Professional Documents
Culture Documents
ETC7 110 BrushPorosity Modelling
ETC7 110 BrushPorosity Modelling
110
ABSTRACT
The present paper discusses three related approaches to efficient brush seal CFD modeling
by using a porous medium approach. The corresponding sets of resistance coefficients are
reviewed and resulting characteristics essential to steam turbine applications are evaluated.
The important radial clearance between bristle pack and rotor is used to calibrate the
numerical model according to experimental data. The influence of a changing clearance
height is studied with respect to the percentage of flow leaving the bristle pack.
As already shown, the porous medium approach proves to be suitable for brush seal
modeling without the necessity to resolve the flow around individual bristles. While previous
works have focused on leakage and the simulation of pressure drops in the porous region to
provide a basis to calculate contact forces, the current porous model is employed to study the
influence of the brush on swirl across the blade shroud of a steam turbine.
It is shown by comparison to standard swirl breakers that the brush effectively reduces
swirl and thus contributes to rotor-dynamic stability.
NOMENCLATURE
a, A, A viscous resistance factor or matrix v flow velocity
Acyl bristle or cylinder cross section V total volume
bbrush axial width of bristle pack Vfluid fluid volume fraction
b, B, B inertial resistance factor or matrix Vsolid solid volume fraction
C porosity factor used in Table 1 Vcyl bristle or cylinder volume
d bristle or cylinder diameter z, r, θ axial, radial, circumferential coord.
~
d equivalent sphere diameter α viscous resistance coefficient
dh hydraulic diameter β inertial resistance coefficient
D porosity factor used in Table 1 ∆p pressure difference
L length in direction of flow ε porosity
Lpipe, Lbristle pipe or bristle length ϕ bristle lay angle
n bristle packing density λL pipe friction coefficient
N number of bristles or cylinders µ dynamic viscosity
Re Reynolds number ρ density
Subscripts
n, z normal to bristles s in bristle direction
INTRODUCTION
Brush seals are widely used in the turbomachinery industry today to obtain leakage reductions
and improved efficiency (e.g. Neef et al., 2006). The theoretical modeling of brush seal behavior,
however, is still lagging behind. A governing approach to predict brush seal performance is
currently not available. The contemporary open literature on CFD modeling of brush seals may be
distinguished by the approach to simulate the bristle pack in the flow path. One method is to treat
the bristle pack as a porous medium which simulates the pressure drop across the bristles and
further effects of the brush on the fluid flow. A second method is to calculate the complete fluid
flow field around the individually packed bristles.
General equations
Early investigations of flows through porous media have been performed by Darcy (1856) and
Carman (1937). The influence of friction in such flow is adequately reflected by equation (eq. 1),
which takes into account the friction of the cylinder surface and the dissipation of the flow energy
due to obstacles in the flow path:
∆p
= A ⋅ µ ⋅ v + B ⋅ ρ ⋅ v2 (eq. 1)
L
The pressure drop ∆p occurs across the length L of the packed bed of cylinders in the direction
of flow. The coefficients A and B in the viscous and the inertial resistance terms, respectively, are
defined by Ergun (1952) as follows:
α ⋅ (1 − ε )
~ 2
A= ~ (eq. 2)
ε3 ⋅ d 2
~
β ⋅ (1 − ε )
B= ~ (eq. 3)
ε3 ⋅ d
~
Variable d is the equivalent sphere diameter that is related to the cylinder or bristle diameter d by
~ 6 × cylinder volume
d= = 1.5 d (eq. 4)
cylinder surface
neglecting the cross-sectional area for long bristles. ε represents the porosity of the medium, which
is defined by the quotient of the fluid volume Vfluid divided by the total volume V enclosing the
porous region including the solid bristle volume Vsolid:
V V
ε = fluid = 1 − solid (eq. 5)
V V
Based on measurements, Ergun suggests values for the empirical resistance coefficients ~ α and
~ ~ ~
β as α =150 and β = 1.75 for spherical particles. For bristles or packed cylinders, making use of
~
the equivalent sphere diameter in eq. 4 the coefficients are α = ~ α / d² ⋅ d² = 66.7 and
~ ~
β = β / d ⋅ d = 1.17 . For a packed bed of cylinders, others have derived the constants α = 80 and β =
1.16, see for example Pröstler (2005) and Chew and Hogg. (1997).
