Chapter 8: Convection in External Turbulent Flow: - (A) Mixing Processes
Chapter 8: Convection in External Turbulent Flow: - (A) Mixing Processes
8.1. Introduction
• Turbulent flow is disordered, with random and unsteady velocity fluctuations; hence,
exact predictions cannot be determined.
• Turbulence affects local velocity distribution, drag force, and heat transfer in both
natural and industrial processes.
• Understanding of turbulent flow leads to the ability to make design improvements to
either reduce or enhance turbulent effects
• Our understanding still relies on empirical data and rudimentary conceptual drawings,
and, more recently, computer simulations.
• Exact solutions are not possible.
• Chapter Focus: Wall-bounded shear flows
(a ) 2 = (a ) 2 (8.7c) a a′ = 0 (8.7d)
∂a ∂a
a + b = a +b (8.7g) = (8.7h)
∂x ∂x
∂a ∂a
=0 (8.7i) =0 (8.7j)
∂t ∂t
• An example proving identities (8.7a) and (8.7g) is given.
• Assumptions
• (1) Two-dimensional flow
• (2) Incompressible flow
• (3) Constant properties
• (4) Body forces are neglected
• (5) Flow is, on average, steady-state, so u and v are constant
• Formulation
Equations (2.10x) and (2.10y) become:
⎛ ∂u ∂u ∂u ⎞ ∂p ⎛ ∂ 2u ∂ 2u ⎞
ρ ⎜⎜ +u + v ⎟⎟ = − + μ ⎜⎜ 2 + 2 ⎟⎟ (8.10x)
⎝ ∂t ∂x ∂y ⎠ ∂x ⎝ ∂x ∂y ⎠
⎛ ∂v ∂v ∂v ⎞ ∂p ⎛ ∂2v ∂2v ⎞
ρ ⎜⎜ +u +v ⎟⎟ = − + μ ⎜⎜ 2 + 2 ⎟⎟ (8.10y)
⎝ ∂t ∂x ∂y ⎠ ∂y ⎝ ∂x ∂y ⎠
From the product rule of differentiation, u (∂u / ∂x) and v(∂u / ∂y ) in the x-
momentum equation become:
∂u ∂u 2 ∂u
u = −u (a)
∂x ∂x ∂x
∂u ∂ (u v) ∂v
v = −u (b)
∂y ∂y ∂y
The terms (A) and (B) in (c) above are combined, and by conservation of mass:
⎛ ∂u ∂v ⎞
− u ⎜⎜ + ⎟⎟ = 0 (d)
⎝ ∂x ∂y ⎠
The x-momentum equation reduces to:
⎛ ∂u ∂u 2 ∂ (u v ) ⎞ ∂p ⎛ ∂ 2u ∂ 2u ⎞
ρ ⎜⎜ + + ⎟⎟ = − + μ ⎜⎜ 2 + 2 ⎟⎟ (8.11)
⎝ ∂t ∂x ∂y ⎠ ∂x ⎝ ∂x ∂y ⎠
• Equations (8.12) are identical to equations (8.10) with two notable exceptions:
• The transient terms ∂u / ∂t and ∂v / ∂t disappear.
• More importantly, new terms indicating fluctuating velocity are introduced.
• Equation (8.14) is almost identical to the steady-state Energy equation (8.10) with two
notable exceptions:
• The transient term ρ c p ∂T / ∂t disappears
• Two terms indicating fluctuating velocities and temperature are introduced, and
come out of the convective terms on the left side of the equation.
• y-momentum:
⎛ ∂v ∂v ⎞ ∂p ⎛ ∂2v ∂2v ⎞ ∂u ′v ′ ∂ ( v′) 2
ρ ⎜⎜ u + v ⎟⎟ = − + μ ⎜⎜ 2 + 2 ⎟⎟ − ρ −ρ (8.12y)
⎝ ∂x ∂y ⎠ ∂y ⎝ ∂x ∂y ⎠ ∂x ∂y
• Energy:
⎛ ∂T
ρ c p ⎜⎜ u +v
∂T ⎞ ⎛ ∂ 2T ∂ 2T
⎟⎟ = k ⎜⎜ 2 + 2
⎞
⎟⎟ − ρ c p
( )
∂ u ′T ′
− ρcp
( )
∂ v ′T ′
(8.14)
⎝ ∂x ∂y ⎠ ⎝ ∂x ∂y ⎠ ∂x ∂y
∂ (u ′) 2 (u ′) 2
~ (a)
∂x L
∂ (u ′) 2 ∂u ′v′
<< (8.22)
∂x ∂y
Using the simplifications (8.17) and (8.22), the x-momentum equation for the turbulent
boundary layer reduces to:
⎛ ∂u ∂u ⎞ dp ∂ 2u ∂u ′v′
ρ ⎜⎜ u +v ⎟⎟ = − +μ 2−ρ (8.23)
⎝ ∂x ∂y ⎠ dx ∂y ∂y
(ii) Turbulent Energy Equation
• The derivation for laminar boundary layer equations is again followed.
