Integrated Stave Mechanics/Cooling
-
Upload
ivor-cochran -
Category
Documents
-
view
47 -
download
3
description
Transcript of Integrated Stave Mechanics/Cooling
M. Gilchriese
Integrated StaveMechanics/Cooling
June 5, 2008CERN
M. Gilchriese
Outline• Additional information at
– http://www-physics.lbl.gov/~gilg/ATLASUpgradeRandD/StaveReview/stave_draft_note.pdf
• Concept• Prototype construction/test• Thermal performance• Structural studies• Material• Questions• What if…..• Development plan• Production cost/schedule
2
M. Gilchriese
Concept• Approximate dimensions of mechanical/cooling core
– Short-strips: length about 1 m + end-of-stave card (2 m possible)
– Long-strips: about 2 m long + end-of-stave card
– About 11 cm wide
– Thickness 3-3.5 mm (CO2) or 5.5-6.5 mm(C3F8)
3
Bus cable
Hybrids Coolant tube structure
Carbon honeycomb or foam
Carbon fiberfacing
Readout IC’s
Silicon sensors
M. Gilchriese
Prototypes• Prototype stave structures were fabricated and tested
(thermal/mechanical) starting Fall ‘06 up to about one year ago.
• The design of the prototypes was fixed before choice of 10 x 10 cm2 detectors and the prototypes are therefore 7 cm wide.
• Goals: gain experience with fabrication, thermal performance, simple mechanical properties and build 1 m object for modules
4
PrototypeNumber
Facing Length (m)
Facing Material # of Plies per Facing
Tube Type Purpose
1 0.343 CN60 10 Flattened Assembly trial
2 0.343 K13D2U 10 Flattened Thermal prototype
3 1.07 K13D2U 10 Flattened For modules
4 0.343 K13D2U 3 4.8 mm round/ POCO foam
Thermalprototype
5 0.343 K13D2U 3 2.8 mm round/ POCO foam
Thermal prototype
M. Gilchriese
Prototype Construction
5
Honeycomb core
Prototype #4
Prototype #3
Facing
Carbon foam
Carbon foam
Prototype #5
Honeycomb 5 mm thick for all prototypes
M. Gilchriese
Prototype Testing• Thermal performance
– Simulated heat loads(e.g. 3.3 W/”hybrid”)
– IR imaging. Water coolant. Compare to FEA
– Before & after T cycling -35 to 20C
• As built-accuracy (CMM scans)
• Deflection measurements– Compare to expected properties
• “Module” removal trials– Attach dummy silicon with adhesive, cure,
remove, replace
• Detailed weights -> material estimates
6
Bus cableAlumina
Heaters 0.3mm silicon
Thermal measurements of prototypes
Dummy detector removal
M. Gilchriese
Prototype Lessons• Fabrication straightforward
– Obviously some learning but no surprises
• Thermal performance (T/Watt) similar for all three tube types, 4.8 mm tube+foam being best, flattened tube or small tube about the same
• Thermal performance in good agreement with FEA within errors of measurement based on expected materials properties (and their errors).
