1. Graham Beck, LBL September 2012 1 Stave Core (“Plank”) Thermal QA - UK Motivation for Thermal QA: Building 24 modules to populate a plank costs a lot of effort and money (1 stave ≡ 40 transatlantic return flights*) => We should test the plank thoroughly (within reason). If it looks good and behaves well mechanically, it will probably be ok thermally. (?) But of course you can’t see inside it. Maybe you want to do some of: - Thermal QA of every plank (had better be fast: ~ 2h per plank?)….. the first few …. a selected sample - Check out thermally: irradiated / thermally cycled / otherwise stressed …planks Infrared Thermography (of the surface) is worth ~ 10 5 individual contact measurements. Is it feasible? (have had useful discussion with Dave L). * “only 3 if you travel First Class” (Adrian Bevan)

2. Graham Beck, LBL September 2012 2 IR Thermography At low T: -30C black body radiation intensity is still appreciable: ~ 40% room temperature level. The peak is well within the LWIR window, commonly used by the cheaper IR cameras * (e.g. QMUL, Liverpool)! Energy Balance + Kirchoff says that: emissivity + reflectivity + transmissivity = 1. =>Thermography works best on high emissivity surfaces: - Bigger signal - Lower reflected ambient radiation (inc. camera!) * Nowadays you can buy a radiometric IR camera for ~ 4 transatlantic flights (10% of a stave). The (default) plank surface is 25  m Kapton over (refective) Aluminium shield (if present). Kapton is somewhat transmissive. For bus tape  I measure ~ 90% (above room temperature); DaveL measures ~83% at low T (-25C?).  Expect plank thermography can work.  Sensitivity to faults will depend on the “noise” due to surface quality – uniformity, flatness etc.  Even if the bus gets buried (?) there should be an electrically insulating (so high  ) surface.

3. Graham Beck, LBL September 2012 3 How would you make a measurement? Flow coolant through the pipe, at ~ -30C. Actively heat the plank? It turns out that natural convection and radiation from ambient supplies about the right thermal load. Rough estimates: Convection: Assume htc = 5W/m 2 K (consistent w. R.Bennett calculations and Geneva experiments). (ignore dependence on plank orientation here). Radiation: W/A = .  * (T 4 – Ta 4 ) A=Area, Stefan’s constant  = 5.67E-8,  =emissivity: assume = 1 ≈ 4 .T mean 3 *  T ; (  T << T) i.e. (linear approx). Radiative htc ≈ 4 .T mean 3 e.g. +20C (ambient) => -25C (T mean ≈ 270K) => 4.5 W/m 2 K - very comparable to convection! Ambient load per module length of plank: ≈ 125 mm × 98mm × ~ 10 W/m 2 K × 45C => 5.5W ≈ Hybrid Power.

4. Graham Beck, LBL September 2012 4 FEA Thermal Model of a section of plank Quick & Dirty: Stave130 Module FEA with sensors+hybrids stripped away. Fast and informative … not particularly accurate ! Same htc top & bottom (clearly inaccurate) Cable Bus top and bottom 2mm id Ti pipe. Material Specific Heats assigned (used in transient case) - but Facing Glue Cv omitted. - Corresponds rather roughly to Tim’s planklet measurement : +20C, 10 W/m 2 K Surface Film Load

5. Graham Beck, LBL September 2012 5 PERFECT PLANK in STEADY STATE: comparison with Tim’s IR plot. After a couple of iterations to tie down CO2 temperature (approx. -38.5C) (FEA Ambient Thermal Load now ~ 6.5W). ABAQUS FEA Tim’s IR (Stavelet3+masking tape) & Fit -32 ° C -25°C - °C Scale: Abaqus  T looks ~ 14% low. => Some confidence in FEA feasibility studies …

6. Graham Beck, LBL September 2012 6 Now introduce a FAULT into the FEA model: GLUE missing between the foam and pipe Effect on STAVE (module) performance: Mean T C Headroom Sensor T (degrees) Baseline, No faults-25.23 C21.6 Missing from one pipe, top half,  Z 10mm -25.2121.5 " " "  Z 30mm -25.1421.3 " FULL circumference, 30mm -24.619.7  Choose as a benchmark test: glue missing over 30mm length, half the pipe circumference.  Has only a small effect on STAVE thermal performance (heat spreads around the fault)  Maybe you would want to know about it anyway. It might deteriorate with thermal cycling, etc.

