- Email: [email protected]
Propane liquefaction with an active magnetic regenerative liquefier
Cryogenics 100 (2019) 69–76
Contents lists available at ScienceDirect
Cryogenics journal homepage: www.elsevier.com/locate/cryogenics
Research paper
Propane liquefaction with an active magnetic regenerative liquefier a,⁎
b
b,c
b
T
b
John Barclay , Kriston Brooks , Jun Cui , Jamelyn Holladay , Kerry Meinhardt , Evgueni Polikarpovb, Edwin Thomsenb a b c
Emerald Energy NW, LLC. (EENW), Bothell, WA 98012, United States Pacific Northwest National Laboratory (PNNL), Richland, WA 99254, United States Ames National Laboratory, Iowa State University, Ames, IA 50010, United States
A R T I C LE I N FO
A B S T R A C T
Keywords: Propane Liquefaction Active-magnetic-regenerator Load-curve Coil-fin-tube-heat-exchanger Gd
Magnetic refrigeration is a well-known cooling technique based on the magnetocaloric effect (MCE) of certain solids as they enter or leave a high magnetic field. An active magnetic regenerator (AMR) uses certain ferromagnetic materials simultaneously as MCE refrigerants and as a regenerator. An effective active magnetic regenerative refrigeration cycle consists of four steps: adiabatic magnetization with no heat transfer gas flow; heat transfer gas flow at constant high field; demagnetization with no heat transfer gas flow; and heat transfer gas flow at constant low field. The first heat transfer gas flow step from a cold-to-hot temperature in this cycle rejects heat from the magnetized regenerator to a hot sink and the second reverse heat transfer gas flow step from a hotto-cold temperature absorbs heat from a cold source. Our primary objectives of the present work were to demonstrate an AMR-cycle liquefier, determine the cooling power of a magnetic refrigerant executing an AMR cycle, and understand the impact of intermittent cooling of the AMR cycle of a reciprocating, dual regenerator design with continuous liquefaction and parasitic heat leaks. This article describes how an AMR-cycle refrigerator using Gd regenerators moving through ∼2.7 T changes at 0.25 Hz was used to liquefy pure propane at two different supply pressures. The measured rates of liquefaction and elapsed times were measured and used to determine the volume collected and derive cooling power at liquefaction conditions for both runs. These results were compared to those obtained from cool-down temperature vs. time data during the same run. The agreement between the two, independent cooling-power results was excellent after the duty cycle of the AMR cycle cooling was properly treated. No direct measurements of the efficiency were made.
1. Introduction 1.1. Prior work While magnetic refrigeration has long been used for refrigeration well below 1 K in research laboratories, use of the technology for refrigeration above 1 K is more recent. In 1966, van Geuns [1] provided the first of numerous studies of magnetic refrigeration for use between ∼4 K and ∼20 K using paramagnetic refrigerants whose total entropy is largely magnetic suitable for magnetic Carnot cycles with adiabatic temperatures from changing magnetic induction in sequence with isothermal connections to thermal sources or heat sinks. Devices using magnetic Carnot cycles successfully liquefied helium. Hakuraku and Ogata [2] produced superfluid helium at 1.8 K with a rotating demagnetized paramagnetic refrigerant that was adiabatically heated up to 4.2 K by the MCE from increasing magnetic induction; at angle of rotation the refrigerant was put in good thermal contact with boiling
⁎
helium to reject heat isothermally as magnetic induction increased to its maximum value. At this point the refrigerant was thermally isolated from the boiling helium before the MCE by partial demagnetization adiabatically cooled the refrigerant until it reached ∼1.8 K where is was put into good thermal contact with a separate bath of liquid helium to be isothermally cooled at ∼1.8 K as demagnetization was completed. Nakagome et al. and Numazawa et al. also used a Carnot-cycle magnetic refrigerator [3,4] operating between ∼20 K and 4.2 K to re-condense helium gas at 0.1013 MPa. Controlled magnetization and demagnetization of paramagnetic gadolinium gallium garnet or dysprosium gallium-iron garnet between high and low magnetic induction with intermittent contact with the 20 K heat sink or 4.2 K heat source. To liquefy cryogens with bubble-point temperatures above ∼20 K, recuperative or regenerative magnetic cycles capable of > 100 K temperature spans were necessary to exploit the MCE. Refrigerator prototypes based on recuperative cycles were demonstrated by Brown in
Corresponding author at: 17505 7th Ave W, Bothell, WA 98012, United States. E-mail address: [email protected] (J. Barclay).
https://doi.org/10.1016/j.cryogenics.2019.01.009 Received 28 August 2018; Received in revised form 30 November 2018; Accepted 26 January 2019 Available online 24 April 2019 0011-2275/ © 2019 Elsevier Ltd. All rights reserved.
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
room-temperature [24]. This requirement was adopted for the magnetocaloric hydrogen liquefaction (MCHL) project at Pacific Northwest National Laboratory (PNNL) [25]. Several experiments with an AMRR prototype with a ∼280 K heat sink temperature have been analyzed, designed, built, and tested to better understand how AMRRs work in comparison to predicted performance. This project has helped understand AMRL optimization for high FOM and the important relationship between specific cooling power of magnetic refrigerant and flows of heat transfer gas through ferromagnetic regenerators during the hot-tocold blow steps of the AMR cycle [26]. In addition, this project has helped appreciate the effect of intermittent cooling of reciprocating dual regenerators on the average measured cooling power with continuous thermal loads of parasitic heat leaks and cooling/liquefaction a process gas. This paper describes results of successful experiments to liquefy propane at two different pressures using a reciprocating AMRR with dual 1.05 kg ferromagnetic Gd regenerators [27,28], moving between 3.3 T and 0.6 T at 0.25 Hz to cool helium gas from ∼280 K to ∼218 K with only parasitic heat leak. The direct cooling and liquefaction of propane with precisely known feed initial gas temperature, pressure, and mass flow rate and final equilibrium liquid temperature, pressure and amount of liquid produced give an independent measurement of cooling power in liquefier mode. Understanding intermittent vs. continuous cooling in this AMRR help guide design of more complex AMRLs for ∼280 K to ∼120 K for LNG, or from ∼280 K to ∼20 K for LH2. No prior report of propane liquefaction with an AMRR has been published to our knowledge.
1976 [5] and Steyert, Jr. in 1978 [6]. A major advance was initiated by invention of the AMR by Barclay and Steyert, Jr. in 1982 [7] wherein the regenerator function of a regenerative refrigeration cycle was combined with the active refrigerant function of the cycle. Early analytical cooling-performance models [8] of active magnetic regenerative refrigerators (AMRRs) explored their use as active magnetic regenerative liquefiers (AMRLs) with potential to substantially increase liquefier efficiency and thereby reduce work required to liquefy key cryogens such as liquid hydrogen (LH2) or liquid natural gas (LNG). Several groups worked on hydrogen liquefaction at 20 K using AMRLs with liquid nitrogen (LN2) at 77 K as the hot heat sink. For example, Astronautics Corporation of America (ACA) developed a hydrogen AMRL design operating between 77 K and 20 K with dual, two-stage, stationary regenerators and two reciprocating high-field superconducting magnets. This device used LN2 as a heat sink and to simultaneously pre-cool hydrogen from room temperature to 77 K [9]. This innovative design, guided by extensive design calculations, component tests, and numerical modeling, incorporated two porous layers of ferromagnetic intermetallic-compound refrigerants, GdPd and GdNi2, and used helium heat transfer gas. This device was the first application of bypass flow of cold helium heat transfer gas [10] to augment liquefier efficiency. The projected AMRL-stage efficiency was over 50% of ideal. Unfortunately, this AMRL prototype was never built and tested by ACA. Barclay and Brook analyzed and disclosed an AMRL design to span from room temperature to 20 K using optimized AMRR stages connected in series with helium heat transfer gas and cold bypass gas at each stage to step-wise cool and liquefy hydrogen [11]. Other experimental work on AMRLs with bypass flow for hydrogen liquefaction was done at Prometheus Energy [12]. Experimental multi-material AMRRs without bypass flow were developed by A. Rowe et al. at the University of Victoria [13–15] and by Chahine et al. at University of Quebec at Three Rivers [16]. Utaki et al. [17] analyzed three different combinations of conceptual AMRR-based liquefier designs combined in series with several different heat exchange fluids but without bypass flow of heat transfer gas to make LH2. They considered 8 AMRR stages rejecting heat into air at 300 K plus use of a Carnot magnetic refrigerator (CMR) stage for hydrogen condensation near 20 K. This group also considered a 5-stage AMRL from ∼120 K to 20 K with LNG as a heat sink at ∼120 K and a CMR stage at 20 K. They also described a 3-stage AMRL with LN2 as a heat sink plus a CMR stage at 20 K to produce LH2. This analytical thermodynamic modeling [18] led to several years of experimental development by Numazawa et al. on design, fabrication, and testing of reciprocatingregenerator AMRL prototypes without bypass gas to operate between 77 K and 20 K to make LH2 using LN2 as the heat sink and for precooling hydrogen [19]. Simultaneously with the Japanese group’s effort, Jeong et al. in Korea also numerically analyzed, designed, built, and tested several two-stage AMRL prototypes for LH2 using LN2 for pre-cooling and as the heat sink [20]. This group used stationary AMRs within stationary superconducting magnets that were charged and discharged in contrast to designs with reciprocating regenerators and stationary magnets or stationary magnets with reciprocating regenerators. A recent report [21] described this group’s latest test results using multi-layer magnetic regenerators with four different rare-earth intermetallic refrigerants. The design complexities of highly-efficient cryogenic AMRs have become very evident from observed differences between experimental results of lab-scale prototypes and predicted performance from numerical models of associated designs [22,23]. It is well known that very high figure of merit (FOM) for liquefaction of many cryogens can only be achieved by elimination of inefficient unit operations in conventional gas-cycle liquefiers. For example, the FOM for production of LN2 is ∼0.40 so when it is used to pre-cool hydrogen or helium in conventional Claude-cycle LH2 and Collins-cycle LHe liquefiers, the resultant FOM is less than 0.4. This fact established the design basis for highly-efficient, multi-stage AMRL designs with hot heat-sinks near
1.2. Active magnetic regenerative refrigerator The magnetic Brayton cycle has two constant magnetic induction and two constant entropy steps. It is a very effective regenerative cycle that enables an AMR cycle to span large temperatures. In the Brayton cycle an element of ferromagnetic refrigerant near its Curie temperature is first warmed by adiabatic magnetization with no heat transfer fluid flow, then cooled by cold-to-hot heat transfer fluid flow at constant magnetic induction, cooled by adiabatic demagnetization with no heat transfer fluid flow, and then warmed by hot-to-cold flow of a heat transfer fluid in the opposite direction to the first flow at constant magnetic induction. During an AMR cycle, elements of refrigerants in high-performance regenerators at slightly different axial temperatures execute different magnetic Brayton cycles thermally coupled together by the periodic, reciprocately flows of heat transfer fluid to create an AMRR [29] capable of pumping heat from TCOLD to THOT. The AMR cycle is a unique cycle that does not have a direct analogy in gas-cycle refrigeration. Using an several layers of different magnetic refrigerants with different Curie temperatures, creates an AMRR to span many times the ΔTS of individual refrigerants. The cooling power of an AMRR is driven by the MCE of magnetic refrigerants, however, cooling power is only achieved by flow of heat transfer fluid through the magnetic regenerators at two constant magnetic induction steps in the AMR cycle. Quantitatively, with optimum hot-to-cold flow rate of heat transfer fluid (e.g., pressurized helium gas) through the demagnetized regenerator, the cooling power of this regenerator, which occurs once every four steps of an AMR cycle, is given by
QĊ (TCOLD ) = νFrCOLD MS CP (TCOLD, BLow )ΔTCD (TCOLD ) In this equation QĊ (TCOLD ) is the cooling power in W, ν is the AMR cycle frequency in Hz, FrCOLD is a numerical parameter related to the fraction of demagnetized regenerator flow length colder than the steady-state cold temperatureTCOLD . This parameter also includes the time integral of the changing cold helium gas temperature over the flow period. This parameter combines the non-linear hot-to-cold temperature profile of the refrigerant, the total cold thermal load on the regenerator, and mass flow rate of heat transfer gas. MS is the total refrigerant mass in kg, CP is the total heat capacity of the refrigerant at 70
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
8.0
parameter has been calculated for various AMR operating conditions by using a detailed 1-D numerical model for the dual Gd regenerators and other parameters used in this experiment [30]. Typical values of FrCOLD are ∼0.25 to ∼0.4. Evaluation of this equation gives cooling power of one regenerator during the hot-to-cold blow of its AMR cycle. The AMRR used for this experiment is illustrated in Fig. 2. It has two identical opposing regenerators separated by the cold heat exchange region as shown in Fig. 3 to provide cooling during half of each AMR cycle.