• No change in thermal performance after 50 cycles from -35C to 20C
• Deflection measurements in reasonable agreement with expectations (within 20%) but small sample (two prototypes)
• As-built accuracy (planarity of facing plane) somewhat worse than we hoped (1 m prototype). – Deviation from average (rms) 30-60 . All points within ± 100 window
– Why? Non-uniformities in honeycomb as provided by vendor. Can be reduced
• Dummy module removal, clean-up and replacement easy with SE4445 (adhesive used to attach current pixel modules)
7
M. Gilchriese
Models of Thermal Performance
8
Item Thickness Thermal Conductivity
(W/mK)
X/Y/Z
Solid Elements
Tube OD: 2.8mm, ID: 2.1mm 200
Composite Facing Thickness 0.21mm 148/1.3/294
Cable 0.125 0.12
Detector 0.28mm 148
BeO 0.38mm 210
Dielectric Hybrid 0.23mm 5
Chips 0.38mm 148
POCO Foam (0.9mm min) varies 50/125/50
Adhesives
Foam to Tube (CGL) 0.1 mm 1
Foam to Composite Facing (CGL) 0.1 mm 1
Facing to Cable 0.05 mm 0.8
Cable to Detector 0.05 mm 0.8
Detector to BeO 0.05 mm 0.8
BeO to Dielectric hybrid 0.05 mm 1.55
Dielectric to chip 0.05 mm 1.55
Shown for 10 x 10 cm2 detectors¼-model, primarily for thermal runawayAgrees with multi-hybrid model T
Multi-hybrid model. More elements. Vary composition of stave. Assess T change
M. Gilchriese
Nominal Structure Thermal Performance
• Honeycomb core
• ¼ model run as function of tube wall temperature
• Take into account detector heating
• Can already tell from this that C3F8 with Tmin =-25C is problematic
9
M. Gilchriese
Modified Structure Performance
• Relevant for C3F8 with Tmin =-25C
• Add more cooling – triple U-tube
• Or replace honeycomb with thermally conducting foam
10
-35
-30
-25
-20
-15
-10
-5
0
5
10
15
20
-35 -30 -25 -20 -15 -10 -5 0 5
Pe
ak
De
tec
tor
Te
mp
era
ture
oC
Tube Inner Wall Temperature oC
Triple U, 0.3W/chip, 1mW/mm2 @0C
Triple U, 0.5W/chip, 2mW/mm2 @ C
Single U-tube, foam, 0.3W/chip, 1mW/mm2 @0C
Single U-tube, 0.3W/chip, 1 mW/mm2 @0C
M. Gilchriese
More Improvements to Structure?• Vary facing thermal properties. Practically gain 1C in T
• Improve K of bus-cable? Assumed K=0.12. If K=0.38 (estimated from average metal content), gain 1.5C in T.
11
Facing Properties
Fiber K is given (Z/X/Y)
Thickness
(mm)
Lay-up Chip Peak T
(ºC)
Detector Peak T
(ºC)
Cooling Tube
K13D2U(294/148/1.3) 0.21 0/90/0 8.61 7.96 Single U-Tube
K13D2U(221/221/1.3) 0.42 0/60/-60/s 7.31 6.61 Single U-Tube
K13D2U(294/148/1.3) 0.42 90/0/0/s 7.93 7.24 Single U-Tube
K13D2U(352/89/1.3) 0.70 90/0/0/0/0/s 8.30 7.62 Single U-Tube
Carbon-Carbon(314/183/25) 0.42 90/0/0/s 6.88 6.19 Single U-Tube
K1100(367/185/2) 0.21 0/90/0 8.05 7.39 Single U-Tube
K13D2U(386/97/1.44) 0.21 0/90/0 4.34 3.69 Triple U-Tube
K13D2U(352/89/1.3) 0.70 90/0/0/0/0/s 5.25 4.55 Triple U-Tube
Effects on thermal performance from variations in the facing properties assuming a 0oC temperature for the coolant tube inner wall, 0.3 W/chip and no detector heating.
M. Gilchriese
Bridged-Hybrid Models• Some studies but not full thermal
runaway estimates
• See backup note for materials
• Concept uses foam in addition to facings to carry heat from foot of bridge back to cooling tube
• 0.25 W/chip, -28C wall temperature, no detector heating for these results
12
Description Chip Peak Temperature
(ºC)
Bridge Gradient
(ºC)
Sensor Max/Min
( ºC)
a. No air, no wire bonds 4.53 10.2 -17.7/-26.5
b. No air, with wire bonds -2.12 7.81 -19.2/-25.7
c. No wire bonds, with air -1.52 8.2 -15.6/-25.4
d. With air, with wire bonds -5.84 6.58 -20.0/-24.4
M. Gilchriese
Bridged-Hybrid Thermal Results• Effect of air flow studied (not
significant at T and flow studied)
• Nominal stave design (not bridge) at 0.25 W/chip, -28C wall and no detector heating has Tmax -22C
• Bridge -20 to -18C depending on foam K
• Optimization of tube position (closer to bridge foot) not studied, expect would reduce Tmax
13
Model IC Peak
Temp(C)
Bridge
Gradient(C)
Sensor
Tmax(C)
¼ bridge model(baseline foam conductivity) -5.8 6.6 -20
¼ bridge model (bridge foam reduced to 10W/mK) +2.4 5.0 -17.8
Multi-Hybrid bridge model (baseline foam conductivity)
With 0.01m/s air flow @ -15ºC
-6.2 n/a -19.9
Multi-Hybrid bridge model (sandwich foam 1W/mK)
With 0.01m/s air flow @ -15ºC, 1W per Detector
-5.74 n/a -19.4
M. Gilchriese
Two-phase Flow Calculations
• Two-phase flow estimates for CO2 (-35C) and C3F8 (-25C)
• Thermal runaway estimated at entrance (worst case)
14
Entrance( 0)
Exit ( 1)Tfluid -35oC
CO2 Heat Transfer 2.8mm OD tube
-35C Fluid Temperature
0
5000
10000
15000
0 0.2 0.4 0.6 0.8 1
Vapor Quality
He
at
Tra
ns
fer
Co
eff
icie
nt-
W/m
2 K
Combined
Boiling
CO2 Coolant Tube 2.8mm OD -35oC
0
1
2
3
0 0.2 0.4 0.6 0.8 1
Vapor QualityT
em
pe
ratu
re -
oC
Film TemperatureDrop
240 W heat load2 m tube, 2.2 mm ID
Vapor quality ()
Complex calculations!
