Flow Mechanism in Axial Turbine Rim Seals

Review Article
Proc IMechE Part C:

J Mechanical Engineering Science
Flow mechanisms in axial turbine 0(0) 1–21
! IMechE 2018
rim sealing Reprints and permissions:

sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0954406218784612
journals.sagepub.com/home/pic
John W Chew, Feng Gao and Donato M Palermo
Abstract
This paper presents a review of research on turbine rim sealing with emphasis placed on the underlying flow physics and
modelling capability. Rim seal flows play a crucial role in controlling engine disc temperatures but represent a loss from
the main engine power cycle and are associated with spoiling losses in the turbine. Elementary models that rely on
empirical validation and are currently used in design do not account for some of the known flow mechanisms, and
prediction of sealing performance with computational fluid dynamics has proved challenging. Computational fluid dynam-
ics and experimental studies have indicated important unsteady flow effects that explain some of the differences
identified in comparing predicted and measure sealing effectiveness. This review reveals some consistency of investiga-
tions across a range of configurations, with inertial waves in the rotating flow apparently interacting with other flow
mechanisms which include vane, blade and seal flow interactions; disc pumping and cavity flows; shear layer and other
instabilities; and turbulent mixing.
Keywords
Computational fluid dynamics, experiment, intrinsic flow unsteadiness, stability, turbine rim seals
Date received: 22 February 2018; accepted: 24 May 2018
Introduction
internal and film cooling for the turbine blades, dissi-
For every complex problem there is a solution that is pates the windage heating in the cavities, cools the
simple, neat and wrong. – H. L. Mencken discs and controls hot gas ingestion through the
gaps between the rotating blade and stationary vane
platforms. While sufficient air must be supplied to
While the authors would not agree with all of satisfy these requirements, it is also desirable to min-
Mencken’s statements, the above quote does seem imise bleed of cooling air from the compressor and to
especially relevant to rim sealing in axial turbines. reintroduce the air to the main gas path with min-
As will be discussed below, a number of simplified imum aerodynamic losses in the turbine. Relative
models of rim seal flows have been developed since movements of stationary and rotating parts during
research began in this field around 50 years ago. These operation constrain the clearance such that the rim
have had significant success in correlating experimen- sealing requirements may increase flow rates above
tal data, but it is still apparent that the flow physics is those otherwise needed. Without thorough under-
not fully understood. With recent progress having standing of rim seal flow mechanisms and reliable pre-
been made in this area, and heightened interest from dictive methods, manufacturers must rely on previous
the turbomachinery research and design communities, experience and development testing to design the rim
the present review aims to summarise and explain the seal geometry and set cooling flow rates.
current understanding of the flow physics, highlight- Perhaps the first openly published research on rim
ing areas where further research is needed. The review sealing was by Bayley and Owen.2 This presented
does not attempt to give a full account of the consid- equation (1a) as a correlation of experimental results
erable collection of publications available in this area.
Figure 1, from Rolls-Royce plc,1 illustrates the use
of rim sealing or purge flow to prevent or supress Thermo-Fluid Systems UTC, Faculty of Engineering and Physical
Science, University of Surrey, Guildford, UK
ingestion of hot main annulus flow gas into the cav-
ities or wheel spaces between the rotating turbine discs Corresponding author:
and stationary components. Here cooling air, bled Feng Gao, University of Surrey, Guildford GU2 7XH, UK.
from the high-pressure (HP) compressor, provides Email: [email protected]
2 Proc IMechE Part C: J Mechanical Engineering Science 0(0)
Figure 1. A hypothetical turbine cooling and sealing arrangement from Rolls-Royce plc.1 H.P: high pressure; L.P: low pressure.
for the minimum flow rate required to seal a rotor/ The apparent lack of dependence on viscosity is
stator disc cavity with a simple axial clearance at the surprising, considering that the disc pumping causing
outer radius and no external flow the ingestion is a viscous phenomenon. Without
viscosity there would be no disc pumping and no
Cwmin ¼ 0:61 Gc Re ð1aÞ ingestion. Rearranging the equation further and con-
sidering the scaling associated with disc pumping pro-
Here Cw ¼ m=b,_ Gc ¼ sc/b and Re ¼ b2/, vides an explanation for the paradox. Inspired by von
where m_ is the purge flow rate, is fluid viscosity, b Kármán’s3 momentum integral solution for free disc
is the outer disc radius, sc is the seal clearance, is the flow, which gives the radially outward flow in the disc
fluid density and is the angular velocity of the disc. boundary layer (at radius r ¼ b) as Cwfd ¼ 0.219Re0.8,
Cwmin denotes the minimum value of Cw required to a throughflow parameter l ¼ Cw/Re0.8 is defined.
avoid ingestion into the cavity. Use of the mass flow Bayley and Owen’s correlation may then be written
parameter Cw and rotational Reynolds number Re in as equation (1c)
equation (1a) disguises the fact that the minimum
sealing flow rate was found to be independent of lmin ¼ 0:61 Gc Re0:2
ð1cÞ
Reynolds number. Rearranging the equation in
terms of the mean velocity through the seal Um Noting also that von Kármán’s solution gives a
_
(¼ m=2s c b) gives equation (1b) disc boundary layer thickness proportional to
Re0.2 it can be seen that the flow rate scaled with
Um,min =b ¼ 0:0971 ð1bÞ disc pumping, lmin, is directly proportional to the seal
Chew et al. 3
clearance scaled with the disc boundary layer thick- cavity. These rotating, unsteady flow structures were
ness. Thus, the combination of mass flow and bound- unrelated to blade passing. Similar, but not always
ary layer thickness scaling for the free disc flow dominant effects have been identified in other studies.
accounts for the absence of viscosity in equation In the light of the new evidence available it is timely to
(1b). Further use of momentum integral methods review again the understanding of the physical mech-
has been made for rotationally driven ingestion,4 as anisms and consider future research directions.
will be discussed later in this review. A brief account of elementary modelling and evi-
For an outer seal in an axial turbine the flow exter- dence from correlation of test data for rim sealing is
nal to the cavity may be highly swirling and subject to given in the next section, focussing on flow physics
non-axisymmetric disturbances from the stationary rather than seal design. ‘Rotating flow modes and
vanes and rotating blades in main annulus. CFD modelling’ section of the report then focuses
Campbell5 and Abe et al.6 were amongst the first to on intrinsic flow unsteadiness and CFD. The rotating
discuss this in the open literature. The importance of cavity/seal modes had not been identified at the time
pressure asymmetries generated by the blades and of Johnson et al.’s review, and CFD capability has
vanes was highlighted, and Abe et al. reported experi- advanced significantly in recent years. Conclusions
mental measurements of ingestion for a disc cavity and the future outlook are summarised in the final
downstream of a row of vanes. These and many section.
other researchers have described how ingress can be
expected if the cavity pressure is lower than maximum Elementary modelling and correlation
pressure in the main annulus. Many researchers have
of test data
produced ‘orifice models’ of the ingestion process in
which the flow through the rim seal varies circumfer- Notwithstanding the complex flow physics in rim seals
entially and is estimated locally as for a simple orifice it is important to note experimental trends and where
with an assumed discharge coefficient. Such models models based on an elementary interpretation on the
have had considerable success in correlating ingestion flow have had success. Departure of such models from
measurements from rig tests, as well illustrated in a experimental data also gives an indication of when
recent review by Scobie et al.7 However, an alterna- other effects are important.
tive, turbulent transport model was proposed by
Graber et al.8 This concept has recently been further
Rotationally driven ingestion
developed by Savov and Atkins9 and also shown to be
able to correlate ingestion measurements. As stated by For a sufficiently wide rotor/stator disc cavity with
Savov and Atkins, these and all other low-order low purge flow, radial outflow (or pumping flow)
models in the literature ‘describe the variation in occurs in the rotor disc boundary layer. This bound-
seal performance with dimensionless flow rate for a ary layer flow is supplied by the purge flow and recir-
seal which has been experimentally characterised, and culation within the cavity which may include some
do not allow predictive capability, per se’. It has ingested flow. For simple geometries the flow in the
become generally accepted that for outer rim seals cavity (without ingestion) is reasonably well under-
externally driven (or pressure-driven) ingestion is stood and is illustrated in Figure 2. An elementary
likely to dominate over rotation-driven (or disc pump- model to deduce the minimum seal flow required to
ing) ingestion. However, it should be noted that it is prevent ingestion was presented in Chew4 and Chew
often difficult experimentally to separate the two et al.12 Using momentum integral methods to estimate
ingestion mechanisms, and that Reynolds number or
viscous effects may not be apparent for either
mechanism.
The complexity of rim seal flows was recognised in
a discussion of the physical mechanisms involved by
Johnson et al.,10 who considered data available in
1994. In addition to disc pumping and annulus flow
asymmetries discussed above, seal geometry, turbu-
lent transport in overlapping platform regions, eccen-
tricity and manufacturing tolerances emerged as
needing further consideration. Since 1994 a consider-
able number of experimental and computational fluid
dynamics (CFD) studies have been published and
understanding has developed further. A particularly
interesting development occurred in the early 2000s,
with, for example, Cao et al.11 showing experimental
evidence that large-scale unsteady flow features domi- Figure 2. Axisymmetric rot-stator cavity flow for
nated ingestion on a particular two-stage axial turbine Cw > Cwmin.4
Figure 3. Model coefficients for seals considered in Chew4

and Chew et al.12 Figure 4. Simple model of the pressure-driven ingestion
mechanism.
in radial clearance seals with extended overlap of

