International Journal of Heat and Mass Transfer 78 (2014) 1177–1186

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier .com/locate / i jhmt

Multi-nozzle array spray cooling for large area high power devices in aclosed loop system

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.07.0670017-9310/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +65 81227806.E-mail address: mtnwong@ntu.edu.sg (T.N. Wong).

J.L. Xie a, Y.B. Tan a, T.N. Wong a,⇑, F. Duan a, K.C. Toh b, K.F. Choo b, P.K. Chan c, Y.S. Chua c

a School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singaporeb Temasek Laboratories @ NTU, Singaporec DSO National Laboratories, Singapore

a r t i c l e i n f o a b s t r a c t

Article history:Received 27 January 2014Received in revised form 16 July 2014Accepted 18 July 2014Available online 15 August 2014

Keywords:Spray coolingElectronics coolingMulti-nozzle arrayLarge target area

A prototype of a closed loop system was built to study multi-nozzle array spray cooling on high-power,large-area electronic devices. Fifty-four nozzles with an in-lined array of 9� 6 were applied to spray coola simulated 6U electronic card using R134a. Simple drainage concepts were introduced to assist thedrainage of both liquid and vapour on the heated surface. The results indicated a promising prospectof using a multi-nozzle array on large-area power electronics cooling. 16 kW heat was removed fromthe 6U card area by maintaining the mean surface temperature below 26.5 �C. Heat transfer coefficientup to 2.8 �104 W=m2 K was obtained, and liquid usage fraction as high as 0.88 was achieved beforeCHF occurred. It was found that increasing nozzle pressure drop or flow rate enhanced heat transferand gave better surface temperature uniformity. Chamber pressure significantly influenced mean surfacetemperature, but had no observable effects on surface temperature uniformity. The control of chamberpressure can maintain a constant temperature on the heated surface when heat load varied largely.The results also showed that the spray-to-spray interactions had inconspicuous effects on local surfacetemperatures but rather the distance from a location relative to the drainage outlets.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

With the escalating applications of high power electronics, thenecessity for high heat flux cooling over large surface area is grow-ing rapidly. Future electronic devices, including thermoelectricpower generation systems, power conversion systems, and somepower defense electronics, will dissipate high heat fluxes overthe surfaces sizing from tens to thousands square centimetres[1]. Mudawar [2] reviewed the high-heat-flux thermal manage-ment schemes in electronics cooling, such as micro-channel andmini-channel heat sinks, pool boiling, channel flow boiling, jetimpingement, and spray cooling, to show that two-phase thermalmanagement schemes gain two- to three-orders-of-magnitudehigher heat transfer coefficients than that of single-phase thermalmanagement schemes. Among the two-phase thermal manage-ment schemes, spray cooling has been regarded as a most promis-ing high-heat-flux thermal management scheme due to itsprominent advantages on the high heat transfer coefficient, iso-thermal operation, and compact system size due to the low flowrate inventory [3].

The mechanisms of spray cooling that contributed to the highheat flux removal have been extensively studied using a singlenozzle spray cooling over a small surface area [3–5]. The liquid filmformed on the heated surface was reported to have important rolesin the process of spray cooling since most of the heat transfermechanisms, such as droplet impingement, film conduction andevaporation, nucleate boiling, bubble growing and transient con-duction, were involving in the liquid film [3,6,7]. Specially, thethickness of the liquid film dominates the film evaporation, anddetermines whether the impinging droplets can penetrate theliquid film to impact on the heated surface directly [3]. Pais et al.[8] suggested that the optimum heat transfer can be achieved atthe thinnest liquid film by using the smallest possible dropletsand the highest percentage of surface saturation.

In the applications of multi-nozzle array spray cooling, theliquid accumulation between spray cones could generate thickand uneven liquid film on the heated surface. Glassman et al. [9]introduced a liquid management system which employed an arrayof suction tubes to drain liquid from the liquid buildup zones onthe heated surface. Significant heat fluxes increment was obtainedby the liquid managed system as compared to the one withoutliquid management. Glassman [10] also reported that usingmulti-nozzle array spray cooling entailed the liquid management

Nomenclature

A heater surface area, m2

CHF critical heat flux, W=m2

�h mean heat transfer coefficient, W=m2 Khfg latent heat of vaporization, J/kghi:L enthalpy of the inlet liquid, J/kgho:L enthalpy of the outlet liquid, J/kgho:V enthalpy of the outlet vapour, J/kgI current, Ak thermal conductivity of heater block, W/m K_m mass flow rate, kg/s_mi:L inlet liquid mass flow rate, kg/s_mo:L outlet liquid mass flow rate, kg/s_mo:V outlet vapour mass flow rate, kg/s

p pressure, Pa

Q heat load, Wq heat flux, W=m2

Tsat saturation temperature, �CTsurf surface temperature, �CTsurf mean surface temperature, �CV voltage, V

Greek SymbolsDx distance between the thermocouple plane and

heater surface, mDp pressure difference across the nozzles plate, PaDTnon�uniformity surface temperature non-uniformity, �Ce evaporation fraction, �q density, kg=m3

1178 J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186

challenges in both of supplying liquid to the multiple spray nozzlesand draining excess liquid on the heated surface. These challengescan be more crucial when a larger flow rate is demanded to removea higher heat flux level on a large surface, especially in a closedloop system.

