
\
150\
\
Guard vacuum\
\
helium entry\
\
Intercoil,\
\
node 1\
\
Coil extension\
\
nodes, 6 X 2\
\
•Each node has a\
\
convective linkage to a\
\
helium flow node.\
\
FIG. 44.Cooldownandwarmupanalysissmallmodeladdresseslongitudinalthermalgradientsand\
\
cooldown/warmupdurations.\
\
70I'I'I'I'\
\
\\* Flow rate varies linearly-260\
\
to 150 g/s at 100 K\
\
return temp., constant240x.\
\
. C60\_at 1SO g/s to 4.5K.\
\
• Inlet temp, is 80K,-220\
\
1O)\4.5 K at sensor temp.2001E\< 1 0 0 K .2E50^ v• Sensor is c»se temp, at180<3\
\
g\ .\_Y\
\
small radius.160!\
\
o40—^ S \^v\"00\
\
\\\\\\*140\
\
1201"o\
\
o\
\
40\
\
^^^>\
120\
\
ans'\
\
CJ30\\\^\—Maximum^""'""100H\
\
20\
\
-allowable transverse AT\
\
Ii,1,1,1,80\
\
150200250300350\
\
Flow rate at 300 K return temperature (g/s)\
\
PIG. 45.Cooldown flow rate at 300 K return tem\
\
peraturemust not exceed180 g/sforallowable\
\
transverse gradient.\
\
flow rate of 340 g/s. The wide and increasing depar\
\
ture of the sliffener temperature Prom the more uni\
\
formly cooling case temperatures is evident and is\
\
so large as to be unacceptable structurally. This is a\
\
symptom of the extremely rapid cooldown with this\
\
high flow rate.\
\
Figure48 showsthecorresponding340 g s\
\
(high flow) longitudinal temperature gradients in\
\
duced in the lower leg of the yin or yang magnet\
\
from the coolant inlet to the exit region. Duration\
\
into the cooldown is the time parameter. These data\
\
are from the literature. The latter data were em\
\
ployedinthecooldown/warmupmodelforthe\
\
magnet jacket, case, and potting material, respec\
\
tively.Dataforoxygen-freehigh-conductivity\
\
(OFHC) copper was taken from Ref. 74. Reference\
\
75 was employed for helium properties, and N EM A\
\
G-l I epoxypropertieswere takenfromRef. 76.\
\
43\
\
300\
\
250 -\
\
200 -\
\
\*150\
\
100\
\
1012141618\
\
Cooldown time (h)\
\
FIG. 46.Cooldown initial flow rateof136g/s results in acceptable transverse gradients.\
\
300\
\
250\
\
200\
\
K150\
\
100-\
\
01.01.52.02.53.03.54.04.55.0\
\
Cooldown time (h)\
\
FIG. 47.Maximum cooldown flow rate (340 g/s) causes severe transverse temperature gradients.\
\
44\
\
Major\
\
radius\
\
Minor\
\
radius\
\
300\
\
250-\
\
200-\
\
m150\
\
100\
\
1015202530\
\
Distance from inlet(ft\|\
\
FIG. 48.Cooldown is 28 h with 340 g/s initial (low rate\
\
35\
\
Properties of 304L and 316 SS were obtained from\
\
Refs. 77 and 78.\
\
COOLDOWN AND WARMUP\
\
ANALYSIS MODELS\
\
The cooldownand warmup thermalanalysis\
\
modelswereformulatedforanimproved,un\
\
published version of the Convair thermal analyzer\
\
computerprogram: the originalversionis docu\
\
mentedinRef. 79.Therearetwoprogram\
\
modifications affecting the analyses of this report:\
\
(1)provisionforaccepting"block-type"node\
\
representation, and (2) optional explicit,forward-\
\
marching solution of the heat balance equations.\
\
Thelatteroptionwasemployedincooldown/\
\
warmup analysis. All analysis runs were executed\
\
on the National Magnet Fusion Energy Computer\
\
Center (NMFHCC) at LLNL.\
\
Two major models were used during the course\
\
ofthemagnetthermalanalysis.Athree-\
\
dimensional model (shown schematically in Fig. 43)\
\
wasdevelopedprimarilytoexaminetransverse\
\
gradients: this modelalsoproducedlongitudinal\
\
gradients. This model was so large it was deemed\
\
economical to break it down into two models: Large\
\
Model I (644 nodes) in the major radius zone of the\
\
magnet,andLargeModel2 (661nodes) inthe\
\
minorradius, coil-extensionzone of the magnet.\
\
For the large model, the cross section was nodalized\
\
in block-node form, as shown in the cross section in\
\
Fig. 43. so that iransverse temperature differentials\
\
across the magnet cross section could be predicted.\
\
A smaller, essentially one-dimensional model\
\
of a total yin or yang magnet (Fig. 44) provided\
\
longitudinaltemperaturedistributions andboun\
\
daryconditionstoinitiateanddrivethelarge\
\
models of Fig. 43. All the transverse nodalizalion of\
\
the large models were collapsed mathematically into\
\
oneiransversenodeateachstationalongthe\
\
magnet resulting in 44 longitudinally-arrayed nodes\
\
around the entire magnet. This model, having much\
\
fewer nodes, (total of 119), in addition to providing\
\
theboundaryconditionsforthelargemodel,\
\
providedan economicalsolutionforlongitudinal\
\
gradients, cooldown durations, and thermal effects\
\
of the unbalanced How split into the two legs of the\
\
maiznel.\
\
45\
\
Figures43 and44 showtheguard-vacuum\
\
helium (81 % of Ihe total), entering the lower coil ex\
\
tension, flowing through it, and then entering the\
\
guard-vacuumcavitieswhichsurroundthecoil\
\
jacket. While flowing through the coil extension,\
\
57Tt of the guard-vacuumheliumis diverted by\
\
relative hole sizes into and out of the intercoil struc\
\
ture to condition its mass. At the top of the magnet\
\
the guard-vacuum helium similarly flows through\
\
the upper coil extension before exiting the magnet.\
\
The coil helium (19% of the total) enters the magnet\
\
ir;arthe low point and flows up each leg of the\
\
winding to join at the top outlet. Magnet cooldown\
\
was analysed using the small model with transverse\
\
nodes collapsed into a single node: thus at any sta\
\
tionalongthemagnet,thenoderepresentsa\
\
weightedaveragetemperature.Withrapid\
\
cooldown in 28 h. very large longitudinal gradients\
\
and bottom-to-top differentials occurring at about\
\
12 h(.XT=275 K)areevident.Becausethe\
\
longitudinaltemperaturedifferentialcouldbe\
\
driven to such a large value, effortwas thereafter\
\
concentrated on reducing the longitudinal differen\
\
tial.\
\
Since longitudinal differential was known to be\
\
less a function of flow rate than of severity of inlet\
\
temperatureschedule, a mitigatedfour-stepinlet\
\
temperatureprofileforcooldownwasselected.\
\
Figure 49 shows the schedule, the cooldown dura-\
\
lion,andthebottom-to-loplongitudinaltem\
\
perature differentialas functions of constant flow\
\
300\
\
280\
\
260\
\
240\
\
220\
\
200\
\
180\
\
160\
\
140\
\
•Flow rate is constant\
\
1601 ' 1 ' ! ' 1 ' 1\
\
1\
\
1 ' 1' 1\
\
150\
\
140\
\
130\
\
120\
\
-\
\
Inlet tempschedule\
\
i l l\
\
150\
\
140\
\
130\
\
120\
\
-Cooldown\
\
time (h)He supply\
\
temp (K)\
\
i l l\
\
150\
\
140\
\
130\
\
120\
\
-0-18\
\
18-36\
\
36-54\
\
54-100\
\
225\
\
150\
\
75\
\
4.5\
\
i l l\
\
110-\
\
100-\> .iX ,-\
\
90^ • " s . 1-\
\
80\
\
7 f 11 i 1 i1 i !\| i 1• h i > t—\
\
90110130150170->\
\
Helium flow rate (g/s)\
\
FIG. 49.Longitudinal 'temperature differences are\
\
acceptable for 3.6- toS-itycooldown durations.\
\
rates.Thedataconfirmsthelongitudinaltem\
\
perature differentialto be relatively insensitive to\
\
flow rate and to decrease with higher flow rates.\
\
The structuralanalysis was performedfor a\
\
119 g/sflowratethatcorrespondstoa195 K\
\
longitudinal temperature difference and cooldown\
\
in88 handwascertifiedasacceptablewith\
\
moderatestresses.Thelongitudinal-temperature\
\
differentialscanbeacceptedfor3.6-to5-day\
\
cooldowndurations with the four-step inlet tem\
\
peratureschedule.Figure50showsthecorre\
\
sponding detailed magnet longitudinal-temperature\
\
profiles from bottom-to-top with cooldown time as\
\
a parameter. The acceptable maximum longitudinal\
\
differentialof 195 K occurs at about 63 h into the\
\
cooldown.\
\
Figure51 showswarmuptimeandmagnet\
\
longitudinal-temperature differential as functions of\
\
constant warmup flow rate. Warmup is considered\
\
to be completed as an accesible condition at 285 K.\
\
Again, the four-step inlet temperature schedule was\
\
used and is defined in the figure. Because of the less\
\
effective use of the heat capacity of the helium by\
\
this warming profile relative to that of the cooling\
\
profile, this figure shows that the warmup is slower\
\
than thecooldown.Themaximumlongitudinal dif\
\
ferentialsare milder (and, againrelatively insen\
\
sitive toflowrate) andthe transve:segradients\
\
would also be milder. The figure shows warmup can\
\
be achieved in 122 h (approximately 5 days) with a\
\
constantflowrateof150 g/s. All of these mild\
\
longitudinal-temperature differentials during warm-\
\
up havebeendeemedacceptable structurallyby\
\
comparison to the larger differentials of cooldown.\
\
Figure52showsthemagnettemperatures\
\
versus distancefromthe conditioningflowinlet\
\
during warmup with 150 g/s flow and the four-step\
\
inlet temperature schedule. Time into the warmup is\
\
shown as the parameter. The maximum differential\
\
of 120 K occurs about 65 h into the warmup.\
\
Theeffectsontransversegradientsand\
\
longitudinal-temperature differentials of the yin or\
\
yang magnet, or between the two magnets, may be\
\
assessed structurally. The temperature data could\
\
be taken in detail from Refs. 89 and 90. However.\
\
the gradients in the two legs of a magnet or be\
\
tween magnets are essentially independent one to\
\
the other, depending only on the different flow rates\
\
in those individual paths. Therefore,Figs. 45. 49,\
\
and51indicatethedifferencesintemperature\
\
gradientand differentialsfor differentflow rates\
\
46\
\
Major\
\
radius\
\
Minor\
\
radius\
\
152025\
\
Distance from inlet (ft)\
\
FIG. 50.Longitudinal temperature gradients for 119 g/s flow rale with 4-stfycooldown inlet temperature\
\
schedule were structurally evaluated to be acceptable.\
\
• Flow rate is constant\
\
• Warmup is to 285 K\
\
260III!III?nn\
\
240Inlet temp.\
\
WarmupHe supplyi1FtO\
\
220time Ih)temp IK)\
\
220O 18\
\
18 3675\
\
150~16Q\
\
?nn\_36-542 2 5\
\
300S40\
\
180=\_\_\_^120\
\
1fi(1\_\ w\
\
100\
\
140\
\
i ? nIIIIl > - - L\
\
90 100110 120130140150\
\
Helium flow rate (g/s)\
\
FIG. 51.Warmup in 3 to 5 days yields mild longi\
\
tudinal gradients.\
\
(unb>iur.ced How splits) into the differentportions\
\
ofthemagnetsystem.Structuralanalysisfora\
\
severe flow imbalance in the separate legs of a single\
\
magnet in which the helium How rate in one leg was\
\
WJ-greater than that in the other leg results in mild\
\
and acceptable stresses.\
\
CONCLUSIONS AND\
\
RECOMMENDATIONS\
\
1.Cooldownanalysesemployingthelarge\
\
model (Fig. 43) yielded detailed temperaturexcur-\
\
sion and transverse temperature differentialdata.\
\
Resulting thermal stress?s are acceptable to a max\
\
imum flow rale uf 180g/s at 80 K inlet and 300 K\
\
returntemperatures.Resultingcooldowntime is\
\
44 h.\
\
2.Longitudinal temperature differences em\
\
ploying the small model (Fig. 44) have been struc\
\
turally certifiedacceptable at a 119 g/sflowrate\
\
47\
\
Major\
\
radius\
\
3 0 0\
\
250\
\
\*200-\
\
E150\
\
100\
\
152 025\
\
Distance from inlet(ft)\
\
FIG. 52.Warmup to accessible conditions is achieved in 5 days with 150 g/s flow rate and 4-step inlet tem\
\
perature.\
\
willl llic four-step inlcl temperature profile achiev\
\
ingconldownin88 h.Reductionoflongitudinal\
\
gradients with higher flowrates also implies accept\
\
abilityforhigherflowrates.\
\
.1.Withthefollowingfour-stepinlettem\
\
perature schedule:\
\
CooldownHel urn supply\
\
time (h)temperature (K)\
\
0 to 18225\
\
18 to 36150\
\
36 to 5475\
\
>544.5\
\
Total How rates to I 19 g/s and cooldown tn 3.6 to 5\
\
d;;vsaccommodatL'stressconstraintsandare\
\
acceptable.\
\
4.Warmup with the following four-step sup\
\
ply temperaturemanagement schedule:\
\
VVarmupHe ium supply\
\
time (h)temperature (K)\
\
0 to 1875\
\
IS to 36150\
\
36 to 54225\
\
>54300\
\
Totalflowrateo\\'150 g/s yields mild and accept\
\
able temperature gradients. Warmup time to 2S5 K\
\
is122 h (approximately5 days).\
\
5.Sincemagnetgradientsareinfluenced\
\
morebyreducedseverityofinlettemperature\
\
schedule than by flow rale, gradually decreasing (in\
\
creasing) orfinerstep-wise decreasing (increasing)\
\
heliumsupplytemperaturescouldpermiteven\
\
higher How rales and faster conldown (uarmup). If\
\
deemedexpedienttheseoptionswouldrequire\
\
furtherstudyandthermodynamie/strueuiral\
\
analvsis certification.\
\
48\
\
SECTION 6\
\
CRYOGENIC SYSTEM\
\
The cryogenic system for the MFTF magnet\
\
provides liquid helium to the magnet coils to main\
\
tain a superconducting stale. Several interfaces and\
\
effecls must be accounled for in ihe magnet system\
\
designtoachievethiscondition.Moreover,the\
\
magnet system must be able to survive numerous\
\
rapid discharge cycles without detrimentaleffects.\
\
These topics are reviewed in the following text.\
\
INTERFACES\
\
The magnet will be supplied wilh liquid helium\
\
throughdedicatedlinesrunningbetweenme\
\
vacuum vessel ports and ihe magnet. As illustrated\
\
in Fig. 53, each magnel coil has separate supply and\
\
return lines.\
\
The magnet assembly will be supported by two\
\
hanger and five stabilizer struts. Heat conduction in\
\
these slruts is minimized by liquid nitrogen-cooled\
\
barriers.\
\
Liquid niliogen-cooled heat shields will com\
\
pletely envelope the magnets for thermal radiation\
\
protection. Figure 54 shows these shields, wilh ad\
\
ditionalwater-cooledshieldsin critical areas, lo\
\
protect the nitrogen shields from beam and plasma\
\
heating.\
\
Anotherinterfacewiththemagnetisthe\
\
current leads penetrating the vacuumvessel wall,\
\
where the lead temperature must make a 5 lo 300 K\
\
transition. Also, the temperature and strain .sensors\
\
on the magnet will be a source of heat transfer lo the\
\
liquid helium.\
\
LIQUID HELIUM\
\
SYSTEM\
\
The piping system for supplying liquid helium\
\
lo ihe magnet is illustrated in Fig. 55. Each magnel\
\
coil has dedicated supply and return lines and valves\
\
exceptattheheliumDewarpenetrations.Flow\
\
through the magnet is by natural convection with\
\
liquid entering the bottom of ihe magnet, splitting\
\
at the bottom of each coil, (lowing up each half coil\
\
and rejoining beforeexiting out at the top. The prin\
\
cipal reason that the yin-yang pair is oriented ai 45°\
\
is so ihe helium will flow in this manner.\
\
Important to the thermal control of the magnet\
\
is an adequate circulation of LHe. Forced pumping\
\
is not practical, so natural circulation was chosen.\
\
because it has been satisfactory in smaller magnet\
\
systems. A computer modelofihe LHe natural cir\
\
culationwasdevelopedtoestimatesteady-state\
\
\*^ \- V a c u u mvessel\
\
/CurrentleadsHeo u t l e t\
\
L Nbarrier\
\
1—.-\
\
-;Hangar\
\
'•\|strut\
\
~X;\
\
""!;.-.Plasma^\
\
•—-i center \*\
\
Stabili7f?f'\
\
strutHeinlet\
\
,(H e l i u mo u t ' e t\
\
V a c u u mvessel\
\
FIG.53,Schematic of magnets in the uicuum vessel\
\
showing LHe lines, support struts, and current leads.\
\
49\
\
Neutral beam shield\
\
• Water liner\
\
LN-cooled bridge\
\
• LN shieldVacuum\
\
communication port\
\
\- Bladder\
\
-Stancoff\
\
-Case\
\
• Guard vacuum\
\
-LN shield\
\
• Water-cooled\
\
plasma shield\
\
IIC. 54.Sectional ("A-A") view of l.ie magnet showing its construction and thermal shields.\
\
mass llov. rale and vapor quality. Also, a sensitivity\
\
studywas madeto delerminewhichparan eters\
\
have the greatest influence and to estimate the range\
\
of uncertainty for LHe flow rale.\
\
The principal requirement of the LHe system is\
\
1.0 maintain quality with less lhan 10 vol% va\[ or in\
\
themagnel.Heat-transferanalysesindicatethat\
\
signilicantly higher vapor qualities would probably\
\
reduce cr>oslabllll>.\
\
The conductor pack and magnet shape do not\
\
immediatelylendthemselvestoasimpleflow-\
\
modeling approach. Therefore, three approaches to\
\
modelingIhemagnetwereconsidered:(I)the\
\
Blasius friction equation. (2) Darcy's porous media\
\
equation, and (3) u three-dimensional orifice model.\
\
In selecting an appropriatemethod, estimates of\
\
hydraulicdiam,flowtortuosity,porosity,per\
\
meability,frictionfactor,andeffectiveorifice\
\
dimensions were made, and the three approaches\
\
werecomparedbymeansoftheirrespective\
\
pressuredrops.Theporousmediaapproach\
\
resulted in the smallest pressure drop, the Blasius\
\
approach yielded a pressure drop 10 limes greater,\
\
and the orifice approach gave a pressure drop that\
\
50\
\
\- LHe supply Dewar\
\
\- \- \- -, \- —27 m1\
\
" Return valve\
\
T?\
\
zi''Rupture disK-\
\
\- Vent valve\
\
Helium\
\
transfer lines\
\
9 m\
\
j"Magnet\
\
7 m\
\
3 m\
\
Vacuum vessel\
\
FIG. 55.Schematic of (he piping systems for sup\
\
plying LHe to the magnets.\
\
was1000 timesgreater.We rejectedtheorifice\
\
assumptionas unrealistic, andchose theBlasius\
\
frictionmethod over the porous media approach\
\
because it was more conservative. Thus, the magnet\
\
pressure loss \_iP was estimatedwithamodified\
\
Blasius friction equation in the following form:\
\
A P = ( lm, dhM - 7 2 r A 2 02(2)\
\
Flow in the magnet is expected to be laminar\
\
by the Reynolds number definition, so that the fric\
\
tion factor (f) is given as:\
\
procedure provided a means for determining the ef\
\
fect of heat load on flow rate (Fig. 56a). Also, vapor\
\
