Development of a Two-Phase Heat Strap for CubeSat Applications

Transcription

1 46th July 2016, Vienna, Austria ICES Development of a Two-Phase Heat Strap for CubeSat Applications Steven A. Isaacs 1, Diego Arias 2, Mike Hulse 3 i2c Solutions, LLC, Longmont, CO, Derek Hengeveld 4 LoadPath, Albuquerque, NM, and Peter Hamlington 5 University of Colorado, Boulder, CO, Due to the compact and modular nature of CubeSat systems, thermal management is becoming a major bottleneck in system design and performance. In this study, we present the development of a flat, lightweight and efficient two-phase heat strap called FlexCool, currently being developed at i2c Solutions. Using acetone as the working fluid, effective thermal conductivities of up to 2,150 W/m-K were demonstrated experimentally, approximately four times greater than pure copper, using a heat strap that was only 0.86 mm thick. The heat strap was also shown to withstand internal vapor pressures as high as 930 kpa, which demonstrates the ability to withstand the pressure difference experienced while operating in space. A closed-form computer model was developed to predict the thermal resistance and maximum heat load achieved by the heat strap, and it was validated with experimental measurements. A d F g K h lv k eff k eff,w k f k w L eff M M N n P l P v Nomenclature = cross-sectional area = wire diameter = predicted value = gravitational constant = permeability = enthalpy of vaporization = effective thermal conductivity of the thermal strap = effective thermal conductivity of the wick layer = thermal conductivity of working fluid = thermal conductivity of wick material = effective length of heat strap = measured value = molecular weight = mesh number = number of measurements per run = pressure of liquid = pressure of vapor 1 Sr. Thermal Engineer, 500 S. Arthur Avenue, Suite Thermal Program Manager, 500 S. Arthur Avenue, Suite Principal Engineer, 500 S. Arthur Avenue, Suite Senior Engineer, 933 San Mateo Boulevard NE, Suite Assistant Professor, Mechanical Engineering Department, 427 UCB.

2 q = heat load R = thermal resistance R = universal gas constant r eff = effective pore radius of wick S = crimping factor T v = temperature of vapor w = wire spacing (ΔP c) m = maximum capillary pressure gradient ΔP ph,c = pressure gradient across liquid-vapor interface at condenser ΔP ph,e = pressure gradient across liquid-vapor interface at evaporator ΔP w1 = pressure gradient across single layer of wick 1 ΔP w2 = pressure gradient across single layer of wick 2 ΔP II = axial hydrostatic pressure drop ΔP + = normal hydrostatic pressure drop ε = porosity μ = dynamic viscosity of working fluid υ lv = change in specific volume φ = tilt angle ρ = density σ = surface tension of working fluid ˆ = accommodation coefficient θ = contact angle between liquid and wick material I. Introduction HE number of CubeSat launches have continued to increase since their introduction in As of 2013, a total T of 129 missions had been launched by nearly 80 organizations, including universities, national space and defense agencies, private companies and amateur organizations. These missions have had a range of objectives including testing new technology, collecting scientific data, communications, and education. However, there are many challenges limiting the development and success rate of CubeSat missions. One major challenge in the design of CubeSats and small satellites is heat dissipation and thermal management. Just like modern commercial and military electronic devices, CubeSats can have high power densities due to their small compact form factor (10 cm x 10 cm x 10 cm for a 1U configuration). In addition, orbital environments introduce very complex inputs and constraints into the thermal design of a CubeSat. Currently, heat dissipation from onboard electronics is limited to conduction through the printed circuit board (PCB) and chassis, in order to radiate to space from an external surface. Depending on the type of orbital environment, CubeSats can be exposed to very dynamic thermal environments. The thermal management system must be able to efficiently transport the required amount of heat through a small form factor and still maintain component operating temperatures at adequate levels while accounting for the low external pressure and dynamic environment. One popular passive cooling technology that provides a small form factor is a flat heat pipe. Also referred to as vapor chambers, these devices operate on the same principles as a conventional heat pipe but are fabricated in a flat profile. There are currently several manufacturing and research efforts to design these heat pipes to be highly conductive, flexible and interface well with commercial and military electronics 2. However, limited research has focused on CubeSat applications and meeting the unique challenges presented by their orbital operating environments. In this paper, we introduce an efficient, flat, lightweight, and conformable two-phase thermal strap specifically designed for space applications. The physical design and fabrication are discussed along with experimental data. Additionally, a closed-form model that accurately predicts the thermal resistance and maximum heat transport capability is discussed. II. Thermal Analysis of a CubeSat In order to illustrate the current thermal management challenges faced by CubeSats, and to investigate the benefit that a lightweight and efficient two-phase thermal strap provides to small satellites and CubeSats, a couple of systemlevel thermal models were developed. Thermal Desktop 3 was used for modeling a simplified 3U CubeSat shown in Figure 1, including the following key components: standard 3U bus (~10 cm x 10 cm x 30 cm), modular rail structure, deployed solar arrays, body-mounted radiators, one standard (~10 cm x 10 cm) internal rail-mounted printed circuit board (PCB), and a single PCB-mounted component (2 cm x 2 cm). Thermal conductors were used to represent the 2