TD
TU
7.0
6.0
TS [K]
5.0
4.0
3.0
1.3. Magnetocaloric properties of Gd
2.0
The magnetocaloric effect of a magnetic refrigerant is normally characterized by isothermal entropy change and adiabatic temperature change as a function of temperature and magnetic induction [31]. The cooling power equation requires precise values of adiabatic temperature changes and total heat capacity. Using experimentally validated data, the molecular mean field theory equation of state with the Debye model for lattice heat capacity and free-electron model for electronic heat capacity using measured Curie temperature, magnetic moment, Debye temperature, and other known physical properties cab be used to calculate magnetocaloric properties as a function of T and B required for the cooling power equation above. The calculated data agree very well with experimental results [32]. For example, the adiabatic temperature changes upon magnetization (ΔTU) and upon demagnetization (ΔTD) of Gd from the same temperature are shown in Fig. 1. These differences in adiabatic temperature changes are important for calculation of the cooling power vs. T curve of an AMRR.
1.0
0.0 200
210
220
230
240
250
260
270
280
290
300
310
320
Temperature (K)
Fig. 1. Adiabatic temperature changes (ΔTS) for Gd upon magnetization from 0.6 T to 3.3 T (solid blue line) and upon demagnetization from 3.3 T to 0.6 T (dashed red line). The Curie temperature of Gd is ∼293 K. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Actuator drive motor Linear actuator Lid for high vacuum cold box Heat Shield Superconducting Magnet
2. Propane liquefaction using an active magnetic regenerative refrigerator
Cryocooler 1st Stage
2.1. The AMRR apparatus
Cryocooler 2nd Stage
Propane liquefaction temperatures are within the 285–218 K span of a single-stage AMRR [33] prototype at PNNL. Fig. 2 [34] shows an overview of several subsystems of the apparatus attached from the top lid of the cold box (Dewar). The structural supports to the magnet, the cryocooler cold head, and thermal shields, are shown. The super-insulated, air-filled, thin-wall stainless-steel center tube has a welded bottom flange and is hermetically attached to the top lid. It isolates the reciprocating dual magnetic regenerators and cold heat exchanger (CHEX) subsystem from the ∼40-K thermal shield around the conduction-cooled superconducting magnet subsystem at ∼4 K. The cold box is evacuated to ∼1 × 10−7 torr during experiments. Fig. 3 illustrates the reciprocating dual regenerators and condensing heat exchanger used for these runs.
Warm bore: encloses regenerators, heat transfer subsystems and instrumentation
Fig. 2. Cut-away drawing of major subsystems of the AMRR except the reciprocating dual-regenerator-CHEX, the heat transfer pump, the cold box, and the instrumentation/control/DAQ subsystems.
TCOLD and low magnetic induction after demagnetization in J/kg K, andΔTCD is the adiabatic temperature change upon demagnetization from TCOLD (see Fig. 1). All the variables in this equation can be independently measured exceptFrCOLD . For different regenerators this
Fig. 3. Schematic of coiled-fin tube HEX, the propane storage vessel, the dual Gd regenerators, and CHEX subsystem. The regenerators have opposing axial hot-to-cold temperature profiles and alternatively produce cold helium flow through fins of the condensing HEX when one regenerator is in high field and the other is in low field.
71
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
C3 purge & vent line to fume hood Flexible linkages to ss tubes
PSV
F Regulated Lecture Bottle ~0.793 MPa Propane
LabVIEW DAQ for GEN-I
Bypass flow Needle valve
PSV
F Existing Alicat; 1.379 MPa He
Alicat Flow, T, P; 5 SLPM-C3
1.379 MPa He gas flow from/to HTF pump
TC TC
Return He into HTF pump
Reg 1
TC
CHEX
TC TC TC
Reg 2
Coiled fin tube heat exchanger to liquefy/ collect propane via cold He gas from dual magnetic regenerators in AMR cycle 1.379 MPa He gas flow from/to HTF pump
TC
Fig. 4. A simplified process flow diagram with some piping & instrumentation indicated for the AMRR dual regenerator and CHEX subsystem and external subsystems required for safe handling of propane during runs. ‘TC’ indicates multiple thermocouple locations.
2.2. Simplified piping & instrumentation diagram (P&ID) for propane liquefaction in dual regenerator CHEX
Vent
Inlet
Several simple modifications were made to the existing AMRR (referred to as GEN-I in Fig. 4) to safely execute flammable-gas experiments. A schematic of simplified P&ID of the test apparatus is shown in Fig. 4. The temperatures, pressures, flow rates, and actuator controls were measured starting from near room temperature. 2.3. Configuration of AMRR apparatus for propane experiments Storage Vessel
2.3.1. Feedstock supply assembly A lecture bottle filled with 454 g of 99.5+% propane at 0.793 MPa provided the feedstock for several runs. The lecture bottle included a pressure regulator and a manual flow-control valve hermetically-connected to the cold HEX inside the AMRR via a ∼2-m stainless-steel (ss) tube up to the AMRR, a flexible ∼25-cm ss tube, and a ∼75-cm ss tube, all with 0.635 cm o.d. This hermetic tubing supplies propane at desired pressures into the compact ss coiled-fin tube HEX close-fitted around a small, thin-walled ss cylindrical collecting vessel shown in Fig. 5. An Alicat pressure, temperature, and mass-flow measuring meter were installed in the 0.635-cm o.d., ∼3-m long, ss vent line into a nearby fume hood. The hood had a continuous air suction rate large enough to insure excellent dilution of the small amount of propane gas used in these experiments. This vent line also included a check valve and a pressure relief valve (PSV) set to 0.551 MPa.
HEX outlet
Fig. 5. A picture from the top of the compact coiled-fin tube condensing heat exchanger and liquid propane collection vessel. The dimensions of this assembly are in the text.
selected operating conditions. The heat transfer design calculations assumed fully-developed flow. The transient behavior of the oscillatory helium gas flow during the hot-to-cold blow steps of the AMR cycle in the dual regenerators improves the heat transfer and provides design margin. The coil tube and the fin sizes were sized based on convective heat transfer coefficients of the cold helium heat transfer gas flow through the fins brazed to the exterior of the helical tube and propane flow inside the tube. According to standard design calculations for
2.3.2. Cold heat exchanger assembly to condense, liquefy, and collect liquid propane The heat exchanger and collecting vessel assembly was sized to fit into the ∼13.47-cm long by 6.35-cm i.d. of the CHEX. The compact heat exchanger was designed to cool and liquefy propane as it flows through a coiled-fin tube and into a small collecting vessel. This requires removal of 38 W of heat from propane at a pressure of 0.358 MPa at a nominal rate dictated by cooling power from the MCE of the Gd for 72
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
DAQ cards interfaced to the LabVIEW computer. The DAQ program for the AMRR apparatus recorded propane pressures, temperatures, and flow rates and additional temperatures in the CHEX and elsewhere on the AMRR. The total amount of liquid produced was measured with the flow-integration feature of the Alicat sensor for the two different runs. The large LabVIEW data files from these runs were stored in Excel and later transferred to Origin for processing and plotting prior to thermodynamic analysis.
convective heat transfer in tubular heat exchangers [35], the limiting conductance between the propane gas and the inner diameter of the helical tube is ∼12 W/m2/K during precooling. Once propane begins to condense and create a mixed phase inside the tube, the limiting heattransfer conductance is 255 W/m2/K between the ∼1.379 MPa cold helium flowing past the external fins and outer diameter of the tubes. Based on these calculations and dimensional design specifications, a coiled, finned-tube HEX was purchased from Energy Transfer MDE [36] as shown in Fig. 5. The tube and fins are 304 stainless steel with 0.63cm o.d. tube with 0.32 cm high, regularly-spaced 0.05 cm-thick fins brazed onto the tube surface. The effective axial (z) length of eleven close-fitting loops of heat exchanger tubes is ∼14 cm. Each loop is ∼16 cm-long. The total length of ∼180-cm finned-tube was gradually bent into a helical coil, so the exterior edge of the fins on each tube fit snuggly inside the G-10 shell of the CHEX. The interior fin edges of each tube fit snuggly around the stainless-steel collection vessel. The cold helium heat transfer gas in the AMRR is forced to flow through the fins on the coils, resulting in very effective heat transfer. The helium gas pressure drop through these fins was calculated to be less than 8 Pa. The internal volume of the HEX is 27.04 cm3 and it has a mass of 420 g. The propane collection vessel was constructed of a 7.62-cm long, 0.165-cm wall, 316-stainless steel tube section with butt-welded endcaps. The outer diameter of the welded tube was turned down 0.038 cm to form a tight fit inside of the HEX. Small 0.3125-cm diameter holes were drilled into both endcaps and 0.3125 cm, 316 stainless tubes were TIG-welded into each hole. The bottom tube was welded to the outlet of the HEX while the top tube was welded to the vent line. The completed assembly is shown in Fig. 5. The volume of the central collecting and storage vessel is 94.39 cm3. Total liquid propane volume in the coiledtube HEX and collection vessel is 121.43 cm3. The connecting tubing plus the storage vessel have a mass of 190 g. The HEX was held in place within the CHEX with two machined G-10 endcaps epoxy-bonded to both ends of the CHEX tube. The complete assemble with the dual Gd regenerators is illustrated in Fig. 3.
