Sie sind auf Seite 1von 27

Chapter 3.

Heat/Mass Transfer Analogy


- Laminar Boundary Layer

As noted in the previous chapter, the analogous behaviors of heat and mass transfer have
been long recognized. In the field of gas turbine heat transfer, several experimental
studies have been done with mass transfer because of its experimental advantages. In
such cases, it was required that the heat/mass transfer analogy function be well known.
This chapter reviews the conventional heat/mass transfer analogy function and
numerically generates the analogy functions to examine their behaviors under the
influences of streamwise pressure gradients and different thermal boundary conditions
(BCs). For practical reasons, the fluid assumed in the heat transfer is air (Pr = 0.707) and
that of the mass transfer is air with naphthalene being the substance to be transported (Sc
= 2.28). Furthermore, since a typical use of heat/mass transfer analogy is to convert mass
transfer data to heat transfer data, discussions are focused mainly on such a direction of
the conversion.

3.1 Analytical Solutions

3.1.1 Conventional relationship

20
The conventional relationship for laminar boundary layer flow is given by the following
simple expression.
1
Nu  Pr  3
=  (3.1.1)
Sh  Sc 
The conditions under which this equation is applicable are (1) no streamwise pressure
gradient (plane wall, also called flat plate), (2) the same Reynolds number and (3)
uniform temperature BCs. Note that this analogy function does not depend on
streamwise position; in other words, a single value can be applied to all streamwise
positions, provided that the flow is laminar. For this reason, the analogy functions are
sometimes called the analogy factor in the present study, referring to the ratio of Nusselt
number to Sherwood number. This section discusses how this analogy function comes
about.
Heat and mass transfer processes are governed by the partial differential
equations: energy (or temperature) and mass transfer equations, respectively. For a
steady-state and flow, two-dimensional, laminar, incompressible boundary layer, they
reduce to the following expressions.
∂T ∂T ∂ 2T
u +v =α 2 (3.1.2)
∂x ∂y ∂y

∂m ∂m ∂ 2m
u +v =D 2 (3.1.3)
∂x ∂y ∂y
Since the velocity parameters, u and v, are involved, the continuity equation and the
momentum equation of the boundary layer must be introduced.
∂u ∂v
+ =0 (3.1.4)
∂x ∂y

∂u ∂u ∂ 2u
u +v =ν 2 (3.1.5)
∂x ∂y ∂y
The above momentum equation assumes no streamwise pressure gradient, incompressible
flow and constant properties. Now we have four equations and four unknowns, u, v, T
and m. Comparing Eqns. (3.1.2) and (3.1.3), when α = D, the temperature and the mass

21
transfer equations are identical. Upon nondimensionalizing Eqns. (3.1.2) and (3.1.3), the
thermal diffussivity, α, and the diffusion coefficient, D, are replaced by the dimensionless
transport properties: Prandtl number, Pr, and Schmidt number, Sc. They are defined by:
ν ν
Pr = ; Sc = (3.1.6 and 7)
α D

Solution methods are discussed in Eckert and Drake (1987) and Kays and
Crawford (1993). The resulting Nusselt number, Nu, distribution for a laminar boundary
layer as a function of streamwise position is well approximated as follows.
Nu x = 0.332 Re x
1/ 2
Pr 1 / 3 (3.1.8)
Similarly, for the mass transfer equation, the Sherwood number, Sh, distribution yields:
Sh x = 0.332 Re x
1/ 2
Sc 1 / 3 (3.1.9)
Equation (3.1.1) can now be obtained by dividing Eqn. (3.1.8) by Eqn. (3.1.9),
assuming the same x-Reynolds number. As is obvious from Eqns. (3.1.2) and (3.1.3), Nu
is equal to Sh when Pr is equal to Sc. Furthermore, in the case of air (Pr = 0.707) and
naphthalene (Sc = 2.28), the analogy function becomes:
Nu
= 0.677 (3.1.10)
Sh
As noted earlier, this relation assumes the same Reynolds number, the uniform-level BCs
(uniform wall temperature for the heat transfer and uniform mass concentration for the
mass transfer) and no streamwise pressure gradients. Effects of these assumptions are
evaluated in the next section.