quality was determined and appears as a function of\
\
heat load in Fig. 56b.\
\
Because the modeling procedure entails some\
\
uncertainties, we were interested in how sensitive\
\
theresults were to changesin certainvariables.\
\
Tabic 14 shows the maximumexpectedrange in\
\
these flow parameters, and Table 15 shows our es\
\
timate forthe totalLHe-systemheat input. The\
\
resultsofthesensitivitystudyarereflectedin\
\
Figs. 56a and 56b by the uncertainty range curves.\
\
The effects of two-phase flow are significant for\
\
helium mass qualities as low as only 1% and flows\
\
near 300 g/s. Surprisingly, the static head inside the\
\
LHe Dewar is the most significantparameteraf\
\
fecting mass flow rate because of the relatively low\
\
fluid-flow resistance of LHe. Also, pipe friction is\
\
far more influential on flow than either magnet fric\
\
tion or heat input.\
\
Onthe basis of this calculation we selected a\
\
pipe with a 6-in. (15-cm) inside diameterfor the\
\
LHe supply and return lines. This pipe size and our\
\
estimatedsystemheatloadof510 W yields an\
\
equilibriummassflowrateofapproximately\
\
750 g/s. Vapor quality at the top of the magnet is\
\
less than 5 vol% (0.7% by mass), and it is less than\
\
25 vol% at the top of the return line. These results\
\
implythatanadequatesafetymarginhasbeen\
\
. provided in the thermal control of the magnet.\
\
f = 64M Ap0/i) Mdh(3)\
\
Pressure losses in the piping system were sim\
\
plymodeledusinglosscoefficientsforbends,\
\
valves,entrance-exitregions,andothereffects.\
\
These were calculated as functions of the friction\
\
factor f. The effect of two-phase flow was also in\
\
cludedbyusingtheLockharlandMartenelli\
\
correlation: flow in the piping system was assumed\
\
turbulent.\
\
Heliumflowrateswereestimatedbyan\
\
iterative computing method. Using an assigned heat\
\
loadforselectedflowmodelelementsandan\
\
assumed How rale, helium properties were deter\
\
mined by each element node using NBS data. The\
\
resulting pressure imbalance in the flow circuit due\
\
to cumulative contributions of friction, momentdi,\
\
and gravity was computed. The flow rate was read\
\
justedandthecalculationsrepeateduntilthe\
\
pressureimbalancewasacceptablysmall.This\
\
>200300400500600700\
\
Heat load (W)\
\
FIG. 56.Steady-stateLHenaturalcirculation\
\
through the magnets.\
\
51\
\
TABLE 14. Uncertainly range of flow parameters.\
\
ParameterRange\
\
0.56 m\
\
0.95 m\
\
Friction factor, f\
\
Tortuosity factor, y\
\
Porosity,O\
\
LHe Dcwar head, tn\
\
Hydraulic diameter ( dn) , cm\
\
Piping heat leakage, W\
\
Length of piping, m\
\
0.014 to 0.018\
\
1.5 to 2.0\
\
0.30 to 0.35\
\
0.5 to1.0\
\
0.15tu0.35\
\
31) to 80\
\
SO tn HHI\
\
TABLE 15. Liquid-helium-system heat sources.\
\
Heat input.\
\
Parameter\\V\
\
i N shietd radiation\
\
LN shield conduction\
\
Magnet hanger rods\
\
Conductor joints\
\
Instrumentation leads\
\
Helium ducts\
\
Total\
\
Curint leads. ma\\.vent rate\
\
160\
\
90\
\
60\
\
45\
\
70\
\
85\
\
510\
\
1.6 K/s\
\
A plenum space is included at the top and bot\
\
tom of each coi' near the supply and outlet ports, as\
\
illustrated in Fig. 57. These plenums are 6-cm thick\
\
and constructed of 6.5-mm-lhick layers of NEMA\
\
G-l 1 having 5- by 37-mm slots alternately oriented\
\
90° to each other. They provide a 50% bearing sur\
\
face for the conductor and a 0.5 porosity for helium\
\
How.Both plenums distribute flow enteringand\
\
leaving the coil, and the lop plenum also provides a\
\
space for vapor to flow outside the coil so the con\
\
ductor will always be liquid-cooled.\
\
A detail of the helium outlet from the conduc\
\
tor packis shownin Fig. 58. Currentleads con\
\
tained in these ducts are wrapped with Kapton to\
\
prevent voltage breakdown to the conductor pack\
\
and surrounding structure. The supply duct at the\
\
bottom of each coil is identical except for the ab\
\
sence of the current leads.\
\
Outermost turns of the conductor in the out\
\
side flat areas in the large-radius regions of each coil\
\
aresupportedby beveledNEMAG-11" backing\
\
blocks. These blocks support magnetic pressures of\
\
nearly 20 MPa(2900 psi) andare beveled, as il\
\
lustrated in Fig. 59, so that vapor can migrate up\
\
wards through (he channels to the plenum.\
\
2.91 m\
\
Plenum\
\
Helium supply\
\
FIG. 57.Jacketaroundcoilshowingdimensions\
\
and locations of helium plenums.\
\
CRYOSTABILITY\
\
The MFTF superconductor is cryostable from\
\
experimental and analytical studies. The High Field\
\
Test Facility (HFTF) at LLNL has demonstrated\
\
the conductorwill recover if driven to a normal\
\
slate in a horizontal solenoid configuration at the\
\
MFTF peak field and current conditions. Convair/\
\
GeneralDynamicsandGeneralAtomicmade\
\
analyticalstudies ofconductorcryostabilityand\
\
concluded there is sufficientheat transfer for the\
\
conductor to recover in the peak field and current\
\
condition.\
\
A comparison of calculated surface heat flux\
\
on the conductor when in a normal condition to ex\
\
perimentally measured heat flux is shown in Fig. 60.\
\
52\
\
Insulating\
\
divider\
\
Conductor pack\
\
FIG. 58.Helium outlet for magnet.\
\
-Slip planes, mylar\
\
Interlayer insulation, slotted G-11\
\
-Grooves, 2-mm spacing\
\
,/\- Conductor\
\
'^Cooling channels^\
\
v\
\
Support block, G-11\
\
Interturn insulation, G-11 buttons:\
\
FIG. 59.Conductor and support iilocks in outside\
\
layer of large-radius region of magnet coil showing\
\
channels for helium Dow.\
\
S0.\
\
0.25\
\
0.20\
\
0.15\
\
0.10\
\
0.05\
\
;I'I 'I'IT"i—r\
\
-JJL\
\
Heat flux at 7.68 T,\
\
Convair analysis\
\
Available cooling,\
\
Lyons' data\
\
.II!L \_ L\
\
45678910111213\
\
Surface temperature (K)\
\
FIG. 60.Conductor heat flux.\
\
The available cooling exceeds the surface heat flux\
\
by approximately5%.resulting in a small margin\
\
for uncertainties or flaws. Moreover, a calculation\
\
byConvairoftheeffectofsolderflawson\
\
cryoslability indicated that a worst case had prac\
\
tically no effect. (A worst-case solder flaw is defined\
\
as no more than 27 flaws of 40% contact or less per\
\
1.5 m of conductor, or less than 13 flaws per 7 cm,\
\
or less than 8 consecutive flaw-s on one side.)\
\
Plenums and flow channels are provided in the\
\
magnet to inhibit vapor accumulation, as described\
\
intheforegoingsection,anuthushelpassure\
\
cryoslability.\
\
CURRENT LEADS\
\
A pair of copper bus conductors with super\
\
conductorswillcarrycurrentfromthevacuum\
\
vessel wall to each magnet. These conductors will\
\
normally be in liquid helium, but they are designed\
\
to be superconducting in vapor flow. II' the liquid\
\
level in Ihe current lead pipes should be depressed,\
\
the leads will be cooled by cold-end conduction and\
\
by controlled vapor How. A heal transfer analysis of\
\
theseleadshasprovidedadesignthaishould\
\
guarantee cryostability for all operating conditions.\
\
By using 2.5- by 7.5-cm (1- by 3-in.) OFHC copper\
\
conductors having a residual resistivity 'alio greater\
\
than150\. a NbjSn superconductor with a 0.4-g,'s\
\
vaporflowperlead,thevapor-cooledlength\
\
operating at 6.000 A can be over 4 m, which is more\
\
than sufficient. A section view of these conductors is\
\
shown in Fig. 61.\
\
Theportionoftheleadspenetratingthe\
\
vacuum vessel w'ull will have a temperature transi\
\
tionfromapproximately5 K in the ct.rrent lead\
\
53\
\
•Instrumentation\
\
cable\
\
Nb3Sn\
\
superconductor\
\
Copper bus\
\
(1 X 3 i n . )\
\
6-11 blade\
\
(1-in. thick)\
\
LN shield\
\
6-in. sch 10\
\
stainless steel\
\
pipe\
\
FiG. 61.Helium-cooled current bus.\
\
pipe to 300 K outside the vessel. They are designed\
\
lo use near-optimum helium flow and to have a suf\
\
ficient amount of thermal mass for safe operation\
\
by using concentric copper and stainless steel tubes.\
\
The leads will be constructed of approximately 60\
\
copper tubes 0.25-in. o.d. (6.35mm)by 0.025-in.\
\
wall (9.65 mm) by 60-in. long (1.5 m). Each copper\
\
lube is encased in an 0.3I-in.-o.d. (8 mm) stainless\
\
steel tube. This design gives a mass ratio of stain\
\
less steel to copper of about 2:1 and should permit\
\
approximately20 minofadiabaticoperationat\
\
6000 A before a fast discharge of the magnets must\
\
be initiated.\
\
The effect of the stainless steel on the lead elec\
\
tricalresistanceissmallascanbeshownby\
\
analysis. Equating joule heating in the lead to con-\
\
vective cooling gives anexpressionforthel;ad\
\
length in the followingform:\
\
(=(4)\
\
where m is the helium mass flow rate, Ae is the\
\
helium enthalpy change. I is the lead current, Acand\
\
Asare copper and stainless steel cross-section areas\
\
andpis the electrical resistivity integral over the\
\
dimensionless lengthx/C.\
\
• / . ' \- ( ? ) •(5)\
\
Flowcircuitryfortheseleadsis shownin\
\
Fig. 62. A by-pass open-close valve is included lo\
\
maintain flow in the event of a failure of the flow-\
\
controlvalve.Heliumflowwillordinarilybe\
\
regulated proportional to lead current with a con\
\
trollerusingthe lead voltage dropas the input\
\
signal. A 2.5-cm venl line will be connected to the\
\
helium duct, as shown in Fig. 62. to prevent the liq\
\
uid level from being depressed too far.\
\
QUENCH\
\
Intheeventofaquench,themagnetwill\
\
automatically be put into a fast-dump mode by a\
\
quench detection computer. Voltage taps on ihe coil\
\
bundle and current leads will be used to detect nor\
\
mal conduction zones. The contained energy will be\
\
dissipated in an external discharge resistor that will\
\
be sized to limit both voltage and maximum con\
\
ductor temperature.\
\
A conservative method for estimating the max\
\
imumconductortemperatureT,intimeTwith\
\
currentIandresistanceR,istocomputean\
\
adiabatictemperatureriseofeachelement,in\
\
cluding the mass m(. length L, and heat capacity CJ.\
\
Bypass\
\
valve\
\
300 K -\
\
GHe line\
\
Control valve\
\
Controller\
\
Transition lead\
\
Current lead duct\
\
GHe vent line\
\
Superconducting bus\
\
u\
\
Room temperature bus\
\
FIG. 62.Transition lead.\
\
54\
\
Rdr =2Jl-miJqdT.(6)\
\
Instantaneousthermalcommunicationbe\
\
tweentheconductorandadjoiningmaterialsis\
\
assumed. The participating materials are the super\
\
conductor, copper, interlayer and interturn insula\
\
tion, and helium gas.\
\
The current history is defined according to the\
\
dela\ lime 7,1.\
\
'lo. 0 < r <,-d\
\
(<>exp(-rro).T >7-J\
\
(71\
\
Since R can be expressed in terms of resistance\
\
length and area. /iL /A. Hq. (5) can be restated in the\
\
following form:\
\
T W „2)••ZJm\
\
i(c\
\
i-\
\
iJ\
\
T „\
\
i//')dT.(8)\
\
Solution of liq. (8) yields the maximum tem\
\
perature as a functionoidelay until a discharge is\
\
initiated. This decay constant corresponds lo a peak\
\
voltage of 1000 V across a 0.17-!! quench resistor,\
\
with a 12-H inductance in the magnet (i.e.. rn =\
\
lil<)/V„,ax). Copper properties assumed in liq. (8)\
\
arethoseofOFHCcopperhavingaresidual\
\
resistivity ratio equal to 150 and a 7.76-T magnetic\
\
field.\
\
Figure 63 shows that a maximum temperature\
\
of 200 K is reached if a 10-s delay in initialing dis\
\
charge is allowed. Longer decay tinie constanls (i.e..\
\
lowerquenchvoltages)resultinhighertem\
\
peratures. This temperature is considered permissi\
\
ble since it is a conservative estimate and is limited\
\
10 a small region of the coil where both initial tran\
\
sition to normal conduction and peak field could\
\
occur. A 200 K temperature rise from 5 K should\
\
result in less than 0.1% thermal expansion of the\
\
conductor. Also. 10 s is adequate time to initiate a\
\
dischargewithanautomaticquenchdelecting\
\
system.\
\
Magnet structuralmaterials can also develop\
\
resistiveheatingduringafastdischargebya\
\
transformer coupling effect between the magnet coil\
\
and surrounding structure. Because of its proper-\
\
lies, a copper guard-vacuumbladder would show\
\
the greatest temperature increase from this source.\
\
Assuming adiabatic conditions, resistive and sensi\
\
ble heating of the copper can be equated.\
\
For a 1392:1 turn ralio. a peak coil voltage (V)\
\
of 1000 V. and a current-decay time constant (TO) of\
\
69 s. the temperature (T) can be found by Eq. (9).\
\
0.26 r0\
\
Vl-5in\
\
T„,a,\
\
r\
where Lcis the effective length andpcis ihe product\
\
ofresistivity and specific heat.\
\
Solutionof this equationyields a maximum\
\
temperatureof 5 K forthe copperbladder: it is\
\
much less for a stainless steel bladder.\
\
A similaranalysisof thestainless-steelcase\
\
yields a temperature of 10 K. Total energy dissipa\
\
tion in the case, bladder and coil jacket is approx\
\
imately 6 MJ. or less than1.5'V of the cnerg> con\
\
tained by the magnet before discharge.\
\
Heat transfer to the helium occurs at a much\
\
slower rate than the inductive healing and has no ef\
\
fect on conductor stability because any appreciable\
\
amount of vapor is formedlong after the curreni\
\
has decayed to a low value.\
\
Rupture discs in the vent lines will limil \\apor\
\
pressures during a quench 10 approximately SO psi\
\
The gaseous helium recover, swem can accept up\
\
to 16.000 g/s. which will be adequate.\
\
010203040\
\
Discharge delay time (si\
\
FIG. 63.Maximum conductor temperature during\
\
a quench.\
\
55\
\
REGENERATION\
\
All liquid helium cooled surfaces of the magnet\
\
.swetn must be warmed periodically to more than\
\
15 K for boil-off of condensed hydrogen. This is to\
\
be accomplishedby radiating heat to the magnet\
\
case and piping from the nitrogen shields. However,\
\
the shields must fir.it be filled with a warm gas and\
\
themagnetmustbediseharged.Heliumvapor\
\
generationin the magnetsystem will be used to\
\
depress the liquid helium level to the bottom of the\
\
vacuumvessel andpump liquidinto the storage\
\
system. After regeneration, the nitrogen and helium\
\
systems will be returned to their normal stales as\
\
soon as possible.\
\
56\
\
SECTION 7\
\
POWER SUPPLY SYSTEM\
\
Thepowersupplysystemfor the yin-yang\
\
superconducting magnet provides controlled power\
\
and protection in case of quench or other poten\
\
tially destructive conditions. Principal requirements\
\
for the power supply system are: provide a con\
\
trolled current of 0-6000 A in each coil with an\
\
offsetofupto1200 Abetweencoilcurrents;\
\
providea means to energize and deenergize the\
\
magnet within 4 hours; provide detection of quench\
\
or other abnormal condition in the magnet; and\
\
deenergize the magnet, either in a slow or rapid\
\
mode, upon detection of an abnormalcondition.\
\
Two identical power circuits are used for each of the\
\
superconducting coils (Fig. 64). Control and coor\
\
dination of the two power circuits and detection of\
\
an abnormal condition in the power supply system\
\
or the magnet is done locally in the power supply\
\
and magnet protection controller.\
\
SYSTEM DESCRIPTION\
\
The power circuit for each coil consists of a 0-\
\
12 V, 0-6000 A(dc) power supply for energizing the\
\
coil and maintaining the current during steady state,\
\
a 0.0015-ohm resistor in parallel with a pneumatic\
\
controlled switch for slow de-energizing of the coil,\
\
and two 6000-A, 750-V(dc) circuit breakers with a\
\
center tapped 200-MJ resistor for fast de-energizing\
\
of the coil. The dc power supply is a conventional\
\
thyristorphase-controlledrectifierwithafree\
\
wheelingdiodepathratedforfullcurrent.A\
\
pneumatic switch is in parallel with the power sup\
\
ply to bypass the supply if its cooling water fails.\
\
Theslowde-energizingresistorandpneumatic\
\
switch are used to insert resistance in scries with the\
\
magnet coil during operator controlled turndown or\
\
an equipment failure of a noncriticalmagnitude.\
\
Two fast de-energizing dc circuit breakers are used\
\
to interrupt the current from the power supply and\
\
transfer it to the 200-MJ resistor if a critical condi\
\
tion in the magnet is detected, such as a propagating\
\
normal zone or overheated current lead. The 200-\
\
MJresistoris a passively cooledresistor of the\
\
natural air convection type. It is center tapped with\
\
a soft ground to limit the coil voltage to 500 V or\
\
less from ground.\
\
The power supply and magnet protection con\
\
troller interfaces with the MFTF computer system\
\
besides being a complete local system which can\
\
operate independently. Its most significant function\
\
is to: monitor key signals such as voltage taps in the\
\
coils,currentleadvoltages,heliumleveland\
\
pressure,and failureindicatorsin the cryogenic\
\
system; process the information to determine if an\
\
abnormal condition exists and its severity; and then\
\
initiateaholdon the currentrise,a slow de-\
\
energizing command, or a fast de-energizing com\
\
mand.\
\
An uninterruptable power system is used to en\
\
sure operationof the magnetprotectioncompo\
\
nents during a power outage. A 120-V(dc) battery\
\