−
∆p (1 - ε )2
= 32 ⋅ 2 2 ⋅ µ ⋅ v (eq. 14)
L bristle ε d
For the flow inside the bristle pack of a brush, the pipe length is represented by the bristle length
Lbristle. Comparison with equation (eq. 1) leads to the definition of the resistance coefficient in
bristle direction as resulting in:
a s = 32 ⋅
(1 - ε )2 (eq. 15)
ε 2d 2
Although this factor looks similar to the ones given above, it should be noted that the
dependency on the porosity ε has changed.
Friction /viscosity Dissipation /inertial The three different approaches for
resistance coefficient resistance coefficient porosity modeling of a brush seal
a = 66.7 C bn = 1.17 D are listed in Table 1 with respect to
Ergun n the bristle diameter d. It shows
as = an bs = bn
(1952) Ergun’s empirical constants without
az = an bz = bn
considering the anisotropy in
an = 80 C bn = 1.16 D
Chew permeability in contrast to Chew’s
a = 1/60 an = 1.33 C bs = 0
(1995) s and Pröstler’s method of modeling
az = an bz = bn
brush anisotropy. The latter both
a = 80 C bn = 1.16 D
Pröstler n refer to Kay & Neddermann (1974)
as = 32 ε C bs = 0
(2005) for the resistance coefficients in the
az = an bz = bn
transverse direction. Assuming a
1 ⎛1− ε ⎞
2
1 ⎛1− ε ⎞
with C = 3 ⋅ ⎜ ⎟ with D = 3 ⋅ ⎜ ⎟ typical brush seal porosity of 15%,
ε ⎝ d ⎠ ε ⎝ d ⎠ it can be seen that Ergun’s approach
~ will lead to the lowest leakage rates
with d = d /1.5 (see eq. 4)
Table 1: Comparison of the different resistance coefficients followed by Pröstler’s and finally
Chew’s approach. Their relative
and quantitative implications at defined operating conditions, though, need to be assessed
separately. This shall be pursued in the following section.
0,06
accounts for the bristle lay angle,
0.06
and the geometric boundary
conditions account for the existing
0.04
0,04
gap underneath the bristle pack
despite its exact physical value.
0.02
0,02 © Siemens Calculations have then been
∆p = 3.5bar extended to the application of the
0,00
0 other resistance coefficients
0
0,0 2,5 5
5,0 inlet pressure
7,5 [bar] 10
10,0 according to Ergun and Chew,
leaving all other boundary
Figure 2: Comparison of calculated and measured leakage flows
through the brush seal conditions including the free gap
size untouched. Results for the low
pressure differential range are not discussed here due to the stronger dependency of the porosity on
pressure drop which is judged to exceed the scope of this investigation.
As already expected by comparing the similar nature and values given in Table 1 for the three
models, the results do not vary too much, with highest mass flows for Chew’s approach due to
lowest resistance coefficients. It may be noted that the numerical models predict slopes of the curve
in good agreement with sealing theory – linear dependence on pressure with the curve intersecting
the origin of the coordinate system. For conventional fin seals, this is true for pressure drops above
the critical pressure ratio where sonic speed is reached in the gap. The experimental values do not
show this trend, since hysteresis effects in the pressure differential dependent blow down of the
brush seal have a strong influence on the leakage. Moreover, the slope of the measured leakage rate
related to the pressure differential is lower, which points to a noticeable dependency of the radial
clearance on pressure differential not taken into account in the current investigation.
Now, in order to assess the influence of the radial gap on the performance of the seal in more
detail, the flow through the brush seal is divided in two fractions, one flow underneath the bristle
pack and the flow which leaves the seal through the bristles (see Figure 3). For the calibration point
of the model (see Figure 2, long dashed line), the fractional mass flows leaving the bristles relative
to the total mass flow are compared in Figure 4 (left, dashed line). Due to the smaller bristlewise
resistance coefficient, Chew’s porosity values lead to a bristle mass flow of about 30% related to the
total flow for the modeled gap of 0.092 mm. The
isotropic approach of Ergun with the highest resistance
back coefficients leads to about 18% of the full mass flow
front plate through the bristle pack. For smaller values for the free
plate
gap underneath the bristle pack, the fraction of flow
through the bristle pack increases while the total
bristle pack bristle flow quantity of the mass flow through the seal decreases
gap height h gap leakage (Figure 4, right). When the gap is completely closed,
the differences in modeling the permeability become
most apparent. Less resistance in the bristle direction
rotating surface leads to higher leakages for the approaches of Pröstler
Figure 3: Brush seal schematic drawing –
and Chew.
bristle and gap mass flow
Since experimental evidence suggests smaller radial
gaps during operation (see above) than the one used for calibration in this calculation, actual
resistance factors might be lower than the ones given by the three discussed porosity models.
