Vous êtes sur la page 1sur 18

1451

An analytical study on heat transfer performance of


radiators with non-uniform airflow distribution
E Y Ng1,2*, P W Johnson1, and S Watkins1
1Vehicle Aerodynamics Group, School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University,
Melbourne, Australia
2Product Design, Australian Automotive Air, Australia

The manuscript was received on 19 July 2004 and was accepted after revision for publication on 24 August 2005.

DOI: 10.1243/095440705X35116

Abstract: Heat exchangers used in modern automobiles usually have a highly non-uniform
air velocity distribution because of the complexity of the engine compartment and underhood
flow fields; hence ineffective use of the core area has been noted. To adequately predict the
heat transfer performance in typical car radiators, a generalized analytical model accounting
for airflow maldistribution was developed using a finite element approach and applying
appropriate heat transfer equations including the e–NTU (effectiveness – number of heat
transfer units) method with the Davenport correlation for the air-side heat transfer coefficient.
The analytical results were verified against a set of experimental data from nine radiators tested
in a wind tunnel and were found to be within +24 and −10 per cent of the experimental
results. By applying the analytical model, several severe non-uniform velocity distributions
were also studied. It was found that the loss of radiator performance caused by airflow
maldistribution, compared with uniform airflow of the same total flowrate, was relatively minor
except under extreme circumstances where the non-uniformity factor was larger than 0.5.
The relatively simple set of equations presented in this paper can be used independently in
spreadsheets or in conjunction with computational fluid dynamics (CFD) analysis, enabling a
full numerical prediction of aerodynamic as well as thermodynamic performance of radiators
to be conducted prior to a prototype being built.

Keywords: radiators, cross-flow heat exchangers, airflow maldistribution, corrugated


louvred fin surfaces, non-uniformity, specific dissipation

1 INTRODUCTION AND OBJECTIVES vehicle geometry strongly influences the airflow path
entering the radiator (see reference [4]).
The cooling airflow is seldom uniformly distributed There is only limited research published using
across an automotive radiator front face, and con- analytical approaches in determining the heat transfer
sequently the radiator cooling performance may be performance of automotive radiators. Analyses of
impaired [1–3]. The causes of non-uniformity of the radiator performance have often been based on
cooling airflow include objects upstream and down- the inadequate assumption that the airflow has a
stream of the radiator, particularly the upstream uniform distribution over the radiator face. An
bumper and associated crash structure. Figure 1 important deficiency in that assumption is that the
shows a normalized velocity contour measured at the air velocity distribution over the face of a real radiator
radiator front face in a typical passenger car, in in the engine compartment is always highly complex
ram-air condition (i.e. no fan operation), at a vehicle and non-uniform, and the airflow rate cannot easily
driving speed of 100 km/h. It is evident that the be analytically determined.
Heat exchanger performance can generally be
* Corresponding author: 7 Ellerton Court, Donvale, Victoria, VIC determined using either one of the following
3111, Australia. email: eton_ng@hotmail.com approaches [5]: the log mean temperature difference

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering
1452 E Y Ng, P W Johnson, and S Watkins

putation procedure) can lead to further understand-


ing of heat transfer characteristics of radiators and a
more accurate prediction of their effectiveness.

2 BACKGROUND TO THE ANALYSIS

Radiators used for vehicle engine cooling are


usually cross-flow type heat exchangers. The airflow
is generally induced by the moving vehicle (ram air)
and/or the cooling fan(s) (fan air).
Radiators with extended surfaces consisting of
multilouvred fins are commonly employed in
Fig. 1 Normalized airflow velocity contour at the modern vehicles. The extended surfaces provide
radiator face at a vehicle speed of 100 km/h. large enhancement of the heat transfer rate, not
(Reprinted from Ng [4]) only by providing additional surface area but also
by reducing the thickness of the boundary layer by
inducing a series of flat-plate leading edges, interrupt-
(LMTD) method or the effectiveness–NTU (e-NTU) ing the growth of the boundary layer along the fin
method. Selection from these two approaches is often surface [9]. Hosoda et al. [10] compared a plate fin
based on what type of problem is to be solved. Use heat exchanger with one with parallel louvres and
of the LMTD method is more convenient in solving found that the air-side heat transfer performance of
sizing or design problems (e.g. see reference [6]), the louvred fin structure was 60 per cent higher than
whereas for the rating problem (performance pre- that of the plain plate fin equivalent.
diction) the e-NTU method is more effective (e.g. see The direction of the coolant tubes can be either
references [7] and [8]). horizontal (coolant flowing horizontally) or vertical
The work presented in this paper was aimed to (coolant flowing downward from the top tank to
fulfil the following objectives: the bottom tank). According to SAE HS-40 [11], the
(a) to provide a comprehensive literature review horizontal flow type radiator is more commonly used
related to air- and coolant-side heat transfer in passenger vehicles. Coolant tubes are usually based
coefficients applicable to modern automotive on a flat (non-circular) tube design. The advantages
radiators; over circular tube designs include a higher heat
(b) to make use of existing heat exchanger theory to transfer area per unit of flow area; the wakes of the
solve radiator cooling problems where the air- tubes cause less reduction of heat transfer in down-
flow distribution is far from uniform; stream regions; the small projected areas of the tubes
(c) to illustrate a finite element method, where the minimize profile drag; and provision of a higher fin
radiator is mathematically treated as a finite efficiency.
number of small elemental radiators, in order to For the purposes of increasing the boiling point
provide a more accurate prediction of the heat of the coolant and preventing corrosion in the cool-
transfer capability of the entire radiator with ing systems, coolants are generally a mixture of
non-uniform airflow distribution; water, antifreeze (usually ethylene glycol, or EG), and
(d) to validate this method against a set of experi- possibly various corrosion inhibitors. It is noted that
mental data; and the use of glycol mixture generally reduces the heat
(e) to further the understanding of the influence transfer performance compared with pure water.
of airflow maldistribution on radiator cooling The analytical model was developed in this
performance. study specifically for the most common radiator
configuration, which consists of the following features:
The method presented can be carried out with-
(a) air-cooled radiator,
out the need for extensive computing resources
(b) cross-flow configuration,
and requires only some relatively simple iteration in
(c) corrugated louvred fins, and
a spreadsheet. Together with computational fluid
(d) flat coolant tubes with a horizontal flow structure.
dynamics (CFD) simulation predicting flow pheno-
mena through radiators, use of the techniques Nevertheless, the model described here, after minor
described in this paper (including sets of governing amendments, can be used to solve the radiator
equations for heat transfer coefficients and a com- performance of other specified configurations.