• Formulation
Scaling arguments for the thermal boundary layer:
x~L (8.16b)
y ~ δt (8.24)
ΔT ~ Ts − T∞ (8.25)
arguments:
There is no preferred direction for the fluctuations, so:
u ′ ~ v′ (8.20)
u ′T ′ ~ v′T ′ (8.27)
The fluctuating terms compare as follows (it is left to the reader to derive this):
∂ (u ′T ′)
<<
( )
∂ v′T ′
(8.28)
∂x ∂y
• Applying simplifications (8.26) and (8.28), the energy equation for the turbulent
boundary layer reduces to:
⎛ ∂T
ρ c p ⎜⎜ u +v
∂T ⎞ ∂ 2T
⎟⎟ = k 2 − ρ c p
( )
∂ v′T ′
(8.29)
⎝ ∂x ∂y ⎠ ∂y ∂y
T ( x,0) = Ts (8.31e)
T ( x, ∞) = T∞ (8.31f)
T (0, y ) = T∞ (8.31g)
• If we know the velocity field outside of the boundary layer, the pressure can be
determined:
• The pressure gradient is expressed as
dp dp∞
= (8.32)
dx dx
• From inviscid flow theory, outside the boundary layer:
dV∞ 1 dp∞
V∞ =− (8.33)
dx ρ dx
• We are left with three equations, (8.8), (8.30) and (8.31), and five unknowns, u , v , T ,
u ′v′ , and v′T ′ ; this is the closure problem of turbulence.
• The two terms u ′v′ and v′T ′ are nonlinear terms
• There is no exact solution to the turbulence boundary layer equations
• Modeling will facilitate numerical and approximate solutions
8.3.4. Eddy Diffusivity
• Reynolds Stress, based on Boussinesq’s hypothesis, is modeled as
∂u
− ρ u ′v′ = ρε M (8.34)
∂y
• ε M is the momentum eddy diffusivity
• ρε M is referred to as the eddy viscosity
• The Reynolds heat flux is modeled as
∂T
− ρ c p v′T ′ = ρ c pε H (8.35)
∂y
• ε H is the thermal eddy diffusivity
• ρ c p ε H is referred to as the eddy conductivity
• The boundary layer momentum and energy equations are written as:
⎛ ∂u ∂u ⎞ ∂ ⎡
ρ ⎜⎜ u +v ⎟⎟ = −
dp
+ ⎢ (μ + ρε M ) ∂u ⎤⎥ (8.36)
⎝ ∂x ∂y ⎠ dx ∂y ⎣ ∂y ⎦
⎛ ∂T
ρ c p ⎜⎜ u +v
∂T ⎞ ∂ ⎡
⎟⎟ = ⎢ (
k + ρ c pε H)∂T ⎤
∂y ⎥⎦
(8.37)
⎝ ∂x ∂y ⎠ ∂y ⎣
• These are simplified by dividing (8.36) through by ρ and (8.37) by ρ c p :
∂ ⎡
u
∂u
+v
∂u
= ⎢ (ν + ε M ) ∂u ⎤⎥ (8.38)
∂x ∂y ∂y ⎣ ∂y ⎦
∂ ⎡
u
∂T
+v
∂T
= ⎢ (α + ε H ) ∂T ⎤⎥ (8.39)
∂x ∂y ∂y ⎣ ∂y ⎦
• The terms in brackets in (8.38) represent the apparent shear stress:
τ app ∂u
= (ν +ε M ) (8.40)
ρ ∂y
• The terms in brackets in (8.39) represent the apparent heat flux:
′′
qapp ∂T
− = (α + ε H ) (8.41)
ρcp ∂y
• The negative sign in (8.41) assigns the correct direction to heat transfer.