3 poor thermal path offered by traditional standoffs used to connect the PCB and body-mounted radiators to the rails in the CubeSat frame structure. A thermal conductor between the component and the body-mounted radiators was included in the model, representing the additional thermal path created by a conceptual thermal strap (e.g., FlexCool heat strap or plain copper strap for comparison). Although this was a simplified model, it represented the general thermal environment of the PCB-mounted electronic component, and was used to perform parametric analyses of the effect of orbital characteristics, component power and thermal path to the body-mounted radiators. Figure 1. 3U CubeSat model in Thermal Desktop. Hot and cold cases of low Earth orbits (LEO) developed from statistical analysis for robust thermal design 4, as well as a geostationary orbit (GEO), were used for bounding the thermal environments of the electronic components inside the CubeSat. Table 1 lists the main characteristics of the orbits, while Figure 2 shows the hot and cold LEO cases: the hot LEO case represented a polar orbit with negligible eclipses and in which the CubeSat was almost perpendicular to the Sun, while the cold LEO case represented an orbit with eclipse times of approximately half of the orbital period. Figure 2. Thermal Desktop representations of hot and cold LEO cases. Table 1. Orbital Characteristics Hot LEO Cold LEO GEO Orbit angle ( ) Altitude (km) 350 1,000 35,786 Albedo Earth temperature (K) Space temperature (K) Orbit period (h)

4 Two versions of the model were used to analyze the temperatures of the CubeSat components: first, using a steady state model with orbital-average boundary conditions, and second, using a transient model with instantaneous orbital conditions applied during two sequential orbits. The predicted orbital-average temperatures of the electronic component as function of its power dissipation are shown in Figure 3: red and blue lines represent the hot and cold LEO cases, respectively. The solid lines correspond to the component temperature without any heat strap (i.e., using the traditional poor thermal path through the PCB and stand offs), while the broken lines represent the component temperatures if a thermal strap with a conductance of 1.6 W/K is used to connect the electronic component to the radiators. As expected, the component temperatures are colder for the cold LEO case, although the difference with the hot LEO case is relatively small. However, having a conductive path through a heat strap results in much lower component temperatures. While these figures show orbital-averaged temperatures, transient cases were run to estimate the temperature swing in both orbital cases: Figure 4 shows the component temperature over two orbits, for a power dissipation of 10 watts and a thermal strap with a conductance of 0.68 W/K, and for both hot and cold LEO orbits. The electric component experiences temperature swings of approximately 5 and 15 C in the hot and cold LEO orbits, respectively, which can be explained by the fact that the satellite sees solar radiation for most of the hot LEO orbit, while it experiences an eclipse during almost half of the cold LEO case. Figure 3. Orbital-averaged component temperatures as function of power dissipation predicted, for hot and cold LEO cases, and with and without thermal straps. The Thermal Desktop model was used to validate a simplified thermal model of a 1-U CubeSat with one PCB, and with similar thermal connectors to the body structure and body-mounted radiators. The thermal model of the 1U CubeSat, developed in the Engineering Equation Solver (EES), 5 was then used to perform parametric analyses of the effect of different design factors on the temperature of the PCB-mounted component. Figure 5 shows the component temperature as function of component power and thermal strap conductance, for the hot LEO case. The temperatures are bounded by two cases: on one extreme, a traditional case in which no thermal strap is used, and, on the other extreme, a case in which an infinitely-large conductance is used to connect the component to the body-mounted radiator. The no-thermal-strap case predicts that components that dissipate 5 watts or more reach temperatures greater than 320K in the hot LEO environment. On the other hand, the maximum component power dissipation that an infinitely-large thermal strap could provide before reaching 320K is approximately 18 watts, at a point where a larger radiator would be needed. Between these two bounding cases, it is apparent that the use of a thermal strap reduces the component temperature by providing an efficient thermal path to the body-mounted radiators. As described in the following sections, the thermal strap fabricated and testing in this paper demonstrated an average conductance of 3.2 W/K. 4