TP 1oCTwall -35+1.75 -33C
Twall -35+1+2.5 -31C CO2
M. Gilchriese
Thermal Runaway – CO2
• Bulk fluid temperature -34C (entrance)
• Fixed heat transfer (film) coefficient 6833 (calculated at entrance) for 240 W
• Note film coefficient is heat dependent(goes up with more heat), not taken into account by us here
• Headroom OK15
M. Gilchriese
Thermal Runaway – C3F8(Tmin -25C)• Heat transfer coefficient
either calculated at entrance for 240 W(different for single and triple U-tube) or taken as 3000.
• Note that we would calculate value to be 3000 for 500 W (about at thermal runaway)
• Triple U – OK
• Foam(K=15 W/mK) instead of honeycomb OK
• If C3F8(Tmin -25C)+foam, need measurement!
16 -20
-15
-10
-5
0
5
10
15
0 1 2 3 4 5
Pea
k D
etec
tor
Tem
per
atu
re C
mW/mm2 @ 0C
Single U-tube, honeycomb core, 0.3W/chip, h=1436
Single U-tube, honeycomb core, 0.3W/chip, h=3000
Single U-tube, foam core, 0.3W/chip, h=1436
Single U-tube, foam core, 0.3W/chip,h=3000
Triple U-tube, honeycomb core, 0.3W/chip, h=1206
Triple U-tube, honeycomb core, 0.5W/chip, h=1206
M. Gilchriese
Thermal Performance Conclusions• The baseline design with a honeycomb core and a single U-tube does not have
acceptable headroom for Tmin = -25oC, representative of current cooling performance with C3F8
• The baseline design with a triple U-tube and a honeycomb core has acceptable headroom for Tmin = -25oC, representative of current cooling performance with C3F8
• A modified design with thermally conducting carbon foam instead of honeycomb and a single U-tube may have acceptable headroom for C3F8 with Tmin = -25oC (and colder fluids)
• The baseline design has acceptable headroom for a single U-tube and honeycomb core for Tmin -35oC, which could be applicable to CO2 or perhaps mixtures of C3F8 with other fluorocarbons.
• The headroom could be increased by small amounts from optimization of the carbon-fiber facings (gain 1oC) and from improved thermal conductivity of the bus-cable (gain 1- 3oC). These possible gains would be most important to realize if C3F8 with Tmin about -25OC were used.
• The headroom for a bridged-hybrid design with Tmin -35oC is likely to be sufficient (but more precise calculations remain to be done)
17
M. Gilchriese
Structural Studies• Preferred support concept is stave-on-shell
• Stave sag, vibrational modes, etc coupled with number of supports along length, shell design (minimize overall X0) – not studied in detail.