the disc boundary layer flow approaching the seal an rotating and stationary elements the ingestion charac-
expression was then derived for the pressure drop teristics would be dominated by rim geometry rather
across the seal, with the minimum sealing flow corres- than the disc boundary layer. Graber et al. derived an
ponding to the condition when the pressure drop expression of the form of equation (4), which was
equals zero. The model included an empirical factor supported by experimental measurements
(k) that was obtained for a number of different seals
by comparison with published experimental data. " ¼ 1 el ð4Þ
As shown in Figure 3, k was given for six different
seal styles. With an estimate of k for any seal, the Here is a constant for a particular seal and could
minimum sealing flow rate may be evaluated from a be determined experimentally. As noted by Chew
graph of lmin against GcRe0.2.4 At lower values of et al.,12 may be chosen so that differences in effect-
GcRe0.2 linear variation of lmin is observed as in iveness given by equations (3) and (4) are less than
equation (1c). The variation of k with seal type 0.035. Considering scatter in experimental data, and
appears consistent with expectations for rotating variation of sealing effectiveness within the cavity,
disc cavity flows.12 such differences may be considered small. Savov and
For flow rates lower than the sealing minimum, Atkins’9 turbulent transport model gives a further
ingestion occurs and the simple flow model cannot expression for sealing effectiveness involving one
apply. In experiments the ingestion has frequently empirical, seal dependent factor, and this was shown
been measured using gas concentration with the seal- to fit a range of cases that included rotationally driven
ing gas composition modified to differ from that in ingestion.
the external flow. A sealing effectiveness " is then Owen13 derived another model of rotationally
defined from the measured concentration c and induced ingestion based on an axisymmetric flow
values for the cavity inflow and external flow, as assumption through the seal and a radial equilibrium
follows equation that neglects the upstream radial velocity
component, contrasting with Chew.4 Expressions,
" ¼ ðcext cÞ=ðcext cin Þ ð2Þ depending on a number of parameters were produced
for minimum sealing flow and for sealing effective-
Based on published experimental measurements ness. The derived expression for minimum sealing
(which are usually taken on the stator disc surface) flow was not used in comparison with experiment
it was proposed that effectiveness could be estimated but the expression for sealing effectiveness was able
by assuming an ingress level about 20% of the short- to fit experimental data, similarly to equations (3) and
fall in flow rate from the minimum sealing flow, giving (4) and Savov and Atkins’ expression.
equation (3)
" ¼ Cw=ð0:8Cw þ 0:2Cwmin Þ ð3Þ

Pressure-driven ingestion
The importance of external flow, and particularly the
This equation was shown to give a reasonable fit to associated circumferential pressure variations was
measurements from a number of sources. established in early experimental studies by, for exam-
A model of ingestion, based on turbulent trans- ple, Abe et al.,6 Kobayashi et al.,14 Phadke and
port, was derived by Graber et al.8 who argued that Owen,15 Dadkhah et al.16 and Hamabe and Ishida.17
Chew et al. 5
The pressure-driven mechanism for ingestion is shown coefficients for inflow and outflow in Owen’s23 orifice
in Figure 4. This also gives the basis for a simple ori- model. Owen notes that the form of variation of
fice model in which the local seal velocity is estimated effectiveness with flow rate is similar for rotationally
from 1D orifice flow theory, with integration circum- driven and pressure-driven ingestion. These and other
ferentially to obtain the overall inflow and outflow. studies, including Savov and Atkins’9 application of
Hamabe and Ishida compared their experimental their turbulent transport model, confirm that charac-
measurements of ingestion to results from an orifice terising the seal flow as a velocity ratio is helpful in
model. The model neglected swirl velocity, integrating correlating and scaling experimental data for sealing
circumferentially with an assumed external pressure effectiveness. However, the elementary models have
variation, and a specified discharge coefficient. limited predictive capability.
Square wave, sinusoidal and triangular wave func-
tions were considered for the pressure variation, as
Multi-mode ingestion
well as the measured profile. Agreement with meas-
urements was shown, and similarity of the predicted The turbulent transport modelling approach8,9 aver-
variation of sealing effectiveness with flow rate to that ages all the flow effects and represents these through
given by equation (4) was noted. an empirically determined factor. This approach has
The orifice model was further considered by Chew perhaps become more attractive as the complexity and
et al.18 alongside experimental and CFD results. intrinsic unsteadiness of some rim seal flows has
Measurements of sealing effectiveness showed conver- emerged. In Savov and Atkins’9 model account is
gence of data when plotted against the mean velocity taken of seal geometry and flow rate through a
through the seal divided by an external flow velocity. volume fraction and blowing ratio, with a turbulence
Using discharge coefficients from measurements with- mixing length being adjusted to match experimental
out rotation, rather than selecting model parameters data. The model has been shown to fit a range of
to fit the data, it was found that the orifice model experimental data for rotationally driven and pres-
predicted some experimental trends but overestimated sure-driven ingestion. It should be noted that rim
ingestion. The neglect of inertial effects associated seal experiments may be subject to manufacturing tol-
with the swirl velocity component was considered to erances and other uncertainties. In their experimental
be a contributing factor. Further developments of the study Savov and Atkins compared results obtained
orifice model, including vane and blade generated using a 16-piece static rim seal ring assembly with
pressure asymmetries and inclusion of inertial effects those obtained using a single smooth ring.
associated with the swirl velocity are described in Hills Measurements showed a much better sealing perform-
et al.19 and Gentilhomme et al.20 Experimental trends ance for the single machined ring, with the effect
were reproduced by the model, but it was noted that attributed to the steps introduced in the 16-piece
adjustment of the ‘model constants’ was needed to assembly. This result clearly shows an important sen-
match different test cases. Other factors noted in sitivity to geometric tolerances and the care needed in
these studies included possible sensitivities to vane interpretation of results and prediction of seal
and blade design and flow separation, vane/blade performance.
flow interactions, Reynolds number effects in annulus, Owen23 gives an expression for the minimum seal-
seal and cavity flows, and possible unsteady inter- ing flow rate that reproduces his rotationally driven
actions of the sealing and annulus flows. Johnson and pressure-driven results in the appropriate limits,
et al.21 summarised the model in Hills et al.19 and noting that there are insufficient data available to test
Gentilhomme et al.20 as having ‘modest success’, con- the model. The model, with an empirical constant,
cluding that improved models would be useful. was later used by Scobie et al.7 to correlate data for
Calibration of orifice models against experimental the minimum sealing flow rate with variable external
data sets, as described by Scobie et al.,7 Johnson pressure asymmetry.
et al.21 and Scanlon et al.22 has been undertaken to Most recent research has focussed on external flow
support use of such models in design calculations. or pressure-driven ingestion. This has been generally
Scanlon et al. applied potential flow approximations thought to dominate outer rim seals, whereas rota-
in estimating annulus pressure variations and esti- tionally driven ingress may dominate at inner seals
mated discharge coefficients for different seals by fit- in double seal designs. Hamabe and Ishida17 and
ting the orifice model to ingestion measurements. The Chew et al.18 discussed simple order of magnitude
model was found to collapse data well apart from estimates for the two mechanisms. With pumping
conditions where unsteady pressure fluctuations effects characterised by a radial velocity in the disc
were measured in the cavity. Johnson et al. included boundary layer of order 0.1b, rotation might be
a brief review of discharge coefficient data in their expected to have little effect on the seal flow for
paper and recommended use of different discharge small 0.0052b2/Dp where Dp is the magnitude of
coefficients for inflow and outflow. In fitting an exten- the circumferential pressure asymmetry. Under these
sive range of experimental data Scobie et al. adjusted conditions rotational effects are still significant for the
the minimum sealing flow and ratio of discharge disc cavity flow, but may act principally to mix the
ingested and sealing flows. For ‘short seals’, such as velocity Um/b, frequency of the unsteady flow
those in Figure 3, the surface areas in the seal may be modes f/, number of circumferential lobes observed
sufficiently small for the surface drag on the rotor and in the unsteady flow pattern N and angular velocity of
stator to be negligible in the seal flow. Counter to this the pattern !s. Note that the values given are intended
argument is the observation in, for example, Cao to be indicative only. In some cases they are estimated
et al.11 and Jakoby et al.24 that rotating cavity, seal indirectly from the published information. These
and annulus flows may be subject to 3D rotating flow results are discussed further in the following
modes that significantly affect ingestion. For more subsections.
extended seals involving overlapping rotor and The velocity ratio Um/b is used here to character-
stator parts, the rotation and surface drag may also ise purge flow rate as it has been shown to be a useful
affect the flow in the seals as, for example, suggested correlating parameter for both rotationally driven and
by Graber et al. and supported by computations for a pressure-driven ingestion. This differs from many of
chute rim seal in Boudet et al.25 With similar func- the publications cited which often give values of the
tional dependence of ingestion on seal flow rate mass flow parameter Cw. The parameters given in the
observed across a wide range of experimental condi- tables do not characterise the main annulus flow
tions it is difficult to distinguish distinct regimes or to (where present) beyond giving blade and vane num-
separate the effects of individual parameters in these bers. Some previous publications have used an annu-
complex flows. lus flow velocity rather than disc speed to normalise
the seal flow. This may relate more directly to the
driving force in pressure-driven ingestion, but is not
Rotating flow modes and CFD modelling
convenient for cases with no annulus flow. A further
While the elementary modelling described above gives indication of the annulus flow, if present, is given by
some insights into the flow physics, detailed investiga- an annulus flow coefficient or Rossby number Ue /b
tion requires experimental and CFD studies. (where Ue is the mean axial velocity) in some later
Although there are a substantial and growing figure captions where this was available or could be
number of research studies available, these are con- estimated. It should be noted that where vanes are
strained by the complexity of the problem and the fitted the annulus flow will generally have a tangential
wide variety of geometries and conditions of interest. velocity significantly greater than disc speed. A fuller
Comparisons between CFD and measurements are characterisation would also include purge to annulus
subject to experimental uncertainties (including man- flow density ratio, but this is generally modest in the
ufacturing tolerances) and modelling uncertainties research rigs (compared to engine conditions).
(including turbulence modelling and boundary condi-
tions). Such comparisons have shown mixed levels of
agreement, as shown in the literature.11,19,20,24–31
Experimental evidence
While conventional Reynolds-averaged Navier– Experiments by Cao et al.,11 Jakoby et al.,24 Schädler
Stokes (RANS) models sometimes give quantitative et al.,34 Roy et al.,35 Savov et al.,36 Town et al.37 and
or qualitative trends in reasonable agreement with Beard et al.42 have all indicated unsteady rim seal
measurements, considerable uncertainties remain. cavity flow features which are unrelated to blade pas-
These CFD and experimental studies have revealed sing. Figure 5 includes these results with blue squares
the occurrence of rotating flow modes that have been or bars indicating the approximate observed fre-
attributed to either intrinsic unsteadiness of the rim quency range. Owing, perhaps, to the variation in
seal flows or interactions of the sealing and annulus or seal geometries considered it is difficult to identify
cavity flows. In some conditions it is clear that these any clear trends. It may be significant that simple
modes play a dominant role in ingestion. In other higher clearance axial seals show generally low fre-
cases their significance relative to other effects, such quencies, and the higher frequency results are for
as vane and blade pressure generated asymmetries, is more complex tighter clearance seals. Note though
not clear. Understanding of the intrinsic unsteadiness that Town et al.’s results37 do not fit this trend.
has emerged as a key area of research and is the focus Dependency on cavity volume might be expected,
of this section. Figure 5 and Tables 1 to 3 give an but it is again difficult to see any clear trend. Note
indication of the research available in the area, sum- also that Savov et al. observed unsteadiness in two
marising many of the experimental and computa- different frequency ranges. Further discussion of the
tional studies that have identified non-blade passing experimental studies is given in the following subsec-
related unsteadiness believed to be associated with tions, classifying the seals as axial, radial or chute
rotating flow modes. In Figure 5 the principal fre- according to the minimum clearance.
quency of the unsteadiness normalised with the
rotor frequency is plotted against the minimum clear- Axial rim seals. As indicated in Figure 1 and Table 1,
ance between the rotor and stator in the seal. Tables 1 axial clearance geometries have been studied experi-
to 3 give further information including, where avail- mentally by Cao et al.,11 Jakoby et al.24 and Schädler
able, rotational Reynolds number Re, mean seal et al.34 The seal and cavity geometries were quite
Chew et al. 7
Figure 5. Summary plot of experimental and CFD evidence for non-blade passing related rotating flow modes. CFD: computational
fluid dynamics.
Table 1. Summary of axial clearance seal studies showing intrinsic unsteady flow.
Reference Method, sector NGV/blade Re 103 sc =b 103 Um =ðbÞ f = N !s =