supplies control power to the circuit breakers and\
\
input power to an inverter which powers the power\
\
supply and magnet protectioncontroller. The ac\
\
building power is used as a backup in case of in\
\
verter failure.\
\
ENERGIZING AND\
\
DE-ENERGIZING\
\
CHARACTERISTICS\
\
Figure 65 shows the energizing and slow de-\
\
energizingcharacteristicsofthepowersupply\
\
system. With 12 V from the power supply and an in\
\
ductance of 12 H in series with the cable resistance\
\
of 0.5 m£2, the current increases at nearly 1 A\*s\_ I\
\
.\
\
Actual energizing characteristics will be slower and\
\
will use the full 4 hours allotted. This is to minimize\
\
probability of rate-induced normal zones develop\
\
ing in the magnet at high current. With the power\
\
supply turned off and with 2milin series with the\
\
12-H inductance, the magnet current decays com\
\
pletely in 3 hours. A slower rate of decay can be ob\
\
tained by leaving the power supply on; however, it is\
\
not anticipated that this w-'l be needed.\
\
Thefastde-energizingcharacteristicis not\
\
shownin the figure. It wouldbe an exponential\
\
decay with a time constant of approximately 70 s.\
\
POWER COMPONENTS\
\
The major power components are the power\
\
supplies, pneumatically controlled switches, 1.5-mft\
\
resistors, dc circuit breakers, 6-kA bus, the 200-MJ\
\
fast de-energizing resistors, and the battery/inverter\
\
57\
\
480 V\
\
480 V\
\
0.0015 Si 6000 A, 750 V\
\
oo -\
\
120V(dc)\
\
battery\
\
120V(ac)\
\
ups\
\
To cryogenic\
\
system\
\
FIG. 64.Magnet power-supply system.\
\
58\
\
6000iir\
\
^ M a x i m u m rate (12.0 V)\
\
5000\
\
L = 12H\
\
R = 0.5 mS2 (Cable resistance)\
\
4000\
\
<3000\
\
2000\
\
1000\
\
Maximum rate (-2.0 V)\
\
t ( h )\
\
FIG. 65.Energizing and de-energizing characteristics.\
\
59\
\
system. With the exception of the fast de-energizing\
\
resistors and the 6-kA bus, all the power compo\
\
nents are located on the third floor, northeast end of\
\
Bldg. 431. The fast-dumpresistors are located on\
\
top of the southeast vessel support pillar.\
\
Power Supplies\
\
Each power supply will be built to El: Depart\
\
mentSpecificationLES 22249. The supplies will\
\
produce 0-12 V, 0-6000 A with a series connected\
\
load of12 H and 0.5-2.0 miJ. The n.\_smilis the\
\
cableresistanceandthe2.0milisthecable\
\
resistance plus1.5milof the slow de-energi/ing\
\
resistor. A continuously rated free-wheelingdiode\
\
path will be provided so that the magnet current will\
\
not be interrupted if the power .supply or the input\
\
ac line voltage fails.\
\
The power supply is required to producefull\
\
output with 480 V ± 5% input. It will operate con\
\
tinuously at reduced output voltage down to 416 V\
\
and for 0.2 s down to 330 V. This latter voltage oc\
\
curs during starting of motors in the cryosystem.\
\
The power supplies will operate either in a local\
\
or remote mode, but local mode is onhfor opera-\
\
lion of the power supplies into u dummy load dur\
\
ing initial checkout and maintenance. The remote\
\
mode is used during operation of the magnet and\
\
corresponds to operation of the supplies from the\
\
powersupplyandmagnetprotectioncontroller.\
\
Power supply current and voltage demands will be\
\
four 20-mA analog signals provided from the con\
\
troller, while digital inputs and outputs of zero and\
\
24 V will be for on/off,localmode inhibit, and\
\
power supply diagnostic indicators. The power sup\
\
plies will be capable of both current and voltage\
\
regulationwithautomaticcrossover,butthe\
\
magnet will be primarily operated in current regula\
\
tion with the voltage regulator serving as a voltage\
\
limit. The maximum absolute errors are ±60 A and\
\
±0.5 Vwithmaximumrepeatabilityerrorsof\
\
±30 A and ±0.5 V.\
\
The power supply output terminals are isolated\
\
by ±1000 V(dc) from the power supply enclosure,\
\
ac input terminals, and control ground. This is to\
\
prevent a ground fault during the fast de-energizing\
\
mode where ±500 V exist between each magnet ter\
\
minal and ground.\
\
No filtering is required in the power supplies\
\
because of the high magnet inductance. A ripple fre\
\
quencyof350 Hzorgreaterwas specifiedwith\
\
special precautions for minimizing lower frequency\
\
harmonics.\
\
Pneumatically Controlled Switches\
\
The switches which parallel the power supply\
\
outputs and the 1.5 mii slow de-energi/ing resistors\
\
are Square D type DB or equivalent. These switches\
\
are capable of continuous operation at 6000 A and\
\
can interrupt 6000 A up to 20 V. They are mounted\
\
in series with the interconnecting bus bars which arc\
\
used in the magnet power suppharea.\
\
Control of the switches is with a conventional\
\
pneumatic control system employing an air-cylinder\
\
operator, a four-way solenoid valve, and a pressure\
\
regulator. An accumulator will be used with a check\
\
valve to provide a backup source of air to allow\
\
operation with loss of building air. The lbur-wa\\
\
solenoidvalveis connectedtotheswitchusing\
\
pohpropvlene tubing to separate the switch and the\
\
controls by ±IOOOV(dc).\
\
The switches which parallel the power supph\
\
outputs are normally open and close automatically\
\
when commanded by the power supph water How\
\
and overtemperature monitors. The switches which\
\
parallelthe1.5-mi2 resistors arenormally closed\
\
and open when commandedby the power supph\
\
andmagnetprotectioncontroller.ANofthe\
\
switches have limit switches for monitoring by the\
\
controller.\
\
1.5 Milliohm Resistors\
\
Thel.5-mi2resistorsarcusedforslowde-\
\
energi/ing of the magnet. They are rated at 54 kW\
\
and are natural-convection air cooled.\
\
dc Circuit Breakers\
\
Two dc circuitbreakers per EEDepartment\
\
SpecificationLES 22250 are used for each magnet\
\
coil for redundancy, and will be mounted on open\
\
frames in the power supply area. They are capable\
\
of operating at 6000 A(dc) continuously and will in\
\
terrupt up to 300 kJ of stored energy in the induc\
\
tance of the cables and the fast discharge resistors.\
\
Althoughratedto operate in a 750-V(dc) circuit,\
\
they actually can operate up to their arc voltage of\
\
2200 V.\
\
The control circuit of each circuit breaker will\
\
be powered from a I20-V battery. Each breaker will\
\
haveanundervoltagereleasewhichwillauto\
\
maticallyopenthe breakerif the controlcircuit\
\
60\
\
voltagefallsbelow70 V. The primarymeans of\
\
opening the circuit breakers is the shunt trip coil,\
\
which will be controlled from the power supply and\
\
magnet protection controller. Individual driver cir\
\
cuits are used for each breaker. The breaker posi\
\
tion auxiliary switches are used for monitoring by\
\
thecontrollerandalsoforinterconnectingthe\
\
breakersto ensure openingof allfourbreakers\
\
simultaneously (Fig. 66).\
\
6-kA Cable Bus\
\
The cable bus runs from the power supply area\
\
lo the magnet lead exit ports are approximately 100-\
\
ft long. Cable bus will be used which is similar to\
\
cabletraybutwithmaintainedspacingbetween\
\
conduciors for sufficient cooling lo allow the cables\
\
to he sized as single conductors in free air. The ad\
\
vantages of cable bus over rigid bus are: continuous\
\
cable runs withoutneeding expansion joints; and\
\
redundani personnel safetv due to having insulated\
\
cables in an enclosed, grounded duel. Details of the\
\
cable bus design will be done b> the manufacturer.\
\
20U-M.I Fast-Dump Resistors\
\
Duringafastde-energi/edcondition,the\
\
magnetenergvisdissipuiedinthefastdump\
\
resistors. The resistors will be passively cooled using\
\
naturalairconvection.Theresistorswillhave\
\
several parallel paths to provide redundancy.\
\
A cable independento^the 6-kA bus will be\
\
used to connect the magnet l^ads at the top of the\
\
vessel to the dump resistors which are located on\
\
top of the southeast vessel support pillar. Indepen\
\
dent cable is used to ensure that the dir-ip resistors\
\
are connected in ease of damage lo the 6-kA bus.\
\
These cables are .-.i/ed for short time ratings and will\
\
not operate ai 6-kA continuously.\
\
Battery and inverter System\
\
A 120-V. 100-Vhbattery.25-Acharger.anda\
\
2.5-kVAinverlercomprisetheuninterruptable\
\
power supply system(Fig. 67). The batteryis of\
\
lead-calcium construction with a translucent jar for\
\
high reliability and ease in maintenance. The bat\
\
tery is composed of 20 separate units mounted ina\
\
seismic, two step rack. The inverler operates from\
\
the battery and supplies power lo the power supply\
\
and magnet protection controller. It has an elec\
\
tromechanical transferswitLh which transfers the\
\
power source fromthe inverter lo the building ac\
\
powerin case ofinverterfailure.The estimated\
\
mean lime between failure for the combined power\
\
source is more than100,000 h. As shown in the\
\
figure, several conditions in both the charger and in\
\
verter are monitored by the controller.\
\
POWER SUPPLY AND MAGNET\
\
PROTECTIONCONTROLLER\
\
The power supply and magnet protection con\
\
troller is a local control system which can operate in\
\
conjunction with the MFTF computer system or in\
\
dependently. It will be located in the power supply-\
\
area in the northeast corner of Bldg. 431.\
\
Functions\
\
The controller has a means for selecting locai\
\
or remote operation using a keylock switch, check\
\
ing out components prior to application of power to\
\
the magnet, monitoring of component failures, and\
\
selecting a hold, slow, or fast de-energizing com\
\
mand. It also communicates to the Local Control\
\
Computers in Bldg. -439 through a CAMAC crate.\
\
Based on digital commands from the local control\
\
computers it provides analog current andvoltage\
\
demandstothepowersupplies,generatesthe\
\
energizing trajectory, controls insertion of the 1.5-\
\
mi!resistors, and controls on. off commands to the\
\
power supplies.\
\
Themostcriticalcontrollerfunctionisto\
\
providemagnetprotection.Ilmonitors;voltage\
\
taps in the magnet for detection of a normal zone\
\
(quench), voltage taps across the current leads for\
\
detectionofovertemperature.heliumleveland\
\
pressure in the helium supply Dewar for detection\
\
of quench or near quench conditions, and failures in\
\
ihe crvogenie and \\acuum s\ stems to anticipale a\
\
polentiil loss in cooling of the magnet, or large heat\
\
influx. Based on the severity of an abnormal condi\
\
tion, the controller selects a hoid. slow, or fast de-\
\
eiiergi/ing command.\
\
Required Features\
\
Beeau^: uf the critical nature of magnet p: elec\
\
tion, the following features will be included in the\
\
controllerThe quench detection and current lead\
\
voltagemonitoringwillhavefullyredundant\
\
s\ stems. Self-checking will be used in low reliability\
\
components such as microcomputers. Checkout of\
\
ke\componentsprior to operation will be done\
\
automaticallywhere possible. Checkout will con\
\
tinue during early stages of energizing the magnet.\
\
61\
\
120 Vdc\
\
Position sw\
\
From controller\
\
120 Vdc\
\
' — O - O -\
\
Control\
\
circuit\
\
From controller\
\
' — C K O -\
\
120 VdcControl\
\
circuit\
\
From controller\
\
120 VdcControl\
\
circuit\
\
From controller\
\
\- » \- T o\
\
controller\
\
• T o\
\
controller\
\
-To\
\
controller\
\
-To\
\
controller\
\
FIG. 66.Circuit breaker interconnections.\
\
62\
\
To controller\
\
\- \\* \- Ground detector(dc)\
\
Powerfailure(ac)\
\
Inverterfailure\
\
• -Overload\
\
Battery low\
\
FIG. 67.Batterv and inverter system.\
\
De-energize Modes\
\
Asstaledearlier,therearethreemodesof\
\
operationif an abnormal conditionin the power\
\
supply system or the magnet occur: hold, slow de-\
\
encrgi/e. and fast de-energi/e. The hold mode stops\
\
the increase in current during energizing. The slow\
\
de-energi/e mode turns the power supplies off and\
\
inserts 1.5milm series with each magnet coil. The\
\
fastde-energi/emodeopensallfourdccircuit\
\
breakers and sends a signal to the cryogenic system\
\
to close the supplv \* alves from the Dewar and open\
\
the valves to the heliumrecoverysystem.\
\
Table 16 shows the main failure conditions and\
\
the appropriate action which will be initiated. The\
\
basis for this table is the failure modes, effects, and\
\
criticalityanalysisperformedbyIntermagnetics\
\
GeneralCorporation(IGC).S<)\
\
Thistablealso\
\
reflects the use of the fast de-energize mode as the\
\
last line of protection. This is to minimize disrup\
\
tiontotheMFTFoperationandtominimize\
\
probability of voltage failures' in the magnet.\
\
Quench Detection\
\
Detection of a normal zone in the magnet will\
\
be done by monitoring the voltage taps in both coils\
\
and by monitoring helium pressure in the helium\
\
supplyDevvar.Useofthevoltagetap>enables\
\
detectionof normal/ones as shortasI m. iden\
\
tification of their approximate locationinthe coils,\
\
andtheirgrowthbehavior.Useofthehelium\
\
pressure enables detection nf normal regions in the\
\
coil but not their location or growth characteristics.\
\
It is used to cmcr potential blind spots in the elec\
\
tricalquenchdetector,andbecauseit is a com\
\
pletelydifferenttype of detcctoiandwoulonot\
\
have a faultmode in commonwnii the electrical\
\
detector.\
\
Based on temperature rise calculations and ex\
\
perimentalnormal /onepropagationrates of ap\
\
proximately 1 m s. the magnet w ill enter the fast de-\
\
energi/e modema normal /one length of approx\
\
imately10 m. To delect small stable normal /ones\
\
and to monitor growth rates it is required to detect a\
\
normal /one as small as 1 m. Detection of growth\
\
rates is required to discriminate against fast signals\
\
whichoccurdue to conductormotionand elec\
\
tromagnetic interference.\
\
Detection of a l-m length requires detection of\
\
26 mV of resistive voltage (at 6000 A) in the pres\
\
ence of ±12 V of inductive voltage during energiz\
\
ing orde-energizing.A well knownmethodfor\
\
quench detection with a single coil is the balanced\
\
63\
\
W2\
\
3+VL\
\
\- 0 +V\
\
C\- 0 — f c0+vR\- C\
\
L\
\
b1 3/ Mb1Mb\£L\
\
b2^b2\
\
Coil #1\
\
\- 0 +v,B\
\
Coil #2\
\
Detector equation:v0= K,(vc\- vB) \+ K2(vL-vR)\
\
For similar coils:L( 1= Lt 2= L(\
\
K, =\
\
LM„\
\
1L\
\
b \ .^h j\
\
LL /\
\
<=fe\_\
\
1^i\*>\
\
L"L\
\
ForMFTF:L, = 5.79H\
\
L = 11.11H\
\
M = 1.149H\
\
v0= 0.0265 (vc\- vB) \+ 1.869 (vL-vR)\
\
FIG. 68.Quench delilion method.\
\
64\
\
TABLE16.Failureconditionsandappropriate\
\
action.\*\
\
Appropriate action\
\
ConditionFastSlowHold\
\
X\
\
X\
\
X\
\
a\
\
Basedon failure mode\*, effects, and criticalily analysis per\
\
formed by Intermagnelics General Corporation.\
\
Quenchdetector,current-leadvoltagealarms,andhelium\
\
pressure alarm have backup systems.\
\
self-inductance bridge method. Because MFTFhas\
\
two coils which are coupled together with a cou\
\
pling coefficient of approximately 0.1, and because\
\
the large dimensions of the coil cross-section result\
\
in varying coupling depending on voltage tap loca\
\
tion, the self-inductance bridge method cannot be\
\
used in the conventional mannerBased on calcula\
\
tions using the EFFI code, approximately 70 mV\
\
result in the bridge due to mutual inductance im\
\
balance. IGC in their quench detection and magnet\
\
protection studyproposed using current rate in\
\
dicators to compensate for the mutual inductance\
\
imbalance.\
\
AnanalysisdonebyLLNLresultedina\
\
quench detection method which successfully com\
\
pensates for mutual inductance imbalance without\
\
requiring current rale sensors.Two voltages be\
\
tweencoilsareusedinadditiontotheself-\
\
inductancebridgestoformadetectorequation\
\
(Fig. 68). The gains K\| and K? are functions of the\
\
partial mutual and self inductances in the two coils.\
\
HARDWAREIMPLEMENTATION\
\
Detailed design of the hardware for the power\
\
supply and magnet protection controllerhas not\
\
been completed. One implementation was described\
\
intheIGCstudy.8 0\
\
Theselecteddesignuses a\
\
programmable controller with an analog backup for\
\
quench detection and current lead monitoring.\
\
Growing normal zone, {> 10 m)X\
\
Stable normal zone ( < 10 m)X\
\
Current lead overheating\
\
Voltage alarm, kvel 1X\
\
Voltage alarm, level 2X\
\
Low helium level in DewarX\
\
High helium pressure in Dewar\
\
Pressure alarm, level IX\
\
Pressure alarm, level 2X\
\
Valve from helium Dewar closesX\
\
Main vacuum failure, majorX\
\
Guard vacuum failure, majorX\
\
LN systemfailure\
\
Helium refrigeration failure\
\
Magnet-protection controller failureX\
\
MI-IT" computer failure\
\
dc power supply failureX\
\
Battery charger failureX\
\
Inverter failureX\
\
I2(l-V(ac) power failureX\
\
Inverter and I20-V(ac) power failureX\
\
120 V(acl to circuit breaker failureX\
\
65-66\
\
SECTION 8\
\
STRUCTURAL ANALYSIS\
\
INTRODUCTION\
\
A baseline structural analysis was performed in\
\
supportof the preparationof the MFTF design\
\
