External Flows

Patricio F. Mendez
University of Alberta
ChE 314

Contents
1 Boundary Layers 3

2 Laminar Thermal Boundary Layers 3

2.1 Origins of the convection coefficient . . . . . . . . . . . . . . . 4

3 Convection Coefficient for Laminar Flow 7

3.1 Thermal boundary layer much thicker than viscous boundary
layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.1.1 Tabulation based on Nu . . . . . . . . . . . . . . . . . 10
3.1.2 Choice of fluid properties when they are not constant . 11
3.1.3 Local and Average Convection Coefficients . . . . . . . 11
3.2 Laminar Viscous Boundary Layer . . . . . . . . . . . . . . . . 12
3.2.1 Definition of Boundary Layer Thickness . . . . . . . . 13
3.3 Termal boundary layer much thicker than viscous boundary
layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 Thermal boundary layer comparable to the viscous boundary
layer: Reynolds Analogy . . . . . . . . . . . . . . . . . . . . . 15
3.5 Thermal boundary layer much thinner than the viscous bound-
ary layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4 Turbulent Thermal Boundary Layers 17

5 External Flows 19
5.1 Expression for the local Nusselt number . . . . . . . . . . . . 21

1
5.2 Methodology of solving external flow problems . . . . . . . . . 23
5.2.1 Example of calculation of average convection coefficient 23
5.3 Average Nusselt number for spheres . . . . . . . . . . . . . . . 24
5.4 Flow across banks of tubes . . . . . . . . . . . . . . . . . . . . 25

2
Introduction
The objective of this chapter is to predict the convection coefficient h. Be-
cause convection is the combination of conduction and advection, it relates
to the fluid motion, involving consideration of fluid mechanics.
In our predictions, we will cover:
1. Mechanisms: natural and forced convection
2. Behavior: laminar and turbulent
3. Geometries: external (flat plate, cylinder, sphere, bank of tubes) and
internal flow (tubes, plate “sandwich,” and cavities)
4. Orientations: regarding flow and gravity

1 Boundary Layers
Boundary layers are thin layers surrounding the surface of a body immersed
in a fluid. In your Fluid Mechanics class you were exposed to the viscous
boundary layer; however there are other types too. There is also a thermal
boundary layer that will be the focus of this class topic, and diffusion bound-
ary layers, which you will study if you take a mass transfer class. Typically,
boundary layers are considered in steady-state, and we will do so in this class
too.
The behavior of the fluid surrounding the body can be laminar or turbulent.
In turbulent flows, the turbulence introduces local advection (at the micro-
scopic level) that enhances the transport properties (thermal conductivity,
viscosity, diffusivity) as we saw when we first discussed the Peclet number.
Laminar fluids are the simplest to understand and we will start with them.

2 Laminar Thermal Boundary Layers

The thermal boundary layer is a thin layer of fluid over the surface of a
body, through which the dominant heat transfer mechanism is conduction.
A thermal boundary layer is an “insulation blanket” around a body. The
coefficient of convection h captures the heat transfer by conduction through
the thermal boundary layer.

3
Some distance away from the surface, the fluid must be in motion (advection).
If the motion is induced (e.g. by a fan, or flow in a pipe), this is called “forced
convection.” If the motion of the fluid is only due to buoyancy, then it is called
“natural convection.”
We see now why we have stated earlier that convection is not a heat trans-
fer mechanism in itself, but in reality is a combination of convection and
advection.

2.1 Origins of the convection coefficient

Let’s look in detail the temperature distribution of the fluid in the boundary
layer, illustrated in Figure 1. In this figure, x is the coordinate over the
surface of the body in the direction of the fluid flow, and y is the coordinate
perpendicular to the surface, with an origin at the surface.

4
Figure 1: Schematic of the temperature distribution of a cooling case where
Ts > T∞ and the characteristic length for heat penetration yc,t = hk .

