international journal heat mass transfertsl.energy.hust.edu.cn/2021_cuihaichuan_01.pdfh.c. cui, x.t....

15
International Journal of Heat and Mass Transfer 167 (2021) 120839 Contents lists available at ScienceDirect International Journal of Heat and Mass Transfer journal homepage: www.elsevier.com/locate/hmt Overall numerical simulation and experimental study of a hybrid oblique-rib and submerged jet impingement/microchannel heat sink H.C. Cui, X.T. Lai, J.F. Wu, M.Z. Wang, W. Liu , Z.C. Liu School of Energy and Power Engineering, Huazhong University of Science and Technology, Wuhan 430074, China a r t i c l e i n f o Article history: Received 16 July 2020 Revised 15 November 2020 Accepted 16 December 2020 Keywords: Hybrid heat sink Oblique-rib Microchannel and jet Rib arrangement Impingement Overall numerical simulation a b s t r a c t In this work, two submerged jet impingement/microchannel heat sink (JIMHS) models were pro- posed, i.e., straight-rib jet impingement/microchannel heat sink (SJIMHS) and oblique-rib jet impinge- ment/microchannel heat sink (OJIMHS). The heat transfer and flow characteristics of the two models were investigated by overall numerical simulation and experiment. In the numerical simulation, the effects of heat flux, pressure drop and rib arrangement on the internal flow and heat transfer of the heat sink were studied. The results indicate that under the same heat flux and inlet condition, the heat transfer surface of OJIMHS achieves more uniform and lower temperature distribution compared with that of SJIMHS, and the average convective heat transfer coefficient of the OJIMHS is obviously higher than that of SJIMHS in all calculation cases, with an increase of about 20%. In addition, the performance of OJIMHS was tested experimentally. The comparison indicates that the maximum relative errors of average temperature and heat transfer coefficient between simulation and experiment were less than 9%. When the volume flow rate is 0.5 L/min and the heat flux is 100 W/cm², the average temperature of the heat transfer surface is still lower than 60°C. Besides, the averaged heat transfer coefficient of 2.8W/(cm 2 · K) was achieved under the inlet fluid temperature of 283K and volume flow rate of 2.5 L/min in the experiment. © 2020 Elsevier Ltd. All rights reserved. 1. Introduction With the continuous progress of information technology and micromachining technology, most of electronics are developing in the direction of miniaturization. However, the problem of heat dissipation with high heat flux has become a major obstacle to the development of electronics in the future [1–3]. According to Moore’s law, the number of transistors in dense integrated circuits increases exponentially [4]. As the number increases, the heat flux in electronics can reach 100 W/cm 2 [5]. Faced with such high heat flux, most of the conventional cooling methods have failed [6]. In comparison, jet impingement/microchannel cooling technology, which combines the advantages of microchannel cooling and jet impingement cooling technology, has been proved to be capable to afford high heat flux cooling capacity by simulation and experi- ment in recent years [7–12]. Abbreviations: EMF, electromagnetic flowmeter; GCI, grid convergence in- dex; ORP, oblique ribbed plate; SRP, straight ribbed plate; SJIMHS, straight- rib jet impingement/microchannel heat sink; OJIMHS, oblique-rib jet impinge- ment/microchannel heat sink. Corresponding authors. E-mail addresses: [email protected] (W. Liu), [email protected] (Z.C. Liu). Microchannel heat sink has been widely used in the heat dis- sipation of some high heat flux electronics because of its high ef- ficiency and compact structure. Tuckerman and Pease introduced the concept of microchannel heat sink for the first time in 1981, and indicated that this method can significantly reduce heat trans- fer resistance and then enhance heat transfer [13]. In the following decades, a lot of experiments and simulations have been carried out to optimize the shape of microchannel, the internal structure and the flow pattern of the heat sink. Xia et al. studied the heat transfer and flow characteristics of complex wave-like microchan- nel heat sink by experiment and numerical simulation, and com- pared with the rectangular microchannel heat sink. The results in- dicate that the heat transfer performance of the complex wavy-like microchannel heat sink is better than that of the traditional rect- angular microchannel heat sink [14]. Deng et al. compared the heat transfer performance of shape reentrant microchannel heat sink and rectangular microchannel heat sink under different work- ing fluids, deionized water and ethanol, experimentally. The results show that the reentrant microchannels give a better heat transfer performance for both coolants tests [15]. Zhai et al. investigated the effect of parallel flow, counter flow and different channel ge- ometries on the heat transfer performance of the double-layered microchannel heat sink by three-dimension method. The results indicate that the double-layered microchannel heat sink with dif- https://doi.org/10.1016/j.ijheatmasstransfer.2020.120839 0017-9310/© 2020 Elsevier Ltd. All rights reserved.

Upload: others

Post on 26-Jan-2021

3 views

Category:

Documents


0 download

TRANSCRIPT

  • International Journal of Heat and Mass Transfer 167 (2021) 120839

    Contents lists available at ScienceDirect

    International Journal of Heat and Mass Transfer

    journal homepage: www.elsevier.com/locate/hmt

    Overall numerical simulation and experimental study of a hybrid

    oblique-rib and submerged jet impingement/microchannel heat sink

    H.C. Cui, X.T. Lai, J.F. Wu, M.Z. Wang, W. Liu ∗, Z.C. Liu ∗

    School of Energy and Power Engineering, Huazhong University of Science and Technology, Wuhan 430074, China

    a r t i c l e i n f o

    Article history:

    Received 16 July 2020

    Revised 15 November 2020

    Accepted 16 December 2020

    Keywords:

    Hybrid heat sink

    Oblique-rib

    Microchannel and jet

    Rib arrangement

    Impingement

    Overall numerical simulation

    a b s t r a c t

    In this work, two submerged jet impingement/microchannel heat sink (JIMHS) models were pro-

    posed, i.e., straight-rib jet impingement/microchannel heat sink (SJIMHS) and oblique-rib jet impinge-

    ment/microchannel heat sink (OJIMHS). The heat transfer and flow characteristics of the two models were

    investigated by overall numerical simulation and experiment. In the numerical simulation, the effects of

    heat flux, pressure drop and rib arrangement on the internal flow and heat transfer of the heat sink were

    studied. The results indicate that under the same heat flux and inlet condition, the heat transfer surface

    of OJIMHS achieves more uniform and lower temperature distribution compared with that of SJIMHS, and

    the average convective heat transfer coefficient of the OJIMHS is obviously higher than that of SJIMHS in

    all calculation cases, with an increase of about 20%. In addition, the performance of OJIMHS was tested

    experimentally. The comparison indicates that the maximum relative errors of average temperature and

    heat transfer coefficient between simulation and experiment were less than 9%. When the volume flow

    rate is 0.5 L/min and the heat flux is 100 W/cm ², the average temperature of the heat transfer surface is still lower than 60 °C. Besides, the averaged heat transfer coefficient of 2.8 W / ( cm 2 · K) was achieved under the inlet fluid temperature of 283K and volume flow rate of 2.5 L/min in the experiment.

    © 2020 Elsevier Ltd. All rights reserved.

    1

    m

    t

    d

    t

    M

    i

    i

    fl

    I

    w

    i

    t

    m

    d

    r

    m

    s

    fi

    t

    a

    f

    d

    o

    a

    t

    n

    p

    d

    m

    a

    t

    h

    0

    . Introduction

    With the continuous progress of information technology and

    icromachining technology, most of electronics are developing in

    he direction of miniaturization. However, the problem of heat

    issipation with high heat flux has become a major obstacle to

    he development of electronics in the future [1–3] . According to

    oore’s law, the number of transistors in dense integrated circuits

    ncreases exponentially [4] . As the number increases, the heat flux

    n electronics can reach 100 W / c m 2 [5] . Faced with such high heat

    ux, most of the conventional cooling methods have failed [6] .

    n comparison, jet impingement/microchannel cooling technology,

    hich combines the advantages of microchannel cooling and jet

    mpingement cooling technology, has been proved to be capable

    o afford high heat flux cooling capacity by simulation and experi-

    ent in recent years [7–12] .

    Abbreviations: EMF, electromagnetic flowmeter; GCI, grid convergence in-

    ex; ORP, oblique ribbed plate; SRP, straight ribbed plate; SJIMHS, straight-

    ib jet impingement/microchannel heat sink; OJIMHS, oblique-rib jet impinge-

    ent/microchannel heat sink. ∗ Corresponding authors.

    E-mail addresses: [email protected] (W. Liu), [email protected] (Z.C. Liu).

    s

    i

    s

    p

    t

    o

    m

    i

    ttps://doi.org/10.1016/j.ijheatmasstransfer.2020.120839

    017-9310/© 2020 Elsevier Ltd. All rights reserved.

