C
HAPTER
6:
I
NTRODUCTION
TO
C
ONVECTION
For conduction through solids, we derived the general heat diffusion
equation based on all the mechanisms by which heat can be transferred or generated in a solid control volume:
= +
∂ ∂ ∂
∂ +
∂ ∂ ∂
∂ +
∂ ∂ ∂
∂
q z
T k z y
T k y x
T k
x & t
T cp
∂ ∂
ρ
(Ein – Eout)cond + Egen = Estorage
qx qx+dx
qz
qz+dz
qy
qy+dy
What happens if instead we use a fluid
(gas, liquid) control volume? All these terms remain BUT… heat transfer by
bulk fluid flow also becomes possible:
k T q cp Tt + cpu ⋅∇T ∂
∂ =
+
∇2 & ρ ρ
(Ein – Eout)cond + Egen = Estorage + Efluid flow
where u ⋅∇T = u ∂∂Tx + v ∂∂Ty + w ∂∂Tz
u = velocity in x-direction; v = velocity in
y-direction; w = velocity in z-direction
During our studies of conduction, we treated convection as a boundary
condition (i.e. if all the convective
medium has T = T∞ ! dT/dx = dT/dy
= dT/dz = 0 in the fluid itself)
However, if dealing with heat transfer through fluids, the inertial (flow) term becomes important to the analysis.
This is particularly true for convection analysis given that convection
involves heat transfer via both
conduction and advection, the
physical movement of heated (or
cooled) fluid from a solid interface or within a fluid.
Advection may arise from different types of fluid flows:
" External flows
• Forced, i.e. flowing at a
significant velocity (pumps, fans)
• Free, i.e. stagnant flow where
diffusion effects dominate
" Internal flows (pipes & conduits)
Thus, in order to develop a framework for convection analysis, we need to
know something about how fluids flow under these different circumstances ! FLUID MECHANICS
F
LUID
M
ECHANICS
The flow properties of a fluid are
defined by its density (
ρ
) and viscosity(
µ
) [Note: both strong functions of T]Density – mass per unit volume
! Property of atoms/molecules
High density (liquids) ! larger inertial force required to move a given
volume of fluid (F = ma)
Viscosity - internal resistance of a
fluid to flow (unit: Pa s = N/m2s)
! Intermolecular interactions
! Molecular tangling (polymers)
High viscosity (oil, polymers)! large
inertia/resistance to flow
! Flow in highly viscous fluids may generate frictional heat
(viscous dissipation)
Velocity Boundary Layer
When a velocity gradient exists in a flow field, the fluid experiences shear forces. A measure of the shear force per unit area is called shear stress,τ.
∂u/∂y = fluid velocity
For a fluid traveling in the x-direction over a surface, the force diagram
acting on the fluid is shown below:
y u Area
Force
∂ ∂ =
= µ
τ
FLUID FLOW
Friction
Applied Force
SURFACE
Key Feature: No-Slip Surface where net force = 0 (applied flow force =
surface frictional drag) !
! !
! wall velocity u = dx/dt|x=0 = 0
Thus, the velocity profile of a flowing fluid over or within an object exhibits a parabolic shape near the interface:
u∞ = free stream fluid velocity parallel
to the surface (no frictional drag)
δ = boundary layer thickness (defined as y value at which u = 0.99u∞)
The parabolic shape of the velocity profile is called the “velocity
boundary layer” of fluid flow.
δ
uu∞
u∞
Plug Flow Surface Flow
x
y Surface shear stress
Frictional effect No surface effect Bulk fluid flow
Consider the case where a plug flow of fluid contacts a wall (x = 0). With increasing distance from the wall
interface, the viscosity (frictional drag
within fluid) transmits the surface
frictional drag deeper into the fluid ! development of boundary layer
Boundary layer development is
governed by the fluid velocity and the frictional drag at the wall interface,
which can be expressed in terms of
the local friction coefficient Cf:
high Cf = high friction
Plug Flow
δ
δ(x)Fully developed flow Steady State
x
2 2
∞ ≡
u
C f s
ρ τ
Thermal Boundary Layer
Thermal boundary layers form as a result of two contributions:
! Advection contributions to
convection change as the fluid
velocity changes as a function of y
• directly proportional to u profile
! Conductive contributions to
convection vary according to the fluid’s thermal conductivity (kf)
• independent of u profile
• if heat can’t diffuse into fluid (u=0
at surface), advection is useless!