drawings.Thisanalysisdemonstratedthe basic\
\
structural integrity of the MFTF structure using the\
\
best available loads and material data. Subsequent\
\
to the completion of the MFTF design, additional\
\
analyses were performed. These refined the finite\
\
element analysis in the critical stress region, and in\
\
terpretedthe finite element results in light of the\
\
latest material properties and actual structural ef\
\
fects such as stress concentrations. Also investigated\
\
werethe potentialeffectsof assumedstructural\
\
faultsin criticalmagnet structure. These analysis\
\
tasksare summarizedin separatesections. The\
\
following discussion summarizes the structural re\
\
quirements analysis methods and results from the\
\
baseline structural analysis documented in Ref. 10.\
\
STRUCTURAL REQUIREMENTS\
\
Materials\
\
The structuralcase materialfor the MFTF\
\
magnet is 304LN CRES steel with a nitrogen con\
\
tent of 0.14% (minimum). The weld metal is E316L.\
\
The coil-jacket plate material is 316 CRES steel as is\
\
the jacket weld metal. Design stresses for theMFTF\
\
case and weldments were based on anticipated yield\
\
strengths of 120 ksi at4 K and on the expectation of\
\
adequate fracture toughness and flaw growth rates.\
\
The preliminary plate and weld mechanical proper-\
\
lies were obtainedfromthe NationalBureau of\
\
Standards(NBS) foruse inthe casedesign\
\
(Table 17), along with properties for the other struc\
\
tural materials in the magnet.\
\
Factors of Safety\
\
The basic requirement imposed on the magnet\
\
case structure was that there should be a factor of\
\
safely of 1.5 on an anticipated 120-ksi yield strength\
\
for the operating magnetic and thermal loads. The\
\
factors of safety on the remainder of the magnet\
\
structure are consistent with the ASME code as in\
\
Table 18.\
\
Design Load Conditions\
\
TheMFTFmagnetmustwithstandboth\
\
operating and fault conditions. Design loading con\
\
ditionsincludecooldown,warmup,and normal\
\
operating conditions, operating and faull magnetic\
\
conditions, and seismic inertia conditionsIn addi\
\
tion to these conditions, the magnet is also designed\
\
to a 2.0 g handling condition.\
\
TABLE 17.Structural material mechanical properties.\
\
MaterialUsage\
\
Temp,\
\
K\
\
Ultimate\
\
strength,\
\
ksi\
\
Yield\
\
strength,\
\
ksi\
\
Elastic\
\
modulus,\
\
I 06\
\
psiSource\
\
304LSupport strutsRT\
\
4.5\
\
100\
\
245\
\
40\
\
70\
\
28.5\
\
29.5\
\
LLL\
\
304 LNMagnet case and\
\
intercoil\
\
4.5244.6111.8''29.7NBS\
\
E3I6L -IFc = 4.5\
\
.IFc= 9.2\
\
Case weld metal4.5\
\
4.5\
\
193\
\
187\
\
116"\
\
128"\
\
31.9\
\
31.9\
\
NBS\
\
316LJacketRT\
\
4.5\
\
80\
\
200\
\
40\
\
80\
\
29.5\
\
31.9\
\
EPON828/Vtrsamid 125\
\
with chopped glass fiber\
\
Conductor shimming20-14700--GDCEMS\
\
04)096-51\
\
CPRUpjohn\
\
Polycast1009-78\
\
Jackct-to-cascRT-13250-0.387Mfgr Data\
\
A286Support holtsRT1409529.1A M S 5737H\
\
"120 ksi wasuseifor design pending determination or finalallowables.\
\
67\
\
TABLE 18.Magnet structural safety factors.\
\
Factor of safety\
\
StructureYieldUltimate\
\
or conditionstrengthstrengthFactor\
\
Case, jacket, intercoil1.5-4.0\
\
Support structure1.53 . 0a , b\
\
4.0\
\
Shields, shield supports1.53.0"\
\
Seismic safely factor1.252.5C\
\
a\
\
Based nn ASME codes requirements.\
\
Ultimate strength safely factor is 2.0 in welds.\
\
C\
\
ASMK codes allow an increase in allowable working stress for\
\
seismic conditions of 120%.\
\
Magnet loads were determined for both normal\
\
operating (both coils 100% energized) and fault con\
\
ditions(onecoil100%energized,onecoilin\
\
operative). The case plate magnetic pressures for the\
\
more critical normal operating condition are shown\
\
in Fig. 69. Also defined were the ground accelera\
\
tions for the seismic inertia conditions. Magnet ac\
\
celerations and support system loads are a function\
\
of bothmagnetsystemandfusionchamberand\
\
were determinedby a dynamic analysis. Magnet\
\
case loads for seismic condition were calculated by\
\
General Dynamics Corp. (GDC) based on assumed\
\
1.0 g vertical and 0.75 g horizontalaccelerations.\
\
Thermalloads in the magnetdue lo longitudinal\
\
and transverse thermal gradients during cooldown\
\
werecalculatedusingfiniteelementstructural\
\
analyses thermal distribution.\
\
ANALYSIS METHODS\
\
The stress analysis of the MFTF magnet and its\
\
support sys'/t is based on data from three separate\
\
finiteele.i-.itanalyses:alarge7000-degrce-of-\
\
freedomGDSAPanalysisofone-quarterofthe\
\
magnet,detailedMSC/NASTRANmodelsof\
\
typicalcasecross-sections,andasimplebeam-\
\
element model of the complete magnet and its sup\
\
ports.Thelarge GDSAPmodeldeterminedthe\
\
overallstresses anddeflectionsforthemagnetic\
\
loads,quenchpressureloads,andthenormal\
\
o p e r a t i n g4.5 Kt e m p e r a t u r ec o n d i t i o n .\
\
MSC/NASTRANmodels were used to refine the\
\
local case bending stresses for the magnetic load\
\
conditions, and a GDSAP beam element model was\
\
used to determine the overall magnet loads(orthe\
\
unsymmetric seismic and cooldown thermal condi\
\
tions. These models are discussed below and are\
\
documented in Ref. 10.\
\
The 7000-degree-of-freedomplate model of the\
\
MFTF magnet is shown in Fig. 70, representing one\
\
quarterofeachmagnetandtheinterconnecting\
\
load-block structure.\
\
The coil jacket, case structure, andintercoil\
\
structure are represented by linear-strain thin-plate\
\
elements,whichsimulatetheaxialandbending\
\
stiffnesses of the plate structure. The conductors are\
\
representedby six continuousrodelementsthat\
\
representthe lumpedaxialstiffnessof the pack.\
\
These elements are connected to the surrounding\
\
case and jacket structure by other rod elements that\
\
simulate the transverse stiffnessofthe conductor\
\
pack including the conductor, insulation, and effec\
\
tive gaps.\
\
The loading conditions for the GDSAP plate\
\
model are all quarter symmetric. Magnetic loads for\
\
boththenormalandfaultconditionswere\
\
calculated for an idealized (5 X 12) conductor grid\
\
used in an EFFI analysis. The loads on the EFFI\
\
grid were lumped together at the GDSAP conduc\
\
tor nodes and applied to the finite element model.\
\
Loads for the normal operating condition assume\
\
100% operatingcurrentinbothmagnets,while\
\
loads for the EFFIfaultcondition are based on\
\
100% current in one magnet and no current in the\
\
other. A quench pressure condition consisted of a\
\
uniform700 psibursting pressure appliedto the\
\
case plates surroundingtheconductorpack.The\
\
normal operating 4.5 K temperature cordition was\
\
analyzedto determine the residual stresses in the\
\
magnet caused by the differencesin thermal con\
\
traction between the case and the conductor.\
\
The output from this analysis consisted of the\
\
overall stresses and deflectionin the magnet con\
\
ductors, jacket, andcase. However,the element\
\
mesh in this modelwas too coarse to provide a\
\
detailed definitionof the local bending stress dis\
\
tributions in the magnet case plates. Therefore, a\
\
detailed MSC/NASTRAN model of the case cross\
\
sectionwas created to refine the overall GDSAP\
\
analysis results. This NASTRAN model refined the\
\
local case bending stresses as illustrated in Fig. 71\
\
and documented in Ref. 10.\
\
The NASTRAN/node! was usedto analyse\
\
three typical sections in the magnet's major radius.\
\
Models of the case at 0= 24°. 48°, and 72° were\
\
68\
\
Case plate magnetic pressuresRadial magnetic pressures\
\
1,6002,0002,4002,8003,2003,6004,000\
\
Case plate pressure (psi)\
\
a2000\
\
1600\
\
1200\
\
800\
\
400\
\
0= 67.5'\
\
0= 72.5°\
\
mimi\
\
Section A-A\
\
0.70.80.91.011.0\
\
Distance from magnet centerline (m)\
\
FIG. 6°-.The conductor pack exerts high pressures on the magnet case.\
\
Conductor\
\
pack\
\
•Quarter symmetric model\
\
•Plate elements (separate jacket & case)\
\
•Conductor pack modeled\
\
•Reflects current geometry and stiffness\
\
Case\
\
plate elements\
\
Jacket\
\
pla;e elements\
\
Conductor\
\
axial elements\
\
(b) Typical model cross section\
\
FIG. 70.The GDSAP finite element model accurately represents the MKTF structure.\
\
createdwiththeonlydifferencesbetweenthe\
\
models being the thickness of the side plate stif-\
\
fener, the depth h of the guard-vacuum section of\
\
the case, and the applied loads.\
\
The magnetic loads are applied directly to the\
\
NASTRANmodelcaseplatesasdistributed\
\
pressureloads,showninFig. 69.Additional\
\
pressureloadswere appliedtotheNASTRAN\
\
modeltosimulatethe effectiveradialpressures\
\
created by the axial stresses existing in the curved\
\
crossoverplates. These pressures were calculated\
\
fromthe longitudinalstresses obtainedfromthe\
\
GDSAP analysis. By assuming that a longitudinal\
\
stress fLin the curvedplate with radius R and\
\
thickness t would create an effective radial (normal\
\
to the plate) pressure P = fj.t/Rin addition to the\
\
pressure loading, forced displacements were applied\
\
to the NASTRAN model to ensure deflection com\
\
patibility with the overall GDSAP analysis.\
\
The GDSAP beam element mode! illustrated in\
\
Fig. 72 was created to determine the overall magnet\
\
and support loads for the unsymmetric inertia and\
\
thermalconditionandisalsodocumentedin\
\
Rcf.10.Because the inertiaandthermalstresses\
\
were expected to be small, the simple beam element\
\
model was considered adequate to confirm that the\
\
seismic and cooldown conditions were not critical.\
\
The model grid points, which are shown in Fig. 72.\
\
are located at the centroid of the conductor pack.\
\
The mass and axial stiffnessof the coil are repre\
\
sented by rod elements connected to the model grid\
\
and the axial and bending stiffnessesofthe magnet\
\
case are represented by beam elements whose cen-\
\
troids are offset from the conductor thermal condi\
\
tions. The cooldown thermal conditions addresses\
\
temperature gradients due to three symmetric and\
\
three unsymmetric helium flow' distributions during\
\
cooldown.Temperaturedistributionsforthese\
\
70\
\
Effective pressures induce by curvature\
\
and longitudinal stresses.\
\
C sym\
\
I\
\
Forced deflections\
\
>\
\
/)\
\
>\
\
\\
\
zs\
\
-\\* — IMagnetic\
\
pressures\
\
< \| \- r5\
\
B\
\
AAJ—\
\
— (£ stiffeners\
\
KIG. 71.The NASTRAIN finite element model refines (he case bending moments.\
\
variousflowdistributionsarcdocumentedin\
\
Rcf. 72.\
\
ANALYSIS RESULTS\
\
Case Stress Analysis\
\
The case stress analysis can be broken into two\
\
categories: the analysis of the major radius, and the\
\
analysis of the minor radius. In the major radius,\
\
theprimaryin-plane(membrane)stressesare\
\
caused by the C-clamp action of the magnet. The\
\
secondary bending moments are caused primarily\
\
by redistribution of the magnetic pressure loads ap\
\
plied to the side plates as shown in Fig. 73. In the\
\
minorradius,thein-planestressesarecaused\
\
primarily by the spreading or opening up action of\
\
the major radius lobes. The secondary bending mo\
\
mentsare causedby redistributionof the seven\
\
million pound intercoil load and by the effects of\
\
the high in-plane stresses acting in the curved minor\
\
radius plates, also shown in Fig. 73.\
\
Major Radius Analysis\
\
Thecriticalin-planestressesinthemajor\
\
radius case plates were determined entirely by the\
\
GDSAP plate modelfinite element analysis. The\
\
plate secondary bending moments were determined\
\
by the same analysis but were also refined by the\
\
Naslran analysis of typical case cross sections. In\
\
the center crossover plate where the Nastran and\
\
GDSAP moments differed because of the manner in\
\
w'hichthemagneticpressureswereapplied,the\
\
structure was analyzed for the worst combination of\
\
moment and in-plane loads. Fig. 74 shows the com\
\
binedprimaryandsecondarystressesatseveral\
\
locations in the case. All principal stresses were less\
\
than SO ksi.\
\
Peakstressesinthemagnetcaseduring\
\
cooldown arc caused by the temperature gradients\
\
between the case side plates and the external stif\
\
feners. The curvature produced in the stiffener by\
\
thetemperaturegradientscauseshighbending\
\
stresses in the case plate adjacent to the stiffener as\
\
7!\
\
FIG. 72.GDSAI' beam element model determines overall magnet stresses for seismic and cooldown thermal\
\
conditions.\
\
' Redistribution of applied loads\> Geometry changes that cause\
\
effective pressure toads\
\
Intercoil load\
\
( P « 6 . 5 X1 06\
\
l b )\
\
X\
\
1iiiuife^g\
\
Magnetic\
\
pressures\
\
(P = 2500 psi)\
\
FIG. 73.Critical case stresses are caused by secondary bending moments.\
\
72\
\
0 = 4 8 °Note:(N) = N A S T R A Ndata\
\
(S) =G D S A P d a t a\
\
\| fs l l e a r= -7500 psi (S)\
\
lf\
\
l o nG. =\- ^ 1 4 p s i ( S )\
\
22000l b / i n\
\
\
>9061 psi (S)\
\
6 4 5 8 psi (S)\
\
3 2 1 6 2 psi (S)\
\
10.0 inK/in(N)\
\
Fmax=38829 psP\
\
FlC. 74.The major radius stressesan\*lessthanHO ksi.\
\
shown m Kill- 75. However, these stresses do not ex\
\
ceed the vield strength of the structure at the tem\
\
perature at w hich the stress occurs. The case stresses\
\
caused b\ the longitudinal thermal gradients did not\
\
exceed12 ksi.\
\
Minor Radius Analysis\
\
The stresses in the minor radius are determined\
\
primarilyb\the GDSAP plate model. Membrane\
\
stresses are caused by the opening up of the minor\
\
radius due to magnetic loads, but 4.5 K operating\
\
thermal loads subtract slightly from these stresses.\
\
Thesecondanbendingmomentsaredueto\
\
redistributionofthe intercoil load and In elet'.ive\
\
pressurescausedbythecaseplatecurvatureas\
\
shown in Fig. 73.\
\
Figure 76 shows the stresses at several location.-,\
\
in the case minor radius. The stresses were taken\
\
directly from the finite element analysis except for\
\
the secondary moments in the chamferred surface of\
\
the inner case plates. The secondary bending mo\
\
ment in that location was modified to account for\
\
inaccuracies in the finite element analysis caused by\
\
the consiani strain irianguLr elements used in that\
\
portionofthe model. The peak principal stresses\
\
predicted h> the baseline analysis exceeded the in\
\
tended SO-ksi design stress by 2 tob''iin both the in\
\
ner and center crossover plates as shown in Fig. 76.\
\
However,an evaluationofsever.ilknowninade\
\
quacies in the GDSAP anahsis indicated that the\
\
predicledstresses would be red"cedby a refined\
\
unalvsis<>fthis area. The anaKsis refinementtask\
\
22500 psi (235K)\
\
43500 psi (140 K)\
\
Kl(j. 75.Cooldowntemperaturegradients create\
\
significant stresses.\
\
73\
\
\- Coil extension\
\
8 9 6 7 0 inlb/in—p'\
\
Element no. 3 0\
\
12970 inlb/in\
\
19206 psi\
\
62832 psi\
\
40907 psi\
\
25594in. lb/in.}^3 5 4 5 0inlb/in\
\
25387 psi ^ ^ j ,(^~——\_\_\_\
\
1 3 8 6 8 0 inlb/in»-i\
\
125653 inlb/in~-~r<^~~--~^\
\
4 8 1 2 8 psi\
\
79492psi\
\
2733 psi\
\
Element no, 250\
\
Element no1959\
\
8 8 3 4 8 inlb/in\
\
14752 psi\
\
FIG. 76.The peak magnet slresses occur in Ihc minor radius.\
\
was accomplished and is discussed in a separate sec\
\
tion of this report, and documentedinRef. 83.\
\
Intercoil Structure Analysis\
\
The intercoi! structure connects Ihc major lobe\
\
of one magnetwiththe minorradius of the other\
\
magnet. The primary slresses in the iniercoil struc\
\
ture are caused by the normal operatingmagnetic\
\
loads that createaseven million pound compression\
\
loadintheintercoilmember.Additionallow\
\
stresses are producedinthe intercoilstructure by\
\
thermalandinertia conditions.\
\
Design loadsfor this structure were obtained\
\
from the large GDSAPplate model and fromthe\
\
smaller G D S A Pbeam model. Figure 77 shows the\
\
membrane slresses in the intercoil structureforthe\
\
normal operating condition. Slresses for the normal\
\
operatingthermal condition are less than500 psi\
\
everywhere except in the inner plate near themajor\
\
radius iobe where the values are shown in Fig. 77.\
\
Overall intercoilloadsdue to cooldown and inertia\
\
conditions obtained from the GDSAP beam model\
\
arcsuperimposed>.nthesestresses.It-.ho. !dbe\
\
noted that the stresses sliould be .milupi.ed b\I \_5\
\
toaccountlorthediffeiencebetweenthe-.5-in.\
\
thicknessusedinIhcan'.lysiN.uulthe3..U-in.\
\
thicknessused in the final design.\
\
Support Structure Analysis\
\
The supportstructure consists of twosupport\
\
rods (' .2) that support the weight of the magnet and\
\
fivestabilizerrod-(3-7)thatreacthorizontal\
\
seismic loads as show n in Fig. 7S. The struts and the\
\
lugsattachingthestrutstothemagnetwere\
\
analyzed byhand to the loads shownin Table ll\
\
>.\
\
Also shown in Table 19 arc the margins of safety for\
\
variouscomponentsineachsupportstial.The\
\
margins of safely are based on the following expres\
\
sion:\
\
F.s\
\
74\
\
(7231\
\
(719)\
\
1596\
\
-6701\
\
5828\
\
1901\
\
8986\
\
4633\
\
Outer cover\
\
-3285\
\
11024\
\
2918\
\