    Microchannel heat sink has been widely used in the heat dis-

    ipation of some high heat flux electronics because of its high ef-

    ciency and compact structure. Tuckerman and Pease introduced

    he concept of microchannel heat sink for the first time in 1981,

    nd indicated that this method can significantly reduce heat trans-

    er resistance and then enhance heat transfer [13] . In the following

    ecades, a lot of experiments and simulations have been carried

    ut to optimize the shape of microchannel, the internal structure

    nd the flow pattern of the heat sink. Xia et al. studied the heat

    ransfer and flow characteristics of complex wave-like microchan-

    el heat sink by experiment and numerical simulation, and com-

    ared with the rectangular microchannel heat sink. The results in-

    icate that the heat transfer performance of the complex wavy-like

    icrochannel heat sink is better than that of the traditional rect-

    ngular microchannel heat sink [14] . Deng et al. compared the heat

    ransfer performance of � − shape reentrant microchannel heat ink and rectangular microchannel heat sink under different work-

    ng fluids, deionized water and ethanol, experimentally. The results

    how that the reentrant microchannels give a better heat transfer

    erformance for both coolants tests [15] . Zhai et al. investigated

    he effect of parallel flow, counter flow and different channel ge-

    metries on the heat transfer performance of the double-layered

    icrochannel heat sink by three-dimension method. The results

    ndicate that the double-layered microchannel heat sink with dif-

    https://doi.org/10.1016/j.ijheatmasstransfer.2020.120839http://www.ScienceDirect.comhttp://www.elsevier.com/locate/hmthttp://crossmark.crossref.org/dialog/?doi=10.1016/j.ijheatmasstransfer.2020.120839&domain=pdfmailto:[email protected]:[email protected]://doi.org/10.1016/j.ijheatmasstransfer.2020.120839

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    f

    f

    c

    c

    p

    b

    t

    t

    a

    t

    n

    s

    c

    m

    fl

    e

    n

    e

    c

    s

    r

    [

    p

    w

    a

    u

    a

    p

    t

    t

    s

    a

    m

    a

    t

    s

    a

    l

    i

    c

    f

    S

    t

    T

    t

    g

    a

    t

    S

    t

    t

    t

    j

    a

    J

    t

    t

    f

    t

    t

    r

    t

    n

    J

    s

    i

    t

    t

    f

    t

    t

    t

    t

    i

    t

    c

    o

    t

    t

    t

    i

    m

    d

    n

    t

    Nomenclature

    A 1 , A 2 the area of the bottom heating surface and heat

    transfer surface respectively ( m 2 )

    C 1 ε , C 2 ε constants for RNG k − ε model c p specific heat at constant pressure ( J · k g −1 · K −1 ) D diameter(m)

    h, h heat transfer coefficient and the average heat trans-

    fer coefficient, respectively ( W · c m −2 · K −1 ) H height(m)

    k thermal conductivity ( W · c m −2 · K −1 ) or turbulence kinetic energy(J)

    L length (m)

    ˙ m coolant mass flow rate ( kg · s −1 ) N the number of grid

    P pressure (Pa)

    Q the heating power(W)

    q heat flux of the bottom heating surface ( W · c m −2 ) Re Reynolds number

    T , T temperature and the average temperature, respec-

    tively (K, °C) u, v , w velocity ( m · s −1 ) W width ( m )

    Zn jet-to-surface space ratio

    Greek symbols

    ρ density ( kg · m −3 ) ε dissipation rate of turbulence kinetic energy λ thermal conductivity ( W · m −1 · K −1 ) μ dynamic viscosity ( Pa · s ) ηt eddy viscosity coefficient σε , σk turbulent Prandtl numbers for k and ε

    Subscript

    a v laterally-averaged f average value of inlet and outlet

    in the heat sink inlet

    jet jet hole

    max maximum value

    min minimum value

    w the bottom heat transfer surface

    out the heat sink outlet

    erent channel geometries in each layer has a better heat trans-

    er performance and lower pressure drop [16] . Although the mi-

    rochannel heat sink can realize high heat flux dissipation by in-

    reasing heat transfer area or enhancing disturbance [14] , the large

    ressure and temperature gradient along the flow direction have

    ecome the main obstacles restricting its application [ 13 , 17 ].

    As for jet impingement cooling technology, due to the ex-

    remely thin velocity and temperature boundary layer formed after

    he high-speed fluid impinges on the heat transfer surface [ 18 , 19 ],

    very high convective heat transfer coefficient will be generated in

    he stagnation zone of jet. In addition, compared with microchan-

    el cooling technology, jet impingement cooling technology has a

    maller pressure loss [17] . As a consequence, the jet impingement

    ooling technology is considered to be one of the most potential

    eans to deal with the problem of heat dissipation with high heat

    ux and has attracted extensive attention from scholars [20] . Wang

    t al. studied the single-hole jet and jet array by three-dimensional

    umerical simulation and experimental method, and compared the

    ffects of jet orifice diameter and the number of jet orifices on

    ooling effect [21] . Since the heat transfer coefficient will drop

    harply away from the stagnation point, the application of jet ar-

    ay is more common in the field of electronics heat dissipation

    2

    22 , 23 ]. McInturff et al. investigated the effect of hole shape on im-

    ingement jet array heat transfer by experimental method. The test

    as conducted under the jet Reynolds numbers of 90 0, 150 0,50 0 0,

    nd 11,0 0 0 and the results show that the racetrack hole config-

    ration generally provides approximately the same heat transfer

    ugmentation as the circle hole configuration, with slightly better

    erformance, under some conditions. The triangle hole configura-

    ion provides lower heat transfer augmentation, compared to both

    he circle hole and racetrack hole configurations [24] . Parida et al.

    tudied the effect of wall-integrated inclined impingement jets in

    confined environment by numerical simulation and experimental

    ethod. The results show that by tilting the jet hole to a certain

    ngle, two asymmetric secondary flow regions can be formed in

    he channel under the interaction of jet and the wall. Due to the

    econdary flow region, the jets get accelerated and flow out at rel-

    tively higher velocity causing the heat transfer rates to increase

    ocally [25] .

    Compared with microchannel cooling technology and jet

    mpingement cooling technology, jet impingement/microchannel

    ooling technology was expected to have better heat transfer per-

    ormance due to integrating the advantages of two technologies.

    ung et al. used simulation and experimental method to study

    he cooling performance of JIMHS with HFE-7100 as working fluid.

    he results show that the heat conduction along the side wall of

    he microchannel is particularly significant in the single-phase re-

    ion at relatively low jet velocity, and the temperature gradient

    long the microchannel decreases with increasing the height of

    he microchannel at relatively low jet velocity [26] . In addition,

    ung et al. also studied the effects of jet sizes and jet array pat-

    erns on the heat transfer performance of JIMHS. They found that

    he decreasing-jet-size pattern yields the highest convective heat

    ransfer coefficients and lowest wall temperatures, while the equal-

    et-size pattern provides the greatest uniformity in wall temper-

    ture [17] . To further improve the heat transfer performance of

    IMHS, some scholars propose to introduce passive structures in

    he microchannels [27–33] . Husain et al. used numerical simula-

    ion to study the effect of adding pillar structure on heat trans-

    er performance. The results show that pillars in the channel con-

    ributed to enhanced heat transfer coefficient and higher heat

    ransfer coefficient can be achieved with smaller jet-to-jet pitch

    atio, while lower pressure drop can be achieved with larger jet-

    o-jet pitch ratio [30] . Huang et al. carried out a three-dimensional

    umerical simulation to study the heat transfer performance of

    IMHS by introducing the dimples to heat transfer surface. The

    imulation results indicate that JIMHS with convex dimples exhib-

    ted the best cooling performance. In addition, they further studied

    he effect of the positional distribution of the dimples on the heat

    ransfer performance and pressure drop [31–33] .

    According to above literature research, it can be noticed that

    or microchannel cooling technology, special microchannel struc-

    ure can enhance the heat transfer performance by increasing heat

    ransfer area and enhancing disturbance, but the disadvantages are

    hat there is often a large pressure gradient along the flow direc-

    ion, and the temperature distribution of the heat transfer surface

    s not uniform [ 13 , 17 ]. As for jet cooling technology, the advan-

    ages are that the extremely high convective heat transfer coeffi-

    ient can be obtained in stagnation zone and the better synergy

    f temperature field and velocity field [34–36] , but the disadvan-

    age is that the heat transfer coefficient will decrease sharply af-

    er leaving the stagnation zone. Therefore, many scholars propose

    o combine microchannel cooling technology with jet array cool-

    ng to complement each other and design the novel jet impinge-

    ent/microchannel heat sink to meet the needs of high heat flux

    issipation. However, the research in this field mainly focuses on

    umerical simulation of a symmetrical unit or a single channel of

    he heat sink [ 17 , 30 , 31 ], and the overall simulation is less. Accord-

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 1. (a) Assembly diagram of experimental testing system; (b) system loop physical map; (c) distribution of temperature measuring points.

    i

    p

    m

    j

    a

    i

    e

    2

    2

    e

    p

    s

    c

    a

    h

    t

    i

    m

    2

    2

    t

    T

    m

    d

    r

    p

    f

    o

    2

    t

    w

    i

    T

    h

    a

    t

    c

    T

    d

    a

    2

    p

    2

    2

    t

    ±i

    a

    w

    s

    t

    d

    ng to the numerical simulation results, the analysis of the whole

    erformance of heat sink and the essence of flow and heat transfer

    echanism is relatively less. In this work, two different submerged

    et impingement/microchannel heat sink models were proposed,

    nd the heat transfer and flow characteristic were systematically

    nvestigated from the aspect of overall numerical simulation and

    xperiment.

    . Experimental verification

    .1. Introduction of experimental system

    Fig. 1 shows the schematic diagram and physical map of the

    xperimental circuit with deionized water as cooling fluid. The ex-

    erimental system is mainly composed of two parts, one is the

    ystem components, and the other is the measuring and control

    omponents. The system components of the experimental circuit

    re liquid accumulator, micropump, jet impingement/microchannel

    eat sink, heat source, condensation device, valves and pipes be-

    ween the components. Measuring and control components mainly

    nclude electromagnetic flowmeter (EMF), pressure gauge, power

    eter, thermocouple and SSR (solid state relay).

    .2. Introduction of main system components

    .2.1. Jet impingement/microchannel heat sink

    As shown in the Fig. 2 , the novel JIMHS consists of three parts:

    he top jet orifice plate, the frame and the heat transfer rib plate.

    he material of the heat sink is cooper, which is processed by

    achine tools and EDM. Looking at Fig. 2 (a), unlike the three-

    imensional model shown in Fig. 6 , there are raised rectangular

    ibs above the plate, which is used to enhance the mechanical

    roperties of the jet orifice plate and prevent the plate from de-

    3

    orming under pressure. In addition, the dimensional parameters

    f the heat sink can be found in Fig. 6 and Table 2 .

