1. Page 1 Modifications to JLab 12 GeV Refrigerator and Wide Range Mix Mode Performance Testing Results P. Knudsen, V. Ganni, N. Hasan, K. Dixon, R. Norton, J. Creel Cryogenics Group

2. Page 2 Milestones JLab 12 GeV cold box milestones: –Feb 2006: Project is approved to proceed with engineering; cold box design process studies began –Oct 2008: Approval to begin construction phase; cold box RFP issued –Sep 2009: Based upon competitive bid process, award for given to Linde Process Plants, Inc., to deliver new 18 kW 4.5-K equivalent cold box –Mar 2012: Cold box received at JLab –May 2013: Commissioned cold box system –Jun – Aug 2013: Investigate high LN usage –Aug 2013: Re-test cold box system –Nov 2013: CHL2 supporting 2-K ops for LINAC commissioning –Jun – Aug 2015: Modifications to CBX made by Linde PP and aux. CBX installed –Sep 2015: Modified CBX tested –Oct 2015: Resume 2-K operation

3. Page 3 Configuration 12 GeV cold box configuration –Three pressure levels (HP, MP and LP) –LN pre-cooler (300 to 80 K) –Below 80 K, four expansion stages; total of 7 turbines w/ variable brake (3) two-turbine in series (no HX) + (1) JT-expander –Two physically separate cold box vessels w/ interconnecting transfer- lines –Upper cold box (vertical, outside): 300 to 60 K –Lower cold box (horizontal, inside): 60 to 4.5 K –All vertically oriented HX’s in both CBX’s (warm-end on top) –Dual 80 K beds, single 20 K bed – both with bypass valves –Shield load can be supported Using either or both T1-T2 / T3-T4 turbine strings, or, At 80 K –3000 liter helium sub-cooler –10000 liter external helium dewar

4. Page 4 LN Pre-Cooler Configuration Original (2013 Testing) Three Stream He HX: HP/MP/LP + He/GN HX Two Stream He HX: HP/MP + Two Stream He HX: HP/LP + He/GN HX Abandoned Aux. CBX Modified (Re-Designed) (2015 Testing)

5. Page 5 Liquid Nitrogen Usage During spring 2013 commissioning it was found that LN pre- cooler required excessive usage ~3 x design for Mode-1 (max. capacity) ~4 x design for Mode-4 (max. refrigeration) ~4 x design for Mode-2 (nominal) Although this is results in increased operational costs, this does not affect the capacity of the refrigerator Subsequent course of actions –Warm cold box to ambient temperature –Inspect 300 – 80 K HX passes at headers –Perform rapid purging of passes –Re-cool, verifying cleanliness (to avoid any blockage due to contamination) –Re-test

6. Page 6 Liquid Nitrogen Usage At this juncture, having eliminated other reasons, improper flow distribution (or ‘mal-distribution’) was only real alternative to account for the low performance HX’s with high NTU’s (> 30), especially in multi-stream HX’s, are extremely vulnerable to improper flow distribution if not carefully considered in the design Improper flow distribution in high NTU HX’s can be caused by: –Hydrostatic differences (if oriented horizontally) –Insufficient length for longitudinal conduction –Lower flow rates (pressure drop is ~mass flow squared) –Improper layering

7. Page 7 Liquid Nitrogen Usage A simple calculation can show the effect if improper flow distribution on the LN usage; i.e., –If ~7% of the passes are blocked, bypassed (flow biased) or have inadequate heat transfer, this can result in an LN usage rate that is ~3.3 times higher –If the passes were blocked, this would be only a ~16% increase in pressure drop (which would be difficult to measure) Upon analyzing this issue on existing systems further, it appears (perhaps intuitively) that the ‘aspect ratio’ has a significant influence –‘Aspect ratio’ – ratio of the effective length to the square root of the total free flow area (all streams)

8. Page 8 Liquid Nitrogen Usage Consider two HX’s, both with the same heat transfer area Even disregarding longitudinal conduction, it is intuitively plausible to expect that the longer HX will have higher effectiveness because it will have a greater pressure drop This conveys the concept that NTU’s can be thought of as describing the HX length, and, The net thermal rating, or (UA), describing the HX volume Length HP Flow LP Flow