5 Figure 4. Instantaneous component temperatures for hot and cold LEO cases (power dissipation of 10 watts, and thermal strap of 0.68 W/K) 350 Component temperature, K K No thermal strap Infinitely-large conductance Component power, W Strap Conductance [W/K] Figure 5. Orbital-averaged component temperature as function of power dissipation and strap conductance. While the chart in Figure 5 was developed for a hot LEO case; similar design charts may be created for other environments, such as a cold LEO or a GEO case. Individual charts depend on the particular orbital conditions, as well as the CubeSat design (geometry and optical properties of body-mounted radiators). In order to analyze the benefit that a two-phase thermal strap provides, the locus of component power and strap conductance values that yield component temperatures equal to 320K was plotted in Figure 6. This line represents the minimum strap thermal conductance needed to keep the component temperature below 320K for a given component power, in the hot LEO case. The right axis in Figure 6 compares the theoretical mass of copper and two

6 phase thermal straps required to achieve a given conductance. Based on a detailed model of the FlexCool two-phase thermal strap presented in section III, a FlexCool with 3-times the thermal conductivity of copper was used in this analysis. Consequently, for a given component power dissipation, the FlexCool can provide the same thermal conductance of a copper strap with only one-third of its cross sectional area and, therefore, a third of its volume. As the density of the FlexCool is one-third of copper, the FlexCool can achieve the same thermal conductance of a copper strap with one-ninth of the weight. 20 For component temperature of 320K Copper 500 Component power, W FlexCool (3X k copper ) Strap mass, g Strap conductance, W/K Figure 6. Minimum thermal strap conductance and width for component temperatures of 320K. The following sections describe the theoretical models that have been created for designing the FlexCool thermal strap, the successful manufacturing of FlexCool for space applications, and the experimental validation of the theoretical models. III. Detailed Thermal Model of Two-Phase Heat Strap Two separate models were developed in order to predict the maximum heat transport capability and minimum thermal resistance of the FlexCool two phase thermal strap. First, a hydrodynamic model was used to predict the maximum heat transport based on the capillary limit, and, second, a thermal resistance network model was used to predict the minimum thermal resistance based on heat transport through the various layers. Both models were developed in Matlab 6. The following section explains the general assumptions and equations used in both models. A. Hydrodynamic Model For heat pipes that operate under moderate heat fluxes and operating temperatures, the maximum heat transport is typically governed by a balance between the maximum capillary pressure induced by the wicking structure and the pressure losses introduced by the same wicking structure, the vapor channel, the phase change and the body forces. This constraint is called the capillary limit and is defined as 7 Pv Pl ( P c ) m Pph, e Pph, c P PII (1) x x Leff L eff where ΔP is the pressure difference across the length of the heat strap and P x is the integral of the pressure gradient along the main axis of the heat pipe. For the heat strap to operate, the maximum capillary pressure induced by the wicking structure, (ΔP c) m, must be greater than or equal to the summation of the pressure losses across the entire heat strap. The terms on the right hand side of Equation 1 represent the vapor pressure loss, liquid pressure loss, pressure loss due to phase change at the evaporator, pressure loss due to phase change at the condenser, normal hydrostatic pressure loss, and axial hydrostatic pressure loss, respectively. For moderate operating conditions in which vaporization and condensation are relatively mild, pressure losses due to phase change can be neglected. Additionally, this study considers a flat heat pipe so that the normal hydrostatic pressure losses can be neglected. Figure 7 illustrates 6