    .2.2. Heat source

    Fig. 3 shows the physical map and three-dimensional perspec-

    ive view of the heat source. The material of heat source is copper,

    hich is cut and processed by machine tools. There are 16 process-

    ng holes at the bottom, corresponding to 16 heating rods inserted.

    he maximum power provided by each heating rod is 100 W. Four

    oles of different lengths with diameter of 1 mm were machined

    t the upper part near 5 mm of the contact surface for inserting

    hermocouples to measure the temperature of the heat source. The

    ontact area between heat source and heat sink is 40 × 20 mm ². he contact surface is smooth and coated with a layer of heat con-

    ucting paste. After that, the heat source and heat sink are tightly

    dhered and fixed with screws.

    .3. Introduction of measuring components

    The relevant parameters of measuring components and gear

    ump are shown in Table 1 .

    .4. Experimental data measurement

    .4.1. Temperature measurement and flow measurement

    In the experiment, the T-type thermocouples were used for

    emperature measurement, in which the measurement error is

    0.2 °C. A total of 10 temperature measuring points were arranged n the experiment. Four of the temperature measuring points are

    rranged around the heat sink and placed close to the heat sink

    all for temperature measurement and the average of the mea-

    ured values approximately replaces the average temperature of

    he heat transfer surface, like the T 5 and T 6 in Fig. 1 (c). In ad-

    ition, T and T are in the symmetrical position. The other four

    7 8

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 2. (a) Top jet orifice plate(front); (b) top jet orifice plate(back); (c) frame and heat transfer rib plate.

    Fig. 3. (a) Physical diagram of the heat source; (b) the three-dimensional perspec-

    tive diagram of the heat source.

    Table 1

    Specification parameters of system measuring components.

    Component name Specifications and parameters

    Gear pump Model number: MG317XK/DC24W

    Electromagnetic flowmeter Model number: AXG005G, Range: 0-3 L/min

    Pressure gage FUJI, Range: 30-3000 KPa, -0.1-0.2 MPa

    Power meter Model number: PF9800, accuracy class: 0.5

    Data collector Model number: Keithley-2700

    Direct-current main 24 V

    Voltage regulating transformer Model number: TDGC-0.5

    Fixed resistance 4 × 250 �Solid state relay Model number: GJH3-80A

    Thermocouple T-type thermocouple, ±0.2 °C

    t

    t

    i

    5

    m

    m

    o

    (

    h

    w

    2

    1

    1

    2

    v

    W

    2

    c

    n

    c

    Q

    w

    t

    t

    T

    w

    b

    a

    h

    i

    m

    fl

    [

    c

    (

    h

    t

    δ

    δ

    y

    y

    emperature measuring points are arranged at the upper part of

    he heating source close to the contact surface by 5 mm, as shown

    n Figs. 1 (c) and 3 . The depths of the four measuring points are

    , 10, 12 and 15 mm, respectively. The remaining two temperature

    easuring points are respectively arranged inside the liquid accu-

    ulator and at the outlet pipe wall for measuring the inlet and

    utlet temperatures of the heat sink.

    The flowmeter shown in Fig. 1 is an electromagnetic flowmeter

    EMF) used in this experiment to measure the overall flow into the

    eat sink. The measurement range of the flowmeter is 0-3 L/min,

    hich is connected in series to the whole circuit of the system.

    .5. Experimental conditions

    In the experiment, the inlet temperature was constant at

    0 ±0.4 °C. The testing volume flow rate was 0.5 L/min, 0.75 L/min, L/min, 1.25 L/min, 1.5 L/min, 1.75 L/min, 2 L/min, 2.25 L/min and

    .5 L/min, respectively. Nine different heat fluxes were test at each

    olume flow rate, and increased from the lowest 20 W / c m 2 to 100

    / c m 2 with a step increment of 10 W / c m 2 .

    4

    .6. Data reduction and uncertainty analysis

    Assuming that the physical properties of the deionized water

    irculating in the experimental system remain constant with the

    umber of tests, the average heat flux absorbed by the heat sink

    an be calculated by the heat absorption formula of water:

    = c p ˙ m ( T out − T in ) (1) here c p is the specific heat capacity of water, and T in and T out are

    he inlet and outlet temperature of the heat sink, respectively.

    The average temperature of the heat transfer surface can be es-

    imated by the following formula:

    = 1 4

    8 ∑ i =5

    T i (2)

    here T 5 ~T 8 are the external surface temperatures of the heat sink

    ottom.

    The average convective heat transfer coefficient ( h ) is defined

    s:

    = Q A 2 ( T − T f )

    (3)

    Among them, T f is the average temperature of the heat sink

    nlet and outlet and A 2 is the area of the heat transfer surface.

    As shown in Table 1 , the absolute error of the T-type ther-

    ocouple is ±0.2 °C, and the relative error of the electromagnetic owmeter is ±1% . According to the uncertainty transfer formula

    37] , the uncertainty of heat flux and average heat transfer coeffi-

    ient was estimated to be ±15 . 3% and ±15 . 87% at the low heat flux q = 20 W / c m 2 ). Similarly, the uncertainty of heat flux and average eat transfer coefficient was estimated to be ±5 . 75% and ±5 . 81% at he high heat flux ( q = 100 W / c m 2 ).

    δQ

    Q =

    √ √ √ √ ( δ �m �

    m

    ) 2 +

    (δy 1 y 1

    )2 (4)

    δh

    h =

    √ (δQ

    Q

    )2 +

    (δy 2 y 2

    )2 (5)

    y 1 = √

    ( δT in ) 2 + ( δT out ) 2 (6)

    y 2 = √

    8 ∑ i =5

    (1

    4 δT i

    )2 +

    (1

    2 δT in

    )2 +

    (1

    2 δT out

    )2 (7)

    1 = T out − T in (8)

    2 = 1 4

    8 ∑ T i −

    1

    2 ( T in + T out ) (9)

    i =5

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 4. (a) Curve of T with mass flow rate under different heat fluxes; (b) curve of h with mass flow rate under different heat fluxes.

    2

    c

    fl

    h

    h

    d

    t

    1

    s

    t

    r

    o

    t

    fl

    s

    r

    c

    t

    r

    m

    a

    o

    s

    f

    t

    o

    t

    u

    fl

    e

    w

    p

    [

    a

    f

    R

    N

    c

    t

    j

    Fig. 5. Comparison with the previous published works.

    Table 2

    Geometric parameters of SJIMHS and OJIMHS [mm].

    D in/out L L 1 L 2 W W 1 W 2 W 3 H H 1 D jet 2 42 2.85 5 22 3 2.5 0.5 5 1.5 0.5

    s

    s

    o

    n

    p

    3

    O

    h

    b

    c

    o

    fl

    t

    l

    3

    O

    e

    .7. Analysis and comparison of experiment results

    As shown in Fig. 4 (a), the average temperature of the thermo-

    ouples T 5 , T 6 , T 7 and T 8 varies with flow rate under different heat

    ux, which approximately replaces the average temperature of the

    eat transfer surface. It can be seen from the figure that when the

    eat flux is constant, the temperature of the heat transfer surface

    ecreases gradually with the increase of volume flow rate. In par-

    icular, when the volume flow rate is 0.5 L/min and the heat flux is

    00 W/cm ², the average temperature of the heat transfer surface is till lower than 60 °C, which is far lower than the critical tempera- ure of 85 °C for traditional electronics [38] . When the volume flow ate is small, changing the volume flow rate has a great impact

    n the average temperature of the heat transfer surface, but with

    he increase of volume flow rate, the effect of increasing volume

    ow rate on the average temperature of the heat transfer surface is

    maller. The reason for this phenomenon is when the volume flow

    ate increases gradually, the convective heat transfer resistance de-

    reases gradually until it is close to the thermal-conduction resis-

    ance of the heat transfer rib plate. When the thermal-conduction

    esistance dominates, the improvement of the heat transfer perfor-

    ance by further increasing the flow will be smaller.

    As shown in Fig. 4 (b), when the heat flux is constant, the aver-

    ge convective heat transfer coefficient increases with the increase

    f the volume flow rate, but when the volume flow rate is con-

    tant, the change of heat flux has little effect on the heat trans-

    er coefficient, which is consistent with study by Wu et al. [ 12 , 39 ]

    hat minor changes in cooling medium parameters due to changes

    f heat flow can be ignored. Therefore, the main factor affecting

    he heat transfer coefficient of heat sink is the volume flow rate

    nder the single-phase heat transfer state. The higher the volume

    ow rate of the inlet, the faster the velocity of the cooling medium

    jected from the jet hole, the thinner the boundary layer formed

    hen scouring the heat transfer surface.

    In addition, the experimental results in this paper are com-

    ared with the results of the jet array cooling in published paper

    27 , 28 , 39–44 ], as shown in Fig. 5 . The Reynolds number and the

    verage Nusselt number of the heat transfer surface are defined as

    ollows:

    e = ρu D jet μ

    (10)

    u = h D jet λ

    (11)

    In the above formula, ρ , μ, λ are the density, dynamic viscosity oefficient and thermal conductivity of the fluid respectively, u is

    he average velocity of the jet hole and D jet is the diameter of the

    et hole. Observing Fig. 5 , it can be found that the jet-to-surface

    5

    pace ratio (Zn), the type of cooling medium and heat transfer

    urface have a great influence on the heat transfer performance

    f heat sink. When the jet-to-surface space ratio is the same, we

    oticed that the average Nusselt number of heat sink with ribbed

    late is better than that with smooth plate.