2.4. Propane liquefaction experimental results The average temperatures in the CHEX during liquefaction were 262.5 K and 246.0 K at equilibrium pressures in the HEX of 0.338 MPa and 0.188 MPa, respectively. The small pressure differences at the lecture bottle and the HEX are due to the feed tubing into the AMRR. In the first run, the lecture-bottle pressure regulator was adjusted to 0.358 MPa. The feed gas temperature of 295.7 K was measured at the Alicat sensor located near the lecture bottle. With the flow-control valve open, no flow was expected. After the He pump, and regenerator drive actuators were started, cooling begins immediately. The temperatures at the top (propane inlet) and bottom of the HEX where it connects to the small storage vessel also began to decrease with no measurable flow of propane. When the CHEX reached ∼262.5 K (see Fig. 7), the propane flow rate increased rapidly as shown in Fig. 6, but after ∼200 s drops by almost a factor of two and liquefaction continued at an approximately constant accumulation rate for ∼1500 s. As soon as initial liquid is collected in the storage tank, the large volume decrease upon condensation creates a cryopump that continuously draws propane through the helium-cooled HEX. Although the feed pressure was set by the regulator at the lecture bottle, the flow rate was determined by the cold helium gas in the CHEX and cryopumping in the HEX. In the first run, accumulated propane flow as a function of time illustrated in Fig. 7 steadily rose from zero until a maximum value when flow completely stops. Fig. 7 also shows TCOLD of He gas in the CHEX was approximately equal to that in the coiled-fin tube HEX and stabilized at 262.5 K for several minutes of liquefaction before the collection vessel temperature slowly decreased. This subcooling continued until TCOLD reached ∼245 K when the feed supply flow-rate dropped to zero and accumulated liquid volume was constant. At that time, the rate of subcooling noticeably increased. After about an hour, the entire CHEX assembly approached steady-state at ∼218 K where the AMRR cooling was balanced by parasitic heat leak. The average feed rate for this run was obtained by dividing the mass of liquefied propane by the total
2.3.3. Instrumentation, control, DAQ, and experimental procedures To monitor temperatures at key locations as indicated in Fig. 4, four bare-tipped, 36-gauge, tightly-twisted, Teflon insulated type-E thermocouples were inserted through the outer wall of the CHEX and sealed with epoxy. The bare tips of two thermocouples were dipped into diluted epoxy and cured to provide a very thin layer of electrical insulation before they were attached with epoxy to the entrance and exit ports of the condensing HEX. The other two bare-tipped thermocouples were mounted in the helium gas flow path near the bottom and top of the liquid collector vessel shown in Fig. 5 (and schematically in Fig. 3). Prior to beginning runs, the entire process gas assembly shown was filled several times with clean, dry nitrogen to ∼0.414 MPa and vented into the fume hood to ∼0.110 MPa before the assembly was filled with propane to 0.414 MPa and vented into the fume hood. This procedure was repeated several times to reduce oxygen concentration within the liquefaction subsystem to below lower flammable limits of any potential mixture. Finally, the subsystem was pressurized to the desired propane feed pressure using the regulator on the lecture bottle with the supply valve left open during the run. The conduction-cooled, superconducting magnet was charged to 3.3 T and put into persistent mode. Temperatures of both external chillers were adjusted to maintain ∼285 K inlet helium gas into the hot end of top or bottom demagnetized dual regenerator for the hot-to-cold blow steps. Before the reciprocating drive actuator was started, helium gas was pumped back and forth for several minutes until a uniform initial temperature of ∼285 K was established at the condensing HEX, the dual regenerators and CHEX subsystems. A LabVIEW-based data acquisition system (DAQ) was used to control motion of actuators, to monitor all sensors, and record all data during AMR cycles. The actuators were controlled via serial communications and sensors were read using National Instruments compact-
Fig. 6. The instantaneous propane flow rate in standard liters per minute as a function of time during the liquefaction run at 0.358 MPa. Note the time scale in this figure is different from that in Fig. 7. 73
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
0.358 MPa propane 40
290
25 260
20
250
15
240
10 5
Top_propane_HEX Bottom_Propane_HEX
230
Total propane (SL)
30
Cooling power (W)
Total propane
270
T (K)
30
35
280
10
-5 1000
2000
20
0
220 0
Gd cool down Propane liquefaction
210
3000
220
230
240
250
260
270
Temperature (K)
Time (s) Fig. 7. Temperatures on the coiled-fin tube heat exchanger and collection vessel and total volume of propane as a function of time for liquefaction in the first run at 0.358 MPa feed pressure. The red dot is where temperature vs. time data were expanded (not shown) over 3–4 cycles for analysis of Gd cooling power from cool-down rate. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 9. Several data points showing the cooling power from the dual Gd regenerators in AMRR apparatus obtained from the rate of cooling of the CHEX and regenerators given by cool-down temperatures vs. time at specific average cold temperatures are shown in this plot of thermal load vs. temperature. The cooling powers for two propane liquefaction runs are added as green stars at the dew point temperatures. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
elapsed time of liquefaction. In a second run, the supply pressure was reduced to 0.200 MPa and everything was prepared in the same manner as in first run. This time the average CHEX cold temperature dropped to ∼246 K before propane started to flow and liquid accumulated in the collection vessel. The plotted temperatures at the inlet and exit of the HEX and the accumulated volume of propane vs. time are shown in Fig. 8. In the second run liquefaction began at an average temperature of 246 K which is the equilibrium temperature of liquid propane at 0.188 MPa. Liquefaction continued at an approximately constant accumulation rate for 2240 s with subcooling of the collected liquid propane as in the first run. After approximately steady-state conditions were reached at the end of the runs, the actuators were stopped. The vent valve that bypassed the PSV was slowly opened enough to reduce the propane-storage vessel pressure to ∼0.110 MPa as the CHEX assembly gradually warmed up toward ∼285 K over about 2 h.
2.5. Analysis of propane liquefaction experiments 2.5.1. Cool-down curve analysis The cooling power of the AMRR was obtained from the rate of cool down (K/s) of the T vs. time curves in Figs. 7 and 8. The thermal masses of the CHEX, the HEX/storage vessel, and colder portion of the dual regenerators were obtained using known properties. In the first experiment, by expanding T vs. time over 3 complete AMR cycles (12 s) at the red dot indicated at 233.5 K in Fig. 7, the cooling power was 33.4 W. Parasitic heat leaks from radiation and convection of oscillating air inside the central tube in Fig. 2 are automatically included in this result. Similar analyses of T vs. time data were repeated at selected points in Fig. 8. In this AMRR with equal time steps in the AMR cycle, cooling only occurs during the hot-to-cold flow steps, i.e., only ½ of the 4-second period. Hence, to obtain continuous cooling values, the cooldown rate results were divided by 2 and plotted in Fig. 9. A linear leastsquare fit was drawn through the red points for illustration.
0.200 MPa propane 40
280
30
T (K)
260
25
250
20
240
15 10
Total propane (SL)
Total propane
270
2.5.2. Propane liquefaction data analysis The Alicat mass flow meter at the lecture bottle was factory-calibrated for propane with flow rate in SLPM and total flow in SL at standard conditions of 298.16 K and 0.1013 MPa. The density of pure propane at these standard conditions is 1.832 kg/m3 which enables conversion of SLPM to kg/s and SL to kg of propane. In the first run, total propane liquefied was ∼36.0 standard for liters and in the second run total propane liquefied was ∼36.5 standard liters. These values are consistent because due to the slight density differences. The 0.358-MPa run yielded 65.9 g of liquid and the 0.200-MPa run produced 68.4 g of liquid. As expected, the ratio of these values is close to the ratio of the densities of liquid propane at 262.5 K and 246 K. This agreement confirmed liquid filled the coils of the HEX and enclosed collection vessel in both runs. The average rate of propane liquefaction from accumulated-mass data in 0.358-MPa run is 4.40 × 10−5 kg/s. Multiplying this value times the measured total liquefaction time and dividing by density of propane at 262.5 K gives total liquid volume of 121 cm3. Likewise, the same calculation for data from the 0.200-MPa run gives an average liquefaction mass flow rate of 2.98 × 10−5 kg/s and a total volume of 121 cm3. Not only do these results perfectly agree with each other, they agree with the combined internal volumes of the HEX and collection
35
230 5 Top_propane_HEX Bottom_Propane_HEX
220 210 0
1000
2 2000 000
0 -5 3000
Time (s) Fig. 8. Temperature on the coiled-fin tube HEX and total propane as a function of time for liquefaction in the second run at 0.200 MPa feed pressure. The red dots were places where the temperature vs. time data were expanded (not shown) over 3–4 cycles for analysis of the cool-down rate. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 74
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
thermal mass of the heat exchanger, the collection vessel, and the liquefied propane much more than by the intermittent flow of cold helium gas in the CHEX. This observation also corroborates that the process gas flow rate through process heat exchangers in cryogenic liquefiers with external process heat exchangers is not controlled by an inlet valve, but instead is controlled by the cooling power at the dew point temperature of the process gas at the pressure at the heat exchanger. This is commonly known in small-scale LNG liquefaction plants, but it is well reinforced herein. In the propane liquefaction experiments, measurement of temperatures in the CHEX and liquefier HEX vs. time provide a good indication of the changing thermal loads from the propane and cool-down of thermal masses in the CHEX and HEX. For instance, as soon as the dew temperature of the propane is reached for a constant propane pressure, the rate of temperature change in the CHEX and HEX is approximately zero as liquid initially collects. This confirms the latent heat from propane liquefaction is much larger than the sensible heat from gradually cooling the thermal masses in the CHEX/HEX saturates the cooling rate of the cold helium gas flowing through the CHEX. The CHEX/HEX cool-down temperature curve changes again once liquid propane accumulates in the collection vessel and again when propane liquefaction ceases. The temperature continues to drop as the sensible heat of all the cold parts of the CHEX/HEX and the liquid propane are cooled to ∼218 K which is the ultimate cold temperature limit of the AMRR at these operating conditions from the parasitic heat leak of ∼6–7 W. This is consistent with calculated parasitic heat leaks from radiation, convection, and conduction mechanisms in this AMRR apparatus. Another observation from Figs. 7 and 8 is that the