7 the simplified model used in this study where the subscripts w1 and w2 refer to wick 1 and wick 2, respectively, located on opposite sides of the vapor core of the heat pipe. Figure 7. Hydrodynamic model diagram. The Young-Laplace equation was used to model the capillary pressure across the fine wick: ( P c ) m = 2σ cos θ r eff. (2) As the FlexCool heat strap uses commercial woven meshes with wire diameter d and wire spacing w, the effective pore radius, r eff, was estimated as 8 r eff = (d + w) 2. (3) Assuming steady state operation, all the liquid at the evaporator is evaporated and all the vapor at the condenser is condensed, and the pressure gradient across the woven mesh wick can be expressed as a function of the applied heat load, q: P = (μl eff q) (KρAh lv ). (4) This pressure gradient represents the pressure loss through the liquid and vapor along the length of the thermal strap, where and are the dynamic viscosity and density of the continuous phase, h lv is the working fluid latent heat of vaporization and A is the cross sectional area of woven mesh. For this study, the effective length, L eff, is the distance between thermocouples T 0 and T 4 (see Figure 11). The above expression is a form of the porous media equation. To calculate the permeability, K, of the mesh, the following equation was used 8 : K = (d 2 ε 3 ) (122(1 ε) 2 ), (5) where the mesh porosity,, is defined as ε = 1 πsnd 4. (6) For a woven mesh, S = 1.05, and N is the mesh number. For the fine mesh, the pressure gradient terms were calculated using liquid properties. For the coarse mesh, pressure gradient terms were calculated using vapor properties. The axial hydrostatic pressure gradient was calculated as P II = ρ l gl eff sinφ, (7) where ρ l is the density of liquid, g is the gravitational constant, and is the tilt angle. In zero-gravity applications, this pressure gradient reduces to zero. The model was written in Matlab, and the thermophysical properties of the working fluid were calculated using CoolProp. 9 Using the operating temperature as an input, the model calculated the maximum heat transport based on the capillary limit. B. Thermal Resistance Network Model A thermal resistance network model, as illustrated in Figure 8, was developed with the objective of predicting the effective thermal resistance of the heat strap. The model represented the conductive heat transfer across the casing layers and the fine meshes that act as wicks, and it assumed negligible thermal resistance through the vapor core. The nomenclature used in the thermal resistance network model is listed in Table 2. 7

8 Figure 8. Thermal resistance network diagram. Table 2. Subscript descriptions for the thermal resistance network model. 1 st subscript Description 2 nd subscript Description a adiabatic b case bottom c condenser p phase change e evaporator s case side t case top w1 wick 1 w2 wick 2 Thermal resistance values of the copper case and meshes were based on a thermal conductance of copper of 400 W/m-K. Assuming that the wicking mesh was completely saturated with liquid working fluid, the effective thermal conductivity of each liquid wick, k eff,w, was calculated using the expression 7 : k eff,w = k f [k f + k w (1 ε)(k f k w )] [k f + k w + (1 ε)(k f k w )], (8) where the porosity was determined using Equation 6. The effective thermal conductivity of the vapor core mesh was defined using the above equation using thermodynamic properties of vapor. To evaluate the thermal resistance due to phase change at the liquid/vapor interface, an expression for heat transfer coefficient based on kinetic theory at an evaporating/condensing interface was used 8 : 2 2 ˆ h lv M P v lv hi 1. (9) 2 ˆ Tv lv 2 RT v 2hlv Using the above expression with the areas of the evaporator and condenser, a thermal resistance due to phase change at the liquid-vapor interface was calculated. These thermal resistances were combined in order to determine an effective thermal resistance of the overall FlexCool heat strap, k eff, which is plotted in Figure 14. IV. Heat Pipe Fabrication and Experimental Testing Though thermal straps can be fabricated using a wide range of materials, copper was chosen in this study due its high thermal conductivity, malleability, and compatibility with acetone and water. A prototype FlexCool thermal strap, shown in Figure 9, was built, having 12.7-cm-long sides and a total thickness of 0.86 mm. The FlexCool thermal strap is based on an assembly of copper foils (casing layers), fine copper woven meshes (wicks), and coarse copper woven meshes (vapor core). Although the type and number of layers can vary depending on desired performance, the prototype developed for this study is composed of two case layers, two fine mesh layers and one coarse mesh layer, as can be seen in Figure 10. Using thermo-compression bonding, also known as diffusion bonding, all layers were bonded together resulting in strong through-thickness mechanical strength and, therefore, a structure that can withstand the internal vapor pressure while operating in space vacuum. 1/ 2 8

