Sei sulla pagina 1di 11
Seediscussions,stats,andauthorprofilesforthispublicationat: https://www.researchgate.net/publication/260608912 Comparison

Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/260608912

Article in InternationalJournalofThermalSciences·June2014

DOI:10.1016/j.ijthermalsci.2014.01.014

CITATIONS

37

READS

669

3authors,including:

CITATIONS 37 READS 669 3authors ,including: KuntAtal ı k HakanErturk BogaziciUniversity
KuntAtal ı k HakanErturk

22 PUBLICATIONS 170 CITATIONS

65 PUBLICATIONS 293 CITATIONS

Someoftheauthorsofthispublicationarealsoworkingontheserelatedprojects:

ThinFilmsandCoatingswithSpectralAlterationforEnergyApplications Viewproject Viewproject

NumericalThermalCharacterizationofWater-hBNNanofluids Viewproject Viewproject

AllcontentfollowingthispagewasuploadedbyHakanErturkon17May2014.

Theuserhasrequestedenhancementofthedownloadedfile.

International Journal of Thermal Sciences 80 (2014) 83 e 92 Contents lists available at ScienceDirect International

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

journal homepage: www.els evier.com/locate/ijts

Sciences journal homepage: www.els evier.com/locate/ijts Comparison of single and two-phase models for nano fl uid

Comparison of single and two-phase models for nano uid convection at the entrance of a uniformly heated tube

Sinan Göktepe, Kunt Atal ı k, Hakan Ertürk *

Department of Mechanical Engineering, Bo gaziçi University Istanbul, Turkey

Engineering, Bo gaziçi University Istanbul, Turkey article info Article history: Received 12 February 2013

article info

Article history:

Received 12 February 2013 Received in revised form 8 November 2013 Accepted 14 January 2014 Available online

Keywords:

Nanouids Single-phase Two-phase Heat transfer enhancement Laminar ow Forced convection Full multiphase coupled

abstract

Macroscopic modeling of hydrodynamic and thermal behavior of nano uid ows at the entry region of uniformly heated pipe is studied. Single-phase models with and without thermal dispersion effect, Eulerian e Eulerian, and Eulerian e Mixture two-phase models are evaluated by comparing predicted convective heat transfer coef cients and friction factors with experimental results from literature. So- lutions with two different velocity e pressure coupling algorithms, Full Multiphase Coupled, and Phase Coupled Semi-Implicit Method for Pressure Linked Equations are also compared in terms of accuracy and computational cost. Dispersion model that uses velocity gradient to de ne dispersion conductivity is found to be more effective at entry region compared to other single-phase models. However, two-phase models predict convective heat transfer coef cient and friction factor more accurately at the entry re- gion. Moreover, computational cost of Eulerian e Eulerian two-phase model can be reduced up to 50% by implementing Full Multiphase Coupled scheme.

2014 Elsevier Masson SAS. All rights reserved.

1. Introduction

Engineered uids made of a base uid and nano sized particles such as CuO, Al 2 O 3 , or TiO 2 [1] that form colloidal suspensions are referred as nano uids [2] . The most commonly used base uids are water and ethylene glycol due to their use in conventional thermal systems. Measured thermal conductivities of nano uids are found to be exceeding predictions based on the Maxwell s effective me- dium theory that led many researchers to consider nano uids as next generation heat transfer uids [3] . Therefore, nano uids are considered for many engineering applications such as, cooling of electronics [4,5] , vehicle thermal management [6] , and solar energy systems [7,8] . Design and analysis of such systems necessitate ac- curate estimation of hydrodynamic and thermal characteristics of nano uids. Many experimental studies were carried out to quantify thermal and ow characteristics of nano uids for laminar and turbulent ow conditions [2,9 e13] . However, it is important to be able to model nano uid ow accurately in order to design equipment that operates with nano uids. It was observed that addition of

* Corresponding author. Tel.: þ 90 212 359 7356. E-mail addresses: snangoktepe@gmail.com (S. Göktepe), atalik@boun.edu.tr (K. Atalı k), hakan.erturk@boun.edu.tr (H. Ertürk).

1290-0729/ 2014 Elsevier Masson SAS. All rights reserved.

nanoparticles to a base uid, augments convective heat transfer together with an increase in pressure drop due to increased ther- mal conductivity and viscosity [1,2,9 e13] . Therefore, modeling tools should estimate both of these behaviors accurately. Macroscopic models for nano uid ow and heat transfer can be classi ed as single-phase and two-phase models [1 e3,7,9 e15] . Single-phase approaches consider nanoparticles and base uid as a single homogeneous uid with respect to its effective properties [14] . Two-phase approaches handle continuity, momentum and energy equations for particles and base uid using three different methods. One of these methods used in this study is Eulerian e Mixture model (EMM) where momentum and energy equations are solved for mixture phase coupled with continuity equation for each phase, then phase velocities are related by empirical correlations [16,17] . The other method that is used in this study is the Eulerian e Eulerian model (EEM) where separate continuity, momentum, and energy equations for each phase are solved. This approach is sug- gested for ows where interactions between phases are not well de ned [17,18] . Although two-phase models provide a better un- derstanding of both phases, single-phase models are computa- tionally more ef cient, however provide less detail about each phase [18] . Forced convection of Al 2 O 3 e water/EG nano uids in a uniformly heated tube at fully developed laminar and turbulent ow regimes using a homogeneous single-phase model is studied by Maiga et al.