rate of accumulated propane mass during liquefaction is approximately constant during the entire liquefaction time span. At first glance, this is puzzling given cooling from the dual Gd regenerators occurs during hot-to-cold blows of helium gas from the demagnetized regenerator into the CHEX every other second during an AMR cycle. However, this result shows the thermal capacitance of liquid propane creates a ‘vapor-collapse’ condition commonly seen in Dewars with a liquid cryogen when additional vapor is added. In the present case, the condensation rate is steady, and resultant slight warming of liquid is too small to cause a change in the cryopumping pressure difference that would affect the flow rate. This situation also occurs in the CHEX when its thermal mass is large enough to smooth out temperature fluctuations from intermittent cooling during the AMR cycle when an external thermal load from a continuous heater in the CHEX or from parasitic heat leaks are used to determine the load curve of the AMRR. The agreement of the measured cooling power from propane liquefaction runs and rates of cool-down in Fig. 9 is good. The comparison of the expected cooling power from the equation given earlier in the paper is several times larger than the measured results in Fig. 9. This result shows that even the maximum mass flow rate of the helium heat transfer gas during the hot-to-cold blow (∼4 gm/sec at the hot end of the regenerators) is much less than required to remove all potential cooling of 1.05 kg of demagnetized Gd. This requires a helium gas pressure of 2–3 times higher than the present value of 1.38 MPa in the sealed reciprocating pump at room temperature. Alternatively, using smaller Gd regenerators are a possibility, although intrinsic parasitic heat leaks become more problematic if this is done. The calculated Biot values suggest that temperature gradients inside the ss tubes in the HEX and in the G-10 walls of the CHEX are small enough to be ignored for the interpretation of the results of the propane runs. We did not measure the work rate input, or the hot sink heat rejection rate, so we could not experimentally measure the FOM of this AMRR during the liquefaction experiment. Our detailed performance simulation models of AMRRs predict FOMs of ∼0.6 can be achieved. However, it is also clear that when one includes all auxiliary power of a complete liquefier system, e.g., the helium scroll compressor power of
vessel. This is expected because these two containers are the only process stream elements cooled by the cold helium gas. As soon as they fill up, the cryopumping pressure-difference goes to zero and flow from the lecture bottle stops. The enthalpy difference between liquid propane at 262.5 K and 0.338 MPa and gaseous propane at 295.7 K and 0.358 MPa times 4.40 × 10−5 kg/s for this run gives an average cooling power during liquefaction of 19.6 W. Fig. 7 also shows that during the last 1213 s of this run, the liquid was slowly subcooled from 262.5 K to 246.0 K. The average heat capacity of propane times the liquid mass over the subcooling temperature span at 0.338 MPa divided by sub-cooling elapsed time gives an additional cooling power of 2.1 W for this run. The total AMRR cooling power of 21.7 W in this run was plotted as a green star at 262.5 K in Fig. 9. Similarly, the enthalpy difference at 246.0 K and 0.188 MPa and at 295.2 K and 0.200 MPa times the average mass flow rate 2.97 × 10−5 kg/s for this run gives a cooling power of 14.5 W. Subcooling from 246.0 K to 233.0 K requires 1.1 W of cooling power for a total of 15.6 W. This result is also plotted as a green star at 246 K in Fig. 9. The thermal loads into the CHEX control volume from liquefaction, subcooling, and parasitic heat leaks are continuous during the 4-second AMR cycle. 2.6. Eddy current and Biot number calculations Eddy currents are induced in each closed conducting loop from the magnetic induction changes during the AMR cycle within the ss finned tubes in the cooling and condensing HEX shown in Fig. 3. Each of the coiled fin tubes is in electrical contact with the thin-walled ss collection vessel. Calculating the emf caused by the magnetic flux changes, the resistance of each HEX loop and that of the storage vessel using known dimensions and resistivity properties, gives the total eddy current thermal load into the CHEX of ∼134 mW for 3.3 T to 0.6 T magnetic induction changes in 1 s twice per AMR cycle. This thermal load is negligible compared to the cooling power of the AMRR. The Biot number of the ss coiled-fin tube HEX during liquefaction of propane was calculated at 246 K and 262.5 K. During the design of the HEX, standard heat transfer calculations show the limiting thermal conductance changes from the propane inside the tubes to helium outside the tubes once liquefaction begins. The two-phase condensation propane conductance is therefore higher than that of the helium gas conductance of 255 W/m2 K. Using this conductance value for the propane flow and neglecting the brazed thin fins on the outside of the tubes, with the tube wall-thickness as the characteristic length with known ss thermal conductivity at these temperatures, the Biot numbers are 0.0.12 and 0.0.11 at 246 K and 262.5 K, respectively. The wall thickness of the G-10 in the CHEX, its known thermal conductivity, and an estimated value of thermal conductance for convection flow of helium in the CHEX gave Biot numbers of 0.18 and 0.16 for the G-10 walls in CHEX at 246 K and 262.5 K, respectively. 3. Discussion These experiments confirm our hypothesis that the HEX and storage vessel become a cryopump as soon as they are cooled by cold helium gas from the dual Gd regenerators to the condensation temperature of propane at a selected pressure. The cryopump action draws propane from the lecture bottle until the cooled HEX and collection vessel are filled and the pressure differential due to propane condensation drops to zero, stopping the cryopumping, and no further liquefaction occurs. This is exactly what was observed in both experiments. Based on the relatively steady accumulated mass flow rate observed in both experiments, the flow caused by the cryopump action is continuous (at least compared to the DAQ data rate which is typically 0.1 s or less) and does not follow the intermittent cooling of the dual regenerators for the reciprocating AMRR operating conditions. This result shows that thermal time constant of the condensation of propane is driven by the cold 75
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
the two-stage GM cryocooler that cools the superconducting magnet to ∼4 K and the thermal shield to ∼40 K, the FOM of the present size AMRR will be very small. This observation means that cryogenic active magnetic regenerative liquefiers will have a minimum liquefaction capacity such as 1–10 metric ton/day of cryogens before they can achieve their potential FOMs of 0.5 or higher.
1984;29:581–7. [4] Numazawa T, et al. The helium magnetic refrigerator: II – liquefaction process and efficiency. Adv Cryog Eng 1984;29:589–96. [5] Brown GV. Magnetic heat pumping near room temperature. J Appl Phys 1976;47:3673–80. [6] Steyert Jr. WA. Stirling-cycle rotating magnetic refrigerator and heat pump for use near room temperature. J Appl Phys 1978;49:1216. [7] Barclay JA, Steyert WA Jr. Active magnetic regenerator. U.S. Patent 4,332,135; June 1, 1982. [8] Barclay JA. The theory of an active magnetic regenerative refrigerator. Proc of the 2nd biennial conf on refrigeration for cryogenic sensors and electronic systems. Greenbelt, MD: NASA Goddard Space Flight Center; 1982. [9] Astronautics Corporation of America, Final Report to DOE on Contract No. DEAC02-90CE40895 compiled by D. J. Janda, chief project engineer. Magnetic liquefaction of hydrogen; July 2, 1992. [10] Holladay JD, Teyber RP, Meinhardt KD, Polikarpov E, Thomsen EC, Archipley CC, et al. Investigation of bypass fluid flow in an active magnetic regenerative liquefier. Cryogenics 2018;93. https://doi.org/10.1016/j.cryogenics.2018.05.010. [11] Barclay JA, Brook TC. Apparatus and methods for cooling and liquefying a fluid using magnetic refrigeration. U.S. Patent 6,467,274 B2; October 22, 2002. [12] Oseen-Senda K, Skrzypkowski MP, Barclay JA. Active magnetic regenerative hydrogen liquefier. Contract No. DOE SBIR DE-FG02-07ER84783; SBIR phase I final report; 2008. [13] Rowe A, Tura A. Experimental investigation of a three-material layered active magnetic regenerator. Int J Refrig 2006;29:1286–93. [14] Rowe A. Thermodynamics of active magnetic regenerators: Part I. Cryogenics 2012;52:111–8. [15] Rowe A. Thermodynamics of active magnetic regenerators: Part II. Cryogenics 2012;52:119–28. [16] Smaili A, Chahine R. Thermodynamic investigations on optimum active magnetic regenerators. Cryogenics 1998;38:247–52. [17] Utaki T, et al. Research on a magnetic refrigeration cycle for hydrogen liquefaction. Cryocoolers 2007;14:645. [18] Matsumoto K, et al. Numerical analysis of AMRs for hydrogen magnetic refrigerators between 20 K and 77 K. Cryogenics 2014;51:353. [19] Numazawa T, Kamiya K, Utaki T, Matsumoto K. Magnetic refrigerator for hydrogen liquefaction. Cryogenics 2014;62:185. [20] Kim Y, Park I, Jeong S. Experimental investigation of two-stage active magnetic regenerative refrigerator operating between 77K and 20K. Cryogenics 2013;57:113. [21] Park I, Jeong S. Development of the active magnetic regenerative refrigerator operating between 77 K and 20 K with the conduction cooled high-temperature superconducting magnet. Cryogenics 2017;88:106. [22] Yu B, et al. A review of magnetic refrigerator and heat pump prototypes built before the year 2010. Int’l J Refrigeration 2010;33:1029–60. [23] Zimm C, et al. Description and performance of a near room temperature magnetic refrigerator. Adv Cryog Eng 1998;43:1759–66. [24] Barclay JA, Oseen-Senda K, Ferguson L, Cousins A, Pouresfandiary J, Ralph H, et al. Active magnetic regenerative liquefier for hydrogen. Final tech report: DOE Contract Number DE-FG36-08GO18064; 2014. [25] Holladay J, Meinhardt K, Polikarpov E, Thomsen E, Barclay J, Cui J, et al. MagnetoCaloric hydrogen liquefaction. DOE annual merit review, PD131. Tech report; 2017. [26] Holladay J, Teyber R, Meinhardt K, Polikarpov E, Thomsen E, Archipley C, et al. Investigation of bypass fluid flow in an active magnetic regenerative liquefier. Cryogenics 2018. https://doi.org/10.1016/j.cryogenics.2018.05.010. [in press]. [27] Barclay JA, Oseen-Senda K, Skrzypkowski MP. Unique features of liquefaction of hydrogen and natural gas using magnetic refrigeration. Int’l Inst Refrig; Proc of Thermag VI, Victoria, B.C.; 7–10 September 2014. [28] Teyber R, Meinhardt K, Polikarpov E, Thomsen E, Holladay J, Cui J, et al. Performance investigation of a high-field active magnetic regenerator. Appl Energy 2018. https://doi.org/10.1016/j.apenergy.2018.12.012. [submitted for publication]. [29] See reference 7. [30] See references 23 and 11 for more details of numerical model. [31] Tishin A, Gschneidner Jr. K, Pecharsky V. Magnetocaloric effect and heat capacity in the phase-transition region. Phys Rev B 1999;59:503–11. [32] Dan’kov S, Tishin A, Pecharsky V, Gschneidner Jr. K. Magnetic phase transitions and the magnetothermal properties of gadolinium. Phys Rev 1998;57B:3478–90. [33] See references 23 and 24 for more details of this active magnetic regenerative refrigerator prototype. [34] The original version of this drawing was created at Prometheus Energy Group Inc.; see reference 24 for more details. [35] See reference 27 for example. [36] https://www.finnedtube.com.