3.1.2 Analytical solution to the cross-boundary-condition analogy

The conventional relationship was derived assuming both processes having the identical
level boundary condition (BC). This relation, Eqn. (3.1.1), may be sufficient in gas
turbine heat transfer studies as shown by the numerical simulations in the next section.
When results from a mass transfer experiment are used to estimate heat transfer situations
in an actual gas turbine engine, the conversion is from uniform-level-BC mass-transfer
22
data to uniform-level-BC heat-transfer data, for which Eqn. (3.1.1) is most suited. The
gas turbine heat transfer situation is best simulated with a constant wall temperature
boundary condition. This is because variations in temperature at the airfoil surface are
relatively small compared to the difference between the surface and freestream
temperatures (e.g. pp.268-70 in Boyce, 1982). The next few sections show that this
relation is quite accurate for a surface with non-zero streamwise pressure gradients.
Another analytical solution is easy to derive. This is for the cases in which
boundary conditions do not match between the two processes: a uniform level for one and
a uniform flux for the other. Here, the relation is first derived, and discussions are made
on its meaning and possible applications.
Assuming a flat plate and laminar boundary layer, the distribution of Nusselt
number for a uniform heat flux BC is well approximated by the following (Kays and
Crawford 1993).
Nu x = 0.453 Re x
1/ 2
Pr 1 / 3 (3.1.11)
Dividing this by Eqn. (3.1.9) yields the analogy function for the cross-BC processes.
1
Nu  Pr  3
= 1.36  (3.1.12)
Sh  Sc 
Corresponding to Eqn. (3.1.10), the above expression, for air and naphthalene, results in:
Nu
= 0.924 (3.1.13)
Sh
Again, this relationship is not a function of Reynolds number or streamwise position.
Noteworthy is that the value for the cross-BC processes, Eqn. (3.1.13), differs 36% from
the value for the conventional relationship, Eqn. (3.1.10).
The first discussion focuses on a significance of these results. Because the
governing partial differential equations are essentially identical for mass and heat transfer,
Eqn. (3.1.12) can, for example, be written for two processes, both being heat transfer:
1
Nu 6  Pr  3
= 1.36 6  (3.1.14)
Nu 5  Pr5 

23
Here, the subscripts 6 and 5 designate the two different fluids and BCs; fluid 6 is at a
uniform flux BC, and fluid 5 is at a uniform level BC (corresponding to designations
introduced later, see Table 3.1). When these are the same fluids, the ratio of Nusselt
numbers becomes 1.36. This means that the heat transfer coefficient at any location on a
flat plate with uniform heat flux BC is always 36% greater than that of the plate with the
same Reynolds number (same location if ν5 = ν6) with a uniform temperature BC. This
can be realized by comparing Eqns. (3.1.1) and (3.1.12) since the denominators have the
same processes in the both equations.
One possible application of this analogy function is to convert data from a heat
transfer experiment with a uniform flux BC to uniform-temperature BC data. This would
be useful especially since many heat transfer experiments (such as steady state measure-
ments with liquid crystals) are done with uniform heat flux BCs. However, as shown in
the next section, streamwise pressure gradients have a greater effect on Eqn. (3.1.14) than
on Eqn. (3.1.1); the cross-BC functions for an airfoil profile deviate from Eqn. (3.1.14)
more than the matching-BC functions for an airfoil profile deviate from Eqn. (3.1.1). For
this type of an application, it is necessary to study how pressure gradients affect the
analogy functions.
Another possible case in which the cross-BC analogy function, Eqn. (3.1.12), may
be useful is when a heat transfer experiment with a uniform heat flux BC is directly
compared with a mass transfer experiment with a uniform mass concentration BC in
order to determine effects of streamwise pressure gradients. Because of the identical
governing PDEs, the mass transfer experiment is representative of both mass and heat
transfer processes with a uniform level BC. The results of the comparison should be
expressed as magnitudes of deviations from Eqn. (3.1.12). It will be shown that
converting the mass transfer data into the heat transfer data with both at a uniform level
BC should not be a problem even for a case with streamwise pressure gradients by the use
of the conventional relationship, Eqn. (3.1.1). Once the generalized effect of pressure
gradients is determined, this relationship may be applied when boundary conditions do
not match.

24
3.2 Analogy Functions on the CF6 Profile

3.2.1 Setup of cases

This chapter uses the numerical boundary layer code, TEXSTAN, to computationally
determine Nusselt number distributions for various flow and boundary conditions (BCs).
Heat/mass transfer analogy functions are obtained from numerically evaluated Nu and Sh
distributions under the flow and BCs of interest.
In this section, the streamwise pressure (or velocity) gradient of the General
Electric CF6 engine first-stage vane is implemented into the simulations, and results are
compared to the corresponding zero-pressure-gradient case. Also varied are BCs in order
to examine the effects on the analogy function. The primary reason that the different BCs
are tested is that a mismatch in BCs is frequent when heat transfer rates are
experimentally determined from mass transfer experiments. Many heat transfer
experiments are done with the BC of a uniform heat flux. If data from such experiments
are used to calculate the analogy function, consistency with Eqn. (3.1.1) would be lost,
for this equation was derived assuming uniform level BCs in both the mass and heat
transfer experiments. Therefore, it is important that one knows the effects of different
combinations of BCs.
Figure 3.1 shows the velocity profiles of the CF6 and the GE90 profiles on the
suction surfaces. The GE90 profile is used in the next section. The suction side profiles
are shown here, and all the analyses focus on the suction side. This is because suction
sides usually are of more interest, for they are the ones subjected to influences such as
flow separation, transition and three-dimensional secondary flows.
When the input data files to TEXSTAN are prepared, all possible combinations of
the three factors are sought, except cases of mass transfer with flux boundary conditions.
This accumulates the number of cases considered in this section to six, and they are
summarized in Table 3.1.

25
1.4
1.2
1.0
0.8
Us/Uex

0.6

0.4 CF6 profile


GE90 profile
0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
x/C

Figure 3.1 Velocity profiles of the CF6 and GE90 blades, normalized on exit velocity.
The CF6 from Chen (1988) and the GE90 from Wang (1997).