Heat exchanged via convection has the following expression:

00
qconv = h(Ts − T∞ ) (1)

but in reality, it is conduction through the “insulation blanket”:

00 ∂T T∞ − Ts
qconv = qcond,BL = −k s
= −k (2)
∂x yc,t

where yc,t is the characteristic length in the y direction for penetration of heat
in boundary layer. In this class, we will define the thickness of the thermal

5
boundary layer as yc,t , also called the “conduction thickness.”
00 T∞ − Ts
qconv = h(Ts − T∞ ) = −k (3)
yc,t
k
yc,t = (4)
h
W
e.g. for forced air convection: k = 0.03 mK and h = 100 mW
2 K , the thermal

boundary layer thickness can be estimated as:

W
k 0.03 mK
yc,t = = 
W
 = 0.3mm
h 100 2
m 
K

The concept of boundary layer only makes sense when it is thin relative to
the size of the body. If the boundary layer thickness were to be comparable
or larger than the body, then the heat exchange would be dominated by
conduction, and should e calculated using the 2D or 3D considerations we
discussed before. The criterion of thin boundary layer is
Lc
yc,t  Lc or 1 (5)
yc,t
where Lc is the characteristic length of body. Keeping in mind that yc,t = hk :
Lc hLc
= = Nu (6)
yc,t k
where Nu is a dimensionless group called “Nusselt number.” Nu looks very
much like Bi, but they have completely different meanings. In Nu, the char-
acteristic length represents a length along the surface of a body, while for Bi
it represents a length in the direction of heat penetration. Also, the thermal
conductivity k for Nu corresponds to the fluid, while in Bi it corresponds to
the body. We can state, then, that the concept of boundary layer is valid
when the Nusselt number is much larger than 1.
As an example, consider the internal side of a 1 m tall window, h ≈ 10 mW
2K ,

and for the thermal conductivity of air (k =0.03 W/m K), the Nusselt number
is:
W
Lc h 1 × 10
m m
2K
 = 333  1
Nu = =

W
k 0.03 m
K
confirming the boundary layer is thin.

6
3 Convection Coefficient for Laminar Flow
The thermal boundary layer involves fluid flow. In this class we will study
flows against a surface, which involve a viscous boundary layer. We now have
two boundary layers, thermal and viscous, which result in two extreme cases
1. Thermal boundary layer is much thicker than viscous boundary layer.
2. Thermal boundary layer is thinner or comparable to the viscous bound-
ary layer.
For the first case, almost all the thermal boundary layer experiences the
velocity of free flow. For the second case, the thermal boundary layer develops
in the region where the flow is slowed down by the no-slip condition (flow
velocity relative to the surface is zero in direct contact with the surface).
The calculation for each case is different. Let’s start with the first case that
is the simplest.

3.1 Thermal boundary layer much thicker than viscous

boundary layer

Figure 2: Schematic of the thermal boundary layer of a cooling case where

Ts > T∞ .

The boundary layer is thin in the y direction and large in the x direction.
We have seen before that when the spect ratio is so large, conduction in the

7
long direction (x) is negligible and only conduction in y matters. Because of
the motion of the fluid, there is advection in the x direction. In this case we
will build a control volume in Langrangian coordinates, which moves with
the flow at velocity u∞ .
Because the control volume moves with the fluid, there is no advection in the
energy balance, since no flow crosses the boundaries of the control volume.
We are left then with only conduction in y for this control volume. Although
the boundary layer is in steady-state, our Lagrangian control volume is not
in steady-state, because it is moving with the flow and experiencing different
temperatures as it moves.
If we consider that at time t = 0 the Lagrangian control volume is at x = 0,
there is a relationship between position and time:

x = u∞ t (7)

As the control volume moves downstream, the governing equation will be

∂T ∂ 2T
=α 2 (8)
∂t ∂y
which is the same equation we have seen in a semi-infinite solid, but applied
to the fluid now. Initial temperature is T∞ , and the temperature far from the
plate is also T∞ . The third boundary condition can vary, just as it happened
for the semi-infinite solid.
Two common situations for the third boundary condition are constant plate
temperature, and constant heat flux from the plate. We will consider the
first case, of a plate a constant temperature, which is equivalent to the semi-
infinite solid with constant surface temperature.
In this case, the penetration of heat into the control volume will be
√ √
yc,t = π αt (9)

which can incorporate the relationship between time and position of Equa-
tion 7, resulting in
√
r
αx
yc,t = π (10)
u∞

8
and the following estimate of Nu:
r
x 1 u∞ x
Nux = =√ (11)
yc,t π α

where yc,t should always be much smaller than x for the concept of boundary
layer to be applicable. This means the leading edge of the plate (when x → 0)
cannot be accurately captured by this analysis.
We can use Equation 4 to calculate the convection coefficient h
r
k k u∞
h= =√ (12)
yc,t π αx

We see that the value of h is infinity at the location of x = 0. This is

consistent with the infinite heat transfer rate at t = 0 in the semi-infinite
solid.