4. Conclusions The use of an AMRR to successfully liquefy propane has been demonstrated. The excellent agreement between predicted and observed results confirmed our understanding of the design. Allowance for the intermittent cooling of a reciprocating dual regenerator design was proven important for future reciprocating dual regenerator liquefier designs. The results of this work motivated the conceptualization of a new liquefier design using two sets of identical reciprocating dual-regenerator AMRRs with their drive actuators phased 90 degrees apart. The movement of the quadruple combination of active regenerators coupled with properly controlled heat transfer gas flows using several 3-way valves to direct the cold helium heat transfer gas in counterflow with a process gas can provide its continuous cooling. Such a design for liquid hydrogen would require four multi-material magnetic regenerators and two superconducting magnets for a high-FOM ∼280 K to ∼120 K stage and another four multi-material magnetic regenerators and two superconducting magnets for a high-FOM ∼120 K to ∼20 K stage. This conclusion suggests using a continuously rotating wheeltype AMRR configuration with regenerators attached to the rim may be a simpler alternative if the rotary seals and magnet configuration issues can be resolved. Such development helps validate predictive performance models and identify design issues for advanced designs of AMRR prototypes for liquefaction of colder cryogens such as LNG and LH2. Acknowledgements It is a pleasure to acknowledge financial and programmatic support of the Fuel Cell Technology Office of the Energy Efficiency and Renewable Energy Branch of the Department of Energy. The pro-active support of DOE program managers Dr. Erika Gupta and Dr. Neha Rustagi is deeply appreciated. Dr. John Barclay appreciates the opportunity for EENW to participate in the MCHL project at PNNL. Disclaimer statement The views and opinions of the authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. References [1] van Geuns JR. A study of a new magnetic refrigeration cycle. Doctoral Thesis. Univ. of Leiden; 1966. [2] Hakuraku Y, Ogata H. A rotary magnetic refrigerator for superfluid helium production. J Appl Phys 1986;60(9):3266–8. https://doi.org/10.1063/1.337716. [3] Nakagome H, et al. The helium magnetic refrigerator: I – Design. Adv Cryog Eng
76
Contents lists available at ScienceDirect
Cryogenics journal homepage: www.elsevier.com/locate/cryogenics
Research paper
Propane liquefaction with an active magnetic regenerative liquefier a,⁎
b
b,c
b
T
b
John Barclay , Kriston Brooks , Jun Cui , Jamelyn Holladay , Kerry Meinhardt , Evgueni Polikarpovb, Edwin Thomsenb a b c
Emerald Energy NW, LLC. (EENW), Bothell, WA 98012, United States Pacific Northwest National Laboratory (PNNL), Richland, WA 99254, United States Ames National Laboratory, Iowa State University, Ames, IA 50010, United States
A R T I C LE I N FO
A B S T R A C T
Keywords: Propane Liquefaction Active-magnetic-regenerator Load-curve Coil-fin-tube-heat-exchanger Gd
Magnetic refrigeration is a well-known cooling technique based on the magnetocaloric effect (MCE) of certain solids as they enter or leave a high magnetic field. An active magnetic regenerator (AMR) uses certain ferromagnetic materials simultaneously as MCE refrigerants and as a regenerator. An effective active magnetic regenerative refrigeration cycle consists of four steps: adiabatic magnetization with no heat transfer gas flow; heat transfer gas flow at constant high field; demagnetization with no heat transfer gas flow; and heat transfer gas flow at constant low field. The first heat transfer gas flow step from a cold-to-hot temperature in this cycle rejects heat from the magnetized regenerator to a hot sink and the second reverse heat transfer gas flow step from a hotto-cold temperature absorbs heat from a cold source. Our primary objectives of the present work were to demonstrate an AMR-cycle liquefier, determine the cooling power of a magnetic refrigerant executing an AMR cycle, and understand the impact of intermittent cooling of the AMR cycle of a reciprocating, dual regenerator design with continuous liquefaction and parasitic heat leaks. This article describes how an AMR-cycle refrigerator using Gd regenerators moving through ∼2.7 T changes at 0.25 Hz was used to liquefy pure propane at two different supply pressures. The measured rates of liquefaction and elapsed times were measured and used to determine the volume collected and derive cooling power at liquefaction conditions for both runs. These results were compared to those obtained from cool-down temperature vs. time data during the same run. The agreement between the two, independent cooling-power results was excellent after the duty cycle of the AMR cycle cooling was properly treated. No direct measurements of the efficiency were made.
1. Introduction 1.1. Prior work While magnetic refrigeration has long been used for refrigeration well below 1 K in research laboratories, use of the technology for refrigeration above 1 K is more recent. In 1966, van Geuns [1] provided the first of numerous studies of magnetic refrigeration for use between ∼4 K and ∼20 K using paramagnetic refrigerants whose total entropy is largely magnetic suitable for magnetic Carnot cycles with adiabatic temperatures from changing magnetic induction in sequence with isothermal connections to thermal sources or heat sinks. Devices using magnetic Carnot cycles successfully liquefied helium. Hakuraku and Ogata [2] produced superfluid helium at 1.8 K with a rotating demagnetized paramagnetic refrigerant that was adiabatically heated up to 4.2 K by the MCE from increasing magnetic induction; at angle of rotation the refrigerant was put in good thermal contact with boiling
⁎
helium to reject heat isothermally as magnetic induction increased to its maximum value. At this point the refrigerant was thermally isolated from the boiling helium before the MCE by partial demagnetization adiabatically cooled the refrigerant until it reached ∼1.8 K where is was put into good thermal contact with a separate bath of liquid helium to be isothermally cooled at ∼1.8 K as demagnetization was completed. Nakagome et al. and Numazawa et al. also used a Carnot-cycle magnetic refrigerator [3,4] operating between ∼20 K and 4.2 K to re-condense helium gas at 0.1013 MPa. Controlled magnetization and demagnetization of paramagnetic gadolinium gallium garnet or dysprosium gallium-iron garnet between high and low magnetic induction with intermittent contact with the 20 K heat sink or 4.2 K heat source. To liquefy cryogens with bubble-point temperatures above ∼20 K, recuperative or regenerative magnetic cycles capable of > 100 K temperature spans were necessary to exploit the MCE. Refrigerator prototypes based on recuperative cycles were demonstrated by Brown in
Corresponding author at: 17505 7th Ave W, Bothell, WA 98012, United States. E-mail address: [email protected] (J. Barclay).
https://doi.org/10.1016/j.cryogenics.2019.01.009 Received 28 August 2018; Received in revised form 30 November 2018; Accepted 26 January 2019 Available online 24 April 2019 0011-2275/ © 2019 Elsevier Ltd. All rights reserved.
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
room-temperature [24]. This requirement was adopted for the magnetocaloric hydrogen liquefaction (MCHL) project at Pacific Northwest National Laboratory (PNNL) [25]. Several experiments with an AMRR prototype with a ∼280 K heat sink temperature have been analyzed, designed, built, and tested to better understand how AMRRs work in comparison to predicted performance. This project has helped understand AMRL optimization for high FOM and the important relationship between specific cooling power of magnetic refrigerant and flows of heat transfer gas through ferromagnetic regenerators during the hot-tocold blow steps of the AMR cycle [26]. In addition, this project has helped appreciate the effect of intermittent cooling of reciprocating dual regenerators on the average measured cooling power with continuous thermal loads of parasitic heat leaks and cooling/liquefaction a process gas. This paper describes results of successful experiments to liquefy propane at two different pressures using a reciprocating AMRR with dual 1.05 kg ferromagnetic Gd regenerators [27,28], moving between 3.3 T and 0.6 T at 0.25 Hz to cool helium gas from ∼280 K to ∼218 K with only parasitic heat leak. The direct cooling and liquefaction of propane with precisely known feed initial gas temperature, pressure, and mass flow rate and final equilibrium liquid temperature, pressure and amount of liquid produced give an independent measurement of cooling power in liquefier mode. Understanding intermittent vs. continuous cooling in this AMRR help guide design of more complex AMRLs for ∼280 K to ∼120 K for LNG, or from ∼280 K to ∼20 K for LH2. No prior report of propane liquefaction with an AMRR has been published to our knowledge.
1976 [5] and Steyert, Jr. in 1978 [6]. A major advance was initiated by invention of the AMR by Barclay and Steyert, Jr. in 1982 [7] wherein the regenerator function of a regenerative refrigeration cycle was combined with the active refrigerant function of the cycle. Early analytical cooling-performance models [8] of active magnetic regenerative refrigerators (AMRRs) explored their use as active magnetic regenerative liquefiers (AMRLs) with potential to substantially increase liquefier efficiency and thereby reduce work required to liquefy key cryogens such as liquid hydrogen (LH2) or liquid natural gas (LNG). Several groups worked on hydrogen liquefaction at 20 K using AMRLs with liquid nitrogen (LN2) at 77 K as the hot heat sink. For example, Astronautics Corporation of America (ACA) developed a hydrogen AMRL design operating between 77 K and 20 K with dual, two-stage, stationary regenerators and two reciprocating high-field superconducting magnets. This device used LN2 as a heat sink and to simultaneously pre-cool hydrogen from room temperature to 77 K [9]. This innovative design, guided by extensive design calculations, component tests, and numerical modeling, incorporated two porous layers of ferromagnetic intermetallic-compound refrigerants, GdPd and GdNi2, and used helium heat transfer gas. This device was the first application of bypass flow of cold helium heat transfer gas [10] to augment liquefier efficiency. The projected AMRL-stage efficiency was over 50% of ideal. Unfortunately, this AMRL prototype was never built and tested by ACA. Barclay and Brook analyzed and disclosed an AMRL design to span from room temperature to 20 K using optimized AMRR stages connected in series with helium heat transfer gas and cold bypass gas at each stage to step-wise cool and liquefy hydrogen [11]. Other experimental work on AMRLs with bypass flow for hydrogen liquefaction was done at Prometheus Energy [12]. Experimental multi-material AMRRs without bypass flow were developed by A. Rowe et al. at the University of Victoria [13–15] and by Chahine et al. at University of Quebec at Three Rivers [16]. Utaki et al. [17] analyzed three different combinations of conceptual AMRR-based liquefier designs combined in series with several different heat exchange fluids but without bypass flow of heat transfer gas to make LH2. They considered 8 AMRR stages rejecting heat into air at 300 K plus use of a Carnot magnetic refrigerator (CMR) stage for hydrogen condensation near 20 K. This group also considered a 5-stage AMRL from ∼120 K to 20 K with LNG as a heat sink at ∼120 K and a CMR stage at 20 K. They also described a 3-stage AMRL with LN2 as a heat sink plus a CMR stage at 20 K to produce LH2. This analytical thermodynamic modeling [18] led to several years of experimental development by Numazawa et al. on design, fabrication, and testing of reciprocatingregenerator AMRL prototypes without bypass gas to operate between 77 K and 20 K to make LH2 using LN2 as the heat sink and for precooling hydrogen [19]. Simultaneously with the Japanese group’s effort, Jeong et al. in Korea also numerically analyzed, designed, built, and tested several two-stage AMRL prototypes for LH2 using LN2 for pre-cooling and as the heat sink [20]. This group used stationary AMRs within stationary superconducting magnets that were charged and discharged in contrast to designs with reciprocating regenerators and stationary magnets or stationary magnets with reciprocating regenerators. A recent report [21] described this group’s latest test results using multi-layer magnetic regenerators with four different rare-earth intermetallic refrigerants. The design complexities of highly-efficient cryogenic AMRs have become very evident from observed differences between experimental results of lab-scale prototypes and predicted performance from numerical models of associated designs [22,23]. It is well known that very high figure of merit (FOM) for liquefaction of many cryogens can only be achieved by elimination of inefficient unit operations in conventional gas-cycle liquefiers. For example, the FOM for production of LN2 is ∼0.40 so when it is used to pre-cool hydrogen or helium in conventional Claude-cycle LH2 and Collins-cycle LHe liquefiers, the resultant FOM is less than 0.4. This fact established the design basis for highly-efficient, multi-stage AMRL designs with hot heat-sinks near
1.2. Active magnetic regenerative refrigerator The magnetic Brayton cycle has two constant magnetic induction and two constant entropy steps. It is a very effective regenerative cycle that enables an AMR cycle to span large temperatures. In the Brayton cycle an element of ferromagnetic refrigerant near its Curie temperature is first warmed by adiabatic magnetization with no heat transfer fluid flow, then cooled by cold-to-hot heat transfer fluid flow at constant magnetic induction, cooled by adiabatic demagnetization with no heat transfer fluid flow, and then warmed by hot-to-cold flow of a heat transfer fluid in the opposite direction to the first flow at constant magnetic induction. During an AMR cycle, elements of refrigerants in high-performance regenerators at slightly different axial temperatures execute different magnetic Brayton cycles thermally coupled together by the periodic, reciprocately flows of heat transfer fluid to create an AMRR [29] capable of pumping heat from TCOLD to THOT. The AMR cycle is a unique cycle that does not have a direct analogy in gas-cycle refrigeration. Using an several layers of different magnetic refrigerants with different Curie temperatures, creates an AMRR to span many times the ΔTS of individual refrigerants. The cooling power of an AMRR is driven by the MCE of magnetic refrigerants, however, cooling power is only achieved by flow of heat transfer fluid through the magnetic regenerators at two constant magnetic induction steps in the AMR cycle. Quantitatively, with optimum hot-to-cold flow rate of heat transfer fluid (e.g., pressurized helium gas) through the demagnetized regenerator, the cooling power of this regenerator, which occurs once every four steps of an AMR cycle, is given by
QĊ (TCOLD ) = νFrCOLD MS CP (TCOLD, BLow )ΔTCD (TCOLD ) In this equation QĊ (TCOLD ) is the cooling power in W, ν is the AMR cycle frequency in Hz, FrCOLD is a numerical parameter related to the fraction of demagnetized regenerator flow length colder than the steady-state cold temperatureTCOLD . This parameter also includes the time integral of the changing cold helium gas temperature over the flow period. This parameter combines the non-linear hot-to-cold temperature profile of the refrigerant, the total cold thermal load on the regenerator, and mass flow rate of heat transfer gas. MS is the total refrigerant mass in kg, CP is the total heat capacity of the refrigerant at 70
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
8.0
parameter has been calculated for various AMR operating conditions by using a detailed 1-D numerical model for the dual Gd regenerators and other parameters used in this experiment [30]. Typical values of FrCOLD are ∼0.25 to ∼0.4. Evaluation of this equation gives cooling power of one regenerator during the hot-to-cold blow of its AMR cycle. The AMRR used for this experiment is illustrated in Fig. 2. It has two identical opposing regenerators separated by the cold heat exchange region as shown in Fig. 3 to provide cooling during half of each AMR cycle.