11
Cao et al. Experiment 50/67 5 7 106 13.1 11–13 10–16 11–14 100%
URANS, 90 , 360 0/0 6 106 13.1 12 8–12 6–12 90–97%
6.6 24 – 22 –
Boudet et al.25 URANS, 90 0/0 1:6 106 13.1 12 1.7 56 3%
Jakoby et al.24 Experiment 16/32 2:2 106 29.6 17–37 2.7 – –
URANS, 360 0/0 17 2.4 3 80%
Rabs et al.32 URANS, 22:5 0/0 2:2 106 29.6 17–73 – – –
Zhang and Ma33 URANS, 22:5 1/2 2:2 106 29.6 37–61 13–20 – –
Schädler et al.34 Experiment 36/54 3 105 6.1 – 7–19 – –
URANS, 360 – 6.6 8 82%
– 21 22 93%
NGV: nozzle guide vane; URANS: unsteady Reynolds averaged Navier–Stokes equation.
different in these three cases, but all these cases were rotor speed increases, frequencies of the related flow
tested within turbine stages with a high swirling annu- modes increase proportionally whereas the non-blade
lus flow. In all three cases, fast response instrumenta- passing related frequencies only increase slightly.
tion indicated unsteadiness at non-blade passing Further analysis of the signal indicated intermittency
frequencies, in a relatively low-frequency range of of wave patterns. There was a very low purge flow
3–19 . Cao et al.’s geometry might be viewed as rate in these experiments, corresponding to 10%
an open narrow disc cavity rather than a rim seal. of free disc entrainment at the rig conditions.
In this experiment, fast response pressures measure- Jakoby et al.’s24 paper principally reports CFD
ments on the face of the stator disc were recorded for studies but includes results from the University of
a range of rotor speeds as shown in Figure 6. As the Aachen for the unsteady pressure in the disc cavity
Table 2. Summary of radial clearance seal studies showing intrinsic unsteady flow.
Reference Method, sector NGV/blade Re 103 sc =b 103 Um =ðbÞ f = N !s =
Roy et al.35 Experiment 22/28 5 9 105 10.2 0–320 22 – –

Wang et al.29 URANS, 360 22/28 5:9 105 10.2 42–83 15–17 15–17 100%
42 10 12 86%
225 28 28 100%
Mirzamaghodam et al.30 Experiment 22/28 6:1 105 13.3 30–120 – 4–8 –
URANS, 360 – 8–13 –
Savov et al.36 Experiment 40/96, 40/0 2 8 106 3.3 3–190 3–5, 25–40 – –
Town et al.37 Experiment 29/36 1:1 106 8.9 45 5–15 15 78%
URANS, 99:3 8/10 12 14.5 82%
Julien et al.38 URANS, 74 9/12 2 106 6 0 26 30 87%
50 24 30 80%
200 22 35 63%
Boutet-Blais et al.39 URANS, 180 0/0 0:2 1 106 – – – 24,28 –
URANS, 24 3/4 7 4–80 23 29 79%
Chilla et al.40 URANS, 36 4/7 – – – >70 – –
URANS, 5:1 0/1 – – – – – –
Berg et al.41 Experiment 48/96 2 4 106 29, 71, 112
3:9 106 35, 128, 332
NGV: nozzle guide vane; URANS: unsteady Reynolds averaged Navier–Stokes equation.
Table 3. Summary of chute seal studies showing intrinsic unsteady flow.
Reference Method, sector NGV/blade Re 103 sc =b 103 Um =b f = N !s =