9
Figure 3: Schematic of the thermal boundary layer yc,t and convection coef-
ficient h as a function of x.

3.1.1 Tabulation based on Nu

Convection coefficients are typically tabulated in dimensionless form. The di-
mensionless number capturing the convection coefficient is the Nusselt num-
ber Nu we discussed earlier in Equation 13
hx
Nux = (13)
k
where the subscript x indicates the characteristic length considered. The
dependence of Nu is typically based on the Reynolds number Re
u∞ x
Rex = (14)
α
10
which relates the relative magnitude of inertial forces to viscous forces in the
flow, and the Prandtl number Pr
ν
Pr = (15)
α
which relates the fluids ability to transmit heat to its ability to transmit
momentum through shear. Applying equations 13, 14, and 15 to 12 we
obtain
1
Nux = √ Re1/2 Pr1/2 = 0.564Rex1/2 Pr1/2 (16)
π
which is the exact solution to the laminar thermal boundary layer when it is
much thicker than the viscous boundary layer. It is remarkable we arrived to
the exact solution without solving any differential equation. The traditional
treatment of this problem involves solving a system of four linked non-linear
partial differential equations.

3.1.2 Choice of fluid properties when they are not constant

If the thermal boundary layer (TBL) covers range from Ts to T∞ , what
temperature to use for fluid properties in calculation?
• It will be said explicitly in the problem description (typically T∞ , and
calculations involve a correction factor based on Ts ).
• If not said, use the “Film Temperature”:
Ts + T∞
TF = (17)
2

3.1.3 Local and Average Convection Coefficients

When we calculate the heat transfer rate via convection using qconv = Ah(Ts −
T∞ ), an average h is used. However, we have seen that h varies with position
and it is necessary to differentiate the local and average value:
Notation:
• h̄: average convection coefficient
• h: local convection coefficient
• Nu = h̄Lc
k
: average Nusselt number

11
 • Nu = hLc
k
: local Nusselt number
For the thermal boundary layer studied:
r
k k u∞
h= =√ (18)
yc,t π αx
hx x 1
Nux = = = √ Pex 1/2 (19)
k yc,t π

calculation of the average convection coefficient over a length of L:

Z Lc
u∞ Lc dx
r Z
1 1 k
h̄Lc = hdx = √ √ (20)
Lc 0 Lc π α 0 x
r
1 k u∞ x1/2 Lc
= √ (21)
Lc π α 1/2 0
h̄Lc = 2hLc (22)
NuLc = 2NuLc (23)

An important conclusion from this analysis is that we should not confuse

local and average values. Average values are used more often.

3.2 Laminar Viscous Boundary Layer

Fluid mechanics and heat transfer are very similar transport problems in-
volving the diffusion of a quantity: thermal energy for heat transfer, and
momentum for fluid flow.
For heat transfer the associated diffusivity is the thermal diffusivity α =
k/(ρcp )
For momentum transfer the associated diffusivity is the kinematic viscosity
ν = µ/ρ, where the viscosity coefficient µ is typically independent of pressure
but heavily dependent on temperature.
The ratio of the two diffusivities represent the ratio between the transport of
heat and momentum. Both have units of m2 /s, resulting in a dimensionless
number. Because of its importance, it is given a name: the Prandtl number:
ν
Pr = (24)
α
12
Similarly, in mass transport the Schmidt number is defined similarly to Pr:
Sc = ν/D.
The thickness of viscous boundary layer for incompressible flow was deter-
mined by Prandtl and his student PhD Blasius. The exact solution involves
infinite series and is very tedious; so tedious that upon graduating Blasius
decided to never return to research and became a teacher. From Blasius suf-
fering, however, we can infer the following thickness for the viscous boundary
layer (defined using the tangent at the surface, as we usually do in this class):
r
νx
yc,v = 3 (25)
u∞
r
x 1 u∞ x
= ≈ 0.33Rex 1/2 (26)
yc,v 3 ν

Figure 4: Schematic of laminar velocity boundary layer (VBL).

