Two-Phase Pumped Cooling System of a HVDC Power Converter: System Analyses and Experimental Results ^†

Abstract

The number of electronic components requiring dedicated cooling is set to increase for the next generation of medium-to-large-sized more-electric aircraft. In the EU-funded H2020 ADENEAS project, the usage of a two-phase mechanically pumped loop for the cooling of a high-power HVDC (High-Voltage Direct Current) converter, in an aircraft case, is analysed. After a fluid selection, R1233zdE appears as one of the premier candidates for aircraft two-phase pumped cooling systems. A comparison between a R1233zdE two-phase pumped cooling system and liquid cooling system with propylene–glycol water as working fluid is performed. The result of the system comparison is a 40% system mass reduction in favour of a two-phase pumped cooling system. By testing, a successful demonstration of the discussed cooling system is shown in nominal and non-nominal scenarios. With the achieved results presented here, a leap to TRL4 of a two-phase mechanically pumped loop for aircraft power electronics cooling is shown.

1. Introduction

The introduction of the more-electric aircraft (MEA) platform will cause a significant increase in the amount of electronic power being consumed by aircraft systems, which in turn will bring about a significant increase in the heat being dissipated by those same systems. The power dissipation from electrical components will be orders of magnitude higher than for conventional aircraft, even if high levels of efficiency are achieved.

In the ADENEAS project, the goal is to reduce mass for distributed MEA architectures. A key optimisation can be found when reconsidering technologies used in existing cooling systems. Here, we introduce a two-phase pumped cooling system as an option for cooling aircraft power electronics. Potential benefits of a two-phase cooling system compared to conventional cooling systems include (1) enabling high-heat-flux cooling of on-board power electronics; (2) transporting heat to ambient air instead of bleed air; and (3) more modular cooling system designs. Two-phase cooling technology was earlier shown to be applicable for two-phase cooling of the AMS02 particle detector on the ISS [1], two-phase cooling of high-heat-flux power electronics [2], and two-phase cooling of fuel cells for aerospace application [3].

In this paper we analyse working fluids for a two-phase cooling system; the system masses of a two-phase and liquid mechanically pumped loop; system analyses of a two-phase pumped loop; and a two-phase cold plate design for high-heat-flux power electronics cooling. A successful technology demonstration for nominal and non-nominal scenarios is shown by testing.

2. Fluid Selection, System Analyses, and System Mass Comparison

2.1. System Description

For the heat transport from power electronics to ambient air, an evaporator, a two-phase mechanically pumped loop (two-phase MPL), and a fluid–air heat exchanger are used, as schematically shown in Figure 1. In the evaporator, the latent heat of the refrigerant is used for absorbing the dissipated heat from the power electronics. The evaporator design will be discussed in Section 3.2. The two-phase MPL transfers heat using a pump in the liquid part of the loop. Evaporation results in a vapour–liquid mixture in the line from the evaporator to the fluid–air heat exchanger. Using mass flow control, the vapour mass fraction in this two-phase line is 0.7 maximum. For a vapour mass fraction of 0.7, the two-phase flow regime is annular [4], which ensures a high heat transfer coefficient at the channel wall in the evaporator. In the fluid–air heat exchanger, the absorbed heat is emitted to the airflow by condensing the two-phase fluid in minichannels. Note that the air heat exchanger sizing has a direct effect on the experienced air drag by the aircraft.

2.2. Fluid Selection

The boiling temperature in the two-phase MPL is controlled by the accumulator pressure. Therefore, the p-T saturation characteristic of fluids is taken into account in the fluid selection process. Ultimately, the fluid selection has a strong effect on system mass. In the fluid selection, various aspects of potential fluids have been analysed, concerning physical fluid properties, system characteristics (i.e., mass flow, tubing diameter), and health, safety, and environmental (HS&E) aspects. The considered fluids are sourced from NIST’s REFPROP database [5]. The physical fluid properties can be combined into a single Figure of Merit for each fluid based on the methodology described in [6].

Table 1. Relevant properties of potential working fluids at 70 °C for two-phase MPL. In blue, the cooling system analysis results are shown. In green, the HS&E aspects of the fluids are shown.