    . Numerical model and calculation method

    According to the experimental results, it is verified that the

    JIMHS can achieve spectacular heat dissipation capacity with high

    eat flux. In order to better understand the interaction mechanism

    etween the flow and heat transfer of JIMHS, the overall numeri-

    al simulation is adopted. In the numerical simulation, the effects

    f heat flux, pressure drop and rib arrangement on the internal

    ow and heat transfer of the heat sink were studied. In addition,

    he working fluid and the heat sink material used in all the simu-

    ations are the same with the experiment.

    .1. Geometric model of SJIMHS and OJIMHS

    Fig. 6 shows the three-dimensional structure of SJIMHS and

    JIMHS, whilst Table 2 lists the main dimensions corresponding to

    ach part. It should be noted that all the orifices have a diameter

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig.6. Jet impingement/microchannel heat sink 3D model exploded view in (a) and 3D model diagram of each part in (b-1), (b-2) and (b-3).

    o

    a

    a

    o

    2

    c

    t

    3

    t

    t

    p

    d

    b

    i

    f

    c

    w

    fi

    p

    p

    fl

    t

    u

    p

    c

    i

    η

    S

    C

    p

    o

    f

    t

    t

    f

    d

    f

    W

    t

    t

    e

    a

    3

    h

    I

    p

    w

    t

    c

    i

    t

    1

    o

    v

    b

    f 0.5 mm, which are arranged in a triangular staggered manner,

    nd the jet to target space is 2 mm. For oblique-rib plate, the size

    nd number of ribs are the same as that of straight-rib plate, it

    nly has a different angle between ribs and horizontal direction of

    7 °. In addition, considering the effect of the outlet reflux on the alculation results, the final fluid outlet length is extended to 6-8

    imes the outlet diameter to eliminate the backflow phenomenon.

    .2. Numerical calculation model and boundary conditions

    According to literature investigation, the two-equation k − εurbulence model is widely used in the numerical simulation of

    he jet impingement/microchannel heat sink and verified by ex-

    eriment [ 17 , 31 ]. Considering the interaction between jets and the

    isturbance of the fluid by the ribs, many vortex structures will

    e formed in the heat sink. The RNG k-epsilon model, which is an

    mprovement of the standard k-epsilon model, shows superior per-

    ormance in predicting the flow features include strong streamline

    urvature, vortices and rotation. Therefore, the RNG k − ε model ith moderate computational economy and acceptable accuracy is

    nally selected for all the numerical simulations. In order to sim-

    lify the calculation, the flow was assumed to be steady, incom-

    ressible and continuous.

    In this work, the commercial software Fluent 16.0 is used as

    ow solver and the governing equations of incompressible flow in

    he jet impingement/ microchannel heat sink are as follows:

    Continuity equation:

    ∂ ( ρu i )

    ∂ x i = 0 (12)

    Momentum conservation equation:

    ∂(ρu i u j )

    ∂ x j = − ∂ p

    ∂ x i + ∂

    ∂ x j

    ∂ u i ∂ x j

    − ρu ′ i u ′

    j

    )(13)

    Energy conservation equation:

    ∂ ( u i T )

    ∂ x j = ∂

    ∂ x j

    ρc p

    ∂T

    ∂ x j

    )(14)

    Kinetic energy equation:

    ∂ x i (ρk u j ) =

    ∂ x j

    [(η + ηt

    σk

    )∂k

    ∂ x j

    ]+ 2 ηt ∂ u i

    ∂ x j S i, j − ρε (15)

    Dissipation rate equation:

    ∂ x i (ρk u j ) =

    ∂ x i

    [(η + ηt

    σε

    )∂ε

    ∂ x i

    ]+ 2 C 1 ε ε

    k ηt

    ∂ u i ∂ x j

    S i, j − C 2 ε ρε 2

    k

    6

    (16)

    In the above formula, ρ is the liquid density, p is the pressure, i is the speeds in the x, y, and z directions, c p is the constant

    ressure specific heat capacity, and ηt is the eddy viscosity coeffi- ient. S i, j , ηt and C 1 ε is given by the following equation, where C 2 ε s 1.68 and C μ is 0.085.

    t = ρC μ k ε

    (17)

    i, j = 1

    2

    (∂ u i ∂ x j

    + ∂ u j ∂ x i

    )(18)

    1 ε = 1 . 42 − η(1 − η/ 4 . 38) 1 + 0 . 015 η3

    (19)

    As for the boundary conditions, the inlet boundary is set as

    ressure inlet with a temperature of 283 K. The outlet is pressure

    utlet, and the relative pressure is 0 Pa. The bottom heating sur-

    ace is set to a constant heat flux, which increases from 40 W/cm ²o 100 W/cm ² with a step increment of 10 W/cm ². Only the heat ransfer is considered in the solid domain, the solid-liquid inter-

    ace is set as no slip, and the rest of the walls are adiabatic.

    The pressure-velocity coupling adopts SIMPLE method, and the

    iscretization scheme is set as the standard for the pressure, QUICK

    or energy, and second-order upwind for the remaining period.

    hen the residuals of continuum equation and momentum equa-

    ion reach 10 −5 , the convergence residuals of turbulence dissipa- ion term reach 10 −4 , and the convergence residuals of energy quation are less than 10 −7 , the convergence is considered as chieved.

    .3. Mesh processing and verification of grid independence

    The overall mesh of the heat sink is shown in Fig. 8 (a) and the

    exahedral mesh of each part is generated by commercial software

    CEM 16.0. Due to the complexity of the whole model, different

    arts of the model are partitioned into blocks. Since the enhanced

    all treatment function was used to calculate the near wall region,

    he y + value is required to be around 1, and the maximum value annot exceed 5. To ensure that the y + value near the wall region s around 1 under the maximum flow condition, the grid height of

    he first layer is set to 0.01mm, and the grid growth factor is set to

    .3. Fig. 9 shows the grid of fluid region inside the heat sink with-

    ut outlet extension. By slicing the grid longitudinally, the zoom-in

    iews of grid in the near wall region and the jet hole region can

    e obtained.

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 7. Comparison of laterally-averaged wall temperature of the heat transfer surface for three different mesh densities along streamwise direction (a) SJIMHS, (b) OJIMHS.

    Fig. 8. The overall mesh of the heat sink in (a) and precise regional mesh information of different part in (b) and (c).

    Fig. 9. The zoom-in views of grid in the near wall region.

    o

    m

    F

    o

    h

    i

    t

    t

    r

    t

    t

    d

    e

    a

    o

    t

    t

    r

    c

    In the process of grid independence verification, the number

    f grids is changed by controlling the maximum grid size while

    aintaining the same refined conditions of the near-wall region.

    ig. 7 shows the comparison of laterally-averaged wall temperature

    f the heat transfer surface for three different grid numbers of two

    eat sink models. The change is observed to be less 1% by increas-

    ng the grid number between the models for N 1 and N 2 . In addi-

    ion, the Grid Convergence Index (GCI) method was used to test

    he grid independence [45] . As shown in Table 3 , three successive

    efined grids for each heat sink model were conducted to ensure

    7

    he accuracy of the numerical solutions. The refinement ratio be-

    ween meshes N 2 and N 1 as well as between meshes N 3 and N 2 are

    efined by r 21 and r 32 , respectively. The solution quantity of the av-

    rage temperature of heat transfer surface are presented by φ1 , φ2 nd φ3 for each respective mesh. In addition, p is the observed rder of convergence between the studied meshes. According to

    he calculation formula of GCI [46] , the Grid Convergence Index for

    he N 2 mesh of each model ( GCI 21 fine ) were calculated. Based on the

    esults of this Grid-independent test and weighing computational

    osts, the grid of SJIMHS with 8,285,374 elements was adopted for

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Table 3

    Grid Convergence Index (GCI) verification.

    φ = T [K], SJIMHS φ = T [K], OJIMHS N 1 , N 2 , N 3 20,106,763; 8,285,374; 4,565,147 29,536,274; 14,849,951; 6,675,643

    r 21 1.34 1.26

    r 32 1.22 1.31

    φ1 308.98 303.98

    φ2 309.94 304.93

    φ3 310.91 306.41

    p 1.67 1.18

    GCI 21 fine 0.61% 1.26%

    Fig. 10. (a) Comparison of simulated and experimental values of T and h under different heat flux;(b) comparison of simulated and experimental values of T and h under

    different mass flow rate.

    s

    s

    3

    c

    c

    u

    a

    e

    F

    t

    c

    t

    w

    s

    a

    h

    m

    u

    t

    i

    v

    e

    b

    c

    t

    4

    4

    t

    l

    t

    f

    S

    s

    T

    t

    a

    m

    p

    h

    f

    i

    r

    fl

    c

    a

    w

    t

    i

    fl

    l

    o

    i

    n

    w

    a

    t

    c

    f

    h

    w

    a

    h

    t

    s

    t

    v

    e

    ubsequent simulations. Similarly, the grid number of 14,849,951 is

    elected to calculate OJIMHS.

    .4. Comparisons between experimental results and simulation results

    As shown in Fig. 10 (a), the average temperature and average

    onvective heat transfer coefficient of the heat transfer surface cal-

    ulated by simulation is compared with the experimental results

    nder the same inlet mass flow rate. It can be seen that the vari-

    tion trend of the simulation results is the same as that of the

    xperimental results and there is little difference between them.

    urthermore, the maximum relative error of the average tempera-

    ure is only 0.33% and the maximum relative error of the average

    onvective heat transfer coefficient is only 2.1%.