• ε M and ε H are properties of the flow and are dependent on the velocity and temperature
fields, respectively
• The number of unknowns has not been reduced; the velocity fluctuation terms have been
replaced with expressions containing different unknowns
• Finding ways to evaluate ε M and ε H is one of the main goals of turbulence research
∂u
u′ ~ A (b)
∂y
• Again assuming that the velocity fluctuations have no preferred direction, u′ ~ v′ , we
have
∂u
v′ ~ A (c)
∂y
• By the above scales, one argues that the scale of the turbulent stress term − u ′v′ is:
2
2 ⎛ ∂u ⎞
− u ′v ′ ~ (u ′)( v ′) ~ A ⎜⎜ ⎟⎟ (d)
⎝ ∂y ⎠
• Therefore, solving equation (8.34) for eddy viscosity gives:
- u ′v′ ∂u
εM = ~ A2 (8.42)
∂u / ∂y ∂y
• The absolute value on the derivative in equation (8.42) ensures the eddy diffusion
remains positive.
• Note that the mixing length itself must still be modeled, and is dependent on the type
of flow.
• Prandtl’s model for mixing length for flow over a flat plat is:
A =κy (8.43)
• where κ is some constant
• This implies that the mixing length approaches zero as y approaches zero.
• Thus equation (8.42) becomes Prandtl’s mixing-length model:
∂u
ε M = κ 2 y2 (8.44)
∂y
• However:
• κ is unknown
• No single value for κ is effective throughout the boundary layer.
• Before developing a suitable model for κ , boundary layer behavior must be further
understood. This is presented next.
+ yu*
y ≡ (8.50)
ν
• Similarly, v + = v / u * , and x + = xu * /ν .
• All of the above variables are dimensionless, and are called wall coordinates.
• Note: Refer to Figure 8.13 to see a plot of data produced by Clauser using these
coordinates.
• The boundary layer extends out to approximately δ ≈ 2000 to 5000
• The plot includes data obtained from flow in a pipe as well as flow along a flat plate,
so the profile appears to be universal.
(iii) Near-Wall Profile: Couette Flow Assumption
• The turbulent boundary layer momentum equation, (8.38), is invoked to develop a model
of the velocity profile near the wall.
• Assumption: Flow is over a flat plate, so dp / dx = 0 .
• The momentum equation is simplified by recognizing that the flow is nearly parallel to
the wall, so v ~ 0 .
• Conservation of mass implies that ∂u / ∂x ~ 0 ; therefore the u component of velocity
does not change significantly along the wall.
• This scaling argument suggests that the convective terms in the momentum equation,
u ∂u / ∂x and v ∂u / ∂y , are each approximately zero, therefore, in (8.38):
∂ ⎡
⎢ (ν + ε M ) ∂u ⎤⎥ ~ 0 near the wall.
∂y ⎣ ∂y ⎦
• The bracketed term is the apparent shear stress, τ app / ρ (Eqn. 8.40)
• The above relationship implies that the apparent stress is approximately constant (with
respect to y), resulting in the Couette Flow Assumption:
τ app ∂u
= (ν +ε M ) ~ constant (8.51)
ρ ∂y
• Recall from Chapter 3: this result is similar to Couette Flow.
• Since the local shear is constant in (8.51), we can replace τ app with its value at the wall,
τo .
• The presence of the eddy diffusivity makes (8.51) different from Couette flow, implying
that the shear stress is constant, but not necessarily linear.
• The Couette Flow Assumption is used to develop a velocity profile at the wall:
• Substituting the definitions of u + and y + into (8.51), it can be shown that:
+
⎛ ε M ⎞ ∂u
⎜1 + ⎟ + =1 (8.52)
⎝ ν ⎠ ∂y
• After rearranging and integrating:
y+
dy +
u+ =
∫(
0
1 + ε M /ν )
(8.53)
• This is a general expression for the universal velocity profile in wall coordinates
• This integral can be evaluated more simply by dividing the boundary layer into two
near-wall regions:
• (1) a region very close to the wall where viscous forces dominate
• (2) a region where turbulent fluctuations dominate
(iv) Viscous Sublayer
• The wall tends to damp out or prevent turbulent fluctuations, so viscous forces dominate
very close to the wall: ν >> ε M .
