1. Trang chủ
  2. » Luận Văn - Báo Cáo

Heat transfer engineering an international journal, tập 31, số 4, 2010

81 424 0

Đang tải... (xem toàn văn)

Tài liệu hạn chế xem trước, để xem đầy đủ mời bạn chọn Tải xuống

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Định dạng
Số trang 81
Dung lượng 19,08 MB

Các công cụ chuyển đổi và chỉnh sửa cho tài liệu này

Nội dung

R., Fabbri, M., Michel, B., Calmi, D., and Kloter, U., High Heat Flux Flow Boiling in Silicon Multi-Microchannels—Part I: Heat Transfer Characteristics of Refriger-ant R236fa, Internatio

Trang 2

CopyrightC Taylor and Francis Group, LLC

ISSN: 0145-7632 print / 1521-0537 online

DOI: 10.1080/01457630903311652

e d i t o r i a l

Recent Developments in Flow Boiling and Two-Phase Flow in Small

Channels and Microchannels

JOHN R THOME and ANDREA CIONCOLINI

Heat and Mass Transfer Laboratory, ´Ecole Polytechnique F´ed´erale de Lausanne, Lausanne, Switzerland

Microscale two-phase flow is at present one of the hottest topics of heat transfer research, both in academia and in

the industry The miniaturization of two-phase flow systems, which has led to numerous experimental and theoretical

challenges not yet completely resolved, is primarily related to the dissipation of high heat duties typical of compact

systems such as CPU (central processing unit) chips, electronic devices, micro chemical reactors, and micro fuel cell

combustors.

Among the areas concerned with CPU (central processing

units) chips cooling [1], in particular, data centers have become

common and are found in nearly every sector of the economy,

such as manufacturing, universities, financial services,

govern-ment institutions, etc The increasing demand during the past 10

years for computer resources has led to a considerable increase

in the number of data centers and their corresponding energy

consumption Energy considerations are becoming essential, as

the International Panel on Climate Change (IPPC) and the

Ky-oto treaty show that drastic reductions of CO2 emissions are

urgently needed In this context, information technology has a

key role to play as the energy consumed in data centers

rep-resents almost 2% of the world electricity consumption and is

growing by 15% annually, while the current efficiency of such

systems is usually less than 20% In addition to environmental

considerations, the rise in energy costs is a key motivator of

technological change, as the cooling process becomes the major

part of the data center operating costs

Address correspondence to Professor John R Thome, Heat and Mass

Trans-fer Laboratory, ´ Ecole Polytechnique F´ed´erale de Lausanne,

EPFL-STI-IGM-LTCM, Station 9, 1015 Lausanne, Switzerland E-mail: john.thome@epfl.ch

The market for cooling of personal computers (PCs), datacenters, and telecom equipment is at a crossroads between oldair-cooling technology and more effective solutions, mainly liq-uid and two-phase cooling It appears that liquid cooling is thepreferred near-term solution because of its higher ease of im-plementation, but two-phase microscale cooling is of particularinterest due to evident performance advantages For instance, thelatent heat allows operation at a lower mass flow rate than single-phase cooling, and thus can reduce pumping power require-ments, resulting in a more energy-efficient system The boilingprocess takes place at an almost constant temperature, leading

to a small temperature gradient along the chip surface, which isadvantageous for thermal interface durability Finally, primarytrends in boiling in multi-microchannels [2–4] show that theboiling heat transfer coefficient increases with heat flux anddecreases slightly with increasing vapor quality Consequently,two-phase cooling is intrinsically well adapted to hot-spot man-agement, which is a critical point for the electronics industry andfor obtaining a uniform operating temperature along the chip.This issue collects seven papers originally presented at the5th International Conference on Transport Phenomena in Mul-tiphase Systems, HEAT 2008, June 30–July 3, 2008, Bialystok,Poland These studies address several aspects of flow boiling

255

Trang 3

256 J R THOME AND A CIONCOLINI

and two-phase flow in small channels and microchannels, both

experimentally and theoretically

REFERENCES

[1] Thome, J R., and Bruch, A., Refrigerated Cooling of

Micropro-cessors With Micro-Evaporation Heat Sinks: New Developments

and Energy Conservation Prospects for Green Datacenters, Proc.

Institute of Refrigeration 2008–2009, 2–1.

[2] Agostini, B., Thome, J R., Fabbri, M., Michel, B., Calmi, D.,

and Kloter, U., High Heat Flux Flow Boiling in Silicon

Multi-Microchannels—Part I: Heat Transfer Characteristics of

Refriger-ant R236fa, International Journal of Heat and Mass Transfer, vol.

51, pp 5400–5414, 2008

[3] Agostini, B., Thome, J R., Fabbri, M., Michel, B., Calmi, D.,

and Kloter, U., High Heat Flux Flow Boiling in Silicon

Multi-Microchannels—Part II: Heat Transfer Characteristics of

Refriger-ant R245fa, International Journal of Heat and Mass Transfer, vol.

51, pp 5415–5425, 2008

[4] Agostini, B., Revellin, R., Thome, J R., Fabbri, M., Michel, B.,

Calmi, D., and Kloter, U., High Heat Flux Flow Boiling in

Sili-con Multi-Microchannels—Part III: Saturated Critical Heat Flux

of R236fa and Two-Phase Pressure Drops, International Journal

of Heat and Mass Transfer, vol 51, pp 5426–5442, 2008.

John R Thome is a professor of heat and mass

trans-fer at the Swiss Federal Institute of Technology in Lausanne (EPFL), Switzerland, since 1998, where he

is director of the Laboratory of Heat and Mass fer (LTCM) in the Faculty of Engineering Science and Technology (STI) His primary interests of re- search are two-phase flow and heat transfer, covering boiling and condensation of internal flows, external flows, enhanced surfaces, and microchannels He re- ceived his Ph.D at Oxford University, England, in

Trans-1978 and was formerly a professor at Michigan State University He is the

author of several books: Enhanced Boiling Heat Transfer (1990), Convective

Boiling and Condensation (1994), and Wolverine Engineering Databook III

(2004) He received the ASME Heat Transfer Division’s Best Paper Award in

1998 for a three-part paper on flow boiling heat transfer published in the Journal

of Heat Transfer.

Andrea Cioncolini is a postdoctoral researcher in the

Laboratory of Heat and Mass Transfer (LTCM) at the Swiss Federal Institute of Technology in Lausanne, Switzerland (EPFL) He received his Laurea degree and Ph.D in nuclear engineering at the Polytechnic University of Milan, Italy He joined LTCM after 2 years as a senior engineer at Westinghouse Electric Company, Science and Technology Department, in Pittsburgh, Pennsylvania.

heat transfer engineering vol 31 no 4 2010

Trang 4

CopyrightC Taylor and Francis Group, LLC

ISSN: 0145-7632 print / 1521-0537 online

DOI: 10.1080/01457630903311678

Flow Patterns and Heat Transfer for Flow Boiling in Small to Micro

Diameter Tubes

TASSOS G KARAYIANNIS1, DEREJE SHIFERAW1, DAVID B R KENNING1,

and VISHWAS V WADEKAR2

1School of Engineering and Design, Brunel University, West London, United Kingdom

2HTFS, Aspen Technology Ltd., Reading, United Kingdom

An overview of the recent developments in the study of flow patterns and boiling heat transfer in small to micro diameter

tubes is presented The latest results of a long-term study of flow boiling of R134a in five vertical stainless-steel tubes of

internal diameter 4.26, 2.88, 2.01, 1.1, and 0.52 mm are then discussed During these experiments, the mass flux was varied

from 100 to 700 kg/m 2 s and the heat flux from as low as 1.6 to 135 kW/m 2 Five different pressures were studied, namely,

6, 8, 10, 12, and 14 bar The flow regimes were observed at a glass section located directly at the exit of the heated test

section The range of diameters was chosen to investigate thresholds for macro, small, or micro tube characteristics The

heat transfer coefficients in tubes ranging from 4.26 mm down to 1.1 mm increased with heat flux and system pressure,

but did not change with vapor quality for low quality values At higher quality, the heat transfer coefficients decreased

with increasing quality, indicating local transient dry-out, instead of increasing as expected in macro tubes There was

no significant difference between the characteristics and magnitude of the heat transfer coefficients in the 4.26 mm and

2.88 mm tubes but the coefficients in the 2.01 and 1.1 mm tubes were higher Confined bubble flow was first observed in the

2.01 mm tube, which suggests that this size might be considered as a critical diameter to distinguish small from macro tubes.

Further differences have now been observed in the 0.52 mm tube: A transitional wavy flow appeared over a significant range

of quality/heat flux and dispersed flow was not observed The heat transfer characteristics were also different from those in

the larger tubes The data fell into two groups that exhibited different influences of heat flux below and above a heat flux

threshold These differences, in both flow patterns and heat transfer, indicate a possible second change from small to micro

behavior at diameters less than 1 mm for R134a.

INTRODUCTION

Modeling and design of micro devices of high thermal

perfor-mance, including electronic chips and other systems containing

compact and ultra-compact heat exchangers, require a

funda-mental understanding of thermal transport phenomena for the

ultra-compact systems In this emerging area of great practical

interest, systematically measured boiling heat transfer data are

The authors thank Professor Andrea Luke of Hannover University and her

team, who carried out the surface roughness measurements for the 0.52 mm

tube, and acknowledge the contributions of Drs Y S Tian, L Chen, and X.

Huo to the earlier part of this long-term study.

Address correspondence to Prof Tassos G Karayiannis, Brunel University,

School of Engineering and Design, West London, Uxbridge, Middlesex, UB8

3PH, United Kingdom E-mail: tassos.karayiannis@brunel.ac.uk

required to understand the mechanisms of flow boiling in

small-to micro-diameter passages

Channel Size Classification

Identifying the channel diameter threshold below which themacro-scale heat transfer phenomena do not fully apply is im-portant in validating and developing predictive criteria for thethermal-hydraulic performance of small- to micro-scale chan-nels However, there is no clear and common agreement onthe definition and classification criteria for the size ranges insmall/mini/microchannel two-phase flow studies One reasoncould be the lack of comprehensive heat transfer data cover-ing a wide range of channel diameters Mehandale et al [1]defined channel size ranges as follows: microchannel (1–100257

Trang 5

258 T G KARAYIANNIS ET AL.

µm), mesochannel (100 µm–1 mm), macrochannel (1–6 mm),

and conventional (dh > 6 mm) Kandlikar and Grande [2]

sug-gested the classification of microscale by hydraulic diameter,

given as conventional channels (dh ≥ 3 mm), minichannels

(200µm ≤ dh < 3 mm), and microchannels (10µm ≤ dh <

200µm) These methods based only on size do not consider the

physical mechanisms and the variation of fluid properties with

pressure The absence of stratified flow in horizontal

microchan-nels, and hence the fact that the orientation of the channel has

virtually no effect on two phase flow patterns, indicates the

pre-dominance of surface tension force over gravity Consequently,

a number of attempts to define macro–micro transition have used

surface tension force as a base to formulate a nondimensional

criterion These include E¨otv¨os number (E¨o> 1) recommended

by Brauner and Moalem-Maron [3] and confinement number

(Co= 0.5) by Cornwell and Kew [4] Thome [5] in his review

of boiling in microchannels indicated the importance of

consid-ering the effect of channel size on the physical mechanisms and

discussed the use of bubble departure diameter as a preliminary

criterion He also mentioned the effects of shear on bubble

de-parture diameter and the effect of reduced pressure on bubble

size that should be considered in addition to surface tension

forces A comprehensive definition for normal and small size

tubes is required that considers all the fundamental phenomena,

based on experimental data for a wide range of conditions The

research presented here addressed this requirement by

system-atic measurements of flow boiling of R134a over wide ranges of

pressures, flow rates, and heat fluxes in five tubes with diameters

ranging from 4.26 to 0.52 mm This choice of size range was

based on an initial assessment using the confinement number

proposed by Cornwell and Kew [4]

Flow Patterns

Flow pattern studies in small/micro tubes have clearly shown

that there is a considerable difference in the flow pattern

char-acteristics compared with conventional size channels These

include the predominance of surface tension force over gravity,

the absence of stratified flow pattern in horizontal channels, and

the appearance of additional flow patterns that are not common

in normal-diameter tubes In the past some researchers have

proposed several flow pattern classes, probably more than is

necessary for modeling Although there are arguments on the

classification of flow patterns, the most commonly identified

flow patterns so far are bubbly flow, slug flow, churn flow, and

annular flow Barnea et al [6] classified the flow patterns as

dispersed bubble, elongated bubble, slug, churn, and annular

Elongated bubble, slug, and churn were considered as

intermit-tent flow Dispersed flow and elongated bubble were replaced

by bubbly flow in the Mishima and Hibiki [7] classification

Kew and Cornwell [8] experimentally observed flow regimes

during their flow boiling tests in small-diameter channels using

R141b, and proposed only three distinct flow regimes They

de-fined the flow patterns as isolated bubble flow, conde-fined bubble

flow, and annular-slug flow Identification of flow patterns issubject to uncertainty, which is not straightforward to quantifyand can also be significantly influenced by the experimentaltechnique used Besides, the transition from one flow pattern

to another may be a gradual rather an abrupt transition, as isoften reported Hence, flow patterns may possess characteris-tics of more than one flow pattern during transition Chen et

al [9] reported the results of a detailed study of flow ization experiments with R134a for a pressure range of 6–14bar and tube diameter from 1.1, 2.01, 2.88, and 4.26 mm withthe same test rig as the present one The typical flow patternsobserved in the four tubes are presented in Figure 1 They in-cluded dispersed flow, bubbly flow, confined flow, slug flow,churn flow, annular flow, and mist flow The flow patterns inthe 2.88 and 4.26 mm tubes exhibit characteristics found inlarge tubes The flow patterns in the 2.01 mm tube demonstratesome “small tube characteristics,” e.g., the appearance of con-fined bubble flow at the lowest pressure of 6 bar, and slimmervapor slug, thinner liquid film, and a less chaotic vapor–liquidinterface in churn flow Confined flow was observed at all pres-sures when the diameter was reduced to 1.1 mm, indicating

visual-Figure 1 Flow patterns for R134a at 10 bar pressure: (a)d= 1.10 mm, (b)

d = 2.01 mm, (c) d = 2.88 mm, and (d) d = 4.26 mm (Chen et al [9]).

heat transfer engineering vol 31 no 4 2010

Trang 6

T G KARAYIANNIS ET AL 259

a potential transition range for heat transfer between 2 and 1

mm

Studies of even smaller diameter tubes are described next

Serizawa et al [10] studied two phase flow in microchannels and

reported the visualization results for air–water and steam–water

flows in circular tube of 20, 25, and 100µm and 50 µm internal

diameter, respectively They found several additional features

to those observed in small-diameter tubes For air–water

two-phase flow in a 25µm silica tube the special flow pattern features

found included liquid ring flow and liquid lump flow The

liq-uid ring flow was described as the appearance of a symmetrical

liquid ring with long gas slugs passing in the middle Serizawa

et al hypothesized that the liquid ring flow could develop from

slug flow when the gas slug velocity is too high and the liquid

slug is too short to form a stable liquid bridge between

consecu-tive gas slugs At this condition, liquid lump flow appeared with

further increases in the gas flow rate According to Serizawa

et al., “the high-speed core gas entrains the liquid phase and

liq-uid lumps are sliding on the wall.” Experiments using the same

fluid but in a 100µm quartz tube gave similar results as for the

25µm silicon tubes except that small liquid droplets in gas slug

flow were sticking on the tube wall, indicating the absence of

a liquid film at these locations between the slug and the wall

Stable liquid ring flow and liquid lump flows were also reported

for the 100µm tube Flow patterns similar to those of air–water

flow in the 25µm silica tube were observed in the case of steam–

water flow in a 50µm silica tube, with the only difference being

the absence of liquid lump flow, which, according to Serizawa et

al., was not a main flow but transition type flow However, liquid

ring flow was still found, which may indicate that the difference

in the method of forming the two-phase flow, i.e., boiling or

adiabatic mixing of air–water, seems to have no considerable

effect, at least for these sizes

Kawahara et al [11] studied two-phase flow characteristics

of nitrogen and deionized water in a 100 µm diameter tube

made of fused silica, and noted the absence of bubbly and churn

flow as one of the differences between their results and results

for larger diameter tubes They reported mainly intermittent and

semi-annular flows Recently, Xiong and Chung [12] studied

ex-perimentally adiabatic gas–liquid flow patterns using nitrogen

and water in rectangular microchannels with hydraulic

diame-ter of 0.209, 0.412, and 0.622 mm They observed four

differ-ent flow patterns: bubbly-slug flow, slug-ring flow (liquid-ring

flow), dispersed-churn flow, and annular flow in the 0.412 and

0.622 mm microchannels The bubbly-slug flow developed to

fully slug flow They reported that dispersed and churn flows

were absent in the 0.209 mm channel

Effect of Diameter on Transition Boundaries

The effect of tube diameter on flow pattern transition

bound-aries was also studied by various researchers Damianides and

Westwater [13] studied the flow regimes in horizontal tubes

of 1 to 5 mm inside diameters using air–water They reported

that reducing the tube diameter shifted the transition boundariesbetween intermittent-dispersed bubbly and intermittent-annularflow toward lower liquid velocity and higher gas velocity, re-spectively Also, they did not observe stratified flow regimeinside the 1 mm diameter tube In the study of air–water flowpatterns in tubes of 0.5 to 4.0 mm inside diameter, for verticalflow, Lin et al [14] observed that decreasing the tube diametershifted the slug-churn and churn-annular transition boundariestoward lower vapor velocity

Recently, Chen et al [9] noted that the diameter influencesthe transition boundaries of dispersed bubble-bubbly, slug-churnand churn-annular flow Also, the slug-churn and churn-annularboundaries are weakly dependent on superficial liquid veloc-ity and strongly dependent on superficial vapor velocity Thereseems to be no effect of diameter at the boundaries of dispersedbubble-churn and bubbly-slug flow The flow pattern transitiondata of Chen et al are plotted on a mass flux versus qualitygraph in Figure 2 for pressures of 6 and 8 bar As shown in thefigure, when the diameter is reduced, the slug-churn and churn-annular transition lines shift toward higher quality The change

is more pronounced for moderate and low mass fluxes There

is no obvious effect on the bubbly/slug transition line The flowregime boundaries are shifted to significantly lower qualities asthe mass flux increases At higher quality, the transition linesfor different tubes merge into a single line Chen et al reportedthat the Weber number may be the appropriate parameter todeduce general correlations to predict the transition boundariesthat include the effect of diameter

Recently, new correlations for transition of non-adiabaticflow patterns were introduced by Revellin and Thome [15]

They identified three main flow patterns, named (a) the isolated

bubble regime, which includes bubbly flow and short slugs—in

this regime coalescence is not significant; (b) the coalescing

bubble regime, where slug flow is the main flow with some of

the bubbles coalescing to form a longer slug; and (c) the annular

regime According to their observations, churn flow is a tion from coalescing bubble to annular flow, and it is considered

transi-an indication of the end of coalescing bubble flow The flowpattern maps were plotted as mass flux versus quality graphs.Revellin and Thome proposed flow pattern transition correla-tions, which give the quality at which the transition occurs Forthe transition from the isolated bubble to the coalescing bub-ble regime, their correlation contained the Reynolds, Boiling,and Weber numbers, as in Eq (1) A similar correlation for thetransition from the coalescing bubble to the annular regime con-tained only the Reynolds number and the Weber number, as in

Trang 7

260 T G KARAYIANNIS ET AL.

Figure 2 Flow pattern transition boundary lines for the four tubes (Chen

et al [9] data): (a) 6 bar and (b) 8 bar pressure.

transition from coalescing bubble to annular flow regime, which

is equivalent to churn to annular transition, shifts to lower

qual-ity with decreasing diameter This is contrary to the results of

Chen et al [9] and could be due to the fact that the correlation

was developed using tests with a single tube diameter rather

than a range of tube diameters For instance, at a mass flux of

400 kg/m2-s and pressure of 8 bar, the transition qualities for the

2.01 and 1.10 mm tubes arex = 0.38 and x = 0.32, respectively.