Proc. IMechE Vol. 219 Part D: J. Automobile Engineering D13604 © IMechE 2005
Heat transfer performance of radiators 1453

3 METHODOLOGY OF RADIATOR ANALYSIS (d) coolant flowrate, and


(e) basic radiator dimensions (refer to section 3.1).
The primary use for this model was to investigate
how airflow maldistribution affected the radiator 3.1 Definitions of the radiator geometry
performance under known conditions. The outlet
temperatures and the air temperature distribution at With reference to Fig. 2, the basic radiator dimensions
the radiator exit were the aspects to be evaluated. that were required in this model are listed below:
For this reason, it was deemed to be more appro-
Core height (B )
priate to utilize the e-NTU method. In this analysis, H
Core width (B )
several assumptions were made, which are stated in W
Core thickness (B )
Appendix 2. This model is based on the following T
Number of rows of tubes in the core depth
conditions being known:
dimension (N )
r
(a) ambient temperature, Number of coolant tubes in one row (N )
ct
(b) airflow velocity distribution across the radiator, Number of profiles (N )
p
(c) coolant inlet temperature (top tank temperature), Number of fins per metre (N )
f

Fig. 2 Definitions of the radiator dimensions

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering
1454 E Y Ng, P W Johnson, and S Watkins

Louvre pitch (L ) air-side convection, fouling on the air side, wall con-
p
Louvre length (L ) duction, fouling on the coolant side, and coolant-side
l
Fin thickness (F ) convection. In mathematical form
t
Fin height (F )
h
Fin pitch (F ) 1 1 Dx 1
p = +R + +R + (10)
Fin end radius (R ) UA (g hA) f,a (kA) f,c (g hA)
f 0 a w 0 c
Angle of fin (a )
f
Coolant tube length (Y ) It is noted that without the use of extended surfaces
l
Coolant tube cross-section length (Y ) (i.e. prime surface) in the coolant tubes in typical
cl
Coolant tube cross-section width (Y ) radiators, the value of g becomes unity.
cw 0,c
Coolant tube thickness (Y ) For simplicity, the fouling resistances on both
t
Coolant tube pitch (Y ) sides were assumed to be small and not considered
p
Coolant tube end radius (R ) in this analysis. Also, coolant tubes are often made
t
of aluminium or copper, which have large thermal
3.2 Calculation of relevant heat transfer areas conductivity, and the wall thickness is small. There-
Based on the preceding definitions, derivations of fore, the thermal resistance of the wall conduction
various surface areas, which are relevant to heat term (Dx/kA) in comparison to the other two terms
w
transfer, are given below: was treated as negligible. It is noted that this wall
term does not include fin conduction, which is
Fin length, F treated separately in the next section.
l
F −2R With these stated restrictions and combining
F =pR + h f (1)
l f cos a the nomenclature used in this study, equation (10)
f becomes
Radiator core frontal area, A
fr,r
A =B B (2) 1 1 1
fr,r H W = + (11)
Coolant tube frontal area, A UA g h A hA
fr,t fr,r 0,a a a c c
A =Y Y N (3) Under typical operating conditions, the air-side
fr,t cw l ct
Fin frontal heat transfer area, A resistance is dominant (i.e. the controlling resistance).
fr,f Davenport [9] showed that at higher water flowrates,
A =F F N YN (4)
fr,f t l f(per metre) l p the coolant-side thermal resistance contributed only
Fin heat transfer area, A 5 per cent of the total resistance.
f
A =2B F N YN (5) Equation (11) is the equation that governs the heat
f T l f(per metre) l p transfer performance of a cross-flow type radiator.
Total heat transfer area on the air side, A Detailed derivations of the air-side and coolant-side
a
A =A +2N Y N [(Y −2R )+(2pR )] (6) heat transfer coefficients (i.e. h and h ) and their
a c
a f ct l r cl t t governing equations, which involve an extensive
Total heat transfer area on the coolant side, A
c review from the literature, are presented in section 6.
A =[2p(R −Y )+2(Y −2R )]Y N N (7)
c t t cl t l ct r
Total air pass area, A
p,a 4.2 Air-side fin efficiency and total surface
A =A −A −A (8) efficiency
p,a fr,r fr,f fr,t
Total coolant pass area, A Fins attached to the coolant tubes are primarily used
p,c
A =[p(R −Y )2+(Y −2Y )(Y −2R )]N N for increasing the surface area and consequently
p,c t t cw t cl t ct r increasing the heat transfer rate. As temperature
(9) gradients along the fins extending into the air create
conduction resistance, the temperature efficiency of
the surface is reduced. Hence, the air-side heat con-
4 HEAT TRANSFER ANALYSIS
ductance term (hA) must be multiplied by a factor
a
to account for the temperature gradient in the fin.
4.1 Thermal circuit and overall heat transfer
The fin efficiency (g ) is defined as the ratio of the
coefficient f
actual heat transfer rate through the fin base (Q )
f
The overall heat transfer resistance for a radiator divided by the maximum theoretical heat transfer
can be considered to derive from the following terms: rate through the fin base (Q ), corresponding to
f,max

Proc. IMechE Vol. 219 Part D: J. Automobile Engineering D13604 © IMechE 2005
Heat transfer performance of radiators 1455