    In order to further verify the reliability of the simulation results,

    he experimental and simulated results at different mass flow rate

    ere compared. The heat flux used in the cases of experiment and

    imulation is 100 W / c m 2 . As shown in Fig. 10 (b), the curves of the

    verage temperature and convective heat transfer coefficient of the

    eat transfer surface measured in the experiment with different

    ass flow rate are basically the same as that measured in the sim-

    lation. Besides, the maximum relative error of the average heat

    ransfer surface temperature between experiment and simulation

    s about 1% and the maximum relative error of the average con-

    ective heat transfer coefficient is about 9%, which are within the

    rror tolerance. Therefore, it can be proved that the RNG k − ε tur- ulence model selected in the calculation has high accuracy in cal-

    ulating the jet impingement/microchannel heat sink models and

    he simulation results have high reliability.

    . Analysis of numerical simulation results

    .1. Comparison of temperature fields between SJIMHS and OJIMHS

    As shown in Fig. 11 , the temperature distribution diagrams of

    he heat transfer surface under different heat flux and the same in-

    et pressure are obtained by numerical simulation. It is noticeable

    hat under the same heat flux condition, the bottom heat trans-

    er surface temperature of OJIMHS is generally lower than that of

    8

    JIMHS. The detailed temperature parameters of the heat transfer

    urface of SJIMHS and OJIMHS are depicted in Table 4 . T max and

    min represent the maximum and minimum temperature respec-

    ively. T is the average temperature of the heat transfer surface

    nd �T is the difference between the maximum and the mini-

    um temperature. Observing Table 4 , it can be found that each

    arameter of OJIMHS is lower than that of SJIMHS under the same

    eat flux condition. In view of the difference of heat transfer per-

    ormance, we speculated that the inclined arrangement of ribs will

    ntensify the turbulence of the fluid, which is beneficial to inter-

    upt the boundary layer and accelerate the mixing of cold and hot

    uids, thus enhancing the convective heat transfer between the

    ooling medium and the channel surface, and lowering the temper-

    ture of the heat transfer surface. As shown in Fig. 12 , k is volume

    eighted average of turbulent kinetic energy, which can reflect the

    urbulence intensity of fluid. It can be found that the k of OJIMHS

    s obvious higher than that of SJIMHS under all the calculated heat

    ux when the inlet pressure is constant.

    Moreover, as shown in Fig. 11 , the temperature is generally

    ower near the upper of the picture, and higher near the bottom

    f the picture, especially in SJIMHS. The upper part of the picture

    s near the heat sink inlet and the lower part of the picture is

    ear the heat sink outlet. The more details can be found in Fig. 13 ,

    hich shows the local averaged heat transfer characteristics. T w,a v nd T f,a v are the laterally-averaged wall temperature of the heat

    ransfer surface and laterally-averaged fluid temperature in the mi-

    rochannel, respectively. As for the local area-averaged heat trans-

    er coefficient, the formula for calculating it is defined as:

    a v = q a v T w,a v − T f

    (20)

    here q a v is the averaged heat flux of the heat transfer surface

    nd T f is the average temperature of the inlet and outlet of the

    eat sink. The local average zone is in the middle of the rib and

    he adjacent two rows of ribs, as shown in Fig. 13 . For the two heat

    ink models, the fluid and heat transfer surface local area-averaged

    emperature at the heat sink inlet are relatively low, and the con-

    ective heat transfer coefficient reaches the maximum value. How-

    ver, the heat transfer surface temperature and fluid temperature

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 11. Comparison of temperature distribution diagrams at different heat flux.

    Table 4

    Numerical simulation result of properties of SJIMHS and OJIMHS.

    Heat

    flux/(W •cm −2 ) SJIMHS OJIMHS

    T max / (K) T min / (K) T / (K) �T / (K) T max / (K) T min / (K) T / (K) �T / (K)

    40 295.42 290.27 293.21 5.15 293.94 288.93 291.63 5.01

    50 298.52 292.09 295.77 6.43 296.68 290.41 293.79 6.27

    60 301.63 293.90 298.32 7.72 299.41 291.89 295.95 7.52

    70 304.73 295.72 300.87 9.01 302.15 293.37 298.11 8.78

    80 307.84 297.54 303.43 10.30 304.86 294.84 300.25 10.02

    90 310.97 299.37 306.01 11.60 307.61 296.33 302.43 11.28

    100 314.02 301.16 308.52 12.86 310.36 297.81 304.58 12.55

    Fig. 12. (a) The curve of k under different heat fluxes; (b) the curve of k under different inlet pressure.

    Fig. 13. Laterally-averaged wall temperature, fluid temperature and heat transfer coefficient of SJIMHS and OJIMHS along streamwise direction ( q = 100 W ·cm -2 , P in = 50 0 0 0 Pa).

    9

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 14. (a) T and h of the cooling surface under different heat fluxes( P in = 50 0 0 0 Pa ); (b) T and h of the cooling surface under different inlet pressure( q = 100 W · c m −2 ).

    g

    m

    t

    b

    s

    r

    t

    i

    c

    s

    w

    f

    b

    o

    c

    a

    j

    s

    g

    o

    z

    c

    m

    t

    s

    c

    c

    f

    s

    a

    g

    i

    f

    t

    t

    w

    t

    a

    r

    w

    t

    l

    r

    t

    s

    c

    I

    f

    s

    f

    h

    w

    a

    e

    fi

    o

    a

    f

    t

    S

    O

    h

    c

    i

    i

    f

    f

    a

    t

    fi

    r

    H

    p

    m

    c

    t

    t

    t

    fl

    t

    f

    c

    b

    c

    F

    u

    4

    e

    v

    m

    radually increase along the flow direction and reach the maxi-

    um value at the outlet. The main reason for this difference is

    he “drift phenomenon” caused by horizontal crossflow, which can

    e seen clearly in Fig. 16 . When the horizontal crossflow from up-

    tream flows in the channel, the downstream vertical jet cannot di-

    ectly wash down the cooling surface and drift toward the horizon-

    al flow direction, which weakens the heat transfer performance

    n the downstream jet area and increases the temperature of the

    ooling surface [ 10 , 31 ]. However, the temperature of heat transfer

    urface is decreasing at the edge of the surface, as shown in Fig. 11 ,

    hich is agreed with the study by Wu et al. [39] that the influence

    rom the adjacent wall enhance the heat transfer.

    In comparison, it can be noticed that the temperature distri-

    ution of the OJIMHS is much more uniform and there are fewer

    verheated areas near the lower part of the picture. The reason

    ould be inferred that the enhancement of turbulence is capable to

    chieve more sufficient heat transfer between fluids in the whole

    et passage, which makes the temperature distribution of fluids in-

    ide the heat sink more uniform and forms a higher temperature

    radient with the heat transfer surface [47] , then avoids the local

    verheating phenomenon and reduces the adverse effect of hori-

    ontal crossflow to some extent. As shown in Fig. 13 , it can be seen

    learly that the local area-averaged fluid temperature of OJIMHS is

    uch more uniform in the middle part of the heat sink. In addi-

    ion, the local area-averaged heat transfer coefficient of OJIMHS is

    ignificantly higher than that of SJIMHS at all measuring zones. In

    onsequence, compared with the straight ribs, the oblique ribs can

    ause better uniform flow field temperature to enhance heat trans-

    er performance.

    By analyzing the local heat transfer performance of the heat

    ink in Fig. 13 , it is noticeable that the heat transfer performance

    t the rib region is observably better than that at the rib free re-

    ion for SJIMHS and OJIMHS. Especially for OJIMHS, the difference

    s most obvious, which is reflected in the fluctuation of heat trans-

    er surface temperature. It is speculated that two reasons are con-

    ributed to this difference. Firstly, the existence of ribs increases

    he heat transfer area, and the fluid will take away more heat

    hen it passes through the rib region. Secondly, the ribs will break

    he boundary layer and restrain the development of the bound-

    ry layer, thus enhancing the heat transfer. Compared with straight

    ibs, the direction of the fluid flow forms a certain inclination angle

    ith the oblique ribs, so that the oblique ribs are more effectively

    han straight ribs in inhibiting the development of the boundary

    ayer. In the flow field analysis in Section 4.2 , the effect of rib ar-

    angement on the flow field will be discussed in detail.

    As shown in Fig. 14 (a), the average temperature of the heat

    ransfer surface of OJIMHS is lower than that of SJIMHS under

    even different heat fluxes, and the temperature difference be-

    omes more and more obvious with the increase of heat flux.

    n addition, with the increase of heat flux, the heat transfer sur-

    t

    10

    ace average temperature of the two heat sinks increases gradually,

    howing a linear change. As for the average convective heat trans-

    er coefficient, the formula for calculating it is as follows:

    = q A 1 A 2

    (T w − T f

    ) (21) here, q is the heat flux of the bottom heating surface, A 1 is the

    rea of the bottom heating surface( 40 × 20m m 2 ), and T w is the av- rage temperature of the heat transfer surface. As indicated in the

    gure, the average convective heat transfer coefficient of OJIMHS is

    bviously higher than that of SJIMHS under seven heat flux, with

    n increase of about 20%. But the change of heat flux has little ef-

    ect on the convective heat transfer coefficient for two models.