∂u +
• The Couette Flow Assumption reduces to =1.
∂y +
• Integrating, with boundary condition u + = 0 at y + = 0 yields:
(
u+ = y+, 0 ≤ y+ ≤ 7 ) (8.54)
• This relation compares well to experimental data from y+ ≈ 0 to 7, which is called the
viscous sublayer.
• Equation (8.54) is illustrated in Figure 8.13. Note that the curvature in the plot results
from the semi-logarithmic coordinates.
(v) Fully Turbulent Region: “Law of the Wall”
• Further away from the wall, turbulent fluctuations (i.e., Reynolds stresses) dominate, so
ε M >> ν .
• The Couette Flow Assumption (8.52) therefore becomes:
ε M ∂u +
=1 (8.55)
ν ∂y +
• As before, τ is constant; this is the same value as in the viscous sublayer (which is the
value at the wall), so discontinuity between regions is avoided.
• Substituting wall coordinates into Prandtl’s mixing length theory (8.44) yields:
2 + 2 ∂u +
εM = κ (y ) ν (8.56)
∂y +
• Equation (8.56) is substituted into (8.55):
+ 2
+ 2⎛
⎞
⎜ ∂u
κ ( y ) ⎜ + ⎟⎟ = 1
2
⎝ ∂y ⎠
• Solving the above for the velocity gradient:
∂u + 1
= (8.57)
∂y +
κ y+
• Integrating the above results in the Law of the Wall:
1
u + = ln y + + B (8.58)
κ
• κ is called von Kármán’s constant, and experiments show that κ ≈ 0.41 .
• B is a constant of integration, and is found as follows:
• The viscous sublayer and the Law of the Wall region appear to intersect at roughly
y + = u + ~ 10.8 .
⎦ ∂y
(8.61)
• As y approaches zero, the eddy diffusivity approaches zero, leaving pure viscous
shear.
• Transforming (8.61) into wall coordinates, and solving for ∂u + / ∂y + , yields:
∂u + 2
= (8.62)
∂y +
1 + 1 + 4κ y2 +2
(1 − e − y + / A+
)
2
• For flow over a smooth, flat plate, van Driest used κ = 0.4 and A+ = 26.
• When integrated numerically, Equation (8.62) produces the curve shown in Fig.
8.13.
• D.B. Spalding’s model is commonly used for both flat plates and pipe flow:
⎡ + (κu + ) 2 (κu + ) 3 ⎤
y + = u + + e −κB ⎢eκu − 1 − κu + − − ⎥ (8.63)
⎢⎣ 2 6 ⎥⎦
u+ =
1
κ
( ) ⎡ +
ln 1 + κy + + C ⎢1 − e − y / X −
y + −0.33 y + ⎤
X
e ⎥ (8.64)
⎣⎢ ⎦⎥
• where κ = 0.40, C = 7.8, and X = 11.
(vii) Effect of Pressure Gradient
• The velocity profiles modeled up to this point assume the pressure gradient is zero.
• Figure 8.14 depicts how the velocity profile is affected by a pressure gradient.
• The plot, for flow over a flat plate, shows that in the presence of an adverse pressure
gradient, the velocity profile beyond y + ≈ 350 deviates from the Law of the Wall
model.
• The deviation is referred to as a “wake.”
• The region y + > 350 is commonly referred to as the wake region.
• The region where the data continue to adhere to the Wall Law is called the overlap
region.
• The wake increases with adverse pressure gradient, until separation, where the
velocity profile deviates even from the overlap region.
• A slight wake exists for zero or even a strong favorable pressure gradient, although
the difference between the two sets of data is negligible.
• Wake models are not addressed in this text.
• The Law of the Wall-type models developed earlier model flat plate flow reasonably
well in the presence of zero pressure gradient.
• A favorable pressure gradient is approximately what we encounter in pipe flow; this is
one reason why the models developed here apply as well to pipe flow.
8.4.3. Approximate Solution for Momentum Transfer: Momentum Integral Method
• To obtain drag force on the surface of a body, the momentum integral equation is
invoked, as in Chapter 5.
• In independent works, both Prandtl and von Kármán used this approach to estimate the
friction factor on a flat plate.
