Prediction of Ingestion Through Turbine Rim Seals-Part II, Externally Induced and Combined Ingress

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Prediction of Ingestion Through

Turbine Rim Seals—Part II:


Externally Induced and Combined
Ingress
Ingress of hot gas through the rim seals of gas turbines can be modeled theoretically
J. Michael Owen using the so-called orifice equations. In Part I of this two-part paper, the orifice equations
Department of Mechanical Engineering,
were derived for compressible and incompressible swirling flows, and the incompressible
University of Bath,
equations were solved for axisymmetric rotationally induced (RI) ingress. In Part II, the
Bath BA2 7AY, UK
incompressible equations are solved for nonaxisymmetric externally induced (EI) ingress
and for combined EI and RI ingress. The solutions show how the nondimensional ingress
and egress flow rates vary with ⌰0, the ratio of the flow rate of sealing air to the flow rate
necessary to prevent ingress. For EI ingress, a “saw-tooth model” is used for the cir-
cumferential variation of pressure in the external annulus, and it is shown that ␧, the
sealing effectiveness, depends principally on ⌰0; the theoretical variation of ␧ with ⌰0 is
similar to that found in Part I for RI ingress. For combined ingress, the solution of the
orifice equations shows the transition from RI to EI ingress as the amplitude of the
circumferential variation of pressure increases. The predicted values of ␧ for EI ingress
are in good agreement with the available experimental data, but there are insufficient
published data to validate the theory for combined ingress. 关DOI: 10.1115/1.4001178兴

1 Introduction Phadke and Owen 关2–4兴 determined Cw,min, the minimum seal-
ing flow rate needed to prevent ingress, in a simple rotor-stator
In Part I of this two-part paper, the orifice equations were de-
system with a number of different rim-seal geometries, with and
rived for compressible and incompressible swirling flows, and the
without an external flow of air. 共It should be noted that there were
incompressible equations were solved for rotationally induced
no vanes or blades in the external annulus of their rig; the authors
共RI兲 ingress. The solutions showed that ␧, the sealing effective-
created the circumferential pressure asymmetries referred to be-
ness, depends on three nondimensional parameters: ⌰0 the ratio of
low by blocking sections of the annulus with wire mesh.兲 They
the sealing flow rate to the minimum flow rate to prevent ingress,
observed both RI ingress 共where, with no external flow, Cw,min
⌫c the ratio of the discharge coefficients for ingress and egress,
increased with increasing Re␾兲 and EI ingress 共where, with non-
and ⌫␤ a swirl ratio. In Part II, the incompressible equations are
axisymmetric external flow, Cw,min was independent of Re␾ and
solved for externally induced 共EI兲 and combined EI and RI in-
increased with increasing Rew, the axial-flow Reynolds number in
gress.
the external annulus兲.
One of the main features of this paper is that the 共somewhat
For the case of quasi-axisymmetric external flow 共with little or
long兲 solutions are analytical: The resulting algebraic equations
no circumferential pressure variation in the annulus兲, Phadke and
are simple to use and should be of value to designers. To make the
Owen found that Cw,min could decrease with increasing external
paper more “reader friendly,” much of the analysis is restricted to
the Appendices; this leaves more space in the main body of the flow, and their measured variation of Cw,min with Rew is shown in
paper to discuss the significance of the solutions. Fig. 1 for an axial-clearance seal with Gc = 0.01. For Rew = 0,
A review of EI ingress is given in Sec. 2, the orifice equations where RI ingress occurs, Cw,min ⬀ Re␾; for large values of Rew,
for EI ingress are solved in Sec. 3, and the solutions are compared where EI ingress dominates, Cw,min ⬀ Rew; for intermediate values
with the experimental data in Sec. 4. In Sec. 5, the equations are of Rew, a minimum value of Cw,min can be seen to occur. 共This
solved for combined ingress, and the conclusions are presented in figure is an example of combined ingress, which is discussed fur-
Sec. 6. ther in Sec. 5. The computations of Vaughan 关5兴 and the experi-
ments of other authors discussed below suggest that the external
flow caused a reduction in the discharge coefficients for the seals
and a consequent reduction in Cw,min.兲
2 Review of EI Ingress Phadke and Owen 关4兴 correlated their results for EI ingress,
Abe et al. 关1兴, who used a turbine rig with vanes in the annulus based on flow visualization for a number of different seal geom-
upstream of the rim seal, were the first to show that ingress could etries, by
be dominated by the external flow in the annulus rather than by
the rotational speed of the disk. The authors tested several rim- Cw,min = 2␲KGc Pmax1/2 共2.1兲
seal geometries and identified three things that affected ingress: where
the ratio of the velocities of the sealing air and the flow in the
annulus, the rim-seal clearance, the shape of the rim seal. 1
Pmax = 2 C p,max Rew2 共2.2兲
C p,max is a nondimensional pressure difference in the external an-
Contributed by the International Gas Turbine Institute of ASME for publication in
the JOURNAL OF TURBOMACHINERY. Manuscript received July 23, 2009; final manu-
nulus and K is an empirical constant; the data were correlated with
script received August 10, 2009; published online November 12, 2010. Editor: David K = 0.6, as shown in Fig. 2. It should be noted that, although Pmax
Wisler. is the controlling parameter, for a given external geometry or in an

Journal of Turbomachinery Copyright © 2011 by ASME JULY 2011, Vol. 133 / 031006-1

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 1 Variation of Cw,min with Rew for axial-clearance seal,
Gc = 0.01 †4‡. For Re␾ / 106, 䊊:ⴝ0, 䊐:ⴝ0.2; 䉮:ⴝ0.4; 䉭:ⴝ0.6; ⴛ:
ⴝ0.8; 䉲:ⴝ1.0; 䉱:ⴝ1.2.