From the experimental results of Chen et al [9], shown in

Fig-ure 2b, the corresponding values are 0.22 and 0.24, respectively

From the preceding review, it appears that small-diameter

tubes exhibit flow pattern characteristics different from those

for large diameter tubes, e.g., the appearance of confined flow

at about 2 mm for R134a, which may indicate a threshold forchange from large to small diameter For the same fluid theCornwell and Kew [4] criterion gives a critical diameter of 1.7

mm for P= 6 bar pressure Flow pattern studies for even smallertubes (near or less than 1 mm) revealed the existence of a number

of different flow pattern types, e.g., ring flow and lump liquidflow, which have not been found in larger diameter tubes This

is indicative of a possible further change in flow patterns andhence in thermal characteristics at these even smaller diameters.This is discussed later in the article in light of the recent resultsfrom our own investigations

Heat Transfer

Nucleate boiling, forced convection, and a combination ofthe two are the main mechanisms often reported in the litera-ture for flow boiling heat transfer in large-diameter tubes, e.g.,Kenning and Cooper [16] These have also been adopted inidentifying the heat transfer mechanism in small-diameter tubesand microchannels, although different conclusions have beendrawn by researchers as to their prevalence Some researchersconcluded that nucleate boiling is the dominant heat transfermechanism when it was observed that the heat transfer coef-ficient is more or less independent of vapor quality and massflux, while it is strongly dependent on heat flux—e.g., Lazarekand Black [17], Wambsganss et al [18], Tran et al [19], Bao

et al [20], Yu et al [21], and Fujita et al [22] On the otherhand, some experimental studies have also reported an effect

of the mass velocity and vapor quality but not of the heat flux

on the heat transfer coefficient The interpretation given to this

is that forced convective boiling is the dominant heat transfermechanism—e.g., Carey et al [23], Oh et al [24], Lee and Lee[25], and Qu and Mudawar [26] Some researchers reported acombined effect of both mechanisms, i.e., nucleate boiling atlow quality and forced convective boiling at high quality region,

in a way similar to that observed in large-diameter tubes—e.g.,Kuznestov and Shamirzaev [27], Lin et al [28], Sumith et al.[29], and Saitoh et al [30] However, it is worth noting here thatmacroscale boiling heat transfer correlations and models did notpredict well the heat transfer coefficient in small-diameter tubes,

as noted by Qu and Mudawar [26], Owhaib and Palm [31], andHuo et al [32]

More complex behavior and differences dependent on thefluid tested were reported by other researchers For example,Dı’az and Schmidt [33] investigated transient boiling heat trans-fer in 0.3× 12.7 mm microchannels using infrared thermogra-phy to measure the wall temperature For water, the heat transfercoefficient decreased with quality near the zero quality region,followed by a uniform heat transfer coefficient However, forethanol at high quality, an increase in heat transfer coefficientwith quality was found to be independent of applied heat flux

A similar behavior, i.e., an increase in the heat transfer cient with quality, was observed by Xu et al [34] and Lie et al.heat transfer engineering vol 31 no 4 2010

Trang 8

coeffi-T G KARAYIANNIS ET AL 261[35] Lie et al [35] investigated experimentally evaporation heat

transfer of R134a and R407c flow in horizontal small tubes of

0.83 and 2.0 mm internal diameter The fluid was preheated to

an inlet quality that varied from 0.2 to 0.8 The heat transfer

co-efficient was observed to increase with quality almost linearly,

except at lower mass flux and heat flux It also increased with

heat flux, mass flux, and saturation pressure Saitoh et al [30]

studied the effect of tube diameter on boiling heat transfer of

R134a in horizontal tubes with inner diameter of 0.51, 1.12,

and 3.1 mm The heated lengths were 3.24, 0.935, and 0.550 m

respectively The heat flux ranged from 5 to 39 kW/m2, mass

flux from 150 to 450 kg/m2s, saturation pressure from 3.5 to 4.7

bar, and inlet vapor quality from 0 to 0.2 For the 3.1 mm tube,

when the quality was less than 0.6, the heat transfer coefficient

was strongly affected by heat flux and was not a function of

mass flux and quality For quality greater than 0.5, heat transfer

coefficient increased with mass flux and quality, but was not

affected by heat flux This quality limit shifted to 0.4 for the

1.12 mm tube The 0.51 mm results did not exhibit the same

heat transfer characteristic as the rest of the tubes When the

quality was less than 0.5, the heat transfer coefficient seemed to

increase with quality and heat flux and slightly with mass flux

In this region, the heat transfer coefficient was slightly higher

than the 1.12 and 3.1 mm tubes There was also an early dry-out

compared with the other tubes, and the region of decreasing

heat transfer coefficient with quality is not such a sharp drop

as the rest They observed flow instabilities in the two larger

tubes (3.1 and 1.12 mm), but not in the 0.51 mm tube Agostini

and Thome [36] categorized the trends in the local heat

trans-fer coefficient versus vapor quality and its relation to heat and

mass flux after reviewing 13 different studies They noted that

in most of the cases reviewed that at low quality (<0.5) the heat

transfer coefficient increases with heat flux and decreases or is

relatively constant with vapor quality, and at high vapor quality

it decreases sharply with vapor quality and is independent of

heat flux or mass flux

Initiation of Boiling

Flow boiling in very-small-diameter tubes is usually

associ-ated with high initial liquid superheat required to initiate boiling

Yen et al [37] conducted flow boiling experiments in 0.19, 0.3,

and 0.51 mm inside diameter tubes using R123 and FC-72 They

observed a high liquid superheat that reached up to 70 K in their

experiments In the low quality region, the heat transfer

coeffi-cient was observed to decrease with quality up to approximately

x= 0.25 and then became almost constant with further increase

in quality Hapke et al [38] investigated boiling in a 1.5 mm

internal diameter tube and reported that the onset of boiling

oc-curred at higher liquid superheat than required for conventional

tubes Peng and Wang [39] and Peng et al [40], based on their

observations of boiling in microchannels of hydraulic diameter

200–600µm, argued that nucleation can hardly be seen in

mi-crochannels They proposed a hypothesis of “evaporating space”

to explain the phenomenon They also suggested a theoretical

model to predict the superheat temperature The unusually highsuperheat in micro tubes was also reported to be related to thereduction of active nucleation sites and vapor nucleation insidevery small channels, by Zhang et al [41] and Brereton et al.[42]

Temperature and Pressure Fluctuations

Microchannel flow boiling studies have demonstrated a crease in heat transfer coefficient with increasing quality, oftenaccompanied by fluctuating wall temperatures—e.g., Lin et al.[28], Yan and Lin [43], Wen et al [44], and Huo et al [32].These have been attributed to transient dry-out, particularly atlow mass flux, and relatively high heat flux Kenning et al [45]suggested that there are two different mechanisms of dry-outaround individual bubbles in microchannels These are dry-out

de-as a result of depletion of the film thickness below a certainminimum by complete evaporation of the liquid film beneaththe confined bubble and dry-out due to surface-tension-driven

“capillary roll-up” on partially wetted surfaces with finite tact angles Experimental studies also indicated fluctuations inpressure and wall temperature Yan and Kenning [46] inves-tigated water boiling at atmospheric pressure in a 2 × 1 mmchannel They showed that the pressure fluctuations were caused

con-by the acceleration of liquid slugs con-by expanding confined bles, confirming a model of Kew and Cornwell [47], and thatthe corresponding fluctuations in saturation temperature were ofmagnitude similar to the mean superheat causing evaporation,

bub-so they could not be neglected

Effect of Decreasing Diameter

There are a limited number of experiments that have tested awide range of tube diameter to investigate the heat transfer trendwith channel size Studies that have considered the effect of di-ameter are reviewed briefly here Yan and Lin [43] conductedexperiments with R134a using a single tube of internal diame-ter 2.0 mm and heated length 100 mm They claimed that theevaporation heat transfer coefficient increased by 30% to 80%compared with conventional diameter tubes Oh et al [24] ex-perimentally investigated the evaporation heat transfer for threedifferent copper tubes of diameter 0.75, 1.0, and 2.0 mm usingR134a For vapor quality less than 0.6, they found the heat trans-fer coefficient for the 1.0 mm tube to be higher than that of the2.0 mm tube by approximately 45% However, decreasing thetube diameter shifted to a lower quality the point at which theheat transfer coefficient started to decrease axially, presumablydue to dry-out Owhaib et al [48] studied experimentally evap-orative heat transfer using R134a in vertical circular tubes ofinternal diameter 1.7, 1.22, and 0.83 mm, and a uniform heatedlength of 220 mm Other parameter ranges are: mass flux 50–

400 kg/m2-s, heat flux 3–34 kW/m2, and pressure 6.5–8.6 bar.They concluded that the heat transfer coefficient increased withdecreasing tube diameter

heat transfer engineering vol 31 no 4 2010

Trang 9

262 T G KARAYIANNIS ET AL.

In general, experimental results indicate an increase in the

heat transfer coefficient as the diameter decreases However,

some contradictory results are also available For example,

Kuwahara et al [49] experimentally studied the flow boiling

heat transfer characteristic and flow pattern inside 0.84 and

2.0-mm diameter tubes using R134a and found no difference in the

heat transfer characteristics between the two tubes Baird et al

[50] conducted boiling experiments on tubes of 0.92 and 1.95

mm diameter and found no significant effect of diameter on the

heat transfer coefficient Khodabandeh [51] studied boiling in

a two-phase thermosyphon loop with iso-butene as a working

fluid with tubes ranging from 1.1 to 6 mm in diameter He also

concluded that the effect of diameter was small and not clear

In the work of Saitoh et al described earlier, there was no

obvi-ous effect of diameter on heat transfer coefficient or it was not

straightforward to deduce the influence

A theoretical three-zone model for predicting the local

dy-namic and local time-averaged heat transfer coefficient was

pre-sented by Thome et al [52] and Dupont et al [53] The model is

based on convective heat transfer in the confined bubble regime

without a contribution from nucleate boiling The model

pre-dictions indicate that the heat transfer coefficient increases with

diameter for quality greater than 0.18, while it decreases with

diameter for quality less than 0.04 Dupont and Thome [54]

compared the model results with the experiments of Owhaib

et al [48] The model did not predict the trend of increasing

heat transfer coefficient with decreasing diameter Instead an

opposite prediction was observed in the quality range covered

Dupont and Thome [54] noted the lack of adequate

experimen-tal data covering a wide range of tube diameter for boiling heat

transfer The model predictions were also compared with

ex-perimental data for R134a and tubes of 2.01 and 4.26 mm in

diameter by Shiferaw et al [55]; they reported that the model

predicts that the diameter has an opposite effect on the heat

transfer coefficient compared to the measured data

The preceding brief overview indicates that a lot of work

is still necessary to elucidate the effect of diameter on the rate

and mechanism of heat transfer, including the possible diameter

thresholds for distinguishing macro, small and microscale

char-acteristics Although more than two tubes were used in some

of the past studies, it was not possible to identify the influence

of diameter because different conditions were used for different

diameter tubes Therefore the experimental facility described in

the next section was used to determine the heat transfer

coeffi-cients for R134a in five tubes of different diameter for similar

wide ranges of heat and mass fluxes and pressure, combined

with flow visualization at the exit from the test section

EXPERIMENTAL FACILITY AND PROCEDURE

The experimental facility consists of three main systems,

which are the R134a main circuit, data acquisition and control,

and the R22 cooling system The main facility, which is shown

in Figure 3, was designed to allow testing of different fluids and

Figure 3 Schematic diagram of the experimental system.

a wide range of flow conditions Details of the experimental tem were given in Huo et al [32] The test sections were made

sys-of stainless-steel cold-drawn tubes The dimensions sys-of the fivetest tubes are given in Table 1 They were heated by the directpassage of alternating electric current The outer wall temper-atures for the 4.26 mm to 1.1 mm tubes were measured using

15 K-type thermocouples that were spot-welded to the outside

of the tube at a uniform spacing The first and last ple readings were not used in the analysis so as to avoid errorsdue to thermal conduction to the electrodes Ten thermocoupleswere spot-welded on the 0.52 mm tube; the two at each endwere located sufficiently far from the electrodes to be used inthe data analysis The pressures and temperatures at the inletand outlet were measured using pressure transducers and T-typethermocouples A differential pressure transducer was installedacross the test section to provide the pressure drop measure-ment At the exit from the heating section, a borosilicate glasstube for flow pattern observation was located A digital high-speed camera (Phantom V4 B/W, 512× 512 pixels resolution,

thermocou-1000 pictures/s with full resolution and maximum 32,000 tures/s with reduced resolution, 10 ms exposure time) was used

pic-to observe the flow patterns

A series of flow boiling tests was then performed at differentmass flux and heat flux During these tests, the inlet temperaturewas controlled at a subcooling of 1–5 K by adjusting the capacity

of the chiller and heating power to the preheater The flow rate

Table 1 Range of experiment parameters Parameters Range Diameter 4.26, 2.88, 2.01, 1.10, and 0.52 mm Wall thickness 0.245, 0.15, 0.19, 0.247, and 0.15 mm Heated length 500, 300, 211, 150, and 100 mm Roughness 1.75, 1.54, 1.82, 1.28, and 1.15 µm Mass flux 100–700 kg/m 2 -s

Heat flux 1.6–150 kW/m 2

Vapor quality 0–0.9 Pressure 6, 8, 10, 12, and 14 bar

heat transfer engineering vol 31 no 4 2010

Trang 10

Heat transfer coefficient 6–12.5%

was set to the required value and the heat flux was increased in

small steps until the exit quality reached about 90% The data

were recorded after the system was steady at each heat flux,

which normally took about 15 min but sometimes longer Each

recording was the average of 20 measurements The next test

was then performed at a different flow rate All the instruments

used were carefully calibrated Tables 1 and 2 summarize the

range and uncertainties of the important parameters

DATA REDUCTION

The local heat transfer coefficientα(z) at each thermocouple

position was calculated using local values of the inside wall

temperature and the saturation temperature and is given by:

(T wi)z − (T s)z (3)whereq is the inner wall heat flux to the fluid determined from

the electric power supply to the test section and the heat loss

T wiis the local inner wall temperature, which can be determined

using the internal heat generation and radial heat conduction

across the tube wall as given by:

T wi = T woq · d i

4

(d i /d o)2− 2 ln(d i /d o)− 1

1− (d i /d o)2

(4)

T sis the local saturation temperature, deduced from the local

fluid pressure assuming a linear pressure drop across the test

section The local specific enthalpy,h i, at each thermocouple

position was determined from the energy balance in each heated

section considering the losses:

h i = h i−1+ L i

˙

where the heat transfer (Q) is the total electric heat input, which

is equal to the product of the voltage and the current applied

di-rectly to the test section (Q) is the heat loss determined using

the loss coefficient obtained from single-phase test before each

series of boiling tests Therefore, the local vapor quality can be

calculated from the local specific enthalpy at each thermocouple

position and is given as:

con-in Figure 4b agree very well with Dittus and Boelter [57] and

Figure 4 Single-phase results ford= 4.26 mm at 7.5 bar: (a) friction factor

vs Re, (b) Nusselt number vs Re.

heat transfer engineering vol 31 no 4 2010

Trang 11

264 T G KARAYIANNIS ET AL.

Petukhov [58] correlation—again, below the uncertainty limit

The preceding results verified the overall accuracy of the

exper-imental system Experexper-imental accuracy becomes an increasing

difficult challenge as the size of the passages decreases and

ei-ther laminar or turbulent flow may exist, depending on the mass

flow rate Therefore, additional single-phase experiments were

performed with the 0.52 mm tube to assess the ability of the

test rig to produce accurate results at this small diameter The

comparisons of the experimental results with past results and

known correlations were presented in Shiferaw et al [59] The

results agreed fairly well with the modified Gnielinski [60] and

Adams et al [61] for the turbulent regime and Choi et al [62]

in the laminar regime The reproducibility of the boiling tests

was also verified The test results (4.26–1.1 mm tubes) were

mostly within the range of uncertainty of the data; see Shiferaw

et al [63] for the 1.1 mm tube The reproducibility of the

0.52 mm tube tests was acceptable in the lower range of heat flux,

had differences in the intermediate range, and was acceptable

again at high heat fluxes [64] This could be due to the sparse

and/or unstable nucleation sites at this small size and will be

examined further The preceding set of experiments confirmed

the adequate accuracy and validity of the present results

EXPERIMENTAL RESULTS AND DISCUSSION

Flow Pattern Results

Figure 5a and b, presents the flow patterns observed during

the boiling test at a mass flux of 400 kg/m2-s and pressure

8 bar for the 0.52 mm tube and should be compared with the

results of Chen et al [9], obtained with the same test facility and

procedure depicted in Figure 1 These flow patterns were taken

simultaneously with the heat transfer tests presented hereinafter

Figure 5 (a) Flow patterns in 0.52 mm tube at 400 kg/m 2 s and 8 bar; (b)

sequence of flow patterns showing coalescence.