3.2.1 Definition of Boundary Layer Thickness

There are many different definitions, and all of them are of the same order
of magnitude. We use yc,t and yc,v because:

∂T T∞ − Ts
=
∂y yc,t
∂u u∞
=
∂y yc,v

13
Most common definition of boundary layer in textbooks ispδ such that u(δ) =
0.99u∞ and the resulting boundary layer thickness δ = 5 uνx∞
, as illustrated
below:

Figure 5: Schematic of the most common way of defining velocity boundary

layer: u(δ) = 0.99u∞ .

Exactly same analysis applies to diffusion boundary layers replacing Pr by

the Schmidt number Sc, which is typically  1.

3.3 Termal boundary layer much thicker than viscous

boundary layer
As derived before, for yc,t  yc,v , Ts =constant, and Pe  1, the thickness of
thermal boundary layer is:
√
r r
αx αx
yc,t = π ≈ 1.77
u∞ u∞
x 1
= Nux = √ Rex 1/2 Pr1/2 ≈ 0.564Rex 1/2 Pr1/2
yc,t π

Now we can compare:

yc,t 1.77 α
= ≈ 0.6Pr−1/2
yc,v 3 ν

14
We can see that yc,t  yc,v is valid only when Pr  1, which is the case of
molten metals and plasmas. The exponent 1/2 in the Prandtl number is the
telltale sign that the expression corresponds to low Pr, even in configurations
different than a flat plate. Because this happens typically for molten metals,
expressions with the 1/2 exponent in Pr are uncommon.

3.4 Thermal boundary layer comparable to the viscous

boundary layer: Reynolds Analogy
For the case when yc,t = yc,v , both boundary layers are the same and have
the same solution. This is called the Reynolds Analogy.
r
αx
yc,t = 3 (27)
u∞

which is the exact solution to the problem, valid for Pe=1, which is close to
the values for air and other common gases.

3.5 Thermal boundary layer much thinner than the

viscous boundary layer
For the case when yc,t  yc,v , the Reynolds analogy still holds approximately,
but the maximum velocity that the thermal boundary layer experiences is
smaller than the velocity of free flow u∞ . We can do this analysis by com-
paring triangles ABC and ADE:
DE BC
=
EA CA
u∞ uc
=
yc,v yc,t
yc,t
uc = u∞
yc,v

15
Figure 6: Schematic of velocity boundary layer and thermal boundary layer
for Pr  1.

As yc,v = 3 uνx and yc,t = 3 αx

p p
∞ uc
:
αx αx yc,v
yc,t 2 = 9 =9
uc u∞ yc,t
αx
yc,t 3 = 9 yc,v
u∞
 3
yc,t αx αx u∞
 1
=9 2
= 9

=
yc,v u∞ yc,v u
 ∞9
ν
x Pr
yc,t
= Pr−1/3 valid for Pr  1 and Pr ≈ 1
yc,v

calculation of h:
k k yc,v k yc,v
h= = =
yc,t yc,t yc,v yc,v yc,t
yc,t
combing yc,v = 3 uνx = Pr−1/3 :
p
∞
and yc,v

k k
h= = p νx Pr1/3
yc,t 3 u∞

16
tabulating in the form of Nux :
r
hx 1 u∞ x 1/3
Nux = = Pr
k 3 ν
≈ 0.33Re1/2 Pr1/3

which is the exact solution for Pr=1, and a close approximation for Pr&1,
which includes air, water and oil. The exponent 1/3 in the Prandtl number is
the telltale that the expression corresponds to Pr&1, even in configurations
different than a flat plate. Because this happens typically for aqueaous liq-
uids, gases, and oil, expressions with the 1/3 exponent in Pr are ubiquitous.

4 Turbulent Thermal Boundary Layers

Laminar flows are ordered and have clear streamlines while turbulent are
chaotic, and microscopic motion enhances transport:

Figure 7: Schematic of the development of viscous boundary layers: laminar,

transition region and turbulent

The onset of turbulence is when Rex > Recritical .

The local Reynolds number is defined: Rex = u∞ν x and Recritical depends on
the geometry of the system and other factors. For a flat plate: Recritical =
5 × 105 . (This can vary with other conditions.)

17
In the turbulent area, transport is dominated by advection from turbulent
⇒ yc,t ≈ yc,v , δt ≈ δ. The behavior of transport is much different in laminar
and turbulent region.