84

S. Göktepe et al. / International Journal of Thermal Sciences 80 (2014) 83e92

[14] . While their predictions underestimate measured heat transfer coef cients, results indicate that addition of nanoparticles en- hances convective heat transfer coef cient of Al 2 O 3 e water nano- uid with 10% particle concentration by 60% at a Reynolds number of 250 [14] . Experimental studies such as [19] reported that the increase in convective heat transfer coef cient exceeds that of effective thermal conductivity. This indicates that there are different mechanisms in heat transfer enhancement for forced convection other than the enhancement in thermal conductivity. Single-phase thermal dispersion models are introduced in Refs. [20] and [21] to account for energy transport by random movement of nanoparticles, that is also known as thermal dispersion effects. Using thermal dispersion model presented in Ref. [20] , Ozerinc et al. [22] studied fully developed laminar forced convection of Al 2 O 3 e water nano uid by considering temperature dependent properties. The reported increase in convective heat transfer coef- cient of 2.5% Al 2 O 3 e water nano uid is 36% at a Peclet number of 6500. The results are in good agreement with the experimental data in the literature, suggesting that single-phase models considering thermal dispersion and temperature dependent properties are capable of predicting heat transfer behavior more accurately. Moraveji et al. [23] showed that convective heat transfer increases as particle size decreases for developing Al 2 O 3 e water nano uid ow using single phase models. Mirmasoumi et al. [24] investigated mixed convection of Al 2 O 3 e water nano uid in a horizontal tube using a two-phase EMM. They have shown that particle concentration is higher near the wall and bottom of the tube, hence uniform particle distribution is not valid for all cases. Nano uid forced convection in developing ow in a tube subjected to constant heat ux and temperature was studied by Bianco et al. [25] by using single and two-phase models including volume of uid (VOF), EMM, EEM considering both constant and temperature dependent properties. According to their results difference between homogeneous single-phase and two- phase mixture model becomes signi cant at 11% volume concen- tration. Moreover, consideration of temperature dependent prop- erties gives a better estimation of convective heat transfer coef cient. They observed that convective heat transfer coef cient for 2.5% Al 2 O 3 e water nano uid at a Reynolds number of 250 in- creases up to 17%. Kalteh et al. [18] numerically studied CuO e water nano uid laminar forced convection in a micro-channel by two-phase EEM. Although velocity and temperature differences between phases are negligible, EEM estimates convective heat transfer coef cient more accurately with respect to single-phase models. They also showed that particle e particle interactions have negligible effect on Nusselt number for laminar ow. Lotet al. [16] evaluated homogeneous single-phase model, EMM, and EEM for Al 2 O 3 e water nano uid. The study neglected temperature dependency of properties and did not include thermal dispersion models. They reported that two- phase models overestimate fully developed heat transfer co- ef cients and EMM is the most accurate model among three two- phase models (EEM, EMM, VOF). Akbari et al. [26] compared sin- gle and two-phase models for mixed convection heat transfer of Al 2 O 3 e water nano uid. Their study covers homogeneous single- phase and two phase models (VOF, EMM, EEM) with temperature dependent properties. It is reported that estimated convective heat transfer coef cients by two-phase models are similar. Two-phase models provide more accurate prediction of convective heat transfer coef cient with an overestimation, whereas single-phase model underpredicts convective heat transfer coef cient. Although single-phase models are found to be less accurate, it should be noted that the study did not include thermal dispersion models. Single and two-phase models for Al 2 O 3 e water nano uid are also studied by Frad et al. [15] . They showed that two-phase

models provide more accurate prediction of heat transfer of nano uids for fully developed ow. Although they are more accurate in predicting heat transfer, two-phase models are computationally more expensive than single-phase models due to the increased number of equations to be solved. Despite being expensive, the Phase Coupled Semi Im- plicit Method for Pressure Linked Equations (PC-SIMPLE) algorithm is widely used in literature due to its robustness [16,18,24,26] for EEM. Computational cost of two-phase EEM can be reduced by using Full Multiphase Coupled (FMC) algorithm for velocity and pressure coupling where equations are solved simultaneously rather than in a segregated manner like PC-SIMPLE. Considering the literature, there is no complete study that considers recent state-of-the art single and two-phase models for laminar forced convection of nano uids. This study aims at evalu- ating single-phase and two-phase models by considering the effect of temperature dependent properties, and dispersion effects for single-phase models, together with the rst time use of FMC for two-phase EEM of nano uid forced convection. Results are compared with experimental data available in literature in terms of error and required CPU time. Al 2 O 3 e water nano uid with 42 nm nanoparticles is considered throughout the study due to availability of experimental data in the literature.