42
Beard et al. Experiment 0/0 2 3 106 4.2, 7.0 0–60 20–23 26–29 79–80%
Gao et al.43–45 URANS, 30 0/0 2:2 106 4.2 0 16, 22 36, 48 45%
LES, 13:33 0 23, 36 54, 81 44%
Boudet et al.25,26 URANS, 120 0/0 1:8 106 6.5 63 17 21 80%
URANS, 13:3 1/2 24, 6.5, 30 – –
O’Mahoney et al.27,28 URANS, 26:7 2/4 1:8 106 6.5 63 24, 6.5, 30 – –
URANS, 360 27/54 24, 6.5, 30 – –
LES, 13:3 1/2 2.2 106 67 32, 22, 35 – –
LES, 40 3/6 32 – –
LES: large eddy simulation; NGV:; URANS: unsteady Reynolds averaged Navier–Stokes equation.
with a simple axial clearance seal. Low-frequency har- effect on aerodynamic pressure losses was high-
monics at about one order of magnitude below the lighted, lessening the expected reduction in losses
rotor blade passing (corresponding to 3 ) fre- as purge flow rate was reduced. The aerodynamic
quency were detected in the disc cavity, as is shown measurements in the annulus show interaction of
in Figure 7. This condition corresponds to a relatively the purge flow with secondary flow generated by
low flow rate. Low-frequency content was not the upstream vane. The unsteady modes were also
observed at higher flow rates. Jakoby et al. concluded considered to have a significant effect on noise gen-
that the low-frequency phenomena significantly influ- eration for the turbine. Schädler et al. observed that
enced the mainstream ingestion. the low frequencies occurred as a band of frequen-
Schädler et al.34 reported unsteady flow measure- cies, as was the case for the Aachen tests in
ments in a stage and a half turbine with a more Figure 7, and concluded that this implied the
geometrically complex axial rim seal than those con- modes were not linked to a geometrically triggered
sidered by Cao et al. and Jakoby et al. ‘Hub cavity acoustic mode.
modes’ in the frequency range 7–19 were The three independent experimental studies for the
detected and found to propagate up to 30% of axial clearance seals provide strong evidence of intrin-
span in the main flow annulus. These modes were sic unsteadiness of rim seal flows at low purge flow
suppressed at higher purge flow rates and their rates. The studies also indicate that this unsteadiness
Chew et al. 9
Figure 6. Fourier analysis results for a pressure transducer on the stator disc in Cao et al.’s experiment, Re 6 106,
Um/b 0.012, Ue/b 0.4.11
Figure 7. Frequency spectrum for pressure in the front cavity of a test turbine at the University of Aachen, Re 2 106,
Um/b 0.017.24
has significant implications for rim seal ingestion and of these harmonics decreases as the dimensionless
aerodynamic performance. coolant mass flow rate Um/b rises, whereas the
energy content of blade passing frequency increases
Radial rim seals. Radial rim seals were studied experi- with Um/b. At low Um/b the low frequency and
mentally by Roy et al.,35 Savov et al.36 and Town blade passing frequencies have similar amplitudes.
et al.,37 as shown in Figure 5 and Table 2. The seal Roy et al. conjectured that the low-frequency signal
and cavity geometries were quite different in these may correspond to the rotating flow mode identified
cases, but all tests were conducted with a highly swir- by Cao et al. For a modified version of the rig,
ling annulus flow. Higher non-dimensional frequen- included an inner seal, Mirzamaghodam et al.30 later
cies values were found by Roy et al. and Savov reported that PIV velocity measurements showed
et al. compared to the axial seal clearance experi- variation in time with either one or two main
ments. However, Savov et al. found two separate fre- ingestion zones appearing in an approximately 90
quency ranges with lower frequency activity in the sector.
range 3–5 . Town et al.’s measurements show similar Savov et al.36 studied the performance of two rim
frequencies to those of the axial clearance seals. seal geometries varying the dimensionless seal purge
The earliest report of subblade passing frequency flow rate, for both bladed and unbladed configur-
unsteadiness in radial rim seals was given by Roy ations. Fast response pressure transducers were
et al.35 In this experiment unsteady pressure measure- fitted to the stator in the main disc cavity, the outer
ments were recorded within the cavity up to the rim rim seal cavity or trench (radially outboard of the
seal region at different purge flow conditions. In minimum clearance), and in the cavity feeding the
Figure 8, a graph showing the auto-spectral density purge flow. Experiments were conducted with vanes
of blade pressure measurements in the rim seal region only fitted (upstream of the seal) and with the add-
is reproduced. A low-peak energy harmonic detected ition of rotating tear drop shaped elements (down-
at 22 (1090 Hz), which is about 78% of the rotor stream of the seal) to simulate the blade pressure
blade passing frequency, can be identified. field. Unsteadiness in the ranges 25–35 and 30–40
Furthermore, it is noteworthy that the energy content was observed in the outer rim seal cavity or trench for
900
cw = 0 (1400 Hz)
800 cw = 1574
cw = 8656
700 cw = 14952
APSD (Pa2/Hz)
600
500
400
300
200
100
(1090 Hz)
0
900 1000 1100 1200 1300 1400 1500 1600
Frequency (Hz)
Figure 8. Auto-power spectral density function for the stator rim seal reported by Roy et al., Re 7 105, Um/b 0, 0.03, 0.18,
0.32, Ue/b 0.6.35
the two seals. This unsteadiness was reduced with the Chute rim seals. While earlier CFD studies had indi-
addition of model blades and was attributed by Savov cated intrinsic unsteadiness in chute seals, this was
et al. to Kelvin–Helmholtz type shear layer instability confirmed experimentally by Beard et al.42
arising from interaction of the purge and mainstream The configuration tested had a relatively simple
flows. At off-design conditions this unsteadiness was chute seal geometry with no imposed external flow
considered to have an important influence on inges- and no blades or vanes in the annulus surrounding
tion. Savov et al. also observed intermittent unsteadi- the disc cavity. Thus, rotationally driven ingestion
ness in the 3–5 frequency range. This was compared was expected. Unsteady pressure sensors in the
to low-frequency unsteadiness observed by Cao et al. cavity showed a distinct signal at 22 , with ampli-
and its effect on ingestion was said to be not tude diminishing slightly with increasing radius. This
understood. mode was still evident just under the rim seal. In the
Town et al.37 reported measurements in the buffer rim seal region, a broader band of unsteadiness with
cavity formed between two radial clearance seals in a increased amplitude was observed. Similar trends
turbine test rig. Two fast response pressure trans- were observed by Beard et al. at all speeds, and for
ducers were spaced 5 apart. Unsteadiness in the fre- a larger seal clearance for which there was very little
quency range 5–15 was observed with analysis of axial overlap between the two surfaces of the chute.
the phase of the signals suggesting a flow pattern with While Taylor–Couette instability might contribute to
15 structures circumferentially rotating at 77.5% of the flow unsteadiness in the chute seal, results for the
rotor speed. As the observed unsteadiness covers a larger clearance indicate that this is not needed to
broad range, some intermittency or variation of the excite the cavity mode.
flow pattern could be inferred. Beard et al.42 and Gao et al.43 analysed results
While the above-mentioned radial seal studies pro- from an unevenly distributed circumferential array
vide further evidence of rotating flow modes, the sig- of sensors to investigate the flow structure in the
nificance of these modes for rim sealing and turbine cavity. An example of the cross correlation between
performance is not so clear as in axial rim seal studies. two sensors is shown in Figure 9(b). In this case the
Savov et al.’s study suggests the possibility of different angle between the sensors is 5 . The peaks appearing
modes associated with either the purge flow/main- with relative high frequency are consistent with a
stream interaction or the seal and cavity flows, lobed flow pattern, as in Figure 9(a), rotating relative
although this is somewhat speculative. The uncer- to the sensors. The time interval between peaks cor-
tainty regarding the flow physics is further illustrated responds to the time taken for the mode to rotate one
by Berg et al.’s41 study. These workers attribute an angular period ( ¼ 2/N for a mode with N lobes).
unsteady signal at 29 in their experiments on a The times for each peak (or lag times) correspond to
radial seal to an axisymmetric Helmholtz mode. the time intervals between similar events occurring at
Further signals at 71 and 112 are attributed to the two probes. Thus, for a perfectly symmetrical
axisymmetric shallow cavity modes. While their inter- rotating mode with N lobes and > , as in
pretation was supported by comparison with results Figure 9(a), the lag time for the first peak is the
for canonical configurations, the possibility of rotat- time taken for one lobe to pass from one sensor to
ing flow modes causing these signals was not the next. Lower frequency effects apparent in the
eliminated. figure may be associated with variation or
Chew et al. 11
(a)
sensor 2
sensor 1
Stator Rotor
Flow structures
(b)
Figure 10. Instantaneous pressure contours on the rotor
surface, Re ¼ 1.6 106, Um/b 0.012, Ue/b 0, Autef.46
inherent unsteadiness indicating further complexity

of the flow physics and challenging the assumptions
used in elementary models. The discussion is orga-
nised in terms of seal type, as was the case for experi-
mental studies in ‘Experimental evidence’ section.
Axial rim seals. The first evidence of inherent unsteady