δ
tq x
y
Ts u = 0
u∞
Fl
u
id
La
y
er
s
T∞ T∞ No heat flow between these two layers
δ
t is the thickness of the thermal boundary layer, defined as the yposition at which
↑
δ
t ! ↓∂
T/∂
y ! slower heat transferQ: Will
δ
t andδ
be the same value?At the surface, u=0 ! conduction only Fourier’s Law applies:
-heat must first diffuse out of solid surface!
Newton’s Law of Cooling (convection from surface):
By energy balance, q”s1 = q”s2; thus:
h depends on kf
(conduction) and boundary
layer temperature gradient
(advection)
Thus, h varies with the fluid velocity
profile defining the boundary layer
99 . 0 = −
− ∞
T T
T T
s s
0 "
1
=
∂ ∂ −
=
y f
s
y T k
q
(
− ∞)
= h T T qs"2 s
(
− ∞)
=∂ ∂ −
=
T T
y T k
h
s
y f
0
We have already applied different h
values in the same convective fluid for fin analysis (comparing h vs. hL)
h = average convection coefficient
hL = localized convection coefficient
To account for localized h values due to boundary layer variation, the total heat transfer rate over any arbitrary surface is the integral of the local heat fluxes:
Define the heat transfer rate in terms of h , the average convection
coefficient over surface: q = hAs(Ts −T∞)
where =
∫
⋅s
A
s s
dA h
A
h 1
(general) or
(for a flat plate) !main problem of convection analysis
(
)
∫
∫
= − ∞=
A
s A
s
s T T hdA
dA q
q "
∫
=L
dx x h L h
0
) ( 1
C
ONVECTION
P
ROBLEMS
Step 1. Solve the appropriate
equations of fluid mechanics to define the velocity profile
1) Start with the general continuity equation for fluid flow through a
differential control volume (2D flow):
(flow in = flow out)
2) Apply Newton’s 2nd Law of Motion
(ΣF acting on control volume = net rate at which momentum leaves the control volume). In the x-direction:
0 = ∂ ∂ + ∂ ∂ y v x u X y u x v x P y v v x u
u +
∂ ∂ + ∂ ∂ + ∂ ∂ − = ∂ ∂ + ∂ ∂ 2 2 2 2 µ ρ
Net rate of = momentum
flow from CV
Net + pressure force
Step 2. Solve the appropriate form of the heat diffusion
equation to define the temperature profile
For steady, 2D flow of an
incompressible fluid with constant properties:
where the viscous dissipation term (mechanical friction during fluid flow that generates heat) is given by:
… solutions can be very ugly!
Φ + + ∂ ∂ + ∂ ∂ = ∂ ∂ + ∂ ∂ µ ρ q y T x T k y T v x T u
cp &
2 2 2
2
Net rate of heat = flow due to bulk fluid flow
(advection)
Net heat + flow by conduction Net heat generation by viscous dissipation Net heat +
Step 3. Estimate the convection
heat transfer coefficient h based
on the energy balance
Since the velocity at the wall surface is zero, Ts,wall = Tfluid, y=0 ! heat
transfer by conduction only at surface. Thus, by energy balance at y>0:
(
wall fluid)
wallT T
h y
T k
q" = − ∞,
∂ ∂ −
=
Solve for h to get average convection coefficient over a given surface
Step 4. Use Newton’s Law of
Cooling to estimate heat flow/flux
or
Heat Flux Heat Flow
We need
δ
T/δ
y ∝δ
u/δ
y ! functional approximations of v/T profile solutions(
Twall T fluid)
h
q"= − ∞, q = hAs
(
Twall −T∞, fluid)
L
AMINAR
/T
URBULENT
F
LOW
The velocity and thermal boundary layers strongly depend on the nature of the fluid flow over the surface.