2. Mathematical models

2.1. Single-phase model

Single-phase models assume that base uid and nanoparticles have the same temperature and velocity eld. Therefore, continu- ity, momentum and energy equations can be solved as if the uid were a classical Newtonian uid by using effective properties of nano uid. Effective properties are functions of particle size ( d p ), type, shape, and particle volume concentration ( f p ) and tempera- ture [2,14,27] . In this study, for the homogeneous single-phase (SPM) Al 2 O 3 e water nano uid model with constant properties, nano uid thermal conductivity ( k nf ) is determined by the correlation reported by Hamilton e Crosser [28] to take a simpler thermal conductivity model into account together with other advanced models. The formulation can be given as;

k

nf

k

bf

k p þ ð n 1 Þ k bf ð n 1 Þ k bf k p f p

¼

k p þ ð n 1 Þ k bf þ k bf k p f p

(1)

where, k bf and k p are the base uid and nanoparticle thermal conductivities, respectively and n ¼ 3 is the shape factor for spherical particles. This correlation does not consider temperature and particle size dependency of thermal conductivity. To account for temperature and particle size dependency of thermal conduc- tivity of Al 2 O 3 e water nano uid, correlation suggested by Chon et al. [29] is used. This is a more advanced correlation compared to that of presented by Hamilton and Crosser [28] . The formulation can be given as;

k

nf

k

bf

¼

1 þ 64 : 7 f 0 : 7460

p

d bf

d

p

0 : 3690

bf ! 0 : 7476

k

p

k

Pr 0 : 9955 Re 1 : 2321 disp

(2)

where, d bf is molecular diameter of base uid (0.29 nm, for water). For this study we considered Al 2 O 3 particles with 42 nm diameter. The Prandtl number ( Pr ) and the dispersion Reynolds number ( Re disp ) in this correlation are de ned as;

S. Göktepe et al. / International Journal of Thermal Sciences 80 (2014) 83e 92

85

Pr ¼

m

bf bf

r bf a

Re disp ¼

r bf k B T

3

pm

2

bf l bf

(3)

(4)

where, r bf is the density of base uid, k B is the Boltzmann constant, T is the absolute temperature, l bf is the mean free path between base uid molecules (0.17 nm, for water), a bf is the thermal diffu- sivity of base uid and m bf is the temperature dependent base uid viscosity (Pa-s) that is de ned in Ref. [29] as follows;

m bf

¼ 2 : 414 10 5 10

247

: 8

T

140

(5)

In this paper homogeneous single-phase model with tempera- ture dependent thermal conductivity is referred as SPT. For the homogeneous single-phase models, effective thermal conductivities are de ned as;

k eff ¼ k nf

(6)

Density ( r nf ) and heat capacity of nano uid ( c p,nf ), on the other hand, are estimated by using classical mixture models [10,14,25] as follows;

c p ; nf ¼ 1 f p r bf c p ; bf þ r p f p c p ; p

r nf

r nf ¼ 1 f p r bf þ f p r p

(7)

(8)

Maiga et al. [14] presented a 2nd degree polynomial curve t to experimental data for Al 2 O 3 e water nano uid viscosity as;

m

nf

m bf

2

¼ 123 f p þ 7 : 36 f p þ 1

(9)

which is used for all single phase models in this study.

2.2. Single-phase dispersion model

Single-phase dispersion model and homogeneous models are differentiated by how the nano uid effective conductivity is de ned. Xuan and Roetzel [20] introduced small perturbations in the momentum and energy equation to account for dispersion ef- fects. Dispersion effects are then represented as an additional term in nano uid effective conductivity and referred as dispersion con- ductivity ( k d ) and can be de ned as in Refs. [20] and [21] . For dispersion models, effective nano uid conductivity is given

as;

k eff ¼ k nf þ k disp

(10)

where, k nf is the nano uid thermal conductivity determined by a correlation such as Eq. (1) or (2) . Formulation suggested by Xuan and Roetzel [20] (SPD1) can be given as;

k disp ¼ C 1 rc p nf f p d p Ru

(11)

where, C 1 is an empirical constant that calibrates model to exper- imental data, u is the ow velocity in x direction, and R is the radius of pipe. A second formulation is suggested by Mokmeli and Saffar- Avval [21] (SPD2) and it can be given as;

k disp ¼ C 2 rc p nf f p

v u R

v

r

d

p

(12)

where, C 2 is an empirical constant that calibrates model to exper- imental data. When C 1 or C 2 are set to zero, both models are reduced to the homogeneous single-phase model.

2.3. Two-phase models

For two-phase nano uid models, it is assumed that base uid and nanoparticles can have different velocity and temperature elds. Volume of Fluid (VOF), Eulerian e Mixture model (EMM), and Eulerian e Eulerian model (EEM) are the three common two-phase models used in modeling of nano uids [26] . Volume of uid model is not considered in this study since it is suggested for free surface ows [30] .

2.3.1. Eulerian e Mixture two-phase model In EMM, continuity, momentum, and energy equations are solved for the mixture-phase and the phase velocities are deter- mined by empirical correlations. The continuity equation for EMM is given as;

V $ ð r m ! m Þ ¼ 0

v

(13)

where, the mass-averaged velocity or mixture velocity, ! v m ; for two-phase mixture is de ned as;

! m ¼ r p f p ! p þ r bf f bf ! bf

v

v

v

r m

(14)

where, ! v p is the particle velocity, ! v bf is the base uid velocity, and r m is the mixture density for two-phase mixture that is de ned as;

r m ¼ f p r p þ f bf r bf

(15)

The steady state momentum equation for two-phase mixture is;

r

m

!

v m $ V v

!

m

¼ V P þ m m V v þ V v

m

!

!