Up to date, the studies involving the use of multi-nozzle arraysin closed loops to cool large surfaces are limited. Lin et al. [11,12]tested the spray cooling performance of multi-nozzle arrays byscaling up the cooling surface area from 2.0 to 19.3 cm2 with thenumber of multiple spray nozzles increasing from 8 to 48, respec-tively. The experimental results show that both the heat transfercoefficient and critical heat flux (CHF) decrease for the larger sur-face in comparison to that of the smaller surface. The liquid accu-mulation zones occurred between the spray cones were suggestedto account for the degradation of heat transfer performance overthe larger surface. Yan et al. [13,14] modified a normal refrigerantsystem to test the spray cooling performance of a 2� 2 nozzlesarray in the normal and inclined spray configurations usingR134a as the working refrigerant. The tested surface was sized as15.1 � 13.5 cm2 and four gas-assisted nozzles were used. At anaverage refrigerant mass flux of 0.8 kg=m2 s, a heat flux of5:0� 104 W=cm2 was achieved by keeping the surface temperaturebelow 25 �C. The experimental results indicate that the inclinedspray configuration performs a better heat transfer performancein terms of the heat transfer coefficient. However, the surface tem-perature uniformity is worse as compared to the normal configura-tion. They attributed the better heat transfer performance ofinclined spray configuration to the increased radial dropletsmomentum, which promoted the liquid drainage on the heatedsurface. Tan et al. [15] and Xie et al. [16] used a miniaturized noz-zle plate with 3� 2 nozzles array to cover a surface area of2.0 � 1.0 cm2. R134a was the working fluid. At the same surfacetemperature of 25 �C, a much higher heat flux of 4� 105 W=m2

was achieved when the provided fluid mass flux was 41 kg=m2 s.In their experiments, the heated surface was probably flooded bythe liquid.

Shedd [17] suggested that future closed loop spray cooling sys-tems require a large and carefully engineered volume in the drain-age region so that the excess liquid and generated vapour can bedriven more quickly, uniformly, and smoothly from the heated sur-face. With this motivation, the present study aims to develop aprototype of a high power spray cooling system which, instead ofintegrating a complex liquid management system, intends tovaporize most of the liquid sprayed on the heated surface thusavoiding the liquid accumulation zones on the heated surface.Moreover, simple drainage concepts were adopted to assist thedrainage of both liquid and vapour on the heated surface. The

experiments simulating the application of using a multi-nozzlearray to spray cool a 6U electronic card (23.3 � 16.0 cm2) underhigh heat loads were conducted. The experimental results showthe promisingly high liquid evaporation fractions at high heatloads when maintaining low surface temperatures.

2. Experimental setup and procedure

The closed loop in the present study was developed based on aDorin condensing unit using R134a as the working refrigerant.Fig. 1 shows the schematic of the overall experimental setup,which mainly consists of a condensing unit (including a variablespeed compressor and an air-cooled condenser), a liquid–vapourseparator, a closed loop spray chamber, a heat exchange accumula-tor, and other necessaries. To ensure the system is in proper work-ing conditions, several sight-glasses were installed to track thenature of flow in the closed loop.

The closed loop spray chamber aims to establish proper condi-tions for the multi-nozzle array spray cooling to take place over alarge surface area. As shown in Fig. 2, the closed loop spray cham-ber is assembled vertically with a liquid feeding chamber, a nozzleplate, a spray chamber, a heater block and Teflon insulators. Hence,gravity is applied to assist the liquid drainage on the heated sur-face. The heater block is a rectangular copper block with the overalldimension of 25 cm � 18 cm � 8 cm. After being assembled withthe spray chamber, the exposed surface area (cooling area) of theheater block in the spray chamber is 23.3 � 16.0 cm2 which isthe exact size of a 6U electronic card. Twelve cartridge heaterswere inserted into the copper block in two planes from two sides,as shown in Fig. 2(a). Each cartridge heater is in cylindrical shapewith 1.91 cm in diameter and 15.2 cm in length, and is capableof delivering 2.4 kW heat. These cartridges are connected in paral-lel and the supplied power is regulated by a 3-phase variac trans-former. Before fabricating the heater block, a 3D steady-statethermal simulation in ANSYS was performed to decide the loca-tions of cartridges in the copper block. The simulated temperaturecontours are shown in Fig. 2(a). This simulation aims to ensurethat: (1) under the designed heat loads, the maximum temperaturein the heater block will not exceed the working temperature of theTeflon insulators and that (2) isotherms below the heated surfaceare one-dimensional so that an easy determination of Tsurf is fabri-cated. The thickness of the Teflon insulator was determined by thesame steady-state thermal simulation that the copper block andinsulators are coupled and the boundary condition at the outsidewalls of the insulators is set as a natural convective heat transfercoefficient (h ¼ 25 W=m2 K). By assuming the environment

Fig. 1. Schematic of the experimental setup.

J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186 1179

temperature of 26 �C, the predicted heat loss under the 2-cm-thickTeflon insulators was less than 1% when the heat load varied from2 to 20 kW.

As shown in Fig. 2 (b), there are 37 temperature sensing points(red spots) designed in the heater block 0.4 cm below the heatedsurface to estimate the local surface temperatures. The thermocou-ples used are K-type sheathed thermocouples with the outer diam-eter of 0.1 cm.