Fig. 3 Variation of Cw,min with Cp,max †14‡ „symbols represent


engine, C p,max is expected to depend only weakly on either Re␾ or experimental data; lines correspond to correlations…
Rew. Hence Cw,min ⬀ Rew, as Eqs. 共2.1兲 and 共2.2兲 show.
Hamabe and Ishida 关6兴 made measurements of the sealing ef-
fectiveness in a turbine rig fitted with upstream guide vanes but found that at low sealing flow rates the addition of the rotating
with no downstream blades. For EI ingress through a simple axial- blades reduced the ingress compared with the case when only
clearance seal, they correlated the effectiveness with a nondimen- vanes were used. Gentilhomme et al. 关10兴 made ingress measure-
sional parameter similar to that used by Phadke and Owen 关4兴, and ments and carried out computations for a single-stage turbine rig
the results of their orifice model were in reasonable agreement with both vanes and blades. The circumferential pressure in the
with their measurements. 共They also showed that, as discussed annulus, obtained by CFD, was used in conjunction with an orifice
above, the discharge coefficient for egress decreased as the exter- model to estimate the effectiveness.
nal flow rate increased.兲 In their model, they approximated the The Aachen group 关11–15兴 conducted many ingress studies in
shape of the external distribution of pressure using simplified turbine rigs with vanes and blades. Of particular relevance, here is
wave forms, which showed that the prediction of ingress de- the paper by Bohn and Wolff 关14兴 who presented a correlation for
pended on the shape used. This is discussed further in Sec. 3. the sealing effectiveness, ␧, in terms of Cw,0, Gc, and C p,max, and
The Sussex group 关7–10兴 conducted many EI ingress studies. their correlations for the four seal geometries shown in Fig. 3
Dadkhah et al. 关7兴 made ingress measurements for both the up- display the linear variation of Cw,min with C p,max1/2 that was found
stream and downstream wheel spaces in a gas-turbine rig without by Phadke and Owen 关4兴; the value of K = 0.6 suggested by the
vanes or blades. Chew et al. 关9兴 extended this work with upstream latter authors produces a conservative estimate for Cw,min.
nozzles in the annulus; their orifice model performed less well The Hartland team 关16–21兴 made a significant contribution to
than that of Hamabe and Ishida 关6兴 in predicting the effectiveness the ingress literature. Recently, Johnson et al. 关19兴 used an orifice
but their steady 3D CFD 共computational fluid dynamics兲 compu- model to obtain good predictions of the effectiveness measure-
tations gave encouraging results. They also measured the dis- ments in the turbine rig of Bohn et al. 关15兴. For the external
charge coefficients, Cd, for the rim seal when there was no disk circumferential pressure distribution in their model, they used the
rotation. The measurements, like those of Hamabe and Ishida 关6兴, values obtained from 2D time-dependent CFD, which allowed the
showed that, for outflow, Cd,e decreased monotonically with in- effects of the vane wakes and the blade bow waves to be taken
creasing external flow rate. However, for inflow Cd,i reached a into account. A modified version of their orifice model was also
minimum at a particular ratio of sealing-to-annulus velocity above successfully applied by Johnson et al. 关21兴 to the ingress measure-
which it increased with increasing external flow rate. The first ments made on a turbine rig in Arizona State University. These
published data for a turbine rig with both vanes and blades were papers are discussed further in Sec. 4.
presented by Green and Turner 关9兴; somewhat surprisingly, they CFD has been used with some success for the ingress problem,
but care must be exercised in computing these unsteady 3D flows.
Although some measurements have shown that the nonaxisym-
metric pressure distribution created in the annulus can be attenu-
ated after the ingested flow has passed through the rim seal, the
combined experimental and computational study of Cao et al. 关22兴
showed that the swirling ingress can destabilize the rotating flow
in the wheel space. Of particular interest is that the instabilities
create a series of cells with a circumferential periodicity different
from that of the vanes or blades. 共This phenomenon may be re-
lated to that found in buoyancy-induced flow, in which the rotat-
ing cells create the circumferential distribution of pressure needed
to generate the necessary Coriolis forces.兲 The authors suggest
that a “segmental domain” in CFD 共where computations are con-
ducted over an angular segment corresponding to, for example,
the vane pitch兲 may be unable to capture these instabilities; how-
ever, their computations for a 90 de segment showed largely good
agreement with those for the full 360 deg domain.
Recently, Mirzamoghadam et al. 关23兴 used a steady 3D CFD
code to study EI ingress for a representative turbine geometry.
Fig. 2 Variation of Cw,min with Pmax †4‡ „symbols represent ex- They computed the variation of “ingestion mixing efficiency”
perimental data; line corresponds to Eq. „2.1…… with superposed flow rate and also determined the ingested flow

031006-2 / Vol. 133, JULY 2011 Transactions of the ASME

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 4 Arbitrary circumferential variation of pressure and
radial velocity in external annulus

rate when the superposed flow was zero. 共Despite the difference
between the authors’ definition of ingestion mixing efficiency and
the definition of effectiveness used below, there is qualitative
agreement between the results from the orifice model for EI in-
gress and the authors’ computations.兲
The above review is by no means exhaustive but it sets the
scene for the following analysis. The reader is referred to the
reviews of Johnson et al. 关16兴 and Chupp et al. 关24兴 if more
background information is required.