(i) Prandtl-von Kármán Model
• Considering a flat, impermeable plate exposed to incompressible, zero-pressure-gradient
flow, the integral momentum equation reduces to equation (5.5):
δ ( x) δ ( x)
∫ ∫
∂u ( x,0) d d
ν = V∞ u dy − u 2 dy (5.5)
∂y dx dx
0 0
• This equation applies to turbulent flows if the behavior of the flow on average is
considered, and the flow properties are interpreted as time-averaged values
• Recall that the integral method requires an estimate for the velocity profile in the
boundary layer.
• Prandtl and von Kármán used Blasius’ model for the shear at the wall of a circular pipe,
based on dimensional analysis and experimental data:
C f ≈ 0.07910 Re D−1 / 4 , (4000 < Re D < 10 5 ) (a)
• Based on the above, the velocity profile in the pipe could be modeled as
1/ 7
u ⎛ y⎞
= ⎜⎜ ⎟⎟ (b)
uCL ⎝ ro ⎠
• where y is the distance from the wall, and u CL is the centerline velocity
• This is the well-known 1/7th Law velocity profile, further discussed in Chapter 9.
• In an external flow a mean velocity is not defined, nor is ro,, so the following
adjustments are made:
• ro is approximated by the edge of the boundary layer δ
• Recall, u CL is modeled as V∞
• Note: Refer to Fig. 8.15 for additional considerations for the boundary layer over a
flat plate.
• Note that, according to this model, the turbulent boundary layer δ /x varies as
Rex−1 / 5 , as does the friction factor C f ; this is contrast to laminar flow, in which δ /x
and C f vary as Rex−1 / 2 .
u 2 ⎛ yV Cf ⎞
= 2.44 ln⎜ ∞ ⎟ + 5 .0
V∞ Cf ⎜ ν 2 ⎟
⎝ ⎠
• Any y value within the wall law layer would satisfy this expression, but a useful
value to choose is the edge of the boundary layer, where u ( y = δ ) = V∞ , so:
1 ⎛ Cf ⎞
= 2.44 ln⎜ Reδ ⎟ + 5 .0 (8.72)
Cf /2 ⎜ 2 ⎟
⎝ ⎠
• where Reδ = V∞δ / ν
• Equation (8.72) relates the skin friction to the boundary layer thickness, and can be used
in the integral momentum equation, though it’s cumbersome.
• By curve-fitting values obtained from (8.72) over a range of values from
Reδ ≈ 10 4 to 10 7 , yielding the approximate relation:
• The above is used to estimate wall shear using the integral method, along with the 1/7th
power law for the velocity profile, resulting in the following solutions to the integral
momentum equation:
δ 0.16
= (8.74)
x Re1x / 7
Cf 0.0135
= (8.75)
2 Re1x / 7
• The above equations replace the less accurate Prandtl-von Kármán correlations; White
recommends these for general use.
• Kestin and Persen used Spalding’s law of the wall for the velocity profile to develop a
more-accurate, but more cumbersome correlation; White simplified their model to:
0.455
Cf = 2
(8.76)
ln (0.06 Re x )
• According to White, the above relation is accurate to Kestin and Persen’s model to
within 1%.
(iii) Total Drag
• Total drag is found by integrating the wall shear along the entire plate.
• Assumptions
• Laminar flow exists along the initial portion of the plate.
• The plate has width w.
• Formulation
• The total drag over the entire plate is:
xcrit L
FD =
∫ (τ )
0
o lam wdx +
∫ (τ )
xcrit
o turb wdx (8.77)
• Dividing by 1
2
ρV∞2 A = 12 ρV∞2 wL , the drag coefficient C D is:
⎡ xcrit L ⎤
∫ ∫
1⎢ ⎥
CD = ⎢ C f ,lam dx + C f ,turb dx ⎥ (8.78)
L⎢ ⎥
⎣ 0 xcrit ⎦
• Substituting Equation (4.48) for laminar flow and using White’s model (8.75) for
turbulent flow, we obtain with some manipulation:
0.0315 1477
CD = − (8.79)
Re1L/ 7 ReL
∂ ⎡
u
∂T
+v
∂T
= ⎢ (α + ε H ) ∂T ⎤⎥ (8.82b)
∂x ∂y ∂y ⎣ ∂y ⎦
• The boundary conditions are
u ( y = 0) = 0 , T ( y = 0) = Ts , (8.83a)
u ( y → ∞ ) = V∞ , T ( y → ∞) = T∞ (8.83b)
• The variables are normalized as follows:
u v T − Ts x y
U= ,V= ,θ= , X = , and Y =
V∞ V∞ T∞ − Ts L L
• Equations (8.82) and boundary conditions (8.83) become:
∂U ∂U 1 ∂ ⎡ ∂U ⎤
U
∂X
+V
∂Y
=
V∞ L ∂Y ⎢⎣(ν + ε M ) ∂Y ⎥⎦ (8.84a)
∂θ ∂θ 1 ∂ ⎡ ∂θ ⎤
U +V = ⎢⎣ (α + ε ) (8.84b)
∂Y ⎥⎦
H
∂x ∂y V∞ L ∂Y
U (Y = 0) = 0 , θ (Y = 0) = 0 (8.85a)
U (Y → ∞) = 1, θ (Y → ∞) = 1 (8.85b)
• Note that normalizing the boundary conditions has made them identical.