Table 3.1 Summary of cases for the CF6 profile

Case CF6-1 CF6-2 CF6-3 CF6-4 CF6-5 CF6-6


Us=constant
Profile CF6
“plane wall”
Sc or Pr 2.28 0.707 2.28 0.707

B.C. Tw qw Tw qw

26
The following is a description of each of the cases simulated in this section. This
list can be applied also to the cases simulated in the next section with the GE90 profile.

Case CF6-1 --- simulation under the same conditions as Chen (1988; and Chen
and Goldstein, 1992), an experiment with the CF6 profile, a uniform mass
concentration and a naphthalene/air mixture being the fluid
Case CF6-2 --- case CF6-1 converted by Eqn. (3.1.1) to heat transfer to air
Case CF6-3 --- simulation under the same conditions as Chung (1992; and Chung
and Simon, 1993), an experiment with the CF6 profile, a uniform wall heat
flux and air being the fluid
Case CF6-4 --- the CF6-1 case with the pressure profile replaced with a uniform
velocity, a flat plate case
Case CF6-5 --- case CF6-4 converted by Eqn. (3.1.1) to heat transfer to air
Case CF6-6 --- Case CF6-3 with the pressure profile replaced with a uniform
velocity of a flat plate case

Two other possible cases, processes with uniform mass flux BCs, were not included
because, when a mass transfer experiment is done with naphthalene, the BC is of a
uniform mass concentration (a uniform level, equivalent to uniform temperature), and
usually, the conversion is performed from mass transfer to heat transfer. Thus, the cases
with uniform mass flux are not important.
Since two of the above cases correspond to heat transfer processes and the others
mass transfer, there are 8 combinations for which to produce some kind of analogy
functions. However, only the practical cases are analyzed. They are described below. As
the original analogy function is expressed in terms of the ratio, Nu/Sh, the numerators
below correspond to heat transfer processes and the denominators to mass transfer.

CF6-5/CF6-4 --- reproducing Eqn. (3.1.1) with the same boundary conditions.
CF6-2/CF6-1 --- same BCs as CF6-5/CF6-4 but with the CF6 pressure profile

27
CF6-3/CF6-1 --- the conditions used to compare the Chen’s and the Chung’s
experiments as done in the next chapter
CF6-6/CF6-4 --- same as CF6-3/CF6-1 but with absence of pressure gradient
effects on both the mass and the heat transfer processes.
CF6-4/CF6-1 --- both mass transfer processes. This shows effects of the
streamwise pressure gradient.
CF6-2/CF6-3 --- both heat transfer processes. This shows effects of the shift in
BCs. If this were a simple function, one could easily convert the data from
an experiment with a uniform heat flux BC to those of a uniform
temperature BC, the latter being more representative of actual engine
conditions.

For each of the above six different cases, an input data file to TEXSTAN was
prepared. They are presented in Appendix A. To introduce the computational code used
in the present study, TEXSTAN was originally developed as the STAN5 program by
Crawford and Kays (1976). The reference here provides theoretical backgrounds and
instructions for STAN5, and basically the same operations apply to TEXSTAN. These
programs solve transport equations such as momentum, energy and mass for two-
dimensional, internal and external boundary layer flows. The user of these programs is to
prepare an input file that lists all necessary parameters such as boundary conditions and
physical properties and directions for the outputs. The user solves the momentum
equation and, if desired, heat and/or mass transport equations.

3.2.2 Results and discussions

Figure 3.2 presents the variations of heat/mass transfer analogy functions versus
streamwise positions. In a numerical simulation of a boundary layer, heat (or mass)
transfer rates are calculated based on temperature (or mass concentration) gradients at the
surface. The calculation cannot proceed after the flow separates, at which point the
velocity gradient becomes zero. This is what happens to cases 3/1 and 2/1 near x/C of

28
0.95. Looking at the velocity profile on Fig. 3.1, one finds that the flow separates just
after the mainstream flow starts to decelerate at x/C of 0.8.
There are two main trends seen in Fig. 3.2. Both the functions 2/1 and 5/4 are
essentially the same, as Eqn. (3.1.1) implies, the Fref value of Fig. 3.2, while functions 3/1
and 6/4 differ.
First of all, looking at the functions that are consistent with Eqn. (3.1.1), one
realizes that each pair consists of the same pressure profile (either the CF6 or flat plate)
and the same level BC (temperature and mass concentration). Also, the pair 2/1 is only
very weakly dependent on the non-zero pressure gradient. At this point, it can be
concluded that using Eqn. (3.1.1) to convert mass transfer data would yield a perfect
estimate of flat-plate heat transfer with uniform temperature BC if the mass transfer were
done with a flat plate. Similarly, it would yield an almost perfect estimate of heat transfer
with the CF6 profile and uniform temperature BC if the mass transfer were also done
with the CF6 profile. In other words, a single value from Eqn. (3.1.1) based only on Pr
and Sc can be used to convert Sh to Nu at any and all streamwise locations, assuming that
the heat transfer process has the same pressure profile, the same Re and a uniform-
temperature BC.
Second of all, the other two lines presented in Fig. 3.2, 3/1 and 6/4, form another
group; the 3/1 line appears to vary in the vicinity of the 6/4 line. Referring to Table 3.1,
one realizes that each line consists of heat transfer with a uniform flux BC and mass
transfer with a uniform level BC – cross-BC analogy functions. To define the cross-BC
analogy functions, I would like to refer to heat transfer with a uniform flux BC and mass
transfer with a uniform level. Since a mass transfer experiment with naphthalene
provides a uniform level BC, this is often the case in converting mass transfer data to get
heat transfer coefficients using the analogy functions. Contrary to the behavior when
numerator and denominator had level BC’s, cross-BC conversions cannot be
approximated by Eqn. (3.1.1) or Eqn. (3.1.12). Moreover, line 3/1 (both on the CF6
profile) is not as close to the line 6/4 (both on flat plate) as line2/1 (both on the CF6
profile) is to the line 5/4 (both on flat plate). Therefore, even when the line 6/4 is