The dependence of mass flow on radial gap height (Figure 4) can also be used to introduce a more
accurate leakage model which takes into account the closure of the free gap for higher pressure
drops. This was already discussed for Figure 2, where experimental values show lowest flow rates
for higher pressures, where the blow down tends to close remaining gaps between bristles and rotor.
100 0,05
0.05
∆p = 3.5 bar Ergun ∆p = 3.5 bar
80 0,04
0.04
Pröstler
total mass flow [kg/s]
h = 0.092mm
bristle flow [%]
60 0,03
0.03
Chew
40 0,02
0.02
h = 0.092mm
20 0,01
0.01
© Siemens © Siemens
0 0
0 0.02 0,04
0,02 0.04 0.06 0,08
0,06 0.08 0.1
0,1 gap height h [mm] 0 0.02
0,02 0.04
0,04 0.06
0,06 0.08
0,08 0.1
0,1
Figure 4: Percentage of flow leaving the bristle pack (left) and total mass flow through the seal (right) as a
function of free gap height (performed at 3.5 bar pressure drop)
blade
shroud outlet
main flow
r © Siemens
inlet blade foil
θ z
Figure 5: Computational blade shroud seal model
To compare the influence of the brush, the effect of a standard swirl breaker (see Fig. 6) is also
examined. The swirl (or circumferential velocity component) within the seal influences both,
turbine efficiency and rotordynamic stability. Efficiency is mainly affected when the leakage flow
mixes with the main flow. Differing circumferential velocities will lead to dissipation and
unfavorable changes in incidence to succeeding blade rows. The rotordynamic stability of the
turbine is influenced because of the swirl in the leakage flow. This swirl is a driver of the ‘steam
whirl effect’ that contributes to the excitation of rotor vibrations.
swirl breaker
seal fins z
θ © Siemens
Swirl Analysis
Figure 7 draws a comparison between a labyrinth seal with regular fins (gap = 0.8 mm), a
labyrinth seal including a thick fin with a small gap representing the geometric dimensions of a
blocked bristle pack (gap = 0.1 mm), and a labyrinth seal including a brush seal modeled as a
porous medium (gap = 0.1 mm). The plot shows the average swirl component at distinct axial
locations corresponding to the geometric representation of the seal at the top of the graph.
Due to the porosity modeling, the swirl velocity in the third configuration is reduced
considerably from 180 m/s to a minimum of about 30 m/s. Due to wall friction at the rotor surface,
the average circumferential velocity subsequently increases to a value of 108 m/s at the outlet of the
seal. This is close to half the shroud surface speed of approx. 280 m/s. The two brushless labyrinth
seals are shown for comparison. The thick fin with a radial gap corresponding to the brush seal
demonstrates that the swirl reduction of the brush seal results from the porous bristle pack rather
than the small clearance as might be expected. The influence of the thick fin with a small clearance
on swirl is larger than that of the regular labyrinth, but much smaller than that of the brush seal.
Corresponding provisional measurements of the swirl component of the fluid behind a brush seal,
which shall not be discussed here in more detail, provide evidence that the level of swirl reduction
is reasonably captured by the numerical model.
250
without brush b=0.3mm
gap=0.8mm
circumferential velocity [m/s]
200
150
without brush b=2mm
gap=0.1mm
100
50 © Siemens brush
0
10 20 30 40 50 60
distance to seal entrance [mm]
Figure 7: Influence of the brush on the swirl
Figure 8 shows the function of different swirl breakers within the leakage path of the seal in
comparison as well as in conjunction with a brush seal. The circumferential velocity component is
most significantly reduced in the seal with a swirl breaker only, and by omitting a subsequent brush.
The long version A of the swirl breaker is slightly more effective since the ratio of the radial swirl
fin extension relative to the remaining gap is better than in the case of the short swirl breaker B. A
combination of swirl breaker and brush does not result in a difference for the swirl velocity level at
the outlet of the seal. This is an interesting result, since it has been previously suggested that swirl
breakers may be used in combination with brush seals in order to improve rotor stability. Here, it is
shown that the brush itself reduces the swirl and that the addition of a swirl breaker upstream of a
brush seal has no additional influence even for the chosen free gap under the bristle pack of 0.1
mm. While the swirl breaker only provides partial flow guidance to reduce axial velocity, the
porous bristle pack governs the seal flow by its high flow resistance and the corresponding total
pressure drop. Additionally, the brush pack acts like a flow straightener. Further investigation has
shown that increasing the gap beneath the bristle pack lessens the swirl breaking effect, but it
remains noticeable even for large radial clearances of up to 0.4 mm.