Slower Velocities:
LAMINAR FLOW –
! fluid will travel along continuous stream-lines
Adjacent layers of fluid move pass one another along parallel paths.
Faster Velocities: TURBULENT FLOW
! fluid flow lines are discontinuous
around the object Swirling flow regions (“eddies”) occur Random 3D motion of large fluid V
How does the flow regime impact surface friction and h?
Laminar: Ordered flow;
∂
u/∂
y ↓ as x↑since
δ
increases as a function of xTurbulent: Disordered flow; typically
comprised of distinct layers
# Close to surface: viscous sublayer
– diffusion dominant, linear
∂
u/∂
y# Intermediate: buffer layer – both
diffusion/turbulent mixing occur
# Outer layer: turbulent mixing
predominant, near-zero
∂
u/∂
yHow do these flow regimes impact heat transfer?
!
Compare the surface velocity profiles in laminar and turbulent flow:
Laminar Flow: Velocity
gradient (y direction) occurs over full
boundary layer
Turbulent Flow:
Significant y velocity gradient observed only in viscous sublayer
>
(turbulent) (laminar) >
(turbulent) (laminar) 0 = ∂ ∂ = y s y u µ
τ 2 0
2 2 2 ∞ = ∞ ∂ ∂ = ≡ u y u u
C f s y
Laminar-to-turbulent flow is triggered by surface roughness, surface
vibrations, or stream fluctuations.
The transition between laminar and turbulent regimes is identified based on a dimensionless number, the
Reynolds Number (Re), the ratio
between inertial and viscous forces.
µ
ρ
µ
ρ
µ
ρ
vLL v
L v v
dx v d
dx dv v
F F
viscous
inertial = ≈ =
=
2 2
2
Re
v = bulk fluid velocity over object/surface
In a tube: Re = ρuµ∞D
Laminar Transition Turbulent
Re < 2100 2100 – 10000 > 10000
Over a plate: Re = ρuµ∞x
Laminar Transition Turbulent
Re < 5x105 5x105 – 3x106 > 3 x 106
D
IMENSIONLESS
N
UMBERS
See Table 6.2 for other dimensionless numbers used in convection analysis.
For laminar flow, the thickness of the velocity boundary layer (
δ
) and the temperature boundary layer (δ
t) are related by the Prandtl Number (Pr)Notes: cp, k values are for the fluid
ν = µ/ρ is the kinematic viscosity
Pr therefore represents the relative effectiveness of momentum transport
in the velocity boundary layer to that
of energy transport (by heat diffusion)
in the thermal boundary layer.
f p
p
f k
c
c k
y diffusivit thermal
y diffusivit
momentum µ
ρ ρ
µ
= =
= Pr
The boundary layer thicknesses in the laminar flow regime are related to the Prandtl number by the expression:
n = positive exponent (often ~1/2)
For gases, low k, cp,
µ
; Pr ~ 1 !δ
≅δ
t For water, mid k,µ
; high cpPr ~ 7 !
δ
>δ
t# Limiting step of heat transfer is conduction of heat into water (i.e. through no-slip layer at surface) For oils, low k, high cp,
µ
Pr > 1000 !
δ
>>δ
t# Advection effective, but relatively little heat diffuses into oil
For molten metal, high k, mid cp,
µ
Pr < 0.01!δ
<<δ
t# Large amount of heat diffuses into metal but advection is less important for heat dissipation
n
t
Pr ≈ δ
δ
Dimensionless characteristic lengths, velocities, and temperatures can also be defined, as in Chapter 5:
Dimensionless length L = characteristic length
Dimensionless velocity
V = upstream velocity = u∞
Dimensionless Temperature
These dimensionless parameters are useful since they reduce every
variable to a unitless variable with a value between 0-1
# simplify boundary condition evaluation in ODE solving # enable combination of these
parameters in any form (e.g.
RePr would also be dimensionless – we will use this later)
L x
x* ≡
L y
y* ≡
V u
u* ≡
V v
v* ≡
s s
T T
T T
T
− − ≡
∞ *
We can also apply these definitions to simplify the expressions defining other dimensionless parameters.