• Equations (8.84) can then be made identical if (ν + ε M ) = (α + ε H ) , which is
possible under two conditions:
• The kinematic viscosity and thermal diffusivity are equal:
ν =α (8.86)
• This condition limits the analogy to fluids with Pr = 1
• This also suggests that the velocity and thermal boundary layers are
approximately the same thickness, δ ≈ δ t
• The eddy diffusivities are equal
εM = εH (8.87)
• This is justified by arguing that the same turbulent mechanism—the motion
and interaction of fluid particles—is responsible for both momentum and heat
transfer; Reynolds essentially made this argument, so equation (8.87) is
sometimes referred to as the Reynolds Analogy.
• This also means that the turbulent Prandtl number Prt is equal to 1
• The analogy is now complete; the normalized velocity and temperature profiles,
U ( X , Y ) and θ ( X , Y ) are equal.
• By imposing conditions (8.86) and (8.87), the terms in parenthesis cancel and, after
substituting the dimensionless variables into (8.88):
′′
qapp c p (Ts − T∞ )∂θ / ∂Y
= (8.89)
τ app V∞ ∂U / ∂Y
• Since the dimensionless velocity and temperature profiles are identical, their
derivates cancel
• ′′ / τ app is constant throughout the boundary layer, so this ratio can be
The ratio qapp
represented by the same ratio at the wall , and equation (8.89) becomes:
qo′′ c p (Ts − T∞ )
=
τo V∞
• This can be recast into a more convenient form by substituting qo′′ = h(Ts − T∞ ) and
τ o = 12 C f ρV∞2 into the above, and rearranging:
h Cf
=
ρV∞ c p 2
• The terms on the left side can also be written in terms of the Reynolds, Nusselt, and
Prandtl numbers
Nu x Cf
St x ≡ = (8.90)
Rex Pr 2
• where St x is called the Stanton number
• Equation (8.90) is commonly referred to as the Reynolds Analogy; it can also be
derived for laminar flow over a flat plate for Pr =1.
• The Reynolds Analogy is limited to Pr = 1 fluids; it is appropriate for gases, but not for
most liquids.
(ii) Prandtl-Taylor Analogy
• The Reynolds analogy doesn’t account for the varying intensity of molecular and
turbulent diffusion in the boundary layer
• Very close to the wall, molecular forces are expected to dominate:
ν >> ε M , and α >> ε H (8.91)
• Further away from the wall, turbulent effects dominate:
ε M >> ν , and ε H >> α (8.92)
• Neither condition above restricts us to Pr = 1 fluids.
• Prandtl and Taylor independently divided the boundary layer into two regions:
• A viscous sublayer where molecular effects (8.91) dominate
• A turbulent outer layer, where (8.92) is assumed to hold
• In order for an analogy to exist, the momentum and boundary layer equations, and their
boundary conditions, must be identical in both regions, so:
• The viscous sublayer is defined as the portion of the boundary layer beneath y = y1,
where y1 is some threshold value, with boundary conditions:
u (0) = 0, T (0) = Ts ,
u ( y1 ) = u1 , T ( y1 ) = T1
• The following normalized variables make the boundary conditions and equation
(8.82) identical:
u v T − Ts x y
U= , V = ,θ= , X = and Y = .