Figure 8: Schematic of the development of viscous boundary layers and vari-

ations on the local Nusselt number as a function of Re and Pr.

The average Nusselt number NuL for combined laminar and turbulent bound-
ary layers can be estimated as:
 
NuL = 0.037ReL − A Pr1/3 Eq. 7.31
4/5

A = 0.037Rex,c 4/5 − 0.664Rex,c 1/2

where Rex,c is the critical Reynolds number, the same as the notation Recritical .

18
5 External Flows
Considering common external flows in radial systems (e.g. cylinders or
spheres):

Figure 9: Schematic of boundary layer formation and separation on a circular

cylinder in cross flow.

The occurrence of boundary layer transition is dependent on the Reynolds

number defined as:
ρV D VD
ReD = =
µ ν

Critical ReD of the onset of the boundary layer transition: ReD = 2 × 105

19
Figure 10: Schematic of the transition from laminar boundary layer to tur-
bulent boundary layer and the effect of turbulence on separation.

The behavior of the external flows strongly influence the drag force acting
on the cylinder FD , which consists of two components: the boundary layer
surface shear stress and the other part results from a pressure differential in
the flow direction. In dimensionless form, drag coefficient CD is defined as:
FD
CD =
Af (ρV 2 /2)
where Af is the cylinder frontal area projected perpendicular to the velocity
of the free stream. The influence of Re on CD are presented below:

Figure 11: Drag coefficients as a function of Re for a smooth cylinder and a

sphere, respectively.

20
5.1 Expression for the local Nusselt number
At the forward stagnation point (θ = 0) for Pr & 0.6, the local Nusselt
number has the following expression and is most accurate at low Reynolds
number:

NuD (θ = 0) = 1.15ReD 1/2 Pr1/3 Eq. (7.43)

Figure 12: Local Nusselt number for airflow normal to a circular cylinder.

However, overall average conditions are more widely used in engineering cal-
culation. An empirical correlation has been modified and is widely used for
Pr & 0.7:

h̄D
NuD = = CReD m Pr1/3 Eq. (7.44)
k
where constants C and m can be checked from the below table for the circular
cylinder in cross flow. (Table 7.2 in the textbook):

21
Figure 13: Constants of Equation 7.44 for the circular cylinder in cross flow

Equation 7.44 can also be applied for external flows over cylinders of non-
circular cross section with the characteristic length D and constants used in
calculation:

Figure 14: Constants of Equation 7.44 for the non-circular cylinders in cross
flow of a gas

Note: in working with Equations 7.43 and 7.44, all properties are
evaluated at the film temperature.

22
Another correlation of the average Nusselt number for the circular cylinder in
cross form is valid for 0.7 . Pr . 500, 1 . ReD . 106 and has the following
expression:
 1/4
h̄D m n Pr
NuD = = CReD Pr Eq. (7.45)
k Prs

where all properties are evaluated at T∞ , except Prs which is evaluated at

Ts . Constants of n in Equation 7.45 is dependent on the value of Pr:
• If Pr . 10, n = 0.37
• If Pr & 10, n = 0.36

Figure 15: Constants of Equation 7.45 for the circular cylinder in cross flow

5.2 Methodology of solving external flow problems

Step1 Select appropriate formula for geometry
Step2 Assess Re , Pr and other factors at appropriate reference temperature
Step3 Calculate Nu: don’t confuse local Nu and average Nu
Step4 Calculate h: don’t confuse local h and average h

5.2.1 Example of calculation of average convection coefficient

Calculate average h for a 20 mm diameter cylinder immersed in cross flow of
water at 1m/s. Thermophysical properties for water: thermal conductivity
2
W
k = 0.6 mK , thermal diffusivity α = 0.15 × 10−6 ms and Pr = 7

23
Step1 Select appropriate formula for geometry
cylinder, Pr = 7 & 0.1 to calculate average convection coefficient⇒using
Equation 7.44 NuD = h̄D k
= CReD m Pr1/3
Step2 Assess Re , Pr and other factors at appropriate reference temperature

Pr = 7
m2 m2
ν = αPr = 0.15 × 10−6 × 7 = 1.05 × 10−6
s s
m
u∞ D 1 s
× 0.02m 4
ReD = = m 2 = 1.9 × 10
ν −6
1.05 × 10 s