flow features in rim seals was, to the present authors’
knowledge, CFD predictions reported by Autef,46 as
mentioned by Chew et al.47 A simple axial seal con-
Figure 9. (a) A hypothetical rotating flow structure and (b) figuration (without vanes and blades) was simulated
cross correlation of pressure apart over one rotor revolution with 2D RANS and 3D URANS models with a weak
for two sensors 5 apart, Re 3 106 , Um/b ¼ Ue/b ¼ 0.42 axisymmetric cross flow in the outer annulus. In com-
parison with the 2D axisymmetric RANS model, the
3D URANS model that predicted rotating flow struc-
intermittency of the flow pattern and are sensitive to tures in the rim seal achieved better agreement with an
the width of the filter applied in processing the results. empirically based sealing effectiveness correlation for
Further analysis over multiple time periods indicated rotationally driven ingestion.12 These results were
a most probable flow structure for this case of a 29- later published in Boudet et al.25 The rotating flow
lobed flow structure rotating at 80% of disc speed. structure illustrated had 56 lobes rotating slowly and
Similar results were obtained at other flow conditions. giving a distinct frequency of 1.5 for pressure fluc-
tuations. Figure 10 shows pressure contours on the
rotor surface with alternating low and HP regions
RANS CFD modelling
appearing in the seal on the rotor lip.
CFD modelling has largely focussed on pressure- Cao et al.11 reported RANS and URANS solu-
driven ingestion, following the pattern of experimen- tions, comparing with measurements from the
tal studies in moving from relatively simple models to ALSTOM UK’s two-stage turbine rig. Models for a
more complete representations. This review discusses 90 sector with circumferential periodicity and a full
unsteady CFD models as steady models cannot (in 360 sector without vanes and blades were presented.
general) capture the combined effects of blades and URANS simulations were performed as RANS solu-
vanes or the intrinsic unsteadiness. An early applica- tions showed poor convergence. The URANS solu-
tion of unsteady Reynolds-averaged Navier–Stokes tions showed good agreement with mean cavity
(URANS) modelling, including stationary vanes, pressure measurements and predicted unsteady flow
rotating ‘pegs’ (used to simulate the blades in an asso- patterns with modes unrelated to the main annulus
ciated experiment), and the disc cavity, was given by blade passing. Examples of radial velocity and
Hills et al.19 The unsteady model predicted consider- pressure contours on an axial plane are shown in
ably more ingestion than steady CFD models, but still Figure 11. These 12-lobed patterns rotated at slightly
gave somewhat higher sealing effectiveness than was less than disc speed. The 360 URANS models
measured. This and subsequent studies have illu- also showed the scale and strength of unsteady struc-
strated effects such as interaction of the vane and tures reducing, with the number of lobes increasing,
blade potential pressure fields, axial decay of the pres- as the axial seal gap is reduced. As a result of the
sure asymmetries upstream and downstream of the CFD predictions, fast response instrumentation
vanes or blades, and inertial effects. As will be was fitted to the rig as described in ‘Axial rim seals’
described below, some studies have also identified section. The measurements showed unsteadiness
Figure 11. Instantaneous radial velocities (left 70 to 50 m/s) and pressure (right 15 to 15 kPa), Re 6 106, Um/b 0.012,
Ue/b 0.4, Cao et al.11
at similar frequencies and amplitude to the URANS predicted unsteady flow modes were not compared
solutions. with unsteady experimental data. The predicted vor-
The above studies were concurrent with, but inde- tices were strongest at high purge flow rates.
pendent of, the ICAS-GT and ICAS-GT2 research With care taken to resolve the rim seal region, the
programmes reported by Smout et al.48 Smout et al. sealing effectiveness in the outer part of the cavity
stated that fast response pressure measurements on a was lower than that given by Volvo’s 22.5 3D
turbine rim sealing rig designed by Bohn et al.49 at model24 but still well above the measurements at
RWTH Aachen University had shown low-frequency lower cavity radii. Rabs et al. also noted that similar
unsteadiness unrelated to blade passing. The presence results were obtained with different turbulence
of large-scale unsteady structures was suggested as a models. Zhang and Ma33 investigated the effect of a
possible cause of discrepancies between measurements circumferentially varying seal clearance, with a 22:5
and URANS modelling for a 22.5 periodic sector. sector including one vane and two blades. Unsteady
Results of CFD studies from the ICAS-GT2 pro- flow features were captured with low frequency from
ject were reported by Jakoby et al.24 These included 13.3 to 20 for both uniform and varying seal
CFD solutions from three project partners, ALSTOM clearances.
(Switzerland), Volvo (Sweden) and MTU (Germany), Schädler et al.34 complemented their experimental
simulating Bohn et al.’s 1.5-stage turbine rig. studies with a full 360 annulus URANS model
ALSTOM’s 22:5 and 360 models did not include including 36 vanes and 54 blades. Low-frequency
vanes and blades, whereas Volvo and MTU made unsteadiness was captured but showed a broader fre-
22:5 full models with vans, blades. Only quency band than the measurements. The experiment
ALSTOM’s 360 URANS model was able to capture showed unsteady flow frequency at 7–19 . At a low
a dominant low-frequency mode at 2.4 , close to purge flow condition, unsteady flow patterns with
the measurement at 2.7 . As shown in Figure 12, eight lobes were detected, rotating at 82% disc
the solution showed a three-lobed unsteady flow speed. At a higher purge flow rate a 22-lobed flow
structure rotating at 80% of disc speed in the main structure was found rotating at 93% of disc speed.
disc cavity. Note also that Figure 12 shows some evi- Schädler et al. reported that increasing the purge
dence of flow structures in the seal having additional mass flow rate can decrease the hub cavity mode fre-
components. The predicted ingestion was less than quency amplitude ratio to that of the blade
measured but significantly greater than small sector passing and increase the frequency. They attributed
models. The failure of the sector models was attribu- the source of the inherent unsteady flow modes to
ted to the periodic boundary assumption suppressing the vortical structures in the outer trough (or cavity)
the large-scale flow structure. of the seal.
More recent CFD studies have also considered The CFD studies for axial seals show that URANS
Bohn et al.’s rig. Amongst these are investigations models can reproduce many of the flow features
by Rabs et al.32 and Zhang and Ma.33 Rabs et al.32 observed experimentally although some discrepancies
performed URANS simulation on a 22:5 sector with- between CFD and measurement remain. The CFD
out vanes and blades. They associated predicted solutions have also given more insight into the flow
unsteady flow modes unrelated to the blade passing physics, showing the unsteadiness is associated with
with the Kelvin–Helmholtz shear layer instability rotating flow modes. There is evidence of flow struc-
in the rim seal gap region, where the main annulus tures associated with the seal or trough region in all
flow interacts with the purge flow. However, the configurations considered above. For the Aachen rig
Chew et al. 13
Figure 12. Instantaneous pressure, Re 2 106, Um/b 0.017, Jakoby et al.24
Figure 13. Instantaneous pressure, Re 6 105, Um/b 004, Ue/b 0.8, Wang et al.29
there is also evidence of a three-lobed, large-scale flow was achieved in the 360 CFD model which overesti-
pattern in the main disc cavity. mated ingestion compared to the experiment.
Figure 13 shows pressure contours for a low flow
Radial rim seals. Zhou et al.50 reported URANS solu- case. The influence of the 22 vanes is clear in the
tions for a 14.4 sector model with blades and vanes outer annulus, the outer seal rotating mode is also
for a simple double seal configuration tested at visible and a less regular structure is seen in the
Arizona State University.35 Ingestion was shown to outer disc cavity. Further results presented by Wang
be underpredicted compared to measurements and et al. show a four-lobed flow structure in the inner
the authors suggested that this may be due to rotating disc cavity.
low-pressure zones being supressed by the small sector Mirzamaghodam et al.30 also reported 360
periodicity. In a related further investigation Wang URANS solutions for the Arizona State turbine rig
et al.29 used a full 360 URANS model for the same but with modified seal geometry. The clearance and
configuration. They reported 15–17 pairs of unsteady overlap of the outer radial seal differed in the two
flow structures rotating at disc angular speed in the cases, and the inner axial clearance seal for Wang
outer seal gap for the two modest purge flow rates. et al.’s case was replaced by a much tighter radial
For a higher purge flow rate 28 pairs of flow features seal in Mirzamaghodam et al.’s case. These authors
rotated at about disc speed. In the cavity inboard of highlighted the slow evolution of the flow and sug-
the seal, species concentration contours revealed 12- gested this may lead to overprediction of ingestion
lobed structures rotating at about 86% of disc speed by the CFD. Figure 14 shows radial velocity contours
for the two low purge flow cases. Better agreement on at different times with varying number of ingress
the sealing effectiveness with the experimental data regions.
Figure 14. Radial velocity at seven revolutions (left) and 16 revolutions (right), Re 6 106, Um/b 0.03, Ue/b 0.8,
Mirzamaghodam et al.30
Town et al.37 compared URANS solution for a 99 and 24.8 in the rotating frame. The existence of the
sector with blades and vanes with their experimental inherent unsteady flow structures was believed to
measurements at Pennsylvania State University. The increase the loss generation in the turbine. By increas-
computations and unsteady pressure measurements ing the sealing mass flow rate and sealing tangential
were consistent with about 15 flow cells rotating at velocity independently, it was found that the rim seal
about 80% disc speed in their shallow disc cavity. flow became more stable as ingestion was prevented
Julien et al.48 reported a URANS study based on and the tangential velocity difference between the two
an industrial test rig geometry described by Feiereisen flows was minimised. They associated the rim seal
et al.51 As shown in Figure 5, a radial seal is combined unsteady flow patterns with Kelvin–Helmholtz
with an outer chute region. A 74 sector was used vortex formation due to the interaction of the main
including nine vanes and 12 blades. Unsteady cavity annulus and rim seal flows.
modes were reported at 26, 24 and 22 for no purge The URANS studies for radial seals have provided
flow, low purge flow and higher purge flow cases, further evidence of very significant effects of rotating
respectively. It was reported that these cavity modes flow modes on ingestion and on turbine performance,
were dominant compared with the blade passing for a range of configurations including engine repre-
mode, and that the strengths and characteristic fre- sentative geometries. In addition, Wang et al.’s results
quencies of the cavity flow structures were suppressed indicate that different rotating flow modes may occur
as the purge mass flow was increased. Further studies in the rim seal region and the disc cavities, and the