250
without brush
200
circumferential velocity [m/s]
150
brush
100
swirl breaker A
50
swirl breaker B
0 © Siemens
CONCLUSIONS
The present paper discusses different approaches to brush seal CFD modeling and shows how
experimental data can be used to calibrate a numerical model for a wide range of operating
parameters. Despite valuable results regarding leakage prediction of brushes, the authors clearly
show the limitations of the porous medium approach. This approach is not suitable to capture
effects related to movement of the bristles such as flutter, blow down or blow over (at high pressure
differentials). This challenges the leakage results in the low pressure differential range. The
approach is also limited when wear or the durability of brushes are studied (see Neef et al., 2006).
Although the treatment with the porous medium approach cannot represent the flexible nature of
the brush, it is suitable to capture general features such as leakage or the capability of the brush to
break inlet swirl of seal flows. This is especially important when the influence of a brush on rotor-
dynamic stability is to be evaluated. Since the anisotropic permeability models can be employed to
account for predominant flow paths in the bristle pack, the sensitivity to the swirl breaking
capability of the brush can be investigated. Finally, it is shown by comparison to standard swirl
breakers that a brush can effectively reduce swirl and can thus contribute to rotor-dynamic stability.
ACKNOWLEGEMENTS
The authors would like to thank Klaus Kwanka († 2006) and Martin Deckner from the Lehrstuhl
für Energiesysteme at the TU Munich for providing the experimental data used for calibration. The
work presented in this paper was supported by the Bundesministerium für Wirtschaft und Arbeit in
Germany under the support code 0327714D as part of the COOREF-T project. Responsibility for
the content of the publication rests with the authors.
REFERENCES
Braun, M., Hendricks, R.C.; Canacci, V.A.: Flow visualisation in a simulated brush seal. ASME-Paper 90-GT-217,
1990
Carman, P.: Fluid Flow through Granular Beds. Trans. Inst. Chem. Engineers Vol. 15, pp. 150-166, London, 1937
Chen, L; Wood, P.; Jones, T.; Chew, J.: An iterative CFD and mechanical brush seal model and comparison with
experimental results. ASME-Paper 98-GT-372, 1998
Chew, J.; Lapworth, B.; Millener, P: Mathematical modeling of brush seals. International Journal of Heat and Fluid
Flow, 1995, Vol. 16, pp. 493-500
Chew, J.; Hogg, S.: Porosity modeling of brush seals. Journal of Tribology. 119, Oct. 1997, pp. 769-77
Darcy, H: Les Fontaines publiques de la ville de Dijon. Paris, 1856 - in: Carman (1937)
Dogu, Y.: Investigation of brush seal flow characteristics using bulk porous medium approach. Journal of Engineering
for Gas Turbines and Power. January 2005, Vol. 127, pp. 136-144
Ergun, S.: Fluid Flow through Packed Columns. Chemical Eng. Progress Vol. 48, pp. 89-94, 1952
Ferguson, J.G.: Brushes as High Performance Gas Turbine Seals ASME-Paper 88-GT-182, 1988
Kay, J.M.; Neddermann, R.M.: An Introduction to Fluid Mechanics and Heat Transfer, Camb. Univ. Press, 3rd ed., 1974
Lelli, D.; Chew, J.; Cooper, P.: Combined 3D fluid dynamics and mechanical modeling of brush seals. ASME-Paper
GT2005-68973, 2005
Neef, M; Sürken, N.; Sulda, E.; Walkenhorst, J.: Design Features and Performance Details of Brush Seals for Turbine
Applications. ASME-Paper GT2006-90404, 2006
Pröstler, S.: Modellierung und numerische Berechnung von Wellenabdichtungen in Bürstenbauart. PhD Thesis
University of Bochum, 2005
Schettel, J.; Deckner, M.; Kwanka, K; Lüneburg, B; Nordmann, R.: Rotordynamic coefficients of labyseals for turbines
– comparing CFD results with experimental data on a comb-grooved labyrinth. ASME-Paper GT2005-68732, 2005
Turner, M.; Chew, J.; Long, C.: Experimental investigation and mathematical modeling of clearance brush seals.
ASME-Paper 97-GT-282, 1997
Urlichs, K.: Ermittlung der Eigenschaften adaptiver Dichtungen für hohen Druck und hohe Temperatur für die
Anwendung in Dampfturbinen, Abschlußbericht Vorhaben 414/00, Bayr. Forschungsstiftung, München, 2005