For example, rewriting the friction
factor Cf in terms of the dimensionless parameters u = u∞u* and y = Ly*:
# Reduce the number of variables in a given problem.
o Three variables (ρ,
µ
, u∞) arebe reduced to one single,
dimensionless variable with a universal applicability to various fluids/flow conditions (Re).
# Differential boundary conditions
now u*=0 at y* = 0 and u* = 1 at
y* = 1 ! easier to evaluate
0 * 0
* 2
0 * 2
* *
Re 2
* * 2
2
= =
∞ ∞
=
∞ ∂
∂ =
∂ ∂ =
∂ ∂ ≡
y L
y y
f
y u y
u u
L u y
u u
C
ρ
µ
ρ
µ
Alternately, substituting dimensionless parameters into expressions may
expose new dimensionless constants of use to further simplify problems.
Substituting T = T*(T∞ - Ts) + Ts and
y = Ly* into the definition of the
convection coefficient derived earlier:
!
!
!
!
Simplifying,
Nusselt Number
(Nu)
May be expressed locally (Nu=hx/kf) or averaged over a surface (Nu=hL/kf)
f s
f
s
k hL L
T T
k
T T
h
Nu =
− −
=
∞ ∞
/ ) (
) (
Physical definition: Nu is the ratio of
convection to conduction heat transfer through the fluid normal to surface
Compare to the Biot Number:
s
k hL
Bi =
Biot Number – conduction in solid
# Use k value for solid, L is the dimension over which the
conduction occurs (x for a plate, r
for infinite cylinder/sphere, or
V/As for other shapes)
Nusselt Number – heat conduction
occurs through liquid phase
# Use k value for fluid, L is the dimension associated with
boundary layer development (L
for plates, D for cylinder/sphere, or V/As for an arbitrary shape)
We can apply these dimensionless
numbers to evaluate the velocity and temperature profiles and calculate an average h value for convection.
Re-write the velocity and temperature profile equations in terms of
dimensionless parameters to simplify boundary conditions for evaluation (see Table 6.1 for full expressions):
Momentum:
Temperature:
dP*/dx* = pressure gradient on surface (determined by geometry of object)
We can then solve these functions with reference to dimensionless
parameters to reduce the number of variables which need be considered.
(
*, *,Re , * *)
*
dx dP
y x f
u = L
(
*, *,Re ,Pr, * *)
*
dx dP
y x f
T = L
For a prescribed geometry, dP*/dx* is fixed (pressure distributions can be
independently obtained by considering flow conditions in the free stream)
Thus, at y* = 0 (surface) and remembering that
the thermal boundary layer can be described as a function of Nu:
Or, integrating over all x for an average surface property:
Thus, a problem involving 7 variables
(h, kf, cp, ρ,
µ
, V, L) is reduced to onecontaining 3 universally applicable
dimensionless constants (Nu, Re, Pr).
(
x*,ReL,Pr)
fNu =
Nu k
hL y
T
f y
= =
∂ ∂
=0 * *
*
(
)
f L
k L h f
Nu = Re ,Pr =
The basic dimensionless parameter correlations developed here will be applied to solve for h (and thus
effectively solve convection problems) using expressions of the form
Nu=ARemPrn, where A, m, and n are
all constants for a given geometry (i.e. fixed dP*/dx*), flow regime, etc.
Such expressions are derived:
(i) From rigorous mathematical
solutions for a limited number of problems (laminar flow)
(ii) From semi-empirical analysis (turbulent flow)
(iii) Correlations of experimental
data (for majority of problems)
We will learn specific correlations for different geometries and flow
conditions (laminar vs. turbulent,
external vs. internal) in Chapters 7-8.
EXAMPLE: Consider a car windshield of length L=800mm, thickness t=6mm, and
k = 1.4W/mK.
The car moves at 111km/h through
ambient air at -15°C. Hot air at 50°C is blown on the inside of the window to
defrost the windshield, a process which requires that the inner surface
temperature of the windshield remain at least at 10°C. If the average convection coefficient on the outside of the
windshield is correlated by the expression
NuL = 0.030ReL0.8 Pr1/3, what is the
smallest value of the internal convection coefficient (hi) which will maintain a
defrosted windshield?