	NH3	R134a	R1234zeE	R1234yf	R236fa	R245fa	R1233zdE
Figure of Merit	1.2 × 1013	1.3 × 1012	1.0 × 1012	9.2 × 1011	7.3 × 1011	7.0 × 1011	5.9 × 1011
Pressure at 70 °C (bar)	33.0	21.1	16.1	20.4	9.8	6.1	5.1
Pressure at 20 °C (bar)	8.5	5.7	4.3	5.9	2.3	1.2	1.1
Triple point (°C)	−78	−103	−104	−150	−93	−103	−78
Enthalpy hlv (kJ/kg)	940	125	124	97	115	162	165
Spec. heat cp (kJ/kgK)	5.47	1.80	1.65	1.83	1.39	1.44	1.31
Liq. density ρl (kg/m3)	527	997	986	884	1185	1205	1145
Liq./vap. density ratio	19.9	8.6	10.8	6.7	17.4	35.9	42.4
∂Tsat/∂psat (°C/bar)	1.3	2.1	2.7	2.3	4.1	6.1	7.5
Req. massflow (g/s)	9.1	68.8	69.1	88.7	74.6	52.9	52.0
Req. vol. flow (lpm)	1.0	4.1	4.2	6.0	3.8	2.6	2.7
Tube diameter (mm)	7.7	12.5	13.0	13.4	14.2	14.3	14.7
NFPA (flam-tox-stab)	3-1-0	1-0-1	2-1-0	2-2-0	1-0-0	2-1-0	2-0-0
Global Warming Pot.	0	1530	1.4	0.5	8690	962	3.9
Future availability	Good	Phase-out	Good	Prob. good	Phase-out	Phase-out	Good

shows the results from the pre-selection of fluids. Considering a maximum NFPA rating of two or lower [7], a low global warming potential (GWP), and good future availability, R1234zeE, R1234yf, and R1233zdE appear as primary choices. These HFOs have similar cooling performance. However, R1233zdE has a significantly lower pressure at the desired operating saturation temperature of 70 °C. System analyses were conducted with R1233zdE as the working fluid.

2.3. Two-Phase Mechanically Pumped Loop System Analyses

In Figure 2 and Figure 3, the conducted system analyses on the two-phase MPL with heat source and heat sink are shown, with R1233zdE as the working fluid. A saturation temperature of 70 °C is considered, resulting in a controlled system pressure of 5.1 bar(a). Three parallel heat sources are shown, each 2 kW. These three parallel heat sources each reflect the combined dissipated heat of two heat sources, with one on the topside and one on the bottom side of the cold plate section in a sandwich configuration. Three sections with 2 kW cooling performance result in a 6 kW total heat dissipation capacity. A flow of 52 g/s ensures a vapour mass fraction of a maximum of 0.7 at the cold plate outlet. In Figure 3, the vapour mass fraction is shown, with a local maximum lower than 0.7 due to the 15 °C of subcooling at the cold plate inlet.

The simulation results show the steady-state performance prediction of the system at the maximum ambient temperature of 55 °C. A heat transfer of 6 kW is established. In the condenser, the fluid transfers its heat to the aircraft’s surrounding air, which results in subcooled R1233zdE at the outlet, required to ensure no cavitation in the pump.

2.4. System Mass Comparison

For the system mass comparison between liquid and two-phase mechanically pumped loops, a range of heat dissipation was considered, as shown in Figure 4. For the two-phase MPL, two working fluids were considered, namely R1233zdE and ammonia. As shown in Table 1, ammonia (NH₃) has superior performance but is toxic. Comparing mass on logarithmic scale, it can be seen that the system mass can be reduced by 40% when opting for a two-phase MPL with R1233zdE compared to a liquid MPL with propylene–glycol water. This mass reduction is 70% when considering ammonia as the two-phase MPL working fluid.

Additional to system mass, the following differences appear when comparing a liquid and two-phase mechanically pumped loop:

• The required mass flow for a two-phase MPL is an order of magnitude smaller, resulting in a lower electrical power consumption of the pump;
• Refrigerants suitable for a two-phase MPL generally have a low triple point temperature. For cooling system operation in an aircraft, a triple point below the minimum ambient temperature in flight is required, which is −55 °C;
• The air drag penalty induced by the fluid–air heat exchanger is lower for a two-phase MPL because of a higher refrigerant temperature inlet.