The spray chamber with the internal volume of23:3� 16:0� 3:73 cm3 is attached to the copper surface using anO-ring as the sealing. Two outlets were fabricated at the top andbottom walls of the spray chamber to drain the vapour and liquid,respectively. In order to prevent the buildup of liquid and vapour atthe corners of the spray chamber (which are farther away from theoutlets), two long slots (18.5 � 2.3 � 1.2 cm3) were fabricated atthe top and bottom walls to serve as reservoir buffers to drainthe liquid and vapour on the heated surface more uniformly(instead of converging all the liquid and vapour to the relativelysmall orifices of the outlets). The nozzle plate, which comprises54 full cone pressure swirl nozzles with an in-lined array of9 � 6, is installed between the feeding and sprays chambers to per-form liquid atomization and distribute uniform liquid on theheated surface. Meanwhile, the nozzle plate ensures that most ofthe heated surface will be directly impacted by the spray droplets.As shown in Fig. 2(b), these spray nozzles are spaced out that thespray overlapping between the nozzles is prevented. In such anarrangement, 76.8% of the heated surface is allowed to be sprayeddirectly by the spray cones. The feeding chamber attached to thenozzle plate acts as a liquid reservoir which supplies liquid refrig-erant uniformly to all the spray nozzles.

As shown in Fig. 1, when operating the system, the single-phaseliquid is supplied to the spray nozzles, which atomize the liquidinto fine and high velocity droplets in the spray chamber. Mostof the droplets are vaporized after impacting on the heated surface,and the unevaporated droplets directly flow back to the

heat-exchange accumulator. There, the liquid coming from thespray chamber continues to evaporate due to the integrated heatexchange coils. As such, the entire refrigerant is transported backto the compressor in the form of low pressure vapour. After under-going the compression in the compressor, the low pressure vapourbecomes the high temperature and high pressure vapour. Follow-ing which, the compressed vapour is converted into the liquidrefrigerant after being cooled by the air-cooled condenser. The con-verted liquid has a relatively higher temperature and is thereafterchannelled to the heat exchange accumulator to assist the liquidevaporation in the accumulator. Eventually, the liquid is furthercooled down before going into the separator for the next coolingcycle. In this system, the primary function of the heat-exchangeaccumulator is to prevent the excess liquid from flooding in theaccumulator and being carried over to the compressor directly.

During the experiment, the refrigeration system was first acti-vated before the cartridge heaters were switched on. When fully-filled liquid was observed from the sight-glass installed upstreamof the closed loop spray chamber, the supply of heat powerbegan. To test the spray cooling performance under different heatloads, the power input to the heater block was graduallyincreased until CHF was reached. At each power level, the read-ings of all sensors (temperature and pressure) were monitoredusing two Agilent data acquisition systems (Model: 34972A)which were incorporated to the PCs. When a steady state casewas reached, the experimental data for the power level weresaved and a new power input was regulated to attain a newsteady state. The data recorded in each steady state were aver-aged for data analysis.

The experimental liquid flow rate, nozzle inlet pressure, andchamber pressure were adjusted by regulating the compressor fre-quency and the manual valves installed in the system. In particular,the liquid flow rate was measured by a Brooks rotameter (Model:GT1000), rated from 28 to 260 ml/s, with an accuracy of 2%. Thenozzle inlet pressure and the spray chamber pressure were

Fig. 2. Schematic of the closed loop spray chamber.

1180 J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186

measured using two STS absolute pressure transducers (Model:232-XX-13-01-47-0-1-U), rated from 0 to 1 MPa, with accuracyof 0.5%.

3. Data reduction and uncertainty analysis

According to the ANSYS simulation, the heat loss to environ-ment is negligible. Thus, the heat load applied to the heater surfaceis calculated by

Q ¼ V � I ð1Þ

where V is the voltage and I is the current in the electric circuit ofcartridge heaters. Therefore, the heat flux applied on the heated sur-face is calculated as

q ¼ Q=A ð2Þ

where A indicates the heated surface area (23.3 � 16.0 cm2).According to the isotherms below the heated surface, local surfacetemperatures can be determined as

Tsurf :i ¼ Tthermo:i � qDx=k ð3Þ

where Tthermo:i is the local temperature reading of the ith thermocou-ple, k is the thermal conductivity of the heater block, and Dx is thedistance from the thermocouple plane to the heated surface. The

mean surface temperature is thereby calculated as the arithmeticmean value of the local surface temperatures,

Tsurf ¼XN

i¼1

Tsurf :i=N ð4Þ

where N is the number of local surface temperatures. The meanheat transfer coefficient �h is thereby calculated as

�h ¼ q=ðTsurf � TsatÞ ð5Þ

where Tsat is the saturation temperature corresponding to thechamber pressure.