9 Figure 9. a) General dimensions of thin flat heat strap, and b) measured total thickness. Figure 10. a) Diagram and b) picture of heat strap cross-section. The prototype FlexCool heat strap developed for this study was filled with 30% of acetone, which is equal to 70%-fill, based on the void fraction of the two layers of fine mesh. A one-dimensional thermal test bench, illustrated in Figure 11, was used to measure the maximum heat transport and thermal resistance of the heat strap. A resistance wire heater was placed at the evaporator end of the sample, and a 30V/20A DC power supply was used to provide the heat input. The evaporator was enclosed by 2.54-cm-thick fiberglass insulation. The condenser (6.4 cm x 12.7 cm) was exposed to ambient conditions and was cooled by natural convection. The adiabatic section measured 1.2 cm x 12.7 cm. A LabView-based data acquisition system was used to collect temperature data from five K-type thermocouples located along the heat strap. At the beginning of each run, the sample was oriented horizontally (i.e., =0 ). A low heat input of 4.83 watts was applied to the evaporator. Once the temperature measurements reached a steady state, the sample was tilted by 10. Steady state operation was considered to be achieved when the fluctuations in temperature measurements changed by less than 2 C. This process was continued at 10 intervals until the temperature at the evaporator, T 0, was greater than the temperature T 1. This jump in the evaporator temperature resulted from an increase in the local thermal resistance due to dryout of the working fluid, and indicated the point at which the device reached the capillary limit and the maximum heat transport. At each heat load the total thermal resistance was calculated as R = (T 0 T 4 )/q. (10) The effective thermal conductivity was cacluated as k eff = L/(RA), (11) where the area, A, is the total cross-sectional area of the thermal strap. 9

10 Figure 11. One-dimensional thermal test setup. An example of the temperature distribution measurements achieved during a set of measurements at constant power input is shown in Figure 12. For tilt angles between 0 and 50, the temperature distribution along the sample is relatively uniform. The inconsistencies in temperatures along the strap were attributed to the uncertainty in the thermocouple measurement (+1 C). A noticeable shift in the operating temperature (i.e., average temperature) of the sample heat strap was apparent, decreasing for an increase in tilt angle. This is a result of the change in heat transfer coefficient at the condenser. At a tilt angle of 60 the temperature at the evaporator began to increase and reached a temperature of 49 C at a tilt of 70. Correspondingly, the capillary limit at 4.8 watts was identified at a 60 tilt angle. The tests were performed for increasing power inputs up to approximately 16 watts. Figure 12. Temperature distribution along heat pipe length at 4.8 watts of heat input. 10