3 Solution of Incompressible Orifice Equations for EI


Ingress
It should be noted that although the analysis below is based on Fig. 5 Circumferential variation of pressure and radial velocity
steady incompressible flow, the results should be applicable to for saw-tooth model: „a… circumferential variation of p1 and p2;
unsteady flow providing the appropriate time-average pressures „b… circumferential variation of Vr
are used. This is discussed in Appendix A3 of Part I.
3.1 Circumferential Variation of Pressure in Annulus. In a
turbine with N nozzle guide vanes, it is convenient to define
f = 1 – 2␪ 共3.2兲
N when 0 ⱕ ␪ ⱕ 1 / 2 and
␪= ␾ 共3.1兲
2␲
f = 2␪ − 1 共3.3兲
where ␾ is the angular coordinate in the 360 deg annulus and ␪ is
the coordinate between the N vanes. Hence 0 ⱕ ␾ ⱕ 2␲ / N and 0 when 1 / 2 ⬍ ␪ ⱕ 1. Using Eqs. 共A11兲 and 共A12兲 in Appendix A,
ⱕ ␪ ⱕ 1, and it is assumed that p2 = p2共␪兲 is the appropriate time- and the symmetry of this distribution, it follows that for gⴱ ⱕ g
average static pressure 共see Appendix A3 in Part I兲 between each ⱕ1


pair of vanes. For simplicity, it is assumed that p1, the pressure at 1/2
Ie = 2冑2
2
some suitable location in the wheel space, is axisymmetric. 共␪ − ␪⬘兲1/2d␪ = g3/2 共3.4兲
3
An arbitrary variation of p2 with ␪ is shown in Fig. 4, where ␪⬘


p2,max and p2,min are the maximum and minimum values of p2 at
␪⬘
Ii = 2冑2
some location in the external annulus. Ingress occurs where p2 2
共␪⬘ − ␪兲1/2d␪ = 共1 − g兲3/2 共3.5兲
⬎ p1 共i.e., 0 ⬍ ␪ ⬍ ␪⬘兲 and egress occurs where p2 ⬎ p1 共i.e., ␪⬘ 3
0
⬍ ␪ ⬍ 1兲. Although the locations for the measurement of p1 and p2
do not feature explicitly in the equations derived below, they may 2
affect the value of the discharge coefficients derived from the I0 = Ie − Ii = 关g3/2 − 共1 − g兲3/2兴 共3.6兲
experimental measurements. 3
The orifice equations for EI ingress are discussed in Appendix where
A for the case where the flow is incompressible and the effects of
swirl in the annulus and wheel-space are negligible. The so-called 1
␪⬘ = 共1 − g兲 共3.7兲
saw-tooth model is considered in Sec. 3.2, and the theoretical 2
results are discussed in Sec. 3.3.
when g = f. As I0 = Ie = Imin when g = 1, it follows that
3.2 Saw-Tooth Model for EI Ingress. Hamabe and Ishida 关6兴 2
solved the orifice equations numerically using three different cir- Imin = 共3.8兲
cumferential distributions for the pressure in the annulus; their 3
“triangular wave form” is what is referred to here as a “saw- Hence, from Eqs. 共A13兲–共A17兲,
tooth.” Figure 5 shows a saw-tooth distribution of p2 and the 2
corresponding distribution of the radial velocity, Vr, through the ⌽min = 3 Cd,e⌬C p1/2 共3.9兲
seal clearance; p1ⴱ is the value of p1 when Cw,0 = 0.
From the definitions of f and g given in Appendix A, it follows Cw,e
⌰e = = g3/2 共3.10兲
that Cw,min

Journal of Turbomachinery JULY 2011, Vol. 133 / 031006-3

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 6 Variation of Cw,e , Cw,i and ε with Cw,0 for saw-tooth
model „Cd,i = Cd,e…

Fig. 7 Effect of Cd,i / Cd,e on variation of ε with Cw,o. Solid line,


saw-tooth model; dashed line, RI ingress.
Cw,i
⌰i = = ⌫c共1 − g兲3/2 共3.11兲
Cw,min

Cw,0 not identical, show that ␧ increases as Cd,i / Cd,e decreases. For a
⌰0 = = g3/2 − ⌫c共1 − g兲3/2 共3.12兲 given value of Cd,e, the value of Cd,i / Cd,e does not alter the value
Cw,min
of Cw,min but it does affect the shape of the ␧ versus Cw,0 / Cw,min
As ⌰0 = 0 when g = gⴱ, it follows that curves. This characteristic could be useful in matching the theory
⌫c2/3 with experimental data; however, as shown below, a value of
gⴱ = 共3.13兲 Cd,i / Cd,e = 1 provides a good fit with the available data.
1 + ⌫c2/3 It is now appropriate to compare the results of the saw-tooth
and so gⴱ = 2 when ⌫c = 1. Also as model with the experimental data. It should be emphasized that
1
the saw-tooth model gives a convenient approximation; for more
⌰i accurate results, it would be necessary to evaluate the integrals
␧=1− 共3.14兲
⌰0 Ie , Ii and Io numerically using the time-average values of the mea-
sured pressure distribution.
it follows that, for gⴱ ⱕ g ⱕ 1,