þ V $

f r

bf bf

!

v

dr ; bf

!

v

dr ; bf

m

þ

T

f r

p p

!

v

dr ; p

!

v

dr ; p

(16)

where, P is the pressure, m m is the viscosity of mixture that is equal to base uid viscosity ( m m ¼ m bf ), ! v dr ; p ; and ! v dr ;bf are the drift

velocity of particles and base uid, respectively. Here, ! v dr ; p and

! v dr ; bf for two-phase mixture are de ned as;

! dr ; p ¼ v

v

! p

! v m

! dr ; bf ¼ v

v

! bf v

! m

(17)

(18)

The steady state energy equation for two-phase mixture is given

as;

V $ f p v

! p r p i p þ f bf

! bf r bf i bf ¼ V $ k eff V T

v

(19)

where, i bf and i p are the enthalpy of base uid and particles, respectively. Effective thermal conductivity for two-phase mixture model k eff is de ned as;

k eff ¼ f p k p þ f bf k bf

(21)

The volume fraction equation for two-phase mixture is;

V $ f p r p

! v m ¼ V $ f p r p v

! dr ; p

(22)

86

S. Göktepe et al. / International Journal of Thermal Sciences 80 (2014) 83e92

The slip velocity or relative velocity, represents velocity of the particles ð ! v p Þ relative to base uid ð ! v bf Þ and it is de ned as;

! bf ; p ¼ v

v

! bf v

! p

(23)

The relation between drift velocity and relative velocity can be presented as;

! dr ; p ¼ ! v p ; bf

v

r p f p

r

m

! bf ; p

v

(24)

The relative velocity is given by Manninen et al. [31] through Schiller and Naumann [32] drag formulation as follows;

! v bf ; p ¼

d

2

p

r p r m

18 m bf f d

r

p

! a

f d ¼

1 þ 0 : 15 Re 0 : 687 0 : 0183 Re p

p

!

a ¼ g

! ð v

! m $ V Þ v

! m

Re p 1000 Re p > 1000

(25)

(26)

(27)

where, ! a and ! g are the particle s and gravitational acceleration, respectively. The particle Reynolds number ( Re p ) for EMM is de ned as;

Re p ¼ U m d p r m

m

m

(28)

2.3.2. Eulerian e Eulerian two-phase model In EEM, momentum and energy equations are solved for each phase. Interactions between phases are de ned by additional terms that represent momentum and heat exchange between phases [33] . The steady continuity equation for each phase is given as;

v f bf r bf u bf

þ

1

v f bf r bf v bf

v

x

r

v

r

¼ 0

v f p r p u p

v r f p r p v p

 

þ 1

 

v

x

r

v

r

¼ 0

(29)

(30)

where, u and v are the velocity components in axial ( x ) and radial ( r ) directions, respectively. For steady ow, axial-momentum equations for base uid and particles can be given, respectively as;

u bf v f bf r bf u bf þ u bf v f bf r bf v bf

v

v

v x þ v

þ ð F d Þ x þ ð F vm Þ x

x

P

v

r

x f bf m bf v u bf

v

v

x

þ 1 r v

¼ f bf

r f bf m bf r v u bf

v

v

r

u p v f p r p u p

v x

þ

u p v f p r p v p ¼ f p

v

r

P

v

v x þ v

x f p m p

v

v

u p

v

x

f p m p r v u p

þ 1 r v

ð F vm Þ x þ ð F col Þ x

v

r

v

r

ð F d Þ x

(31)

(32)

Similarly, the radial-momentum equations can be written as;

v bf v f bf r bf u bf þ v bf v f bf r bf v bf

v

x

v P

v

r

v

r

x f bf m bf v v bf

v

þ v

þ

v

x

ð F d Þ r þ ð F vm Þ r

þ 1 r v

¼ f bf

m bf v bf 2

r

r f bf m bf r v v bf

v

v

r

(33)

v p v f p r p u p

v x

þ

v p v f p r p v p

v r

¼ f p v r þ v x f p m p v x

v

P

v

v

v p

r f p m p

þ 1 r v

ð F v m Þ r þ ð F col Þ r

v

v

v p

v

r

m p

v

p

2

r

ð F d Þ r

(34)

where, ! F vm is the virtual mass force due to relative acceleration of phases [17] . Since it has negligible effect on heat transfer as shown

in Ref. [18] , it is neglected in this study. ! F col is particle e particle

collision force, and F d is the drag force between the uid and particle phases de ned by phase interaction equations. The drag

force between phases is de ned as;

!

!

F d

¼ b v

! bf v

! p

(35)

where, b is the friction factor, which depends on particle volume concentration and particle size. For dilute solutions ( f bf 0.8), b is de ned by Syamlal and Gidaspow [33] as;

b ¼

3

4

C d f p f bf

d

p

! v bf ! v p

f

2 : 65 bf

(36)

where, C d is the drag coef cient and can be predicted by;

C d ¼

8

<

:

p 1 þ 0 : 15 Re 0 : 697

24

Re

p

0 : 44

Re p 1000

Re p < 1000

(37)

Here, particle Reynolds number, Re p , is de ned as;

Re p ¼

f bf r bf

! v bf v d p

!

p

m bf

(38)

Collision force of particles, ! F col ; is de ned by Bouillard et al. [34] as;

! F col ¼ G f bf V f bf

(39)

where, G ( f bf ) is the particle e particle interaction modulus and it is

de ned as;

G f bf ¼ exp 600 f bf 0 : 376

(40)