TD
TU
7.0
6.0
TS [K]
5.0
4.0
3.0
1.3. Magnetocaloric properties of Gd
2.0
The magnetocaloric effect of a magnetic refrigerant is normally characterized by isothermal entropy change and adiabatic temperature change as a function of temperature and magnetic induction [31]. The cooling power equation requires precise values of adiabatic temperature changes and total heat capacity. Using experimentally validated data, the molecular mean field theory equation of state with the Debye model for lattice heat capacity and free-electron model for electronic heat capacity using measured Curie temperature, magnetic moment, Debye temperature, and other known physical properties cab be used to calculate magnetocaloric properties as a function of T and B required for the cooling power equation above. The calculated data agree very well with experimental results [32]. For example, the adiabatic temperature changes upon magnetization (ΔTU) and upon demagnetization (ΔTD) of Gd from the same temperature are shown in Fig. 1. These differences in adiabatic temperature changes are important for calculation of the cooling power vs. T curve of an AMRR.
1.0
0.0 200
210
220
230
240
250
260
270
280
290
300
310
320
Temperature (K)
Fig. 1. Adiabatic temperature changes (ΔTS) for Gd upon magnetization from 0.6 T to 3.3 T (solid blue line) and upon demagnetization from 3.3 T to 0.6 T (dashed red line). The Curie temperature of Gd is ∼293 K. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Actuator drive motor Linear actuator Lid for high vacuum cold box Heat Shield Superconducting Magnet
2. Propane liquefaction using an active magnetic regenerative refrigerator
Cryocooler 1st Stage
2.1. The AMRR apparatus
Cryocooler 2nd Stage
Propane liquefaction temperatures are within the 285–218 K span of a single-stage AMRR [33] prototype at PNNL. Fig. 2 [34] shows an overview of several subsystems of the apparatus attached from the top lid of the cold box (Dewar). The structural supports to the magnet, the cryocooler cold head, and thermal shields, are shown. The super-insulated, air-filled, thin-wall stainless-steel center tube has a welded bottom flange and is hermetically attached to the top lid. It isolates the reciprocating dual magnetic regenerators and cold heat exchanger (CHEX) subsystem from the ∼40-K thermal shield around the conduction-cooled superconducting magnet subsystem at ∼4 K. The cold box is evacuated to ∼1 × 10−7 torr during experiments. Fig. 3 illustrates the reciprocating dual regenerators and condensing heat exchanger used for these runs.
Warm bore: encloses regenerators, heat transfer subsystems and instrumentation
Fig. 2. Cut-away drawing of major subsystems of the AMRR except the reciprocating dual-regenerator-CHEX, the heat transfer pump, the cold box, and the instrumentation/control/DAQ subsystems.
TCOLD and low magnetic induction after demagnetization in J/kg K, andΔTCD is the adiabatic temperature change upon demagnetization from TCOLD (see Fig. 1). All the variables in this equation can be independently measured exceptFrCOLD . For different regenerators this
Fig. 3. Schematic of coiled-fin tube HEX, the propane storage vessel, the dual Gd regenerators, and CHEX subsystem. The regenerators have opposing axial hot-to-cold temperature profiles and alternatively produce cold helium flow through fins of the condensing HEX when one regenerator is in high field and the other is in low field.
71
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
C3 purge & vent line to fume hood Flexible linkages to ss tubes
PSV
F Regulated Lecture Bottle ~0.793 MPa Propane
LabVIEW DAQ for GEN-I
Bypass flow Needle valve
PSV
F Existing Alicat; 1.379 MPa He
Alicat Flow, T, P; 5 SLPM-C3
1.379 MPa He gas flow from/to HTF pump
TC TC
Return He into HTF pump
Reg 1
TC
CHEX
TC TC TC
Reg 2
Coiled fin tube heat exchanger to liquefy/ collect propane via cold He gas from dual magnetic regenerators in AMR cycle 1.379 MPa He gas flow from/to HTF pump
TC
Fig. 4. A simplified process flow diagram with some piping & instrumentation indicated for the AMRR dual regenerator and CHEX subsystem and external subsystems required for safe handling of propane during runs. ‘TC’ indicates multiple thermocouple locations.
2.2. Simplified piping & instrumentation diagram (P&ID) for propane liquefaction in dual regenerator CHEX
Vent
Inlet
Several simple modifications were made to the existing AMRR (referred to as GEN-I in Fig. 4) to safely execute flammable-gas experiments. A schematic of simplified P&ID of the test apparatus is shown in Fig. 4. The temperatures, pressures, flow rates, and actuator controls were measured starting from near room temperature. 2.3. Configuration of AMRR apparatus for propane experiments Storage Vessel
2.3.1. Feedstock supply assembly A lecture bottle filled with 454 g of 99.5+% propane at 0.793 MPa provided the feedstock for several runs. The lecture bottle included a pressure regulator and a manual flow-control valve hermetically-connected to the cold HEX inside the AMRR via a ∼2-m stainless-steel (ss) tube up to the AMRR, a flexible ∼25-cm ss tube, and a ∼75-cm ss tube, all with 0.635 cm o.d. This hermetic tubing supplies propane at desired pressures into the compact ss coiled-fin tube HEX close-fitted around a small, thin-walled ss cylindrical collecting vessel shown in Fig. 5. An Alicat pressure, temperature, and mass-flow measuring meter were installed in the 0.635-cm o.d., ∼3-m long, ss vent line into a nearby fume hood. The hood had a continuous air suction rate large enough to insure excellent dilution of the small amount of propane gas used in these experiments. This vent line also included a check valve and a pressure relief valve (PSV) set to 0.551 MPa.
HEX outlet
Fig. 5. A picture from the top of the compact coiled-fin tube condensing heat exchanger and liquid propane collection vessel. The dimensions of this assembly are in the text.
selected operating conditions. The heat transfer design calculations assumed fully-developed flow. The transient behavior of the oscillatory helium gas flow during the hot-to-cold blow steps of the AMR cycle in the dual regenerators improves the heat transfer and provides design margin. The coil tube and the fin sizes were sized based on convective heat transfer coefficients of the cold helium heat transfer gas flow through the fins brazed to the exterior of the helical tube and propane flow inside the tube. According to standard design calculations for
2.3.2. Cold heat exchanger assembly to condense, liquefy, and collect liquid propane The heat exchanger and collecting vessel assembly was sized to fit into the ∼13.47-cm long by 6.35-cm i.d. of the CHEX. The compact heat exchanger was designed to cool and liquefy propane as it flows through a coiled-fin tube and into a small collecting vessel. This requires removal of 38 W of heat from propane at a pressure of 0.358 MPa at a nominal rate dictated by cooling power from the MCE of the Gd for 72
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
DAQ cards interfaced to the LabVIEW computer. The DAQ program for the AMRR apparatus recorded propane pressures, temperatures, and flow rates and additional temperatures in the CHEX and elsewhere on the AMRR. The total amount of liquid produced was measured with the flow-integration feature of the Alicat sensor for the two different runs. The large LabVIEW data files from these runs were stored in Excel and later transferred to Origin for processing and plotting prior to thermodynamic analysis.
convective heat transfer in tubular heat exchangers [35], the limiting conductance between the propane gas and the inner diameter of the helical tube is ∼12 W/m2/K during precooling. Once propane begins to condense and create a mixed phase inside the tube, the limiting heattransfer conductance is 255 W/m2/K between the ∼1.379 MPa cold helium flowing past the external fins and outer diameter of the tubes. Based on these calculations and dimensional design specifications, a coiled, finned-tube HEX was purchased from Energy Transfer MDE [36] as shown in Fig. 5. The tube and fins are 304 stainless steel with 0.63cm o.d. tube with 0.32 cm high, regularly-spaced 0.05 cm-thick fins brazed onto the tube surface. The effective axial (z) length of eleven close-fitting loops of heat exchanger tubes is ∼14 cm. Each loop is ∼16 cm-long. The total length of ∼180-cm finned-tube was gradually bent into a helical coil, so the exterior edge of the fins on each tube fit snuggly inside the G-10 shell of the CHEX. The interior fin edges of each tube fit snuggly around the stainless-steel collection vessel. The cold helium heat transfer gas in the AMRR is forced to flow through the fins on the coils, resulting in very effective heat transfer. The helium gas pressure drop through these fins was calculated to be less than 8 Pa. The internal volume of the HEX is 27.04 cm3 and it has a mass of 420 g. The propane collection vessel was constructed of a 7.62-cm long, 0.165-cm wall, 316-stainless steel tube section with butt-welded endcaps. The outer diameter of the welded tube was turned down 0.038 cm to form a tight fit inside of the HEX. Small 0.3125-cm diameter holes were drilled into both endcaps and 0.3125 cm, 316 stainless tubes were TIG-welded into each hole. The bottom tube was welded to the outlet of the HEX while the top tube was welded to the vent line. The completed assembly is shown in Fig. 5. The volume of the central collecting and storage vessel is 94.39 cm3. Total liquid propane volume in the coiledtube HEX and collection vessel is 121.43 cm3. The connecting tubing plus the storage vessel have a mass of 190 g. The HEX was held in place within the CHEX with two machined G-10 endcaps epoxy-bonded to both ends of the CHEX tube. The complete assemble with the dual Gd regenerators is illustrated in Fig. 3.