␧ = 1 − ⌫c 冋 册
1−g
g
3/2
共3.15兲 4 Comparison Between Saw-Tooth Model and Experi-
mental Data
These results are referred to here as the saw-tooth model for EI
ingress, and their significance is discussed below. In the experiments discussed below, the effectiveness was de-
termined in terms of a velocity ratio, u0 / ⍀b, where u0 is the
3.3 Discussion of Theoretical Results for Saw-Tooth average velocity of the superposed flow in the seal clearance such
that
Model. The above equations express the flow parameters and ␧ in
terms of g, the pressure parameter. However, most experimental ṁ0
studies express ␧ in terms of the superposed flow rate, and so it is u0 = 共4.1兲
2␲␳bsc
useful to show how ␧ and the other parameters vary with ⌰0. This
is shown in Fig. 6 for the case where Cd,i = Cd,e 共that is, ⌫c = 1兲. It therefore follows that
It should be noted that, as ⌰i = ⌰i,max when g = gⴱ, Eq. 共3.11兲 u0
shows that ⌽0 = 共4.2兲
⍀b
ⴱ 3/2
⌰i,max = ⌫c共1 − g 兲 共3.16兲 Johnson et al. 关19兴 used a time-average circumferential distribu-
and when ⌫c = 1 共i.e., Cd,i = Cd,e兲 tion of pressure computed at the upstream edge of the seal clear-
ance, downstream of the vanes. Their pressure coefficient, C p,2, is
1
⌰i,max = ⌰e,min = 共3.17兲 defined by
2 冑2
p 2 − p 1ⴱ
or C p,2 = 共4.3兲
1/2␳⍀2b2
Cw,i,max ⬇ 0.35Cw,min 共3.18兲
p1ⴱ being the value of p1 when Cw,0 = 0.
It should also be noted that Eq. 共3.17兲 is identical with Eq. 共4.7兲 of Bohn et al. 关15兴 conducted experiments at Aachen University in
Part I for RI ingress. The importance of this result is that the a turbine rim-seal rig with axial-clearance seals, using concentra-
maximum ingress, which occurs when the sealing flow rate is tion measurements to determine the sealing effectiveness. Johnson
zero, is approximately 35% of the flow rate needed to seal the et al. 关19兴 used an orifice model to calculate numerically the mea-
system. Also of interest to designers is the fact that, for both EI sured effectiveness for two different geometrical configurations.
and RI ingress, ␧ ⬇ 0.8 when ⌰0 = 0.5; that is approximately 50% The two configurations 共which they referred to as Conf 1a and 1c兲
of Cw,min achieves an 80% effectiveness. had different axial spacing between the trailing edge of the nozzle
Figure 7 shows the effect of Cd,i / Cd,e on the variation of ␧ with guide vanes and the leading edge of the turbine blades, Conf 1a
Cw,0 / Cw,min for the saw-tooth model; also shown are the curves having a smaller spacing 共and consequently a larger level of ⌬p兲
for RI ingress, which correspond to those in Fig. 5 of Part I. As than Conf 1c. The seal clearance was midway between the vanes
expected, the results for EI and RI ingress, which are similar but and blades, where the pressure created by the vane wakes and the

031006-4 / Vol. 133, JULY 2011 Transactions of the ASME

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 8 Circumferential distribution of Cp,2: Solid line is for Fig. 10 Variation of ε with ⌽0. Symbols represent experimen-
Conf 1a of Johnson et al. †19‡ Dashed line is equivalent saw- tal data of Johnson et al. †21‡; solid line represents saw-tooth
tooth distribution model with ⌫c = 1.

blade bow waves was close to a minimum. For both configura- saw-tooth model, with an assumed value of ⌽min = 0.20, provides
tions, an axial-clearance seal with Gc = 0.0296 was used and the a good fit to the experimental data but, as the value of ⌬C p is not
rotational speed was 4500 rpm. specified in the paper, it is not possible to calculate the value of
Figure 8 shows a comparison between the circumferential dis- Cd,e.
tribution of C p,2 determined by Johnson et al. 关19兴 for Conf 1a and The authors in Ref. 关21兴 solved their own orifice model numeri-
the equivalent saw-tooth distribution. Figure 9 shows a compari- cally, and they computed ␧ for various values of Cd,e and Cd,i; the
son between the theoretical and experimental values of effective- respective values of 0.27 and 0.20 were found to give the best fit
ness for Confs 1a and 1b. to their data. Their computations, which—for clarity—are not
According to the saw-tooth model, ␧ depends on ⌽0 / ⌽min, and shown in Fig. 10, straddle the curve for the saw-tooth model.
so to compare the theory with the measurements it is necessary to Although the assumption that Cd,e = Cd,i provides a good fit be-
know, or to assume, a value for ⌽min. To fit the experimental data tween the saw-tooth model and the above experimental data, there
in Fig. 9, the values of ⌽min = 0.22 and 0.11 were, respectively, is no reason to believe that this assumption will be valid in all
chosen for Confs 1a and 1c; this was an arbitrary choice and linear cases. It is recommended that, for future comparisons between the
regression could be used to find a better fit. It can be seen that the model and experimental data, a least-squares fit should be used to
agreement between the data and analytical saw-tooth model is as determine suitable values for Cd,e and Cd,i.
good as that for the numerical model of Johnson et al. 关19兴. Having demonstrated that the simple EI and RI models produce
The discharge coefficient can be calculated from Eq. 共3.9兲 for results that are in good agreement with the published experimental
the saw-tooth model, where data, it is now appropriate to look at the case of combined ingress.
3 ⌽min
Cd,e = 共4.4兲 5 Combined EI and RI Ingress
2 ⌬C p1/2
5.1 Relationship Between EI and RI Ingress. It was shown
For Conf 1a, Fig. 8 shows that ⌬C p ⬇ 0.55, hence Cd,e ⬇ 0.44 for
in Sec. 3 that there is a strong similarity between the theoretical
the saw-tooth model; Johnson et al. 关19兴 used Cd,e = 0.40 in their
variation of ␧ with ⌰0 for the special cases of EI and RI ingress
numerical model. The difference between these two discharge co-
共where Cd,i = Cd,e and the external swirl is zero兲. For these cases,
efficients is attributed to the difference between the two pressure
␧ = ␧共⌰0兲 where ⌰0 = ⌽0 / ⌽min, and from Eq. 共4.1兲 of Part I, for RI
distributions shown in Fig. 8. 共It is unclear in the experimental
paper which value of ⌬p should be used to calculate Cd,e for Conf ingress
1c.兲 ⌽min,RI = Cd,e⬘C␤11/2 共5.1兲
Figure 10 shows the comparison between the saw-tooth model
and the effectiveness measured by Johnson et al. 关21兴. The con- where the prime is used to distinguish the discharge coefficient for
centration measurements were made for Re␾ = 5.86⫻ 105 in a rig RI flow 共in which the variation in external pressure is zero兲 from
with radial-clearance seals 共the value of Gc was not given兲. The that in the general case where the pressure variation is nonzero.
From Eq. 共3.9兲 for EI ingress,
2
⌽min,EI = 3 Cd,e⌬C p1/2 共5.2兲
It was shown in Part I that, for RI ingress, the Bayley–Owen 关25兴
correlation, Cd,e⬘C␤ 1/2 = 0.097, gave a reasonable fit to the data of
1
Graber et al. 关26兴. Although this correlation cannot be regarded as
a general result for all seals, for simplicity it is used below. It
therefore follows that, using the above results,
⌽min,EI Cw,min,EI
= ⬇ 7Cd,e⌬C p1/2 共5.3兲
⌽min,RI Cw,min,RI
The solution of the saw-tooth model for combined ingress, where
both EI and RI ingress are significant and the external swirl is not
negligible, is given in Appendix B; the case when external swirl is
negligible is discussed below.
Fig. 9 Variation of ε with ⌽0 for Confs 1a and 1c. Rectangular
blocks represent range of experimental data †15‡; dashed line 5.2 Saw-Tooth Model for Combined Ingress When Exter-
represents orifice model of Johnson et al. †19‡; solid line rep- nal Swirl Is Negligible. As stated in Part I, contrary to the pre-
resents saw-tooth model with ⌫c = 1. diction of the theoretical model, the experimental data of Graber