Once momentum equations for each phase and the related

phase interaction equations are de ned, energy equations for each phase can be presented as follows;

v

x f bf r bf u bf c p ; bf T bf þ

v

v

r f bf r bf v bf c p ; bf T bf

v

¼

x f bf k eff ; bf v T bf

v

v

v

x

h v T bf T p

þ 1 r v

v

r f bf k eff ; bf r v T bf

v

r

(41)

S. Göktepe et al. / International Journal of Thermal Sciences 80 (2014) 83e 92

87

v

v

x

f p r p u p c p ; p T p þ

v

v

r

f p r p v p c p ; p T p

¼

v

v

x

f p k eff ; p

v

T p

v

x

þ 1 r v

v

r

f p k eff ; p r v T p

v

r

þ h v T bf T p

(42)

where, k eff,p and k eff,bf are the effective thermal conductivity of base uid and particles, and h v is volumetric interphase heat transfer coef cient between phases. The temperature for particle and base uid phases are T p and T bf , respectively. Volumetric interphase heat transfer coef cient is de ned as;

h v ¼ 6 1 f bf

d p

h p

(43)

For mono dispersed particles, particle heat transfer coef cient, h p , is estimated by a correlation presented by Wakao and Kaguei [35] through particle Nusselt number ( Nu p ) as;

Nu

p ¼ h p d p ¼ 2 þ 1 : 1 Re 0 : 6 Pr 1 = 3

k

bf

p

(44)

where, Pr is the Prandtl number of base uid. The particle Reynolds number is de ned by Eq. (38) . The effective conductivities for particle and base uid phases are presented as [18] ;

3.1. Single-phase model

For single-phase model, momentum, energy, and continuity

equations are solved with effective nano uid properties for prob- lem domain in Fig. 1. Third order power-law scheme is used in discretization of momentum, energy and pressure equations. Semi Implicit Method of Pressure Linked Equations (SIMPLE) algorithm is used for velocity pressure coupling. A convergence criterion is used so that residuals for all equations are less than 1 10 6 . Boundary conditions for governing equations of single-phase model are de ned considering the no-slip condition at the wall;

u ð x ; R Þ ¼ v ð x ; R Þ ¼ 0

and uniform inlet velocity

u ð 0 ; r Þ ¼ U ;

v ð 0 ; r Þ ¼ 0

with constant wall heat ux ð q w Þ condition, expressed as;

00

k eff

v

T

v

r

r ¼ R

00

¼ q w

(52)

(53)

(54)

Dispersion model empirical constants C 1 and C 2 in Eqs. (14) and (15) are determined according to the experimental data provided by Ref. [12] .

k eff ; bf ¼ k b ; bf

f

bf

and

k eff ; p ¼ k b ; p

k eff ; p ¼ k b ; p

f

p

where, k b,p and k b,bf are de ned as;

k b ; bf

¼ 1

p

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1 f bf

k bf

k b ; p ¼

p

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1 f bf

ð uA þ ½ 1 u GÞ k bf

(45)

(46)

(47)

(48)

with u ¼ 7.26 10 3 for spherical particles and G is de ned as;

G ¼

0

@

2

B ð A 1 Þ

2 ln A

1 B

B

A

A 1

B
A

B 1 B þ 1

1 B

2

A

1 A

(49)

3.2. Two-phase models

Eulerian e Eulerian model (EEM) and EMM are used with the third order power-law scheme for solving momentum and energy equations. The third order Quadratic Upstream Interpolation for Convective Kinetics (QUICK) scheme is used to solve volume frac- tion equation. For EMM, SIMPLE algorithm is used for pressure and velocity coupling. However, for EEM, in addition to Phase Coupled SIMPLE (PC-SIMPLE) algorithm that is widely used in literature, a different algorithm, namely Full Multiphase Coupled (FMC), is used for coupling scheme to assess its accuracy and ef ciency. Phase Coupled-SIMPLE is a well-established, widely used, and robust algorithm, where phase velocities are solved in a segregated manner with a pressure correction. Velocity, shared pressure, and volume fraction corrections are coupled simultaneously in FMC algorithm. Although FMC is expected to be more ef cient, diver- gence may occur at high particle volume concentrations due to volume fraction equation [17] . Boundary conditions for EMM are de ned based on no-slip

condition at wall for mixture phase as;

u m ð x ; R Þ ¼ v m ð x ; R Þ ¼ 0

(55)

where, parameters A and B for spherical particles are given as;

A ¼ k p

k

bf

B ¼ 1 : 25 1 f bf

f

bf

! 10

9

and uniform inlet velocity for base uid and particle phases, respectively;

(50)

u bf ð 0 ; r Þ ¼ U bf ;

v bf ð 0 ; r Þ ¼ 0

(56)

(51)

u p ð 0 ; r Þ ¼ U p ;

v p ð 0 ; r Þ ¼ 0

(57)

where, U p ¼ U bf are mean velocities of base uid and particles. Uniform constant heat ux at wall is applied as,

3. Problem statement and numerical method

Finite Control Volume method is used to solve numerically the equations for nano uid ow and heat transfer. The problem domain is presented in Fig. 1 that is discretized by a uniform structured grid.

k eff

v

T

v

r

r ¼ R

00

¼ q w

(58)

Similarly, the boundary conditions for EEM are de ned with respect to no-slip at boundaries for base uid and particle phases as;