were performed by Boutet-Blais et al.39 These authors possible significance of Kelvin–Helmholtz type
reported large-scale unsteady flow structures in both instability in the rim seal/annulus region has been fur-
simplified and real engine geometries with rig and real ther illustrated.
engine operating conditions. In the 180 simplified
model without vanes and blades, 28 or 24 flow struc- Chute rim seals. Following the early studies of axial
ture cells were captured under different operating con- seal geometries described in ‘Axial rim seals’ section,
ditions. In the 24 real engine model with three vanes CFD investigations at the University of Surrey
and four blades a 29-lobed flow structure rotating at focussed on a chute seal design investigated experi-
about 79% of disc speed was found, giving a domin- mentally at the University of Sussex by
ant mode frequency of 23 . Gentilhomme et al.52 and at Oxford by Beard
Chilla et al.40 reported CFD solutions for another et al.42 Boudet et al.25,26 modelled the Sussex turbine
industrial turbine rig having a radial seal combined rig with an outer chute seal and an inner axial seal, as
with an outer chute region. In these models the in Figure 5, both with and without main annulus
close radial clearance seal gap was taken as an inlet vanes and blades. A 120 sector model excluding
boundary. A 36 sector URANS model with four vanes and blades detected a dominant, 21-lobed flow
vanes and seven blades showed a dominant unsteady structure in the seal region with a frequency of 16.7
flow structure mode at frequency 12 in the rotating rotating at 80% of the disc speed. The flow structure
frame of reference. The number of lobes and rota- is illustrated in Figure 15. In a 13:33 sector URANS
tional speed of the flow structure were not given, model, including one vane and two blades, a domin-
but the frequency in the stationary frame was stated ant unsteady pressure frequency at 23.8 was
to be greater than the blade passing frequency of 70 . detected in the rim cavity. It was suggested that
Using a 5:14 URANS model of an isolated rotor restriction to the 13.33 sector accounted for the dif-
blade with the rim seal outer cavity, Chilla et al. inves- ferent frequency in the bladed model. Other frequen-
tigated the influence of the rim geometry. It was found cies, representing interactions with blade passing were
that the unsteady mode frequency varied between 14 observed. This is shown in Figure 16 where the 23.8
Chew et al. 15
Figure 15. Radial velocity for a chute seal without blades or vanes, Re 2 106, Um/b 0.06, Ue/b 0.8, Boudet et al.25
Figure 16. Spectral content of pressure signal for outer cavity, Re 2 106, Um/b 0.06, Ue/b 0.8, Boudet et al.26
mode is shown at 44% of blade passing frequency and different operating condition showed a shift in pre-
modes at 56 and 12% blade passing frequency arise dicted frequencies.
from nonlinear interaction with the blade passing. Gao et al.43 reported URANS results for a chute
The mode at 12% of blade passing frequency seal configuration that was recently experimentally
(¼6.5 ) dominated the inner cavity.26 It was further tested, without blades and vanes, at the University
conjectured that instabilities due to inflections in the of Oxford.42 A 30 sector model with zero purge
rim seal velocity profiles and Taylor–Couette instabil- flow showed inherent unsteady flow structures in the
ity may be involved in formation of the unsteady flow rim seal region with frequencies of 16.2 and 21.7 .
structures. The URANS models predicted less inges- These correspond to 36 or 48 unsteady flow cells
tion than was measured but correctly captured some rotating at about 45% of disc speed. The lower rota-
experimental trends. tional speed of the structures compared to those pre-
As part of a later investigation, O’Mahoney dicted by Boudet et al. might be due to the lack of a
et al.27 extended the URANS studies described by highly swirling external flow in this case. However, the
Boudet et al. to computations on a 26.7 sector and analysis of the detailed unsteady pressure measure-
on a full 360 annulus with blades and vanes. Both ments available for this case suggested flow structures
the peak frequencies and amplitudes of the unsteady rotating at about 80% disc speed. This discrepancy is
remained the same as reported by Boudet et al., discussed further below. Gao et al. also reported the
suggesting that the sector size has little effect on predicted unsteady flow characteristics with URANS
the inherent unsteady flow structures for the two can be affected by the inner step convergence as the
cases studied. However, it was also noted that dual-time stepping approach is used.
it was possible that these larger models were Similar unsteady flow features to those for rim
run for insufficient time for different unsteady fre- seals have been identified in a stationary shroud
quencies to develop. Calculations at a somewhat cavity supplying cooling air to the main annulus
upstream of the turbine blades through a chute Increasing the simulation time by additional disc
shaped channel by Tang et al.53 This geometry corres- revolutions did not change ingestion levels but
ponded to a HP turbine stage of a helicopter engine showed the development of a more distinct frequency
designed by Safran Turbomeca. The full 360 annulus at 32 (compared to a URANS frequency of
URANS model includes NGVs, blades and the cavity 19 ). Improving the mesh resolution gave a better
above the rotor shroud, with the cavity’s axisymmet- prediction of sealing effectiveness at the chute seal but
ric slot upstream of the rotor blades. This cavity is did not improve the sealing effectiveness in the inner
unlike that of the turbine disc rim cavity, as it is a cavity. The finer mesh simulation also showed signs of
completely stationary component of the engine. Tang an additional distinct frequency developing closer to
et al. identified unsteady flow modes unrelated to the that given by URANS.
vane and blade interaction modes. The dominant Although Chilla et al.40 do not give details in their
mode had 19 lobes and rotated at 37% of the rotor paper they mention that LES was applied to their test
speed, with a frequency of 7 . They also reported case. Discussing the occurrence of dominant frequen-
that the unsteady ingestion and egestion flow struc- cies in the rim seal region it was reported that fre-
tures may be associated with 1:3 % stage efficiency quency spectra from LES showed a less
deficit. deterministic behaviour than URANS for low fre-
As for axial and radial clearances, the chute seal quencies. This is consistent with O’Mahoney et al.’s
studies show significant effects of the intrinsic observations.
unsteadiness. Boudet et al. demonstrated interaction Recently, Gao et al.43–45 presented LES for the
of rotating flow modes with blade passing, leading to chute seal studied experimentally at the University
generation of lower frequency unsteadiness that pro- of Oxford. A prescribed sector size of 13:33 was
pagated into and dominated the inner cavity. Tang selected, since unsteady flow structures with 27
et al.’s shroud cavity study further indicates that the lobes were detected in the experiment. The LES
rotating flow modes occur under a wide range of model achieved very good agreement with the experi-
conditions. mentally measured mean pressure distribution, par-
ticularly within the chute seal clearance. The
inherent unsteady flow frequency observed in the
Large eddy simulation (LES) experiment at 23.5 was also accurately captured
One source of uncertainty in URANS modelling by the LES. However, significant discrepancies were
described above is the turbulence modelling. The stu- found between the experiment and the LES. These
dies described above have used a variety of popular included the amplitude of the peak frequency and its
one and two equations turbulence models, without propagation into the cavity, broadband frequency in
revealing any great sensitivity to the choice of model the experiment within the chute seal gap and the speed
but with little systematic investigation. As all of unsteady flow structures. While factors that may
Reynolds-averaged turbulence models are subject to contribute to these differences between the LES and
well-known shortcomings in predicting complex, experiment have been identified, uncertainties remain.
intrinsically unsteady flows such as those considered Gao et al.’s LES showed 54- or 81-lobed structures
here, their predictions must be treated with caution. (corresponding to two or three lobes in the 13.33
At the cost of increased computing requirements LES domain) rotating at about 45% of disc speed (which
offers improvements in predictive reliability and this is close to the mean flow tangential velocity in the
has motivated the studies described below. seal). This compares with a 29-lobed structure rotat-
The first openly published large-eddy simulation of ing at about 80% disc speed indicated by the meas-
turbine rim seal flows appears to be those by urements at the rig condition simulated. The
O’Mahoney et al.27,28 who considered the chute seal procedure to deduce the unsteady flow structure char-
geometry for the University of Sussex test rig. acteristics used to analyse the measurements was
However, this followed a preliminary study by affirmed by application of the same procedure to the
Autef54 which suggested that LES could offer LES. Considering that the measured data have con-
improvements over URANS techniques. In compari- siderably more noise than the LES it is possible that
son with URANS, LES predicted a higher level of modulation of the flow structure affects the estimated
ingestion, in closer agreement with the experimental rotation speed. Otherwise it may be that this is sensi-
measurement of sealing effectiveness. LES for a 13.3 tive to the difference between the rig and LES model,
sector did not show the very distinct low-frequency such as the boundary conditions for the outer annu-
peak in cavity pressure observed in URANS but lus. The seal flow given by the LES is illustrated by the
showed increased amplitude over a range of frequen- contour plots in Figure 17. The bottom part of the
cies.27 Further studies investigated the sensitivity of figure shows pressure and radial and tangential vel-
LES to the simulation time, computational sector ocity components on a plane of constant radius at two
size and mesh resolution.28 Increasing the computa- instants in time. The left-hand set shows a two-lobed
tional sector size from 13:3 to 40 had a very small structure and the right-hand set shows a three-lobed
influence on the average and unsteady results. structure, corresponding to 54 and 81 lobes over the
Chew et al. 17
Figure 17. Circumferential vorticity and velocity vectors on a circumferential plane, and contours on the indicated radial plane of
radial velocity, circumferential velocity and pressure at two instants, showing two lobes (left) and three lobes (right) in a 13.3 sector,
Re 2 106, Um/b ¼ 0, Ue/b ¼ 0, Gao et al.43–45
full 360 . Some interesting flow features may be maximum Reynolds number based on inner cylinder
observed. These include boundary layers on the speed and radial gap was around 2000. Considering,
rotor and stator and a tendency for ‘axial streaks’ to for example, Beard et al.’s42 chute seal the radius ratio
form. The latter is consistent with the Taylor– is around 0.995 and a typical equivalent Reynolds
Proudman theory predicting zero axial gradients in number is 12,000, so there is no direct correspond-