3. Experimental Setup and Cold Plates Under Test

3.1. System Demonstrator

A two-phase MPL demonstrator was built in order to validate the two-phase MPL system performance and individual cold plate performance. The demonstrator system is shown in Figure 5 and Figure 6. In this setup, cold plate designs can be tested by drop-in and several modularity configurations can be replicated. The system has shown its ability to transfer up to 10 kW of heat from heat source to sink, with R1233zdE as the working fluid.

3.2. Cold Plate Design and Manufacturing

The individual items under test were two manufactured cold plates. One of them was manufactured by employing AlSi10Mg additive manufacturing with in-house machining capabilities at NLR, as shown in Figure 7 and Figure 8. The other cold plate was manufactured by conventionally machining aluminium, as shown in Figure 9 and Figure 10. The orange-coloured blocks are copper blocks that serve as thermal emulators of HVDC converters. Embedded cartridge heaters in the copper blocks generate heat in a controlled manner. A heat flux of 7.7 W/cm² is applied on each cold plate cooling section of 10 by 13 cm on both sides. At 1.0 mm above the interface between copper blocks and the cold plate, thermocouples are placed to monitor the local temperature.

Based on design trade-off and flow simulations, the machined cold plate (Figure 9) consists of three parallel sections for three pairs of heat sources. Note that a pair consists of a heat source on the top and bottom of the cold plate in a sandwich structure. The flow enters the cold plate at the liquid inlet, bottom right in Figure 10, and leaves at the two-phase outlet, top right. Due to higher pressure drop for two-phase flow compared to liquid flow, the diameter of the two-phase channel is 50% larger than for the liquid channel.

Flow balancing between the parallel sections is achieved by placing restrictions upstream of each section. These serve as the main pressure drop for each section, i.e., the restriction pressure drop is five times larger than the pressure drop over a two-phase minichannel by design. This appeared effective for flow balancing across a cold plate [2]. Furthermore, acceleration effects are negligible by designing the minichannels such that inertial forces are stronger than acceleration forces.

For the detailed design of both cold plates, the internal geometry dimensions were the same. For the additively manufactured cold plate, however, a reduction to 2/3 of the size of the machined cold plate was applied due to the manufacturing envelope. Therefore, its heat absorption capacity was reduced to 4 kW. The additively manufactured cold plate was manufactured as a single piece, employing its inherent manufacturing benefits. The machined cold plate was assembled as a two-piece design, with a braze line along the midplane of the design.

4. Technology Demonstration by Testing

4.1. Performance Test of Additively Manufactured Cold Plate at Maximum Heat Load

The additively manufactured cold plate was tested with a heat load of up to 4 kW. A mass flow of 35 g/s was established to achieve a vapour mass fraction at the outlet of 0.70. In Figure 11, the measured temperatures in the two-phase MPL demonstrator system are displayed. In Figure 12, the vapour mass fraction in the two-phase MPL demonstrator is displayed. The saturation temperature was kept constant at 70 °C by pressure control. The condenser outlet is at 10 to 15 °C due to the cooling capacity of the building’s general water circuit. In the demonstrator system, a recuperator was used to heat the cold flow upstream of the cold plate by interfacing with the hot two-phase flow downstream of the cold plate via a plate heat exchanger. This results in a high evaporator inlet temperature of about 65 °C. This low amount of subcooling results in two-phase cooling performance near the entrance of the heated section of the cold plate. The evaporator outlet is at saturation temperature, which is 70 °C.

In Figure 13, the temperature at the cold plate interface is shown. When the preheater (between pump and cold plate) and pump are switched on, an initial temperature rise is observed. A further increase in temperature occurs from t = 0.16 h, when the heat loads on the cold plate sections are switched on to 2 kW, i.e., 500 W per heat source. Initially, this results in single-phase heating. Once above saturation, a superheat peak is observed. This phenomenon is, in general, ill-understood. A lack of nucleation sites can be a reasonable explanation; however, for the additively manufactured cold plate, the suppression of superheat would be expected due to the relatively high wall roughness at the minichannel perimeter. The superheat peak is sustained for a minute. Once superheat is suppressed, the temperature sensors show a uniform temperature distribution across the cold plate interface. For 3 kW, similar performance is found, albeit at a higher temperature due to a higher heat flux.