the entire fin surface being at the base temperature radiator on air and coolant sides are typically in
the range of a 20–35 °C rise and a 4–10 °C drop
Q
g= f (12) respectively. Hence
f Q
f,max
C =C and C =C
For a straight fin of uniform cross-section, the fin min a max c
efficiency can be expressed as For a typical automotive radiator it is assumed
tanh(mL) that the approaching air stream is directed into a
g= (13) large number of separate passages with no cross
f mL
mixing when the air is travelling through the radiator.
where L=effective fin length and m=fin efficiency The same assumption is commonly applied to
parameter. For the fins extending from wall to wall coolant flow. Hence, radiators are often considered
F as a type of cross-flow arrangement heat exchanger
L= h with both fluids unmixed [13, 14]. The appropriate
2
e-NTU relationship for this type can be found in
For thin sheet fins (see reference [5]) graphical form in reference [5], and mathematically
in reference [15].
m=
S 2h
kF
a
t
Radiators are typically of a finned-tube con- 5 FINITE ELEMENT APPROACH
struction. Hence, heat transfer takes place in both the
fins and the unfinned (primary) surface at the same To account for the effects of airflow maldistribution,
time. For this reason, the total surface efficiency (g ) a finite element approach was used, which divides
0
is introduced to account for the weighted mean the entire radiator into a number of independent
efficiency of the composite structure consisting of small radiators (cells). The e-NTU method was applied
the fins and the base structure. The total surface to each cell, with the outlet coolant temperature
efficiency can be calculated from of the upstream cell being the inlet temperature of
the downstream cell, in order to determine its heat
A
g =1− f (1−g ) (14) dissipation rate (see Fig. 3). As a coolant horizontal-
0 A f flow radiator was considered, the coolant flow into
a
This equation is based on the assumption that the each cell was equal to the total coolant flow divided
air-side heat transfer coefficient is unchanged by by the number of rows. The heat dissipation of
addition of the fins. the entire radiator was equal to the sum of the heat
dissipation of all cells.
4.3 The e-NTU method This approach may be similar to techniques
used in some commercial software such as KULI,
The radiator effectiveness can be expressed as a
FLOWMASTER and GT-COOL, and also techniques
function of two dimensionless groups, NTU and C ,
r developed in the European project V-THERM, except
for a given flow arrangement (such as counter-flow
that this procedure can be carried out where no
or cross-flow) [5]
extensive computing resources or software packages
e= f (NTU, C , flow arrangement) (15) are available. This enables engineers to perform
r
analysis or performance predictions in house by use
where
of simple manipulations.

P
UA 1 The number of cells, including the number of rows
NTU= fr,r = U dA
C C fr,r (m) and the number of columns (n), is given by the
min min A
user. It is assumed that the coolant flow is uniformly
C=heat capacity rate=ṁ×c distributed through the radiator and equally divided
p
C between the different rows. Given the top tank
C = min temperature (T ) and the coolant mass flow, and
r C ci
max assuming the tanks on both sides of the radiator are
C =ṁ ×c =A r V ×c well insulated, it can be demonstrated that the inlet
a a p,a a a a p,a
C =ṁ ×c =A r V ×c coolant temperature of each cell in the first column
c c p,c c c c p,c is equal to the top tank temperature; i.e.
Under normal driving conditions, SAE J1393 [12]
states that the temperature changes across the T =T (16)
ci ci(i,1)

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering
1456 E Y Ng, P W Johnson, and S Watkins

Fig. 3 Finite element approach

where T refers to the inlet temperature of the cell Because of the complexity of airflow over louvred
ci(i,1)
(i, 1), with i=1, … , m. fins, it is difficult to determine the heat transfer
Except for the cells in the first column, the coolant coefficient from conventional convection heat trans-
inlet temperature of any cell is equal to the tem- fer equations, as the coefficient is a function of both
perature at the exit of the upstream cell in the same fin geometry and flow conditions. Because the heat
row transfer characteristics of the louvres are closely
related to the flow structure around them, the follow-
T =T (17)
co(i,j) ci(i,j+1) ing subsection provides a brief discussion of the flow
with i=1, … , m and j=1, … , (n−1). The value of phenomena occurring in the louvre array.
the bottom tank temperature (T ) is taken as the
co
average value of the summation of the last cells for 6.1.1 Flow structure in louvre fins
all rows
Following Beauvais [17] and Wong and Smith [18],
m who discovered that louvres act to realign the air-
∑ T
co(i,n) flow in a direction parallel to their own planes,
T = i=1 (18) Davenport [9] performed a detailed investigation on
co number of rows
corrugated louvred fin heat exchangers, demon-
Similarly, the coolant mass flowrate of each cell strating that the flow structure within the louvred
becomes array was a function of Reynolds number.
ṁ Figure 4 illustrates a section through a louvre array
ṁ = c (19) in which two possible extreme flow directions are
c(i,j) number of rows
indicated. Davenport found that at low Reynolds
for all (i, j). numbers the flow did not pass through the louvres
but travelled axially through the fins and behaved like
duct flow (duct-directed flow). He explained that the
6 EQUATIONS FOR HEAT TRANSFER developing boundary layers on the louvres became
COEFFICIENTS

6.1 Air side


Airflow over a louvred fin array is complex. Rather
than acting as surface roughness or turbulent
generators, the louvres are used to deflect airflow
from their incident direction and consequently the
flow becomes aligned with the planes of the louvres.
They enhance the heat transfer by providing multiple Fig. 4 Section through louvred fin indicating possible
leading edges, associated with high heat transfer flow directions. (After Achaichia and Cowell
coefficients [16]. [19])

Proc. IMechE Vol. 219 Part D: J. Automobile Engineering D13604 © IMechE 2005
Heat transfer performance of radiators 1457

sufficiently thick to block off the gaps between louvres. corrugated louvred-fin surfaces found in the publicly
This gradually changed to an almost complete align- available literature. The correlations are empirical
ment with the louvres as the Reynolds number was using a multiple regression technique. The corre-
increased. At high Reynolds numbers, the flow was lation for heat transfer for corrugated louvred fin
directed by the louvres flowing nearly parallel to surfaces is valid when 100<Re<4000 and the
them, behaving like flat-plate flow (louvre-directed recommended Colburn modulus j factor correlation
flow). Additional studies conducted by Achaichia and was given as
Cowell [19, 20] and Kajino and Hiramatsu [16] agreed