u1 u1 T1 − Ts y1 y1
• For the viscous sublayer, the ratio of the apparent heat flux and apparent shear stress
(Eqn. 8.86) leads to the following:
qo′′
Ts − T1 = Pr u1 (8.93)
τo cp
• The outer layer closely resembles the Reynolds Analogy, with ε M = ε H (or Prt = 1 ),
but this time we assume that the turbulent effects outweigh the molecular effects; this
region has boundary conditions:
u ( y1 ) = u1 , T ( y1 ) = T1 ,
u ( y → ∞ ) = V∞ , T ( y → ∞) = T∞
• The following normalizing variables make the analogy valid in this region:
u − u1 v - u1 T − T1 x y
U= , V= ,θ= , X = , and Y =
V∞ − u1 V∞ − u1 T∞ − T1 L L
• For the outer region, the ratio of the apparent heat flux and apparent shear stress
(equation 8.86) leads to:
qo′′
T1 − T∞ = (V − u ) (8.94)
τocp ∞ 1
qo′′ Cf /2
St = =
ρV∞ c p (Ts − T∞ ) ⎡ u1 ⎤
⎢V ( Pr − 1) + 1⎥
⎣ ∞ ⎦
• The velocity at the edge of the viscous sublayer, u1 , is estimated using the universal
velocity profile (Fig. 8.13); a value of u + = y + ≈ 5 is chosen to approximate the edge
of the viscous sublayer.
• From the definition of u + :
u1 2
u+ = 5 = , or
V∞ Cf
u1 Cf
=5 (8.95)
V∞ 2
• Thus the Prandtl-Taylor analogy is:
Nu x Cf /2
St x ≡ = (8.96)
Rex Pr ⎡ C f ⎤
⎢5 ( Pr − 1) + 1⎥
⎢⎣ 2 ⎥⎦
(iii) von Kármán Analogy
• Theodore von Kármán extended the Reynolds analogy even further to include a third
layer – a buffer layer – between the viscous sublayer and outer layer:
Nu x Cf /2
St x ≡ = (8.97)
Rex Pr Cf ⎧ ⎡ 5 Pr + 1⎤ ⎫
1+ 5 ⎨( Pr − 1) + ln ⎢ ⎬
2 ⎩ ⎣ 6 ⎥⎦ ⎭
• Note: Refer to Appendix D for development.
(iv) Colburn Analogy
• Colburn proposed a purely empirical modification to the Reynolds analogy that accounts
for fluids with varying Prandtl number, using an empirical fit of available experimental
data:
Cf
St x Pr 2 / 3 = (8.98)
2
• The exponent (2/3) is entirely empirical
• The Colburn Analogy yields acceptable results for Rex < 107 (including the laminar flow
regime) and Prandtl number ranging from about 0.5 to 60.
• An example, Example 8.3, is presented in which the average Nusselt number for heat
transfer along a flat plate of length L with constant surface temperature is determined,
using White’s model for turbulent friction factor and assuming the flow over the plate
has an initial laminar region.
• Equations developed within the example:
• The average heat transfer coefficient from Equation (2.50) is split into laminar and
turbulent regions:
⎡ xc L ⎤
∫ ∫
1⎢ ⎥
hL = ⎢ hlam ( x)dx + hturb ( x)dx ⎥ (8.99)
L⎢ ⎥
⎣0 xc ⎦
• After several manipulations, substitutions and assumptions, the equation for the
average Nusselt number for heat transfer along a flat plate of length L with constant
surface temperature and initial laminar region is:
( )
Nu L = 0.0158 ReL6 / 7 − 739 Pr1/ 3 (8.100)
• If the laminar region had been neglected, the above would be:
Nu L = 0.0158 ReL6 / 7 Pr1 / 3 (8.101)
+ ρ c p u*
T ≡ (Ts − T ) (8.105)
qo′′
• The above is cast into a simpler form:
∂T + ν
= (8.106)
∂y + (α + ε H )
• Though the above can now be integrated, ε H is unknown, but substituting ε M into
(8.106) by using the definition of the turbulent Prandtl number (8.81) yields:
y+
ν dy +
T+ =
∫
0
α + εH
(8.107)
∫
ν
T + = T1+ + dy + (8.109)
εH
y1+
2 + 2 ∂u +
εM = κ (y ) ν (8.110)
∂y +
• The partial derivative ∂u + / ∂y + can be found from the Law of the Wall
• Substituting the above and (8.58) into (8.109) yields:
y+
∫κ
Prt
T+ = +
dy + (8.111)
y
y1+
T+ =
Prt⎛ y+ ⎞
ln⎜ + ⎟ , y + > y1+
κ ⎜⎝ y1 ⎟⎠
( ) (8.112)
• The temperature profile defined by Equations (8.108) and (8.112) depends on the fluid
(Pr), as well as the parameters Prt and κ .