The surface temperature non-uniformity is defined as the dif-ference of the maximum local surface temperature minus the min-imum local surface temperature on the heated surface,

DTnon�uniformity ¼ Tsurf :max � Tsurf :min ð6Þ

In order to study the liquid usage efficiency in the experiments,evaporation fraction e which represents the proportion of liquidevaporated in the spray and heat transfer process with respect tothe total liquid flow rate, is evaluated [16]. Considering the spraychamber as a control volume, the evaporation fraction is definedaccording to the energy and mass conservations in the spraychamber,

Fig. 3. Visualization of multi-nozzle array spray cones impinging on a flat surface: Dp = 0.3 MPa. (a) Footprints of the 54 spray cones impinging on the acrylic surface, (b)Schematic of the liquid drainage and spray-to-spray interactions on the impinged surface.

J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186 1181

_mi:Lhi;L þ Q ¼ _mo:Lho:L þ _mo:V ho;V ð7Þ_mi:L ¼ _mo:L þ _mo:V ð8Þ_mo:V ¼ _mi:L � e; _mo:L ¼ _mi:L � ð1� eÞ ð9Þ

e ¼_mi:Lðhi:L � ho:LÞ_mi:Lðho:V � ho:LÞ

þ Q_mi:Lðho:V � ho:LÞ

ð10Þ

where, _mi:L and hi:L are the inlet liquid mass and enthalpy, respec-tively, _mo:L and ho:L are the outlet liquid mass and enthalpy, respec-tively, _mo:V and ho:V are the outlet vapour mass and enthalpy,respectively. In Eq. (10), the first term on the right hand side isthe evaporation rate due to the nozzle expansion, and the secondterm is the evaporation rate attributed by the applied heat load.

The critical quantities of interest in evaluating the heat transferperformance are the heat load and surface temperature, which aredependent on the accuracy of the measurements of voltage andcurrent from the power control panel, thermocouple readings fromthe data acquisition system and distance from the thermocoupleplane to the heated surface. The uncertainty of the electrical powerthrough the power analyzer (Carlo Gavazzi, Type WM14–96) is±0.5% for voltage and current measurements. The error of the dis-tance from the thermocouple plane to the heated surface is lessthan 0.1 mm. The thermocouples used in this study are all cali-brated in a constant temperature bath before being installed inthe system. The accuracy of the calibrated thermocouples is a func-tion of the thermo-calibrator, which has the accuracy within� 0:1�C in the experimental temperature range. The thermalconductivity value for copper was taken as 398 � 3 W=m K. The

heat loss to environment is less than 1.0% based on the ANSYSsteady state thermal simulation.

4. Results and discussion

4.1. Open loop visualization

An open loop visualization experiment was set up to explorethe impingement pattern of the multi-nozzle array spray conesimpinging on a flat surface. An acrylic sheet with a thickness of5 mm was fabricated to replace the heater block so that the foot-prints of the multi-nozzle array spray could be observed fromthe other side of the acrylic sheet. As shown in Fig. 3(a), the insideand outside views of the multi-nozzle array spray cones are pre-sented. It is observed that the footprints of the 54 spray conesare approximately the same, which suggests a uniform liquid feed-ing to all the spray nozzles. For each spray cone, the idea footprintshape should be a circle with diameter of 26.5 mm. However, in theexperiment, the footprint shape is not a circle but rather like asquare. This is possibly due to the spray-to-spray interactionsbetween the neighbouring spray cones, as well as the gravitationaleffect. Note also that, between the neighbouring rows and columnsof the spray cones, there are horizontal and vertical liquid bridgeswith width around 4 mm formed to assist liquid drainage.

Fig. 3(b) schematically depicts the flow pattern of a segment ofthe liquid flow on the impinged surface. From the perspective ofspray-to-spray interactions, the horizontal and vertical liquid

Fig. 4. Effect of nozzle pressure drop on: (a) mean surface temperature, (b) heattransfer coefficient.

1182 J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186

bridges are formed due to the collision of the surrounding spraycones and gravity. With the aid of gravity and the specified liquiddrainage design, no liquid accumulation has been observed in thespray chamber.

4.2. Effects of nozzle pressure drop

By maintaining the chamber pressure, the mean surface tem-perature and heat transfer coefficient as a function of nozzle pres-sure drop are illustrated in Fig. 4. Two different heat loads weretested at their respective chamber pressures. The nozzle pressuredrop was varied from 0.24 to 0.42 MPa, resulting in the flow rateranging from 78 to 99 ml/s accordingly. At Q = 8.5 kW, it isobserved that the mean surface temperature decreases from 19.7to 16.3 �C as the nozzle pressure drop increases from 0.24 to0.37 MPa (see Fig. 4(a)), and correspondingly the heat transfercoefficient increases from 1:61� 104 to 2:11� 104 W=m2 K (seeFig. 4(b)). At Q = 12.6 kW, the same observation is found. As thenozzle pressure drop increases from 0.37 to 0.42 MPa, the meansurface temperature is reduced from 24.7 to 22.2 �C with the cor-responding heat transfer coefficient increasing from 2:38� 104 to2:81� 104 W=m2 K. This demonstrates the positive effects of noz-zle pressure drop on the heat transfer effectiveness of spraycooling.