A B
with Davenport’s investigation. L 1.1
j=0.249Re−0.42 L0.33 l F0.26 (20)
Lp h F h
h
6.1.2 Air-side Reynolds numbers
This correlation is valid for the ratio of the louvre
The flow path over louvres is dependent on Reynolds length to fin height ranging between 0.62 and 0.93,
numbers for a given louvred fin array. However, it which is applicable to modern automotive radiators.
seems that the characteristic length is rather arbitrary The Davenport correlation was reasonably accurate
for louvred fin surfaces. Davenport [21], after testing and easy to apply, and approximately 95 per cent of
32 samples of multilouvred fin surfaces, suggested the data were correlated within ±6 per cent. The j
that the fin pitch and hydraulic diameter made curve (i.e. plot of the j factors versus Reynolds
no contribution to the correlation of Colburn’s numbers) had a mean gradient of −0.42 from the
modulus j factor. He concluded that, although the regression analysis compared with −0.5 for the classic
hydraulic diameter is relevant to heat transfer in theoretical Pohlhausen equation. Also, the values
plain fins, using the louvre-pitch-based Reynolds of the j curve were about 35 per cent below the
number (Re ) is more appropriate to describe the Pohlhausen line. Davenport explained that this was
Lp
heat transfer on louvred fin surfaces. Most of the later due to the entire heat transfer surface, not all of
research has been consistent with this finding and which was louvred. This study revealed that the heat
has used Re as a basis – which is the approach transfer behaviour over louvred fins has a general
Lp
used here. similarity with Pohlhansen solutions for flow over a
flat plate, suggesting the existence of laminar bound-
6.1.3 Air-side heat transfer coefficient ary layers on the louvres, which is consistent with
Publication of the heat transfer performance of the the discussions in the previous sections.
corrugated louvred fin geometry, which is the basic Since the Davenport study, questions have arisen
structure of modern radiators, has been very limited, as to whether his correlation is still valid, as the
perhaps for commercial reasons. Apart from the core structures he tested were noticeably different
corrugated geometry, there are some other types from the ones used nowadays in terms of material
of heat exchanger geometry available including: a and fin geometry. A recent study conducted by Webb
flat tube and louvred plate fin [19], a corrugated et al. [24] revealed that the correlations are still appli-
louvred fin with a rectangular channel [22], a cable to current automotive radiator cores. Further-
corrugated louvred fin with a splitter plate and more, a large number of recent researchers have still
rectangular channel [23], and a corrugated louvred employed Davenport’s experimental data as a base
fin with a splitter plate and triangular channel [22]. source to validate their findings, including references
However, these geometries are not normally used in [24] to [27].
automotive applications; thus discussion on them is Another study conducted by Aoki et al. [28],
outside the scope of this paper. who measured the local heat transfer for individual
One of the important studies on extended louvres, also suggested laminar heat transfer being
heat transfer surface was conducted by Kays and present on the louvres. A correlation between the
London [5]. However, those louvre designs are very Nussult number and the Reynolds number based on
different from the multilouvred surfaces, which are the louvre pitch was presented in one equation for
widely used nowadays for automotive radiator cores, different louvre pitches
and the data are of little relevance when applied Nu =0.87Re1/2 Pr1/3 (21)
to modern heat exchangers. Davenport [21], after Lp Lp
testing with louvred fin cores with systematically Based on the work of Aoki et al., Webb [29]
varied louvre geometry, presented correlations of recommended that the theoretical Pohlhausen
heat transfer and flow friction characteristics. This solution for laminar flow over a flat plate with con-
is the only set of experimental data published for stant heat flux (see reference [30]) can be used to

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering
1458 E Y Ng, P W Johnson, and S Watkins

predict the heat transfer coefficient on louvres, i.e. is relatively insensitive to water-side heat transfer.
The coolant-side heat transfer coefficient can be
Nu =0.906Re1/2 Pr1/3 (22)
av Lp evaluated by applying appropriate well-established
Based on the above findings, Sahnoun and heat transfer equations for flow inside tubes. The
Webb [25] and Dillen and Webb [26] developed coolant tubes in typical radiators are a flat-oval
an analytical model and a semi-analytical model shape. Therefore, the hydraulic diameter (D ) is used
respectively, to predict the heat transfer coefficient h
to substitute for the characteristic physical diameter.
of the corrugated louvred fin geometry. Both of Invalid use of the hydraulic diameter has been found
the models were based on dividing the louvred only when calculating tubes with sections having
fin surface into four regions: the louvred areas, the very sharp corners (e.g. triangular), which gives
plain leading and trailing areas, the plain middle unacceptably large errors.
area, and the end region areas. In their models, the There has been a large amount of fundamental
heat transfer coefficient in the unlouvred regions work in understanding flow characteristics and heat
is predicted using a fully developed laminar flow transfer in tubes; therefore only several represent-
solution, while using the Pohlhansen solution for ative equations are considered here. Starting with
laminar flow over a flat plate [equation (22)] to pre- the Reynolds experiments in 1883, it has been
dict louvred areas. A summation of the heat transfer demonstrated that laminar flow becomes unstable as
coefficients calculated from different regions was the velocity of flow increases in a given tube. The
used to determine the heat transfer coefficient of the transition from laminar to turbulent flow occurs at a
entire louvred fin. value of Reynolds number near 2300. The transition
Instead of dividing the louvred fin surface into to turbulent flow generally takes place in the range
four regions, an earlier work of Beard and Smith [6] of Reynolds numbers from about 2300 to 10 000, and
developed a simple method that approximated the a fully turbulent flow mostly occurs at a Reynolds
effects of louvres by calculating the heat transfer number above 10 000. It has also been found that the
coefficient from a mean of two coefficients obtained transition of the flow is greatly affected by the tube
from louvred fin regions and unlouvred fin regions. inlet configuration and surface roughness.
Since the proposed benefit of using louvred fins was
to give a series of leading edges to the airflow creating 6.2.1 Heat transfer in laminar flow
laminar boundaries, the heat transfer characteristics
In the laminar flow regime, the heat transfer flux is
of each louvre could be similar to flow over a flat
strongly dependent on the thermal boundary con-
plate. For that reason, their model was developed
dition along the whole length of the tube, while less
using the theoretical Pohlhausen equation for flow
dependent in the turbulent flow regime for fluids
over a flat plate to predict the heat transfer coefficient
with Pr1. The thermal boundary condition refers to
on louvres.
the set of specifications describing temperature con-
This simple method was shown to be valid as the
ditions and/or the heat transfer rate at the inside wall
calculated results were found to give satisfactorily
of the tube. According to Shah and London [31], the
close agreement with a series of wind-tunnel test
thermal boundary condition of automotive radiators
results with a maximum error of approximately 10
can be classified as the constant wall temperature
per cent.
peripherally as well as axially, since one fluid has a
6.1.4 Summary of preferred air-side heat transfer very much higher capacity rate than the other.
equations Knudsen and Katz [32] reported an equation
proposed by Hansen in 1943 as representing the
Essentially, there have been only four models avail- Graetz solution for constant surface temperature,
able in the existing literature to predict the heat fully developed laminar flow, and parabolic velocity
transfer coefficient for radiators (corrugated louvred distribution. The well-known Hansen equation, which
fin surfaces). Among them, it appears that the has been widely used for the mean Nusselt number
Davenport correlation [equation (20)] is relatively easy over the entire length of the tube, was used in this
to use without compromising accuracy. Therefore, this study to predict the coolant-side heat transfer
correlation was chosen in this study to determine the coefficient in the laminar flow regime
air-side heat transfer coefficient (h ).