    Fig. 14 (b) shows the average temperature and convective heat

    ransfer coefficient of the heat transfer surface of OJIMHS and

    JIMHS with the same heat flux as a function of inlet pressure.

    bserving the figure, it is obvious that changing the inlet pressure

    as a great influence on the average temperature and the average

    onvective heat transfer coefficient. The higher the inlet pressure

    s, and the higher the average convective heat transfer coefficient

    s, the lower the average temperature is. In addition, under six dif-

    erent inlet pressures, the average temperature of the heat trans-

    er surface of OJIMHS is lower than that of SJIMHS, and the aver-

    ge convective heat transfer coefficient of OJIMHS is higher than

    hat of SJIMHS. Comparing the change of mass flow rate, we can

    nd that when the inlet pressure is small, the inlet mass flow

    ate of two heat sinks is not much different, basically the same.

    owever, when the inlet pressure is larger, the mass flow rate dis-

    arity between the two heat sinks gradually increases, and the

    ass flow rate of SJIMHS is slightly higher than that of OJIMHS. It

    ould be speculated that the oblique ribs exacerbate the fluid dis-

    urbance, which causes an increase in the flow resistance. When

    he inlet pressure is small, the velocity of the fluid is slow and

    he disturbance caused by oblique ribs is relatively weak, so the

    ow resistance caused by the disturbance is relatively small and

    he inlet mass flow rate of the two heat sinks is not much dif-

    erent. However, when the inlet pressure is large, the disturbance

    aused by oblique ribs is intensified, so the flow resistance caused

    y disturbance is relatively large and the mass flow rate of the

    ooling medium flowing through OJIMHS is lowered. As shown in

    ig. 12 (b), the difference of k between SJIMHS and OJIMHS is grad-

    ally increasing with the increase of inlet pressure.

    .2. Analysis of velocity field

    In Section 4.1 , the temperature distribution of the two mod-

    ls under different inlet pressures and heat fluxes is analyzed. In

    iew of the heat transfer performance difference between the two

    odels, we propose that the difference of temperature distribu-

    ion is mainly caused by the fluid disturbance. In order to better

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 15. Relative position of slices in JIMHS.

    Fig. 16. The comparison of A-type velocity gradient diagram and streamline diagram between OJIMHS and SJIMHS: (a) q = 100 W ·cm -2 , P in = 50000 Pa, SJIMHS; (b) q = 100 W ·cm -2 , P in = 50 0 0 0 Pa, OJIMHS.

    u

    f

    w

    o

    p

    p

    4

    t

    j

    z

    [

    p

    d

    t

    f

    h

    s

    a

    i

    j

    I

    f

    fl

    b

    i

    t

    S

    M

    f

    s

    i

    m

    b

    T

    t

    n

    l

    s

    t

    l

    nderstand the interaction mechanism between heat transfer per-

    ormance and fluid disturbance, the velocity fields of two models

    ere analyzed. Fig. 15 shows the relative position of the two slices

    f SJIMHS and OJIMHS, wherein the lateral slice located in the XZ

    lane is named slice A, and the longitudinal slice located in the ZY

    lane is named slice B.

    .2.1. Comparison of flow fields between SJIMHS and OJIMHS

    Observing Fig. 16 , it can be seen for two models, the closer to

    he inlet of the heat sink, the smaller the entrance velocity of the

    et hole, the less obvious the “drift phenomenon” caused by hori-

    ontal crossflow. This is consistent with the study by Huang et al.

    31] that the jet velocity near the outlet is larger and the “drift

    henomenon” is more obvious. The main reason for the velocity

    ifference is the different length of the path from the jet hole to

    he outlet of the heat sink. When the fluid enters the heat sink

    rom the jet hole near the inlet, the length of the path from the jet

    ole to the outlet is longer than that of the jet hole near the outlet,

    o the pressure loss caused by the on-way resistance is larger. As

    result, at the same inlet pressure, the greater the flow resistance

    s, the smaller the jet velocity will be when it passes through the

    et hole. This was verified by Cui et al. [10] in visual experiment.

    11

    n single-phase heat transfer regime, they found the crossflow ef-

    ect was quite small in experiments due to the short path in cross-

    ow direction. Consequently, they propose that measures should

    e taken to diminish the adverse effect of crossflow when design-

    ng the cooling applications for large areas.

    It should be noticed from Fig. 16 that the fluid velocity near

    he bottom heat transfer surface of OJIMHS is higher than that of

    JIMHS, which is more evident near the outlet of the heat sink.

    eanwhile, comparing the velocity gradient diagrams, it can be

    ound that for OJIMHS, the fluid disturbance inside the heat sink is

    tronger than that in SJIMHS, which is manifested that the change

    n velocity gradient is more obvious, especially near the left and

    iddle parts of the heat sink. Higher velocity means stronger tur-

    ulence, which leads to higher convective heat transfer coefficient.

    his explains why the heat transfer surface temperature distribu-

    ion of OJIMHS is more uniform than that of SJIMHS, and there is

    o local overheating area near the outlet of the heat sink.

    Fig. 17 shows the B-type velocity gradient diagram and stream-

    ine diagram comparison between OJIMHS and SJIMHS. It can be

    een clearly that for SJIMHS, a pair of symmetrical vortex struc-

    ures formed by the jet near the heat transfer surface directly be-

    ow each jet hole. As for OJIMHS, the vortex structures formed un-

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 17. The comparison of B-type velocity gradient diagram and streamline diagram between OJIMHS and SJIMHS: (a) q = 100 W ·cm -2 , P in = 50000 Pa, SJIMHS; (b) q = 100 W ·cm -2 , P in = 50 0 0 0 Pa, OJIMHS.

    Fig. 18. Comparison of the isosurface vortex core region between OJIMHS and SJIMHS.

    d

    s

    n

    fl

    v

    s

    t

    r

    fl

    fl

    t

    t

    e

    i

    t

    O

    a

    l

    s

    e

    O

    c

    b

    a

    e

    i

    4

    o

    t

    o

    l

    t

    t

    l

    j

    s

    a

    n

    l

    i

    g

    s

    er the jet hole are completely asymmetrical, and one of the vortex

    tructures is obviously shifted toward the upper part of the chan-

    el. Fig. 18 shows the distribution of isosurface vortex cores in the

    ow field of the two heat sink models. It can be found that the

    ortex structure in SJIMHS is more symmetrical, while the vortex

    tructure in OJIMHS obviously deflects to one side, especially near

    he outlet of heat sink. We speculated that the inclination of the

    ibs makes the flow direction in the channel inclined to the main

    ow direction. Under the joint action of the main stream and the

    ow in the channel, the vortex structure formed by the jet deviates

    o one side. The closer to the heat sink outlet, the more significant

    he influence of horizontal crossflow. Therefore, the mutual influ-

    nce between the main stream and flow in the channel is more

    ntense.

    In addition, due to the direction of the main flow forms a cer-

    ain inclination angle with the ribs, the internal disturbance of

    JIMHS is stronger and the interaction between the jet holes is

    lso more significant. We can see clearly in Fig. 17 that the ve-

    ocity in the gap between ribs and the upper cover of SJIMHS is

    maller than that of OJIMHS, which verifies that the mutual influ-

    nce between the adjacent jet holes of SJIMHS is inferior to that of

    JIMHS.

    Owing to the interaction between the jet holes, the mixing of

    old and hot fluids is more sufficient, thus the temperature distri-

    g

    12

    ution inside the heat sink is more uniform and a higher temper-

    ture gradient with the heat transfer surface is formed, which will

    nhance the heat transfer and avoid the occurrence of overheating

    n the bottom heat transfer surface.

    .2.2. Influence of pressure drop and heat flux

    Since the change laws of changing inlet pressure and heat flux

    n the flow field of two heat sinks are basically the same, only

    he flow field of OJIMHS is selected for analysis. By analyzing the

    verall change of Fig. 19 , it can be seen clearly that as the in-

    et pressure increases, the velocity of the fluid increases and the

    urbulence becomes more and more obvious, which is reflected in

    he change of the velocity gradient level. In addition, as the in-

    et pressure continues to increase, the velocity unevenness of the

    et holes becomes more and more obvious. This is agree with the

    tudy by Husain et al. [30] that the velocity distribution of jet holes

    long the crossflow direction is uneven, especially at the large Re

    umber. The main causes of this phenomenon are different path

    engths and the increasing turbulence intensity as inlet pressure

    ncreases, which were discussed in Section 4.2.1 .

    As shown in the Fig. 20 , the A-type velocity gradient dia-

    rams of OJIMHS are obtained by numerical simulation under the

    ame inlet pressure and different heat flux. Observing the dia-

    ram, changing heat flux has little effect on the internal flow of

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 19. A-type velocity gradient diagrams of OJIMHS under different pressure drops: (a) q = 100 W ·cm -2 , P in = 10 0 0 0 Pa; (b) q = 10 0 W ·cm -2 , P in = 30 0 0 0 Pa; (c) q = 100 W ·cm -2 , P in = 50 0 0 0 Pa.

    Fig. 20. A-type velocity gradient diagrams of OJIMHS under different heating heat flux: (a) q = 40 W ·cm -2 , P in = 50 0 0 0 Pa; (b) q = 100 W ·cm -2 , P in = 50 0 0 0 Pa.

    t

    t

    h

    r

    s

    a

    s

    m

    m

    m

    t

    c

    c

    d

    5

    w

    i

    f

    (

    (

    (

    he heat sink. According to previous literature research, there were

    hree states of flow in microchannel heat sink with the change of

    eat flux, namely, single-phase flow, subcooled boiling and satu-

    ated boiling. In the case of single-phase convection heat transfer

    tate, the flow and heat transfer characteristics of the heat sink

    re mainly related to the flow rate. The faster the flow is, the

    tronger the disturbance is, the better the heat transfer perfor-

    ance is [ 4 8 , 4 9 ]. Moreover, the flow rate of the internal fluid is

    ainly related to the pressure drop between inlet and outlet.

    For the case simulated in this work, when the heat flux is the

    aximum, q = 100 W / c m 2 , the maximum temperature of the heat ransfer surface does not exceed 340K, which is in the single-phase

    onvection heat transfer state. Therefore, the flow and heat transfer

    haracteristics of the heat sink are mainly related to the pressure

    rop.