The enhanced heat transfer performance by increasing nozzlepressure drop is due to the improved spray characteristics asobserved in the previous work [4]. Increasing the nozzle pressuredrop consequently increases the flow rate, droplet flux and outfitsthe droplets with higher impact energy [4]. First, increasing flowrate improves the spray cooling performance by intensifying theconvective flow and sufficiently wetting over the heated surface[14,18,19]. Second, an increase in droplet flux could enhance heattransfer in the boiling regime of spray cooling by providing moresecondary nuclei [20]. Last, the higher impact energy of dropletswould assist droplets to impact on the heated surface, which atfirst enhances the heat transfer effectiveness of droplet impinge-ment cooling [21]. Besides, the droplets with higher impact energystrongly agitate the thermal boundary layer of the formed liquidfilm on the heated surface, which enhances the heat transfer per-formance as well [17].

4.3. Effect of chamber pressure

By regulating the control valves at the outlets of spray chamber,the chamber pressure can be adjusted over a wide range. Mean-while, the flow rate can be maintained in a narrow range by con-trolling the expansion valve and compressor running speed inthe closed loop. Fig. 5 presents the effects of chamber pressureon the mean surface temperature and heat transfer coefficient,respectively. Four different heat loads were studied by maintainingthe flow rate ranging from 94 to 98 ml/s.

Fig. 5(a) shows that for all the tested heat loads, the mean sur-face temperatures increase almost linearly with increasing cham-ber pressure. Meanwhile, the chamber saturation temperatureincreases linearly with the chamber pressure at the similar slopeof the surface temperature curve. This implies that the chamberpressure affects mainly the mean surface temperature by varyingthe saturation temperature inside the spray chamber. By regulat-ing the chamber pressure, the same mean surface temperaturecan be achieved even if the applied heat load varies largely. Forinstance, as shown in Fig. 5(a), the same mean surface temperatureof Tsurf = 22 �C is obtained when the heat load increases from 10.3to 14.2 kW (with 37.9% heat load increment). The result revealsthat controlling the chamber pressure is an effective way to main-tain the same operating temperature of the electronic device whenit works under different heat loads.

Fig. 5(b) demonstrates that increasing chamber pressure andthus the saturation temperature of the refrigerant, is appreciablyconducive to the heat transfer performance in spray cooling. Anincrease in chamber pressure results in the increase of heat trans-fer coefficient, which agrees with the findings by Lin et al. [12] andYan et al. [13] that the heat transfer at a higher saturation temper-ature is more effective than that at a lower saturation temperature.Rainey et al. [22] found that increasing chamber pressure resultedin a wider range of cavity radius that can be activated at a givenwall superheat in pool boiling. Hence, the improved heat transferperformance observed in Fig. 5(b) is probably due to the increasedactive nucleation site density as a result of increasing chamberpressure.

4.4. Critical heat flux

A series of CHF experiments were conducted by slowly increas-ing heat load until a sudden temperature jump (CHF) wasobserved. As shown in Fig. 6, the transition process from the laststeady state point to the incidence of CHF is illustrated. Three dis-tinct regimes can be divided in the transition process: the steadystate regime, transition regime, and film boiling regime. In thesteady state regime where Q = 16.1 kW, the thermocouple readingsare stable and constant. As the heat load increases slightly to16.5 kW, the thermocouple readings shoot up rapidly from 30 to75 �C in 10 min which is so-called the transition regime. Thereaf-ter, a sudden temperature jump occurs to reach the film boilingregime, and at that instant, heater is powered off immediately.

Fig. 5. Effect of chamber pressure on: (a) mean surface temperature, (b) heattransfer coefficient.

Fig. 7. Effect of heat load on: (a) mean surface temperature, (b) heat transfercoefficient, in CHF studies.

J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186 1183

The heater surface remains in film boiling until the stored heat inthe heater block is not enough to sustain a high surface tempera-ture for film boiling. When film boiling finally disappears, normalspray cooling scenario returns and the surface temperaturedecreases rapidly. In spray cooling, the incidence of CHF is attrib-uted to two mechanisms. Pais et al. [23] suggested that CHF inspray cooling occurs due to the combination of neighbouring

Fig. 6. Transition process from steady state to CHF.

bubbles on the heated surface which creates dry vapour stemsand hot spots to finally form a large vapour film. However, Linand Ponnappan [11] suggested that CHF in spray cooling wascaused by the inability of the impinging droplets to reach theheated surface due to the vapour flow uprising from the heatedsurface. Xie et al. [24] experimentally found that the impingingdroplets could be blown away by the uprising vapour flow beforeimpacting on the heated surface to perform heat transfer whenthe surface temperature is high.

Fig. 7(a) shows the variations of mean surface temperature andevaporation fraction with increasing heat load during the CHF test.As expected, the mean surface temperature as well as evaporationfraction monotonically increases with increasing heat load. WhenQ < 15 kW, the surface temperature increases gently with theincrease of heat load, thereafter increases more rapidly until CHFoccurs. As can be seen from Fig. 7(a), although the mean surfacetemperature obtained from the last steady state point is very low(Tsurf = 26.5 �C at Q = 16.1 kW), CHF occurs when heat load isslightly increased to 16.5 kW (see Fig. 6). This is probably due tothe insufficient liquid supply on the heated surface. For example,assuming the steady state working conditions at Q = 16.5 kW aresimilar to that at Q = 16.1 kW, the estimated liquid evaporationfraction at Q = 16.5 kW reaches 0.91. Such evaporation fraction istoo high to maintain a steady state heat transfer situation therebyleads to the incidence of CHF.