A B
a D
0.0668 h,c Re Pr
6.2 Coolant side Y Dh,c c
Nu =3.66+ l (23)

CA B D
It is emphasized that the dominant thermal resist- c D 2/3
1+0.04 h,c Re Pr
ance is air-side convection and that the modelling Y D h,c c
l

Proc. IMechE Vol. 219 Part D: J. Automobile Engineering D13604 © IMechE 2005
Heat transfer performance of radiators 1459

All properties appearing in the equation are required 104<Re <5×106 and 0.5<Pr<2000. Thus this
D
to be evaluated at the average value of the mean equation h,cwas chosen for this study in the case of
temperature. The Hansen equation is valid for the turbulent flow.
constant surface temperature condition over the Gnielinski [37] modified the Petukhov equation in
entire tube length. Considering a sufficiently long order to derive a new equation that is applicable in
tube, it is noted that the Nusselt number approaches the transition flow region (2300<Re<104). It was
a value of 3.66, which is the analytical solution for noted that in the transition region the Gnielinski
laminar, fully developed conditions with a constant equation satisfactorily reproduced the decrease in the
surface temperature [30], i.e. Nu =3.66. heat transfer coefficient with decreasing Reynolds
c
number. Thus, the following equation was employed
6.2.2 Heat transfer in turbulent and transition flow in this study
Due to the complicated nature of turbulent flow, ( f /8)(Re −1000)Pr
Nu = D
c 1+12.7 √ fh,c/8(Pr2/3−1)
which is transient and possesses highly irregular (27)
fluctuations, and the fact that heat is generally
transferred by convection as well as conduction, This equation is valid for 0.5<Pr<2000 and
empirical correlations of turbulent heat transfer data 2300<Re <104, and was compared with approxi-
D
for flow in tubes are often preferred for simplicity. mately 800h,c experimental data points, with nearly
The correlations have been obtained based on 90 per cent of the data falling within ±20 per cent.
several dimensionless groups, including the Reynolds
number, Nusselt number, Prandtl number, and
Stanton number. 6.2.3 Summary of preferred coolant-side heat
There were three famous correlations proposed transfer equations
in the 1930s, which were the Dittus and Boelter Flow regime Equation used
equation [33], the Colburn equation [34], and the Laminar (2300>Re ) Hansen equation
Sieder and Tate equation [35]. Incropera and DeWitt Dh
Transition Gnielinski equation
[36] reported that the use of the above equations, (10 000Re 2300)
D
although they may be easily applied, may lead to Turbulent (Re h>10 000) Petukhov equation
errors as large as 25 per cent. For this reason, these Dh
correlations were not chosen for this study. To examine the consistency of these three heat
Prandtl in 1944 was the first to present an equation transfer equations selected to be used in the analytical
for heat transfer in tubes that was related to the model, the heat transfer coefficient was calculated in
pressure drop. The equation was in the form of different flow regimes ranging from Reynolds number
of 400 to 13 500, as shown in Fig. 5. Discontinuities
Nu f /8
= (24) were found at the transitions between the regimes
Re Pr 1+8.7 √ f /8(Pr−1)
(i.e. at Reynolds numbers of 2300 and 10 000); never-
where f is the friction factor in the tube. Knowledge theless, these discontinuities of the coolant-side
of the friction factor in the tube is required before heat transfer coefficient are relatively unimportant,
applying this equation, and value of the friction since the dominant thermal resistance is air-side
factor can be directly obtained from the Moody convection.
diagram. Alternatively, the friction factor can be
calculated from the following equation for isothermal
flows in smooth tubes [37] 7 EXPERIMENTAL VALIDATION OF THE
f =[0.79 ln(Re )−1.64]−2 (25) ANALYTICAL MODEL
Dh,c
Since then, the Prandtl equation has been further A set of experimental data obtained using the RMIT
improved. From the basic form, a correlation recom- University cooling test facility was used to validate
mended by Petukhov [38] offered better accuracy for the analytical model.
turbulent tube flow. The equation is expressed as
( f /8)Re Pr 7.1 Test procedure
Nu = Dh,c (26)
c 1.07+12.7 √ f /8(Pr2/3−1)
The detailed test procedure and equipment can be
This equation predicted experimental data with found in reference [39], and only a brief description
an accuracy of 5–6 per cent over the range of is provided here. Nine sections of corrugated louvred

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering
1460 E Y Ng, P W Johnson, and S Watkins

Fig. 5 Coolant heat transfer coefficients in different flow regimes

fin radiator cores with various fin pitch, tube spacing, To measure the heat dissipation rate of each
and fin width (listed in Table 1) were tested indi- radiator, hot water was supplied by an external
vidually in a small open-circuit, closed-test-section heat bench and thermocouples were set at various
wind tunnel with a cross-section of 0.3 m (width)× locations, including the radiator inlet and outlet,
0.35 m (high). Air flowed perpendicularly to, and upstream and downstream of the radiator core. In
uniformly through, a section of radiator that was addition, the water flowrate was monitored and
located in the middle of the 0.8 m long test section. recorded via a magnetic flowmeter. The supplied
Airflow velocity of 9.35 m/s was set for each of the water was set at 1 l/s and approximately 75 °C. Data
radiators. were taken at the equilibrium state.

Table 1 Radiator dimensions, experimental data, and analytical predictions


Radiator sample numbers

1 2 3 4 5 6 7 8 9 Unit

B 352 352 352 352 352 352 352 352 352 mm


H
B 302 302 302 302 302 302 302 302 302 mm
W
B 38 32 32 16 40 66 38 38 32 mm
T
N 2 2 2 1 3 4 2 2 2
r
N 41 35 29 35 33 29 41 41 34
ct
N 42 36 30 36 34 30 42 42 35
p
N 16 16 16 16 16 14 14 18 21 in
f
L 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 mm
p
L 5.1 6.7 8.7 6.7 7.2 8.7 5.1 5.1 6.7 mm
l
F 0.058 0.058 0.058 0.058 0.058 0.058 0.058 0.058 0.058 mm
t
F 6.146 7.746 9.746 7.746 8.246 9.746 6.146 6.146 7.746 mm
h
F 1.588 1.588 1.588 1.588 1.588 1.814 1.814 1.411 1.210 mm
p
R 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 mm
f
a 25 25 25 25 25 25 25 25 25 degree
f
Y 302 302 302 302 302 302 302 302 302 mm
l
Y 12.2 12.2 12.2 12.2 12.2 12.2 12.2 12.2 12.2 mm
cl
Y 2 2 2 2 2 2 2 2 2 mm
cw
Y 0.127 0.127 0.127 0.127 0.127 0.127 0.127 0.127 0.127 mm
t
Y 8.4 10 12 10 10.5 12 8.4 8.4 10 mm
p
R 1 1 1 1 1 1 1 1 1 mm
t
Experimental 556 434 408 282 506 555 510 579 493 W/K
Prediction 509 502 506 347 555 625 489 526 541 W/K
Error −8.6 15.7 24.0 23.0 9.6 12.7 −4.1 −9.1 9.7 %