9. Page 9 Liquid Nitrogen Usage Quantitatively, this can be explained as follows: –Ratio of Colburn ‘j’ factor to (Fanning) friction factor, (j H / f) is relatively invariant –Heat transfer coefficient, h = (C p · G/Pr 2/3 ) · (j H /f ) · f –Stream specific NTU; N h = (UA)/C h and N l = (UA)/C l –So, N h and N l ~ f · L (i.e., Fanning friction factor times length) –Taking the simplest case of, C h = C l NTU = 1/(1/N h + 1/N l )  T hl =  T max / (1 + NTU) –Where,  T max = T h,in – T l,in –That is, NTU ~ L and,  T hl ~ 1/L –And, (UA) ~ U ·( A c · L/r h ), where U=(UA)/A, and, A c is the free flow cross-sectional area –So, (UA) ~ A c · L, that is, the volume of the core

10. Page 10 Liquid Nitrogen Usage However, NTU’s and (UA) are NOT geometrical factors! So, next consider two HX’s, both are the same length, but different cross-sections; the left-hand one having a smaller (UA) than the right-hand one Length HP Flow Length Further, lets assume they have the same NTU’s But do they have the same effectiveness ? –Theory says ‘yes’, but empirical data suggests otherwise

11. Page 11 Liquid Nitrogen Usage Aspect ratio for selected systems that use LN pre-cooling

12. Page 12 Liquid Nitrogen Usage Neither the cold box or HX manufacturer could explain the poor performance; original design had, –28% thermal margin (including longitudinal heat conduction) in Mode-1 (max. capacity) –13.9 NTU’s per meter (this is higher than 10 NTU/m), but not too far off –Adequate pressure drop ratios (all greater than 3) Present theory is not adequate for this demanding HX application!

13. Page 13 Liquid Nitrogen Usage The 12 GeV re-design of the sensible heat portion of the LN pre-cooler uses two new cores –Core A: HP, MP streams, 5.5 m long with HP & MP stream re-mixing headers –Core B: HP, LP streams, 5.5 m long with HP & LP stream re-mixing headers –NTU’s per meter and aspect ratios for these are in the range of HX’s that have performed acceptably –Aspect ratio is ≈ 5 Existing HP – nitrogen (gas) core is being re-used Warm (300 K) flow balancing valves control the flow split

14. Page 14 Liquid Nitrogen Usage In systems that use LN pre-cooling, it is quite common to see high LN usage, as compared to the design goal Even in systems that do not, it is not unusual in multi-stream HX’s for the medium pressure stream exiting the cold box (i.e., 300 K temperature level) to be considerably colder than the design Many times this is due to the challenging task of balancing the number of HP stream passes paired to the MP vs. LP streams, while striving to minimize the pressure drop from the load return through the LP stream This is especially the case for a system operating over a wide range of modes (e.g., liquefier to refrigerator) and a wide range of capacities, whether this was planned in the process design or not

15. Page 15 Liquid Nitrogen Usage MSU-FRIB plans to use two HX’s (cores) for the sensible heat portion of the LN pre- cooler –Core A: HP, MP and nitrogen streams –Core B: HP and LP streams –This was same as the option reqested in the JLab 12 GeV cold box RFP –Warm (300 K) flow balancing valves are intended to control the flow split Due to existing practical limitations, this configuration was not used for the 12 GeV 300 – 80 K HX re-design

16. Page 16 Wide Range Testing After the implementation of the re-design by the manufacturer was complete, the 12 GeV cold box under went extensive testing by JLab staff over a wide range of capacities and modes, including design modes The following presents the results of the LN consumption –Met design (within ~15%) for max. refrigeration case and, –Better than design at all other modes tested

17. Page 17 Wide Range Testing As done for 12 GeV commissioning in 2013 and for SNS, testing of cold compressor load on 4.5 K cold box did not require sub-atmospheric helium Helium supply pressure to cold box ranged from 19.5 to 6.5 bar without throttling turbine inlet valves –Except for max. refrigeration test w/out shield; to protect turbine T6 outlet temperature from going below 5 K Transition between different capacities and modes was very stable, requiring minimal operator interaction