At 4 kW operation, a spread in temperatures is shown in Figure 13. This indicates sub-optimal flow balancing. Because of the additive manufacturing process, small agglomerations of AlSi10Mg particles have been formed, which affect flow distribution in the plenums. Therefore, higher temperatures occur at the temperature sensors furthest away from the inlet. Even though imbalance occurs, the operation at 4 kW is stable, indicating that two-phase cooling for this cold plate has been achieved for all heat dissipation levels.

4.2. Performance Test of Machined Cold Plate at Maximum Heat Load

Performance testing on the machined cold plate has been performed up to 6 kW. Again, the heat distribution among the six heat sources is even, i.e., max 1 kW per heat source.

The pressure is controlled at 5.1 bar (a), i.e., at a saturation temperature of 70 °C. This pressure control is performed on the outlet of the cold plate, as this is the most stable control method. The pressure differential over the cold plate stays within 0.2 bar along the test duration.

The vapour mass fraction of the cold plate outlet is shown in Figure 14. Due to the thermal inertia of the system, the vapour mass fraction is increasing during the test phase at 2 kW between t = 0.25 h to t = 0.50 h. The recuperator is causing this inertia. At 4 kW and 6 kW heat loads, constant vapour mass fractions of x = 0.45 and x = 0.69 are observed.

The measured temperatures at the cold plate interface are shown in Figure 15. Remarkably, the superheat peak is small. This might be suppressed by the presence of the midline braze, which can act as a primary nucleation site. The temperature variation between sensors is low. Unlike the additively manufactured cold plate, the machined cold plate shows excellent flow balancing properties, confirming the working of the applied pressure balancing method. Furthermore, the highest temperature observed is only 80 °C during 6 kW operation, indicating a high heat transfer coefficient at the fluid–wall interface.

4.3. Performance Test of Machined Cold Plate with Unbalanced Heat Load

As part of the test campaign on the system’s modularity, a test was carried out with unbalanced heat load. Here, the machined cold plate was used for testing. An alternating sequence of the heat loads on the six heat sources was applied. The following sequence was applied: 1-1-1, 1-1-0, 1-0-1, 0-1-1, 1-1-1, 1-0-0, 0-1-0, 0-0-1, 1-1-1, where “1” indicates full load and “0” indicates zero load on the section. As highlighted in Section 3, a section consists of a sandwich configuration of two heat loads. Hence, a section load of 2 kW total is determined.

In Figure 16 and Figure 17, the system temperatures and vapour mass fraction show performance as expected for the applied heat load profile. The vapour mass fraction shows a spike at t = 1.17 h, following incorrect shutdown of the system by human mistake.

The results of the cold plate temperature in Figure 18 show expected behaviour, in line with the cold plate performance shown in Section 4.2. A maximum temperature of 81 °C is shown. Even though heat load imbalance is introduced, no signs of flow imbalance are noted. This is a result of the pressure balancing method by restrictions, which mitigate the difference between liquid and two-phase pressure drop in the cold plate minichannels.

5. Conclusions

In the presented study, a successful technology demonstration of a two-phase mechanically pumped loop for power electronics cooling in more-electric aircraft is reported. Successful cold plate and system performance have been shown in both nominal and non-nominal operation. Furthermore, the choice for R1233zdE as a high-performance refrigerant with little HS&E disadvantages has been substantiated. The two-phase MPL, with R1233zdE as a working fluid, has a 40% lower system mass than a liquid MPL. This can be enhanced to 70% reduction when ammonia is considered as a working fluid.

With the achieved results presented here, the two-phase mechanically pumped loop for aircraft power electronics cooling has progressed to TRL4. This paper shows potential cooling system mass saving. Therefore, incorporation of two-phase cooling technology for power electronics cooling should be considered in next-generation aircraft system design.