Proc. IMechE Vol. 219 Part D: J. Automobile Engineering D13604 © IMechE 2005
Heat transfer performance of radiators 1461

7.2 Test parameter transfer process in radiators is dominant and the


resultant error on the coolant side becomes minor.
The specific dissipation (SD) technique was adopted
In another study of radiator airflow maldistribution
in the measurements. This technique is commonly
in a real vehicle, this analytical model can predict
used within the automotive industry to evaluate the
the same trend as the experimental measurements
cooling performance of a vehicle in non-climatic
within 10 per cent for various configurations (see
wind tunnels. The major advantage of using this
reference [4]). As this model is developed for use
technique is that the SD parameter, which relates to
in predicting radiator performance trends before a
a radiator’s effectiveness, normalized heat transfer
prototype becomes available rather than predicting
with respect to the driving temperature difference,
absolute values, this achieved accuracy is deemed
was proved to be insensitive to variations of air and
to be acceptable. This model also accounts for the
coolant temperatures [14]. SD is defined as the heat
correct cooling airflow distribution in a vehicle,
transfer rate of the radiator (Q) divided by the temper-
giving more accurate predictions. It is emphasized
ature difference between the water and air entering
that analytical studies are always a supplement to,
the radiator (T −T ) (note that this is the maximum
ci ai not a substitute for, performance testing in product
temperature difference available to the radiator)
development.
ṁ c (T −T )
SD=e(ṁ c )= c p,c ci co (28)
a p,c T −T
ci ai
8 APPLICATIONS
where e is the radiator effectiveness
The model can be used as a general tool (with some
7.3 Experimental versus analytical results modifications to suit other configurations of heat
exchanger applications) in early radiator design,
Table 1 shows the experimental data along with those before a prototype is built. This can allow engineers
results calculated under the same test conditions to study the effect of parameter changes, such as fin
by the analytical model described previously. The pitch or tube sizes. In the early design and develop-
results indicate that the analytical model provides ment phase, an absolute value is not normally
predictions of the heat transfer performance of a required and the accuracy of this analytical model is
range of geometrically different radiators within +24 sufficient for the purpose of the initial conceptual
and −10 per cent. It is also noted that the model study. Furthermore, this modelling can be done with-
predicts the coolant temperature at the radiator out- out the need of extensive computing resources and
let well (within 1.2 °C), compared to the measured skills. In this section, two examples are presented to
values. illustrate the use of the model.
The predicted values are generally correlated well
with the measured values with an average deviation
of 8.1 per cent and a mean deviation of 12.9 per cent, 8.1 Effect of airflow maldistribution on heat
but in two cases errors are overpredicted by more than transfer performance
23 per cent compared to the test data. Since this To understand the effect of airflow maldistribution on
modelling is developed based primarily on analytical the heat transfer performance of a radiator, several
derivation, except for the heat transfer coefficients severe non-uniform velocity distributions were con-
for air and coolant flow, it is thought that such error sidered (see Figs 6 and 7). Four airflow profiles were
would arise from either or both empirically based artificially simulated, but in each case the average
correlations. Davenport’s air-side correlation should flow velocity through the radiator was unchanged
have given fairly accurate predictions according to and equal to 5 m/s. The coolant inlet temperature,
previous publications (95 per cent of experimental ambient temperature, and coolant flowrate were all
data correlated within 6 per cent). When Reynolds fixed at 60 °C, 25 °C and 1 l/s respectively.
numbers range from 2300 to 10 000, the coolant To quantify the non-uniformity of airflow across
flow condition in the tube is transitional, which is the radiator face, a parameter is needed. Given that
the case here. Applying the Gnielinski equation in the radiator is segmented into a finite number of
transition flow can lead to a rather large error in equal-area cells, a non-uniformly factor was defined
predicting the coolant-side heat transfer coefficient as
(see section 6.2.2). On the other hand, good pre-
dictions of coolant outlet temperatures can be 1 n |V −V |
i= ∑ local,k average (29)
explained because the air-side resistance in the heat n V
k=1 average

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering
1462 E Y Ng, P W Johnson, and S Watkins

Fig. 6 Reduction in SD due to airflow maldistribution

Fig. 7 Increase in coolant outlet temperature due to airflow maldistribution

where flow being the same. In a case of a non-uniformity


parameter of 0.4 (the typical value for passenger
n=number of cells
vehicles is generally less than 0.5), the SD value,
V =local velocity
local and hence the radiator cooling performance, would
V =average velocity through the entire
average reduce by 8 per cent. An increase of about half a
radiator
degree Celsius at the bottom tank would result as a
The finite element method was utilized for a number consequence. Except under extreme circumstances,
of non-uniformity factors and distributions, as shown it appears that the effect of airflow maldistribution on
in Figs 6 and 7. The figures clearly indicate that the SD is relatively minor (less than 10 per cent penalty
cooling capacity of a radiator is influenced by non- in SD for i=0.5). Nevertheless, it is suggested that
uniformity of airflow, despite the total amount of air- the factor of non-uniformity should be realistic in

Proc. IMechE Vol. 219 Part D: J. Automobile Engineering D13604 © IMechE 2005
Heat transfer performance of radiators 1463

any radiator performance analysis, in order to pro- has been used to evaluate a radiator’s effectiveness
vide accurate calculations of SD as well as radiator in non-climatic type wind tunnels [14], and reflect
performance. changes in cooling airflow [4].
This analytical study reveals that:
8.2 Study of the SD parameter
1. Coolant flowrates (Fig. 8). When testing at low
This analytical model has been also used in studying coolant flowrates, SD is fairly sensitive to changes
the effect of sensitivity of coolant flowrates, ambient in coolant flow and small fluctuations in coolant
temperature drift, and coolant inlet temperature drift flowrate may cause a considerable variation in SD.
on specific dissipation (SD) [see equation (28)]. This 2. Ambient temperature drift (Fig. 9). A drift of about
parameter has been proven to be insensitive to 4 per cent in SD would occur when the ambient
changes in ambient and coolant temperatures, and temperature changed from 10 to 40 °C.