    . Conclusions

    Two kinds of jet impingement/microchannel heat sink models

    ere proposed in this work, which were studied by overall numer-

    cal calculation and experiment. On the basis of the results, the

    ollowing conclusions are drawn:

    1) Under the same inlet pressure and heat flux, the cooling per-

    formance of OJIMHS is better than that of SJIMHS, which is re-

    13

    flected in that the temperature distribution of OJIMHS is more

    uniform and lower than that of SJIMHS. In addition, the average

    convective heat transfer coefficient of the OJIMHS is obviously

    higher than that of SJIMHS, with an increase of about 20%.

    2) According to the analysis of temperature field and velocity field,

    adding oblique ribs at the bottom heat transfer surface can

    cause better fluid disturbance. The increase of turbulence is ca-

    pable to interrupt the boundary layer development and acceler-

    ate the mixing of cold and hot fluids, which makes the temper-

    ature distribution inside the heat sink more uniform and forms

    a higher temperature gradient with the heat transfer surface,

    then enhance the heat transfer between the fluid and the wall,

    avoid the local overheating phenomenon and reduce the ad-

    verse effect of horizontal crossflow on the heat transfer perfor-

    mance to some extent.

    3) The experimental results are in good agreement with simula-

    tion. Comparing the simulated and experimental vales of h and

    T under different heat flux, the maximum relative error of T

    between the simulated values and the experimental values is

    only 0.33%, and the maximum relative error of h is only 2.1%.

    In addition, when exploring the influence of mass flow rate

    on heat transfer performance, the maximum relative error of

    h between the simulated values and the experimental values is

    9%, and the maximum relative error of T is 1%. When the vol-

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    Fig. 21. (a) T of the cooling surface under different heat fluxes( P in = 50 0 0 0 Pa ); (b) comparison of wetted areas of three heat transfer plates; (c) h of the cooling surface under different heat fluxes( P in = 50 0 0 0 Pa ).

    D

    b

    a

    C

    F

    d

    D

    w

    m

    P

    A

    d

    A

    C

    r

    h

    m

    b

    s

    t

    i

    m

    s

    t

    f

    i

    h

    i

    t

    h

    s

    r

    p

    s

    t

    a

    t

    h

    b

    p

    o

    v

    S

    c

    a

    h

    f

    s

    r

    h

    c

    o

    c

    d

    R

    ume flow rate is 0.5 L/min and the heat flux is 100 W/cm ², theaverage temperature of the heat transfer surface is still lower

    than 60 °C. Besides, the averaged heat transfer coefficient of 2.8 W / ( cm 2 · K) was achieved under the inlet fluid temperature of 283K and volume flow rate of 2.5 L/min in the experiment.

    eclaration of Competing Interest

    The authors declared that there is no conflict of interest. To the

    est of our knowledge and belief, neither I nor any coauthors have

    ny possible conflicts of interest.

    RediT authorship contribution statement

    H.C. Cui: Conceptualization, Methodology, Software, Validation,

    ormal analysis, Investigation, Data curation, Writing - original

    raft, Writing - review & editing. X.T. Lai: Methodology, Software,

    ata curation. J.F. Wu: Software, Data curation. M.Z. Wang: Soft-

    are, Data curation. W. Liu: Resources, Supervision, Project ad-

    inistration, Funding acquisition. Z.C. Liu: Resources, Supervision,

    roject administration, Funding acquisition.

    cknowledgment

    This work was supported by the National Natural Science Foun-

    ation of China (grant no. 51736004 ).

    ppendix

    omparison of heat transfer performance between smooth plate and

    ibbed plate

    To show the effect of the increase of heat transfer area on the

    eat transfer performance, we calculated the heat transfer perfor-

    ance of heat sink with smooth plate under some conditions. The

    asic size parameters of the heat sink with smooth plate is the

    ame as SJIMHS and OJIMHS except that there is no rib array on

    he heat transfer surface. The size parameters have been shown

    n Fig. 6 and Table 2 . The same meshing strategy and turbulence

    odel are used to perform numerical simulation on heat sink with

    mooth plate. Fig. 21 (a) shows the average temperature of the heat

    ransfer surface varies with heat flux on the bottom heating sur-

    ace. It can be found that the average temperature of smooth plate

    s significantly higher than that of SJIMHS and OJIMHS. When the

    eat flux is 100 W · c m −2 , the average temperature of smooth plate s about 20 K higher than that of OJIMHS or SJIMHS, which shows

    he heat transfer performance of heat sink can be significantly en-

    anced by introducing ribs on the heat transfer surface. Fig. 21 (b)

    hows the wetted area of three different types of heat transfer

    ibbed plate. The wetted area of oblique-rib plate and straight-rib

    late is the same, which is 1.875 times of smooth plate. Fig. 21 (c)

    14

    hows the average convective heat transfer coefficient of the heat

    ransfer surface under different heat flux. It can be seen that the

    verage convective heat transfer coefficient of OJIMHS is higher

    han that of smooth plate, and that of smooth plate is slightly

    igher than that of SJIMHS, which is similar with the study by Tra-

    old et al [27] that in some cases, the Nusselt number of smooth

    late is higher than that of ribbed rough plate. Through the previ-

    us analysis, we have explained the reason why the average con-

    ective heat transfer coefficient of OJIMHS is higher than that of

    JIMHS. As for the situation that the average heat transfer coeffi-

    ient of smooth plate is higher than that of straight-rib plate, we

    nalyze the main reasons as follows. The introduction of ribs en-

    ances the heat transfer performance by increasing the heat trans-

    er area and enhancing the fluid disturbance. However, due to the

    ame direction of rib arrangement and fluid flow in SJIMHS, the

    ibs have little disturbance to the fluid and the improvement of

    eat transfer performance of heat sink is mainly realized by in-

    reasing the heat transfer area. On the other hand, the increase

    f heat transfer area leads to an increase of flow resistance in the

    hannel, and the flow velocity decreases with the same pressure

    rop, resulting in the decrease of heat transfer coefficient.

    eferences

    [1] X.B. Luo , R. Hu , S. Liu , K. Wang , Heat and fluid flow in high-power LED pack-

    aging and applications, Prog. Energy Combust. Sci. 56 (2016) 1–32 . [2] S.M.S. Murshed , C.A.N. de Castro , A critical review of traditional and emerging

    techniques and fluids for electronics cooling, Renew. Sustain. Energy Rev. 78 (2017) 821–833 .

    [3] R.C. Chu , R.E. Simons , M.J. Ellsworth , R.R. Schmidt , V. Cozzolino , Review of

    cooling technologies for computer products, IEEE Trans. Device Mater. Reliab. 4 (4) (2004) 568–585 .

    [4] N.S. Kim , T. Austin , D. Baauw , T. Mudge , K. Flautner , J.S. Hu , M.J. Irwin , M. Kan-demir , V. Narayanan , Leakage current: Moore’s law meets static power, Com-

    puter 36 (12) (2003) 68–75 . [5] T. Anderson , I. Mudawar , Microelectronic cooling by enhanced pool boiling of

    a dielectric fluorocarbon liquid, ASME J. Heat Transfer 111 (3) (1989) 752–759 .

    [6] C. Leng , X.D. Wang , T.H. Wang , W.M. Yan , Optimization of thermal resistanceand bottom wall temperature uniformity for double-layered microchannel heat

    sink, Energy Convers. Manage. 93 (2015) 141–150 . [7] I. Mudawar , Assessment of high-heat-flux thermal management schemes, IEEE

    Trans. Compon. Packag. Technol. 24 (2) (2001) 122–141 . [8] M.A. Ebadian , C.X. Lin , A review of high-heat-flux heat removal technologies,

    ASME J. Heat Transfer 133 (11) (2011) 110801 .

    [9] F.J. Hong , C.Y. Zhang , W. He , P. Cheng , G. Chen , Confined jet array impingementboiling of subcooled aqueous ethylene glycol solution, Int. Commun. Heat Mass

    Transf. 56 (2014) 165–173 . [10] F.L. Cui , F.J. Hong , P. Cheng , Comparison of normal and distributed jet array

    impingement boiling of HFE-70 0 0 on smooth and pin-fin surfaces, Int. J. Heat Mass Transf. 126 (2018) 1287–1298 .

    [11] A.J. Robinson , R. Kempers , J. Colenbrander , N. Bushnell , R. Chen , A single phasehybrid micro heat sink using impinging micro-jet arrays and microchannels,

    Appl. Therm. Eng. 136 (2018) 408–418 .

    [12] R.K. Wu , T. Hong , Q.Y. Cheng , H. Zou , Y.W. Fan , X.B. Luo , Thermal modelingand comparative analysis of jet impingement liquid cooling for high power

    electronics, Int. J. Heat Mass Transf. 137 (2019) 42–51 . [13] D.B. Tuckerman , R.F.W. Pease , High-performance heat sinking for VLSI, IEEE

    Electron Device Lett. 2 (5) (1981) 126–129 .

    https://doi.org/10.13039/501100001809http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0001http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0001http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0001http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0001http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0001http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0002http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0002http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0002http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0003http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0003http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0003http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0003http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0003http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0003http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0004http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0005http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0005http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0005http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0006http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0006http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0006http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0006http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0006http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0007http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0007http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0008http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0008http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0008http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0009http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0009http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0009http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0009http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0009http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0009http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0010http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0010http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0010http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0010http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0011http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0011http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0011http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0011http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0011http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0011http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0012http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0012http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0012http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0012http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0012http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0012http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0012http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0013http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0013http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0013

  • H.C. Cui, X.T. Lai, J.F. Wu et al. International Journal of Heat and Mass Transfer 167 (2021) 120839

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [

    [14] G.D. Xia , D.D. Ma , Y.L. Zhai , Y.F. Li , R. Liu , M. Du , Experimental and numericalstudy of fluid flow and heat transfer characteristics in microchannel heat sink

    with complex structure, Energy Convers. Manage. 105 (2015) 848–857 . [15] D.X. Deng , W. Wan , Y. Tang , Z.P. Wan , D.J. Liang , Experimental investigations on

    flow boiling performance of reentrant and rectangular microchannels–a com- parative study, Int. J. Heat Mass Transf. 82 (2015) 435–446 .