at each value of heat flux They represent the more frequentlyobserved flow pattern for the particular heat flux However,more than one type of flow pattern occurred intermittently insome cases Image 1 shows bubbly flow Confined bubble flow(images 2 and 3) was observed at low heat flux or exit quality Asthe heat flux increased, the bubbles grew in length and becameelongated Further increase in heat flux resulted in the liquidslug between the bubbles being “pushed” onto the downstreambubble, creating coalescence of the bubbles and a wavy film

A similar phenomenon was observed by Revellin et al [65].Figure 5b shows a sequence of how three relatively short bubblescoalesce in the adiabatic viewing section to form an elongatedbubble, leaving the liquid film interface wavy Note that theseobservations were carried out at the exit of the test section andcoalescence may be different in the heated section As shownagain in Figure 5a, when increasing the heat flux even further,

a type of wavy film flow, similar in appearance to what wasdescribed earlier as liquid ring flow (Serizawa et al [10]), isobtained for a relatively wide range of quality (images 4–6) Inthis case, the film interface is highly nonuniform and can lead

to a transition to annular flow (image 7), since further increase

in heat flux reduces the wave irregularity and distributes thewaves almost uniformly: annular flow (images 8–10) At highheat flux, the annular flow patterns have small-scale roughness

of very short amplitude and wavelength

Overall, the flow patterns observed in the smaller tube ofinternal diameter 0.52 mm were different from those observed

in the larger tubes by Chen et al [9] As mentioned earlier,these differences include the absence of dispersed flow and theappearance of a transitional wavy film flow In this tube, liquidlump flow (see Serizawa et al [10]) was not observed

Heat Transfer Results

Typical experimental data for the five tubes are plotted asgraphs of heat transfer coefficient vs quality, the presentationconventionally used for large tubes This implies that heat trans-fer depends only on local flow conditions and not on how theflow is developed, so that the convective component depends

on the local flow pattern The relationship between flow patternobservations in an adiabatic section at the exit from the tube andthe flow pattern within the heated section at the same qualitymay require examination for the particular conditions in smalltubes, in which the growth of an individual bubble may influence

a considerable length of the tube

Data at a pressure of 8 bar and a mass flux of 400 kg/m2s

in the tubes with diameters 4.26–0.52 mm are plotted in Figure6a–e As seen in, for example, Figure 6a for the 4.26 mm tube,

at a qualityx < 0.5 approximately and moderate heat flux, the

heat transfer coefficient is constant within±10% at a value thatincreases with heat flux and pressure, but that is independent ofquality Huo et al [32] and Shiferaw et al [55] reported similartrends at 8 bar and a mass flux of 300 kg/m2-s in the 4.26 and2.01 mm tubes Within this range, the local variations appear toheat transfer engineering vol 31 no 4 2010

Trang 12

T G KARAYIANNIS ET AL 265

Figure 6 Local heat transfer coefficient as a function of vapor quality at mass flux 400 kg/m 2 -s and pressure 8 bar: (a) 4.26 mm; (b) 2.88 mm; (c) 2.01 mm; (d) 1.10 mm; (e) 0.52 mm.

follow a pattern associated with the axial positions of the

mea-suring stations As the variations do not appear in single-phase

flow experiments, they are not associated with individual

ther-mocouples or wall roughness that would affect the liquid flow

They may indicate variations in wall characteristics that affectbubble nucleation or the stability of thin liquid films round con-fined bubbles At higher quality and/or heat flux, these patternschange to a general tendency for the heat transfer coefficientheat transfer engineering vol 31 no 4 2010

Trang 13

266 T G KARAYIANNIS ET AL.

to decrease with increasing quality and to converge on a single

line that is independent of heat flux This trend cannot be fully

confirmed in these experiments with a fixed heated length for a

given diameter of tube, since high quality cannot be achieved at

low heat flux However, one can also observe that the quality at

which the heat transfer coefficient becomes independent of heat

flux and decreases with quality moves to lower values of quality

as the diameter is reduced (e.g., at approximatelyx = 0.5 for

d = 4.26 mm and x = 0.3 for d = 2.01 mm).

At very high heat flux, the heat transfer coefficient may

de-crease with heat flux The effect is particularly marked in the

2.01 mm tube, Figure 6c forq= 95–134 kW/m2 The heat flux

and quality at which this occurs both decrease with decreasing

tube diameter Shiferaw et al [55] and Huo et al [32] reported

that the tube wall temperature was highly unstable in this

par-ticular region, which could indicate the occurrence of partial

(intermittent) dry-out with a long time scale Lin et al [28] and

Sumith et al [29] observed wall temperature fluctuations that

increased as the heat flux increased This was assumed to be

related to time varying local heat transfer coefficient and local

pressure; see Lin et al [28] and Wen et al [44]

The behavior in the 0.52 mm tube at the same pressure and

mass flux is significantly different, as in Figure 6e For this

tube, the liquid-only Reynolds number is 1100, which should

correspond to laminar flow at the inlet, unlike the liquid-only Re

numbers in the 4.26, 2.88, 2.01, and 1.1 mm tubes which were

9500, 6400, 4500, and 2500, respectively There is a different

dependence of the heat transfer coefficient on heat flux and

vapor quality below and above a heat flux of 17.9 kW/m2 This

heat flux threshold coincides with the appearance of the wavy

film flow—see image 4 in Figure 5a—and the disappearance

of the small superheat that is recorded by the thermocouple in

the exit flow At the low heat fluxes, the heat transfer coefficient

does not depend on heat flux and decreases slightly with quality

However, it must be noted that the data here are limited tox <

0.15 At these low heat flux values a longer tube would be

required to reach high exit quality There is an abrupt increase

in the heat transfer coefficient and a change in its trend with

quality and heat flux at heat fluxes of 17.9 kW/m2and above At

these heat fluxes, the heat transfer coefficient initially increases

rapidly with quality, as in Figure 6e The data points for all

heat fluxes converge on approximately the same line as far as

the third thermocouple in zone I The initial variations may be

influenced by the small differences in the low inlet subcooling

In zone II, between the third and fourth thermocouples, the

heat transfer coefficient levels off at a maximum value that

depends on the heat flux This is followed by a large reduction

in heat transfer coefficient in zone III between the fourth and

fifth thermocouples After that, the data fall on another line

of increasing heat transfer coefficient that, within experimental

error, is almost independent of heat flux in zone IV At the

highest heat flux only, there is a large fall in the heat transfer

coefficient at the last measuring point at a qualityx = 0.71 This

is not reproduced in other runs at nearly the same conditions,

so it may indicate that the system is on the threshold of the

Figure 7 Heat transfer coefficient vs axial distance at mass flux 400 kg/m 2 -s and pressure 8 bar for 0.52 mm tube Heat flux values as in Figure 6e.

transient dry-out that is thought to cause the reduction in heattransfer coefficient with increasing quality in the larger tubes.When plotted againstz/L, Figure 7, the pattern of variation of

the heat transfer coefficient appears to be related to the axialpositions of the measuring stations more strongly than for thelarger tubes

Figure 8 is a plot similar to Figure 6e for the same 0.52

mm tube at a lower mass flux of 300 kg/m2-s (liquid-only Renumber 720) and a lower pressure of 6 bar, reported in Shiferaw

et al [59] It confirms that the heat transfer characteristics ofthis tube are indeed different from the larger tubes There areagain two groups of data, this time separated by a threshold heatflux of 12.5–14.8 kW/m2, which also appears to coincide withthe change of slug or confined flow to the wavy film type flowmentioned earlier at the exit from the heated section At thelow heat fluxes, the heat transfer coefficient is approximatelyindependent of heat flux although, in contrast to Figure 6e,

it initially falls significantly with quality and then exhibits aweak increase At values higher than 14.8 kW/m2, theα versus

x plot follows the same general pattern of axial development

through zones I–IV seen in Figure 6e, except that the values ofaxially increasing heat transfer coefficient in zone I depend on

Figure 8 Heat transfer coefficient vs quality at mass flux 300 kg/m 2 -s, sure 6 bar in the 0.52 mm tube.

pres-heat transfer engineering vol 31 no 4 2010

Trang 14

T G KARAYIANNIS ET AL 267

Figure 9 Heat transfer coefficient vs axial distance at mass flux 300 kg/m 2 -s,

pressure 6 bar in 0.52 mm tube (The markerx= 0 indicates the position where

saturation is achieved for q = 1.6 kW/m 2 For the rest, this happens at/before

the first thermocouple position.) The symbols and colors are the same as for

Figure 8.

heat flux and the influence of heat flux extends into zone IV,

where the heat transfer coefficient again increases axially The

test section is not long enough at low heat fluxes to show for

certain whether the data converge on a line independent of heat

flux at high quality For heat fluxes above the threshold value,

the pattern of variations inα again appears to depend on the

fraction of heated lengthz/L, as in Figure 9, but the pattern is

not exactly the same as in Figure 7 The maximum heat transfer

coefficient now occurs at thermocouple 3 instead of 4 The

subsequent reduction in zone III is less abrupt, still continuing

to thermocouple 5 There are also differences in the detail of the

pattern in zone IV If the pattern depends on the effect of local

roughness on local nucleation of bubbles, the effect appears to

be moderated by the changes in flow conditions and system

pressure

Yet another way of plotting the same data in Figures 8 and

9 is as boiling curves at measuring points 3–8, as in Figure 10;

see Shiferaw et al [59] The plots look like pool boiling curves

for increasing heat flux in a system with nucleation hysteresis

at 12.5 kW/m2 If the nucleation characteristics vary axially, it

Figure 10 Wall superheat vs heat flux at each measuring station,D= 0.52

mm, mass flux 300 kg/m 2 -s, pressure 6 bar.

is unlikely that the same threshold would apply at all stations.Alternatively, nucleation may occur at upstream sites, and down-stream positions are influenced by the growth of individual con-fined bubbles that may cover a long axial length It is impossible

to observe local nucleation in a metal tube and the observations

of flow patterns are restricted to the tube exit Confined bubbleflow with smooth liquid films round long bubbles, as assumed

in the Thome et al [52] convective model, is observed with lowheat transfer coefficients just below the threshold heat flux, as

in Figure 5a, image 3, at 400 kg/m2-s, and wavy film flow justabove the threshold The large increase in heat transfer coeffi-cient above the threshold occurs throughout the length of thetube and particularly near the inlet in zones I and II, so it cannot

be caused by a gradual progression from the exit toward theinlet of a flow regime transition at a particular quality Furtherinvestigation is required of whether nucleation is triggered at asingle site, which could exert downstream influence through thebubble frequency that is an important parameter in the Thome

et al model for convective evaporation, or at more widely tributed sites The availability of sites may become subject tolarge statistical variability as the surface area decreases withdecreasing tube diameter, as in Zhang et al [41] and Brereton

dis-et al [42]

A further special feature of the 0.52 mm tube is the decrease

in the heat transfer coefficient in zone III, commencing at aquality that increases with increasing heat flux, followed byconstant or increasing heat transfer coefficient in zone IV, with

a fall very close to the tube exit in some runs It is therefore likely

to have a different mechanism from the axial decrease in heattransfer coefficient observed in the larger tubes of this study,which commences at a quality that decreases with increasingheat flux and is then maintained to the end of the tube Because

of its association with a particular axial length of the tube,the heat transfer in zone III of the 0.52 mm tube may depend

on interactions between nucleation sites and the changing flowregime From the observations of the exit flow, as in Figure5a, the flow in zone IV is annular, with intensive disturbances

to the liquid film that decrease in scale with increasing heatflux and quality It is not possible to determine directly whethernucleation occurs in the film

Conventionally, the relative importance of nucleate boilingand convective evaporation are deduced from the dependence ofthe heat transfer coefficient on heat flux or mass flux and quality.Thome et al [52] showed that this could be misleading in smallchannels Shiferaw et al [55] found that the Thome convectivemodel, which includes cyclic dry-out of the thin films round con-fined bubbles, provided satisfactory estimates for heat transfer

in the 4.26 and 2.01 mm tubes of this study under conditions parently dominated by nucleate boiling, possibly because bothmechanisms involve the cyclic creation and evaporation to dry-ness of thin liquid films It must be noted from the flow vi-sualization by Chen et al [9], Figure 2, and for the 0.52 mmtube in this article, that the regime for which the Thome model

ap-is valid (thin, undap-isturbed films around dap-iscrete confined bles) is restricted to low qualities Convective models for highheat transfer engineering vol 31 no 4 2010

Trang 15

The experimental heat transfer coefficients in the 4.26–1.10

mm tubes all exhibit at low quality “apparently nucleate

boil-ing” characteristics, being nearly independent of quality and

mass flux, if the region of heat transfer coefficient decreasing

with quality, indicative of transient dry-out, is excluded For the

0.52 mm tube, the heat transfer coefficient is nearly independent

of quality and mass flux in zone II All these data are shown

in Figure 11 on a plot of heat transfer coefficient vs heat flux

for a mass flux of 400 kg/m2s at 8 bar pressure The data were

fitted by a power-law equation of the formα = Cq n, as is

con-ventional for nucleate boiling As mentioned earlier, this could

be due to the fact that both mechanisms (pool and transient

film evaporation) involve the cyclic creation and evaporation of

thin liquid films The exponentn is kept constant at 0.62 and

the values of the constantC for the 4.26, 2.88, 2.01, 1.10, and

0.52 mm diameter tubes are 14.3, 14.5, 16.6, 19.5, and 33.7,

respectively The heat transfer coefficients for the 4.26 and

2.88 mm diameter tubes are almost the same; the increases for

the 2.01, 1.10, and 0.52 mm tubes are 15, 35, and 134%,

respec-tively This last figure exaggerates the benefit from decreasing

diameter, because it is based on the peak values in zone I and

the improvement averaged over zones I plus II is about 90%

This approach may be useful for the design of cooling systems

for minimum temperature difference, achieved by operating at

low exit quality to avoid dry-out

The dependence of the heat transfer coefficient on mass

flux and local quality is shown in Figure 12 for a heat

flux of 54 (4.26 to 1.1 mm tubes) and 58 kW/m2

(0.52-mm tube) and 8 bar pressure At low qualities, the

ap-proximately constant values of the heat transfer coefficient

are almost independent of mass flux within the

experimen-tal uncertainty for the four larger diameter tubes For the

4.26 mm tube, afterx = 0.15, the heat transfer coefficient

de-creases slightly with mass flux, which could be related to an

influence on film thickness However, this is not repeated in the

2.88 to 1.1 mm tubes As also noted earlier, further experiments

are required to resolve the issue, using longer heated lengths toachieve larger exit qualities, subject to any limitations imposed

by pressure drop The results for the smallest diameter tube inFigure 12e are clearly different There is a significant effect ofmass flux in zone IV (increasing trend of heat transfer coeffi-cient with quality) In this region, the heat transfer coefficientincreases with increasing mass flux and, as seen in Figure 6e,there is no obvious effect of heat flux especially at high quality.This, with the observations at the visualization section, appar-ently supports the previous speculation that convective evapora-tion of the annular flow dominates the heat transfer mechanism

at high quality (Lin et al [28], Sumith et al [29], and Saitoh et

al [30]) However, when plotted against axial distancez/L in

Figure 13, the data for the 0.52 mm tube collapse onto a gle line independent of mass flux but with large axial variations,suggesting that time-averaged quality is not the controlling vari-able By contrast, the data for the 1.1 mm tube follow a line ofnearly constantα at high mass flux, with lines of decreasing αbranching off at points that move toward the tube inlet as themass flux is reduced It appears that quality is the relevant vari-able for the assumed process of transient dry-out in the largertubes of this study

sin-The influence of system pressure is illustrated in Figure 14

by plots of heat transfer coefficient vs quality for all the tubes

at the same mass flux of 400 kg/m2-s and heat flux of 54 kW/m2(4.26, 2.88, 2.01, and 1.10 mm tubes) and 58 kW/m2 (0.52

mm tube) (These are almost the same as plots ofα vs z/L.)

For qualityx < 0.3, the heat transfer coefficient increases with

system pressure for the 4.26 to 1.10 mm tubes The effects ofpressure at higher qualities at various values of heat flux andmass flux were reported in Shiferaw et al [55] For the 4.26

mm diameter tube, the effect of pressure was less significant

at higher qualities (x > 0.5), while for the 2.01 mm diameter

tube there was a rather uniform increase in the coefficient withpressure throughout the experimental range of quality (x < 0.7).