• Simple calculation of sag (< 75 in horizontal position, worst) with support every 50 cm
• Stave distortions upon cool-down from 25C to operating temperature– Quick look taking artificially bad case of alternating
modules top and bottom. Result is 11 microns out of plane for 50C temperature change
– Should be less with balanced structure
• Shear stress between Al tube and foam estimated and looks OK – see ATLAS note
• Clearly much more structural analysis needed18
M. Gilchriese
Material• Material estimates for simple stave only. Does not include
coolant, bus-cable, modules, end-of-stave cards, support points, strain relief…
• Based partly on prototype weights (scaled) and from calculation
• Uncertainty in facing thickness/density, adhesive choices, tube diameters => plausible range below for different configurations
• Top three for nominal design (modules glued to bus-cable). Bottom estimate for bridged-hybrid
19
Description X0 (%) Single Al U-tube (3mm OD, 2.2mm ID), honeycomb core 0.30-0.35
Single Al U-tube (6mm OD, 5.3mm ID), foam core ( = 0.2 g/cc) 0.45-0.50
Triple Al U-tube (6mm OD, 5.3mm ID), honeycomb core 0.65-0.70
Single Al U-tube (3mm OD, 2.2 mm ID), foam core ( = 0.2 g/cc) 0.35-0.40
C3F8 CO2
M. Gilchriese
Questions• Is it credible to assume the use of conducting carbon foam
around the tube in the baseline design (with honeycomb core)?– Yes. Foam of density 0.5 g/cc (as used in prototypes) is available from
at least two vendors. Production (batch size) is 150,000-200,000 cc, far more than we would need
• One of the design alternatives uses low density carbon foam ( ≤ 0.2 g/cc). Is this credible?– We think so. We are actively working with three vendors (for pixel staves)
on conducting foam with the appropriate properties and have samples in hand from all three. The production rate is claimed to not be driven by .
• Are there any other “non-standard” materials proposed for use?– No.
• Could you make a 4 m stave for the long-strip layers?– Not in my opinion
20
M. Gilchriese
What If….• What if the short-strips staves were 2 m long instead of 1 m?
– Fabrication of 2 m stave cores would not be significantly more challenging than 1 m stave cores. Could be cheaper (less labor) since fewer parts.
– CO2 cooling at about -35C would work with a 4 m single U-tube but probably would increase tube ID by small amount (tenths of mm)
– Structurally would be same as 1 m since supported along length (e.g. every 50 cm) except possibly for fixation scheme that accounts for CTE difference between stave and shell support but even this goes away if 2 m is fixed at center and 1 m fixed at an end.
– Good experience handling 1 m prototype, including wire bonding. 2 m harder, but not by much
– Survey of modules on 2 m stave harder, may require cross reference at 1 m scale, depends on survey capability. Not showstopper.
• What if stainless steel pipes were used?– Impact on thermal performance small (< 1C)
– Bending (for larger diameter for C3F8) – not sure
– Radiation length increase CO2 (C3F8) 0.3(0.5)% x ratio of wall thickness to Al
21
M. Gilchriese
Development Plan• These four principal activities would occur largely in parallel
• Thermal ( 1 yr once coolant testing available)– Selection of coolant essential to make progress (or need to carry multiple design options)– Small-scale prototypes likely to be needed– Design, fabricate and test full-length prototype(s)
• Structural ( 1.5 yrs)– Also coolant dependent. Once coolant selected…..– Combined design of stave and supporting structure (obviously also coupled with thermal
design) => baseline design that meets thermal and structural requirements. – Build prototypes and test (in addition to thermal prototypes)
• Module interface ( 2 yrs)– Define and prototype module mounting requirements: temporary holding for module
mounting, survey, testing (boxes, how to cool), shipping (boxes), etc…
• Production planning interface ( 2 yrs)– Tooling, procedures, who builds what, etc..
• Durations shown ignore resource constraints!
22
M. Gilchriese23
Production• A preliminary estimate of production cost and duration made earlier this
year: http://www-physics.lbl.gov/~gilg/ATLASUpgradeRandD/StaveReview/Cost%20Estimate%20for%20Integrated%20Stave%20Mechanics.doc
• Covers barrel and simple extrapolation to disks. All staves/petals.• Material and equipment costs in U.S. $. • Cost and manpower range estimated.• Includes contingency (but not escalation)
• Materials and equipment: $2-4M
• Engineering labor: 8-12 FTE years
• Technical labor: 28-48 FTE years
• Rough schedule 2 years design/prototype 1 year pre-production 2 years production
• Resource constraints not included!
Costs in U.S. ‘08 $