(719)\
\
(703)\
\
1862\
\
-130320966\
\
-8748185285846\
\
-1496-200654672\
\
-24180-4370\
\
-3875\_\
\
-1917-19752108\
\
-21237\
\
-3167-26839-24243-3700\
\
-281652505-3629\
\
160\
\
20701789-3692\
\
-46893300229776-24319\
\
\- 33345\
\
-2718-1672-94-1755\
\
-6776673125-5989\
\
50050-3603633159-33278\
\
-3029186713874844\
\
Side plate\
\
Location 2\
\
Typical (727)\
\
{521\
\
(703)\
\
5589592524949934J—\
\
31881271173053933972\
\
1455156541433-4238\
\
/-1060 V\
\
-10\
\
\330 /\
\
Inner cover.—L\
\
1(52)\
\
stressesfor4.5° operating case\
\
FIG. 77.The stresses in (he intcrcoil structure are acceptable.\
\
TABLE 19.Support struts have adequate margins of safet\\.\
\
RodLLNI."tDesignSlrutSlnilBearingStrutCle>isfle»i»\
\
No.loadloadtensioncompressioncapacil)end lugpinIllS\
\
1+6H7\+ X(H>\
\
--.175\+ 11.09+(1.95\+ 0.15-0.1(1-0.2K\- 0.52\
\
2\+ 724\
\
-16R\
\
\+ K0O\
\
-175\
\
3\+ 5IK\+ 5211\
\
-2.17-275\+ 0.6K\+ 1.M\+ 11.17-11.14\+ 11.99-11.15\
\
4\+ 424\
\
-2\*1\
\
+5211\
\
-275\
\
5\+ .11(1\
\
-18.1\
\
\+ 175\+ 1.1.1-0.95\+ 11.12-11.11-11.76•11.15\
\
6\+ .159\+ 52(1-0.C.S\+ (1.19-11.17\- 0 . 1 4\- 0.99-11.15\
\
7-242\
\
-5115\
\
\+ 5211\
\
'Rvfcrutrv Tclvcon MFTf-SHW-M-105 of17 Octubcr i«)7H.\
\
when:\
\
FALLOW 'sl n t ;\
\
lesserof\
\
( FU/ 3 ) X1.2 = FI u.2.5and\
\
\
lb)\
\
(c)\
\
Thelugswerecheckedfortensionand\
\
shear •'bearing failures using an industry-wide stan\
\
dard lug analysis documentedin Lockheed Stress\
\
MemoNo. 1. Bolls were checkedforshearand\
\
bendingfailures.Thestrutsthemselveswere\
\
checked for tension and for beam column compres\
\
sion loads. In compression, an initial bowof1 in.\
\
was assumed in all rods.\
\
FRACTURE ANALYSIS\
\
A fracture analysis was performed in support\
\
ofthe MFTF designto confirmthe selection of\
\
304LNforthestructuralcasematerialandto\
\
demonstrate adequate life al the 80-ksi design stress\
\
level.\
\
A linear elastic fracture analysis of the 304 LN\
\
plateand316Lweldfillerwasdocumentedin\
\
Ref. 10. The analysis was based primarily on frac\
\
ture toughness and flaw growth rate data obtained\
\
fromNBS.Wheredatawasnotavailable,the\
\
anaKsisusedmaterialproperlyestimatedIr.im\
\
data lor similar materials obtainedIrom literature\
\
sources.TheanaKsiswasperformedusingthe\
\
FLAGRO II anaKsis program, a linear Hastic IT.K-\
\
ture mechanics program developed hv Rockwell In\
\
ternational Corporation.\
\
The case parentmaterialwas anaK/edlor a\
\
t\\pieal surface flaw (0.150-m. long X 0.0~5-indeepi\
\
and for a i>pical corner Haw (0.050 X 0.0501 at the\
\
edge of a penetration hole\parametric stud) with\
\
various stress levels was made for both Haw con\
\
figurationsusingNHSmaterialdataandthe\
\
FLAGRON analysisprogram.TheanaKsis in\
\
dicated that the required four design lives could be\
\
obtainedwith a stress level of 90 ksi in the basic\
\
plat- material and wiih a stress level of 43 ksi at\
\
t>pis.al penetrations.\
\
The3I6Lweldmaterialwasanah /edfor\
\
ivpical surface flaws, thatwere 0.150-m. long and\
\
0.075-in.deep.However,duringthebaseline\
\
analysis, there was no Haw growth data available\
\
forthis material. As it uasfellthaithe charac\
\
teristics of F.3I0 filler would be \\er\ similar to those\
\
of316L. the analysis was performedusingF310\
\
filler (law-growthdata. An anaKsis u>.ing 304LN\
\
flaw-growthdata was also performed,resulting in\
\
design life stress of 90 ksi with 304LN flaw-growth\
\
data and100 ksi with I£31() data.\
\
76\
\
FK;. 78.MKTF supports are designed for magnetic,\
\
rods. \ o s .3-7are stabilizer rods.)\
\
C O N C USIONS\
\
TheMt-TI-baseline -analysis confirmedthe\
\
basic structuralintegrityof magnet, and that the\
\
support system mei ASME factor of safety criteria.\
\
Cooldown thermal stresses were acceptable for the\
\
cooldoun time of 3.6 days which met design objec\
\
tives, fracture analvses showed acceptable magnet\
\
I,and seismic loads.(RodNos.1 and 2 are support\
\
lifeatihe designstresslevelofHOksi. Analysis\
\
showed that stress levels in the magnet were within\
\
the SO-ksi design limit everywhere except for several\
\
locations in the minor radius where peak stresses\
\
reachedS6 ksi.A subsequentrefinementofthe\
\
finite analysis which is discussed elsewhere in this\
\
reportandis documentedin Ref. S3 reduced the\
\
calculated peak stress levels to less than 80 ksi.\
\
77-78\
\
SECTION9\
\
STRUCTURAL FINITE ELEMENT ANALYSIS REFINEMENT\
\
SUMMARY\
\
The baseline structural analysis of the MFTF\
\
magnetconfirmedthe structuralintegrity of the\
\
magnet. However, the peak stresses at several loca\
\
tionsin themagnetmirrorradiusexceededthe\
\
allowable design limits of 80 ksi by 2 to 6%. These\
\
stresses were determined by a GDSAP finite ele\
\
ment analysis and are documented in Ref. 10. It was\
\
recognizedat.:.ecompletionofthebaseline\
\
analysis that a refined analysis of the critical stress\
\
regionusingtheNASTRANprogramwould\
\
probablyreducethecalculatedpeakstresses.\
\
Therefore, a revised finite element analysis of the\
\
MFTF magnet was conducted and documented in\
\
Rcf. 83. This analysis incorporated a refined mesh,\
\
updatedmaterialthicknessesandaNASTRAN\
\
plate elementthat accounts for the outof plane\
\
shearflexibilitynotrepresentedintheGDSAP\
\
analysis.\
\
MODEL DESCRIPTION\
\
For this analysis the case, jacket, and ini\_;coil\
\
structureweremodeledbyisoparametric,\
\
quadrilateral, plate elements (Fig79). The conduc\
\
tor pack was modeled by "lumping" the stiffness of\
\
the individual conductor strands into six equivalent,\
\
continuous, axial rod elements. Rod elements in the\
\
Magnet\
\
case\
\
•Quarter symmetric model\
\
•Plate elements (separate jacket & case\
\
•Conductor pack modeled\
\
•Reflects current geometry and stiffness\
\
-Case\
\
plate eleme' ts\
\
-Jacket\
\
plate elements\
\
-Conductor\
\
axial elements\
\
(b) Typical model cross section\
\
PIG. 79.Finite element model—typical cross section.\
\
79\
\
transverse directions were aKo i::ck;dcd ui mode!\
\
theconductorpackstiffnessinthosedirections\
\
(Fig. 80).\
\
Theloadingconditionsanalyzedincluded\
\
'hoseloadsduetoelectromagneticforces,the\
\
residual thermal loads after eooldown to 4 K. the\
\
internal pressure load resulting from a quench, and\
\
the critical combination of noima! operating elec\
\
tromagnetic loads and residual thermai cooldown\
\
U ads. The electromagneticloaddistributionwas\
\
determinedb\LLN'L anduaspropor'ionedac\
\
cording to the nodesofthe conductor pack.\
\
The structuralanahsiswas accomplishedin\
\
two steps. First, the haseline analysis (Ref. 79), us\
\
ing the G D S \ P finite element program, identified\
\
those area-, where peak stresses exceeded the KO-ksi\
\
allowable stress level. This allowable stress level ac\
\
counts for a ,'actor of safen of 1.5 on \\ield strength\
\
(120ksi) for the 304LN ease material. Regions oi\
\
criticalstress were foundin the inner andinter\
\
mediate crossover plates in the minor radius sec\
\
tion.\
\
The model refinements included: increasing the\
\
numberof elementsinthe min^rradius region.\
\
changing the case plate thickness from 3.00 in. lo\
\
.1.20 in. (to account for the actual thickness of the\
\
received case material), and modeling the offset of\
\
the mid-surfaces at the transition from 3.20-jn. to\
\
5.00-in. crossover plate (at the transition from ma\
\
jor to minor radius) (Fig\*. 81 and 82). Also, in addi\
\
tion to a GDSAP analysis, the model was analysed\
\
usingtheNASTRANUniteelementprogram.\
\
NASTRAN accounts for transverse shear stiffness\
\
in plates and also utilizes some higher order ele\
\
ments. The refined model has fewer triangular ele\
\
ments than the baseline model and no highly dis\
\
torted quadrilateral elements.\
\
Coil\
\
one\
\
"-is&ir\
\
Rod elements modeling\
\
conductor pack\
\
Complete model -plate elements\
\
IK;. SO.Refined finite clement model—c»se structure and conductor pack.\
\
SO\
\
InnerANALYSIS RESULTS\
\
Narrow row of\
\
inclined elements\
\
at transition^\
\
Inner case\
\
plate\
\
Fid. HI.Refined model accounts for nonalignment\
\
of neutral surfaces at transition from 3.2-in.-thf-;k\
\
plate to 5.0-in.-thick plate.\
\
The maximumprincipal stresses predicted in\
\
the criticalregionsoftheminorradiusshowed\
\
reductionsofWlo8%forNASTRAN and 07 to\
\
1%for GDSAP from those predicted b\ the baseline\
\
model.Thelargeststressesoccurredwithelec\
\
tromagnetic loading applied by itself.\
\
ForthisloadingtheNASTRANanalysis\
\
found the peak principal stress of SI.5 ksi occurring\
\
ul 45°fromthe axis ofsymmetryintheinner\
\
crossover plate of the minor radius {Fig. S3). This\
\
was the only location showing a stress larger than\
\
the 80-ksi allowable stress level for the NASTRAN\
\
analysis. The GDSAP anaKsis found stresses above\
\
80 ksi over a 25° span with a peak stress of S3.4 ksi\
\
(Fig. 83),Thesestresses represent reductions of 77\
\
byNASTRANand57b>GDSAPfromthe\
\
previously predicted peak principal stress of Ks ksi.\
\
GDSAPstressesalsoexceededtheSO ksi:\
\
e\ el\
\
slight!) lor the intermediate crossover plnte ove; i\
\
10° sp;;n at the axis of symmetry with a peak stress\
\
of81.$ ksi(Fig. S4).NASTRANpredictedall\
\
stresses below SO ksi in this area.\
\
V ^\>\} T\
\
Transition being modeled\
\
Neighboring elements in tension\
\
Desired deformation if unconstrained\
\
Sid? plate resist longi\
\
tudinal bending at edges\
\
-Positive\
\
longitudinal\
\
bending\
\
moment\
\
Actual deformations which induce transverse bending\
\
moments due to side plate constraints\
\
Longitudinal\
\
direction\
\
Positive transverse\
\
bending moment\
\
(NASTRAN)\
\
PIC, 82.Schematic of behavior at transition from 3.2-in. In 5.0-in. crossuvcr place.\
\
81\
\
\\^N\
\
90\
\
80\
\
70\
\
60\
\
50\
\
40\
\
30\
\
20\
\
10\
\
0\
\
Major radius0\
\
Maximum allowable stress —y\
\
4>\
\
Minor\
\
radius\
\
Baseline GDSAP\
\
analysis\
\
Refined NASTRAN\
\
Refined GDSAP\
\
.0\
\
Major\
\
. radius\
\
sym\
\
\_L\_L\
\
20406075\|02545\
\
Perimeter position (degrees)\
\
6585\
\
S\
\
1\
\
FIG. 83.Principal stress distribution—inner crossover plate—electromagnetic load only.\
\
\
\
Minor\
\
radius\
\
6075025\
\
Perimeter position (degrees)\
\
Refined NASTRAN\
\
Refined GDSAP\
\
85\
\
^\
\
I\
\
FIG. 84.Principal stress distribution—inner crossover plate—combined electromagnetic and residual thermal loads.\
\
Whenthe residual thermalloads due to the\
\
cooldownfromroomtemperatureto4.5 K\
\
operatingtemperaturearesuperimposedonthe\
\
electromagneticloads, a.slight reduction of stresses\
\
is oburned (Figs. 85 and 86). It is important to note\
\
for this load combination that NASTRAN showed\
\
all stresses in the magnet to be below the 80-ksi\
\
design allowable stress level. GDSAP predicted a\
\
peak principal stress of 81.2 ksi. with a span of 10°\
\
in the inner crossover plate at stresses over 80 ksi\
\
(I-'ig.H5).The intermediate crossover plate, for this\
\
load combination, showed a peak principal stress of\
\
80.6 ksi at the axis of.symmetry (Fig. 86;.\
\
Discontinuitiesare evidentin the stress dis\
\
tributions predicted with the refined model al the\
\
transition from 3.20-in.- to 5.00-in.-ihiek crossover\
\
plate material. These discontinuities had not shown\
\
up in the baseline analysis. The cause of the discon\
\
tinuities was that the refined model accounted for\
\
the offset of the mi Surfaces al the transition from\
\
3.20-in. to 5.00-in. plate material. This offset, in\
\
duces both longitudinal and transverse bending mo\
\
ments as is shown schematically in Fig. 82. These\
\
discontinuities, which are discussed in more detail\
\
in Rcf. 83, were found to not have any critical effect\
\
on the stresses.\
\
CONCLUSIONS\
\
The refined finite element analysis verified the\
\
structuralintegrityof theMFTFmagnet system\
\
with a factor of safety of 1.5 on yield stress for the\
\
304LN case materia). This verification is based on\
\
theNASTRANfiniteelementprogramforthe\
\
worst compatible load case (normal operating con\
\
dition)ofelectromagneticplusresidualthermal\
\
loading.\
\
84\
\
Baseline GDSAP\
\
analysis\
\
Refined NASTRAN\
\
Refined GDSAP\
\
Perimeter position (degrees)\
\
CIY(b)\
\
n o . 85.Principal stress distribution on intermediate crossover plate--electromagnetic load only.\
\
V^N\
\
Refined NASTRAN\
\
Refined GDSAP\
\
025\
\
Perimeter position (degrees)\
\
f l\
\
/\
\
.0\
\
Major\
\
. radius\
\
sym\
\
FIG. 86.Principal stress distribution on intermediate crossover plate—combined electromagnetic and residual thermal loads.\
\
SECTION 10\
\
STRUCTURAL CASE FAULT ANALYSIS\
\
INTRODUCTION\
\
A case faultanalysis documentedin Ref. 84\
\
was performedto assess the criticality of selected\
\
failuresin the magnetcase, jacketandinlercoil\
\
structure. The locations of the five failures selected\
\
for analysis wr\
\
ere determinedonthe basis of the\
\
manufacturingprocesses, structural geometry and\
\
local stress levels at these locations and, ultimately,\
\
on the possible consequences of the failures. Since\
\
this task followed the program to confirm the struc\
\
tural integrity of the MFTF magnet case, jacket and\
\
shield system(Rtif.10), every effortwas made to\
\
utilize analysis methods and tools developed earlier.\
\
The large, 7000-degree-of-freedomGDSAP model\
\
developed during that program was used extensively\
\
throughoutthisanalysistasktodeterminethe\
\
overall stresses anddeflectionsin the vicinity of\
\
assumedfaults. Sub.;jquentto thefinite element\
\
analyses, hand analyses were used to determine the\
\
stresses normal to the approaching fault front. Us\
\
ing this stress data a Mode I linear elastic fracture\
\
analysis was performed to quantify the criticalily of\
\
eachfault.\
\
ANALYSIS METHODS\
\
A common methodology was used to analyze\
\
each of the five selected faults. The analysis used' lie\
\
7000-degree-of-freedomGDSAPmodeltodeter\
\
mine the overall changes in stresses and deflections\
\
causedbytheassumedfaults.Datafromthis\
\
analysiswasthenusedtodeterminethelocal\
\
stresses normal to the fault front and a simple Mode\
\
Ilinearelasticfractureanalysisquantifiedthe\
\
criticality of the faults.\
\
The7000-degree-of-freedomfiniteelement\
\
model used in the fault analysis is basically the same\
\
as the model used in the original MFTF analysis\
\
and documented in Ref. 10- However, several minor\
\
changes that are documented in Ref. 84 were made\
\
to the model prior to the Mart of the fault analysis.\
\
The faults were simulated in the GDSAP analysis\
\
by additional nodes coincidental to existing nodes\
\
in the baseline finite element model and changing\
\
fault boundary plate connectivities. No attempt was\
\
made to refine the finite element mesh adjacent to\
\
thefaultboundariessubsequenttothe analysis.\
\
Model deflectiondatawas usedto calculatethe\
\
local slresses normal to the assumed faultfront.\
\
A simpleModeI fractureanuKsis was per\
\
formed using the stresses calculated from the finite\
\
element data to determine the criticality of the fault.\
\
The fracture analysis used the stress intensity solu\
\
tions for a finite crack in an infinite plate and for a\
\
0.5-in. edge crack in an infinite plate to approximate\
\
the stress intensities resulting from the faults. The\
\
expressions for the stress intensities in these cases\
\
are:\
\
Kj =\
plate (Ref. S5)\
\
K \| =1 .J2I5fT\\/7ra0.5-..edgecrackinin\
\
finite plate (Kef. ,S5J\
\
wherea =nominal stress, normal to fault, and\
\
a = half-cracklength.\
\
The stress intensities resulting from the anahsis\
\
werecomparedtotheplane-slrainfracture\
\
toughness (Ku)datausedin the originalMIT1\
\
fracture analysis. This data i^ shown in Table 20.\
\
ANALYSIS RESULTS\
\
A total of five fault conditions were an;il\\/ed\
\
duringthisstudy.Isometricsketches,selection\
\
rationale,andanalysisresultsaresummarized\
\
below for each of the following faults:\
\
1.Minor radius intermediate 3- to 5-in. plate\
\
intersectioninthe chamfer region atcaucrline of\
\
symmetry.\
\
2.Conductor pack jacket close-outweld in\
\
the major-to-minor radius transition.\
\
3.Inlercoil member shear transfer weld, bot\
\
tom closure plate to side plate corner weld.\
\
4.The 3-in. close-out weld at the center line\
\
of symmetry in the minor radius at the bottom of\
\
the 9 section.\
\
5.The 3- to 5-in. transition butt weld joint at\
\
the major to minor radius transition on the conduc\
\
tor case top plate.\
\
Fault number ! is shown in Fig. S7 along with\
\
the rationale for selecting it for analysis. In the finite\
\
87\
\
TABI.K211.I'aultfracture analyses Here based on preliminary material data used during the MFTF design.\
\
1II4I.N IN.12';N \]a\
\
I JK.I« d d filler\
\
. l i e4.5"\
\
i t i-v.2a\
\
I llll-ll. » d dfiller'1\
\
I MSIIHIS. S \ l \ «I\
\
Temp.\
\
K\
\
4.5\
\
4.5\
\
Case-material fracture-analvsis properties\
\
ksi\
\
lit.\
\
!2H\
\
da/dn\
\
(low groMih rate.\