Again, the 0.52 mm tube behaves differently, as in Figure14e Increasing pressure causes a much larger increase in theheat transfer coefficient at smallx in zones I and II, compared to

zone IV at higherx, and the decrease in heat transfer coefficient

in zone III becomes sharper There is a drop in heat transfercoefficient at the last measuring point for 8 and 10 bar pressure,which might indicate the onset of thin film dry-out

DISCUSSION

This article is based on flow visualization studies and heattransfer measurements obtained over a period of 6 years for fivetubes of different diameters Some of the data are new and somehave been published previously When some data sets were ex-tended in range, the heat transfer coefficients were found to bereproducible within±5%, even after intervals of 3 years Thedata for the 4.26 to 1.1 mm tubes have some features that areconventionally and appropriately presented as functions of localquality, combined with a weak dependence on the axial positionheat transfer engineering vol 31 no 4 2010

Trang 16

T G KARAYIANNIS ET AL 269

Figure 12 Effect of mass flux on heat transfer coefficient versus quality at heat flux (q= 54 and 58 kW/m 2 ) and pressure (P= 8 bar): (a) 4.26 mm; (b) 2.88 mm; (c) 2.01 mm; (d) 1.1 mm; (e) 0.52 mm.

within a particular test section This axial dependence was much

stronger in the data for the 0.52 mm tube These axial patterns

are not present in single-phase tests, so they are consequences

of boiling Very recent tests on this tube have shown that the

patterns tend to be stable during a series of tests on a particular

day but there may be a different pattern on other days In the

parametric studies of heat flux, mass flux, and pressure reported

in this article, examples have been chosen from tests performed

at similar times Similar variability on different days was

ob-served in the “apparently nucleate” regime during flow boiling

of water in a large (9.6 mm) tube but not in the “apparentlyconvective” regime of Kenning and Cooper [16] In that study,polishing the tube surface also modified the nucleate but not theconvective boiling regime Surface roughness has a large influ-ence on bubble nucleation in pool boiling, so axial variations insurface roughness may influence local nucleation The influence

of surface conditions on boiling in small metal tubes has as yetreceived little attention Surface roughness may also affect aparameter in the convective boiling model of Thome et al [52]for microchannels, namely, the minimum stable thickness of theheat transfer engineering vol 31 no 4 2010

Trang 17

270 T G KARAYIANNIS ET AL.

Figure 13 Effect of mass flux on the heat transfer coefficient versus axial

distance at 8 bar: (a) 0.52 mm tube at 58 kW/m 2 and (b) 1.1 mm tube at 54

kW/m 2

evaporating liquid film round confined bubbles Shiferaw et al

[55] showed that the predictions of the Thome et al model were

improved if the experimental measurements of roughness were

used instead of the recommended film thickness The surface

roughness of samples from the four larger tubes used in this

study was measured after sectioning by scanning in an axial

direction with a conventional contact stylus; values are given in

Table 1 The surface roughness of the 0.52 mm tube was

ob-tained from a three-dimensional (3-D) sample, captured using a

high-resolution non-contact probe

In the experiments described here, and in those performed

earlier by Chen et al [9], flow patterns were observed at the

exit of the test section Observations within a tube are possible

for transparent tubes with transparent thin-film heaters, as in the

experiments of Owhaib et al [48], but the nucleation

charac-teristics are different and it is difficult to obtain simultaneous

accurate measurements of the wall temperature The flow

pat-terns observed at the exit from the 0.52 mm tube were certainly

different from those observed earlier in the relatively larger

di-ameter tubes (4.26–1.1 mm) by Chen et al [9] These differences

include the absence of dispersed bubble flow and the

appear-ance of a transitional wavy film flow Thus, there were further

differences between the flow patterns leaving the 2.88 and4.26 mm diameter tubes and those from the 2.01 and 1.1 mmtubes, which exhibited confined flow, slimmer vapor slugs, thin-ner liquid films, and smoother vapor–liquid interfaces Thesedifferences coincided with the progressive transition to higherheat transfer coefficients in the 2.01 and 1.10 mm tubes Usingthe confinement number (Cornwell and Kew [4]), the deviationfrom large-tube characteristics should be observed at diame-ters of 1.4 to 1.7 mm at 6–14 bar pressure for R134a, which isroughly in agreement with the present heat transfer results andflow visualization observations “Small-tube characteristics” in1.1 mm tubes were reported in the previous studies of Damian-ides and Westwater [13] and Mishima and Hibiki [7]

Flow maps such as Figure 2, based on observations at theexit from the 4.26 to 1.10 mm tubes, show that, at the lowmass fluxes covered in the present heat transfer tests, the tran-sition to annular flow shifts to higher qualities approaching

x ≈ 0.5 While the information on flow regimes cannot betransferred with certainty to upstream locations, it is likelythat slug/churn flow is the typical flow pattern in the region

of near-uniform high heat transfer coefficient dependent marily on heat flux This could be at least one of the rea-sons for the increase in the heat transfer coefficient with areduction in the channel size The relative importance of nu-cleate and convective boiling in this region is still unclear.However, there are claims that suggest that, for small pas-sages, the same behavior, i.e., uniform heat transfer coeffi-cient dependent on heat flux and independent of quality, can

pri-be explained if transient evaporation of the thin liquid filmsurrounding elongated bubbles, without nucleate boiling con-tribution, is the dominant heat transfer mechanism (Thome et

al [52]) One may argue that the variations in heat transfercoefficient with axial position, evident in Figure 6, especiallyfor the larger tubes, may indicate some dependence on nucle-ate boiling Kenning and Yan [66] observed cyclic triggering

of nucleate boiling in smooth films around confined bubbles inwater associated with pressure fluctuations This needs furtherinvestigation

The heat transfer results of the smallest diameter tube (0.52mm) demonstrated different characteristics than the rest of thetubes, particularly at the high quality region It is the only tubefor which the incoming liquid flow is laminar, and this mayinfluence the initiation of confined bubble (slug) flow Unlikethe larger tubes that were examined in this study, which exhibitdry-out phenomena at high quality as the heat flux increaseswith a drop of the heat transfer coefficient with quality, a mono-tonic increase in heat transfer coefficient was observed near theexit for the smallest diameter tube This could be related tolaminar flow and domination of surface tension force over mo-mentum, providing more uniform liquid film thickness alongthe circumference, with less interfacial waves and disturbances,which improves wetting of the wall (Shiferaw et al [59]) Inaddition, the dependence of the heat transfer coefficient on axialposition is much stronger in the 0.52 mm tube, as in Figure

13, extending to high quality in the annular flow regime Theheat transfer engineering vol 31 no 4 2010

Trang 18

T G KARAYIANNIS ET AL 271

Figure 14 Effect of pressure on heat transfer coefficient vs quality,G= 400 kg/m 2 s,q= 54 and 58 kW/m 2 : (a) 4.26 mm; (b) 2.88 mm; (c) 2.01 mm; (d) 1.1 mm; (e) 0.52 mm.

experiments in this tube are currently being repeated for the

complete range of variables and further confirmation of these

characteristics

These observations indicate additional changes as the size

diminishes further into microscales In general, the complex

de-pendence of the heat transfer rate on various parameters suggests

the difficulty of interpreting the heat transfer mechanisms using

simple conventional terms and the challenge of heat transfermodeling

CONCLUSIONS

Flow boiling patterns and heat transfer results with R134aand five tubes of diameter 4.26, 2.88, 2.01, 1.10, and 0.52 mmheat transfer engineering vol 31 no 4 2010

Trang 19

272 T G KARAYIANNIS ET AL.

were presented in this article It was anticipated that the wide

range of data at different diameters could be used to identify

the threshold(s) where the small or micro diameter effects

be-come significant The major conclusions that can be drawn from

the current part of this long-term study are as follows:

1 In the 4.26 and 2.88 mm diameter tubes, the heat transfer

coefficient increases with heat flux and system pressure, but

does not change with vapor quality when the quality is less

than about 40% to 50%, for low heat flux The boundary

moves to 20–30% for the 2.01 and 1.10 mm diameter tubes

The actual quality values depend also on the heat flux In this

region, there is no significant difference in the magnitude of

the heat transfer coefficient of the 4.26 and 2.88 mm tubes

However, there is an increase of 15% and 35% when the tube

diameter is reduced to 2.01 and 1.10 mm, respectively

2 The heat transfer coefficient behavior of the tubes (4.26–

1.1 mm) at low quality could be interpreted as the evidence

that nucleate boiling is the dominant heat transfer

mecha-nism However, transient evaporation of the thin liquid film

surrounding elongated bubbles, which is a dominant flow

pattern in small passages, without a nucleate boiling

con-tribution, may also result in similar heat transfer coefficient

dependence and magnitude For higher vapor qualities, the

heat transfer coefficient becomes independent of heat flux

and decreases with vapor quality This could be caused by

partial (intermittent) dry-out in the convection-dominated

region This leads to the design recommendation that exit

qualities be kept low (Zhang et al [67, 68])

3 Chen et al [9] concluded that flow patterns for the 4.26 and

2.88 mm diameter tubes exhibit flow pattern characteristics

similar to those of “normal” diameter tubes, while “small

tube characteristics,” e.g., the appearance of confined flow,

were observed when the tube diameter was reduced to 2.01

mm and further to 1.10 mm This is consistent with a

crite-rion based on the ratio of surface tension and gravitational

forces The change in behavior may be progressive, rather

than occurring at a sharp threshold

4 The heat transfer data suggest that there is some deviation

from “normal” behavior even for the 4.26 mm tube, because

the expected increase in heat transfer coefficient with

in-creasing high quality was replaced by a decrease attributed

to intermittent dry-out This may indicate that film stability

in the heated zone depends on the ratio of surface tension to

other forces This cannot be detected by flow visualization at

the exit from the test section

5 As the tube diameter decreased further down to 0.52 mm,

different flow and heat transfer characteristics were

ob-served, indicating a possible further change as the size

dimin-ished These include: (a) The flow patterns observed in the

0.52 mm tube are different, i.e., absence of dispersed bubble

flow, and the appearance of a wavy film type flow that leads

into annular flow (b) The dependence of the heat transfer

co-efficient on quality, heat flux, and mass flux changes sharply

in character at a threshold value of heat flux In the low heatflux region, there is no significant effect of heat flux but theheat transfer coefficient decreases (at low mass flux and pres-sure) or remains constant (at higher mass flux and pressure),then increases gradually with quality At moderate and highheat flux, in the front part of the channel, the heat transfercoefficient increases with increasing heat flux and also de-pends in a complex way on quality It reaches a maximum at

an intermediate quality, which might be caused by transientpartial dry-out or dry patches in the confined bubble regime

At higher quality, toward the test section exit, the heat transfercoefficient gradually increases again with quality but there

is no clear effect of heat flux The heat transfer coefficientalso increases with mass flux in this region According to theconventional interpretation, this is evidence for a convectiveboiling dominant heat transfer mechanism in annular flow

An alternative plotting of heat transfer coefficient suggeststhat it is more dependent on the surface conditions associ-ated with particular axial positions than on quality Thesemight influence bubble nucleation or the stability of thin liq-uid films The slender evidence as yet available may indicatesome variability in the activation of the small population ofnucleation sites available in a channel of small surface area.The results of the 0.52 mm tube are currently being repeated

The complexity of interpreting heat transfer tics and understanding the prevailing mechanisms and, con-sequently, the difficulty of developing generalized models areverified by the work presented in this article Phenomenologicalmodels that are based on the local flow structure may be de-veloped for clearly specified ranges Therefore, it is important

characteris-to identify the range of applicability of dominant flow regimes.Current results also indicate that much more research is needed

to understand the different characteristics associated with crotubes and channels

Trang 20

T G KARAYIANNIS ET AL 273

T temperature, K

t time (s)

U gs superficial gas velocity, m/s

U ls superficila liquid velocity, m/s

[1] Mehendale, S S., Jacobi, A M., and Shah, R K., Fluid Flow

and Heat Transfer at Micro- and Meso-Scales with Application to

Heat Exchanger Design, Applied Mechanics Reviews, vol 53, no.

7, pp 175–193, 2000

[2] Kandlikar, S G., and Grande, W J., Evolution of Microchannel

Flow Passages—Thermohydraulic Performance and Fabrication

Technology, Heat Transfer Engineering, vol 25, no 1, pp 3–17,

2003

[3] Brauner, N., and Moalem-Maron, D., Identification of the Range

of Small Diameter Conduits, Regarding Two-Phase Flow

Pat-tern Transitions, InPat-ternational Communications in Heat and Mass

Transfer, vol 19, pp 29–39, 1992.

[4] Cornwell, P A K., and Kew, P A., Boiling in Small Parallel

Channels, in Energy Efficiency in Process Technology, ed P A.

Pilavachi, Elsevier Applied Science, London, pp 624–638, 1993

[5] Thome, J R., Boiling in Microchannels: A Review of Experiment

and Theory, International Journal of Heat and Fluid Flow, vol.

25, pp 128–139, 2004

[6] Barnea, D., Luninski, Y., and Taitel, Y., Flow Pattern in Horizontal

and Vertical Two Phase Flow in Small Diameter Pipes, Canadian Journal of Chemical Engineering, vol 61, no 5, pp 617–620,

1983

[7] Mishima, K., and Hibiki, T., Some Characteristics of Air-Water

Two-Phase Flow in Small Diameter Vertical Tubes, tional Journal of Multiphase Flow, vol 22, no 4, pp 703–712,

Interna-1996

[8] Kew, P A., and Cornwell, K., Correlations for the Prediction

of Boiling Heat Transfer in Small Diameter Channels, plied Thermal Engineering, vol 17, no 8–10, pp 705–715,

[10] Serizawa, A., Feng, Z., and Kawara, Z., Two Phase Flow in

Mi-crochannels, Experimental Thermal Fluid Sciences, vol 26, pp.

[13] Damianides, D A., and Westwater, J W., Two-Phase Flow

Pat-terns in a Compact Heat Exchanger and in Small Tubes, Second

UK National Conference on Heat Transfer, vol 11, Sessions 4A–

6C, pp 1257–1268, 1988

[14] Lin, S., Kew, P A., and Cornwell, K., Two-Phase Flow Regimes

and Heat Transfer in Small Tubes and Channels, Heat Transfer

1998, Proceedings of 11th IHTC, vol 2, August 23–28, Kyongju,

Korea, pp 45–50, 1998

[15] Revellin, R., and Thome, J R., A New Type of Diabatic Flow

Pattern Map for Boiling Heat Transfer in Microchannels, Journal

of Micromechanics and Microengineering, vol 17, pp 788–796,

2007

[16] Kenning, D B R., and Cooper, M G., Saturated Flow Boiling of

Water in Vertical Tubes, International Journal of Heat and Mass Transfer, vol 32, pp 445–458, 1989.

[17] Lazarek, G M., and Black, S H., Evaporative Heat Transfer,Pressure Drop and Critical Heat Flux in a Small Vertical Tube

With R-113, International Journal of Heat and Mass Transfer,

[20] Bao, Z Y., Fletcher, D F., and Haynes, B S., Flow Boiling Heat

Transfer of Freon R11 and HCFC123 in Narrow Passages, national Journal of Heat and Mass Transfer, vol 43, pp 3347–

Inter-3358, 2000

[21] Yu, W., France, D M., Wambsganss, M W., and Hull, J R., Phase Pressure Drop, Boiling Heat Transfer, and Critical Heatheat transfer engineering vol 31 no 4 2010

Trang 21

Two-274 T G KARAYIANNIS ET AL.

Flux to Water in a Small Diameter Horizontal Tube, International

Journal of Multiphase Flow, vol 28, pp 927–941, 2002.

[22] Fujita, Y., Yang, Y., and Nami, F., Flow Boiling Heat Transfer and

Pressure Drop in Uniformly Heated Small Tubes, in Proc 12th

International Heat Transfer Conf., Grenoble, vol 3, pp 743–748,

2002

[23] Carey V P., Tervo, P., and Shullenberger, K., Partial Dryout in

Enhanced Evaporator Tubes and Its Impact On Heat Transfer

Per-formance, SAE Technical Paper 920551, 1992 Available online

at http://www.SAE.org

[24] Oh, H K., Katsuta, M., and Shibata, K., Heat Transfer

Character-istics of R134a in a Capillary Tube Heat Exchanger, in Proc 11th

IHTC, vol 6, pp 131–136, 1998.

[25] Lee, H J., and Lee, S Y., Heat Transfer Correlation for Boiling

Flows in Small Rectangular Horizontal Channels With Low

As-pect Ratios, International Journal of Multiphase Flow, vol 27,

pp 2043–2062, 2001

[26] Qu, W., and Mudawar, I., Flow Boiling Heat Transfer in Two

Phase Microchannel Heat Sinks I Experimental Investigation

and Assessment of Correlation Methods, International Journal of

Heat and Mass Transfer, vol 46, pp 2755–2771, 2003.

[27] Kuznetsov, V V., and Shamirzaev, A S., Two Phase Flow Pattern

and Flow Boiling Heat Transfer in Non-Circular Channel With

a Small Gap, Two-Phase Flow Modeling and Experimentation,

Pisa, Italy, vol 1, pp 249–253, 1999.

[28] Lin, S., Kew, P A., and Cornwell, K., Two-Phase Heat Transfer

to a Refrigerant in a 1 mm Diameter Tube, International Journal

of Refrigeration, vol 24, pp 51–56, 2001.

[29] Sumith, B., Kaminaga, F., and Matsumura, K., Saturated Flow

Boiling of Water in a Vertical Small Diameter Tube, Experimental

Thermal Fluid Sciences, vol 27, pp 789–801, 2003.

[30] Saitoh, S., Daiguji, H., and Hihara, E., Effect of Tube Diameter

on Boiling Heat Transfer of R134a in Horizontal Small Diameter

Tubes, International Journal of Heat and Mass Transfer, vol 48,

pp 4973–4984, 2005

[31] Owhaib, W., and Palm, B., Flow Boiling Heat Transfer in a

Ver-tical Circular Microchannel Tube, Eurotherm Seminar, no 72,

Valencia, Spain, 2003

[32] Huo, X., Tian, Y S., and Karayiannis, T G., R134a Flow

Boil-ing Heat Transfer in Small Diameter Tubes, Advances in

Com-pact Heat Exchangers, R T Edwards, Philadelphia, pp 95–111,

2007

[33] D´ıaz, M C., and Schmidt, J., Experimental investigation of

tran-sient boiling heat transfer in microchannels, Int Journal of Heat

and Fluid Flow, vol 28, pp 95–102, 2007

[34] Xu, J., Shen, S., Gan, Y., Li, Y., Zhang, W., Su, Q., Transient

flow pattern based microscale boiling heat transfer mechanisms,

Journal of Micromechanics and Microengineering, vol 15, pp

1344-1361, 2005

[35] Lie, Y M., Su, F Q., Lai, R.L., Lin, T F., Experimental

Study of Evaporation Heat Transfer Characteristics of

Refrig-erants R-134a and R-407C in Horizontal Small Tubes,

Interna-tional Journal of Heat and Mass Transfer, vol 49, pp 207–218,

2006

[36] Agostini, B., and Thome, J R., Comparison of an Extended

Database of Flow Boiling Heat Transfer Coefficient in

Multi-Microchannel Elements With the Three-Zone Model, in ECI

In-ternational Conference on Heat Transfer and Fluid Flow in

Mi-croscale, Castelvecchio Pascoli, Italy, 2005.