Fig. 8 Effect of coolant flowrates on SD. Analytical predictions versus wind-tunnel test data.
(Reprinted from Ng [4])

Fig. 9 Calculated heat rejection rates and SD values at ambient temperatures between 10 and
45 °C in three uniform flow cases. (Reprinted from Ng [4])

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering
1464 E Y Ng, P W Johnson, and S Watkins

Fig. 10 Calculated heat rejection rates and SD values at coolant inlet temperatures between 30
and 120 °C in three uniform flow cases. (Reprinted from Ng [4])

3. Coolant inlet temperature drift (Fig. 10). A 30 °C ACKNOWLEDGEMENTS


change in coolant temperature (from 50 to 80 °C)
results only in a small variation in SD and the The authors would like to acknowledge the Depart-
consequent error is less than 1 per cent. ment of Mechanical and Manufacturing Engineering,
RMIT University, for financial and technical support.

9 CONCLUSIONS
REFERENCES
A generalized analytical model was developed based
on applying relevant heat exchanger theory to pre- 1 Olson, M. E. Aerodynamic effects of front end
dict the heat dissipation rates and to study the effects design on automobile engine cooling systems. SAE
of airflow maldistribution on the performance of a technical paper 760188, 1976.
radiator. The model employed the e-NTU method in 2 Chiou, J. P. The effect of the flow nonuniformity on
the sizing of the engine radiator, SAE technical paper
combination with the Davenport correlation for the
800035, 1980.
prediction of the air-side heat transfer coefficient. 3 Williams, J. An automotive front-end design
According to the flow regime, the coolant-side approach for improved aerodynamics and cooling.
heat transfer coefficient was calculated from the SAE technical paper 850281, 1985.
Hansen equation, Gnielinski equation, or Petukhov 4 Ng, E. Y. Vehicle engine cooling systems: assess-
equation. The model was validated against a set of ment and improvement of wind-tunnel based
experimental data. evaluation methods. PhD Thesis, School of Aero-
space, Mechanical and Manufacturing Engineering,
The loss of engine cooling performance caused
RMIT University, Melbourne, 2002.
by airflow maldistribution, compared with uniform 5 Kays, W. M. and London, A. L. Compact heat
airflow of the same total flowrate, was found to be exchangers, 3rd edition (Reprint edition with
relatively minor – in a typical automobile radiator corrections), 1998 (Krieger Publishing Company).
airflow cases are less than 10 per cent, or typically 6 Beard, R. A. and Smith, G. J. A method of calculating
half a degree increase in coolant temperature at the the heat dissipation from radiators to cool vehicle
radiator outlet. Nevertheless, improvement in flow engines. SAE technical paper 710208, 1971.
7 Emmenthal, K. D. and Hucho, W.-H. A rational
uniformity, in particular in areas influenced by the
approach to automotive radiator systems design.
bumper wake, is expected to increase the overall SAE technical paper 740088, 1974.
vehicle energy efficiency since the heat transfer area 8 Eichlseder, W. and Raab, G. Calculation and design
can be used more effectively, which can lead to a of cooling systems. SAE technical paper 931088,
reduction in aerodynamic drag. 1993.

Proc. IMechE Vol. 219 Part D: J. Automobile Engineering D13604 © IMechE 2005
Heat transfer performance of radiators 1465