18. Page 18 Wide Range Testing For performance data, LP stage compressor bypass adjusted so that there was no bypass –In normal operation, a modest amount of bypass is needed so that if a MP or LP stage compressor trips, the cold compressors can remain operating while the trip issue is addressed –The adjustment was made based upon data taken from a test of the LP stage over a range of slide valve positions and pressure ratios –Plots below show the ratio of the part load to full load efficiency; volumetric and isothermal

19. Page 19 Wide Range Testing Table below summarizes the tests performed

20. Page 20 Wide Range Testing Below is the 4.5-K cold box inverse coefficient of performance (COP inv ) and system exergetic efficiency vs. the equivalent 4.5-K refrigeration load (on an equal exergy basis) As can be seen, the performance changes little down to ~1/3 of the maximum capacity

21. Page 21 Wide Range Testing These extremes reflect different modes –Isothermal 4.5-K refrigeration –Liquefaction (4.5 to 300 K) –Mixed, 50% liquefaction + 50% refrigeration –Cold compressor load (~30 K, 1.15 bar)

22. Page 22 Wide Range Testing Turn-down performance ranges from 19.5 to 6.5 bar supply pressure to the cold box No turbine shut-down or turbine inlet (pressure) throttling is required, and, System can adjust w/ little operator intervention

23. Page 23 Wide Range Testing In regards to the Floating-Pressure Process, note the direct correlation between capacity and HP supply pressure Below ~9 bar supply pressure to the cold box, operator intervention is required to shut down the excess compressor capacity Any three of the five compressors can support the minimum turn- down (at 6.5 bar discharge pressure)

24. Page 24 Wide Range Testing Cold box design has approximately equal ‘Carnot-step’ (expansion stage) mass flows, allowing good performance and turn-down capability Original JLab CHL (4 GeV system) and SNS cold box designs did not utilize the equal ‘Carnot-step’ design basis –Manifest limited efficient turn-down capability –Both have a highly dominate turbine stage mass flows compared to the other turbine stages that require throttling of turbine inlet valves at reduced capacities

25. Page 25 Wide Range Testing Unlike the original (4 GeV) CHL… –Shield load can be supported and transition seamlessly between T1-T2 string, T3-T4 string, or both (hybrid) –Also, the full shield load can be supported at nominal (cold compressor) capacity without running the shield turbines (though at an elevated temperature)

26. Page 26 Wide Range Testing Unlike the original (4 GeV) CHL… While supporting LINAC at 2-K, cold compressors did not trip when, –Either or both of the upper turbine strings (exhausting to the MP stream) trip-off –Any one of the two lower turbine strings would trip off –LP stage compressor would trip off These trips were very rare and due mostly to instrumentation issues

27. Page 27 Wide Range Testing Refrigeration system is usually operated as a fixed inventory system, with a natural floating pressure operation –Automatically monitors for system leakage Apparent by loss in refrigerator’s dewar liquid level –Component or load performance issues Apparent by changes in the HP supply pressure to the cold box

28. Page 28 Wide Range Testing Not withstanding good component efficiencies at design conditions, the primary factors allowing good efficiency over a wide range of capacities and modes are: –Floating – Pressure Process –A compressor system/skid design allowing a wide range of operation (developed at JLab) –An equal ‘Carnot-step’ (expansion stage) cold box design –Variable turbine brake –Proper HX specification/design –Proper sizing of internal piping/components (for a wide range of operation)

29. Page 29 Pulsed Load Test Pulsed load test conducted using the heaters in the 3 m 3 cold box sub-cooler and the 10 m 3 external dewar Intent to impose a pulsed load similarly scaled to that expected to be imposed on the ITER cryo-plants –Peak-to-peak to mean heat load ratio of ~30% –Period of 1800 seconds –Ramp rate of 30 W/s No special or additional instrumentation, controls, algorithms or equipment was implemented for this test (i.e., same system used throughout testing and normal operation) During test compressor bypass valves did not open and there was no mass in or out (to gas storage), nor liquid accumulation