Journal of Turbomachinery JULY 2011, Vol. 133 / 031006-5

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use


mensional sealing flow rate, and Cw,min, the minimum value of
Cw,0 needed to prevent ingress兲, on Cd,i / Cd,e 共the ratio of the
discharge coefficients for ingress and egress兲, and on the swirl
ratio of the external fluid. The variation of ␧ with ⌰0 for EI
ingress is similar to, but not identical with, that for RI ingress.
For the special case, or basic theory for EI ingress, where
Cd,i / Cd,e = 1 and the external swirl is zero, the solutions show
that—as for RI ingress—the maximum nondimensional ingress
共which occurs when the sealing flow rate is zero兲 is approximately
0.35Cw,min. Also, for both EI and RI ingress, ␧ ⬇ 0.8 when ⌰0
= 0.5; that is, approximately 50% of Cw,min achieves an 80% ef-
Fig. 11 Variation of Cw,min,com / Cw,min,RI with ⌫⌬p1/2 according to fectiveness.
Eq. „5.7… with Cd,e / Cd,e⬘ = 1 Assuming that Cd,i / Cd,e = 1, the theoretical predictions of ␧ for
EI ingress are in good agreement with published measurements,
although there is insufficient information in the literature to cal-
et al. 关26兴 showed no systematic effect of external swirl on the culate the actual values of the discharge coefficients. There are
measured effectiveness. For simplicity, therefore, it is assumed also insufficient published data to validate the theoretical results
that ⌫␤ = 0 in Eq. 共B10兲 of Appendix B so that for combined ingress. There is a need for more experimental mea-
surements in this important subject, and all relevant data should
2 关1 + ⌫⌬p兴3/2 − 1 be included in the publication of the experimental results.
⌽min,com = Cd,eC␤11/2 共5.4兲
3 ⌫⌬p It is hoped that this two-part paper will be helpful to the de-
signers of internal air systems and to those involved in computa-
where ⌫⌬p is the ratio of the driving forces for EI and RI ingress.
tional and experimental research into the ingress problem. In par-
In the limit when ⌫⌬p = 0 共such that Cd,e = Cd,e⬘兲, ticular, the model could be used to inform the design of
⌽min,com = Cd,e⬘C␤11/2 共5.5兲 experimental rigs for ingress measurements, to help to interpret
the results of these rigs 共and the results of CFD兲, and to extrapo-
which is identical to Eq. 共5.1兲 for RI ingress. Also as ⌫⌬p → ⬁ late the rig results to engine conditions. The accuracy of this ex-
2 trapolation will depend on the limitations of both the theoretical
⌽min,com = Cd,e⌬C p1/2 共5.6兲 model and the experimental rig.
3
which is identical to Eq. 共5.2兲 for EI ingress. Equation 共5.4兲 can be Acknowledgments
expressed as I wish to thank Gill Green for producing most of the figures in
⌽min,com Cw,min,com 2 Cd,e 关1 + ⌫⌬p兴3/2 − 1 this two-part paper, and I am grateful to my colleagues, particu-
= = 共5.7兲 larly Olly Pountney, Carl Sangan, and Kunyuan Zhou, for check-
⌽min,RI Cw,min,RI 3 Cd,e⬘ ⌫⌬p
ing the analysis. I also wish to thank Dr. B.V. 共Bruce兲 Johnson for
where, as stated above, Cd,e⬘ is the value of Cd,e when ⌫⌬p = 0. his helpful comments and for the many discussions on ingress we
Figure 11 shows the EI and RI asymptotes together with the varia- have had over the years. Last but not the least I thank the review-
tion of ⌽min,com / ⌽min,RI according to Eq. 共5.7兲 when Cd,e / Cd,e⬘ ers for their constructive comments.
= 1.
As stated in Sec. 2, Chew et al. 关9兴 measured the effect of the Nomenclature
external flow on the discharge coefficients and showed that Cd,e b ⫽ radius of seal
decreased as the ratio of the axial velocity in the annulus to the Cd,e , Cd,i ⫽ discharge coefficients for egress, ingress
radial velocity through the seal increased. It can be shown that this
Cd,e⬘ ⫽ value of Cd,e for RI ingress
velocity ratio is proportional to the parameter Gc⌫⌬p1/2 / Cw,e, and
pressure coefficient 共=共p2 − p1兲 / 2 ␳⍀2b2兲
1
when Cw,e = Cw,min this parameter takes on a constant value for EI
Cp ⫽
pressure coefficient 共=共p2 − p1ⴱ兲 / 2 ␳⍀2b2兲
1
ingress 共where Cw,min ⬀ ⌫⌬p1/2兲. That is, Cd,e initially decreases as C p,2 ⫽
pressure coefficient 共=⌬p / 2 ␳W2兲
1
⌫⌬p1/2 increases and then tends to a constant value as the EI as- C p,max ⫽
ymptote is approached. This could explain the results of Phadke Cw ⫽ nondimensional flow rate 共=ṁ / ␮b兲
and Owen 关3,4兴 shown in Fig. 1 where Cw,min initially decreases Cw,min ⫽ minimum value of Cw,0 to prevent ingress
from the RI limit as the axial flow increases and, after reaching a Cw,0 ⫽ nondimensional sealing flow rate
minimum value, increases as it approaches the EI asymptote. C ␤1 ⫽ modified internal swirl ratio
In future ingress experiments, it is important to establish where 共=␤12 / 共1 − r12 / r22兲兲
the measurements are located in the EI-RI domain, particularly if
the data are to be extrapolated from rig to engine conditions.
C ␤2 ⫽ modified external swirl ratio 共=␤22 / 共r22 / r12
There are insufficient published data to validate the theory for − 1兲兲
combined ingress, and it is recommended that tests should be f ,g ⫽ pressure terms in EI saw-tooth model
conducted to determine the relationship between Cw,min and ⌫⌬p; f ⬘ , g⬘ ⫽ pressure terms in combined-ingress model
this would involve measuring the effect of external flow on the G ⫽ gap ratio 共=s / b兲
discharge coefficients. Having determined Cw,min, ␧ could then be Gc ⫽ seal clearance ratio 共=sc / b兲
calculated using the appropriate equations for EI or RI ingress. I ⫽ integrals for EI ingress
I⬘ ⫽ integrals for combined ingress
6 Conclusions ṁ ⫽ mass flow rate
The incompressible orifice equations have been solved for EI N ⫽ number of vanes
and combined ingress using a saw-tooth model to represent the p ⫽ static pressure
circumferential distribution of pressure in the annulus. Pmax ⫽ nondimensional pressure parameter
As for RI ingress, the analytical solutions for EI ingress show 共= 21 Cp,max Rew2兲
that the nondimensional ingress and egress and the sealing effec- r ⫽ radius
tiveness, ␧, depend on ⌰0 共which is the ratio of Cw,0, the nondi- ReW ⫽ axial Reynolds number in annulus 共=Wb / ␯兲