11 A comparison between the predicted and measured maximum heat load as function of tilt angle is shown in Figure 13. The mean absolute error (MAE) 6 between modeled and measured data was calculated as 13.6%. In general, the model slightly over-predicted the maximum heat load values, which can be explained by imperfections that may have been introduced during the manufacturing process, as well as the fact that the model assumes that all layers are perfectly flat, while in reality the sample contains minor bends and dimples, increasing the flow resistances. At the highest heat load level of 15.5 watts, the evaporator temperature was 86.4 C. Although the model prediction fell within the uncertainty of the data point measurement, it is possible that for the high heat load data point boiling is occurring within the heat strap and introducing phenomena not captured by the current model. The presence of boiling would provide an extra heat transfer mechanism and would explain the under-prediction of the model. Figure 13. Experimental data vs. model prediction for maximum heat transport. The effective thermal conductivity based on the effective length and cross-sectional area of the heat strap is shown in Figure 14 as function of heat loads. These data were taken with the heat strap sample at a 0 -orientation and were tested under varying heat loads. Using the distance between thermocouples T 0 and T 4, the thermal strap width (12.7 cm) and the thermal strap total thickness (0.86 mm), the effective thermal conductivity of the thermal strap was calculated for each test point, as shown in Figure 14. It was found that the effective thermal conductivity was relatively constant with an average of 2,150 W/m-K, for heat loads between 4.5 and 15 watts. For reference, Figure 14 compares these values with the thermal conductivity of pure copper (i.e., approximately 400 W/m-K). On average, the thermal strap was found to be more than 400% more conductive than a pure copper spreader. The effective thermal conductivity prediction using the thermal resistance network was roughly two-times higher than the measured values. This discrepancy may have been due to the thermal resistance through the vapor or contact resistances introduced by insufficient bonding between the various layers. 6 The mean absolute error, used to quantify the difference between measured, M, and predicted, F, values, is defined as: n 1 M i Fi MAE 100. n M i 1 11 i

12 Finally, the maximum internal pressure carrying capability of the FlexCool heat strap was evaluated by attaching the prototype to a shop air compressor. It was found that the FlexCool heat strap withstood an internal gauge pressure up to 930 kpa without any failure. It should be noted that the heat strap did not fail at these high internal pressures, and that the test was limited by the pressure output of the compressor. Figure 14. Effective thermal resistance measurements of the FlexCool heat strap at 0 o orientation compared with copper equivalent and model prediction. V. Conclusions This study introduced and evaluated an efficient, flat and lightweight two-phase heat strap for CubeSats applications. A prototype heat strap with dimensions 12.7 cm x 12.7 cm x 0.86 mm was built, and experimental results demonstrated an effective thermal conductivity of up to 2,150 W/m-K. In addition, the heat strap was shown to withstand internal pressures as high as 930 kpa (135 psig), which demonstrates the capability of withstanding internal vapor pressure while operating in space vacuum. The closed-form theoretical model of maximum heat transport was validated through one-dimensional thermal testing. Based on thermal system analyses, the FlexCool two phase heat strap can offer solutions for thermal management of electronic components in small form factor satellites. Figure 15 illustrates a potential configuration for CubeSats in which the heat strap can be designed to fit the small form factor at the same time that transfers thermal energy from electronic components to the body mounted radiators. Future work will include, among others, designing of attachment mechanisms, size standardization for off-the-shelf availability and flight qualification testing. 12

13 Figure 15. FlexCool heat strap development and conceptual integration into a 1U CubeSat. Acknowledgments This work is based on the work supported by NASA under the SBIR Program NNX15CM45P. The authors would like to acknowledge Stephanie Mauro and Jeff Farmer for the support provided during the development of this project. References 1 M. Swartwout, "The First One Hundred CubeSats: A Statistical Look," JoSS, Vol.2, No. 2, 2013, pp Bar-Cohen, A., Matin K., Jankowski, N., and Sharar, D., "Two-Phase Thermal Ground Planes: Technology Development and Parametric Results." Journal of Electronic Packaging, Vol. 137, No. 1, 2015, Panczak, T., Ring, S., Welch, M., Johnson, D., Cullimore, B., and Bell, D., Thermal Desktop User s manual. June Hengeveld, D. W., Hot- and Cold-Case Orbits for Robust Thermal Control, J. Spacecraft and Rockets, Vol. 46, No. 6, 2009, F-Chart Software, EES Engineering Equation Solver for Windows Operating Systems, MATLAB Version 8.6 (R2015b) Natick, Massachusetts: The Mathworks Inc., Peterson, G. P., An Introduction to Heat Pipes: Modeling, Testing and Applications, John Wiley & Sons, Inc., 1994, 8 Faghri, A., Heat Pipe Science and Technology, Taylor & Francis, Bell, I. H., Wronski, J., Quilin, S., and Lemort, V., Pure and Pseudo-pure Fluid Thermophysical Property Evaluation and the Open-Source Thermophysical Property Library CoolProp, Industrial & Engineering Chemistry Research, Vol. 53, No. 6, 2014, pp