Fig. 7(b) demonstrates the corresponding heat transfer coeffi-cient with increasing heat load. It reveals that when Q < 15 kWthe heat transfer coefficient increases with increasing heat loadand subsequently decreases when Q > 15 kW. This agrees with

Table 1Evaporation fractions under different CHF tests.

CHF tests Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 error

Flow rate (ml/s) 88 90 91 93 94 97 ± 2.0%Last steady state load Q (kW) 15.5 15.6 15.8 15.7 16.1 16.0 ± 0.5%Heat load at CHF Q (kW) 15.8 16.0 16.1 16.1 16.5 16.7 ± 0.5%Last steady state Tsurf ð�C) 25.3 25.1 25.3 25.6 26.5 25.6 ± 0.2

Evaporation fraction (e) at last steady state point 0.89 0.88 0.88 0.88 0.89 0.88 ± 0.018

Fig. 8. Effects of heat load and flow rate on the temperature non-uniformity.

1184 J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186

the results obtained by Bostanci et al. [25] which reported that theheat transfer coefficient decreased before the incidence of CHF.When Q > 15 kW, the evaporation fraction is too high (e > 0.85),which makes a continuous liquid film spreading over the heatedsurface to be questionable. On the contrary, it is more likely thatat some locations on the heated surface, the sprayed liquid is com-pletely vaporized to create some local ‘‘dryout’’. These ‘‘dryout’’locations resulted in some hot spots on the heated surface, andhence caused the rapid increase in surface temperature and adecrease in heat transfer coefficient before CHF occurs.

Table 1 lists the CHF tests under different flow rates. It can beseen that for all the CHF tests, the evaporation fraction obtainedat the last steady state point is almost at the same level (e =0.88). When the applied heat load increases slightly, CHF occurs.This implies that the evaporation fraction of 0.88 could be the crit-ical value to separate the heat transfer situation from the steadystate operation to the incidence of CHF. Therefore, it is believedthat increasing flow rate can delay the incidence of CHF by main-taining the evaporation efficiency below the critical value.

Fig. 9. Effect of chamber pressure on the surface temperature non-uniformity.

4.5. Surface temperature non-uniformity

Another important criterion to evaluate the reliability of usingmulti-nozzle array spray cooing technique on large area electronicscooling is the surface temperature non-uniformity. As shown inFig. 8, the surface temperature non-uniformity as a function of heatload is demonstrated. It indicates that the surface temperaturenon-uniformity is sensitive to the applied heat load, especiallywhen the flow rate is lower. This is likely due to the higher liquidconsumption at a higher heat load. As discussed above, a higherevaporation fraction is easier to cause local ‘‘dryout’’ on the heatedsurface, which consequently result in some hot spots with highlocal surface temperature. This phenomenon is more evident whenthe applied heat load is near to the incidence of CHF as shown inFig. 8.

When the flow rate ranges from 91 to 93 ml/s, the surface tem-perature non-uniformity suddenly increases from 5.9 to 9.1 �C asthe heat load increases from 14.2 to 15.7 kW (0:79 6 e 6 0:88).The same phenomenon is also observed at the flow rate from 94to 98 ml/s when the heat load increases from 14.4 to 16 kW(0:80 6 e 6 0:89). The other possible reason that accounts for thelarger surface temperature non-uniformity at a higher heat loadis the effect of vapour generated from the heated surface. Xieet al. [4,24] reported that the vapour uprising from the heated sur-face influenced the spray cone formation in spray cooling, whichcould change the liquid distribution on the heated surface anddeprave the surface temperature uniformity. Fig. 8 also manifeststhat increasing liquid flow rate is beneficial to maintain a uniformsurface temperature (e.g., 97 < flow rate < 100 ml/s). This isbecause a higher flow rate creates a more stable liquid film onthe heated surface and hence, a more uniform surface temperatureon the heated surface.

Although the chamber pressure has a distinguished influenceon the mean surface temperature, Fig. 9 shows no detectableeffects of chamber pressure on the surface temperature

non-uniformity. It is reasonable since the chamber pressure onlyinfluences the surface temperature by changing the saturationtemperature inside the spray chamber.

Furthermore, to characterize the surface temperature distribu-tion in detail, the heated surface has been divided into five regions(from A to E), as shown in Fig. 10. The mean surface temperaturesof these five regions are illustrated in Fig. 11.

As can be seen that, region B has the highest surface tempera-ture, followed by regions A and C, while regions D and E havethe lowest surface temperature. Region B being at the highest sur-face temperature could be due to the fact that it is situated at theplace directly facing to the liquid and vapour outlets at the bottomand top of the spray chamber. As a result, some of the sprayed

Fig. 10. Temperature regions divided on a 6U card area. (a) Temperature regions onthe 6U surface, (b) Thermocouple locations in Region I, (c) Thermocouple locationsin Region II.

Fig. 12. Local surface temperatures at different measuring points: (a) Region I, (b)Region II.

J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186 1185

liquid at this region could drift quickly to the outlets beforeimpinging on the heated surface. It is suspected that the sprayedliquid directing to region B may not be fully utilized before beingdrained to regions D and E (near to the outlets). Therefore, regionB has the highest surface temperature while regions D and Ereceiving more refrigerant obtain the lowest surface temperature.In addition, the lower surface temperature obtained at regions Dand E could be due to their relatively lower local chamber pressureas well. As shown in Fig. 10(a), regions D and E are closest to theoutlets and hence, experience a lowest local chamber pressure ascompared to regions A, B and C (noted that the generated vapourin spray chamber flows from regions A, B and C to regions D andE). A lower chamber pressure results in a lower saturation temper-ature which causes a lower surface temperature.