88

S. Göktepe et al. / International Journal of Thermal Sciences 80 (2014) 83e92

International Journal of Thermal Sciences 80 (2014) 83 e 92 Fig. 1. Problem geometry and boundary

Fig. 1. Problem geometry and boundary conditions.

u bf ð x ; R Þ ¼ v bf ð x ; R Þ ¼ 0

u p ð x ; R Þ ¼ v p ð x ; R Þ ¼ 0

(59)

(60)

with uniform inlet velocity for base uid and particle phases as;

u bf ð 0 ; r Þ ¼ U bf ;

v bf ð 0 ; r Þ ¼ 0

u p ð 0 ; r Þ ¼ U p ;

v p ð 0 ; r Þ ¼ 0

(61)

(62)

and constant wall heat ux for particle uid mixture as;

f p k eff ; p

v

T p

v

r

r ¼ R þ f bf k eff ; bf v T bf

v

r

3.3. Model validation

r ¼ R ¼ q

00

w

(63)

A grid independence study is carried out in order to ensure grid independent solutions. Three different grid resolutions are compared with correlation given by Shah [36] and experimental data reported in Ref. [12] . The results indicate that 15 2000 grid yielded a grid independent solution. For both single and two phase models the local Nusselt number for nano uids ( Nu nf,x ) de ned as;

Nu nf ; x ¼ h ð x Þ D

k

nf

q

00

w D

k nf ð T w T mean Þ

¼

(64)

Variation in local Darcy friction factor ( f x ) was also checked, and results indicate that 15 2000 grid has 0.7% error with respect to the theoretical value of Darcy friction factor for the fully developed region. Darcy friction factor used here is de ned as;

f x ¼ 8 s w ð x Þ

r bf U 2

(65)

where, s w ( x ) is local wall shear stress, U is the mean axial velocity. Reynolds number for nano uids is de ned as;

Re ¼ r nf DU

m nf

4. Results and discussion

(66)

The goal of this study is to assess effectiveness of single and two- phase models for predicting thermal and hydrodynamic charac- teristics of nano uids. Therefore, solutions for single-phase ho- mogeneous, single-phase dispersion, two-phase Eulerian e Mixture, and two-phase Eulerian e Eulerian models are compared with experimental data presented by Wen and Ding [12] at x / D ¼ 63 and

x / D ¼ 115 for a Reynolds number of 1050. Comparison of single- phase models that are used here is presented in Table 1 to clarify any ambiguity regarding de nition of single-phase models. Comparisons of local Nusselt numbers for 1.6% Al 2 O 3 e water nano uid are presented in Figs. 2 and 3 . Single-phase models un- derestimate Nu nf, x at the entry region of circular tube, whereas both two-phase models overestimate Nu nf, x as can be seen from Figs. 2 and 3 . Error with experimental values are 31.4% and 19.6% at x / D ¼ 63 and x / D ¼ 116, respectively for SPM. For EEM, error values are 0.9% and 7.7% at the same locations, respectively. Nu x pre- dictions of our EMM and EMM by Ref. [26] are very close to each other. Slight difference between models might be due to variation

of thermal properties of Al 2 O 3 used. In this study thermophysical

properties of Al 2 O 3 are taken from National Institute for Standards and Technology [37] . An additional comparison with experimental data from Ref. [12] in terms of local convective heat transfer coef- cient ( h x ) versus x / D is also presented in Figs. 4 e7 to prevent any possible ambiguity due to de nition of nano uid thermal conductivity. Fig. 4 indicates that at the very beginning of entry region, effect of temperature dependency of properties is limited since the ow is not fully affected by the thermal boundary conditions and the temperature difference is not signi cant. The effect increases as the ow develops as it can be observed from comparison of SPM and SPT. Table 2 suggests that use of temperature dependent nano uid conductivity model increases solution accuracy up to 4% for single- phase models. Comparison of dispersion models reveals that, SPD2, that uses dispersion conductivity formulation by Eq. (12) , is more accurate in predicting heat transfer coef cient at entry region compared to SPD1, which uses formulation by Eq. (11) . As shown in Table 2 , SPD2 model is approximately 8% more accurate in pre- dicting heat transfer coef cient compared to SPD1. In Table 2 and Fig. 5 , it is also shown that for volume fraction of 1.6% at Reynolds number of 1050, predictive accuracy of EMM becomes superior to that of EEM as the ow develops. However the difference between two models is very small and as far as the accuracy of the models are considered, such difference can be neglected. It is also observed that both EEM and EMM start over-predicting as ow develops. Based on Figs. 2 and 4 , and Tables 2 e 4 , it can be concluded that the best single-phase model is SPD2. However, there is no clear indication on the better two-phase model, since both EEM and

Table 1 Single-phase models and corresponding effective property models.

Model name

Thermal conductivity model(s)

Viscosity model

k nf

k disp

m nf

SPM

Eq. (1)

0

Eq. (9)

SPT

Eq. (2)

0

Eq. (9)

SPD1

Eq. (2)

Eq. (11)

Eq. (9)

SPD2

Eq. (2)

Eq. (12)

Eq. (9)

S. Göktepe et al. / International Journal of Thermal Sciences 80 (2014) 83e 92

89

Journal of Thermal Sciences 80 (2014) 83 e 92 89 Fig. 2. Comparison of single-phase models

Fig. 2. Comparison of single-phase models with experimental data [12] for 1.6% Al 2 O 3 ewater nano uid at Re ¼ 1050.

