You are on page 1of 9

Influence of Fluid Flow Regime

Noël Brunetière
e-mail: bruneti@lms.univ-poitiers.fr on Performances of
Bernard Tournerie Non-Contacting Liquid Face Seals
Jean Frêne
Some non contacting mechanical face seals are running near the laminar boundary flow
Laboratoire de Mécanique des Solides, limit. A modification of operating conditions leads to a non laminar fluid flow in the seal
UMR CNRS 6610, interface while inertia forces remain negligible. A numerical model has been developed to
S.P.2M.I., BP 30179, determine pressure and velocity fields in the sealing dam for laminar to turbulent regime.
86962 Futuroscope Chasseneuil cedex, France The turbulent viscosity determination is based on the Elrod and Ng model. Evolutions of
seal characteristics (opening force, friction torque, leakage rate . . . ) and fluid film dy-
namic coefficients versus running conditions are presented. Numerical results show that
great variations appear in the transition to turbulence. 关DOI: 10.1115/1.1456453兴

1 Introduction numerically determined by a small perturbation technique. The


effect of an increase in seal clearance and in angular velocity on
A mechanical face seal is basically composed of a rotating seat the seal behavior is studied.
and a static ring separated by a thin fluid film. The seal ring is
flexibly mounted in order to dynamically track rotor misalign-
ments 共There are other possible seal layouts, however兲. The dy-
namic behavior of face seals was largely studied in the last two 2 Determination of Fluid Flow Regime
decades. The non-linear character of the equations to be solved led In a non contacting face seal, the fluid flow in the film has two
several authors to use relatively complex iterative techniques, components. A circumferential Couette flow due to the seat rota-
such as Adams method 关1–3兴. Green and Etsion 关4兴 proposed to tion and a radial Poiseuille flow due to the pressure gradient. Both
linearize the problem by using a small perturbation technique components are at the origin of flow regime transition. We will
which proved to be simple and accurate to study stable dynamic first study the effect of each component.
tracking. This method was then generalized and used by many If the seal is running without a pressure gradient, the fluid flow
authors 关5–7兴. These works showed that the fluid film dynamic in the interface is a pure Couette flow. Then, the transition to
coefficients control the seal dynamic behavior, leading generally, turbulence is governed by the Reynolds number Rec :
to a stable dynamic tracking mode 关7,8兴. All the studies quoted ␳r␻h
previously were carried out for a laminar fluid flow in the film. Rec ⫽ .
Very few workers have been concerned with seals operating in the ␮
non laminar flow regime. Within the framework of two phase When Rec⬍Recl the sealed flow remains laminar and when Rec
seals, Yasuna and Hughes 关9兴 were interested in the axial dynamic ⬎Rect turbulence appears. In the case of seals, no experimental
behavior for both laminar and turbulent thin film fluid flow. They study has been undertaken to determine the limits Recl and Rect .
point out that their study is limited by the fact that no attempt is However, there are some theoretical studies treating of turbulent
made to resolve the transition from laminar regime to turbulence. flows. Lin and Yao 关14兴 note that the flow in the interface of a face
On the other hand, several authors have studied dynamic behavior seal becomes turbulent for Rec⬎1000. Yasuna and Hughes 关9兴
of annular seals operating in transition to turbulence regime assumed that turbulence appears when Rec⬎3000. Turbulent
关10,11兴. The results highlight significant variations of fluid film flows have been studied in the context of hydrodynamic thrust
dynamic coefficients in the transitional zone. bearings, which present a geometry rather close to that of the
Some face seals, such as in reactor coolant pump primary seals mechanical seals, and where Couette flow is dominant. Experi-
共hydrostatic seals 关12兴兲, are designed to firstly insure a pressure mental studies made it possible to observe the transition to turbu-
drop. Generally, they are large-sized and operating with a rela- lence as discontinuities or abnormal evolutions of the dissipated
power and temperature 关15兴. In 1977, Frêne 关16兴 carried out ex-
tively thick film that leads to high leakage rate. Subsequently, the
periments and observed that thin film fluid flows become turbulent
fluid flow in the sealing dam is close to the limit of the laminar
when Rec min⬎800 and 4100⬍Rec max⬍5000, with Rec min and
flow regime. The purpose of this paper is to study the influence of Rec max as the minimum and maximum values of the Couette Rey-
running condition variations that leads to a non-laminar flow on nolds number in the film. In his theoretical study, Frêne consid-
the behavior of such seals. ered that the fluid flow is not laminar for Rec⬎1000 and becomes
In our study, we present a numerical method to determine pres- turbulent when Rec⬎2000. Souchet 关17兴 recommended the values:
sure and velocities for laminar to turbulent isothermal flow in the
mechanical seal interface. Classical lubrication assumptions are Recl ⫽900 and Rect ⫽1600.
considered to be valid. An original technique is proposed to char- If the seal operates with a large pressure gradient and without
acterize the fluid flow regime in the film while taking into account rotating velocity, its performance is equivalent to that of a hydro-
simultaneously the Couette and Poiseuille flows. An interpolation static thrust bearing. The fluid flow in the film is of Poiseuille type
law allows to extend the use of the Elrod and Ng turbulent model characterized by the Reynolds number Rep :
关13兴 to the transition zone. Fluid film dynamic coefficients are
␳ Ṽ r h
Rep ⫽ with Ṽ r as the mean radial velocity across
Contributed by the Tribology Division for publication in the ASME JOURNAL OF ␮
TRIBOLOGY. Manuscript received by the Tribology Division February 27, 2001;
revised manuscript received August 7, 2001. Associate Editor: S. Wu. the film.