Fig. 11. Mean surface temperature at different regions on the 6U card area.

To characterize the effects of spray-to-spray interactions onlocal surface temperatures, two sub regions (see regions I and IIin Fig. 10) in regions D and E are investigated. The proposed liquidflow pattern in these sub regions is depicted in Fig. 3(b). Regions Iand II are identical to each other, except Region I is near to thevapour outlet (top) and Region II is near to the liquid and vapouroutlet (bottom). Six thermocouples were used to measure the localsurface temperature in each region. Particularly, the surface tem-peratures of the interested locations were measured as shown inFigs. 10(b) and (c), such as the stagnation points of spray cones(I-1, I-2, II-1, II-2), colliding points of four spray cones (I-3, I-4, I-6, II-3, II-4, II-6), and the colliding points of two spray cones (I-5,II-5). The local surface temperatures under different heat loadsfor regions I and II are illustrated in Fig. 12, respectively. It is notedthat at Q < 11:8 kW, the local surface temperatures at I-1 to I-4have no distinct discrepancies between each other in region I, aswell as II-1 to II-4 in region II. While, the surface temperatures atI-5 to I-6 are higher at all the tested heat loads in region I, as wellas that at II-5 to II-6 in region II. The results imply that rather thanthe spray-to-spray interactions, the distance of a location relativeto the outlets has more significant effect on the local surface tem-perature. For instance, at Q = 16.1 kW, the local surface tempera-ture is highest at I-6 and II-6 which are the farthest points to theoutlets, followed by I-5 and II-5 (second farthest), then I-4 and II-4 (third farthest) as shown in Fig. 12.

1186 J.L. Xie et al. / International Journal of Heat and Mass Transfer 78 (2014) 1177–1186

5. Conclusions

A closed loop high power spray cooling system has been builtand tested using R134a as the working coolant. The experimentedsurface was a flat copper surface with the 6U electronic carddimensions (23.3 cm � 16.0 cm). Fifty-four pressure swirl nozzleswere assembled on a holding plate with an in-lined array of9 � 6 to cover a large ratio of the copper surface. Parametric effectson the surface temperature, heat transfer coefficient, surface tem-perature non-uniformity, and CHF were investigated. The key find-ings are as follows.

1. Experimental results indicate that increasing nozzle pressuredrop enhances the heat transfer performance. At the heat loadof 12.6 kW, the mean surface temperature reduces from 24.7to 22.2 �C and the corresponding heat transfer coefficientincreases from 2.38 to 2.81 �104 W=m2 K with increasing thenozzle pressure drop from 0.37 to 0.42 MPa.

2. Chamber pressure significantly affects the mean surface tem-perature by changing the saturation temperature inside thespray chamber. Hence, by controlling the chamber pressure,the system is able to maintain a constant operating temperaturefor a power device working under different heat loads. Besides,a higher chamber pressure results in a higher heat transfercoefficient.

3. Increasing heat load degrades the surface temperature unifor-mity, especially when the flow rate is low. While, increasingflow rate appears to achieve a uniform surface temperature dis-tribution. The results show that chamber pressure has nodetectable effects on the surface temperature non-uniformity.

4. CHF occurs when the liquid evaporation fraction is beyond acritical value (e = 0.88) even if the surface temperature is stilllow. Therefore, increasing flow rate will delay the incidence ofCHF as long as the evaporation fraction is maintained belowthe critical value.

Conflict of interest

We wish to draw the attention of the Editor to the followingfacts which may be considered as potential conflicts of interestand to significant financial contributions to this work. We confirmthat the manuscript has been read and approved by all namedauthors and that there are no other persons who satisfied the cri-teria for authorship but are not listed. We further confirm thatthe order of authors listed in the manuscript has been approvedby all of us.

We confirm that we have given due consideration to the protec-tion of intellectual property associated with this work and thatthere are no impediments to publication, including the timing ofpublication, with respect to intellectual property. In so doing weconfirm that we have followed the regulations of our institutionsconcerning intellectual property. We understand that the Corre-sponding Author is the sole contact for the Editorial process(including Editorial Manager and direct communications with theoffice). He/she is responsible for communicating with the otherauthors about progress, submissions of revisions and final approvalof proofs. We confirm that we have provided a current, correct

email address which is accessible by the Corresponding Authorand which has been configured to accept email from xiej0010@e.n-tu.edu.sg or mtnwong@ntu.edu.sg.

References

[1] A. Bhunia, C.L. Chen, Liquid micro-jet array impingement cooling for large areasystems, in: Proceedings of the Asme International Mechanical EngineeringCongress and Exposition – 2009 – Advances in Aerospace Technology, 2010.

[2] I. Mudawar, Assessment of high-heat-flux thermal management schemes, IEEETrans. Compon. Packag. Technol. 24 (2) (2001) 122–141.

[3] J.H. Kim, Spray cooling heat transfer: the state of the art, Int. J. Heat Fluid Flow28 (4) (2007) 753–767.