031006-6 / Vol. 133, JULY 2011 Transactions of the ASME

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Re␾
s
sc



rotational Reynolds number 共=␳⍀b2 / ␮兲
axial space between rotor and stator
seal clearance
Vr,i = Cd,i 再 2⌬p

共f − g兲 冎 1/2
共A5兲

u0 ⫽ average velocity of superposed flow when f ⱖ g 共i.e., 0 ⱕ ␪ ⱕ ␪⬘兲.


Vr ⫽ radial component of velocity Equations 共A4兲 and 共A5兲 can be integrated to give


W ⫽ axial velocity in external annulus 1
z ⫽ axial distance ṁe = 共2␳⌬p兲1/2AcCd,e 共g − f兲1/2d␪ 共A6兲
␣ ⫽ area ratio ␪⬘
␤ ⫽ swirl ratio 共=V␾ / ⍀r兲
⌫c ⫽ ratio of discharge coefficients 共=Cd,i / Cd,e兲 and


⌫p ⫽ pressure parameter 共=C p / C␤1兲 ␪⬘
⌫T ⫽ summation of ⌫ parameters ṁi = 共2␳⌬p兲1/2AcCd,i 共f − g兲1/2d␪ 共A7兲
⌫␤ ⫽ swirl parameter 共=C␤2 / C␤1 ⬇ 共␤2 / ␤1兲2兲 o
⌫⌬p ⫽ ratio of driving forces for EI and RI ingress In an engine ⌬p1/2 is proportional to rotational speed, and so it is
共=⌬C p / C␤1兲 useful to express these two equations as
⌬C p ⫽ external pressure coefficient 共=⌬p / 2 ␳⍀2b2兲
1

⌬p ⫽ maximum external pressure difference 共=p2,max ⌽e = Cd,e⌬C p1/2Ie 共A8兲


− p2,min兲 and
␧ ⫽ sealing effectiveness 共=1 − Cw,i / Cw,e兲
␩ ⫽ nondimensional axial distance 共=z / sc兲 ⌽i = Cd,i⌬C p1/2Ii 共A9兲
␪ ⫽ angular coordinate between vanes where
␪⬘ ⫽ value of ␪ where velocity is zero
⌰ ⫽ sealing flow ratio 共=Cw / Cw,min兲 ⌬p
␮ ⫽ ⌬C p = 共A10兲
dynamic viscosity 1/2␳⍀2b2
␯ ⫽ kinematic viscosity


␳ ⫽ density 1
␾ ⫽ angular coordinate in annulus Ie = 共g − f兲1/2d␪ 共A11兲
⌽ ⫽ nondimensional sealing parameter ␪⬘
共=Cw / 2␲Gc Re␾兲
and
⍀ ⫽ angular velocity of rotating disk
Subscripts
com
e


combined ingress
egress
Ii = 冕0
␪⬘
共f − g兲1/2d␪ 共A12兲

EI ⫽ externally induced ingress It should be noted that, as shown in Fig. 4, ␪⬘ is the value of ␪
i ⫽ ingress where Vr = 0.
max ⫽ maximum As ⌽0 = ⌽e = ⌽min when ⌽i = 0 共i.e., when g = 1兲, it follows that
min ⫽ minimum
o ⫽ superposed flow ⌽min = Cd,eImin⌬C p1/2 共A13兲
RI ⫽ rotationally induced ingress
where
1,2 ⫽ locations in wheel-space and annulus



⫽ value when Cw,0 = 0 1
Imin = 兵1 − f其1/2d␪ 共A14兲
o

Appendix A: Solution of Incompressible Orifice Equa- Using Eqs. 共A10兲–共A14兲,


tions For EI Ingress When Swirl Is Negligible
Ie
Define ⌰e = 共A15兲
Imin
p2 − p2,min
f共␪兲 = 共A1兲 Ii
⌬p ⌰i = ⌫c 共A16兲
Imin
and
and
p1 − p2,min
g= 共A2兲 I e − ⌫ cI i
⌬p ⌰0 = 共A17兲
Imin
for the range 0 ⱕ f共␪兲 ⱕ 1 and gⴱ ⱕ g ⱕ 1, where gⴱ is the value of
g when Cw,0 = 0 and ⌬p = p2,max − p2,min. Hence where ⌫c = Cd,i / Cd,e. As

p1 − p2 = ⌬p共g − f兲 共A3兲 ⌰i
␧=1− 共A18兲
⌰e
and so, from the equations for incompressible flow derived in Part
I, it follows from Eqs. 共A15兲 and 共A16兲 that