rotationally dominated flow. Further flow features ence of conditions. Nevertheless, the instability of the
are apparent in the contours of circumferential vorti- base flow will apply, and for more effective seals with
city and velocity vectors at the top of the figure. This low purge flow the seal flow and turbulence will be
instantaneous plot is at a circumferential position dominated by rotation. As suggested by Graber
where there is little net inflow or outflow. Structures et al.,8 radial clearance seals with extended overlap of
resembling Taylor–Coutte vortices are observed. rotating and stationary elements the ingestion charac-
teristics could be dominated by rim geometry.
Mathematically, susceptibility of rotating flows to
Instability and wave theory
waves is demonstrated by considering a small perturb-
It is well known that rotating flows are susceptible to ation of a flow rotating as a solid body. This subject
waves, and this is illustrated by Andereck et al.’s55 has received much attention from the atmospheric and
study of flow regimes in the annulus between differen- oceanographic research communities as indicated, for
tially rotating cylinders. According to Rayleigh’s sta- example, by Greenspan.56 In the natural cylindrical
bility criterion the base flow between rotating cylinders coordinates for the present problem, possible solutions
becomes unstable to small perturbations if the radial may be found with an arbitrary number of sinusoidal
derivative of angular momentum is negative. With suf- circumferential lobes, radial variation given by Bessel
ficiently high inner cylinder rotation a wide variety of functions and sinusoidal variation in the axial direc-
waveforms were identified. These include classic and tion. These inertial waves are associated with the
turbulent Taylor vortices (as illustrated in Figure 18), Coriolis force which has a restoring effect in a har-
and further spiral, interpenetrating, modulated and monic motion. Frequency of these waves is limited to
wavy flows. The inner to outer radius ratio in <2 in the rotating frame of reference. Thus, for a
Andereck et al.’s experiment was 0.883 and the wave with N circumferential lobes, the angular speed
Figure 18. Classic flow structures: (a) Taylor vortex streamlines between rotating flow cylinders, (b) rotating inertial wave pressure
contours and (c) Kelvin–Helmholtz shear layer vortices.
of the pattern in the rotating frame will be <2 /N, Kelvin–Helmholtz instability has been suggested as
limiting the departure from the mean flow rotation a cause of rim seal flow unsteadiness by several
speed. These solutions do not account for non-linear authors, with URANS CFD showing evidence of
effects or strong departures from the forced vortex base structures representative of this effect. The instability
flow. Nevertheless, there is a qualitative correspond- is associated with inflection in the velocity profile and is
ence to many of the rotating flow structures observed associated with characteristic billows or vortices as
and discussed above. There may also be quantitative shown in Figure 18. For an unconstrained flow
correspondence on the rotational speed of the mode in with uniform density, the instability will apply to any
some cases. For example, recent LES for a chute seal shear layer. Computations with circumferentially uni-
with no imposed external flow shows lobed patterns form imposed annulus flow indicate that the high shear
rotating roughly at the mean flow speed,43 as is con- between purge and annulus flows supports the occur-
sistent with the theoretical frequency restriction. Other rence of such instability.32,40 Further to this Savov
studies, with highly swirling flow in the annulus, show et al.36 associated increased unsteadiness in their seal
faster angular speeds of the flow pattern perhaps outer cavity or trough with increased shear between
reflecting a higher mean flow tangential velocity purge and main annulus flows at off-design conditions.
induced by the mixing of the purge and mainstream It appears that the shear flow instability may overcome
flows in the rim seal. The inertial wave illustrated in any stabilising effect (according to Rayleigh’s criterion)
Figure 18 has 5 high-pressure/low-pressure cell pairs of the angular momentum increasing with radius from
and associated radial inflow and outflow regions the purge flow to the main annulus flow. As the billows
distributed circumferentially, with an inner to outer or vortices generated by Kelvin–Helmholtz instability
radius ratio of 0.65, and can be compared to the will travel with the mean flow (which has a strong cir-
CFD results in Figures 10 to 15. The linear analysis cumferential component) this mechanism could per-
allows many such solutions with variable radius ratios, haps combine or interact with inertial rotating waves.
etc. In practice these will be constrained by boundary As noted above some workers have sought to
conditions, viscous dissipation and nonlinear effects, explain experimental results with axisymmetric acous-
but there are clear similarities with the CFD solutions. tic and shallow cavity modes.41 In engine conditions,
Rotating acoustic waves may also occur in disc cav- density differences in purge and main annulus flows
ities. Catalfamo’s57approximate analysis of small, com- might also give rise to Rayleigh–Taylor type instabil-
pressible flow disturbances in 2D rotating flow shows ities due to an unstable stratification in the centrifugal
relatively high-frequency modes for an example with force field. The turbine environment may support a
three circumferential lobes. Compared with the meas- number of potentially interacting modes and compet-
urements published by Jakoby et al.24 it is reported that ing sources of instability that could feed into turbu-
the predicted frequency was 3.7 times higher than the lence or large-scale flow structures.
experimental value. Given the agreement between
measurements and incompressible CFD solutions in,
Conclusions
for example, Cao et al.,11 it is clear acoustic wave effects
are limited in some cases. Nevertheless, the occurrence After several decades of research a considerable body
of rotating acoustic waves and their interaction with of knowledge on turbine rim seal flows is available
other phenomena at some conditions is feasible. from full turbine and simplified test rigs. Much of
Chew et al. 19
the available data for sealing effectiveness is reason- from limitations on computing power. These include
ably correlated for particular configurations by a limited simulation time, mesh dependency, use of
simple velocity ratio, and the relative performance restricted domains, sensitivities to solution method,
of different geometries generally follows expected numerical smoothing and choice of time step.
trends. However, comparisons of CFD predictions These uncertainties are often not fully addressed
with measurements have often shown disparities, lim- in pioneering, computationally expensive CFD
iting confidence in design predictions and motivating investigations.
more detailed experimental and computational LESs have demonstrated sensitivity of rim seal
research. flows to the treatment of turbulence in CFD and are
Research using both CFD and experiments has a promising research approach. However, computa-
established the intrinsically unsteady nature of many tional requirements for LES are even more restrictive
ingesting rim seal flows. URANS CFD shows rotating than for URANS, and this will remain an issue for the
large-scale flow structures unrelated to the rotating foreseeable future. Thus, accurate prediction in engine
blades in various configurations, including examples design will continue to be challenging, and develop-
with and without a forced main annulus flow. The ments in modelling will continue to rely on calibration
occurrence of such flow structures has been confirmed against experimental data.
by experiments and LES, although compared to One factor not discussed in this review is the esti-
URANS models these tend to show less distinct fea- mation and control of engine seal running clearances.
tures and frequencies with more modulation of the Advances in this area, including improved aero-
flow structure. In some cases the intrinsic unsteadiness thermo-mechanical modelling, are expected to reduce
is confined to a quite small rim seal region, in other uncertainty in clearance estimation and new designs
cases unsteadiness (particularly lower frequency com- may allow reduction in seal clearance. Together with
ponents) propagates further into the disc cavity and/ moves to smaller, hotter engine cores in future aircraft
or annulus, possibly interacting with blade or vane propulsion, this may increase interest in rim sealing still
passing. The intrinsic unsteadiness dominates inges- further. The present review highlights the need for care-
tion in some cases, including examples with represen- ful experimental and computational research with
tative turbine main annulus flows, but its significance increasing attention to detail. In addition to factors
in some conditions is questionable. already mentioned, systematic investigation of the
Larger scale rotating flow modes identified in the effects of purge to main flow density ratio and eccen-
URANS CFD studies show clear similarities to inertial tricity effects would be useful, as these could affect the
waves arising from the Coriolis force and predicted as a scaling of test rig data to engine conditions. A full
linear perturbation of the inviscid flow in solid body understanding of rim seal ingestion requires further
rotation. Kelvin–Helmholtz type vortices associated detailed study of unsteady flow features.
with free shear layers have been observed in CFD at Understanding developed will also clarify links to
the interface of the purge and main annulus flows as other aspects of engine performance such as turbine
the purge flow enters the annulus. This effect has been aerodynamics, noise and vibration.
observed to be strongest and clearest at high purge flow
rates and has been associated with unsteady measure- Acknowledgements
ments in the outer seal region. The shear layer instabil- The authors gratefully acknowledge contributions to the
ity could clearly affect mixing of the two streams in the work described above from colleagues at the Thermo-
annulus although its role in low purge flow, ingesting Fluid System University Technology Centre, Rolls-Royce
conditions is less clear. For seals with extended over- plc and other collaborating institutions.
lapping radial clearances, flow instability is expected
from Rayleigh’s criterion and Taylor-type vortex struc- Declaration of Conflicting Interests
tures have been detected in LES studies, illustrating the The author(s) declared no potential conflicts of interest with
complexity of the flow structure. respect to the research, authorship, and/or publication of
Considering the various turbulent flow mechanisms this article.
and the non-linear effects and interactions involved in
rim sealing it is not surprising that CFD modelling Funding
has proved challenging. It should also be noted that The author(s) disclosed receipt of the following financial
uncertainties arise in matching experimental bound- support for the research, authorship, and/or publication
ary conditions, and from manufacturing tolerances, of this article: Financial support for associated research
with clearance variation due to eccentricity and from Rolls-Royce plc, the Engineering and Physical
other effects possibly being significant. Sciences Research Council, the Department of Trade and