30. Page 30 Pulsed Load Test Top two curves go to left hand axis: total heat load profile (red) and high pressure supply to cold box (blue) Bottom two curves go to right hand axis: low pressure load return (green) and high pressure mass flow to cold box (magneta)

31. Page 31 Pulsed Load Test It should be kept in mind that the primary design mode for the 12 GeV refrigerator was for a cold compressor load w/ a modest amount of 4.5 K (to 300 K) liquefaction (although one of the design modes was a maximum refrigeration load to ensure ‘well- rounded’ HX’s)

32. Page 32 Pulsed Load Test If it were primarily designed for a pure refrigeration load the turbine nozzle coefficients would have been selected differently Also, if the objective is to prevent any variation in the load return pressure, the displacement of the LP stage compressor would have been selected differently

33. Page 33 Pulsed Load Test Three pulsed load test were performed with similar results The table below summarizes the variations This test not only demonstrated the capability of a well-balanced refrigerator system design (compressors and cold box), but also the effectiveness of the Floating – Pressure Process

34. Page 34 Conclusions Re-design of LN pre-cooler has been successfully implemented by manufacturer, producing excellent results Full realization of the Ganni Cycle – Floating Pressure Process has been successfully demonstrated in the JLab 12 GeV cold box and compressor system –This allows a very wide range of operation (19.5 to 6.5 bar supply to the cold box) with good efficiency (changing little down to ~1/3 of max. capacity) Similar successful results for implementation of this process on the NASA-JSC 20-K refrigeration system used for the James Webb project have been presented Additionally, this process is being used for the MSU- FRIB project and is anticipated to be used in other projects

35. Page 35

36. Page 36 Configuration Original Modified

37. Page 37 Design Capacity Specification contained six design ‘modes’ –Modes 1, 3 and 4 are guaranteed by the cold box manufacturer, while –Modes 2, 5 and 6 are not; rather they are termed manufacturer ‘expected’ performance

38. Page 38 Liquid Nitrogen Usage Visual inspection of HX passes at headers –Some debris found in HP stream –Discoloration of MP passes HP, MP and LP passes shown below: Cold box and HX manufacturer, as well as, JLab did not believe this was related to the high LN usage unless this was residual from moisture (which caused blockage of some of the passes)

39. Page 39 Liquid Nitrogen Usage During the re-test in Aug. 2013, four methods were used to cross-check the LN usage (#1 to #3 were used in the spring) –Note: Due to operational constraints, it was not possible to measure the LN storage tank depletion #1 – Energy balance using in-stream temperature measurements at 300 and 80-K temperature levels #2 – Energy balance using surface mounted temperature measurements at the 300 K temperature level and in-stream measurements at 80-K #3 – Energy balance using in-stream temperature measurements at inlet and outlet of helium-nitrogen boiler #4 – Cold box LN vessel depletion rate (i.e., supply is closed) Results similar to the spring testing were obtained with no improvement in the LN consumption

40. Page 40 Liquid Nitrogen Usage The following simple calculation illustrates the effect if improper flow distribution on the LN usage

41. Page 41 Liquid Nitrogen Usage The table below summarizes key HX specifications required by JLab to address proper flow distribution in the RFP Items #3 and #4 are often claimed by manufacturers to be an excessive design specification The aspect ratio was not part of the original HX specification

42. Page 42 Liquid Nitrogen Usage The following table shows a comparison of the original design and re-design that was selected and implemented

43. Page 43 Wide Range Testing A particularly useful component implemented in the 12 GeV compressor system design was a bypass control valve between the recycle return (or MP) stream and the load return (or LP) stream; shown as “BYPML”

44. Page 44 Wide Range Testing This valve allows any excess LP stage capacity, which would otherwise be recirculated by the “BYPL” bypass, to be used to handle recycle load return flow The exergy loss across this valve is small, since the MP and LP streams are usually close in pressure, and is outweighed by the increased pressure ratio across the upper two turbine strings and the inefficiency in implementing the compressor slide valve