Vr,e = Cd,e 再 2⌬p



共g − f兲 冎 1/2
共A4兲 ␧ = 1 − ⌫c
Ii
Ie
共A19兲

when g ⱖ f 共i.e., ␪⬘ ⱕ ␪ ⱕ 1兲, and In general, the integrals in the above equations need to be

Journal of Turbomachinery JULY 2011, Vol. 133 / 031006-7

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use


computed numerically. However, as explained in Sec. 3.2, the ⌽min,com = Cd,eC␤ 1/2
共1 − ⌫␤兲1/2 共B11兲
1
simple saw-tooth model can be used to provide analytical solu-
tions.
which is identical to the result for RI ingress 共see Eq. 共B16兲 of
Appendix B in Part I兲. In the limit as ⌫⌬p → ⬁ Eq. 共B10兲 becomes

Appendix B: Solution of Saw-Tooth Model for Com- 2


bined Ingress ⌽min,com = Cd,e⌬C p1/2 共B12兲
3
For RI and EI ingress, the radial velocities are assumed to vary
with z and with ␪, respectively. For combined ingress, the problem which is identical to the result for EI ingress 共see Eq. 共A13兲
is two dimensional as the radial velocities will vary with both z above, where Imin = 2 / 3兲. It therefore follows that for EI ingress,
and ␪. The calculation of effectiveness therefore requires numeri- there is no effect of ⌫␤ on ⌽min. The special case where ⌫␤ = 0 is
cal integration over both z and ␪, which is beyond the scope of discussed further in Sec. 5.2.
this paper. A one-dimensional approach is used below to calculate
Cw,min; knowing this, the effectiveness could be estimated by us-
ing the appropriate EI or RI equations.
Referring to Eqs. 共3.7兲 and 共3.8兲 of Part I for inviscid flow,

冑 冑
References
Vr,e p 1⬘ − p 2 Vr,i p 2⬘ − p 1 关1兴 Abe, T., Kikuchi, J., and Takeuchi, H., 1979, “An Investigation of Turbine
= and = 共B1兲 Disk Cooling: Experimental Investigation and Observation of Hot Gas Flow
⍀b 1/2␳⍀2b2 ⍀b 1/2␳⍀b2 Into a Wheel Space,” 13th CIMAC Congress, Vienna, Paper No. GT30.
where 关2兴 Phadke, U. P., and Owen, J. M., 1988, “Aerodynamic Aspects of the Sealing of
Gas-Turbine Rotor-Stator Systems, Part 1: The Behaviour of Simple Shrouded
p1⬘ = p1 + 1/2␳⍀2b2C␤1, p2⬘ = p2 − 1/2␳⍀2b2C␤2 共B2兲 Rotating-Disk Systems in a Quiescent Environment,” Int. J. Heat Fluid Flow,
9, pp. 98–105.
关3兴 Phadke, U. P., and Owen, J. M., 1988, “Aerodynamic Aspects of the Sealing of
By analogy with Eqs. 共A1兲 and 共A2兲 above, define Gas-Turbine Rotor-Stator Systems, Part 2: The Performance of Simple Seals in
a Quasi-Axisymmetric External Flow,” Int. J. Heat Fluid Flow, 9, pp. 106–
p2⬘ − p2,min p1⬘ − p2,min
f ⬘共 ␪ 兲 = and g⬘ = 共B3兲 112.
关4兴 Phadke, U. P., and Owen, J. M., 1988, “Aerodynamic Aspects of the Sealing of
⌬p ⌬p
Gas-Turbine Rotor-Stator Systems, Part 3: The Effect of Nonaxisymmetric
Similarly, by comparison with Eqs. 共A8兲 and 共A9兲, External Flow on Seal Performance,” Int. J. Heat Fluid Flow, 9, pp. 113–117.
关5兴 Vaughan, C., 1987, “A Numerical Investigation Into the Effect of an External
⌽e = Cd,e⌬C p1/2Ie⬘ and ⌽i = Cd,i⌬C p1/2Ii⬘ 共B4兲 Flow Field on the Sealing of a Rotor-Stator Cavity,” D.Phil. thesis, University
of Sussex, UK.
where 关6兴 Hamabe, K., and Ishida, K., 1992, “Rim Seal Experiments and Analysis of a

冕 冕
Rotor-Stator System With Nonaxisymmetric Main Flow,” ASME Paper No.
1 ␪⬘ 92-GT-160.
I e⬘ = 共g⬘ − f兲1/2d␪ and I i⬘ = 共f ⬘ − g兲1/2d␪ 共B5兲 关7兴 Dadkhah, S., Turner, A. B., and Chew, J. W., 1992, “Performance of Radial
Clearance Rim Seals in Upstream Rotor-Stator Wheelspaces,” ASME J. Tur-
␪⬘ 0
bomach., 114, pp. 439–445.
关8兴 Green, T., and Turner, A. B., 1994, “Ingestion into the Upstream Wheelspace
Now ⌽i = 0 when p1 = p2,max − 2 ␳⍀2b2C␤2, which occurs when g⬘
1
of an Axial Turbine Stage,” ASME J. Turbomach., 116, pp. 327–332.
= 1 + 共C␤1 − C␤2兲 / ⌬C p. 关9兴 Chew, J. W., Green, T., and Turner, A. B., 1994, “Rim Sealing of Rotor-Stator
By comparison with eq 共A13兲, Wheelspaces in the Presence of External Flow,” ASME Paper No. 94-GT-126.
关10兴 Gentilhomme, O., Hills, N. J., Turner, A. B., and Chew, J. W., 2002, “Mea-
⌽min = Cd,eImin⬘⌬C p1/2 共B6兲 surement and Analysis of Ingestion Through a Rim Seal,” ASME Paper No.
GT2002-30481.
where 关11兴 Bohn, D., Johann, E., and Krüger, U., 1995, “Experimental and Numerical