9 Davenport, C. J. Heat transfer and fluid flow in 27 Chang, Y.-J. and Wang, C.-C. A generalized heat
louvred triangular ducts. PhD Thesis, Department transfer correlation for louvre fin geometry. Int.
of Mechanical Engineering, Coventry (Lanchester) J. Heat and Mass Transfer, 1997, 40(3), 533–544.
Polytechnic, 1980. 28 Aoki, H., Shinagawa, T., and Suga, K. An experi-
10 Hosoda, T., Uzuhashi, H., and Kobayashi, N. Louvre mental study of the local heat tranfer characteristics
fin type heat exchangers. Heat Transfer – Jap. Res., in automotive louvred fins. Expl Thermal and Fluid
1977, 6(2), 69–77. Sci., 1989, 2, 293–300.
11 SAE HS-40 Principles of engine cooling systems, 29 Webb, R. L. Principles of enhanced heat transfer,
components and maintenance, 1991 (Society of 1994 (Wiley Interscience, New York).
Automotive Engineers, Warrendale, Pennsylvania). 30 Kays, W. M. and Crawford, M. E. Convective heat
12 SAE J1393 On-Highway Truck Cooling Test Code, June and mass transfer, 2nd edition, 1980 (McGraw-Hill,
1984 (Society of Automotive Engineers, Warrendale, New York).
Pennsylvania). 31 Shah, R. K. and London, A. L. Laminar flow forced
13 Yan, W.-M. and Sheen, P.-J. Heat transfer and friction convection in ducts. Advances in Heat Transfer
characteristics of fin-and-tube heat exchangers. Int. (Suppl. 1), 1978.
J. Heat and Mass Transfer, 2000, 43, 1651–1659. 32 Knudsen, J. G. and Katz, D. L. Fluid Dynamics and
14 Lin, C. Specific dissipation as a technique for Heat Transfer, McGraw-Hill Series in Chemical
evaluating motor car radiator cooling performance. Engineering (Ed. S. D. Kirkpatrick), 1958 (McGraw-
PhD Thesis, Department of Mechanical and Manu- Hill, New York).
facturing Engineering, RMIT University, Melbourne, 33 Dittus, F. W. and Boelter, L. M. K. Heat transfer in
1999. automobile radiators of the tabular type. University
15 Holman, J. P. Heat transfer, 7th edition, 1992 of California, Berkeley Publications on Engineering,
(McGraw-Hill, New York). 1930, 2(13), 443–461.
16 Kajino, M. and Hiramatsu, M. Research and 34 Colburn, A. P. A method of correlating forced con-
development of automotive heat exchangers, in vection heat transfer data and a comparison with
Heat transfer in high technology and power engineer- fluid friction. Trans. Am. Inst. Chem. Engrs, 1955,
ing (Eds W. J. Yang and Y. Mori), 1987, p. 420–432 29, 174.
(Hemisphere, Washington, DC). 35 Sieder, E. N. and Tate, C. E. Heat transfer and
17 Beauvais, F. N. An aerodynamic look at automotive pressure drop of liquids in tubes. Ind. Engng Chem.,
radiators. SAE technical paper 650470, 1965. 1936, 28, 1429.
18 Wong, L. T. and Smith, M. C. Airflow phenomena 36 Incropera, F. P. and DeWitt, D. P. Fundamentals of
in the louvred-fin heat exchanger. SAE technical heat and mass transfer, 4th edition, 1996 (John
paper 730237, 1973. Wiley, New York).
19 Achaichia, A. and Cowell, T. A. Heat transfer 37 Gnielinski, V. New equations for heat and mass
and pressure drop characteristics of flat tube and transfer in turbulent pipe and channel flow. Int.
louvred plate fin surfaces. Expl Thermal and Fluid Chem. Engng, 1976, 16(2), 359–367.
Sci., 1988, 1, 147–157. 38 Petukhov, B. S. Heat transfer and friction in
20 Achaichia, A. and Cowell, T. A. A finite difference turbulent pipe flow with variable physical proper-
analysis of fully developed periodic laminar flow in ties. In Advances in heat transfer (Eds J. P. Hartnett
inclined louvre arrays. In Proceedings of the 2nd UK and T. F. J. Irvine), 1970, 503–564 (Academic Press,
National Heat Transfer Conference, Glasgow, 1988. New York).
21 Davenport, C. J. Correlations for heat transfer and 39 Blatti, A. Drag and heat dissipation from passenger
flow friction characteristic of louvred fin. AIChE vehicle radiators. BEng Thesis, Department of
Symp. Ser., 1983, 79(225), 19–27. Mechanical and Manufacturing Engineering, RMIT
22 Webb, R. L. and Jung, S.-H. Air-side performance University, Melbourne, 2002.
of enhanced brazed aluminium heat exchangers.
ASHRAE Trans., 1992, 98(2), 391–401.
23 Rugh, J. P., Pearson, J. T., and Ramadhyani, S. A
study of a very compact heat exchanger used for
passenger compartment heating in automobiles. APPENDIX 1
ASME Symp. Ser. HTD, 1992, 201, 15–24.
24 Webb, R. L., Chang, Y.-J., and Wang, C.-C. Heat
Notation
transfer and friction correlations for the louvre fin
geometry. Proc. Veh. Thermal Managmt System, A surface area for heat transfer
1995, 2, 533–541. A total heat transfer area on the air side
25 Sahnoun, A. and Webb, R. L. Prediction of heat a
A total heat transfer area on the coolant
transfer and friction for the louvre fin geometry. c
J. Heat Transfer, 1992, 114, 893–900. side
26 Dillen, E. R. and Webb, R. L. Rationally based heat A fin heat transfer area
f
transfer and friction correlations for the louvre fin A fin frontal heat transfer area
fr,f
geometry, SAE technical paper 940504, 1994. A radiator core frontal area
fr,r

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering
1466 E Y Ng, P W Johnson, and S Watkins

A coolant tube frontal area Y coolant tube cross-section width


fr,t cw
A total air pass area Y coolant tube length
p,a l
A total coolant pass area Y coolant tube pitch
p,c p
B core height Y coolant tube thickness
H t
B core thickness
T
B core width a angle of fin
W f
c specific heat capacity e effectiveness
p
C heat capacity rate g fin efficiency
f
D hydraulic diameter g total surface efficiency of an extended fin
h 0
f core friction factor surface
F fin height r density
h
F fin length
l
F fin pitch Subscripts
p
F fin thickness
t a air
h heat transfer coefficient
av average
i non-uniformity parameter
c coolant
j Colburn factor
i inlet
k thermal conductivity
Lp louvre pitch
L effective fin length
max maximum
L louvre height
h min minimum
L louvre length
l o outlet
L louvre pitch
p r ratio
m fin efficiency parameter
w wall
m number of rows
ṁ mass flowrate
n number of columns
n number of cells
N number of coolant tubes in one row
ct APPENDIX 2
N number of fins
f
N number of profiles
p List of assumptions
N number of rows of tubes in the core
r
depth dimension 1. The cooling system operates under steady state
NTU number of heat transfer units conditions, i.e. a constant coolant flowrate and
Nu Nusselt number fluid temperatures at both inlets.
Pr Prandtl number 2. Heat carried by the coolant only transfers to
Q overall heat transfer rate of the the airflow that travels through the radiator. Any
radiator other heat losses to, or heat gains from, the
Q actual heat transfer rate through the fin surroundings are negligible.
f
base 3. There are no phase changes in the fluid streams
Q maximum theoretical heat transfer rate flowing through the radiator.
f,max
through the fin base 4. Longitudinal heat conduction in the fluid and in
Q thermodynamically permitted maximum the wall is negligible.
max
possible heat transfer rate 5. The coolant flowrate in coolant tubes is uniformly
Re Reynolds number distributed through the radiator, with no flow
R fouling resistance maldistribution, flow stratification, flow bypass-
f
R fin end radius ing, or flow leakage occurring.
f
R coolant tube end radius 6. Coolant flow is in a fully developed condition in
t
SD specific dissipation each tube.
St Stanton number 7. Both fluids are considered incompressible flow,
T temperature and unmixed at any cross-section between passes.
U overall heat transfer coefficient The term ‘unmixed’ is defined as each fluid
V velocity through the radiator that behaves as if it was
Dx wall thickness divided into a large number of separate passages
Y coolant tube cross-section length with no lateral mixing.
cl

Proc. IMechE Vol. 219 Part D: J. Automobile Engineering D13604 © IMechE 2005
Heat transfer performance of radiators 1467

8. All dimensions are uniform throughout the 11. The thermal conductivity of the radiator material
radiator and the heat transfer surface area is is constant.
consistent and distributed uniformly. 12. The thermal resistance (fouling) induced by fluid
9. There are no heat sources and sinks in the impurities, rust formation, or other reactions
radiator walls or fluids. between the fluids and the material is assumed
10. Pure water is used as the coolant. to be small.

D13604 © IMechE 2005 Proc. IMechE Vol. 219 Part D: J. Automobile Engineering

Vous aimerez peut-être aussi