[37] Yen, T., Kasagi, N., and Suzuki, Y., Forced Convective Boiling

Heat Transfer in Microtubes at Low Mass and Heat Fluxes, ternational Journal of Multiphase Flow, vol 29, pp 1771–1792,

In-2003

[38] Hapke, I., Boye, H., and Schmidt, J., Onset of Nucleate Boiling

in Minichannels, International Journal of Thermal Sciences, vol.

[40] Peng, X F., Hu, H Y., and Wang, B X., Boiling Nucleation

During Liquid Flow in Microchannels, International Journal of Heat and Mass Transfer, vol 41, pp 101–106, 1998.

[41] Zhang, L., Koo, J., Jiang, L., Goodson, K E., Santiago, J G.,and Kenny, T W., Study of Boiling Regimes and Transient Signal

Measurements in Microchannels, Proc Transducers ’01, Munich,

Germany, pp 1514–1517, 2001

[42] Brereton, G J., Crilly, R J., and Spears, J R., Nucleation in

Small Capillary Tubes, Chemical Physics, vol 230, pp 253–265,

1998

[43] Yan, Y Y., and Lin, T F, Evaporation Heat Transfer and

Pres-sure Drop of Refrigerant R-134a in a Small Pipe, International Journal of Heat and Mass Transfer, vol 41, pp 4183–4194,

[46] Yan, Y., and Kenning, D B R., Pressure and Temperature

Fluc-tuations During Boiling in Narrow Channel, Eurotherm 62: Heat Transfer in Condensation and Evaporation, Grenoble, pp 107–

1223, 1998

[47] Kew, P A., and Cornwell, K., On Pressure Fluctuations During

Boiling in Narrow Channels, 2nd European Thermal-Science and 14th UIT National Heat Transfer Conference, Rome, pp 1323–

1327, 1996

[48] Owhaib, W., Martin-Callizo, C., and Palm, B., Evaporative Heat

Transfer in Vertical Circular Microchannels, Applied Thermal gineering, vol 24, pp 1241–1253, 2004.

En-[49] Kuwahara, K., Koyama, S., and Hashimoto, Y., Characteristics ofEvaporation Heat Transfer and Flow Pattern of Pure Refrigerant

HFC134a in a Horizontal Capillary Tube, in Proc 4th KSME Thermal Engineering Conference, pp 385–390, 2000.

JSME-[50] Baird, J R., Bao, Z Y., Fletcher, D F., and Haynes, B S., LocalFlow Boiling Heat Transfer Coefficients in Narrow Conduits, in

Boiling: Phenomena and Engineering Applications, Multiphase

Science and Technology, vol 12, no 3–4, pp 129–144, 2000.[51] Khodabandeh, R., Influence of Channel Diameter on Boiling HeatTransfer in a Closed Advanced Two-Phase Thermosiphon Loop,

Proc 5th International Boiling Conference, Montego Bay,

Ja-maica, 2003

[52] Thome, J R., Dupont, V., and Jacobi, A M, Heat Transfer Modelfor Evaporation in Microchannels, Part I: Presentation of theheat transfer engineering vol 31 no 4 2010

Trang 22

T G KARAYIANNIS ET AL 275

Model, International Journal of Heat and Mass Transfer, vol.

47, pp 3375–3385, 2004

[53] Dupont, V., Thome, J R., and Jacobi, A M., Heat Transfer Model

for Evaporation in Microchannels, Part II: Comparison With the

Database, International Journal of Heat and Mass Transfer, vol.

47, pp 3387–3401, 2004

[54] Dupont, D., and Thorne, J R., Evaporation in Microchannels:

Influence of the Channel Diameter on Heat Transfer, Journal of

Microfluidics and Nanofluidics, vol 1, no 2, pp 119–125, 2005.

[55] Shiferaw, D., Huo, X., Karayiannis, T G., and Kenning, D B

R., Examination of Heat Transfer Correlations and a Model for

Flow Boiling of R134a in Small Diameter Tubes, International

Journal of Heat and Mass Transfer, vol 50, pp 5177–5193,

2007

[56] Blasius, H., Das Ahnlichkeitsgesetz bei Reibungsvorgangen in

Flussigkeiten, Forschungs-Arbeit des lgenieur-wesens, vol 131,

no 1, 1913

[57] Dittus, F W., and Boelter L M K., Heat Transfer in Automobile

Radiators of Tubular Type, University California, Berkeley, Publ.

Eng 2/13, pp 443–461, 1930

[58] Petukhov, B S., Heat Transfer and Friction in Turbulent Pipe Flow

With Variable Physical Properties, in Advances in Heat Transfer,

vol 6, Academic Press, New York, pp 503–564, 1970

[59] Shiferaw, D., Mahmoud, M M., Karayiannis, T G., and Kenning,

D B R., Experimental Flow Boiling Study in a 0.52 mm

Diam-eter Vertical Tube Using R134a, 5th European Thermal-Sciences

Conference, Eindhoven, the Netherlands, 2008.

[60] Gnielinski, V., VDI-W¨armeatlas, Springer-Verlag, Berlin, 1997.

[61] Adams, T M., Abdel-khalik S I., Jeter S M., and Qureshi Z H.,

An Experimental Investigation of Single-Phase Forced

Convec-tion in Micro-Channels, InternaConvec-tional Journal of Heat and Mass

Transfer, vol 41, no 6–7, pp 851–857, 1998.

[62] Choi, S B Barron, R F., and Warrington, R O., Fluid Flow and

Heat Transfer in Microtubes, Micromechanical Sensors,

Actua-tors and Systems, ASME DSC, vol 32, pp 123–128, 1991.

[63] Shiferaw, D., Karayiannis, T G., and Kenning, D B R., Flow

Boiling in a 1.1 mm Vertical Tube With R134a: Experimental

Results and Comparison With Model, International Journal of

Thermal Sciences, doi:10.1016/j.ijthermalsci.2008.02.009, 2008.

[64] Shiferaw, D., Two-Phase Flow Boiling in Small- to

Micro-Diameter Tubes, Ph.D Thesis, Brunel University, 2008

[65] Revellin, R., Dupont, V., Ursenbacher, T Thome, J R., and Zun,

I., Characterization of Diabatic Two-Phase Flows in

Microchan-nels: Flow Parameter Results for R-134a in a 0.5 mm Channel,

International Journal of Multiphase Flow, vol 32, pp 755–774,

2006

[66] Kenning, D B R., and Yan, Y., Saturated Flow Boiling of Water

in a Narrow Channel: Experimental Investigation of Local

Phe-nomena, IChemE Trans A, Chemical Engineering Research and

Design, vol 79, pp 425–436, 2001.

[67] Zhang, L., Goodson, K E., and Kenny, T W., Silicon

Microchan-nel Heat Sinks, Theories and Phenomena, Springer, Berlin, chaps.

5–7, 2004

[68] Zhang, L., Wang, E N., Goodson, K E., and Kenny, T W., Phase

Change Phenomena in Silicon Microchannels, International nal of Heat and Mass Transfer, vol 48, pp 1572–1582, 2005.

Jour-Tassos G Karayiannis is a professor of thermal

en-gineering in the School of Enen-gineering and Design of Brunel University, UK, where he is co-director of the Centre for Energy and Built Environment Research.

He obtained a B.Sc in mechanical engineering from City University (UK) in 1981 and a Ph.D from the University of Western Ontario (Canada) in 1986 He has carried out research in single-phase heat transfer, enhanced heat transfer, and thermal systems He has been involved with research in two-phase flow and heat transfer for about 20 years He is a fellow of the Institution of Mechanical Engineers and the Institute of Energy.

Dereje Shiferaw received his M.Sc in sustainable

energy engineering in 2004 from the Royal Institute

of Technology (Sweden) and his Ph.D from Brunel University (UK) in 2008 He won an award for the best master’s thesis from the Swedish Center for Nu- clear Research His research interests include single- and two-phase flow heat transfer in microchannels, nanofluids, compact heat exchangers, cooling of elec- tronics, and renewable energy systems.

David Kenning graduated in mechanical sciences

from Cambridge University in 1957 and worked for the UK Atomic Energy Authority for 3 years before returning to Cambridge to start his career of research

on multiphase flows and boiling heat transfer He joined Oxford University in 1963 and was a univer- sity lecturer in engineering science and tutorial fellow

of Lincoln College from 1967 until his official ment in 2003 He then joined the research group of Professor Tassos Karayiannis as a visiting professor, first at London South Bank University and now at Brunel University.

retire-Vishwas V Wadekar is Technology Director, HTFS

Research, at AspenTech Ltd, UK In addition to aging HTFS research, he chairs the HTFS Industrial Review Panel on Compact Heat Exchangers He has authored or co-authored a number of technical and re- search papers in the area of compact heat exchangers, multiphase flow heat and mass transfer, and boiling.

man-He has held visiting faculty positions in a number of universities United Kingdom and abroad Currently

he serves as a visiting professor at the University of Hamburg in Germany and the Lund Institute of Technology in Sweden He obtained his B.Chem.Eng and Ph.D degrees from UDCT, Bombay University.

He is a member of AIChE and is actively involved in organizing technical sions at AIChE and ASME conferences He is also a member of the Eurotherm

ses-Committee and an associate editor of Heat Transfer Engineering.

heat transfer engineering vol 31 no 4 2010

Trang 23

CopyrightC Taylor and Francis Group, LLC

ISSN: 0145-7632 print / 1521-0537 online

DOI: 10.1080/01457630903311694

A New Method for Determination

of Flow Boiling Heat Transfer

Coefficient in Conventional-Diameter Channels and Minichannels

DARIUSZ MIKIELEWICZ

Faculty of Mechanical Engineering, Gdansk University of Technology, Gdansk, Poland

Presented in this article are considerations regarding modeling of flow boiling in conventional, small-diameter channels and

minichannels A concise survey of available methods for prediction of heat transfer coefficient in saturated boiling regime is

given and in that light a modified author’s own model is presented The presented model, contrary to other approaches, finds

application in the cases of both conventional and small-diameter channels The results of calculations are compared with

some experimental data available from literature on conventional-size tubes and also minichannels Obtained agreement is

satisfactory.

INTRODUCTION

Boiling heat transfer, as one of the most efficient techniques

for removing high heat fluxes, has been studied and applied in

practice for a very long time Nowadays, rapid development

of practical engineering applications for devices,

micro-systems, advanced material designs, manufacturing of compact

heat exchangers, high-capacity micro heat pipes for spacecraft

thermal control, and electronic microchips increases the demand

for better understanding of small and micro-scale transport

phe-nomena The last decade of the 20th century witnessed rapid

progress in the research into micro- and nano-scale transport

phenomena, which bore important applications in modern

tech-nologies such as microelectronics

Despite numerous applications, the theoretical approaches

to modeling of flow boiling still require substantial progress, as

determination of heat transfer and pressure losses is done mostly

by means of empirical correlations Their drawback is that they

feature fluid-dependent coefficients and thus are not general

Such correlations must be therefore used with special care and

precautions, only in the range of conditions that such

correla-tions have been developed for That fact also disables direct

ap-plication of correlations developed for conventional channels to

Address correspondence to Professor Dariusz Mikielewicz, Faculty of

Me-chanical Engineering, Gdañsk University of Technology, ul G Narutowicza

11/12, 80-952 Gdañsk, Poland E-mail: Dariusz.Mikielewicz@pg.gda.pl

small-diameter channels Recently there has also been progressattained using structure-dependent modeling using flow maps,which is, however, tedious and not very convenient for engi-neering applications

Following a brief presentation of available approaches tomodeling of flow boiling, a model is presented here that is de-veloped on the basis of considerations of dissipation of energy

in the flow and is deemed to be applicable to both tional and small-diameter channels, laminar and turbulent flowregimes, and flow with bubble generation and without it

conven-REVIEW OF EXISTING FLOW BOILING CORRELATIONS

The topic of flow boiling predictions has been scrutinized forover half a century, as the interest in that kind of heat transferstarted in the early 1960s In the present article it is not theintention of the author to provide a survey of all available meth-ods for that purpose, but only to indicate the major approaches

to modeling of flow boiling heat transfer in conventional andsmall diameter channels For an extensive literature survey offlow boiling in conventional-size channels the reader is referred

to Thome [1] or for small-diameter channels to Bergles et al [2],Kandlikar [3], or Thome [4] In general, all existing approachesare either the empirical fits to the experimental data, or form276

Trang 24

D MIKIELEWICZ 277

an attempt to combine two major influences to heat transfer,

namely, the convective flow boiling without bubble generation

and nucleate boiling Generally that is done in a linear or

nonlin-ear manner Alternatively, there is a group of modern approaches

based on models that start from modeling a specific flow

struc-ture and in such a way postulate more accurate flow boiling

models, usually pertinent to slug and annular flows

The empirical correlations suggested to date are based on

reduction of a restricted number of authors’ own experimental

data or form a generalization of a greater number of

experimen-tal data from various authors In the latter case correlations are

usually of less accuracy in predicting the heat transfer

coeffi-cient, or the pressure drop, due to the fact that each experiment

contributes with its own systematic measurement error;

how-ever, such correlations are of more general character All such

empirical correlations, however, are the fits to experimental data,

which restricts their generality In the case of a lack of generation

of bubbles the experimental data in conventional-size tubes, i.e.,

having diameters greater than 3 mm [3], are usually modeled as

a function of the Martinelli parameter Xttin the form:

αTP

αL = a (Xtt) b

(1)

Dengler and Addoms [5] suggested values of a and b to be a=

3.5 and b= –0.5, respectively, whereas Guerrieri and Talty [6]

determined as a = 3.4 and b = –0.45 In the case when bubble

generation in the flow is present, such an approach encounters

some limitations, and in order to alleviate that an approach

based on incorporation of the Bond number, Bo= q w /(Gh lg),

into the correlation is often used A general form of heat transfer

coefficient with account of bubble generation yields:

αTP

αL = a

Schrock and Grossman [7] recommended a correlation where

coefficients a, b, and m assume values of a= 7400, b= 0.66,

and m = 0.00015 Collier and Pulling [8] suggested another

set of coefficients in Eq (2): a = 6700, b = 0.66, and m =

0.00035 On the basis of that approach several other correlations

for conventional channels have been developed, and here is just

a mention of those due to Shah [9], Kandlikar [10], and Gungor

and Winterton [11]

The most popular approach, however, to model flow boiling

is to present the resulting heat transfer coefficient in terms of

a combination of nucleate boiling heat transfer coefficient and

convective boiling heat transfer coefficient:

αTP= [(αcbF)n+ (αPBS)n]1/n (3)

where αPBis the pool boiling heat transfer coefficient, and αcb

the liquid convective heat transfer coefficient, which can be

evaluated using, for example, the Dittus–Boelter type of

corre-lation Exponent n is an experimentally fitted coefficient

with-out recourse to any theoretical foundations Function S is the

so-called suppression factor, which accounts for the fact that

to-gether with increase of vapor flow rate the effect related to forced

convection increases, which on the other hand impairs the tribution from nucleate boiling, as the thermal layer is reduced

con-The parameter F accounts for the increase of convective heat

transfer with increase of vapor quality That parameter alwaysassumes values greater than unity, as flow velocities in two-phase flow are always greater than in the case of single-phaseflow The approach represented by Eq (3) is usually dedicated

to Rohsenow [12], who suggested a linear superposition with

n= 1, which has been later modified by Chen [13], who

incor-porated the suppression and enhancement functions, S and F ,

respectively The correlation due to Chen is used up to date with

a significant appreciation in the case of flows in size tubes It was also Kutateladze [14] who recommended asuperposition approach, but combined in a geometrical rather

conventional-than linear manner with the value of exponent n = 2 A ilar summative nonlinear approach was recommended later by

sim-Steiner and Taborek [15] with n = 3 There is also an issue

of the choice of appropriate correlation selection for tion of pool boiling heat transfer coefficient, as Chen [13] usedthe model due to Forster and Zuber [16], whereas later studiestend to use rather the more general correlation due to Cooper[17], which enables calculations of pool boiling heat transfercoefficient for different modern fluids

calcula-Kattan et al [18] concluded that only the models based ondistinguishing between flow regimes should be genuinely con-sidered for a general use in prediction of heat transfer coefficient

in channels A model for flow boiling in horizontal tubes hasbeen developed by him based on a flow map It was assumed that

in the flow the upper part of the tube inside is usually contactingvapor and the remaining part is flooded with liquid The heattransfer coefficient for such case was taking that fact into ac-count and the heat transfer for these two regions was evaluated

in the following manner:

αTPB= [R · θdryαv + R(2π − θdry)αwet]/2πR (4)The heat transfer coefficient for a wetted part of the tube, αwet,

is to be obtained from the form of Eq (3) with n = 3 Thenucleate boiling heat transfer coefficient αPBis calculated fromthe Cooper’s [17] correlation, whereas αcbis from this author’sown empirical correlation Some refinement to that approachwas introduced by Wojtan et al [19] That kind of approach

to modeling seems to be promising and falls to the class ofthe regime-dependent models It requires prior knowledge ofthe particular flow regime, as well as the proportion of liquidand gaseous phase in the flow For that reason such a model isdifficult to apply in engineering practice Following that briefsurvey of superposition models, the question arises of how to

select the appropriate value of exponent n, in Eq (3) Should there be a value of n= 1 or 2 or 3 assumed? Or maybe someother value of that exponent ought to be looked for? Mikielewicz[20] provided an answer to that question, showing on the ba-sis of consideration of energy dissipation in the two-phase

flow that the exponent should be n = 2 A modified version

of that model will be examined in the course of the presentstudy

heat transfer engineering vol 31 no 4 2010

Trang 25

278 D MIKIELEWICZ

A completely different kind of approach to model flow

boil-ing is presented in tacklboil-ing the problem from the principles of

conservation of mass, momentum, and energy, and subsequently

through a numerical solution to such problem An example of

such an approach is a four-field two-fluid model due to Lahey

and Drew [21] In such an approach the two phases of fluid can

exist in continuous or dispersed form, leading to the occurrence

of four fields, namely, continuous liquid–dispersed gas and

con-tinuous gas–dispersed liquid Some success has been obtained in

modeling of bubbly flows and annular flows; however, the major

challenge is to predict the flow development and transformation

through consecutively developing flow structures (Podowski

[22]) The closure models and jump conditions form a very

difficult task to be implemented into the calculation procedure,

similar to the transient conditions Apparently the research is

still being exercised in that direction but there is still some time

to go before useful results are to be obtained of simulation of

the entire transformation from subcooled liquid to superheated

vapor

For that reason most of the approaches to date use

empir-ical correlations Presented next in brief is a model due to

Mikielewicz [20], which, as mentioned earlier, has been

de-veloped on the basis of consideration of dissipation in the flow

and recently modified to its final form, in Mikielewicz et al

[23] In the present article its further extension to transitional

and laminar flows is presented Such cases are often found in

minichannels, i.e., channels with diameters ranging from 600

µm to 3 mm [3] Empirical correlations known to date

can-not distinguish between the laminar and turbulent flow regimes

They are specially tailored to either one of the flow regimes

MODEL OF FLOW BOILING

The principal idea in the development of the model by

Mikielewicz [20] was a hypothesis that evaluation of energy

dissipation of major contributions in the flow boiling process

with bubble generation will lead to determination of heat

trans-fer in such flow Energy dissipation results from the friction in

the flow, which, on the basis of thermal hydraulic analogy, is

linked to heat transfer The flow is considered as an equivalent

liquid flow with properties of a two-phase flow

A fundamental hypothesis in the original model under

scrutiny [20] here is the fact that heat transfer in flow

boil-ing with bubble generation, treated here as an equivalent flow

of liquid with properties of a two-phase flow, can be modeled as

a sum of two contributions leading to the total energy

dissipa-tion in the flow, namely, energy dissipadissipa-tion due to shearing flow

without the bubbles, ETP, and dissipation resulting from bubble

generation, EPB:

Energy dissipation under steady-state conditions in the

two-phase flow can be approximated as energy dissipation in the

laminar boundary layer, which dominates in heat and tum transfer in the considered process Expressed as a powerlost in a unit volume of a boundary layer of two-phase flow ityields [20]:

momen-ETP= τTP2

µL = ξ2TPρ2

Lw4 TP

Analogically the energy dissipation due to bubble generation

in the two-phase flow can be expressed with velocity wTPandfriction factor ξPB:

EPB=ξ2PBρ2

Lw4 TP

In the Russian literature there are a number of contributionswhere investigations into flow resistance caused merely by thegeneration of bubbles on the wall are reported [24], which con-firm that the modeling approach presented in this article is pos-sible

The final term in Eq (5), E TPB, is modeled as the total energy

dissipation in the equivalent two-phase flow with velocity wTP

and some friction factor ξTPB, which after Eq (6) can be modeledas:

Making use of the analogy between the momentum and heat

we can generalize the preceding result to extend it over to heattransfer coefficients to yield heat transfer coefficient in flowboiling with bubble generation in terms of simpler modes of heattransfer, namely, heat transfer coefficient in flow without bubblegeneration and heat transfer coefficient in nucleate boiling:

Heat Transfer in Flow Boiling Without Nucleation

Let’s focus our attention first on the case without nucleation,which will lead to determination of a problem where convectiveboiling is dominant From the definition of the two-phase flowmultiplier, the pressure drop in two-phase flow can be related to

the pressure drop of a flow where only liquid at a flow rate G is

present:

The pressure drop in the two-phase flow without bubble ation can also be considered as a pressure drop in the equivalentheat transfer engineering vol 31 no 4 2010

Trang 26

The pressure drop of the liquid flowing alone can be determined

from a corresponding single-phase flow relation:

pL= l

dξLρLw

2 L

In the present article the most up-to-date relations for the friction

coefficient ξ, required for calculation of pressure drop, both in

laminar and turbulent flows will be used In case of turbulent

flow we use the Blasius equation for determination of the friction

In Eq (14) ReTP = (wTP d )/ν Land ReL = (w L d )/ν L The

defi-nition of the two-phase flow multiplier therefore leads us to the

following relation between the velocity in a two-phase flow and

Substituting relations (12) to (14) into (11), with relations

rele-vant to turbulent flow, we obtain a relation between the velocity

of two-phase flow and liquid-only:

wTP= R 1

1.75 wL ≈ R 0.571 wL (16)

An expression for the heat transfer coefficient, in the case of

tur-bulent flow, can be obtained if Eq (16) is used in determination

of the Reynolds number in expressions of the Dittus–Boelter

type, where ReLOis raised to the power 0.8, valid for turbulent

flows:

αTP

αL =

ReTPReL

0.8

=R 0.5710.8 ∼

= R 0.45 (17)where αLO is a heat transfer coefficient in the flow of only

liquid and the relation stemming from Eq (16), namely, ReTP =

R0.57ReLO, was used Use of the Dittus–Boelter equation with

properties referred to liquid is possible as the heat transfer in the

flow is governed merely by the flow of liquid The bubbles are

forming only a void and hence influence only the velocity of the

flow The scaling of velocity into the two-phase flow velocity is

done by means of application of a two-phase flow multiplier to

a single flow velocity (a result of rearrangement from definition

of a two-phase flow multiplier (11)) In case of laminar flow the

friction factor can be evaluated from the expression:

ξTP= 16

ReTP ξL= 16

Reynolds numbers are defined in a same way as in Eq (14)

Substituting Eq (18) into the definition of the two-phase flow

As can be noted, a linear relation between velocities results:

Utilizing Eq (20) in a correlation for the heat transfer coefficient

in laminar flow, where the Nusselt number is a function of asquare of the Reynolds number, we obtain a relation for two-phase heat transfer coefficient without nucleation in laminarflow:

of flows in channels Therefore, a value of 0.5 was selected forthe present modeling in laminar and transitional flows in tubes.Also in laminar flows past plates there are both analytical andempirical relations incorporating the Reynolds number raised tothe power 0.5

The following final form of correlation for flow boiling insmall diameter channels can be now postulated:

In the case of flow boiling the boundary layer is thinner andhence the gradient of temperature is more pronounced, whichsuppresses generation of bubbles in flow boiling That is thereason why heat flux is included into modeling That term ismore important for conventional size tubes, but cannot be totallyneglected in small-diameter tubes in the bubbly flow regime,where it is important A postulated form of correction has a formpreventing it from assuming values greater than 1, which was

a fundamental weakness of the model in earlier modifications

In the case of turbulent flow the exponent n assumes a value of

0.9, whereas in the case of laminar flow that exponent assumes

a value of n= 2

The two-phase flow multiplier RMSdue to M¨uller-Steinhagenand Heck [27] is recommended for use in the case of refriger-ants (Ould Didi et al [28]) It should be noted, however, that theheat transfer engineering vol 31 no 4 2010

Trang 27

280 D MIKIELEWICZ

choice of a two-phase flow multiplier to be used in the

postu-lated model is arbitrary In the activities presented in this article

the Muller-Steinhagen and Heck model [27] has been selected

for use as it is regarded best for refrigerants such as

hydrocar-bons; however, a different model could be selected, such as the

Lockhart–Martinelli model, where the two-phase flow

multi-plier is a direct function of the Martinelli parameter The latter

model is often found in correlations of flow boiling without

bubble generation, similar to Eq (1) Another conclusion could

be drawn from the presented model that in correlations of the

type of Eq (1) the two-phase flow multiplier could also be used

for modeling instead of the Martinelli parameter The author’s

experience to date shows that the influence of the two-phase

flow multiplier is very important and each fluid has a different

description of a two-phase resistance, which will be apparent

later when data for fluids other than hydrocarbons will be used

(Sumith et al [29]) In some of recent works by Thome [30],

there are three-zone approaches to devise a proper pressure drop

in the flow, which subsequently are used in heat transfer

calcu-lations Such an approach, in my opinion, is just confirming the

presented model, where it is stated that the flow resistance must

be incorporated directly into modeling of flow boiling heat

trans-fer The more accurate the two-phase flow multiplier, the more

accurate are the heat transfer predictions In the presented model

the RMSacts in the correction P as a sort of convective number,

known from other correlations In the form applicable to

con-ventional and small-diameter channels the Muller-Steinhagen

and Heck model [27] yields:

RMS=



1+2

1

and m = 0 for conventional channels Best consistency with

experimental data, in the case of small-diameter channels and

minichannels, is obtained for m = –1 In Eq (23) f1= (ρLG)

(µLG)0.25 for turbulent flow and f1 = (ρLG)(µLG) for

laminar flow Introduction of the function f1z, expressing the

ratio of heat transfer coefficient for liquid only flow to the heat

transfer coefficient for gas-only flow, is to meet the limiting

conditions; i.e., for x = 0 the correlation should reduce to a

value of heat transfer coefficient for liquid, αTPB= αL, whereas

for x= 1, approximately that for vapor, i.e., αTPB∼= αG.Hence:

f 1z= αGO

where f 1z = (µGL)(λLG)1.5(CpL/CpG) for turbulent flows

and f 1z= (λGL) for laminar flows When we want to consider

a limiting behavior of a correlation for vapor in a turbulent flow

regime then we consider only the abbreviated form of relation

(22), where the Dittus–Boelter relation is used for expressing

respective heat transfer coeffients:

αTPB

αL = R 0.5n

MS =

1

1.5 0.4

(26)

If we substitute Eq (26) into Eq (25) the right-hand side ofthe equation will have the exponent equal approximately 1 (as

n= 0.9 in turbulent flows) and the heat transfer coefficient will

assume a value of a vapor one for x = 1 A similar analysiscan be conducted for the case of laminar flow We should be

now convinced that introduction of the function f1zrenders that

correlation (22) obeying the limiting conditions; i.e., for x= 0,the correlation reduces to a value of heat transfer coefficient forliquid, αTPB = αL, whereas for x= 1, the correlation reduces

to approximately that for vapor, i.e., αTPB∼= αG.

It is possible to compare the proposed model (22) with themethod due to Chen [13] on the basis of comparisons of respec-

tive terms with the intensifying term F , which in the case of the model is approximately the term R 0.5n

MS, or the suppression term

S, which in the case of the present model is (1/(1+P )) 0.5 ple calculations performed for R123 are presented in Figures 1and 2 The character of changes shown by both distributions isconsistent with experimental data, which confirms a good quali-tative behavior of the model Subsequent comparisons, confirm-ing a good quantitative performance of the correlation, will beshown on the basis of comparisons of the results of calculationsagainst the experimental data

Sam-Figure 1 Correction (1+ P ) −0.5for different values of quality in comparison

with the suppression term in Chen’s correlation (term S) for R123.

heat transfer engineering vol 31 no 4 2010

Trang 28

D MIKIELEWICZ 281

Figure 2 Enhancement term F in Chen’s correlation for R123 in comparison

to corresponding term R 0.5n

MS in Eq (22).

APPLICATION TO SMALL-DIAMETER CHANNELS

It is a major drawback of most correlations developed for

calculation of heat transfer in conventional-size tubes that their

accuracy significantly deteriorates when applied to small-size

tubes, regarded here as greater than 600 µm [3] In such a

situa-tion the surface tension effects become more dominant and need

to be reflected in the model Most of the experimental data

in-dicate that most important for small channels is the convective

flow boiling mode, which, as stems from Eq (22), is

depen-dent on the flow resistance In such a case a great deal of care

must be exercised in the use of an appropriate friction model

Again, available correlations of two-phase flow friction fail to

be accurate for small-diameter channels A recent study due to

Tran et al [32] recommends a modification of the Chisholm

model to be applicable to small-diameter channels A very good

consistency with experimental data is reported; however, when

applied to relation (22) an inappropriate behavior is found when

quality x approaches unity, as the heat transfer coefficient for

vapor is not retrieved For that reason the Muller-Steinhagen

and Heck two-phase multiplier correlation [27] was modified to

incorporate the function f 1z, the derivation of which has been

presented in Eq (24) In case of modeling for small-diameter

passages the additional term responsible for surface tension

ef-fects, namely, the constraint number Con, has also been applied

to the discussed method (Mikielewicz et al [23]) The version of

Eq (23) ought to be therefore used with the constraint number,

Con, where the exponent index is m= –1, if calculations are

carried out for tube diameters smaller than 3 mm, and in cases

where the diameter is greater than 3 mm, the version without it

is appropriate, i.e., m= 0

Some effort has also been exercised to extend the correlation

(22) into the subcooled flow regime In the course of activities it

has been assumed that in the case of subcooled flow boiling the

correlation (22) can be adopted to subcooling if we assume that

the two-phase flow multiplier tends to unity, R → 1, and that

the correction function P tends to zero, P → 0 The nucleateboiling heat transfer coefficient in subcooled boiling is related

to the tsat temperature drop in the channel, instead of total

tsat+ t sub , where tsat is the difference between the walltemperature and the saturation temperature (superheating), and

t subis the difference between the saturation temperature andthe mean fluid temperature along the channel (subcooling) Insuch a case the heat transfer coefficient in subcooled boiling isreduced to:

αPBsubαPB

postulated model The empirical correction P in relation (22)

is dependent upon quality, which shows up in Figures 1 and 2,

whereas the suppression factor S in the model due to Chen is

independent of quality The conclusion that can be drawn from

examination of distributions of correction P in Figure 1 is that

the suppression is smaller for higher qualities and it exhibits adecrease with increasing Reynolds number

Next, the available data bank for conventional size tubes hasbeen revisited, featuring about 2500 data points [22], and thecalculations of theoretical values were made for the amended

value of exponent n = 0.9 in (22); see Figure 3 It must benoted that data for conventional-size tubes fall to the turbu-lent flow regime The accuracy of calculations has not wors-ened and again about 65% of points were within ±30% oferror

From examination of Figure 4 we can see that the data spreadaround the expected value of unity is also uniform, and thecorrelation assumes an equal spread in error for consideredqualities

heat transfer engineering vol 31 no 4 2010

Trang 29

282 D MIKIELEWICZ

Figure 3 Data bank for conventional size tubes reduced with correlation (22).

Small-Diameter Tubes and Minichannels

Having shown that correlation behaves reasonably well in the

case of conventional-size tubes, some comparisons were made

with the data for small-diameter channels In order to do that,

the only correlation adjustment, in comparison to

conventional-channels data, was that the version of Eq (23) incorporating the

constraint number Con raised to power m= –1 was used

First, attention was focused on a set of experimental points

due to Karayiannis [34] These investigations were carried out

on stainless-steel tubes with internal diameters of 2 mm and

4.26 mm using R134a as a working fluid and 996 data points

were at the disposal for comparisons The test section was 40

cm long The flow parameters were the following: x= 0.1–0.8,

mass flow rate G = 100–500 kg/m2-s, applied heat flux q =

Figure 4 Data bank for conventional size tubes versus quality reduced with

correlation (22).

Figure 5 Data due to Karayiannis [34] reduced with correlation (22).

11–100 kW/m2, saturation temperature TSAT = 30–46◦C, and

tube diameters d = 2 mm and 4.26 mm In Figures 5 and 6are presented calculations using the author’s own method, Eq.(22), of data due to Karayiannis [34] It is apparent from Fig-ure 6 that the discrepancies exceeding 30% are visible only forthe quality close to zero All other data fall into the envelope

of ±30% That result must be deemed satisfactory In Figure

7, presented for comparison are calculations obtained using theChen [13] method Notably the agreement is worse, and ac-cording to communication with the author of the experimentaldata the correlation due to Chen [13] was performing best of allothers known in literature It is important to note that the bigger

Figure 6 Data due to Karayiannis [34] versus quality reduced with correlation (22).

heat transfer engineering vol 31 no 4 2010

Trang 30

Another data set that has been tested in the course of the

present study was experimental data collected for water in a

tube diameter of d = 1.45 mm due to Sumith et al [29] The

experimental parameters were as follows: quality x= 0.1–0.8,

mass flow rate G= 80–250 kg/m2-s, heat flux= 0.3–80 kW/m2,

and saturation temperature TSAT= 263◦C The results of the

cal-culations are presented in Figure 8 As can be seen, the available

database does not correlate too well The heat transfer

coeffi-cient, αTPB, calculated using Eq (22) returns somewhat lower

Figure 8 Data due to Sumith et al [29] reduced with correlation (22).

Figure 9 Data due to Summith et al [29] reduced with correlation (22) incorporating the two-phase flow multiplier due to Gronnerud [28].

values than the experimental data The first question in the ysis of these data was whether the expression for the two-phaseflow multiplier was adequate for calculations of water heat trans-fer Subsequently, calculations were performed using the Gron-nerud correlation of two-phase flow multiplier [28] rather thanthat of Muller-Steinhagen and Heck [27] In Figure 9 are pre-sented calculations obtained using the Gronnerud correlation[28] In this case the comparison is much better, as more than60% of experimental points are found in the±30% envelope oferror The conclusion that can be drawn here is that the postu-lated correlation still preserves its qualitative consistency andshows only the quantitative discrepancy, which can be reduced ifthe appropriate model for the two-phase flow multiplier is used.However, the obtained results are still acceptable, from the point

anal-of view that the suggested correlation is anal-of a general character.The data set due to Sumith et al [29] shows the necessity for veryaccurate methods of determination of two-phase flow multiplierfor different flow regimes It can also be concluded that if theexperimental data exhibits very strong dependence on the flowresistance then the bubble generation is of secondary impor-tance That finds confirmation in recent works by Thome [30],who states that in small-diameter channels and minichannels,dominant is the convective heat transfer and bubble nucleation

is negligible The models of flow boiling developed by Thomeand his coworkers are looking very detailed at the first glance,but in perspective are aimed at the development of a precisevalue of the pressure drop in the channel, which further con-verts to the two-phase flow multiplier Hence, we can regardthe model presented here as the sort of top-to-bottom approach,whereas the approach by Thome [30] is the bottom-up approach.heat transfer engineering vol 31 no 4 2010

Trang 31

284 D MIKIELEWICZ

We find a necessity of obtaining very accurate models for

the two-phase flow multipliers, whereas Thome’s approaches

are forming one of the ways of postulating the appropriate

pres-sure drop in the channel However, as ought to be remembered,

none of the models presented so far in the literature incorporate

flow resistance, which is a key for success in small-diameter

channels, minichannels, and probably microchannels, which

have a dominant character in convective heat transfer

Another data set under scrutiny here is the nitrogen data due

to Qi et al [35] Nitrogen is practically an ideal wetting fluid on

most solid surfaces It features a small value of surface tension,

which may have a significant influence on the bubble

nucle-ation, growth, and detachment Moreover, liquid nitrogen also

has smaller viscosity, higher thermal conductivity, and smaller

density ratio of liquid phase to vapor phase, compared to the

nor-mal fluids (water and refrigerants) These features undoubtedly

affect the heat transfer and pressure drop characteristics of flow

boiling in both macro- and microchannels The experiments due

to Qi et al [35] have been performed in the stainless-steel

circu-lar tubes with diameters of 0.531, 0.834, 1.042, and 1.931 mm

The range of flow parameters was: x= 0.1–0.8, mass flow rate

G= 195–2000 kg/m2-s, heat flux q= 10–15 kW/m2, and

satu-ration pressure pSAT= 200–800 kPa Data have been presented

in Figure 10 We can see that 60% of results fall into the error

band of±30% That result should be regarded as quite

satisfac-tory Some of the data have been collected for the flows in tubes

with diameters smaller than 1 mm

In the case of other dimensions of the tubes we can observe

that the distribution of the ratio αth /α expis somewhat linear in

relation to quality x That is apparent for the remaining diameters

of tubes However, the slope is more or less the same in all cases

Qi et al [35] in their paper tested other empirical correlations

Figure 10 Data due to Qi et al [35] reduced with correlation (22).