Fig. 5. Comparison of convective heat transfer coef cients ( h x ) of two-phase model with experimental data at Re ¼ 1050 for 1.6% Al 2 O 3 e water nanouid.

¼ 1050 for 1.6% Al 2 O 3 e water nano fl uid. Fig. 3. Comparison

Fig. 3. Comparison of two-phase models with experimental data [12] for 1.6% Al 2 O 3 e water nano uid at Re ¼ 1050.

Al 2 O 3 e water nano fl uid at Re ¼ 1050. Fig. 4. Comparison

Fig. 4. Comparison of convective heat transfer coef cients ( h x ) of single-phase model with experimental data at Re ¼ 1050 for 1.6% Al 2 O 3 ewater nano uid.

Fig. 6. Comparison of convective heat transfer coef cients of SPD2 with experimental data at Re ¼ 1050 [12] for three different volume concentrations (0.6%, 1%, and 1.6%).

three different volume concentrations (0.6%, 1%, and 1.6%). Fig. 7. Comparison of convective heat transfer coef

Fig. 7. Comparison of convective heat transfer coef cients of EEM with experimental data at Re ¼ 1050 [12] for three different volume concentrations (0.6%, 1%, and 1.6%).

90

S. Göktepe et al. / International Journal of Thermal Sciences 80 (2014) 83e92

Table 2 Error in convective heat transfer coef cient for f p ¼ 1.6%.

Model

x / D ¼ 22

x / D ¼ 63

x / D ¼ 116

x / D ¼ 178

EEM

12.1%

7.7%

11.3%

25.6%

EMM

12.8%

8.1%

10.8%

25.0%

SPM

38.4%

37.6%

27.3%

21.3%

SPT

38.2%

36.5%

25.1%

17.5%

SPD1

32.8%

28.0%

12.1%

 

0.0%

SPD2

24.2%

23.1%

 

9.5%

0.1%

Table 3 Error in convective heat transfer coef cient for f p ¼ 1%.

 

Model

x / D ¼ 22

x / D ¼ 63

x / D ¼ 116

x / D ¼ 178

EEM

16.7%

5.7%

4.4%

15.5%

EMM

17.4%

6.2%

3.9%

14.9%

SPM

36.6%

29.5%

23.6%

17.3%

SPT

35.9%

28.7%

22.6%

16.2%

SPD1

31.0%

21.0%

11.7%

1.7%

SPD2

24.9%

17.0%

 

9.5%

1.3%

Table 4 Error in convective heat transfer coef cient for f p ¼ 0.6%.

 

Model

x / D ¼ 22

x / D ¼ 63

x / D ¼ 116

x / D ¼ 178

EEM

14.4%

6.0%

0.6%

8.5%

EMM

19.2%

6.5%

1.2%

7.8%

SPM

31.4%

22.0%

18.8%

12.6%

SPT

12.8%

20.5%

17.0%

10.6%

SPD1

27.8%

16.2%

10.9%

 

2.0%

SPD2

23.8%

13.6%

 

9.6%

1.6%

EMM have different accuracies at different volume fractions as shown in Tables 2 e 4 . For example, for volume fraction of 1% EMM model prediction error is smaller than that of EEM. However, for volume fraction of 0.6%, the opposite is true. Overall, results suggest that at entrance region, EEM performs better with low particle concentrations, whereas EMM performs better at higher particle volume concentrations. The prediction accuracies of SPD2 and EEM are investigated at different concentrations in Figs. 6 and 7 with Tables 2 e 4 . Results indicate that, EEM model underestimates convective heat transfer

that, EEM model underestimates convective heat transfer Fig. 9. Comparison of estimated Darcy friction factors of

Fig. 9. Comparison of estimated Darcy friction factors of single-phase model at Re ¼ 1050 for 1.6%, 1%, and 0.6% Al 2 O 3 e water nanouid.

coef cient at the beginning of entry region where, SPD2 under- predicts until the calibration point ( x / D ¼ 176). However, as ow develops, EEM starts overestimating the convective heat transfer coef cient. Results in Tables 2 e 4 also suggest that calibration constant for SPD2 is independent of volume fraction and SPD2 and SPD1 perform best near the calibration point as expected. Accuracy of models at different Reynolds numbers is also investigated in terms of Nusselt number for volume fraction of 1.6%. In Fig. 8 Nusselt number predictions of models at different Reynolds numbers ( Re ¼ 1050, 1320, 1600, 1810) are presented and the re- sults are compared with experimental data from Ref. [12] . Both dispersion models were calibrated based on the experimental data by Ref. [12] at axial location of x / D ¼ 178 for Reynolds number of 1050 and volume fraction of 1.6%. Comparison of models in Fig. 8 reveals that the most accurate model in the Reynolds numbers range considered is the SPD2. Despite considering each phase individually, both two-phase models underestimate the change in Nusselt number with chang- ing Reynolds numbers. Although SPD2 is calibrated for Reynolds number of 1050, the model can predict the trend in experimental data accurately for other Reynolds numbers as well. Since desired

accurately for other Reynolds numbers as well. Since desired Fig. 8. Comparison of accuracy of models

Fig. 8. Comparison of accuracy of models in predicting Nu nf, x of 1.6% Al 2 O 3 e water nano uid at x/D ¼ 116 for different Reynolds numbers ( Re ¼ 1050, 1320, 1600, 1810).