[4] J.L. Xie, Z.W. Gan, F. Duan, T.N. Wong, S.C.M. Yu, R. Zhao, Characterization ofspray atomization and heat transfer of pressure swirl nozzles, Int. J. Therm. Sci.68 (2013) 94–102.

[5] B. Horacek, K.T. Kiger, J. Kim, Single nozzle spray cooling heat transfermechanisms, Int. J. Heat Mass Transfer 48 (2005) 1425–1438.

[6] G.E. Martinez, J.C. Ramos, R. Anton, R. Khodabandeh, Film thickness and heattransfer measurements in a spray cooling system with R134a, J. Electron.Packag. 133 (1) (2011).

[7] J.L. Xie, R. Zhao, F. Duan, T.N. Wong, Thin liquid film flow and heat transferunder spray impingement, Appl. Therm. Eng. 48 (2012) 342–348.

[8] M. Pais, D. Tilton, L. Chow, E. Mahefky, High heat flux, low superheatevaporative spray cooling, in: Proceedings of the 27th AIAA Aerospace SciencesMeeting, Reno, NV, 1989.

[9] B. Glassman, S. Kuravi, J. Du, Y.-r. Lin, G. Zhao, L. Chow, A fluid managementsystem for a multiple nozzle array spray cooler, in: 37th AIAA ThermophysicsConference 2004, American Institute of Aeronautics and Astronautics, 2004.

[10] B.S. Glassman, Spay cooling for land, sea, air and space based applications, afluid management system for multiple nozzle spray cooing and a guide to highheat flux heater design, in: Department of Mechanical Material and AerospaceEngineering, University of Central Florida, 2005

[11] L. Lin, R. Ponnappan, Heat transfer characteristics of spray cooling in a closedloop, Int. J. Heat Mass Transfer 46 (2003) 3737–3746.

[12] L. Lin, R. Ponnappan, K.Y.A.B. Hager, Large Area Spray Cooling, in: 42nd AIAAAerospace Sciences Meeting and Exhibit, AIAA: Reno, Nevada, 2004

[13] Z.B. Yan, K.C. Toh, F. Duan, T.N. Wong, K.F. Choo, P.K. Chan, Y.S. Chua,Experimental study of impingement spray cooling for high power devices,Appl. Therm. Eng. 30 (10) (2010) 1225–1230.

[14] Z.B. Yan, F. Duan, T.N. Wong, K.C. Toh, K.F. Choo, P.K. Chan, Y.S. Chua, L.W. Lee,Large area impingement spray cooling from multiple normal and inclinedspray nozzles, Heat Mass Transfer (2013) 1–6.

[15] Y.B. Tan, J.L. Xie, F. Duan, T.N. Wong, K.C. Toh, K.F. Choo, P.K. Chan, Y.S. Chua,Multi-nozzle spray cooling for high heat flux applications in a closed loopsystem, Appl. Therm. Eng. 54 (2) (2013) 372–379.

[16] J.L. Xie, Y.B. Tan, F. Duan, K. Ranjith, T.N. Wong, K.C. Toh, K.F. Choo, P.K. Chan,Study of heat transfer enhancement for structured surfaces in spray cooling,Appl. Therm. Eng. 59 (1-2) (2013) 464–472.

[17] T.A. Shedd, Next generation spray cooling: high heat flux management incompact spaces, Heat Transfer Eng. 28 (2) (2007) 87–92.

[18] S.-S. Hsieh, T.-C. Fan, H.-H. Tsai, Spray cooling characteristics of water and R-134a. Part I: nucleate boiling, Int. J. Heat Mass Transfer 47 (2004) 5703–5712.

[19] L.C. Chow, M.S. Sehembey, M.R. Pais, High heat flux spray cooling, Ann. Rev.Heat Transfer 8 (1997) 291–318.

[20] D.P. Rini, R.H. Chen, L.C. Chow, Bubble behavior and nucleate boiling heattransfer in saturated FC-72 spray cooling, J. Heat Transfer Trans. Asme 124 (1)(2002) 63–72.

[21] R.J. Issa, S.C. Yao, Numerical model for spray-wall impaction and heat transferat atmospheric conditions, J. Thermophys. Heat Transfer 19 (4) (2005) 441–447.

[22] K.N. Rainey, S.M. You, S. Lee, Effect of pressure, subcooling, and dissolved gason pool boiling heat transfer from microporous surfaces in FC-72, J. Heat MassTransfer 125 (1) (2003) 75–83.

[23] M.R. Pais, L.C. Chow, E.F. Mahefkey, Surface roughness and its effects on theheat transfer mechanism in spray cooling, J. Heat Transfer Trans. Asme 114(1992) 211–219.

[24] J.L. Xie, Z.W. Gan, T.N. Wong, F. Duan, S.C.M. Yu, Y.H. Wu, Thermal effects on apressure swirl nozzle in spray cooling, Int. J. Heat Mass Transfer 73 (2014)130–140.

[25] H. Bostanci, D.P. Rini, J.P. Kizito, V. Singh, S. Seal, L.C. Chow, High heat fluxspray cooling with ammonia: Investigation of enhanced surfaces for CHF, Int. J.Heat Mass Transfer 55 (13-14) (2012) 3849–3856.