29
analytically derived, it is not a representative of cross-BC analogy functions for cases of
non-zero pressure gradient. The importance of this lies in the fact that many heat transfer
experiments are done with a uniform flux (a uniform heat flux).

1.2
6/4
3/1
1.0

0.8 2/1 Fref


Nu/Sh

0.6

0.4 5/4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
x/C

Figure 3.2 Analogy functions for the CF6 profile and the corresponding flat plate cases.
Rec=171,000 and P/C=0.77 (x/C=1.25 at the trailing edge). Fref corresponds to the value
from Eqn. (3.1.1). Case designations are simplified such as 3/1 for CF6-3/CF6-1.

3.3 Analogy Functions on the GE90 Profile

3.3.1 Setup of cases

In this section, distributions of analogy functions on the GE90 profile and the
corresponding flat plate cases are calculated, using the TEXSTAN program, and
discussed. The velocity profile of the GE90 blade is provided in Fig. 3.1, and the cases of
the computations are summarized in Table 3.2. The GE90 is a newer engine than the CF6
from General Electric, and its profile has a long-lasting acceleration region compared to
the CF6 profile, almost one half of the entire surface length. Table 3.2 is essentially the

30
same as Table 3.1, except that, where non-zero velocity profiles are implemented, the
GE90 replaces the CF6 profile. Therefore, the GE90-1 case and the CF6-1 case, for
instance, are under the same conditions except for the velocity profiles.
The case descriptions in the previous section can still be used, and all the six pairs
of analogy functions that were generated in the CF6 case are regenerated for the GE90
case.

Table 3.2 Summary of cases for the GE90 profile

Case No. Nu or Sh (profile#, fluid, BC )


GE90-1 Sh ( GE90, Sc=2.28, Tw )
GE90-2 Nu (GE90, Pr=0.707, Tw )
GE90-3 Nu (GE90, Pr=0.707, q” )
GE90-4 Sh (f.p.*, Sc=2.28, Tw )
GE90-5 Nu ( f.p.*, Pr=0.707, Tw )
GE90-6 Nu (f.p.*, Pr=0.707, q” )
# free-stream velocity profile
* flat plate profile (uniform free-stream velocity)

3.3.2 Results corresponding to the Wang’s experiment

Figure 3.3 presents four of the six analogy functions generated in this section, and the
other two are discussed in the next section. The cases with the non-zero pressure gradient
again indicate flow separation near x/C of 1.24. By this point, the flow has remained
attached for a quite a long distance against adverse pressure gradients downstream of the
beginning of deceleration near x/C of 0.7. This is remarkably longer than in the CF6 case
in which flow separates just after the pressure gradient sign reverses. This may be

31
because of differences in the degree of deceleration and thinner boundary layers in the
case of the GE90, due to the prolonged strong acceleration.
Now, the results in Fig. 3.3 look quite similar to those in Fig. 3.2: lines 2/1 and
5/4 forming one group and lines 3/1 and 6/4 forming the other. The fist group behaves in
the same way as the one in the CF6 profile cases. The flat plate pair, 5/4, again confirms
the validity of Eqn. (3.1.1), and, together with the line 5/4 on Fig. 3.2, it is shown that the
reference value, Fref = 0.66, is not a function of Reynolds number. The GE90 profile pair,
2/1, varies only a little from the reference value, Fref, compared to lines, 3/1 and 6/4.
The other group of lines, 6/4 and 3/1, has the cross-BC analogy functions. The
flat plate function, 6/4, records the value 0.92, the same as the one from the line 6/4 in
Fig. 3.2. This shows that the analogy function for cross-BC cases with the absence of the
pressure gradient may not depend on Reynolds number. A simple relation such as Eqn.
(3.1.12) may be possible for cross-BC cases.
The deviations of lines 3/1 and 2/1 from the lines 6/4 and 5/4, respectively, in Fig.
3.3 seem greater than the corresponding deviations in Fig. 3.2 although line 2/1 still
deviates from Fref only slightly. Figures 3.2 and 3.3 are not enough to conclude that the
GE90 profile produces larger errors from the corresponding flat plate cases than does the
CF6 profile. This is because the GE90 cases are simulated at a greater chord Reynolds
number. The next section discusses whether these deviations are actually due to the
greater Reynolds number or the difference in the pressure profiles.