\
u i n . / o c l t\
\
1.(154 XId"4\
\
(\_iK ,-T\
\
-5 M\
\
anNot mailable\
\
15Ni>t available\
\
11164.9HAHI-\
\
\\\\ \|AK\
\
b\
\
s«rar^< Kvf. H7.\
\
L'k'iiicni• , ! ^ \- i - .coincidentgridpointswerein\
\
stalled,iiI!K' Jmnilerregion. C'ontinumisconnec\
\
t si t ,'•.!-; "•ided for m the 3-in. pl;itc, while con-\
\
:IL-L;i\ n \mthe^ in. wasinterruptedthroughout;i\
\
."••'A\\-(,ippni\\jinatel\_\6.82in.J.\
\
\ i K i l \ s i so fthe5-in.-platedelleeliondatain-\
\
.ite>' h.nI hestressfieldn o r m a ltoI heap-\
\
p r n . K l i i i i ; :1 n u l lfrontis a p p r o x i m a t e l y25,880psi,\
\
n e u i e c l n i gbending.Assuming an infinite plate w i t h\
\
af i n i t ec r . i c k .\
\
Kj=2^Ml\\.'(\\H2TTpsiv'Tn.(10)\
\
Bav.-.! ,tn the haseline anahsisreportedKjt-\
\
\\.due lot I M6weldnilerUI-e = 9.2)al4.5°Kor95\
\
ksi ^ n... a eoniparisonof calculatedand critical\
\
stress iniensit} values indicates that this fault would\
\
continue to propagate under static loading until a\
\
lower -.tress field is encountered or a crack arrest\
\
mechanism is encountered. However, current NBS\
\
testing predicts a minimumfracturetoughness in\
\
the weld metal of 140 ksi yin. Based UII this data,\
\
the assumed fault would not propagate under static\
\
loading hut would continue to grow under cyclic\
\
loads until failure.\
\
Fault number 2 and its selected rationale are\
\
shown in Fig. 88. For the analysis of fault number\
\
2\. coincident grid points were installed in the model\
\
along the entire 0.5-in. plate "L" section. It was\
\
assumed that the transition from 0.5-in. to 1.0-in.\
\
jacketplatingwouldactasacrackarrest\
\
mechanism. Continuity was provided for an all case\
\
plating while continuityofthe 0.5-in. jacket plate\
\
was interrupted al the plane in question.\
\
L\\aliialion of the finite element analysis results\
\
indicates that the peak tension stress normal to the\
\
assumed fault occurs at the bottom of the "L" sec\
\
tion where the stress is 40,200 psi. This stress results\
\
in a Mode I stress intensity of:\
\
K\| = SOksiv/TrT\
\
Because this stress intensityis less thanthe\
\
predicted KK-for the 316 plate material, this fault\
\
wouldnotpropagatestatically. Thefaultwould\
\
propagate to failure under cyclic loading. But. the\
\
analysis indicates that a total failure of the jacket\
\
plate would have little effecton the overall struc\
\
tural case stresses.\
\
Fault number 3 is illustrated in Fig. 89. For the\
\
analysis of this fault, the finite element model was\
\
modified by installing coincident grid points along\
\
the intercoil member fault corner. The plate connec\
\
tivity was altered so there could be no load transfer\
\
between the bottom and side plate locally along the\
\
faultperimeter.\
\
Theanalysisresultsindicateverylittle\
\
redistributionofstress as a result of the assumed\
\
failure.Because the stresses were primarily com\
\
pression, however, the failure mode of concern was\
\
not Mode I fracture. The primary concern was the\
\
precipitationof an instabilityfailure in the large\
\
plates. However, an updated stability analysis of the\
\
intercoi! plates with revised boundary conditions to\
\
simulate the cracked weld predicted buckling failure\
\
stressesinexcessofthepeakintercoilstresses.\
\
Based on this analysis, there were no significant ef\
\
fects due to the assumed fault.\
\
Joint geometry\
\
Selection rationale\
\
•Weld joint\
\
-Complex unbalanced weld-on-weld joint\
\
-Heat sink gradient\
\
•Joint geometry\
\
-Chamfer region increases welding\
\
difficulty but has no significant effect\
\
on gross stress distribution\
\
• Stress\
\
-Plate membrane stresses parallel to this\
\
weld joint approach 80 ksi\
\
-Principal stresses in the 5-in. plate\
\
approach SO ksi\
\
• Concern\
\
-Loss of the 5- to 3-inch joint (basically\
\
a shear transfer member) may result in\
\
excessive case deflection and subsequent\
\
conductor pack crushing\
\
2,0-in. plate\
\
2.0-in. plate\
\
3.0-in. plate\
\
2.0-in. plateConductor\
\
pack\
\
I'IG. 87.Fault No. 1: minor radius intermediate J- to 5-in. plate intersection in the chamfer region at the ccntcrline of symmetry.\
\
Joint geometry\
\
Selection rationale\
\
•Weld joint\
\
— Close-out weld increases weld inspection\
\
difficulty\
\
•Joint geometry\
\
— Joint oriented normal to significant plate\
\
membrane stresses\
\
•Stresses\
\
— Principal stresses in the area of the joint\
\
approach 50 ksi\
\
•Concern\
\
— Decreased weld inspectability increases\
\
probability of potentially critical flaw\
\
existing\
\
— Previous stress analysis assumed that the\
\
jacket plating contributed to the case\
\
stiffness. Jacket failure will increase the\
\
primary case plate stresses\
\
3.0-in. plate\
\
Fault\
\
(0.5-in. plate)\
\
5.0-in. plate\
\
Section plane at\
\
major-to-minor\
\
radius transition\
\
3.0-in. plate\
\
Vacuum\
\
guard\
\
FIG. 88.Fault No. 2: conductor pack jacket close-out weld in the major-to-m nor radius transition.\
\
G\_ Sym coil no. 2\
\
minor radius\
\
Selection rationale\
\
•Weld joint\
\
— Joint geometry - no significant effect\
\
• Stress\
\
— Compressive only\
\
• Concern\
\
— Loss of joint may cause intercoil\
\
member instability and system\
\
catastrophic failure\
\
4.0-in. plate\
\
wedge\
\
Joint geometry\
\
KIG. 89.Fault Nr.3:intercoil mcmbei shear transfer weld, bottom closure plate to side plnte corner weld.\
\
I-aull numher 4, which is shown in Fig. 90, was\
\
a ..umed to extend fromthe minor radius center line\
\
of\\\mmetr>througha 20° arc (half length=6.82\
\
in.j.Coincidentgridpointswereinstalledinthe\
\
nasch:e model along existingfaultlire gridpoints.\
\
Conlinuitxofthe3-to5-in.-platejointwasin\
\
terruptedb \redefiningexistingC Q l . ' A Delement\
\
LIH! pointL'onneclnilie^.\
\
DeflectiondatanearI heapproachingfault\
\
frontwas extractedfromllic computeroutputfor\
\
use in a ModeI traclurc analysis. I r o mthis deflec\
\
tion data the local plale membrane stress was delr\
\
minedlobe approximate!}r.667psi.1-rom!i,\
\
plague niCLlianics thcors. for an infinite plale w\
\
Unite cenleicrack.\
\
5667 \6.82-26.23 ksiyin.(ID\
\
This stress indicates thatalthoughthis crackis not\
\
catastrophic,itwillcontinuelopropagateunder\
\
cyclicloadinguntilfailureoracrackarrest\
\
mechanismisencountered.\
\
Fault number 5 was the most critical of the live\
\
failuresthatwereanalyzed.Thefault,whichis\
\
shownin Fig. 91, causeda significantreductionin\
\
the overallbendingstrengthof the magnetcase in\
\
the minorradius anda resulting increasein stress.\
\
The finite elementresults indicated a significantin\
\
creaseinstressesapproached99,300psi.Atthis\
\
stress le\\el. the stress intensityfor a 5-in. edge crack\
\
in aninfiniteplale is:\
\
K, =1.22 X 9 9 3 0 0 ^ 2 ^ -= .340ksi y m".112)\
\
ThisKjreflectsacatastrophicfailurewhencom\
\
pared to the critical stress intensities lor 3C4I.N and\
\
3161.welds.\
\
CONCLUSIONS\
\
Resultsof(hisstudyaresummarizedin\
\
Table 21. Oftheliveconditionsanalyzed,oneis\
\
consideredto have structurallycatastrophiceffects;\
\
catastrophic meaning these faults would be reflected\
\
b\suddenandviolentchangesin the geometryof\
\
thestructure.Thefaultconsideredstructurally\
\
catastrophic ( M o d e I analysis) was FaultNo. 5 (the\
\
.3- to 5-in. transition butt weld joint at themajor-to-\
\
minorradiustransitiononthe conductorcasetop\
\
plate).\
\
Faultnumber 4, the 3-in.-plaleclose-outweld\
\
at the center line of symmetryin the minor radius at\
\
thebottomofthe9section,is consideredtobe\
\
TABLE 21.Case fault analysis summary.\
\
Faultdescription\
\
Minor iadius intermediate 3- to 5-in.-plateintersection\
\
in tliu chamferregion at the center line of symmetry\
\
K\| equals K\[(-. Fault will be self propagating.\
\
Catastrophic failureassumed.1 1\
\
Conductor pack jacketdose-outHeld in themajor-to-\
\
minor radius transition\
\
50 ksi v/\
\
hTK\| lower than K \| f .Fault will propagate under\
\
cyclicloading.Overallsystemintegritynot\
\
significantlyaffected.\
\
Imercojlmembersheartransferweld,bottomclosure\
\
plate lo side plate corner weld\
\
M.S.calculatedassumingcompressive-\* icld\
\
allowable of 12(1 ksi.\
\
3-in.-pIate close-out weld at the center line of symmetry\
\
in the minor radius at the bottom of die 9 section\
\
26 ksi\\/in.Kj loner than Kjf.Fault will propagate under\
\
cyclic loading.\
\
53 - to 5-in.-transition butt weld joint at the major to minor441 ksi\\/\\n.\
\
radius transition on Ihc conductor case t<-- jilate\
\
KjhigherthanK\\£.Faultwillbeself-\
\
propagating. Catastrophic failureassumed.\
\
3\
\
Updatedfract.:emechanics data ( K j f% J4M ksi\\/\\n.)Indicates that thi« fault is not initially catastrophic. However, it will propagate\
\
quickly under cyclic load lo the failure poinl.\
\
92\
\
Selection rationale\
\
•Weld joint\
\
— Close-out weld; increases weld\
\
inspection difficulty\
\
• J o i n t geometry\
\
— Not significant\
\
• Stress\
\
— Moderate plate longitudinal membrane\
\
stress\
\
•Concern\
\
— Decreased weld inspectability increases\
\
probability of potentially critical flaw\
\
existing. Joint failure and subsequent\
\
redistribution of stresses may result in\
\
excessive deflections and conductor pack\
\
crushing\
\
2.0-in. plate\
\
2.0-in. plateConductor\
\
3.0-in. plnte\
\
5.0-in. plate\
\
-JFaulti\
\
(1=6.82 in.)-1\
\
FIG. 90.Fault No. 4: 3-in. plate close-out weld at the eenlerline of symmetry in Ihe minor radius at the bottom of the 'I section.\
\
5.0-in plate\
\
Selection rationale\
\
•Weld joint\
\
— One sided butt joint (unbalanced)\
\
— Weld bead reinforcement (K-r)\
\
— Heat sink gradient\
\
•Joint geometry\
\
— 3- to 5-inch plate transition produces\
\
local stress raiser\
\
•Stress\
\
— Plate membrane stresses normal to this\
\
joint appr. -ach 60 ksi\
\
— Transverse secondary bending moments\
\
approach 95 in.- kips/in.\
\
•Concern\
\
— Joint failure causes a significant\
\
reduction in magnet section modules\
\
resulting in redistribution of stresses\
\
and possible system catastrophic failure\
\
5.0-in to 3.0-in plate\
\
transition joint\
\
fault (\
3.0-in plate\
\
Coil extension\
\
structure\
\
tW'Joint geometry\
\
FIG. 91.Fault No. S: major-to-minor radius transition, 3-in, lo 5-in. buil-»cld jointonthe conductor case top plate.\
\
potentially catastrophic; i.e., the calculated stress\
\
intensity factors are lower than the critical stress in\
\
tensity factor but under cyclic loading this fault will\
\
continue to propagate until repaired or a suitable\
\
crack arrest mechanism is encountered.\
\
Fault number1, the 3- to 5-in.-plate intersec\
\
tion in the chamfer region of the minor radius, is\
\
potentially catastrophic, evenwithupdatedNBS\
\
fracture toughness data. At the calculated Kj of 120\
\
ksiy/\\r\\.,the stress intensity is less than the KI Cof\
\
140 ksi\\firT.measured for the 316L weld. This in\
\
dicates tha. the fault would not fail immediately but\
\
wouldcontinueto growunder cyclic loaduntil\
\
failure.\
\
95-96\
\
SECTION11\
\
STRUCTURAL METALLURGY\
\
INTRODUCTION\
\
Following the BaseballII magnet experience\
\
Nilronic40was first selected for the MFTF magnet\
\
structure. However, plate of sufficient size was un\
\
available, and welding across major stress planes\
\
was necessary. Accordingly, an effort was made to\
\
developaweldprocedureusingInconel625: a\
\
nickel-base alloy, which, it was hoped, could match\
\
the parent metal strength and toughness. Because\
\
poorfracturetoughness of both the Inconel 625\
\
weldand the Nilronic40 basemetalprecluded\
\
design stresses above 80 ksi, it was decided to aban\
\
don the this approach and use a tougher and less ex\
\
pensive austenilic stainless steel, 304 LN, with a\
\
ferrite-free 316L weld metal. In this section the ex\
\
periencewithbothbase-metalsystemsare\
\
described.\
\
INITIAL WELDDEVELOPMENT\
\
Initialweld developmentprograms were ori\
\
entedtowarddevelopmentof procedures to join\
\
Nitronic-40 stainless steel for the casestructural\
\
material, whose nominal composition is Fe-21%Cr-\
\
6%Ni-9%Mn (Refs. 87-90). Fusion-welding of thick\
\
sections of this alloy (thicker than 1/4-in.) had been\
\
greatly hampered by the inability to develop a weld\
\
filler metal that matched both the base metal's high\
\
yield strength (5\*180 ksi) and moderatefracture-\
\
toughness («75 ksi v/IrT.) at 4 K (Refs. 87-90). The\
\
filler metals selected for evaluation were a modified\
\
Nitronic 40, with nominal composition Fe-20%Cr-\
\
7%Ni-9.5% Mn, and a nickel-base alloy, known as\
\
"Alloy625," withnominalcompositionsNi-\
\
22%Cr-9%Mo-3.5%Cb-3.5% Fe. The latter material\
\
was selected on the basis of its known good 77 K.\
\
ductility as a weld filler-metal for joining both stain\
\
less steels and nickel-base alloys (Ref. 91). Welding\
\
processes evaluated were shielded-metal arc (SM A),\
\
gas-metal arc (GMA), gas-tungsten arc (GTA), and\
\
submergedarc(SA). Allfourprocesseswere\
\
evaluated using the Alloy 625 filler, while modified\
\
Nitronic filler was used with the GMA and GTA\
\
processes.\
\
MODIFIED NITRONIC-40\
\
FILLER METAL\
\
Details of welding are presented in Ref. 92 and\
\
summarized in Table 22 of this report \[see informa\
\
tion tabulated for welds No. I (GMA) and No. 2\
\
(GTA)\].Weld-metalchemical compositions were\
\
typicalfor thismaterial(Table23). Weld-metal\
\
CharpyV-nolchimpacttests performedat 77 K\
\
were used as a screening tool to avoid extensive and\
\
expensive 4 K lesting of many specimens. Results of\
\
77 K charpy impact tests are presented in Table 24.\
\
Average energy absorption values were 17.5 ft-lb for\
\
the GTA weld and 27.0 ft-lb for the GMA weld, as\
\
compared with typical annealed-base metal values\
\
of 60-70 ft-lb at 77 K (Ref. 90). Microscopic ex\
\
amination of both welds showed the usual duplex\
\
austenite-ferrite microstructure with ferrite contents\
\
of 4.5% for the GMA weld and 8% for the GTA\
\
weld. The disappointing 77 K Charpy V-notch im\
\
pactperformanceconfirmsthe SMA results pre\
\
sentedatthe October1977 VailWorkshop on\
\
Structural Materials for Low Temperature Service,\
\
(Ref. 90),andfurtherworkonevaluationof\
\
modified Nitronic-40 weld metal was discontinued\
\
in favor of increased efforts on the Alloy 625 weld\
\
metal.\
\
ALLOY 625 WELD METAL\
\
Detailsofweldmanufactureandresulting\
\
chemicalcomposition,microstructure,and\
\
mechanical-property performance are discussed in\
\
detailinRefs. 93and94 andsummarizedin\
\
Tables 22 through 26 of this report (see information\
\
tabulated for welds Nos. 3-6 for the SMA, GTA,\
\
GMA,and SA welds, respectively). Welds were\
\
madein2-in.-thickNitronic-40plate,heavily-\
\
restrained by welding the free edges of the plates to\
\
a 4 to 6-in.-thicksteel plate, to simulate the high\
\
degree of restraint that would be imposed on weld-\
\
ments in the actual case structure. Chemical com\
\
positionswereusualforAlloy625 weldments\
\
(Table 23). Welds were evaluated for soundness by\
\
radiography,whichrevealednounacceptable\
\
97\
\
T A B L E 22.Welds made in the multiprocess study.\
\
Process\
\
Flux or\
\
shielding gas\
\
Electrode\
\
WcM joint\
\
Process\
\
Flux or\
\
shielding gasFiller\
\
Diam\
\
(in.)SupplierHeat No.Lot No.\
\
WcM joint\
\
No.Process\
\
Flux or\
\
shielding gasFiller\
\
Diam\
\
(in.)SupplierHeat No.Lot No.TypeOriginPositionmetal\
\
1Gas-metal arcArgon +2% O2Nitronic 400.030Armco91520DouMe-VLLNLFlatNitrouic 40\
\
2Gas-tungsten arcHeliumNitronic 400.030Armco91520Double-VLLNLFlatNitronic 40\
\
3Shielded-melal arcAlloy 6255/32IncoButtLLNLFlatNitronic 40\
\
4Gas-tungsten arcHeliumAlloy 6251/8IncnNX 851 iButtLLNLFlatNitrouic 40\
\
5Gas-metal arcArgon +2%O2Alloy 62S1/16IncoNX7478RuttLLNLFlatNitronic 40\
\
6Submerged arcArcos N82Alloy 6251/16IncoNX7478ButtLLNLFlatNitronic 40\
\
8Gas-metal arcArgon +2% O 2316L1/16UtiibrazeIF2-9201047ButtLLNLFlat304 LN\
\
9Submerged arcHobart HS300316L3/32Johnson17222ButtLLNLFlat304 LN\
\
10Flux-cored arc316 LT-33/32Stoody0388ButtLLNLFlat304 LN\
\
11ElectroslagArcos N82Alloy 6251/16IncoNX 7990ButtLLNLVertical304 L\
\
12ElectroslagArcos N82316L3/32UnihrazcIF2-83I1246ButtLLNLVertical304 L\
\
13ElectroslagArcos N823163/32UnibrazeIF2-8311246ButtLLNLVertical304 LN\
\
17Pulsed gas-metal arc75% He/25% Ar2RM691/16Sandvik743802Oouble-VCBI-HslnVertical304 LN\
\
20Pulsed gas-metal arc75% He/25% Ar316L0.045Double-UFMC San JoseFlat304 LN\
\
21ElectroslagHobart PT2033161.3/32McKay0463431821ButtFMC San JoseVertical304 LN\
\
TABLE 23.Chemical composition of welds made in the multiprocess study (wl%).\
\
Weld\
\
No.MoTiCb+liN2\
\
1Gas-metal arc0.0179.52\
\
2Gas-tungsten arc0.0179.52\
\
3Shieldcd-metal arc0.0460.25\
\
4Gas-tungsten arc0.050.09\
\
5Gas-metal arc0.020.24\
\
6Submerged arc0.020.24\
\
11FJcctroslag0.050.40\
\
12F.lcctroslag0.0301.70\
\
13F.lcdroslag0.0154.60\
\
17\
\
20\
\
Pulsed gas-metal arc\
\
Pulsed gas-metal arc\
\
0.14\
\
0.14\
\
0.58\