7. Graham Beck, LBL September 2012 7 PLANK FEA: GLUE FAULT (STEADY STATE) Plotted bands: -32C=> -36C 0.2C intervals (~ camera resolution) Height of bump is 0.7C (top surface - also 0.4C bump on bottom surface) Would this be visible in practice? - depends on “noise”: bus surface flatness, uniformity, convective fluctuations etc.

8. Graham Beck, LBL September 2012 8 Tim, re Thermal Shock: “What is the typical dT/dt seen during Stavelet Testing with CO2?” Now look at Transient behaviour … Stavelet 3: IR temperature history of two points on the surface, above the pipes, after turning on CO2 (later turning off / back on again) At SP02 (above inlet pipe) T falls ~ 7C/s after turning on CO2… CO2 flows around U-bend, and reaches SP01 after ~43s => speed ~ 12mm/s. -meanwhile, outlet (SP01) is cooled, slowly, by conduction across the plank. (Interesting? ~simultaneous warming when flow stopped). Try to reproduce this in FEA - then assess potential for fault finding...

9. Graham Beck, LBL September 2012 9 Transient FEA… Sledgehammer approach: -Divide pipe surface (crudely!) into 8 x 12.25mm lengths along Z -Transient FEA 1s increments: -Change film conditions for successive sections ( ≡ 12.25mm/s) from +20C, 10W/m 2 K => -38.5C, 8000W/m 2 K. -Apply this to inlet pipe. -After 43s, apply (in reverse sense) to outlet pipe -Allow to settle (a further 43s). (See Perfect Plank Transient.wmv: !Range is +20C to -40C) Propagate change from Ambient to CO2 conditions in 1s steps => Plot history of surface T above centre of each pipe (as Tim)

10. Graham Beck, LBL September 2012 10 PERFECT PLANK: Surface Temperature transient. FEA doesn’t reproduce all features of IR, and peak dT/dt is 70% higher But worth pursuing … IR: Rate [C/s] vs time T vs time FEA:

11. Graham Beck, LBL September 2012 11 FEA TRANSIENT – with and without GLUE FAULT PERFECT Rate [C/s] vs time FAULTY (see Faulty Plank Transient.wmv Range: 0C to -40C) Peak Rate is unchanged on inlet side (~11C/s), but ~25% lower over centre of fault. T vs time

12. Graham Beck, LBL September 2012 12 A Trivial QA Algorithm: Define lines L1,L2 on the surface above the pipes. Record T on L1, L2 for duration of test (this would be ~ 5 mins for a full stave) For each Z, evaluate dT/dt(t). Plot Peak dT/dt vs Z. Look for anomalies (bumps) The plots are derived in this way from the FEA Temperatures at 2x20 nodes along the bus tape. [Fluctuations due to jerky FEA! Ignore Z extremes – wrong BCs!] => Inlet behaviour is fairly level... => Outlet shows expected bump due to fault. So there is an appreciable signal !!...BUT...... there will be background conduction from ± Z...and what is the noise like in practice??? Many Questions: Best answered by experiment!

13. Graham Beck, LBL September 2012 13 Vaporiser (Heater) Alternative (avoiding thermal feedback) - FLOW routed through independent chiller bath - RETURN from stave routed directly to Vaporiser Dial + DAQ. (Re-locate upstream of vaporiser) T1T2 T3 T4 T5 Tn located as per Nikhef Manual. T4 redundant if HEX omitted. T5 could be un-monitored/low precision. T1,T2,T3: 4-wire PT100? (expensive but accurate). OR Pico 8-chan. Thermocouples (incs. Labview) (Poss.re-locate second valve). CYLINDER STAVE QMUL will build a CO2 blow-off cooling system to evaluate these QA ideas (so far have a list of parts to purchase – but need a decision on temperature monitoring)...and I believe Oxford are building an intentionally “faulty” stave.

14. Graham Beck, LBL September 2012 14 BACKUP: Measurement of IR Emissivity of Bus Tape Bus sample + thermal grease on (warmed) block of copper. Black tape added (v. small correction to T, estimated by adding more layers). Spot radiometer measures T of tape and bus: Bus  estimated assuming 94% for black tape.  seems high – as hoped. Note: Kapton thickness is 25  m, cf radiometer sensitive in 8-13  m window. At low temperature (-25C?), Dave Lynn finds ~ 83%. coverlayer bus tape