32
1.4
3/1
1.2
6/4
1.0
2/1
0.8
Nu/Sh

Fref
0.6
5/4
0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
x/C

Figure 3.3 Analogy functions, Nu/Sh, for the GE90 profile and the corresponding flat
plate cases. Rec=540,000 and P/C=0.75 (x/C=1.38 at the trailing edge). Fref corresponds
to the value from Eqn. (3.1.1). Case designations are simplified, such as 3/1 for GE90-
3/GE90-1

3.3.3 Results corresponding to the Chen’s chord Reynolds number

This section examines the cases with the GE90 profile to which Chen’s chord-Reynolds
number is applied, i.e. the Reynolds number used in the section 3.2. The resulting
analogy functions are compared with those in the section 3.2 (the results shown in Fig.
3.3 could not be compared with Fig. 3.2 because of the difference in the Reynolds
numbers). The results of this section are presented in Fig. 3.4, and comparing with Fig.
3.2, will reveal effects of different pressure profiles.
Any difference in the distributions shown in Fig. 3.4 from those in Fig. 3.3 is hard
to distinguish. In fact, the two figures look exactly the same. This is expected from the
discussions in section 3.1. The effects of Reynolds numbers cancel each other when the
numerator and the denominator have the same profile.

33
1.2 3/1
6/4
1.0

0.8 2/1
Nu/Sh

0.6
5/4
0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
x/C

Figure 3.4 Analogy functions, Nu/Sh, for the GE90 profile and the corresponding flat
plate cases. Rec=171,000 and P/C=0.77 (x/C=1.38 at the trailing edge). Fref corresponds
to the value from Eqn. (3.1.1). Case designations are simplified such as 3/1 for GE90-
3/GE90-1

34
3.4 Heat/Heat and Mass/Mass Analogy Functions

3.4.1 Origins of non-uniform analogy functions

In the previous sections, all analyses were based on analogy functions between a fluid
with Pr=0.707 (representing air) and a fluid with Sc=2.28 (representing naphthalene), and
it was demonstrated that some pairs of processes resulted in rather uniform analogy
functions and others do not. This section attempts to further examine different effects
that form a non-uniform distribution of the analogy function. One of the most significant
effects is the difference in the thermal boundary conditions (BCs) as discussed in the
previous section. The first step here is to examine the effect of pressure profiles in cases
with a transport property of 2.28. The second step is to examine the effect of thermal
boundary conditions in cases with a transport property of 0.707.
Obviously, a uniform distribution of the analogy function is advantageous, so it is
important to know under what circumstances the analogy function is not uniform and how
it may behave in such cases. When the function is uniform over a surface such as lines
6/4 and 5/4 in Figs. 3.2, 3.3 and 3.4, a single factor can be applied to the entire surface in
order to convert Sherwood numbers to Nusselt numbers, for example. When the function
is not uniform, e.g. line 3/1 in Figs. 3.2, 3.3 and 3.4, the conversion factor is dependent on
locations on the surface. Even if the variations along the surface cannot possibly be
predicted by knowing the effects of different causes, one should be able to determine
when the analogy function may behave with large deviations.
Pressure profiles may be thought to heavily affect uniformity of the analogy
function distribution. However, this is not always true. For example, line 3/1 in Figs.
3.2, 3.3 and 3.4 is a case with non-zero pressure gradient profiles and deviates from line
6/4 which is the identical case, except for the zero pressure gradient profiles. This large
deviation is not observed between the pairs 2/1 and 5/4. Therefore, the deviation of line
3/1 has a combined effect of thermal BCs and pressure profiles. Pair 4/1 is chosen to
examine the effects of pressure profiles since it is an intermediate case between 3/1 and

35
6/4. Pair 2/3 is chosen to examine the effects of thermal boundary conditions since it may
become useful in converting flux-BC experimental data to level-BC data.

3.4.2 The effects of pressure profiles in mass/mass transfer analogy: line 4/1

Since the transport property of 2.28 was assumed to represent Sc of naphthalene, pair 4/1
might be called a mass/mass analogy. Line 4/1 in Figs 3.5, 3.6 and 3.7 is discussed here
as an example of a pair with level-BCs and identical transport properties. Looking at
each figure, the shape looks exactly like line 3/1 in the previous figures, except near the
leading edge. The resulting distributions clearly show that the trends are similar to lines
3/1 and 2/1 in Figs. 3.2, 3.3 and 3.4, except for the magnitudes of deviation. In other
words, when line 3/1 has a positive slope, line 2/1 also has one. Therefore, it can be
concluded that the shapes of lines 3/1 and 2/1 are due to the pressure gradients, and the
BCs are mainly responsible for the magnitudes of the deviations from the corresponding
plane wall cases.