\
0.28\
\
11.14\
\
11.14\
\
11.30\
\
11.48\
\
0.17\
\
0.007\
\
0.007\
\
0.004\
\
0.004\
\
0.012\
\
0.012\
\
0.010\
\
19.93\
\
19.93\
\
21.48\
\
22.13\
\
11.01221.90\
\
0.01221.90\
\
20.90\
\
18.70\
\
O.tllO24.76\
\
7.17\
\
7.17\
\
Bal.8.850.063.600.05AU0.O4, Fe:3.06\
\
61.68\
\
60.34\
\
8.87\
\
9.56\
\
0.22\
\
0.27\
\
3.33\
\
3.39\
\
0.08\
\
0.04\
\
\1:0.001, Fe:3.06.\
\
Al:0.28\
\
Fc:3.91, AI:0.2I\
\
60.349.560.273.390.05B:0.001\
\
48.4116.800.15700.08Al:0.15, Fe:19.50\
\
12.702.10\
\
•M.4H2.14\
\
T A B L E 24.CharpyV-notch impactperformance at 77K atwelds made in the multiprocess study.\
\
Process\
\
Ferrite number\
\
VenderLI.I.\
\
Energy absorbedLateral expansion\
\
No.Process\
\
Ferrite number\
\
VenderLI.I.Average.\* ft lbRange,ft-lbAverage", milsRange, mils\
\
1Gas-metal arc4.517.516.5 -18.56.04.0 -8.0\
\
2Gas-tungsten arc8.027.724.5 -31.012.512.0 -13.0\
\
JShiclded-mclal arc37.037.0 -37.031.229.5 -32.8\
\
4Gas-tungsten arc45.343.0 - 46.(141.436.7 -46.8\
\
5Gas-metal arc\
\
6Submerged arc70.619.5 -21.518.912.8 -24.6\
\
8Gos-mctal arc5.S26.823.1\- 3 0 . 019.516.6 -23.4\
\
9Submerged arc5.526.921.5 - 31.(113.910.8 -18.4\
\
1(1Flux-cored arc13.1)14.19.8 -17.98.74 . 6 -11.6\
\
11Electroslag88.071.11 -II 1.057.351.0 - 66.4\
\
12Eleclroslag95.480.2 -110.575.263.6 - 75.4\
\
13Kleclroslag103.09 5 . 0 -II 1.0511.844.0 -57.5\
\
17Pulsed gas-metal arc068.263.(1 -79.055.049.0 -63.0\
\
20Pulsed gas-metal arc8.535.93 ! i> -42.522.119.0 -28.6\
\
21Elcclrustag5.034.928.0 -41.530.424.2 -38.0\
\
Average of 3 to 5 specimens.\
\
b\
\
I mil = 0.001 in.\
\
T A B L E 25.4K tensile and fracture-toughness properties of welds made in multiprocess study.\
\
Weld\
\
No.Process\
\
Ultimate\
\
strength,\
\
ksi\
\
Yield\
\
strength,\
\
ksi\
\
Elongation,\
\
Reduction\
\
in area.\
\
ksi vin.\
\
3Shielded-metal arc183.0b\
\
127.030.725.0122.8\
\
4Gas-tungsten arc173.3b\
\
138.018.015.0121.3\
\
5Gas-metal arc157.7b\
\
121.022.726.7136.0\
\
9Submerged arc157.9b\
\
97.260.039.0101.0\
\
11Electroslug1 0 7 . lb\
\
82.611.014.0\
\
12Electroslag2 0 4 . 1b\
\
74.739.926.2146.0\
\
13Electros! ag2(MUC\
\
65.1\
\
20Pulsed gas-metal arc200.5b\
\
118.348.038.5\
\
21ElectroslagI 7 0 . lb\
\
96.235.020.4\
\
193.7°85.835.027.9\
\
183.5d\
\
75.630.526.4\
\
" K Qis an invalid, noncMservtitive, measure of fracture-toughness due to inadequate specimen thickness or improper lest procedure.\
\
" L — Lo\*gitn4mal, parallel to weld axis.\
\
C\
\
L T-Long transverse, across weM axis.\
\
-\
\
ST—Short IraMverse, through weld axis.\
\
99\
\
TABLE26.4 K mechanical properties of 2-in.-thick Nitronic 40 plate used inthe alloy 625 welding study.\
\
Ultimate0.2% offset\
\
yield\
\
strength,\
\
ksi\
\
Elongation\
\
in 1 in..\
\
Reduction\
\
in area.\
\
Fracture-toughness\
\
strength,\
\
psi\
\
0.2% offset\
\
yield\
\
strength,\
\
ksi\
\
Elongation\
\
in 1 in..\
\
Reduction\
\
in area.Longitudinal,3\
\
KQ\
\
Transverse.\
\
ksi v/nT\
\
245\
\
251\
\
245\
\
198\
\
195\
\
196\
\
22\
\
26\
\
2(1\
\
35\
\
30\
\
25\
\
138\
\
1311\
\
131\
\
P.5\
\
136\
\
132\
\
Avenge247.0196.322.7.111.1113.1.(1131.(1\
\
a\
\
K n is a nonconservalive value orfraclurc-toughness, i.e. K Q > Kjc, the minimum-value or "plane-strain fracture-toughness." caused hy\
\
the specimen thickness of 0.5 in. being less than that required to ensure attainment -jf plane-strain conditions at the crack-tip during testing.\
\
defects, and by room temperature side-bend testing,\
\
whichplacestheentirethicknessof aslice cut\
\
through the weld in tension. Side-bend specimens\
\
cutfromSMA, GTA,andSA welds were bent\
\
around a 2t (t = specimen thickness) radius mandrel\
\
to an angle of 180° without any evidence of failure.\
\
However, the GMA side-bend specimens failed at a\
\
bendangleofabout30°.Bothmicroscopic ex\
\
amination and chemical analysis of the failed GMA\
\
bend-test specimens failed to reveal any cause for\
\
this poorperformance.Post-weldheat-treatment\
\
(PWHT)ofanotherGMAbendspecimenat\
\
2150°F for1 h, followedby testing as described\
\
above, resulted in this specimen passing the side-\
\
bendtest.Thisindicatedthatthecauseof\
\
premature failure of the as-welded GMA specimens\
\
was associated with the presence of a metallurgical-\
\
phase formedduring solidificationand/or during\
\
cooling to room temperature after welding, and that\
\
this phase could be removed or rendered innocuous\
\
by a high-temperaturePWHT.However,such a\
\
PWHTis notpracticalduringassemblyofthe\
\
massive case sections, and is certainly not a prac\
\
tical treatment on the case close-out welds once the\
\
magnet was sealed inside the care.\
\
Low temperature evaluation of the Alloy 625\
\
weldmenls consisted of 77 K Charpy V-notch im\
\
pact tests (Table 24) and 4 K tensile andfracture-\
\
toughness tests (Table 25). Notches in the charpy\
\
specimensandcracksinthefracture-toughness\
\
specimenswereorientedsothatthecracks\
\
propagated from the top to the bottom of the weld\
\
metal. Average Charpy V-notch test energy absorp\
\
tion values were 37 fl-lb for the SMA weld, 45.3 fl-\
\
Ib for the GTA weld, and 20.6 ft-lb for the SA weld.\
\
Whilebelowthe60ft-lbvalueforunwelded\
\
Nilronic 40 (Ref. 90). results for the SMA and GTA\
\
welds represented a distinct improvement over 77 K\
\
CharpyV-notchtestenergy-absorptionvalue.\*\
\
shownbythemodifiedNilronic40.veldmetal\
\
(Table 24).\
\
Tensile test results at 4 K of the Alloy 625 weld\
\
metalsaresummarizedin Table25 andofthe\
\
Nitronic-40 base metal in Table 26. All materials ex\
\
hibited satisfactory values of tensile ductility, with\
\
average elongation values for the base metal and\
\
SMA.GTA.andGMAweldmetalsof22.7r\
\
J.\
\
30.7%, 18.0%. and 22.7%. respectively. Average ten\
\
sile yieldstrengthvalues ofthebasemetaland\
\
SMA. GTA. and GMA weld-metals were 196.3 ksi.\
\
127.0 ksi.138.0 ksi. and121.0 ksi.respectively.\
\
Scanning electronmicroscopy examination of the\
\
fracture surfaces of representative tensile specimens\
\
showed no signs of brittle failure, with all fracture\
\
surfaces exhibiting a ductile fracture appearance.\
\
Fracture-toughnesstestresultsoftheweld\
\
metalsaresummarizedinTable27 andofthe\
\
Nitronic-40basemetalinTable26.Whilethe\
\
numerical values are adequate, with average values\
\
for the base metal. SMA. GTA. and GMA weld-\
\
metals of132.0 ksi v1rT 122.8 ksi \\/uT. 121.3 ksi\
\
v/in.. and136.0 ksi \\/irT-. respectively, further ex\
\
amination of the test specimens andtest records\
\
give rise to serious doubts about the validity of these\
\
values for the reasons presented below:\
\
I.While the range of numerical values of the\
\
weld metal fracture-toughnessforthe three weld\
\
processes are high (122.8 to136.0 ksi v'in.), ex\
\
amination of the fracture-toughnesstest methods\
\
100\
\
TABLE 27.Comparisonofminimumfracture-\
\
toughnessspecimenthicknessrequiredtoachieve\
\
plane-straintestconditionsinInconel62S weld\
\
metal.\*\
\
Wild\
\
process\
\
K,°\_\
\
ksiy/'m-ksi\
\
a, Bc a\|c,\
\
in.\
\
"meas'\
\
in.\
\
SUA12.11272..150.5\
\
CTA1211.181.920.5\
\
(,'MA1.16121.1.160.5\
\
lias' metal1.121 %1.1.1(1.5\
\
"From ASIA! K .VW-74. all »2.5 ( K / « y > \ where 3,H are Ihe\
\
wiitlh andthickness nf the fracluie-touRhnessspecimen,K Ihe\
\
measuredvalue nf stress intensity at fracture, andaythe yield\
\
strength.\
\
Average values used in calculations.\
\
Displacement !V)-\*\
\
FIG. 93.Typesofloaddisplacementtracesob\
\
tained during fracture toughness testing.\
\
indicatedthat the values determinedare noncon-\
\
servalive and use of these values for design would\
\
resultinover-estimationoftheseweldmetals'\
\
resistance to brittle fracture. These nonconservativc\
\
resultswereobtainedbecausethesizeofthe\
\
fracture-toughnesssamplesusedwereloosmall\
\
(Table 27 andRef. 95) to achieve conservative or\
\
"planestrain"values.Figure92schematically\
\
represents the variation of stress intensity factor K,\
\
with specimen or component thickness. For condi\
\
tions of plane strain, where the material thickness is\
\
great enough to constrain plastic flow (yielding) in\
\
the plane of a crack or other defect, K = KI C. For\
\
other than conservative or plane strain conditions,\
\
K = KQ > K) c. What had been determined for the\
\
Plane-stress conditions\
\
Plane-strain conditions\
\
" 0" 1\
\
Specimen thickness (B)\
\
FIG. 92.Effect of specimen thickness on fracture\
\
toughness.\
\
Nitronic 40 base metal and the alloy 625 welds was\
\
Kg.\
\
2.Inadditiontothenonconservative\
\
fracture-toughnessvaluesofthevariousweld\
\
metals,examinationoffracture-toughnesstest\
\
proceduresindicatedthattheformoffracture-\
\
toughnesslestload/displacementtracesindicate\
\
thatalithetestresultsexhibitedevidenceof\
\
catastrophiccrackpropagation.Representative\
\
load/displacementcurvesareshowninFig. 93.\
\
Referring to this figure, either Type 1 or Type 2\
\
behaviorisrepresentativeoi'acceptableservice-\
\
behavior, i.e.. propagation of a preexisting flaw re\
\
quires increasing load. However, Type 3 behavior,\
\
where once a flaw begins lo propagate, it continues\
\
to do so even with a decreasing load situation, is un\
\
acceptable from the standpoint of rational brittle-\
\
fracture-resistantdesign practices.\
\
3.Examination of the 4 K tensile test results\
\
shows that the Alloy 625 weld-metal yield strengths,\
\
irrespective of weld process used, fall in the range of\
\
117 to 143 ksi, or about 60-73% of the base metal\
\
value. Design of the magnet case requires that some\
\
of the welds be located in the primary load paths.\
\
Under these conditions, it is good design practice to\
\
matchthebase-metalandweld-metalyield-\
\
strengths to prevent localization of load and strain\
\
in the weaker weld-metals to avoid over-loading of\
\
the weld metal, early development of cracks, and\
\
rapid propagation of the small, preexisting flaws ex\
\
pected in any weld. At this time, late June 1978, it\
\
101\
\
wasrecognizedbytheMFTFmagnetproject\
\
engineer that reconsideration of the selection of the\
\
magnet case structure material and the associated\
\
welding development efforts was needed. Experts in\
\
theassociatedareasofstructuralmaterialsfor\
\
cryogenic service andwelding developmentfrom\
\
withinLLNLandfromotherorganizations\
\
(National Bureau of Standards at Boulder, General\
\
Dynamics/Convair,and the DOE) met atLiver-\
\
more on June 23, 1979 to reexamine the related\
\
issues of selection of the case structural material and\
\
associated welding development. The recommenda\
\
tionsofthisreviewcommitteearesummarized\
\
below and were adopted by the MFTF magnet proj\
\
ect engineer for implementation.\
\
By using as a guide the principle of selecting a\
\
base metahweld metal pair with equal 4 K tensile\
\
yieldstrengths,abasemetal4 Ktensileyield\
\
strengthnotlessthan1.5timestheGeneral\
\
Dynamics/Convairdesign stress of 80 ksi, and a\
\
4 K fracture-toughnessvalue of125 ksi vTrT, the\
\
followingpreliminaryrecommendationswere\
\
proposed:\
\
Type 304LN base metal :316L weld metal, or\
\
Type 316LN base metal :316L weld metal.\
\
Welds should be made by the GM A process for op\
\
timumcontrolof purity, 4 Kfracture-toughness,\
\
and ferrite content. Neither of these approaches en\
\
tails materials costs as high as the Nitronic 40:ln-\
\
conel 625 approach. The need for resolution of the\
\
one openissue, optimizationof weld-metal com\
\
position to prevent weld-metal microfissuring dur\
\
ingproductionoftherestrained,heavy-section\
\
welds (up to 3-in.- thick) in the actual case, versus\
\
limitationofferritecontenttoguardagainst\
\
degradation of 4 K weld-metalfracture-toughness,\
\
was recognized.\
\
WELD DEVELOPMENT\
\
PROGRAMS PERFORMED IN\
\
SUPPORT OF JOINING\
\
TYPE-304LN BASE METAL\
\
Once the choice of type 304LN stainless steel\
\
wasmade,threeseparatebutrelatedweld-\
\
development programs were begun to provide an\
\
initial evaluation of the suitability of the common\
\
welding processes, such as SMA, GMA, GTA, SA,\
\
FCMA (flux-coredmetal arc), and ESW (electro-\
\
slag welding), for deposition of type 3I6L stainless\
\
steelweldmetal.Evaluationofavarietyof\
\
processes was necessary to ensure the eventual selec\
\
tion of one or more processes that would enable\
\
out-of-position (other than flat position) welding to\
\
be done, as well as qualify a process, such as SA or\
\
ESW, that would enable large quantities of weld\
\
metal to be deposited in a short time with a con\
\
current reduction in welding time and costs (Table\
\
28). With this task essentially completed, and the\
\
SMA process selected by Chicago-Bridge and Iron\
\
Company (CBI), the successful bidder on the case-\
\
fabricationcontract,two parallelprograms were\
\
initiated. One of these programs involved qualifica\
\
tionofCBI'sSMAweldingproceduresusing\
\
TABLE 28.Estimated time required lo weld MFTF magnet case assuming 20,000 lb of weld metal deposited.\
\
Weld metal\
\
deposition data\
\
Shicldcd-\
\
mctal\
\
arc\
\
Gas-\
\
metalFlux\
\
core\
\
SubmergedFJectroslag\
\
(Note b)\
\
Deposition data\
\
100% arc time, Ib/h\
\
Est. arc lime. %\
\
Actual deposition rate, Ib/h\
\
Welding time\
\
Total hours to weld\
\
No. man-hours reqHiredc\
\
No. man-months required0\
\
2.59.013.014.515.0\
\
20.0.10.040.050.095.0\
\
0.52.75.207.2514.25\
\
40,0007,407.43,846.22,758.61.403.5\
\
24044.423.116.68.4\
\
203.701.921J80.7\
\
40.0\
\
95.0\
\
38.0\
\
526.3\
\
3.1\
\
0.26\
\
\*Per measured lab rale using a 4H0-A power supply (max). Nolsufficient.\
\
"Attainable with power supply of adequate capacity.\
\
c\
\
V a h e s base\* on available maa-hr/yr -(8 h/JHS d/wk)(5 wk/yr) =2000 maa-hr/yr/man.\
\
102\
\
E316L-15electrodes,andtheotherinvolved\
\
developmentofweldinginformationforuse in\
\
otherareasofmagnetfabrication,suchasthe\
\
magnet jacket, and for a backup weld melal should\
\
difficulties arise with the weld metal deposited by\
\
typeE3I6L-15 electrodes. The status of each of\
\
theseprograms,includingresultstodate,open\
\
issues, and ongoing work, is presented below.\
\
Multiprocess Study\
\
Welds were made by the SMA, GMA, SA,\
\
FCMA, and ESW processes in type 304LN stain\
\
less steel plate ranging in thickness from 1-3/4 to 3\
\
in. using type 3I6L weldmetalin the formap\
\
propriate to the welding process. Details of the joint\
\
designandweldingpositionaresummarizedin\
\
Table 22. Most of the welds were made at Liver-\
\
more, but two welds, one GMA and one ESW, were\
\
made at FMC (San Jose). The fabrication subcon\
\
tractor for the magnet case, CBI, supplied a GMA\
\
weld using a ferrite-free filler, 2RM69, for evalua\
\
tion by LLNL. All the welds made at LLNL were\
\
made under heavily-restrained conditions, i.e., the\
\
plates were first welded to a 4- to 6-in.-thick mild-\
\
steel plate to prevent free thermal expansion and\
\
contraction of the plates and connecting test weld,\
\
thussimulatingthesituationthatexistsduring\
\
assembly of the magnet case.\
\
Thoseweld-metalchemicalcompositions\
\
determined to date are reported in Table 23. Initial\
\
evaluationof welding performancewas by room\
\
temperature side-bend testing. Failures were seen in\
\
both specimens from weld No. 8 (GMA process,\
\
3I6L, filler) and one of two specimens from weld\
\
No. 9 (SA. 316L filler) made at LLNL. Failure in\
\
the latter case was associated with an entrapped-\
\
slag defect in the tension-side of the sample. Such a\
\
defect is associated with incomplete removal of the\
\
fused slag during welding, and is not an inherent\
\
problem in either the choice of weld filler material\
\
or process. However, the two failures in the GMA\
\
welds were traced to thin oxide films on the solidify\
\
ing weld metal, caused by the presence of 2% O2 in\
\
the shielding gas. Oxygenis added to the argon\
\
shielding gas to lower the surface tensionof the\
\
molten stainless steel weld metal and promote flow\
\
of the molten metal to the edge of the weld joint,\
\
thus ensuringcompletewetting alongthe entire\
\
weld-metal-to-base-meta!interface(Ref. 96).As\
\
such, the cause of failure in the GMA samples was\
\
associated with the weld process and was considered\
\
to be sufficientreasonto eliminate conventional\
\
GMAwelding,whichusesAr-02 gasmixtures,\
\
fromfurtherconsideration.\
\
Intergranular cracks of lengths insufficientto\
\
constitute failure accordingto SectionIIIof the\
\
ASM E Boiler and Pressure Vessel Code were seen\
\
in side-bend specimens of Weld No. 17 made by\
\
CBI (Houston),using the pulsedGMAprocess,\
\
2RM69 filler and 75% He-25% Ar shielding gas.\
\
Details of evaluation of this weld conducted by both\
\
LLNLandCBI(Houston)arecontainedin\
\
Refs. 97-99.Briefly,manyintergranular fissures,\
\
with lengths up to 0.008 in., were found in the as-\
\
deposited weld metal. After side-bend testing, these\
\
defects grew to lengths of up to 0.156-in. While the\
\
exact cause of these defects was not determined, the\