While URANS models have given considerable Industry, the European Commission, Alstom Power and the
insight into rim seal flows such predictions involve University of Surrey is also gratefully acknowledged.
considerable uncertainties, with turbulence modelling
perhaps being the most obvious. Several authors have ORCID iD
also discussed numerical uncertainties arising largely Feng Gao http://orcid.org/0000-0002-4989-3992
References 21. Johnson BV, Wang C-Z and Roy RP. A rim seal orifice
1. Rolls-Royce plc. The jet engine. 5th ed. reprinted 1996, model with two Cds and effect of swirl in seals. ASME
1996. paper GT2008-50650, 2008.
22. Scanlon T, Wilkes J, Bohn D, et al. A simple method
2. Bayley FJ and Owen JM. The fluid dynamics of a
for estimating ingestion of annulus gas into a turbine
shrouded disk system with a radial outflow of coolant.
rotor stator cavity in the presence of external pressure
J Eng Power 1970; 92: 335–341.
variations. ASME paper GT2004-53097, 2004.
3. von Kármán T. Uber laminare und turbulente reibung.
23. Owen JM. Prediction of ingestion through turbine rim
Z Angew Math Mech 1921; 1: 204–209.
seals – part 2: externally-induced and combined ingress.
4. Chew JW. A theoretical study of ingress for shrouded ASME J Turbomach 2011; 113: 031006.
rotating disc systems with radial outflow. ASME J 24. Jakoby R, Zierer T, Lindblad K, et al. Numerical simu-
Turbomach 1991; 113: 91–97. lation of the unsteady flow field in an axial gas turbine
5. Campbell DA. Gas turbine disc sealing system design. rim seal configuration. In: Proceedings of the ASME
In: Proc. AGARD conf. on seal technology in gas turbine turbo expo 2004, 2004, paper no. GT2004-53829.
engines, 1978. AGARD-CP-237. 25. Boudet J, Autef VND, Chew JW, et al. Numerical
6. Abe T, Kikuchi J and Takeuchi H. An investigation of simulation of rim seal flows in axial turbines.
turbine disc cooling. In: Third CIMAC congress, Aeronaut J 2005; 109: 373–383.
Vienna, 1979, paper no. GT30. 26. Boudet J, Hills NJ and Chew JW. Numerical simulation
7. Scobie JA, Sangan CM, Owen JM, et al. Review of of the flow interaction between turbine main annulus
ingress in gas turbines. ASME J Eng Gas Turbines and disc cavities. In: ASME turbo expo 2006, 2006,
Power 2016; 138: 120801. paper no. GT2006-90307.
8. Graber DL, Daniels WA and Johnson BV. Disk pump- 27. O’Mahoney TSD, Hills NJ, Chew JW, et al. Large-
ing test. Air Force Wright Aeronautical Laboratories. Eddy simulation of rim seal ingestion. Proc IMechE,
Technical Report AFWAL-TR-87-2050, 1987. Part C: J Mechanical Engineering Science 2011; 225:
9. Savov SS and Atkins NR. A rim seal ingress model 2881–2891.
based on turbulent transport. In: ASME turbo expo, 28. O’Mahoney TSD, Hills NJ and Chew JW. Sensitivity of
2017, paper no. GT2017-63531. LES results from turbine rim seals to changes in grid
10. Johnson BV, Mack GJ, Paolillo RE, et al. Turbine rim resolution and sector size. Prog Aerosp Sci 2012; 52:
seal gas path flow ingestion mechanisms. AIAA paper 48–55.
94, 1994. 29. Wang C-Z, Mathiyalagan SP, Johnson BV, et al. Rim
11. Cao C, Chew JW, Millington PR, et al. Interaction of seal ingestion in a turbine stage from 360 degree time-
rim seal and annulus flows in an axial flow turbine. In: dependent numerical simulation. J Turbomach 2013;
Proceedings of the ASME Turbo Expo 2003, 2003, paper 136: 031007–031007-12.
no. GT2003-38368. Also ASME J Eng Gas Turbines 30. Mirzamaghodam AV, Kanjiyani S, Riahi A, et al.
Power 2003; 126: 786–793. Unsteady 360 computational fluid dynamics validation
12. Chew JW, Dadkhah S and Turner AB. Rim sealing of of a turbine stage mainstream/disk cavity interaction.
rotor-stator wheelspaces in the absence of external flow. ASME J Turbomach 2014; 137: 011008–011008-9.
ASME J Turbomach 1992; 114: 433–438. 31. Da Soghe R, Bianchinin C, Sangan CM, et al.
13. Owen JM. Prediction of ingestion through turbine rim Numerical characterisation of hot gas ingestion through
seals – part 1: rotationally-induced ingress. ASME J turbine rim seals. ASME J Eng Gas Turbines Power
Turbomach 2011; 113: 031005. 2016; 139: 032602–032602-9.
14. Kobayashi N, Matsumato M and Shizuya M. An 32. Rabs M, Benra FK, Dohmen HJ, et al. Investigation of
experimental investigation of a gas turbine disk cooling flow instabilities near the rim cavity of a 1.5 stage gas
system. ASME J Eng Gas Turbines Power 1984; 106: turbine. In: ASME turbo expo, 2009, paper no.
136–141. GT2009-59965.
15. Phadke UP and Owen JM. Aerodynamic aspects of the 33. Zhang J and Ma H. Numerical investigation of improv-
sealing of gas-turbine rotor-stator systems parts 1–3. Int ing turbine sealing effectiveness through slot width
J Heat Fluid Flow 1988; 9: 98–117. modification of the rim seal. In: ASME turbo expo,
16. Dadkhah S, Turner AB and Chew JW. Performance 2013, paper no. GT2013-95570.
of radial clearance rim seals in upstream and down- 34. Schädler R, Kalfas AI, Abhari RS, et al. Modulation
stream wheelspaces. ASME J Turbomach 1992; 114: and radial migration of turbine hub cavity modes by the
439–445. rim seal purge flow. ASME J Turbomach 2016; 139:
17. Hamabe K and Ishida K. Rim seal experiments and 011011–011011-10.
analysis of a rotor-stator system with non-axisymmetric 35. Roy RP, Feng JJ, Nazary DD, et al. Experiments on
main flow. ASME paper 92-GT-160, 1992. gas ingestion through axial-flow turbine rim seals.
18. Chew JW, Green T and Turner AB. Rim sealing of ASME J Eng Gas Turbines Power 2005; 127: 573–582.
rotor-stator wheelspaces in the presence of external 36. Savov SS, Atkins NJ and Uchida SA. Comparison of
flow. ASME paper 94-GT-126, 1994. single and double lip rim seal geometries. ASME J Eng
19. Hills NJ, Chew JW and Turner AB. Computational and Gas Turbines Power 2017; 139: 112601–112601-13.
mathematical modelling of turbine rim seal ingestion. 37. Town J, Averbach M and Camci C. Experimental
ASME J Turbomach 2002; 124: 306–315. and numerical investigation of unsteady structures
20. Gentilhomme O, Hills NJ, Chew JW, et al. Measurement within the rim seal cavity in the presence of purge
and analysis of ingestion through a turbine rim seal. mass flow. In: ASME turbo expo, 2016, paper no.
ASME J Turbomach 2003; 125: 505–512. GT2016-56500.
Chew et al. 21
38. Julien S, Lefrancois J, Dumas G, et al. Simulations of flow 54. Autef V. Aerothermal modelling for axial turbines. PhD
ingestion and related structures in a turbine disk cavity. In: thesis, University of Surrey, 2007.
ASME turbo expo, 2010, paper no. GT2010-22729. 55. Andereck CD, Liu SS and Swinney HL. Flow regimes
39. Boutet-Blais G, Lefrancois J, Dumas G, et al. Passive in a circular Couette system with independently rotating
tracer validity for cooling effectiveness through flow cylinders. J Fluid Mech 1986; 164: 155–183.
computation in a turbine rim seal environment. In: 56. Greenspan, HP. The theory of rotating fluids.
ASME turbo expo, 2011, paper no. GT2011-45654. Massachusetts Inst of Tech Cambridge Dept of
40. Chilla M, Hodson H and Newman D. Unsteady inter- Mathematics, 1968.
action between annulus and turbine rim seal flows. 57. Catalfamo PT. Characterization of turbine rim seal
J Turbomach 2013; 135: 051024–051024. flow and its sealing effectiveness. Master thesis.
41. Berg RA, Tan CS, Ding Z, et al. Experimental and Massachusetts Institute of Technology, 2013.
analytical assessment of cavity modes in a gas turbine
wheelspace. In: ASME turbo expo, 2007, paper no.
GT2017-63734.
42. Beard PF, Gao F, Chana KS, et al. Unsteady flow phe- Appendix
nomena in turbine rim seals. ASME J Eng Gas Turbines
Power 2016; 139: 032501–032501-10. Notation
43. Gao F, Chew JW, Beard PF, et al. Numerical studies of
b disc outer radius
turbine rim sealing flows on a chute seal configuration.
In: Proceedings of 12th European conference on turbo- c species concentration
machinery fluid dynamics & thermodynamics, Cd discharge coefficient
Stockholm, Sweden, 2017, paper no. ETC2007-284. Cw throughflow Reynolds number (¼ m=b) _
44. Gao F, Chew JW, Beard PF, et al. Large-eddy simula- f frequency
tion of unsteady turbine rim sealing flows. Int J Heat Gc seal gap ratio (¼ sc/b)
Fluid Flow 2018; 70: 160–170. k constant in rotationally driven ingestion
45. Gao F, Poujol N, Chew JW, et al. Advanced numerical model
simulation of turbine rim seal flows and consideration m_ mass flow rate
for RANS turbulence modelling, In: Proceedings of the N number of lobes in a rotating flow
ASME Turbo Expo 2018: turbomachinery technical
mode
conference & exposition, Oslo, Norway, 2018, paper
r local radius, radial direction
no. GT2018-75116.
46. Autef V. Computation of disc cavity rim sealing. Re rotational Reynolds number (¼b2/)
University of Surrey/Institut National Polytechnique sc seal clearance
de Grenbole project report, 2002. Ue mean axial velocity in the main annulus
47. Chew JW, Hills NJ, Hornsby C, et al. Recent develop- flow
ments in application of CFD to turbomachinery inter- Um _
mean velocity through seal (¼ m=2s c b)
nal air systems. In: Proceedings of the 5th European useal local seal velocity
turbomachinery conference, Prague, March 2003.
48. Smout PD, Chew JW and Childs PRN. ICAS-GT: a
European Collaborative Research Programme on
Internal Cooling Air Systems for Gas Turbines. In: Greek symbols
ASME turbo expo, 2002, paper no. GT2002-30479.
49. Bohn DE, Decker A, Ma H, et al. Influence of sealing
constant in equation (4) or angle
air mass flow on the velocity distribution in and inside between sensors
the rim seal of the upstream cavity of a 1.5-stage tur- angle between adjacent lobes in flow
bine. In: ASME turbo expo, 2003, paper no. GT2003- structure
38459. Dp pressure difference
50. Zhou DW, Roy RP, Wang C-Z, et al. Main gas inges- " sealing effectiveness
tion in a turbine stage for three rim cavity configur- Cw
l throughflow parameter ¼ Re 0:8
;
ations. J Turbomach 2011; 133: 031023–031023. dynamic viscosity
51. Feiereisen J, Paolillo R and Wagner J. UTRC turbine fluid density
rim seal ingestion and platform cooling experiments. In: !s angular speed of rotating flow mode
AIAA/ASME/SAE/ASEE joint propulsion conference
angular disc speed
and exhibit, 2000, paper no. AIAA-2000-3371.
52. Gentilhomme O. Turbine rim seal ingestion. PhD thesis.
University of Sussex, 2004.
53. Tang E, Trébinjac I, Ngo Boum G, et al. Aerodynamic Subscripts
interactions between a high pressure turbine stage
and a shroud cavity. In: Proceedings of 12th European ext external flow
conference on turbomachinery fluid dynamics & thermo- fd free disc flow
dynamics, Stockholm, Sweden, 2017, paper no. int inlet to cavity
ETC2017-176. min minimum sealing requirement

Flow Mechanism in Axial Turbine Rim Seals

Uploaded by

Copyright:

Available Formats

Flow Mechanism in Axial Turbine Rim Seals

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Flow Mechanism in Axial Turbine Rim Seals

Uploaded by

Copyright:

Available Formats

Review Article

Proc IMechE Part C:

rim sealing Reprints and permissions:

John W Chew, Feng Gao and Donato M Palermo

Date received: 22 February 2018; accepted: 24 May 2018

Figure 3. Model coefficients for seals considered in Chew4

in radial clearance seals with extended overlap of

" ¼ Cw=ð0:8Cw þ 0:2Cwmin Þ ð3Þ

Reference Method, sector NGV/blade Re 103 sc =b 103 Um =ðbÞ f = N !s =

Reference Method, sector NGV/blade Re 103 sc =b 103 Um =ðbÞ f = N !s =

Roy et al.35 Experiment 22/28 5 9 105 10.2 0–320 22 – –

Table 3. Summary of chute seal studies showing intrinsic unsteady flow.

Reference Method, sector NGV/blade Re 103 sc =b 103 Um =b f = N !s =

inherent unsteadiness indicating further complexity

Axial rim seals. The ﬁrst evidence of inherent unsteady

You might also like