冕再 冎 冕再 冎
Investigations of Aerodynamic Aspects of Hot Gas Ingestion in Rotor-Stator
1 1/2 1 Systems With Superimposed Cooling Mass Flow,” ASME Paper No. 95-GT-
C ␤1 − C ␤2 1 − ⌫␤ 1/2
Imin⬘ = 1+ −f d␪ = 1+ −f d␪ 143.
o
⌬C p o
⌫⌬p 关12兴 Bohn, D., Rudzinski, B., Sürken, N., and Gärtner, W., 1999, “Influence of Rim
Seal Geometry on Hot Gas Ingestion Into the Upstream Cavity of an Axial
共B7兲 Turbine Stage,” ASME Paper No. 99-GT-248.
关13兴 Bohn, D., Rudzinski, B., Sürken, N., and Gärtner, W., 2000, “Experimental
and and Numerical Investigation of the Influence of Rotor Blades on Hot Gas
Ingestion Into the Upstream Cavity of an Axial Turbine Stage,” ASME Paper
00-GT-284.
C ␤2 ⌬C p 关14兴 Bohn, D., and Wolff, M., 2003, “Improved Formulation to Determine Mini-
⌫␤ = , ⌫⌬p = 共B8兲
C ␤1 C ␤1 mum Sealing Flow—Cw, min—for Different Sealing Configuration,” ASME
Paper No. GT2003-38465.
Using the saw-tooth model 共where f is given by Eqs. 共3.2兲 and 关15兴 Bohn, D. E., Decker, A., Ma, H., and Wolff, M., 2003, “Influence of Sealing
Air Mass Flow on the Velocity Distribution in and Inside the Rim Seal of the
共3.3兲兲, Eq. 共B6兲 becomes

冕再 冎
Upstream Cavity of a 1.5-Stage Turbine,” ASME Paper No. GT2003-38459.
1/2 关16兴 Johnson, B. V., Mack, G. J., Paolillo, R. E., and Daniels, W. A., 1994, “Turbine
1 − ⌫␤ 1/2
⌽min = 2冑2Cd,e共C␤1⌫⌬p兲1/2
Rim Seal Gas Path Flow Ingestion Mechanisms,” AIAA Paper No. 94-2703.
+␪ d␪ 共B9兲 关17兴 Feiereisen, J. M., Paolillo, R. E., and Wagner, J., 2000, “UTRC Turbine Rim
o
2⌫⌬p Seal Ingestion and Platform Cooling Experiments,” AIAA Paper No. 2000-
3371.
Hence, assuming that ⌫␤ and ⌫⌬p are invariant with ␪ and that 关18兴 Roy, R. P., Feng, J., Narzary, D., Saurabh, P., and Paolillo, R. E., 2004, “Ex-
⌫␤ ⬍ 1, periments on Gas Ingestion Through Axial-Flow Turbine Rim Seals,” ASME
Paper No. GT2004-53394.
2 共1 − ⌫␤ + ⌫⌬p兲3/2 − 共1 − ⌫␤兲3/2 关19兴 Johnson, B.V., Jakoby, R., Bohn, D.E., and Cunat, D., 2006, “A Method for
⌽min,com = Cd,eC␤11/2 Estimating the Influence of Time-Dependent Vane and Blade Pressure Fields
3 ⌫⌬p on Turbine Rim Seal Ingestion,” ASME Paper No. GT2006-90853.
关20兴 Wang, C.-Z., Johnson, B. V., Cloud, D. F., Vashist, T. K., and Roy, R. P. 2006,
共B10兲 “Rim Seal Ingestion Characteristics for Axial Gap Rim Seals in a Closely-
Spaced Turbine Stage From a Numerical Simulation,” ASME Paper No.
where the subscript “com” is used to denote the combined-ingress GT2006-90965.
case. 关21兴 Johnson, B.V., Wang, C.-Z., and Roy, P. R., 2008, “A Rim Seal Orifice Model
In the limit when ⌫⌬p = 0 Eq. 共B10兲 becomes With Two Cds and Effect of Swirl in Seals,” ASME Paper No. GT2008-50650.

031006-8 / Vol. 133, JULY 2011 Transactions of the ASME

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use


关22兴 Cao, C., Chew, J. W., Millington, P. R., and Hogg, S. I., 2004, “Interaction of “Sealing in Turbomachinery,” NASA Report No. NASA/TM 2006-214341.
Rim Seal and Annulus Flows in Axial Flow Turbine,” ASME J. Eng. Gas 关25兴 Bayley, F. J., and Owen, J., 1970, “The Fluid Dynamics of a Shrouded Disk
Turbines Power, 126, pp. 786–793. System With a Radial Outflow of Coolant,” ASME J. Eng. Power, 92, pp.
关23兴 Mirzamoghadam, A. V., Heitland, G., Morris, M. C., Smoke, J., Malak, M., 335–341.
and Howe, J., 2008, “3D CFD Ingestion Evaluation of a High Pressure Turbine 关26兴 Graber, D. J., Daniels, W. A., and Johnson, B. V., 1987, “Disk Pumping Test,
Rim Seal Disk Cavity,” ASME Paper No. GT2008-50531. Final Report,” Air Force Wright Aeronautical Laboratories Report No.
关24兴 Chupp, R. E., Hendricks, R. C., Lattime, S. B., and Steinetz, B. M., 2006, AFWAL-TR-87- 2050.

Journal of Turbomachinery JULY 2011, Vol. 133 / 031006-9

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/19/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like