1.4

1.2
4/1
1.0

0.8
Nu/Sh

0.6
2/3
0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
x/C

Figure 3.5 CF6 Chen cases,


Rec=171,000 and P/C=0.77 (x/C=1.25 at the trailing edge)

36
For case 1, which assumes non-zero pressure gradients, values of the analogy
function in the vicinity of the stagnation point can be analytically calculated using the
wedge flow solution (Eckert and Drake 1987). This solution method may be used where
velocity in the freestream, Us, can be expressed as:
Us = c ⋅ x m (3.4.1)

The exponent, m, is described as follows.


x dU s
m= ⋅ (3.4.2)
U s dx
According to Eckert and Drake (1987), the resulting Nusselt number for a boundary
condition (BC) of uniform wall temperature is:
Nu x = f (m) ⋅ Re x
1/ 2
Pr 1 / 3 (3.4.3)
We know, at this point, that the above equation is the same as:
Sh x = f (m) ⋅ Re x
1/ 2
Sc1 / 3 (3.4.4)
This equation can now be used for line 4/1. Here, the coefficient, f(m), is a constant with
x but which depends on the acceleration parameter, m, in Eqn. (3.4.2), and is tabulated in
Kays and Crawford (1993) as the form of f(m) multiplied by Sc1/3.
One can apply Eqn. (3.4.4) to the stagnation flow region (small x/C) of the line
4/1. The numerator is the flat plate case where m = 0, and the stagnation flow is assumed
for the denominator, m = 1. Therefore, f(0) = 0.41 and f(1) = 0.72 yield (pp.167, Kays
and Crawford 1993). The following results since Sc is the same for the both cases.
Sh 4 f (0)Sc1 / 3 U s , 4 0.41 U s , 4 U s, 4
= ⋅ = ⋅ = 0 . 57 (3.4.5)
Sh 1 f (1)Sc1 / 3 U s ,1 0.72 U s ,1 U s ,1

The numerator of the above equation is for the flat plate case for which m is zero. The
denominator is for the case with the non-zero pressure gradients for which stagnation
flow with m being 1.0 is assumed near the leading edge since the velocity increase is
linear with x. The factor, therefore, is dependent on the ratio of freestream velocities.
The ratio theoretically begins with infinity at the leading edge since the velocity at the
stagnation point, Us,1, is zero. For this reason, long-lasting acceleration, e.g. the GE90
profile, causes a continuous decrease seen until x/C of about 0.3 in Figs. 3.6 and 3.7.

37
1.4

1.2 4/1
1.0

0.8
Nu/Sh

0.6
2/3
0.4
0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
x/C

Figure 3.6 GE90 Wang cases,


Rec=540,000 and P/C=0.75 (x/C=1.38 at the trailing edge)

Once passing the acceleration regions, line 4/1 behaves differently, depending on
the pressure profiles. The GE90 profile, with long-lasting acceleration, tends to have
lower values in this region whereas the CF6 profile jumps from the minimum to a high
value. The differences between the figures seem to be greater between different pressure
profiles (Fig. 3.5 and Fig. 3.6) than between different chord Reynolds numbers (Fig. 3.6
and Fig. 3.7).
The only difference between cases 4 and 1 is the pressure profile, and, in case 1,
there is a point (or two) at which pressure gradient, at least locally, is zero. This can be
used to check the behavior of line 4/1 in the figures. If both had no pressure gradient,
meaning that they are identical cases, the line would remain at Sh/Sh of 1.0 throughout
the airfoil. Though the local mass transfer process depends on the local conditions as
well as the flow history, one can observe that, when local pressure gradient becomes zero,
the Sh/Sh values tend to adjust toward 1.0.

38
1.4

1.2 4/1

1.0

0.8
Nu/Sh

0.6
2/3
0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
x/C

Figure 3.7 GE90 profile with Chen’s Reynolds number,


Rec=171,000 and P/C=0.77 (x/C=1.38 at the trailing edge)

For ease, the GE90 profile can be looked at first. According to Fig. 3.1, the GE90
profile has a point at which no acceleration or deceleration is present (x/C of 0.68). At
this streamwise position, the Sh/Sh values in Figs. 3.6 and 3.7 are about 0.8, moving
toward 1.0. At x/C of 1.0, line 4/1 in Fig. 3.6 crosses the Sh/Sh value of 1.0. This delay
of reaching 1.0 seems to be the effect of the flow history, which has suppressed the
boundary layers from growing. The flow in the GE90 profile has experienced a long
acceleration period when it ends. The boundary layer of such flow is still different from
the one having developed on a flat plate. Following the acceleration period in the GE90
profile is the deceleration period. Deceleration seems to strongly change boundary layer
characteristics, thereby increasing the Sh/Sh values. Finally, it reaches the maximum
value just before the flow separates.
The CF6 pressure profile is much more flat than the GE90 profile, as seen in Fig.
3.1. Therefore, the Sh/Sh values in Fig. 3.5 vary but remain close to closer to 1.0. The

39
profile has a point of zero gradient just after the stagnation flow region near the leading
edge. The Sh/Sh values quickly adjust to near 1.0. From this point on, the profile has a
few points at which the gradient becomes zero and the Sh/Sh values tend to converge to
1.0.
Regardless of the type of pressure gradient, line 4/1 strongly responds to adverse
pressure gradients. Just before the flow separates, it goes beyond 1.0 and even over 1.2
for the GE90 cases. One must note here, however, that the local freestream velocity is
not equal to that of the plane wall case, and the normalized local freestream velocity at
separation of each pressure profile is also different. Therefore, to conclude that the
difference in the pressure profiles caused the GE90 profile to have higher analogy factors
at separation may be too hasty.