2.4. Propane liquefaction experimental results The average temperatures in the CHEX during liquefaction were 262.5 K and 246.0 K at equilibrium pressures in the HEX of 0.338 MPa and 0.188 MPa, respectively. The small pressure differences at the lecture bottle and the HEX are due to the feed tubing into the AMRR. In the first run, the lecture-bottle pressure regulator was adjusted to 0.358 MPa. The feed gas temperature of 295.7 K was measured at the Alicat sensor located near the lecture bottle. With the flow-control valve open, no flow was expected. After the He pump, and regenerator drive actuators were started, cooling begins immediately. The temperatures at the top (propane inlet) and bottom of the HEX where it connects to the small storage vessel also began to decrease with no measurable flow of propane. When the CHEX reached ∼262.5 K (see Fig. 7), the propane flow rate increased rapidly as shown in Fig. 6, but after ∼200 s drops by almost a factor of two and liquefaction continued at an approximately constant accumulation rate for ∼1500 s. As soon as initial liquid is collected in the storage tank, the large volume decrease upon condensation creates a cryopump that continuously draws propane through the helium-cooled HEX. Although the feed pressure was set by the regulator at the lecture bottle, the flow rate was determined by the cold helium gas in the CHEX and cryopumping in the HEX. In the first run, accumulated propane flow as a function of time illustrated in Fig. 7 steadily rose from zero until a maximum value when flow completely stops. Fig. 7 also shows TCOLD of He gas in the CHEX was approximately equal to that in the coiled-fin tube HEX and stabilized at 262.5 K for several minutes of liquefaction before the collection vessel temperature slowly decreased. This subcooling continued until TCOLD reached ∼245 K when the feed supply flow-rate dropped to zero and accumulated liquid volume was constant. At that time, the rate of subcooling noticeably increased. After about an hour, the entire CHEX assembly approached steady-state at ∼218 K where the AMRR cooling was balanced by parasitic heat leak. The average feed rate for this run was obtained by dividing the mass of liquefied propane by the total
2.3.3. Instrumentation, control, DAQ, and experimental procedures To monitor temperatures at key locations as indicated in Fig. 4, four bare-tipped, 36-gauge, tightly-twisted, Teflon insulated type-E thermocouples were inserted through the outer wall of the CHEX and sealed with epoxy. The bare tips of two thermocouples were dipped into diluted epoxy and cured to provide a very thin layer of electrical insulation before they were attached with epoxy to the entrance and exit ports of the condensing HEX. The other two bare-tipped thermocouples were mounted in the helium gas flow path near the bottom and top of the liquid collector vessel shown in Fig. 5 (and schematically in Fig. 3). Prior to beginning runs, the entire process gas assembly shown was filled several times with clean, dry nitrogen to ∼0.414 MPa and vented into the fume hood to ∼0.110 MPa before the assembly was filled with propane to 0.414 MPa and vented into the fume hood. This procedure was repeated several times to reduce oxygen concentration within the liquefaction subsystem to below lower flammable limits of any potential mixture. Finally, the subsystem was pressurized to the desired propane feed pressure using the regulator on the lecture bottle with the supply valve left open during the run. The conduction-cooled, superconducting magnet was charged to 3.3 T and put into persistent mode. Temperatures of both external chillers were adjusted to maintain ∼285 K inlet helium gas into the hot end of top or bottom demagnetized dual regenerator for the hot-to-cold blow steps. Before the reciprocating drive actuator was started, helium gas was pumped back and forth for several minutes until a uniform initial temperature of ∼285 K was established at the condensing HEX, the dual regenerators and CHEX subsystems. A LabVIEW-based data acquisition system (DAQ) was used to control motion of actuators, to monitor all sensors, and record all data during AMR cycles. The actuators were controlled via serial communications and sensors were read using National Instruments compact-
Fig. 6. The instantaneous propane flow rate in standard liters per minute as a function of time during the liquefaction run at 0.358 MPa. Note the time scale in this figure is different from that in Fig. 7. 73
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
0.358 MPa propane 40
290
25 260
20
250
15
240
10 5
Top_propane_HEX Bottom_Propane_HEX
230
Total propane (SL)
30
Cooling power (W)
Total propane
270
T (K)
30
35
280
10
-5 1000
2000
20
0
220 0
Gd cool down Propane liquefaction
210
3000
220
230
240
250
260
270
Temperature (K)
Time (s) Fig. 7. Temperatures on the coiled-fin tube heat exchanger and collection vessel and total volume of propane as a function of time for liquefaction in the first run at 0.358 MPa feed pressure. The red dot is where temperature vs. time data were expanded (not shown) over 3–4 cycles for analysis of Gd cooling power from cool-down rate. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 9. Several data points showing the cooling power from the dual Gd regenerators in AMRR apparatus obtained from the rate of cooling of the CHEX and regenerators given by cool-down temperatures vs. time at specific average cold temperatures are shown in this plot of thermal load vs. temperature. The cooling powers for two propane liquefaction runs are added as green stars at the dew point temperatures. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
elapsed time of liquefaction. In a second run, the supply pressure was reduced to 0.200 MPa and everything was prepared in the same manner as in first run. This time the average CHEX cold temperature dropped to ∼246 K before propane started to flow and liquid accumulated in the collection vessel. The plotted temperatures at the inlet and exit of the HEX and the accumulated volume of propane vs. time are shown in Fig. 8. In the second run liquefaction began at an average temperature of 246 K which is the equilibrium temperature of liquid propane at 0.188 MPa. Liquefaction continued at an approximately constant accumulation rate for 2240 s with subcooling of the collected liquid propane as in the first run. After approximately steady-state conditions were reached at the end of the runs, the actuators were stopped. The vent valve that bypassed the PSV was slowly opened enough to reduce the propane-storage vessel pressure to ∼0.110 MPa as the CHEX assembly gradually warmed up toward ∼285 K over about 2 h.
2.5. Analysis of propane liquefaction experiments 2.5.1. Cool-down curve analysis The cooling power of the AMRR was obtained from the rate of cool down (K/s) of the T vs. time curves in Figs. 7 and 8. The thermal masses of the CHEX, the HEX/storage vessel, and colder portion of the dual regenerators were obtained using known properties. In the first experiment, by expanding T vs. time over 3 complete AMR cycles (12 s) at the red dot indicated at 233.5 K in Fig. 7, the cooling power was 33.4 W. Parasitic heat leaks from radiation and convection of oscillating air inside the central tube in Fig. 2 are automatically included in this result. Similar analyses of T vs. time data were repeated at selected points in Fig. 8. In this AMRR with equal time steps in the AMR cycle, cooling only occurs during the hot-to-cold flow steps, i.e., only ½ of the 4-second period. Hence, to obtain continuous cooling values, the cooldown rate results were divided by 2 and plotted in Fig. 9. A linear leastsquare fit was drawn through the red points for illustration.
0.200 MPa propane 40
280
30
T (K)
260
25
250
20
240
15 10
Total propane (SL)
Total propane
270
2.5.2. Propane liquefaction data analysis The Alicat mass flow meter at the lecture bottle was factory-calibrated for propane with flow rate in SLPM and total flow in SL at standard conditions of 298.16 K and 0.1013 MPa. The density of pure propane at these standard conditions is 1.832 kg/m3 which enables conversion of SLPM to kg/s and SL to kg of propane. In the first run, total propane liquefied was ∼36.0 standard for liters and in the second run total propane liquefied was ∼36.5 standard liters. These values are consistent because due to the slight density differences. The 0.358-MPa run yielded 65.9 g of liquid and the 0.200-MPa run produced 68.4 g of liquid. As expected, the ratio of these values is close to the ratio of the densities of liquid propane at 262.5 K and 246 K. This agreement confirmed liquid filled the coils of the HEX and enclosed collection vessel in both runs. The average rate of propane liquefaction from accumulated-mass data in 0.358-MPa run is 4.40 × 10−5 kg/s. Multiplying this value times the measured total liquefaction time and dividing by density of propane at 262.5 K gives total liquid volume of 121 cm3. Likewise, the same calculation for data from the 0.200-MPa run gives an average liquefaction mass flow rate of 2.98 × 10−5 kg/s and a total volume of 121 cm3. Not only do these results perfectly agree with each other, they agree with the combined internal volumes of the HEX and collection
35
230 5 Top_propane_HEX Bottom_Propane_HEX
220 210 0
1000
2 2000 000
0 -5 3000
Time (s) Fig. 8. Temperature on the coiled-fin tube HEX and total propane as a function of time for liquefaction in the second run at 0.200 MPa feed pressure. The red dots were places where the temperature vs. time data were expanded (not shown) over 3–4 cycles for analysis of the cool-down rate. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 74
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
thermal mass of the heat exchanger, the collection vessel, and the liquefied propane much more than by the intermittent flow of cold helium gas in the CHEX. This observation also corroborates that the process gas flow rate through process heat exchangers in cryogenic liquefiers with external process heat exchangers is not controlled by an inlet valve, but instead is controlled by the cooling power at the dew point temperature of the process gas at the pressure at the heat exchanger. This is commonly known in small-scale LNG liquefaction plants, but it is well reinforced herein. In the propane liquefaction experiments, measurement of temperatures in the CHEX and liquefier HEX vs. time provide a good indication of the changing thermal loads from the propane and cool-down of thermal masses in the CHEX and HEX. For instance, as soon as the dew temperature of the propane is reached for a constant propane pressure, the rate of temperature change in the CHEX and HEX is approximately zero as liquid initially collects. This confirms the latent heat from propane liquefaction is much larger than the sensible heat from gradually cooling the thermal masses in the CHEX/HEX saturates the cooling rate of the cold helium gas flowing through the CHEX. The CHEX/HEX cool-down temperature curve changes again once liquid propane accumulates in the collection vessel and again when propane liquefaction ceases. The temperature continues to drop as the sensible heat of all the cold parts of the CHEX/HEX and the liquid propane are cooled to ∼218 K which is the ultimate cold temperature limit of the AMRR at these operating conditions from the parasitic heat leak of ∼6–7 W. This is consistent with calculated parasitic heat leaks from radiation, convection, and conduction mechanisms in this AMRR apparatus. Another observation from Figs. 7 and 8 is that the rate of accumulated propane mass during liquefaction is approximately constant during the entire liquefaction time span. At first glance, this is puzzling given cooling from the dual Gd regenerators occurs during hot-to-cold blows of helium gas from the demagnetized regenerator into the CHEX every other second during an AMR cycle. However, this result shows the thermal capacitance of liquid propane creates a ‘vapor-collapse’ condition commonly seen in Dewars with a liquid cryogen when additional vapor is added. In the present case, the condensation rate is steady, and resultant slight warming of liquid is too small to cause a change in the cryopumping pressure difference that would affect the flow rate. This situation also occurs in the CHEX when its thermal mass is large enough to smooth out temperature fluctuations from intermittent cooling during the AMR cycle when an external thermal load from a continuous heater in the CHEX or from parasitic heat leaks are used to determine the load curve of the AMRR. The agreement of the measured cooling power from propane liquefaction runs and rates of cool-down in Fig. 9 is good. The comparison of the expected cooling power from the equation given earlier in the paper is several times larger than the measured results in Fig. 9. This result shows that even the maximum mass flow rate of the helium heat transfer gas during the hot-to-cold blow (∼4 gm/sec at the hot end of the regenerators) is much less than required to remove all potential cooling of 1.05 kg of demagnetized Gd. This requires a helium gas pressure of 2–3 times higher than the present value of 1.38 MPa in the sealed reciprocating pump at room temperature. Alternatively, using smaller Gd regenerators are a possibility, although intrinsic parasitic heat leaks become more problematic if this is done. The calculated Biot values suggest that temperature gradients inside the ss tubes in the HEX and in the G-10 walls of the CHEX are small enough to be ignored for the interpretation of the results of the propane runs. We did not measure the work rate input, or the hot sink heat rejection rate, so we could not experimentally measure the FOM of this AMRR during the liquefaction experiment. Our detailed performance simulation models of AMRRs predict FOMs of ∼0.6 can be achieved. However, it is also clear that when one includes all auxiliary power of a complete liquefier system, e.g., the helium scroll compressor power of