Figure 11 Data due to Bohdal [33] reduced with correlation (25).

against their experimental data These were correlations due

to Klimenko, Shah, Chen, and Tran; see [35] None of thesecorrelations was found suitable to be relevant to reflect that type

of experimental data

Subcooled Boiling in Conventional Tubes

Subsequently, our attention is focused on flow boiling insubcooled channels The relation (25), describing that process,which is a reduced form of expression (22), is used in calcula-tions The technical capabilities enabled for experiments in the

following range of parameters: mass flow rate G = 100–1600kg/m2-s, saturation temperature TSAT= –30 to 20◦C, fluid sub-

cooling (TSAT – T f) = 0–10 K, and heat flux, q = 0–30,000

W/m2 In Figure 11 are presented comparisons of model dictions using Eq (25) and experimental data for subcooledflow boiling of R134a, in a vertical channels of 13 mm in in-ner diameter As can be seen, the model underpredicts the heattransfer coefficient by about 30% That is primarily due to thefact that the assumption has been made during the reduction ofthe correlation to the subcooled flow regime that the two-phaseflow multiplier assumes a value of unity, which is not true, asthe presence of bubbles leads to the increase of two-phase flowmultiplier above unity

pre-Therefore, calculations have been repeated with a value of

two-phase flow multiplier artificially set to RMS = 3 The results

of calculations have been presented in Figure 12 We can see

a much better agreement with experimental data We can alsoconclude that the correct form of two-phase flow resistance is

of paramount importance in simulations of flow boiling heattransfer, in both saturated and subcooled regimes

heat transfer engineering vol 31 no 4 2010

Trang 32

D MIKIELEWICZ 285

Figure 12 Data due to Bohdal [33] reduced with correlation (25), RMS = 3.

CONCLUSIONS

The model of flow boiling presented in this article is a

gen-eral model, which can be advised for use in numerous

engineer-ing applications Comparisons with different experimental data

show its robustness and satisfactory accuracy The model,

con-trary to other methods of calculating heat transfer coefficients in

flow boiling, has been developed on a theoretical basis, as a result

of consideration of energy dissipation in the flow, where boiling

occurs That led, first of all, to devising a value of exponent n

in Eq (1), most often used for reduction of experimental data

The presented model gives the value of exponent n = 2 The

postulated model also incorporates another term, which proves

to be very important in case of minichannels and

microchan-nels That is the inclusion of a two-phase flow multiplier, which

models the convective flow without bubble nucleation The

se-lection of the two-phase flow multiplier is also very important

for calculations In the case of refrigerants, the most effective

is a model due to M¨uller-Steinhagen and Heck [27] In the case

of other fluids the latter model may not provide good accuracy

of calculations For example, the data set due to Sumith et al

[29] shows that the best two-phase pressure drop model is due

to Gronnerud [28] The more accurate is the selected model for

the two-phase flow multiplier, the more accurate are the results

obtained In order to get the most accurate results of calculations

we must use the models of pressure drop developed for specific

flow structures That proves that the approach to model heat

transfer in flow boiling starting from precise evaluation of

pres-sure drop is very promising [19, 30], but that is also consistent

with the general structure of the model presented here, with the

two-phase flow multiplier being an important item in it What

can be advised for prospective research into flow boiling is that

the measurements of pressure drop and quantities constituting

the heat transfer coefficient should take place simultaneously,

in order to obtain the most accurate information

The model presented here is applicable both to conventionalchannels and to small-diameter channels It has been tested forchannel diameters greater than 1 mm In the case of smallerchannel dimensions the presented model tends to overpredictthe experimental data Very useful here are the findings ofThome [30], who stipulates that in small-diameter channels,and minichannels and microchannels in particular, the bubblenucleation is not present and only the slug and annular flowstructures are dominant That observation may also be incor-porated into the postulated model (22), as in such cases it may

be recommended to drop the entire nucleate boiling term Such

an operation will lead to the dependence of the two-phase flowheat transfer coefficient merely on the two-phase flow multiplier,which would serve as useful information to other researchers as

to how continue their investigations

The model presented can also be reduced to a form cable for the analysis of subcooled flow boiling Also here wecan notice that the knowledge of precise pressure drop is veryimportant In the case of subcooled boiling there is a lack ofmodels for two-phase flow multipliers By guessing the value oftwo-phase flow multipliers we can significantly improve modelpredictions

appli-In the author’s opinion, the presented model can be suggestedfor a wider use among engineers, but further validation withexperimental data would add value to its robustness

NOMENCLATURE

Bo Boiling number, Bo= q ·l·ρL

ρG·hLG ·µ L

Cp specific heat at constant pressure, J/kg-K

d channel inner diameter, m

g gravity, m/s2

G mass flow rate, kg/m2-s

hLG latent heat of evaporation, J/kg

l bubble characteristic length, channel length, m

P correction in Eq (2)

Pr Prandtl number, Pr=µL·CL

λL

q heat flux density, W/m2

R two-phase flow multiplier

Trang 33

[1] Thome, J R., Boiling of New Refrigerants: A State-of-the-Art

Review, International Journal of Refrigeration, vol 19, no 7, pp.

435–457, 1996

[2] Bergles, A E., Lienhard, V J H., Kendall, G E., and

Grif-fith, P., Boiling and Evaporation in Small Diameter

Chan-nels, Heat Transfer Engineering, vol 24, no 1, pp 18–40,

2003

[3] Kandlikar, S G., Fundamental Issues Related to Flow Boiling

in Minichannels and Microchannels, Proc Experimental Heat

Transfer, Fluid Mechanics and Thermodynamics, pp 129–146,

Thesalloniki, Greece, 2001

[4] Thome, J R., Boiling in Microchannels: A Review of Experiment

and Theory, International Journal of Heat and Fluid Flow, vol.

25, no 2, pp 128–139, 2004

[5] Dengler, C E., and Addoms, J N., Heat Transfer Mechanism for

Vaporisation of Water in Vertical Tube, Chem Eng Prog Symp.

Ser., 52, 18, 95–103, 1956.

[6] Guerrieri, S A., and Talty, R D., A Study of Heat Transfer to

Organic Liquids in Single Tube Natural Circulation Vertical Tube

Boilers, Chem Eng Prog Symp Ser., vol 52, no.18, pp 69–77,

1956

[7] Schrock, V E., and Grossman, L M., Forced Convection

Boiling Studies, University of California, Institute of

Engi-neering Research, Report 73308-UCX-2182, Berkeley, CA,

1959

[8] Collier, J G., and Pulling, D J., Heat Transfer to Two-Phase

Gas–Liquid Systems, Report AERE-R3908, Harwell, UK, 1962

[9] Shah, M M., Chart Correlation for Saturated Boiling Heat

Trans-fer: Equations and Further Study, ASHRAE Trans., vol 88, pp.

185–196, 1982

[10] Kandlikar, S G., A General Correlation for Saturated

Two-Phase Flow Boiling Heat Transfer Inside Horizontal and

Ver-tical Tubes, Journal of Heat Transfer, vol 112, pp 219–228,

1989

[11] Gungor, K E., and Winterton, R H S., A General Correlation for

Flow Boiling in Tubes and Annuli, International Journal of Heat

Mass Transfer, vol 29, pp 351–358, 1986.

[12] Rohsenow, W M., A Method of Correlating Heat Transfer Data

for Surface Boiling of Liquids, Trans ASME, vol 74, pp 969–

976, 1952

[13] Chen, J C., Correlation for Boiling Heat-Transfer to Saturated

Fluids in Convective Flow, Industrial & Chemical Engineering

Process Design and Development, vol 5, no 3, pp 322–339,

1966

[14] Kutateładze, S S., Boiling Heat Transfer, International Journal

of Heat and Mass Transfer, vol 4, pp 31–45, 1961.

[15] Steiner, D., and Taborek, J., Flow Boiling Heat Transfer in Vertical

Tubes Correlated by Asymptotic Model, Heat Transfer ing, vol 23, no 2, pp 43–68, 1992.

Engineer-[16] Forster, H K., and Zuber, N., Dynamics of Vapour Bubbles and

Boiling Heat-Transfer, AIChE J., vol 1, no 4, pp 531–535,

1955

[17] Cooper, M G., Saturation Nucleate Pool Boiling: A Simple

Cor-relation, International Chemical Engineering Symposium, ser no.

[22] Podowski, M., Understanding Multiphase Flow and Heat

Trans-fer: Perception, Reality, Future Needs, Archives of ics, vol 26, no 3, pp 3–20, 2005.

Thermodynam-[23] Mikielewicz, D., Mikielewicz, J., and Tesmar, J., Improved Empirical Method for Determination of Heat Transfer Coefficient

Semi-in Flow BoilSemi-ing Semi-in Conventional and Small Diameter Tubes, ternational Journal of Heat and Mass Transfer, vol 50, pp 3949–

In-3956, 2007

[24] Ananiev, E L., On the Mechanism of Heat Transfer in NucleateBoiling Flow of Water in a Tube Against the Reynolds Analogy,

in Convective Heat Transfer in Single and Two-Phase Flows, ed.

V W Borishansky, Energia, Leningrad, 1964 (in Russian)

[25] Shah, R K., and London, A L., Laminar Flow Forced Convection

in Ducts, Advances in Heat Transfer, Suppl 1, Academic Press,

London, 1978

[26] Park, H S., Dependency of Heat Transfer Rate on the Brinkman

Number in Microchannels, Proc Therminic 2007, Budapest,

Hun-gary, pp 61–65, 2007[27] M¨uller-Steinhagen, H., and Heck, K A., A Simple Fric-tion Pressure Drop Correlation for Two-Phase Flow in Pipes,

Chemical Engineering and Processing, vol 20, pp 297–308,

[29] Sumith, B., Kaminaga, F., and Matsumura, K., Saturated Flow

Boiling of Water in Vertical Small Diameter Tube, Experimental Thermal Fluid Sciences, vol 27, pp 789–801, 2003.

[30] Thome, J R., Wolverine Engineering Databook III, Chapter I,

2007, www.wlv.com/products/databook/db3/DataBookIII.pdf[31] Kew, P., and Cornwell, K., Correlations for the Prediction of Boil-

ing Heat Transfer in Small Diameter Channels, Applied Thermal Engineering, vol 17, nos 8–10, pp 705–715, 1997.

heat transfer engineering vol 31 no 4 2010

Trang 34

D MIKIELEWICZ 287[32] Tran, T N., Chyu, M C., Wambsganss, M W., and France, D.

M., Two-Phase Pressure of Refrigerants During Flow Boiling in

Small Channels: An Experimental Investigation and Correlation

Development, International Journal of Multiphase Flow, vol 26,

pp 1739–1754, 2000

[33] Bohdal, T., Nucleate Boiling Phenomena, Koszalin University of

Technology Publishers, Koszalin, vol 76, 2001 (in Polish)

[34] Karayiannis, T., Private communication, 2004

[35] Qi, S L., Zhang, P., Wang, R Z., and Xu, L X., Flow

Boil-ing of Liquid Nitrogen in Micro-Tubes: Part II—Heat

Trans-fer Characteristic and Critical Heat Flux, International

Jour-nal of Heat and Mass Transfer, vol 50, pp 5017–5030,

2007

Dariusz Mikielewicz is a professor of thermal

sciences at the Gdansk University of Technology, Gdansk, Poland He received his M.Sc degree from the Gdansk University of Technology (1990), and his Ph.D degree from the University of Manchester (1994) In 2002 he presented his habilitational dis- sertation at the Gdansk University of Technology In 1994–1996 he worked as an engineer at the Berkeley Nuclear Laboratories, Gloucestershire, UK He has been teaching at the GUT since 1996 He has been

an elected member to the Science Council in 2004–2008 His research tributions have been in the field of modeling of mixed convection, two-phase flows, and recently renewable energy He is currently working on enhanced heat transfer and condensation in heat exchangers.

con-heat transfer engineering vol 31 no 4 2010

Trang 35

CopyrightC Taylor and Francis Group, LLC

ISSN: 0145-7632 print / 1521-0537 online

DOI: 10.1080/01457630903312049

Mechanisms of Boiling in

Micro-Channels: Critical Assessment

JOHN R THOME and LORENZO CONSOLINI

Laboratoire de Transfert de Chaleur et de Masse, ´Ecole Polytechnique F´ed´erale de Lausanne, Switzerland

Numerous characteristic trends and effects have been observed in published studies on two-phase micro-channel boiling heat

transfer While macro-scale flow boiling heat transfer may be decomposed into nucleate and convective boiling contributions,

at the micro-scale the extent of these two important mechanisms remains unclear Although many experimental studies have

proposed nucleate boiling as the dominant micro-scale mechanism, based on the strong dependence of the heat transfer

coefficient on the heat flux similar to nucleate pool boiling, they fall short when it comes to actual physical proof A strong

presence of nucleate boiling is reasonably associated to a flow of bubbles with sizes ranging from the microscopic scale

to the magnitude of the channel diameter The bubbly flow pattern, which adapts well to this description, is observed,

however, only over an extremely limited range of low vapor qualities (typically for quality less than 0.01–0.05) Furthermore,

at intermediate and high vapor qualities, when the flow assumes the annular configuration and a convective behavior is

expected to dominate the heat transfer process, the experimental evidence yields entirely counterintuitive results, with heat

transfer coefficients often decreasing with increasing vapor quality rather than increasing as in macro-scale channels,

and with a much diminished heat flux dependency compared with would be expected In summary, convective boiling in

micro-channels has been revealed to be much more complex than originally thought The present review aims at describing

and analyzing the boiling mechanisms that have been proposed for two-phase micro-channel flows and confronting them

with the available experimental heat transfer results, while highlighting those questions that, to date, remain unanswered.

INTRODUCTION

The initial purpose of the studies on two-phase heat transfer

in micro-channels was (and still is, to a certain extent) aimed

at understanding the mechanisms controlling the flow boiling

process The earliest results of Lazarek and Black [1] showed

values for the heat transfer coefficient that were unaffected by

vapor quality, but were a function of heat flux, leading them

to conclude that nucleate boiling was the dominant heat

trans-fer mechanism as in macro-scale flow boiling While others

confirmed this trend and adopted their explanation, new trends

arose as the amount of experimentation in the sector grew A

significant number of studies found decreasing curves in the

α–x plane rather than flat ones (with α the heat transfer

coeffi-cient and x the thermodynamic vapor quality), and even some

increasing trends in α versus x, giving rise to a rather puzzling

scenario with respect to the macro-scale knowledge base (these

were documented, for instance, in Agostini and Thome [2])

Address correspondence to Professor John R Thome, EPFL STI IGM

LTCM, ME G1 464, Station 9, 1015 Lausanne, Switzerland E-mail:

john.thome@epfl.ch

The relative importance of nucleate boiling, thin film ration, and convective boiling in the individual flow patterns thatare characteristic to a micro-channel flow is thus still unclear.The studies directed specifically to flow patterns (see, for exam-ple, Tripplett et al [3] and Serizawa et al [4]), many of whichare for air–water flows, found general agreement as to the fourmain regimes: (i) bubbly flow, at very low mass fractions of air orvapor, (ii) slug flow, describing the passage of long bubbles sep-arated by liquid slugs, (iii) churn flow, a transition mode betweenslug flow and fully annular flow, and (iv) annular flow, occurring

evapo-at the highest gas superficial velocities (see Figure 1) Recently,similar flow patterns have also been reported for the flow ofrefrigerants (cf Revellin et al [5]), confirming the absence ofany stratified regime in the micro-scale Cornwell and Kew [6]coupled flow patterns and heat transfer, by arguing that differ-ent flow regimes presented different heat transfer mechanisms,varying from essentially nucleate boiling, to confined bubbleboiling, and finally to purely convective evaporation for annularflows Jacobi and Thome [7] and Thome et al [8] postulatedthat during slug flow, nucleate boiling is completely suppressedand heat is transferred primarily by conduction through the thinevaporating film surrounding the elongated bubbles, while heat288

Trang 36

J R THOME AND L CONSOLINI 289

Figure 1 From top to bottom: bubbly, slug, churn, and annular flows for

R-134a in a 510 µm tube at a mass velocity of 500 kg/m2-s Taken from Consolini

[30].