• Kays et al. assumed Prt = 0.85 and κ = 0.41, but found that the thickness of the
conduction sublayer ( y1+ ) varies by fluid
• White reports a correlation that can be used for any fluid with Pr ≥ 0.7:
Prt
T+ = ln y + + 13 Pr 2 / 3 − 7 (8.113)
κ
• In this model, Prt is assumed to be approximately 0.9 or 1.0
• Figure 8.18 is a plot of this model, along with viscous sublayer, for various values of
Pr
• Note that the temperature profile increases with increasing Prandtl number.
(iv) A 1/7th Law for Temperature
• A simpler 1/7th power law relation is sometimes used for the temperature profile:
1/ 7
T − Ts ⎛ y ⎞
=⎜ ⎟ (8.114)
T∞ − Ts ⎜⎝ δ t ⎟⎠
ρ c p u* ρ c p V∞ C f / 2
T∞+ = (Ts − T∞ ) = (Ts − T∞ ) (8.116)
qo′′ qo′′
• where (8.47) was substituted for the friction velocity u *
• Substituting (8.116) into (8.115) for (Ts − T∞ ) and rearranging yields:
ρ c pV∞ C f / 2 x
Nu x =
T∞+ k
• Then, multiplying the numerator and denominator by ν yields:
Rex Pr C f / 2
Nu x = (8.117)
T∞+
• The universal temperature profile, Equation (8.113),is used to evaluate T∞+ :
Prt
T∞+ = ln y∞+ + 13Pr 2 / 3 − 7 (8.118)
κ
• A precise value for y∞+ is not easy to determine, but a clever substitution is made by
using the Law of the Wall velocity profile.
• Equation (8.58) is evaluated in the free stream as:
1
u∞+ = ln y∞+ + B (8.119)
κ
• Substituting (8.119) into (8.118) for ln y ∞+ , the Nusselt number relation then becomes
Rex Pr C f / 2
Nu x =
Prt (u∞+ − B) + 13Pr 2 / 3 − 7
• The above is simplified using the definition of Stanton number, St = Nu x /( Rex Pr ) ,
selecting B = 5.0 and Prt = 0.9, and noting that the definition of u + leads to
u ∞+ = 2 / C f
Cf /2
St x =
0.9 + 13 Pr ( 2/3
)
− 0.88 C f / 2
(8.120)
• Note how similar this result is to the more advanced momentum-heat transfer analogies,
particularly those by Prandtl and Taylor (8.96) and von Kármán (8.97).
8.5.5. Integral Methods for Heat Transfer Coefficient
• The universal temperature profile allows us to model heat transfer using the integral
energy equation.
• A case of turbulent flow over a flat plate where a portion of the leading surface is
unheated is examined, illustrated in Figure 8.19.
• The simplest solution is to assume the 1/7th power law for both the velocity and
temperature profiles
• This is mathematically cumbersome, and is developed in Appendix E.
• The result of the analysis is:
−1 / 9
Nu x C f ⎡ ⎛ xo ⎞9 / 10 ⎤
St x ≡ = ⎢1 − ⎜ ⎟ ⎥ (8.121)
Rex Pr 2 ⎢⎣ ⎝ x ⎠ ⎦⎥
• where xo is the unheated starting length
• Note that Equation (8.121) reduced to the Reynolds Analogy when xo = 0, (the Prandtl
number in that derivation was assumed to be 1).
• The model has been used to approximate heat transfer for other fluids:
• Equation (8.121) is expressed as:
Nu xo = 0
Nu x = (8.122)
[1 − (x / x) ]
o
9 / 10 1 / 9
• where Nu xo = 0 represents the heat transfer in the limit of zero insulated starting
length
• In this form, other models for heat transfer, like von Kármán’s analogy, could be
used to approximate Nu xo = 0 for Pr ≠ 1 fluids.
St =
Cf
2
[Pr + C (k
t
+ 0.2
s ) Pr
0.44
Cf /2 ]
−1
(8.123)