vessel. This is expected because these two containers are the only process stream elements cooled by the cold helium gas. As soon as they fill up, the cryopumping pressure-difference goes to zero and flow from the lecture bottle stops. The enthalpy difference between liquid propane at 262.5 K and 0.338 MPa and gaseous propane at 295.7 K and 0.358 MPa times 4.40 × 10−5 kg/s for this run gives an average cooling power during liquefaction of 19.6 W. Fig. 7 also shows that during the last 1213 s of this run, the liquid was slowly subcooled from 262.5 K to 246.0 K. The average heat capacity of propane times the liquid mass over the subcooling temperature span at 0.338 MPa divided by sub-cooling elapsed time gives an additional cooling power of 2.1 W for this run. The total AMRR cooling power of 21.7 W in this run was plotted as a green star at 262.5 K in Fig. 9. Similarly, the enthalpy difference at 246.0 K and 0.188 MPa and at 295.2 K and 0.200 MPa times the average mass flow rate 2.97 × 10−5 kg/s for this run gives a cooling power of 14.5 W. Subcooling from 246.0 K to 233.0 K requires 1.1 W of cooling power for a total of 15.6 W. This result is also plotted as a green star at 246 K in Fig. 9. The thermal loads into the CHEX control volume from liquefaction, subcooling, and parasitic heat leaks are continuous during the 4-second AMR cycle. 2.6. Eddy current and Biot number calculations Eddy currents are induced in each closed conducting loop from the magnetic induction changes during the AMR cycle within the ss finned tubes in the cooling and condensing HEX shown in Fig. 3. Each of the coiled fin tubes is in electrical contact with the thin-walled ss collection vessel. Calculating the emf caused by the magnetic flux changes, the resistance of each HEX loop and that of the storage vessel using known dimensions and resistivity properties, gives the total eddy current thermal load into the CHEX of ∼134 mW for 3.3 T to 0.6 T magnetic induction changes in 1 s twice per AMR cycle. This thermal load is negligible compared to the cooling power of the AMRR. The Biot number of the ss coiled-fin tube HEX during liquefaction of propane was calculated at 246 K and 262.5 K. During the design of the HEX, standard heat transfer calculations show the limiting thermal conductance changes from the propane inside the tubes to helium outside the tubes once liquefaction begins. The two-phase condensation propane conductance is therefore higher than that of the helium gas conductance of 255 W/m2 K. Using this conductance value for the propane flow and neglecting the brazed thin fins on the outside of the tubes, with the tube wall-thickness as the characteristic length with known ss thermal conductivity at these temperatures, the Biot numbers are 0.0.12 and 0.0.11 at 246 K and 262.5 K, respectively. The wall thickness of the G-10 in the CHEX, its known thermal conductivity, and an estimated value of thermal conductance for convection flow of helium in the CHEX gave Biot numbers of 0.18 and 0.16 for the G-10 walls in CHEX at 246 K and 262.5 K, respectively. 3. Discussion These experiments confirm our hypothesis that the HEX and storage vessel become a cryopump as soon as they are cooled by cold helium gas from the dual Gd regenerators to the condensation temperature of propane at a selected pressure. The cryopump action draws propane from the lecture bottle until the cooled HEX and collection vessel are filled and the pressure differential due to propane condensation drops to zero, stopping the cryopumping, and no further liquefaction occurs. This is exactly what was observed in both experiments. Based on the relatively steady accumulated mass flow rate observed in both experiments, the flow caused by the cryopump action is continuous (at least compared to the DAQ data rate which is typically 0.1 s or less) and does not follow the intermittent cooling of the dual regenerators for the reciprocating AMRR operating conditions. This result shows that thermal time constant of the condensation of propane is driven by the cold 75
Cryogenics 100 (2019) 69–76
J. Barclay, et al.
the two-stage GM cryocooler that cools the superconducting magnet to ∼4 K and the thermal shield to ∼40 K, the FOM of the present size AMRR will be very small. This observation means that cryogenic active magnetic regenerative liquefiers will have a minimum liquefaction capacity such as 1–10 metric ton/day of cryogens before they can achieve their potential FOMs of 0.5 or higher.
1984;29:581–7. [4] Numazawa T, et al. The helium magnetic refrigerator: II – liquefaction process and efficiency. Adv Cryog Eng 1984;29:589–96. [5] Brown GV. Magnetic heat pumping near room temperature. J Appl Phys 1976;47:3673–80. [6] Steyert Jr. WA. Stirling-cycle rotating magnetic refrigerator and heat pump for use near room temperature. J Appl Phys 1978;49:1216. [7] Barclay JA, Steyert WA Jr. Active magnetic regenerator. U.S. Patent 4,332,135; June 1, 1982. [8] Barclay JA. The theory of an active magnetic regenerative refrigerator. Proc of the 2nd biennial conf on refrigeration for cryogenic sensors and electronic systems. Greenbelt, MD: NASA Goddard Space Flight Center; 1982. [9] Astronautics Corporation of America, Final Report to DOE on Contract No. DEAC02-90CE40895 compiled by D. J. Janda, chief project engineer. Magnetic liquefaction of hydrogen; July 2, 1992. [10] Holladay JD, Teyber RP, Meinhardt KD, Polikarpov E, Thomsen EC, Archipley CC, et al. Investigation of bypass fluid flow in an active magnetic regenerative liquefier. Cryogenics 2018;93. https://doi.org/10.1016/j.cryogenics.2018.05.010. [11] Barclay JA, Brook TC. Apparatus and methods for cooling and liquefying a fluid using magnetic refrigeration. U.S. Patent 6,467,274 B2; October 22, 2002. [12] Oseen-Senda K, Skrzypkowski MP, Barclay JA. Active magnetic regenerative hydrogen liquefier. Contract No. DOE SBIR DE-FG02-07ER84783; SBIR phase I final report; 2008. [13] Rowe A, Tura A. Experimental investigation of a three-material layered active magnetic regenerator. Int J Refrig 2006;29:1286–93. [14] Rowe A. Thermodynamics of active magnetic regenerators: Part I. Cryogenics 2012;52:111–8. [15] Rowe A. Thermodynamics of active magnetic regenerators: Part II. Cryogenics 2012;52:119–28. [16] Smaili A, Chahine R. Thermodynamic investigations on optimum active magnetic regenerators. Cryogenics 1998;38:247–52. [17] Utaki T, et al. Research on a magnetic refrigeration cycle for hydrogen liquefaction. Cryocoolers 2007;14:645. [18] Matsumoto K, et al. Numerical analysis of AMRs for hydrogen magnetic refrigerators between 20 K and 77 K. Cryogenics 2014;51:353. [19] Numazawa T, Kamiya K, Utaki T, Matsumoto K. Magnetic refrigerator for hydrogen liquefaction. Cryogenics 2014;62:185. [20] Kim Y, Park I, Jeong S. Experimental investigation of two-stage active magnetic regenerative refrigerator operating between 77K and 20K. Cryogenics 2013;57:113. [21] Park I, Jeong S. Development of the active magnetic regenerative refrigerator operating between 77 K and 20 K with the conduction cooled high-temperature superconducting magnet. Cryogenics 2017;88:106. [22] Yu B, et al. A review of magnetic refrigerator and heat pump prototypes built before the year 2010. Int’l J Refrigeration 2010;33:1029–60. [23] Zimm C, et al. Description and performance of a near room temperature magnetic refrigerator. Adv Cryog Eng 1998;43:1759–66. [24] Barclay JA, Oseen-Senda K, Ferguson L, Cousins A, Pouresfandiary J, Ralph H, et al. Active magnetic regenerative liquefier for hydrogen. Final tech report: DOE Contract Number DE-FG36-08GO18064; 2014. [25] Holladay J, Meinhardt K, Polikarpov E, Thomsen E, Barclay J, Cui J, et al. MagnetoCaloric hydrogen liquefaction. DOE annual merit review, PD131. Tech report; 2017. [26] Holladay J, Teyber R, Meinhardt K, Polikarpov E, Thomsen E, Archipley C, et al. Investigation of bypass fluid flow in an active magnetic regenerative liquefier. Cryogenics 2018. https://doi.org/10.1016/j.cryogenics.2018.05.010. [in press]. [27] Barclay JA, Oseen-Senda K, Skrzypkowski MP. Unique features of liquefaction of hydrogen and natural gas using magnetic refrigeration. Int’l Inst Refrig; Proc of Thermag VI, Victoria, B.C.; 7–10 September 2014. [28] Teyber R, Meinhardt K, Polikarpov E, Thomsen E, Holladay J, Cui J, et al. Performance investigation of a high-field active magnetic regenerator. Appl Energy 2018. https://doi.org/10.1016/j.apenergy.2018.12.012. [submitted for publication]. [29] See reference 7. [30] See references 23 and 11 for more details of numerical model. [31] Tishin A, Gschneidner Jr. K, Pecharsky V. Magnetocaloric effect and heat capacity in the phase-transition region. Phys Rev B 1999;59:503–11. [32] Dan’kov S, Tishin A, Pecharsky V, Gschneidner Jr. K. Magnetic phase transitions and the magnetothermal properties of gadolinium. Phys Rev 1998;57B:3478–90. [33] See references 23 and 24 for more details of this active magnetic regenerative refrigerator prototype. [34] The original version of this drawing was created at Prometheus Energy Group Inc.; see reference 24 for more details. [35] See reference 27 for example. [36] https://www.finnedtube.com.
4. Conclusions The use of an AMRR to successfully liquefy propane has been demonstrated. The excellent agreement between predicted and observed results confirmed our understanding of the design. Allowance for the intermittent cooling of a reciprocating dual regenerator design was proven important for future reciprocating dual regenerator liquefier designs. The results of this work motivated the conceptualization of a new liquefier design using two sets of identical reciprocating dual-regenerator AMRRs with their drive actuators phased 90 degrees apart. The movement of the quadruple combination of active regenerators coupled with properly controlled heat transfer gas flows using several 3-way valves to direct the cold helium heat transfer gas in counterflow with a process gas can provide its continuous cooling. Such a design for liquid hydrogen would require four multi-material magnetic regenerators and two superconducting magnets for a high-FOM ∼280 K to ∼120 K stage and another four multi-material magnetic regenerators and two superconducting magnets for a high-FOM ∼120 K to ∼20 K stage. This conclusion suggests using a continuously rotating wheeltype AMRR configuration with regenerators attached to the rim may be a simpler alternative if the rotary seals and magnet configuration issues can be resolved. Such development helps validate predictive performance models and identify design issues for advanced designs of AMRR prototypes for liquefaction of colder cryogens such as LNG and LH2. Acknowledgements It is a pleasure to acknowledge financial and programmatic support of the Fuel Cell Technology Office of the Energy Efficiency and Renewable Energy Branch of the Department of Energy. The pro-active support of DOE program managers Dr. Erika Gupta and Dr. Neha Rustagi is deeply appreciated. Dr. John Barclay appreciates the opportunity for EENW to participate in the MCHL project at PNNL. Disclaimer statement The views and opinions of the authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. References [1] van Geuns JR. A study of a new magnetic refrigeration cycle. Doctoral Thesis. Univ. of Leiden; 1966. [2] Hakuraku Y, Ogata H. A rotary magnetic refrigerator for superfluid helium production. J Appl Phys 1986;60(9):3266–8. https://doi.org/10.1063/1.337716. [3] Nakagome H, et al. The helium magnetic refrigerator: I – Design. Adv Cryog Eng
76