Reynolds numbers ( Re ¼ 1050, 1320, 1600, 1810). Fig. 10. Comparison of estimated Darcy friction

Fig. 10. Comparison of estimated Darcy friction factors of Eulerian eEulerian two- phase model at Re ¼ 1050 for 1.6%, 1%, and 0.6% Al 2 O 3 e water nano uid.

S. Göktepe et al. / International Journal of Thermal Sciences 80 (2014) 83e 92

91

Table 5 Computational time [ s ] comparison of FMC and PC-SIMPLE.

f p [%]

Eulerian e Eulerian model

Single-phase

Eulerian e Mixture

FMC

PC-SIMPLE

SIMPLE

SIMPLE

0.6

82.2

157.1

77.6

552.8

1

89.8

158.6

78.1

572.7

1.6

111.7

163.8

81.7

566.5

accuracy can be achieved with a single calibration for different volume fractions and Reynolds numbers, SPD2, is suggested for applications where calibration data is available. On the other hand, for nano uids with no prior experimental studies, one of the two- phase models is suggested. Another objective of this study is to assess the effectiveness of single and two-phase models in the estimation of hydrodynamic characteristics of nano uids. Estimated Darcy friction factors are presented in Figs. 9 and 10 for SPD2 and EEM, respectively for three volume fractions (0.6%, 1%, 1.6%). EMM model in not considered here since, the model showed no change in the friction factor compared to that of homogenous single-phase model. The single- phase model estimated less than 1% change in friction factor with respect to base uid for all three volume fractions. However, this contradicts the data reported by experimental studies such as Ref. [38] . Hwang et al. [38] reported 4.2% increase in friction factor for 0.3% Al 2 O 3 e water nano uid at a Reynolds number of 400. The EEM model estimated 2% increase in friction factor for the same nano- uid at a Reynolds number of 400. Based on estimated results, single-phase models estimate no change in friction factor. The major drawback of two-phase models is their computa- tional expense. The required CPU time was measured as 163.8 s for EEM using PC-SIMPLE algorithm, 77.6 s for single-phase models, and 566.5 s for EMM for 1.6% Al 2 O 3 e water nano uid ow at a Reynolds number of 1050 as shown in Table 5 . The numerical studies presented in this study were performed on a workstation operating at a Quad Core 2.4 GHz CPU and all four cores were used in calculations. Although, PC-SIMPLE is a very robust method, it is computationally expensive. A remedy might be the use of Full Multiphase Coupled (FMC) scheme for dilute nano uid systems. Fig. 11 shows that the predicted h x distribution is identical for FMC and PC-SIMPLE. As shown in Table 5 , computational cost of EEM model can be reduced by approximately 65%.

cost of EEM model can be reduced by approximately 65%. Fig. 11. Comparison of two coupling

Fig. 11. Comparison of two coupling algorithms for 1.6% Al 2 O 3 ewater nano uid at Re ¼ 1050.

5. Conclusion

Single and two-phase models have been investigated for the characterization of laminar forced convection of Al 2 O 3 e water nano uid with various concentrations in a circular tube. Single- phase thermal dispersion model suggested by Mokmeli and Saffar-Avval [21] is found to be the most accurate single-phase model. It is recommended for applications, when calibration data is available, thermal analysis is the objective, and computational ef ciency is important. Furthermore, it is shown that for Reynolds numbers and particle volume fractions considered in this study, calibration constant used in the de nition of dispersion conduc- tivity is independent of Reynolds number and volume fraction of particles. Therefore, the model can be used at varying Reynolds numbers and particle volume concentrations without any re- calibration. It is also observed that Eulerian e Eulerian and Euler- ian e Mixture models under predict heat transfer coef cient at the beginning of the entry region then, as the ow develops both models start to over predict the heat transfer coef cient. Consid- ering its computational ef ciency, Eulerian e Eulerian two-phase model is recommended for applications, when no prior experi- mental data is available and prediction of both heat transfer and pressure drop is important. For the Eulerian e Eulerian two phase model, computational cost can be reduced by the use of Full Multiphase Coupling algorithm without sacri cing solution accu- racy. The transport between phases in Eulerian e Eulerian two phase model is estimated based on correlations derived for macro parti- cles. Developing relations more suitable for nanoparticles are required to further improve the prediction accuracy of Eulerian e Eulerian two phase model.

Acknowledgments

The authors would like to thank The Scienti c and Technological Research Council of Turkey (TUBITAK) for support under the grant 111M1777 of the 1001 Program.

Nomenclature

C 1 , C 2 C d c p

calibration constants for SPD1 and SPD2, respectively drag coef cient heat capacity

d

particle diameter

D

diameter of tube

!

F

force vector

f

Darcy friction coef cient

h

convective heat transfer coef cient

G

particle e particle interaction modulus

!

gravitational acceleration vector

g

i

enthalpy

k

thermal conductivity

k B

Boltzmann constant

L

length

n

shape factor

Nu x

local Nusselt number

q 00

heat ux

P

pressure

Pe

Peclet number

Pr

Prandtl number

R

radius of pipe

Re

Reynolds number

T

temperature

U

axial mean velocity

u , v

x and r velocity components, respectively