3.4.3 The effects of thermal BC in heat/heat transfer analogy: line 2/3

Line 2/3 in Figs. 3.5, 3.6 and 3.7 corresponds to both cases having the same non-zero
pressure gradients, the same transport property of 0.707 (presumably air) but different
BCs. To be precise, the numerator, case 2, is with a uniform level BC and the
denominator, case 3, with a uniform flux BC. If this analysis of line 2/3 yielded a
sufficiently flat profile of the analogy function, the function could be used together with
the analogy 2/1 to form the analogy 3/1, which is the analogy between uniform level mass
transfer data and uniform flux heat transfer data. When this combined process, 2/3 and
2/1, can be simply expressed, an experimental relationship for the cross-BC analogy
functions can be found. This is done in the next chapter. Besides different Reynolds
numbers, deriving a simple analogy function requires understanding the effects of
mismatching BCs. This section addresses the cross-BC effect as augmented by the non-
zero pressure gradients.
Unlike the pair 4 and 1, cases 2 and 3 have the same geometry. Therefore, the
flow fields are identical. If, by any means, the uniform level BC generated a similar
temperature field as that of the uniform flux BC, the Nu/Nu ratio again would be near 1.0.

40
However, this is clearly not the case. The Nu/Nu values are confined in the region
between 0.6 and 1.0, smaller variations than in the previous study (4/1).
In the three figures, some trends can be observed. The distributions always begin
with the ratio of 1.0 at the leading edge and quickly decrease, without sudden changes,
toward values in the vicinity of 0.7 and 0.8. At the point of separation, each analogy
factor reaches the minimum of almost 0.6. Remaining always less than 1.0 is expected
because of the earlier discussion regarding the development of Eqn. (3.1.12). In fact, the
inverse of the equation can be used to explain the major part of the behavior of line 2/3 in
Figs. 3.5, 3.6 and 3.7. Because line 2/3 assumes the same transport property, the ratio is
unity. Therefore, Eqn. (3.1.12) is left only with the coefficient 1.36, and its inverse is
0.735. Originally, the equation is developed for a plane wall case with the Nusselt
number assumed to result from uniform-flux BC and the Sherwood number from a
uniform-level BC. The only difference from line 2/3 is the pressure profile, besides the
inverted ratio. Thus, for a flat plate case, 6/5, we expect a value of 1/1.36 (= 0.735, see
also Eqn. 3.1.14).
A general trend then is, except near the leading edge, that the local analogy factor
is greater than 0.735 when the flow accelerates and less when decelerating. The flow
history seems not to matter much here. When the velocity profiles get flat, the analogy
factors seem to quickly adjust to 0.735. The magnitudes of the deviations from the
reference value of 0.735 seem to be a function of flatness of the velocity profiles; the
flatter CF6 profile deviates less than does the less-flat GE90 profile. Finally, whether this
analogy function is useful in converting flux-BC heat transfer data to level-BC heat
transfer data for laminar flow regions is dependent on the type of pressure profile.

3.5 Considerations on Flow Transition from Laminar to Turbulent

As discussed in the next Chapter, the conventional analogy factor for turbulent boundary
layers is different from that of laminar boundary layers (compare Eqns. (4.1.2) and

41
(3.1.10) ). Therefore, it would be interesting to find how the factors vary in the transition
region.
One way to do this may rely on experimental data. If there were two identical
experiments, except that one was heat transfer experiment and the other mass transfer, the
two data sets could be compared to determine analogy factors in the transition region.
This is a motivation to the next chapter in which the factors in transition regions are
discussed.
Another possibility to study variations of analogy factors in transition regions is to
extend the TEXSTAN calculations performed in this Chapter. The program has a
capacity to simulate transitions and turbulent flows. However, the calculations performed
so far had these features disabled so that no transition or turbulence modeling would be
necessary. Therefore, the calculations stopped when flows reach separation or the trailing
edge, remaining laminar throughout. By implementing transition and turbulence models,
the above calculations can be extended to the factors in transition regions and into
turbulent boundary layer regions. To do this, however, one must know which transition
model is best suited for the flow that is simulated.
An option available in TEXSTAN is to provide the program with a momentum-
thickness Reynolds number as an indicator for the beginning of transition. The Reynolds
number, Rem, increases with the streamwise positions, and transition is expected to begin
where Rem reaches a transition value. Therefore, with this option, the user of the program
must have, at least, an estimate of the transitional Reynolds number.
Momentum-thickness Reynolds number is a simple indicator to predict the
beginning of transition, and it has been discussed quite extensively. McDonald (1973)
establishes a relation between displacement-thickness Reynolds number and freestream
turbulence (Figure 1). Almost twenty years later, Mayle (1991a and 1991b) developed,
from more experimental data, an equation for transitional momentum-thickness Reynolds
number as a function of freestream turbulence intensity. Mayle emphasizes a good
agreement between the McDonald’s relation, also called Abu-Ghannam and Shaw
relation, and the Mayle’s equation, which reads:

42
Re m , tr = 400TI −5 / 8 (3.5.1)

Both McDonald’s and Mayle’s relations are plotted in Fig. 3.8, along with momentum
thickness Reynolds numbers calculated by TEXSTAN for the CF6 velocity profile.
At a first glance of Fig. 3.8, one recognizes that the Rem from TEXSTAN does not
reach either of the transition Reynolds numbers, even at the point of separation. Even at
this point, the TEXSTAN Rem is much lower than the transition values. This then
suggests that the transition take place after flow separation, since we know, from Chung’s
and Chen’s experiments, that the boundary layers near the trailing edge of the CF6 profile
are turbulent. This is consistent with Chen’s discussions in his study (section 5.3 in
Chen, 1988). One reason for the difference in momentum thickness Reynolds numbers
between the models and the TEXSTAN Rem is that the models assume no streamwise
pressure gradients. The TEXSTAN calculation prescribes adverse pressure gradients
toward the trailing edge, which considerably reduce transitional Reynolds numbers. This
analysis raises yet another question regarding the analogy function: What is the analogy
function in the region of flow separation? This is left for further studies in the analogy
function.

43
700

600

500

400 McDonald
Rem

Mayle
300
TEXSTAN
200

100

0
0.0 0.2 0.4 0.6 0.8 1.0
x/C

Figure 3.8: Comparison of critical momentum thickness Reynolds numbers for the CF6
velocity profile that is shown in Fig. 3.1

3.6 Conclusions

The heat/mass transfer analogy function is studied for laminar boundary layers. Transport
of heat and sublimation of naphthalene from heated and naphthalene-coated surfaces,
respectively, are assumed. Analytical and numerical calculations have resulted in various
values and distributions of the analogy function, depending on the combinations of
pressure profiles and thermal boundary conditions. Furthermore, to examine the effects
of these two, a mass/mass analogy function and a heat/heat analogy function are
developed.
First, two analogy factors are developed analytically for plane wall cases. The
analogy factor, Nu/Sh, of 0.677 is a conventional relationship which assumes zero
pressure gradients and uniform-level boundary conditions for both heat and mass transfer
processes. It is a motivation to the present study to examine how the analogy function
44
deviates from this conventional value by changing the boundary conditions and pressure
profiles. The other analogy factor developed is 0.924 (1.36 multiplied by 0.677). This
factor assumes zero pressure gradients for both and flux-BC for heat transfer and level-
BC for mass transfer. The effect is a 36% increase in the analogy factor due to the flux
boundary condition assumed in the heat transfer process. Still, the single value of 0.924
may be applied to an entire surface.
Second, the TEXSTAN program is employed to numerically calculate the analogy
functions for non-zero pressure gradients, namely the CF6 and GE90 profiles. The above
two analytical solutions are verified, and the non-zero gradient cases exhibit deviations
from the corresponding plane wall cases. The magnitudes of the deviations are different,
depending on the thermal BC assumed in the heat transfer processes. The deviations are
larger with a heat flux BC than with a uniform temperature BC. These behaviors are
similarly observed in both the CF6 and the GE90 cases.
Third, two pairs of processes that are not heat/mass analogies are considered in order to
further investigate the effects of pressure gradients and thermal BCs. To see the effects of
pressure profiles, mass/mass analogy is developed. The resulting shape of line 4/1 in
Figs. 3.5, 3.6 and 3.7 suggests that the shapes of lines 3/1 and 2/1 in Figs. 3.2, 3.3 and 3.4
are due to the pressure profiles, but the pressure profiles may not be responsible for the
deviations from the corresponding plane wall cases. A heat/heat analogy is developed to
see the effects of thermal boundary conditions. With the only difference being the
thermal boundary conditions, this analogy function examines a possibility of a conversion
of flux-BC heat transfer data to level-BC heat transfer data for non-zero pressure
gradients. If this analogy function happened to be completely a flat distribution, the
conversion would be very simple. However, this has not been the case. The deviation
from the corresponding plane wall factor of 0.735 in this case is larger than the deviation
of line 2/1 in Figs. 3.2, 3.3 and 3.4 from its corresponding plane wall factor of 0.677.
This implies, for laminar boundary layers, that a mass transfer experiment with
naphthalene sublimation is going to be more accurate than a flux-BC heat transfer

45
experiment since the boundary condition in an actual gas turbine heat transfer is
considered to be a uniform level boundary condition.
Finally, a few thoughts are given to the analogy function in the regions of flow separation
and transition. It should be interesting and useful to uncover how the analogy function
behaves under such conditions. The present study suggests that both numerical and
experimental investigations could be done for this purpose, and the experimental
approach is pursued in the next chapter.

46

Das könnte Ihnen auch gefallen