    [16] Y.L. Zhai , Z.H. Li , H. Wang , J.X. Xu , Thermodynamic analysis of the effect ofchannel geometry on heat transfer in double-layered microchannel heat sinks,

    Energy Convers. Manage. 143 (2017) 431–439 .

    [17] M.K. Sung , I. Mudawar , Effects of jet pattern on single-phase cooling perfor- mance of hybrid micro-channel/micro-circular-jet-impingement thermal man-

    agement scheme, Int. J. Heat Mass Transf. 51 (19-20) (2008) 4614–4627 . [18] J.L. Xie , R. Zhao , F. Duan , T.N. Wong , Thin liquid film flow and heat transfer

    under spray impingement, Appl. Therm. Eng. 48 (2012) 342–348 . [19] Z.B. Yan , K.C. Toh , F. Duan , T.N. Wong , K.F. Choo , P.K. Chan , Y.S. Chua , Exper-

    imental study of impingement spray cooling for high power devices, Appl.

    Therm. Eng. 30 (10) (2010) 1225–1230 . 20] A.C. Kheirabadi , D. Groulx , Cooling of server electronics: a design review of

    existing technology, Appl. Therm. Eng. 105 (2016) 622–638 . [21] E.N. Wang , L. Zhang , L.N. Jiang , J.M. Koo , J.G. Maveety , E.A. Sanchez , K.E. Good-

    son , T.W. Kenny , Micromachined jets for liquid impingement cooling of VLSI chips, J. Microelectromech. Syst. 13 (6) (2004) 1072 .

    22] S. Chunqiang , S. Shuangquan , T. Changqing , X. Hongbo , Development and ex-

    perimental investigation of a novel spray cooling system integrated in refrig- eration circuit, Appl. Therm. Eng. 33 (2012) 246–252 .

    23] G. Liang , I. Mudawar , Review of spray cooling–part 1: single-phase and nucle-ate boiling regimes, and critical heat flux, Int. J Heat Mass Transf. 115 (2017)

    1174–1205 . 24] P. McInturff, M. Suzuki , P. Ligrani , C. Nakamata , D.H. Lee , Effects of hole shape

    on impingement jet array heat transfer with small-scale, target surface triangle

    roughness, Int. J. Heat Mass Transf. 127 (2018) 585–597 . 25] P.R. Parida , S.V. Ekkad , K. Ngo , Experimental and numerical investigation of

    confined oblique impingement configurations for high heat flux applications, Int. J. Therm. Sci. 50 (6) (2011) 1037–1050 .

    26] M.K. Sung , I. Mudawar , Single-phase and two-phase cooling using hybrid micro-channel/slot-jet module, Int. J. Heat Mass Transf. 51 (15-16) (2008)

    3825–3839 .

    27] T.A. Trabold , N.T. Obot , Impingement heat transfer within arrays of circular jets: part II—effects of crossflow in the presence of roughness elements, J. Tur-

    bomach. 109 (4) (1987) 594 . 28] H. Chang , J. Zhang , T. Huang , Experimental investigation on impingement heat

    transfer from rib roughened surface within arrays of circular jets: correlation, in: Proceedings of ASME Turbo Expo20 0 0, Munich, Germany, 20 0 0 .

    29] S. Spring , Y.F. Xing , B. Weigand , An experimental and numerical study of heat

    transfer from arrays of impinging jets with surface ribs, ASME J. Heat. Transfer 134 (8) (2012) 082201 .

    30] A. Husain , M. Ariz , N.Z.H. Al-Rawahi , M.Z. Ansari , Thermal performance anal-ysis of a hybrid micro-channel, -pillar and -jet impingement heat sink, Appl.

    Therm. Eng. 102 (2016) 989–10 0 0 . [31] X.M. Huang , W. Yang , T.Z. Ming , W.Q. Shen , X.F. Yu , Heat transfer enhancement

    on a microchannel heat sink with impinging jets and dimples, Int. J. Heat Mass Transf. 112 (2017) 113–124 .

    15

    32] T.Z. Ming , C.J. Cai , W. Yang , W.Q. Shen , W. Feng , N. Zhou , Optimization of dim-ples in microchannel heat sink with impinging jets-part B: the influences of

    dimple height and arrangement, J. Therm. Sci. 27 (4) (2018) 321–330 . 33] T.Z. Ming , C.J. Cai , W. Yang , W.Q. Shen , T. Gan , Optimization of dimples in mi-

    crochannel heat sink with impinging jets-part A: mathematical model and the influence of dimple radius, J. Therm. Sci. 27 (3) (2018) 195–202 .

    34] Z.Y. Guo , D.Y. Li , B.X. Wang , A novel concept for convective heat transfer en-hancement, Int. J. Heat Mass Transf. 41 (14) (1998) 2221–2225 .

    35] Q. Chen , J. Ren , J.A. Meng , Field synergy equation for turbulent heat transfer

    and its application, Int. J. Heat Mass Transf. 50 (25-26) (2007) 5334–5339 . 36] W. Liu , P. Liu , Z.M. Dong , K. Yang , Z.C. Liu , A study on the multi-field synergy

    principle of convective heat and mass transfer enhancement, Int. J. Heat Mass Transf. 134 (2019) 722–734 .

    37] S.N. Joshi , E.M. Dede , Effect of sub-cooling on performance of a multi-jet twophase cooler with multi-scale porous surfaces, Int. J. Therm. Sci. 87 (2015)

    110–120 .

    38] K. Ebrahimi , G.F. Jones , A.S. Fleischer , A review of data center cooling technol-ogy, operating conditions and the corresponding low-grade waste heat recov-

    ery opportunities, Renew. Sustain. Energy Rev. 31 (2014) 622–638 . 39] R. Wu , Y. Fan , T. Hong , H. Zou , R. Hu , X. Luo , An immersed jet array impinge-

    ment cooling device with distributed returns for direct body liquid cooling of high power electronics, Appl. Therm. Eng. 162 (2019) 114259 .

    40] E.A. Browne , G.J. Michna , M.K. Jensen , Y. Peles , Experimental investigation of

    single-phase microjet array heat transfer, ASME J. Heat Transfer 132 (4) (2010) 041013 .

    [41] S. Caliskan , S. Baskaya , Experimental investigation of impinging jet array heat transfer from a surface with V-shaped and convergent-divergent ribs, Int. J.

    Therm. Sci. 59 (2012) 234–246 . 42] A . Husain , N.A . Al-Azri , N.Z.H. Al-Rawahi , A. Samad , Comparative performance

    analysis of microjet impingement cooling models with different spent-flow

    schemes, J. Thermophys. Heat Transf. 30 (2) (2016) 465–471 . 43] A. Husain , J.-H. Kim , K.-Y. Kim , Performance characterization of laminar flow

    in multiple microjet impingement heat sinks, J. Thermophys. Heat Transf. 28 (1) (2014) 133–141 .

    44] N. Karwa , C. Stanley , H. Intwala , G. Rosengarten , Development of a low ther-mal resistance water jet cooled heat sink for thermoelectric refrigerators, Appl.

    Therm. Eng. 111 (2017) 1596–1602 .

    45] P.J. Roache , Perspective: a method for uniform reporting of grid refinement studies, ASME J. Fluids Eng. 116 (3) (1994) 405–413 .

    46] I.B. Celik , U. Ghia , P.J. Roache , C.J. Freitas , Procedure for estimation and report-ing of uncertainty due to discretization in CFD applications, ASME J. Fluids Eng.

    130 (7) (2008) 078001 . [47] W. Liu , P. Liu , J.B. Wang , N.B. Zheng , Z.C. Liu , Exergy destruction minimization:

    a principle to convective heat transfer enhancement, Int. J. Heat Mass Transf.

    122 (2018) 11–21 . 48] S.S. Bertsch , E.A. Groll , S.V. Garimella , Refrigerant flow boiling heat transfer in

    parallel microchannels as a function of local vapor quality, Int. J. Heat Mass Transf. 51 (19-20) (2008) 4775–4787 .

    49] J. Lee , I. Mudawar , Fluid flow and heat transfer characteristics of low temper-ature two-phase micro-channel heat sinks-part 2. Subcooled boiling pressure

    drop and heat transfer, Int. J. Heat Mass Transf. 51 (17-18) (2008) 4327–4341 .

    http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0014http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0014http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0014http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0014http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0014http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0014http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0014http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0015http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0015http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0015http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0015http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0015http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0015http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0016http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0016http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0016http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0016http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0016http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0017http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0017http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0017http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0018http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0018http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0018http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0018http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0018http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0019http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0019http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0019http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0019http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0019http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0019http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0019http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0019http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0020http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0020http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0020http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0021http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0022http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0022http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0022http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0022http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0022http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0023http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0023http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0023http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0024http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0024http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0024http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0024http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0024http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0024http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0025http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0025http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0025http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0025http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0026http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0026http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0026http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0027http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0027http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0027http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0028http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0028http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0028http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0028http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0029http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0029http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0029http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0029http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0030http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0030http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0030http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0030http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0030http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0031http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0031http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0031http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0031http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0031http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0031http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0032http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0032http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0032http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0032http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0032http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0032http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0032http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0033http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0033http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0033http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0033http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0033http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0033http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0034http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0034http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0034http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0034http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0035http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0035http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0035http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0035http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0036http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0036http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0036http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0036http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0036http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0036http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0037http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0037http://refhub.elsevier.com/S0017-9310(20)33777-7/sbref0037http://refhub.e