According to Table 7.2, C=0.193 and m=0.618

Step3 Calculate Nu

0.618
NuD = 0.193 1.9 × 104 × 71/3 = 163.0

Step4 Calculate h
W
h̄D k 163 × 0.6 mK W
NuD = ⇒ h̄ = NuD = = 4890 2
k D 0.02m mK

5.3 Average Nusselt number for spheres

With constant properties:

NuD = 2 + 0.6ReD 1/2 Pr1/3 Eq. (7.49)

With temperature-dependent properties:

 1/4

1/2 2/3

0.4 µ
NuD = 2 + 0.4ReD + 0.06ReD Pr Eq. (7.48)
µs

where all properties except µs are evaluated at T∞

24
5.4 Flow across banks of tubes
The geometric arrangement of a bank of tubes is typically, one fluid moves
over the tubes while a second fluid of different temperature passes through the
tubes. There are two common arrangement of banks: aligned or staggered in
the direction of the fluid velocity. The configuration is characterized by the
tube diameter D, transverse pitch ST , and longitudinal pitch SL measured
between tube centers. Flow conditions within the bank are dominated by
boundary layer separation effects, wake interactions, and the convection heat
transfer. Flow conditions are quite different and strongly dependent on the
bank arrangement.

Figure 16: Schematic of a tube bank in cross flow and two common arrange-
ments in a bank: aligned and staggered.

25
 Figure 17: Flow conditions for aligned and staggered bank arrangements

The heat transfer rate per unit length q 0 of the tubes can be computed by:
q 0 = N h̄πD∆Tlm


Eq. (7.56)
where N is the total number of tubes in the bank and ∆Tlm is the log-
mean temperature difference used to compute the appropriate temperature

26
difference in calculation:
(Ts − Ti ) − (Ts − To )
Tlm =   Eq. (7.54)
Ts −Ti
ln Ts −To

where Ti and To are the temperatures of the fluid as it enters and leaves the
bank, respectively. In order to calculate ∆Tlm , the outlet temperature To is
needed and can be estimated by:
 
Ts − To πDN h̄
= exp − Eq. (7.55)
Ts − Ti ρV NT ST cp

The average heat transfer coefficient for the entire tube bank has the following
correlation valid for NL & 20, 0.7 . Pr . 500, 10 . ReD,max . 2 × 106
 1/4
Pr
NuD = C1 Rem
D,max Pr
0.36
Eq. (7.50)
Prs

where ReD,max is calculated based on the maximum velocity:

• For aligned arrangement: Vmax = ST
ST −D
V Eq. (7.52)
• For staggered arrangement, choose the higher one between
ST
Vmax = V Eq. (7.53)
2 (SD − D)
ST
Vmax = V Eq. (7.52)
ST − D

Properties involved in Eq. 7.50 are evaluated at the film temperature Tf =

(Ti + To )/2 except Prs , which is evaluated at the temperature of the surface
Ts .

27
 Figure 18: Constants of Equation 7.50 for the tube bank in cross flow

If there are 20 or fewer rows of tubes (NuL 6 20), the average heat coeffi-
cient is typically reduced and a correction factor is necessary to capture the
reduction:

NuD = C2 NuD Eq. (7.51)

(NuL <20) (NuL >20)

where the correction factor C2 can be found in Table 7.6:

Figure 19: Correction factor C2 of Equation 7.51 for NuL < 20

28
Summary
Correlation Geometry Re/Pr Other Temp. Condition
Dimensionless
δ = 5 × Re−1/2
x (7.17) Flat plate Re<5 × 105 - Tf Laminar
−1/2
Cf,x = 0.664Rex (7.18) Flat plate Re<5 × 105 - Tf Laminar,
local
1/2
Nux =0.332Rex Pr1/3 (7.21) Flat plate Re<5 × 105 - Tf Laminar,
Pr & 0.6 local
δt = δ Pr−1/3 (7.22) Flat plate Re<5 × 105 - Tf Laminar
Pr& 0.6
−1/2
C f,x = 1.328Rex (7.24) Flat plate Re<5×105 - Tf Laminar,
average
1/2
Nux = 0.664Rex Pr1/3 (7.25) Flat plate Re<5 × 105 - Tf Laminar,
Pr& 0.6 average
1/2
Nux = 0.564Pex (7.26) Flat plate Re<5 × 105 Pex & 100 Tf Laminar,
Pr. 0.05 local
−1/5
Cf,x = 0.0592Rex (7.28) Flat plate Re. 108 - Tf Turbulent,
local
δ = 0.37 × Re−1/5
x (7.29) Flat plate Re. 108 - Tf Turbulent
4/5
Nux = 0.0296Rex Pr1/3 (7.30) Flat Plate Re. 108 - Tf Turbulent,
0.6 .Pr. 60 local
−1/5
C f,L = 0.074ReL − 1742Re−1 L Flat plate Re. 108 - Tf Mixed,
(7.33) average
4/5
NuL = (0.037ReL −871)Pr1/3 (7.31) Flat plate Re. 108 - Tf Mixed,
0.6 .Pr . 60 average
NuD =CRem D Pr
1/3
(7.44) Cylinder 0.4.Re . 4 × 105 - Tf Average
(Table 7.1) Pr& 0.7
NuD = CRem n
D Pr (Pr/Prs )
1/4
(7.45) Cylinder 1.Re. 106 - T∞ Average
(Table 7.2) 0.7. Pr. 500
1/2
NuD = 0.3+[0.62ReD Pr1/3 Cylinder RePr& 0.2 - Tf Average
×[1+(0.4/Pr)2/3 ]−1/4 ]
×[1+(ReD /282, 000)5/8 ]4/5 (7.46)
1/2 2/3
NuD = 2 + (0.4ReD + 0.06ReD )Pr0.4 Sphere 3.5. Re. 7.6 × 1.0. µ/µs . 3.2 T∞ Average
×(µ/µs )1/4 ) (7.48) 104
0.71 . Pr. 380
1/2
NuD = 2 + 0.6ReD Pr1/3 (7.49) Falling - - T∞ Average
drop
NuD = Tube bank 10 .Re. 2 × 106 - T Average
C1 C2 RemD,max Pr
0.36
(Pr/Prs )1/2 0.7 .Pr . 500
(7.50), (7.51) (Tables 7.3, 7.4)
εj H = 2.06Re−0.575
D (7.72) Packed 90.Re . 4000 - T Average
bed Pr≈ 0.7
of spheres

For tube banks and packed beds, properties are evaluated at the average fluid
temperature, T = (Ti + To )/2

29
 Table 7.1: Constants of Equation 7.44 for the circular cylinder in cross flow
ReD C m
0.4-4 0.989 0.330
4-40 0.911 0.385
40-4000 0.683 0.466
4000-40,000 0.193 0.618
40,000-400,000 0.027 0.805

Table 7.2: Constants of Equation 7.45 for the circular cylinder in cross flow
ReD C m
1-40 0.75 0.4
40-1000 0.51 0.5
103 − 2 × 105 0.26 0.6
5 6
2 × 10 − 10 0.076 0.7

Table 7.3: Constants of Equation 7.50 for the tube bank in cross flow

Configuration ReD,max C1 m
Aligned 10-102 0.80 0.40
Staggered 10-102 0.90 0.40
Aligned 102 − 103 Approximate as a single
Staggered 102 − 103 (isolated) cylinder
Aligned 103 − 2 × 105 0.27 0.63
(ST /SL >0.7)∗
Staggered 103 − 2 × 105 0.35(ST /SL )1/5 0.60
(ST /SL <2)
Staggered 103 − 2 × 105 0.40 0.60
(ST /SL >2)
Aligned 2 × 105 − 2 × 106 0.021 0.84
5 6
Staggered 2 × 10 − 2 × 10 0.022 0.84
*For (ST /SL )<0.7,heat transfer is inefficient and aligned tubes should not be used.

30
Table 7.4: Correction factor C2 of Equation 7.51 for N uL <20 (ReD,max & 103 )

NuL 1 2 3 4 5 7 10 13 16
Aligned 0.70 0.80 0.86 0.90 0.92 0.95 0.97 0.98 0.99
Staggered 0.64 0.76 0.84 0.89 0.92 0.95 0.97 0.98 0.99

Acknowledgement
This set of notes is based on class notes from previous years by Prof. Mendez,
and was first typed and illustrated by PhD student Ying Wang as part of
her Fraser and Shirley Russell Teaching Fellowship.

31