\
factthat they occurred on a weld metal specially\
\
developedforfreedomfromthistypeofdefect\
\
(Ref. 100) and in weld deposits made with a process\
\
(GMA) known to be sensitive to both operator and\
\
process variables (Ref. 96) causes one to view with\
\
concern any use of conventional GMA welds and\
\
2RM69 filler metal for the stringent requirements of\
\
4 K high-stress service.\
\
Low-temperature mechanical-property evalua\
\
tions of all welds included 77 K Charpy V-nolch im\
\
pacttests, followedby 4 K tensile andfracture-\
\
toughnesstestingofselectedwelds.Resultsare\
\
presented in Tables 24 and 25 and discussed below.\
\
77 K Charpy V-Notch Impact\
\
TestResults (Figure 94).\
\
As a function of weld process, from largest to\
\
smallestvalue of average energy absorption,the\
\
resultsfortheLivermore-producedweldswere:\
\
ESW, SMA, SA, GMA, and FCMA. Usingaverage\
\
lateral expansion at the root of the notch as a rating\
\
tool, the results for the Livermore-produced welds\
\
were: ESW, SMA, GMA, SA, and FCMA. Results\
\
are also shown in Fig. 94 for two welds made by\
\
FMC (San Jose). Their ESW weld exhibited much\
\
lower impact performance than the two ESW welds\
\
made by Livermore, but their pulsed-GMA weld\
\
results fell between the SMA and Livermore GMA\
\
results.BasedontheresultsoftheLivermore-\
\
produced welds, the SMA, SA and ESW welds were\
\
selected for evaluation at 4 K.\
\
103\
\
80\
\
8\
\
60\
\
o40\
\
20-\
\
LLL-Weld made byLLL/MFD\
\
FMC-Weld made by FMC-San Jose\
\
—i\
\
#12 LLL\
\
-\
\
#7 LLL\
\
J#13 LLL\
\
DD®#20 FMC\
\
#8 LLL\
\
1x\
\
1\
\
#21 FMC\
\
#10 LLL\
\
X\
\
-\
\
Ld\
\
100\
\
50\
\
#8 LLL\
\
IXI\
\
#9 LLL\
\
Ix\|\
\
#20 FMC\
\
LiJ\
\
-\
\
#8 LLL\
\
IXI\
\
#9 LLL\
\
Ix\|\
\
x\
\
X#13LLL\
\
#12LLL\
\
#20 FMC\
\
LiJ\
\
#7 LLL\
\
#8 LLL\
\
IXI\
\
#9 LLL\
\
Ix\|\
\
21 FM C#20 FMC\
\
LiJ\
\
-X\
\
#8 LLL\
\
IXI\
\
#9 LLL\
\
Ix\|\
\
21 FM C#20 FMC\
\
LiJ\
\
-X\
\
#8 LLL\
\
IXI\
\
#9 LLL\
\
Ix\|(?)\
\
#20 FMC\
\
LiJ#8 LLL\
\
IXI\
\
#9 LLL\
\
Ix\|#10\
\
Ix\|\
\
#10\
\
Ix\|\
\
SMAGMASAESW\
\
Weld process\
\
FCMAPulsed\
\
GMA\
\
FIG.94.Resultsof 77 KCharpyV-notch impact tests on type 316L weld metal deposited by various welding\
\
processes and laboratories.\
\
104\
\
4 KTensile and\
\
Fracture-ToughnessTest\
\
Results (Table25and Fig.95)\
\
Table 25 and Fig. 95 give our 4K tensile and\
\
fracture toughness test results, together with data\
\
from other sources (Refs. 101 and102). Examina\
\
tion of Fig. 95 shows that the SMA weld. No. 7, ex\
\
hibited a superior combination of yield strength and\
\
fracture-toughnessrelative to boththe SA weld,\
\
No. 9. and the ESW weld. No. 12. This provided ad\
\
ditional support for selection of the SMA process as\
\
the primary candidate for fabrication of the magnet\
\
case. Most of the data from Refs. 101 and 102 was\
\
generated after testing of welds 7, 9, and12 was\
\
completed. The consistently superiorperformance\
\
of SMA welds shown in Fig. 95 confirms the choice\
\
made in the basis of data from the three Livermore-\
\
produced welds.\
\
Evaluation of Weld-Metals\
\
Deposited by the SMA\
\
Process Using Type\
\
E316L-15 Electrodes\
\
Welds have been made by both Livermore and\
\
theCB1corporateweldinglaboratory,using\
\
E316L-1S electrodes supplied by both the Teledyne-\
\
McKayCorporationunderthetradename\
\
"Kryokay" and by the ARCOS Corporation. Both\
\
companiessupplyproductto specificationAWS\
\
5.4-75, the industry standard, with an added restric\
\
tion of a ferrite number (FN) of zero, i.e., an all-\
\
austenitic weld metal. All butt welds were made in\
\
3-in.-thick type 304LN plates which were heavily\
\
restrained by first welding the plates to a 6-in.-thick\
\
slrongback(LLNL)orbywelding intoarigid\
\
restraintfixture(CBI-Houston). These welds are\
\
described in Table 29.\
\
Thoseweld-metalchemicalcompositions\
\
determined to date are reported in Table 30. Initial\
\
evaluation of weldment performance was by room\
\
temperature side-bend testing. No failures were ob\
\
served.\
\
Low-temperature mechanical-property evalua\
\
tions included 77 K Charpy V-notch impact tests of\
\
mostwelds,followedby4 Ktensile,fracture-\
\
toughness,andfatigue-crackgrowthtestingof\
\
selected welds. Results are presented in Tables 31\
\
and32\. The77 K CharpyV-notchimpacttest\
\
results are presented in Table 31 and summarized in\
\
Fig. 96. On the basis of these results, there was little\
\
difference between ARCOS and Teledyne-McKay\
\
filler metals, or between welds made in the flat or\
\
verticalpositions. The 4 K tensile andfracture-\
\
toughness test results are presented in Table 30 and\
\
summarized in Fig. 97. Note that the performance\
\
TABLE 29.Shielded-metal-arc welds in type 304LN, using typeE316L-I5electrodes.\
\
«>ldDiam.\
\
(in.)\
\
ElectrodeWeld jointFerrite No.\
\
Vendor\
\
. »y\
\
No.\
\
Diam.\
\
(in.)SupplierHeal No.Lot No.TypeOriginPosition\
\
Ferrite No.\
\
VendorLLNL\
\
75/32McKay021462161376ButtLLNLFlat1.0\
\
143/16McKay213333186938BaitLLNLFlat1.0\
\
163/16ArcosT-12150-2FlatLLNLFlat00\
\
185/32£\
\
1 / \*\
\
McKay3860223086644ButtCBI-HslnVert.0\
\
191/4McKayN/A0631-401ButtLLNLFlat\
\
225/32«\
\
3/16\
\
McKayN/AN/ABut)CBI-HsMFlat\
\
231/4ArcosT-12150-2ButtCBI-HstnFlat\
\
241/8McKay3860223086683ButtCBI-HstnVert.0.5\
\
285/32McKay021462161376Dble-VLLNLVert.0\
\
295/32McKay163293178554Dble-VLLNLVert.0\
\
345/32McKay5861082397723ButtLLNLFlat0\
\
355/32McKay5861082397723DUe-VLLNLVert.0\
\
385/32McKay5861082397723ButtCBI-HstnVert.0\
\
N/A—Data not a«aHaMe.\
\
105\
\
0Y\
\
75100125\
\
Symbol\
\
Yield strength (ksi)\
\
Welding process and detailsBase metalReference\
\
ASMA E316L-16304 L101\
\
VSMA E316L-15316 LN101\
\
•SMA E316L-15, Interpass peening316 LN102\
\
VSMA E316L-15, Interpass peening\
\
plus 1900 F/1 h r / H20 Quench316 LN102\
\
K\
\
HSMA "KRYO-KAY", Horizontal weld304 LNThis Report\
\
ASMAARCOSE316L-15304 LN\
\
KvSMA "KRYO-KAY", Vertical weld304 LN\
\
OSA316L304 LN102\
\
€SA316L304 LN102\
\
»SA316L304 LNThis Report\
\
EESW316L304 LN\
\
DSMA E316L-15 Flat weld304 LN\
\
FIG. 95.4 K plane-strainfracture toughness as a function of yield stress for type 316L weld metal deposited by\
\
various practices.\
\
106\
\
TABLE 30.Chemical compositions of shielded metal-arc welds made in type 304LN using type E316L-15 elec\
\
trodes (wt%).'\
\
WeH\
\
No.CMilSiSPCrNiMoTlCb+T«CuN2V\
\
140.0301.650.3518.013.22.25\
\
160.0332.440.360.0070.03218.2013.812.140.056\
\
0.0332.540.330.0100.03518.0713.812.230.061\
\
180.0232.140.300.0140.02218.1213.222.290.050.020.180.45\
\
230.0332.440.360.0070.03218.2013.812.140.056\
\
240.0251.940.280.0120.02117.8013.002.220.060.020.220.042\
\
280.O301.640.3518.013.22.25\
\
290.0301.650.3518.013.22.25\
\
.140.0362.180.220.0120.01817.8413.S82,100.020.280.084\
\
350.0362.180.220.0120.01813.5813.582.100.0520.020.280.084\
\
380.0362.180.220.0120.01813.5813.582.100.0520.020.280.084\
\
\*As determined on undiluted weld pads deposited in flit position.\
\
of Ihe individual weld metals is independent of both\
\
supplier and weld position and Falls at the upper end\
\
of the scatter-band for all data in type 316L weld\
\
metals deposited by the SMA process. Examination\
\
of the 4 K tensile ductilitydatain Table 32 in\
\
dicated a large variation in both elongation (5.4% to\
\
44.0%) and reduction area (6.8% to 33.5%). Since\
\
low values of both of these quantities are indicative\
\
of some factor in the weld metal being out of con\
\
trol, metallographic examinations of the welds hav\
\
ingbothhighandlow values oftheseductility\
\
parameters were conducted.\
\
Metallographic Examinations\
\
of Welds\
\
Examinations were performed using standard\
\
optical-microscopy techniques, and indicated that\
\
low values of elongation and reduction in area ap\
\
pear to be associated with the presence of extensive\
\
weld-metal defects in or near the plane of fracture.\
\
Such defects include microfissures (Refs. 100, 104-\
\
107) or small (0.OO5-O.O30-in. long) inter-granular\
\
cracks that form during solidification of low-ferrite\
\
orferrite-freeweldmetal(Fig.98)andsuch\
\
operator-related defects as lack of fusion (Fig. 99)\
\
and slag entrapment (Fig. 100), which are caused by\
\
less thansatisfactorywelding practices. A semi\
\
quantitativeratingoftheoccurrenceofmicro-\
\
fissures was made, presented in Table 32, indicates\
\
that the presence of appreciable amounts of this\
\
type of defect is limited to the vertical welds Nos.\
\
18,28,and29madewithTeledyne-McKay\
\
"Kryokay" electrodes. As this fact was discovered\
\
toolateduringthefabrication-cycleof the two\
\
magnet cases to change to another weld metal, cer\
\
tain remedial steps to limit the occurrence of micro-\
\
fissuringandassesstheeffectofheavily\
\
microfissured welds on the expected performance of\
\
simulated case close-out welds were undertaken and\
\
are discussed later.\
\
Fatigue-Crack Growth\
\
Behavior of Type-316L\
\
Weld Metals\
\
Fatigue-crack growth behavior was evaluated\
\
by the National Bureau of Standards and the results\
\
are presented in Fig. 101. Note that the behavior of\
\
type-316L weld metals are consistent in that the\
\
fatigue crack growth (FCG) rates fall with a factor\
\
of about 5 of each other and are about a factor of 3\
\
to-10 less than that of the type-304LN base metal\
\
used for fabricationof the magnet cases. An in\
\
formalreport(Ref. 103)preparedbyGeneral\
\
Dynamics/Convair,usingtheactual4 K\
\
mechanical properties of the base metal and welds\
\
presented in Fig. 101, demonstrated the structural\
\
adequacy of this case at a design stress of 80 ksi at\
\
4 K .\
\
107\
\
TABLE 31.Chirpy V-notch impact performance at77K ofshielued-metalarcweldsintype3041.N plateusing\
\
type E3I6L-15 electrodes.\
\
Energy absorbed\
\
W t M N nAverage, ft-lb\*Range, ft-lb\
\
741.732.1) In 6U.II\
\
1433.829.5 Co 38.11\
\
1639.438.0 to 42.0\
\
1843.734.0 to 53.il\
\
2237.627.0 to48.11\
\
2439.835.0 to45.(?\
\
3440.S39.8 to 41.2\
\
3540.539.8 to 41.2\
\
3846.639.0 to 56.0\
\
Lateral expansion\
\
Average,rails3\
\
'"Range, mils\
\
37.830.8 to 37.J.\
\
30.1118.5 to 54.0\
\
26.1124.0 lo 28.5\
\
30.524.0 to 40.0\
\
29.621.0 In35.0\
\
311.226.11 In 41.0\
\
29.027.(1In31.0\
\
29.027.0 to 31.11\
\
30.1128.11to411.0\
\
'Average of 3 to 5 specimens.\
\
h\
\
lmH =0.001 in.\
\
TABLE32.4 K tensile propertiesof shielded-metal arc welds made in type 3041.N plate using type E316L\*15\
\
electrodes.\
\
Wdd\
\
No.\
\
Spec.\
\
4Ktensile properties\
\
Ultimate\
\
strength,\
\
ksi\
\
YieldKlong.\
\
strength,I in.,\
\
ksi%\
\
Reduction\
\
in area.\
\
k.c"\
\
Occurrence/\
\
extent orMicro-\
\
microfissuresstructure\
\
18\
\
22\
\
23\
\
28\
\
29\
\
35\
\
LT\
\
L\
\
ST\
\
L\
\
ST\
\
L\
\
L\
\
L\
\
LT\
\
ST\
\
L\
\
L\
\
L\
\
189.4152.326.025.7\
\
176.4119.825.025.0\
\
176.2117.044.027.4\
\
153.5111.118.027.5\
\
187.4125.744.033.5146.0\
\
159.4123.85.46.8170.0\
\
183.2112.036.428.1183.0\
\
163.6130.710.010.9\
\
180.7128.516.014.5\
\
171.4123.532.024.3\
\
168.7111.914.516.2\
\
184.2145.39.510.8\
\
188.3111.319.017.6\
\
Some\
\
Some\
\
Auslt'nite\
\
Ferritc\
\
A listen ite\
\
.-"write\
\
SomeA listen.tc\
\
ManyAusten itc\
\
SomeA ustcn ile\
\
Austcnite\
\
Austcnile\
\
\*Sped«eii •TJentitMHi relative to centerline of weld.\
\
L—LMCMUMMI, paraMel lo weld axis.\
\
LT—Long transverse, across weld axis.\
\
ST—Short transverse, through weldthickness.\
\
Determinedhyelastic-plastic J-mtegraltest.\
\
108\
\
-X60\
\
-.TI-,-\|#24\
\
40\
\
20\
\
4»3 4\
\
">H#7\
\
-J #14\
\
J#38\
\
¥22\
\
#18\
\
\_ l \_ \_ JIIIIIIIIIIL\
\
ARCOS-FlatMcKay-FlatMcKay-Vertical\
\
LLLCBILLLCBILLLCBI\
\
Source of filler metal and position of weld\
\
FIG. 96.Results of 77 K Charpy V-notch impact tests on type 3I6L-IS weld metals deposited on type 304LN\
\
base plate.\
\
109\
\
200\
\
-175\
\
c\
\
>\
\
\|150\
\
jI125\
\
100\
\
0\
\
Scatter-band for type\
\
316L weld-metal\
\
deposited by/SMA\
\
process\
\
\\
\
\\
\
\\
\
\\
\
\\
\
\\
\
\\
\
\\
\
\\
\
\\
\
\\
\
\\
\
75100\
\
Yield strength (ksi)\
\
125\
\
LLL\
\
SymbolFiller-metal vendorBase metalCommentswelding\
\
KH\
\
Teledyne-McKay304 LNFlat weld22\
\
AARCOS304 LNFlat weld16\
\
Kv\
\
Teledyne-McKay304 LNVertical weld18\
\
DTeledyne-McKay304 LNFlat weld7\
\
FIG. 97.4 K plane-strain fracture toughness as a function of yield strength for type 316L weld metal used in\
\
fabrication of MFTF.\
\
110\
\
*\
\

"Just-completed and never-used"\

\
The Reagan administration’s decision to mothball the machine came as a gut punch to the researchers. Lawmakers tried to fight for extra funds to throw a lifeline to the program or to salvage parts of the project to perform some science in a more limited scope. The money and time spent on the project was weighing heavy in the comments recorded in hearings held by the House Subcommittee on Energy Research and Production in February of 1986.\
\
Congressman Fortney ("Pete") Stark of California [made a particularly impassioned plea to extend the life of the project. Citing the Reagan Administration's proposed budget, Stark wrote "This proposal would mothball the just-completed and never-used Mirror Fusion Test Facility-B (MFTF-B) at the Lawrence Livermore National Laboratory." Stark continued, "A lot of hard work and money has been invested in the world's largest superconducting, tandem mirror fusion experiment. For close to 8 years some of the finest scientists and engineers in America have dedicated their time and energy to the project. $350 million dollars have been invested in MFTF-B."\
\
The MFTF-B under construction in 1983. Source: Lawrence Livermore National Laboratories.\
\
Stark included some of the photos of the facility for the record, showing its impressive scale. "As you see the facility is truly an incredible accomplishment. Even to those who have not had years of scientific training, the enormous complexity of the project can be appreciated. Please let me repeat: 8 years of dedicated manpower and $350 million dollars have been pumped into the MFTF-B."\
\
In the years following the shutdown of the program, parts of the machine were scavenged for other projects, and the rest was scrapped in 1998. Building 431 at Lawrence Livermore National Laboratory sat empty for a number years and was eventually demolished around 2005 after determining that the site did not meet the threshold of historical significance to be protected on the National Register of Historic Places.\
\
The installation of a magnet into the MFTF-B in 1981. Source: LLNL\
\

Recent breakthroughs reignite hopes for fusion\

\
With the recent news of Lawrence Livermore National Laboratory's National Ignition Facility achieving the first net energy gain nuclear fusion reaction, the outlook for prioritizing fusion research looks bright.\
\
Having achieved "ignition" for the first time on December 5, 2022 by creating a reaction where more energy is released than consumed, it turns out a third approach was the key to success: containing the plasma within high powered lasers. In fact, the successful ignition employed another huge, expensive machine – this one equipped with 192 massive lasers all focused on a tiny pellet, pounding it with 2 million joules of energy, creating a fusion reaction that only lasted for 100 trillionths of a second.\
\
"Crossing this threshold is the vision that has driven 60 years of dedicated pursuit — a continual process of learning, building, expanding knowledge and capability, and then finding ways to overcome the new challenges that emerged. These are the problems that the U.S. national laboratories were created to solve.”\
\
Another huge, expensive fusion machine. But this one worked. The target chamber of the National Ignition Facility at Lawrence Livermore National Labs (LLNL). Source: LLNL\
\
_\\* In 1986, the total cost of the project was described in Congressional testimony as costing $350 million, which would equal $965.4 million according to the_ _BLS_ _._\
\
\
\
#### Mirror Fusion Schematic T-Shirts\
\
Featuring a schematic of the 400 ton "yin-yang" magnet used in the The Mirror Fusion Test Facility at Lawrence Livermore National Laboratories. Source: Lawrence Livermore National Laboratories archives.\
\
Browse the Beautiful Public Data shop on Etsy\
\
🙏🏻\
\
Did you enjoy this post?\
\
You can subscribe to our newsletter to get future posts delivered to your inbox for free. 👉🏻 📫 Subscribe now.\
\
☕️ Buy me a coffee\
\
Say thanks with a one time tip.\
\
Sharing is caring\
\
📣 If you think your followers or friends may like it, please consider sharing it.\
\
👕🖼️ Buy Beautiful Public Data merch on our Etsy shop\
\
Unique t-shirts and prints inspired by our stories\
\
🙋🏻‍♀️ If you have any suggestions, comments or requests, please email them to [email protected]\
\
Thanks for reading!\
\
\- Jon Keegan\
\
Bluesky: @jonkeegan.com\
\
Mastodon: mastodon.social/@jonkeegan\
\
Subscribe\
\
StripeM-Inner