transfer to the liquid slug and any dry zone are only of second

and third importance depending on their respective residence

times

An element added to the discussion concerns the possibility

of periodic dry-out of the channel wall (see Thome et al [8],

Kew and Cornwell [9], and Kandlikar [10]) This mechanism,

however, was not entirely understood, and very few gave a clear

opinion as to how and when it occurred The work presented in

Thome et al [8] was among the few studies that attempted to

address this issue, suggesting the development of a dry zone at

the tail of an evaporating bubble

In recent years, a number of studies provided evidence of

the sensitivity of two-phase micro-channel systems to flow

in-stabilities Oscillating pressure drops and wall temperatures,

and visualizations showing cyclical backflow, were

encoun-tered in many experiments Unfortunately, many of these

data have been mingled in with stable data and thus

cre-ate a confusing situation Bergles and Kandlikar [11]

clas-sified these flows as compressible volume instabilities,

re-lating them to the presence of compressibility prior to the

heated channels Relative to a stable mode, an unstable

sys-tem presents entirely different flow features and may bring

about substantial differences in the heat transfer mechanisms

(see [12])

The discussion that follows is aimed at providing a critical

review of the main conclusions that may be drawn to date on

the mechanisms of heat transfer for boiling in micro-channels,

Figure 2 Experimental heat transfer coefficients from Lazarek and Black [1] for R-113 in a 3.1 mm tube.

focusing primarily on the characteristics of stable two-phaseflows

EXPERIMENTAL HEAT TRANSFER

Among the first studies on flow boiling heat transfer in asingle channel was the one by Lazarek and Black [1], who re-ported experimental heat transfer coefficients for flow boiling

of R-113 in a vertical tube with an inner diameter of 3.1 mm(Figure 2) Their heat transfer coefficients had a strong depen-dency on the applied heat flux, but were essentially independent

of vapor quality Similar results were obtained by Tran et al.[13] and Bao et al [14] Tran and coworkers performed exper-iments on R-12 in circular and rectangular channels with sizesranging from 2.40 to 2.92 mm Their data showed that for a suffi-ciently high wall superheat (above 2.75 K) the values of the heattransfer coefficient were unaffected by vapor quality and massvelocity, but increased significantly with heat flux The flowboiling experiments of Bao et al again confirmed these trendsand presented additional data showing the improvement in heattransfer with increasing saturation pressure, further promotingthe view that nucleate boiling was dominating (no visualization

of the flow was possible in their setup) More recently, Lihong

et al [15] performed experiments on a 1.3 mm circular channelfor refrigerant R-134a, yielding similar results

Added trends in heat transfer were reported in the work byLin and coworkers [16] (Figure 3) In their study on R-141b(1.1 mm tube), they observed three distinct responses in theheat transfer coefficient to changes in heat flux and vapor qual-ity: (1) At low heat fluxes, heat transfer improved with increas-ing vapor quality, (2) for intermediate values from 30 to 53kW/m2 in Figure 3 and vapor qualities within 0.40, the heattransfer coefficient increased with heat flux, much like whatwas observed in the investigations cited previously, and (3) at

the highest heat fluxes, heat transfer gradually fell with x and

heat transfer engineering vol 31 no 4 2010

Trang 37

290 J R THOME AND L CONSOLINI

Figure 3 Experimental heat transfer coefficients from Lin et al [16] for

R-141b in a 1.1 mm channel and a mass velocity of 510 kg/m 2 -s.

tended to heat-flux-independent values While further heating

increased the heat transfer coefficient for vapor fractions up

to 0.40, the correspondence was much less clear beyond this

threshold

Saitoh et al [17] obtained heat transfer data for boiling of

R-134a in a 0.51 mm tube (550 mm long) at 15 and 29 kW/m2

for qualities extending to almost unity Their heat transfer

co-efficients showed an inverted “U” shape in the α–x plane The

heat transfer data increased up to a quality of 0.60, beyond

which the coefficients declined monotonically In another study

on R-134a, Martin-Callizo et al [18] presented results for a

ver-tical 0.64 mm stainless-steel micro-channel, finding that once

again the dominant effect was that of heat flux while mass

velocity was less important They found that their heat

trans-fer coefficients were rather insensitive to vapor quality until

reaching the higher range of their heat flux test range,

where-upon the heat transfer coefficients then decreased

monoton-ically from vapor qualities of about 0.01–0.02 down to

val-ues of about 0.6–0.8 without going through any maximum or

minimum

As for the effect of mass velocity, a number of investigations

have shown heat transfer coefficients to remain unchanged when

varying the fluid flow rate, as in the data from Tran et al [13]

(see Figure 4) Tran and coworkers reported an improvement in

heat transfer with mass velocity only for wall superheats lower

than 2.75 K The experiments from Bao et al [14] on R-11 and

R-123, and from Lihong et al [15] on R-134a, also showed no

change in heat transfer with flow rate, with the latter observing

mild differences only at the lowest heat flux tested One of

the few studies on a single channel that presented a different

outcome was that of Sumith et al [19] for flow boiling of water

in a 1.45 mm vertical tube, reporting heat transfer coefficients

that often decreased when increasing the flow, even at high heat

fluxes

At present, there is general agreement among the studies that

addressed the effect of saturation conditions on heat transfer,

Figure 4 Experimental heat transfer coefficients from Tran et al [13] for R-12 in a 2.46 mm channel at different heat fluxes and mass velocities and wall superheats above 2.75 ◦C.

that a higher saturation pressure/temperature yields higher heattransfer coefficients (see for example [14, 15, 18])

Similar results have also been recently reported in flow ing heat transfer experiments on multi-micro-channel systems,

boil-as in the cboil-ase of Agostini et al [20, 21], who tested refrigerantsR-134a and R-236fa in a 67-parallel-micro-channel evaporator(rectangular channels, 0.223 mm wide, 0.680 mm high, and20.0 mm long, separated by 0.80 mm wide fins) It can be seenfrom Figure 5 that their heat transfer data at low heat fluxes tend

to increase with vapor quality until intermediate heat fluxes,where they first increase and then show nearly no influence ofvapor quality At higher heat fluxes, the heat transfer coefficientsstart to decline with increasing vapor quality While the heattransfer coefficients rise sharply with increasing heat flux, at the

Figure 5 Flow boiling data of Agostini et al [20] for R-236fa in a silicon multi-micro-channel test section at a mass velocity of 810.7 kg/m 2 -s, a nominal pressure of 2.73 bar, and saturation temperature of 25 ◦C The silicon test section

without its cover plate is shown in the inset photograph.

heat transfer engineering vol 31 no 4 2010

Trang 38

J R THOME AND L CONSOLINI 291

highest values (starting at 178.4 W/cm2relative to the surface

area of the heating element) a peak is reached and the heat

trans-fer coefficients begin to decrease with increasing heat flux as

the critical heat flux is approached (but not reached) in this data

set

HEAT TRANSFER MECHANISMS

The heat transfer mechanisms that are active in boiling in

micro-channels can be summarized as follows: (i) In bubbly

flow, nucleate boiling and liquid convection would appear to be

dominant, (ii) in slug flow, the thin film evaporation of the liquid

film trapped between the bubble and the wall and convection to

the liquid and vapor slugs between two successive bubbles are

the most important heat transfer mechanisms, including in terms

of their relative residence times, (iii) in annular flow, laminar or

turbulent convective evaporation across the liquid film should be

dominant, and (iv) in mist flow, vapor-phase heat transfer with

droplet impingement will be the primary mode of heat transfer

For those interested, a large number of two-phase videos for

micro-channel flows from numerous laboratories can be seen in

the free e-book of Thome [22]

Notably, many experimental papers conclude without proof

that nucleate boiling is dominant in their data only because they

find a substantial heat flux dependency; a heat flux dependency,

however, does not prove that nucleate boiling is dominant or

even present For instance, Jacobi and Thome [7] and Thome et

al [8] have argued that the heat flux effect can be explained and

predicted by the thin-film evaporation process occurring around

elongated bubbles in the slug flow regime without any

nucle-ation sites Their model shows that the heat flux dependency

comes mainly through its effect on the bubble frequency and

the thin film evaporation process Thus, simply labeling

micro-channel flow boiling data as being nucleate boiling dominated

is misleading since this seems to only be the case for the bubbly

flow regime, which occurs at very low vapor qualities (typically

for x < 0.01–0.05 depending on the mass velocity, etc.) Some

experimental flow boiling studies that report that nucleate

boil-ing was dominant at low vapor qualities also report that the flow

regime observed at these conditions was elongated bubble flow

without any bubbly flow observed These two conclusions thus

seem to be contradictory One should further contemplate that a

nucleate boiling correlation does not actually model the

nucle-ate boiling process, but is only an empirical relationship relating

the heat transfer coefficient to the heat flux, and hence the actual

mechanism is not actually addressed in the correlation Hence,

in flow boiling in a micro-channel in elongated bubble (slug)

flows, the heat flux dependency in such a correlation most likely

is coming in through the bubble frequency and thin film effects,

not that of nucleate boiling

To date, many types of noncircular micro-channels have

been tested; for instance, results for square, rectangular,

par-allel plate, triangular, and trapezoidal geometries are currently

available in the literature Besides the problems associated withcharacterizing the channel size (e.g., a hydraulic diameter of anoncircular channel has no physical relationship to an annularfilm flow), the rectangular channels tested sometimes have veryhigh aspect ratios whose effect on heat transfer is not well un-derstood Recalling the wedge flows observed by Cubaud and

Ho [23] for air–water, a partially wetted perimeter along andaround elongated bubbles will have an influence on heat trans-fer while the wet corners may tend to better resist completedry-out

EMPIRICAL PREDICTION METHODS

The variety of trends in heat transfer data and the inherentdifficulties in performing experimental work on these small sys-tems have made it very challenging to develop a well-establishedunderstanding of boiling in micro-channels Several authorshave correlated their experimental results through different sets

of generally non-dimensional groups Others, on the other hand,have attempted to either extend methods previously developedfor conventional macro-scale systems to the micro-scale, or todefine new approaches specifically for micro-channel two-phaseflows

Lazarek and Black Correlation

From their heat transfer experiments on R-113, Lazarek andBlack [1] proposed the following nondimensional correlationfor the flow boiling Nusselt number (Nul = αD/k l):

Nul = 30Re0.857

with Relo = GD/µ l, the all-liquid Reynolds number, Bo =

q /(Gh lv ) the Boiling number, and G the mass velocity of the total

flow of liquid and vapor Equation (1) expresses no dependence

of the heat transfer process on the local vapor quality

Tran et al Correlation

As mentioned earlier, in their experiments on 12 and

R-113 Tran and coworkers [13] observed that for wall superheatsabove 2.75 K their heat transfer data expressed a strong α versus

q behavior, assigning this to the macro-scale mechanism ofnucleate boiling The authors therefore modified the correlation

of Lazarek and Black [1], Eq (1), by replacing the Reynoldsnumber with the Weber number, Welo = G2D/(ρl σ), removingviscous effects in favor of surface tension The liquid to vapordensity ratio was added to further account for variations in fluidproperties, so that

Trang 39

292 J R THOME AND L CONSOLINI

The 8.4× 105factor in Eq (2) is dimensional, with units of

W/(m2-K) Equation (2) removes any dependence of the heat

transfer coefficient on mass velocity Furthermore, Eq (2) also

yields the following proportionality between the heat transfer

coefficient and the channel diameter: α∝ D 0.3, which seems to

be the opposite of experimental trends found in later studies

Kandlikar and Balasubramanian Correlation

Kandlikar and Balasubramanian [24] extended the

correla-tion proposed by Kandlikar for convencorrela-tional tubes, where the

local two-phase heat transfer coefficient was determined

accord-ing to the value of the dominant mechanism between nucleate

boiling (nb) and convective evaporation (cv):

α= larger of



αnb

The original correlations for the two coefficients in Eq (3)

were developed for all-liquid Reynolds numbers, Relo, above

3000, and presented the following functional dependencies:

The nondimensional groups in Eq (4) are respectively the

Convection number, Cv, the Boiling number, Bo, the all-liquid

Froude number, Frlo, and the vapor quality For Relo >3000,

Kandlikar suggested using transition (Gnielinski) and fully

tur-bulent (Petukhov and Popov) correlations for the single-phase

liquid heat transfer coefficient, αl, based on the all-liquid

Reynolds number However, for smaller channels the authors

argued that the value of the Reynolds number was generally

lower than 3000, making the preceding single-phase

correla-tions inconsistent Furthermore, the reduced effect of gravity

in micro-channels justified the removal of the Froude number

from Eq (4) In view of both these considerations, Kandlikar and

Balasubramanian proposed the following modified correlations

with F sf a constant that was used to fit the expressions to each

particular surface material–fluid combination For Reynolds

numbers in the range 1600≤ Relo <3000, the authors suggested

interpolating between laminar and transition correlations for αl

On the other hand, for Relo <1600 the flow was considered

laminar, and a laminar correlation of the form Nu= αl D/k l =

constant was deemed applicable Finally, for Relo≤ 100, Eq (3)

was modified to α= αnb, with αnbgiven by Eq (5) ingly, their nucleate boiling and convective boiling heat transfercorrelations in Eq (5) are identical, except for values of the twolead constants and one of the exponents Thus, it is not clearhow one represents nucleate boiling and the other convectiveboiling

Interest-Zhang et al Extension of Chen’s Correlation

Zhang and coworkers [25] analyzed 13 separate databases,confronting them with some of the most widely quoted cor-relations for two-phase heat transfer in conventional systems.Chen’s superposition model gave the best outcome However,the authors observed that for micro-channels, the values of theliquid Reynolds numbers, Rel = G(1 – x)/µ l, were mostly lowerthan 2000, i.e., lower than the laminar–transition threshold,and argued that this was inconsistent with the original form

of Chen’s model (similar to the reasoning of Kandlikar and asubramanian [24]) Chen’s superposition model for convectiveboiling states that heat is transferred by two competing mech-anisms, namely, nucleate boiling and convective vaporization.The overall heat transfer coefficient is given by an additive lawthat combines these different contributions,

The nucleate boiling term in Eq (6) is expressed as theproduct of the nucleate pool boiling value (αnpb) computed

at the corresponding wall superheat through the Forster and

Zuber [26] correlation, and a boiling suppression factor, S, that

accounts for the suppression of bubble nucleation due to theconvective nature of the two-phase system On the other hand,the convective contribution depends on the flow properties and

is given as an all liquid heat transfer coefficient multiplied by a

two-phase correction factor, F That is:

For the all liquid heat transfer coefficients in Eqs (7) and(8), Zhang et al [25] suggested using a laminar or turbulentexpression according to the value of the liquid Reynolds number,

Rel For the two-phase factor, F , they proposed using the larger value of 1 and an expression, F, based on the general form of

the Martinelli parameter, X:

where C is Chisholm’s constant For the suppression factor, S,

they presented a form similar to the one given by Chen:

Trang 40

J R THOME AND L CONSOLINI 293(with a liquid Reynolds number in the place of the two-phase

Reynolds number as originally proposed by Chen), justifying

their choice by assuming the nucleate boiling suppression

mech-anism to remain the same as in the macro-scale

Further Remarks

Some further comments about the methods just described are

in order All of the preceding methods are essentially

modifica-tions of macro-scale flow boiling methods, and thus assume that

nucleate boiling is an important heat transfer mechanism

with-out proof of its existence for the two principal micro-channel

flow regimes: slug (elongated bubble) flow and annular flow

Furthermore, using a tubular single-phase flow correlation to

predict convective heat transfer in an annular flow is not

phys-ically realistic since an annular flow is a film flow and is thus

governed by its film Reynolds number rather than by a tubular

Reynolds number Similar to Nusselt’s laminar film

condensa-tion theory, as long as there are no interfacial waves, the local

laminar annular flow heat transfer coefficient is dependent on

heat conduction across the liquid film thickness, and it is thus

not appropriate to calculate its value in terms of the tubular

so-lution of Nul = 4.36 For instance, no one applies the tubular

flow solution to predict laminar falling film condensation on a

vertical plate, so it does not seem to be appropriate to apply it to

an evaporating laminar annular film flow either For that matter,

turbulent falling film condensation on a vertical plate is

corre-lated based on its local liquid film Reynolds number in which

the film thickness is the active characteristic dimension, and

hence turbulent annular film evaporation should be correlated

in the same manner, not using a tubular flow Reynolds number

Not withstanding the preceding comments, wholly empirical

methods can be fitted to experimental databases for prediction

purposes On the other hand, none of the preceding correlations

is able to predict the diverse trends in the heat transfer coefficient

versus vapor quality described earlier

MECHANISTIC PREDICTION METHOD

Thome et al Three-Zone Evaporation Model for Slug Flow

Thome and coworkers [8] developed a phenomenological

method to describe heat transfer for a purely convective

micro-channel slug flow (no nucleate boiling), based principally on the

following assumptions:

1 The vapor and liquid travel at the same velocity

2 The heat flux is uniform and constant with time along the

inner wall of the micro-channel

3 All energy entering the fluid is used to vaporize liquid Thus,

the temperatures of the liquid and vapor remain at the

satura-tion value, i.e., neither the liquid nor the vapor is superheated

Figure 6 Image of an elongated bubble (top) and a schematic diagram of the three-zone evaporation model (bottom).

4 The local saturation pressure is used for determining the localsaturation temperature

5 The liquid film remains attached to the wall The influence

of vapor shear stress on the liquid film is assumed negligible,

so that it remains smooth without ripples

6 The thickness of the film is very small with respect to theinner radius of the tube

7 The thermal inertia of the channel wall can be neglected

Slug flow was modeled as a cyclical passage of a zone” sequence, comprising a liquid slug, an elongated bubblesurrounded by an evaporating liquid film, and an all-vapor dry-zone at the bubble tail (see Figure 6), respectively referred to

“three-by subscripts L, F , and D The first assumption yields an equal velocity, W , for the vapor and liquid, given by the homogeneous

elongated bubble plus the dry zone (t V = t F + t D), and

the liquid slug, t L, were derived from the definition of vaporquality and through Eq (11) as

with t the passage period of the three-zone structure, and

having neglected the liquid mass within the film The local filmbehavior was modeled as the evaporation of a stagnant liquid

layer; from an energy balance at given axial position, z:

2π R qdz = −2π ρ l (R− δ)

dt h lv dz (13)

heat transfer engineering vol 31 no 4 2010

Ngày đăng: 19/07/2016, 06:15

TỪ KHÓA LIÊN QUAN

🧩 Sản phẩm bạn có thể quan tâm