Journal of Tribology Copyright © 2002 by ASME JULY 2002, Vol. 124 Õ 515

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm


axis. In the present work, a flexibly mounted stator with three
degrees of freedom is studied: axial motion along the z 0 axis and
tilts about the (O 1 ,x 0 ) and (O 1 ,y 0 ) axes with O 1 as the stator
face center. This model is well known in face seal analysis 关4,7兴.
The stator position is given by the centreline clearance h and the
angles ␹ x and ␹ y .
The positions and the velocities of the points M 1 and M 2 共Fig.
2兲 relative to an inertial reference, are defined using the following
relations:

OM 1 1 ⫽xXជ 0 ⫹yYជ 0 ⫹ 共 h⫹ ␤ 1 共 r⫺r e 兲 ⫹ ␹ x y⫺ ␹ y x 兲 Zជ 0 ⫽xXជ 0 ⫹yYជ 0


⫹H 1 Zជ 0

OM 2 ⫽xXជ 0 ⫹yYជ 0 ⫹ ␤ 2 共 r⫺r e 兲 Zជ 0 ⫽xXជ 0 ⫹yYជ 0 ⫹H 2 Zជ 0


Vជ 共 M 1 兲 ⫽ 共 ḣ⫹ ␹˙ x y⫺ ␹˙ y x 兲 Zជ 0 ⫽V Z1 Zជ 0

Vជ 共 M 2 兲 ⫽⫺ ␻ yXជ 0 ⫹ ␻ xYជ 0 ⫽V X2 Xជ 0 ⫹V Y 2 Yជ 0
Conservation of the Momentum. The conservation of the
Fig. 1 Presentation of flow factor ␣ momentum in a Newtonian fluid results in the Navier-Stokes
equations. To study turbulent flows, we use a mean form of these
equations where a turbulent diffusivity ␧ t is introduced as pro-
Bassani and Picigallo 关18兴 consider the flow to become turbulent posed by Boussinesq 共Szeri, 关19兴兲. To determine the importance of
when Rep⬎2300. inertia forces, we compared pressure fields obtained by solving
In practice, seals operates with both a pressure gradient and a Navier Stokes equations 共CFD software兲 and by solving Reynolds
rotating speed. Thus, the fluid flow in the sealing dam is a com- equation for the nominal and the most unfavourable cases pre-
bined circumferential Couette flow and radial Poiseuille flow. The sented in the results section. The results showed that inertia terms
transition to turbulence is governed by both Reynolds numbers remain insignificant in the majority of the cases studied. However,
Rec and Rep . To generalize, we define a vector Reynolds number a variation of about 6 percent on the opening force (F z ) were
with components Rec and Rep . The limits of the flow regime predicted for the largest film thickness. Hence, it seems reasonable
transition are given by ellipses of the same eccentricity 共Fig. 1兲. A to neglect inertia forces. Furthermore, thermal effects are also
flow factor ␣ is defined to denote the flow character: neglected. Thus, classical lubrication assumptions are valid, and

冑冉 冊 冉 冊
the fluid film velocities are expressed as
Rec 2 Rep 2
␣⫽
1600

2300
.

The fluid flow is turbulent for ␣ ⬎ ␣ t ⫽1, and laminar when ␣


V̄ x ⫽
⳵ P̄
⳵x 冉 I1

• I⫺ •J ⫹V X2 • 1⫺
J1
J
J1 冉 冊 (1)

冉 冊 冉 冊
⬍ ␣ l ⫽(900/1600).
⳵ P̄ I1 J
V̄ y ⫽ • I⫺ •J ⫹V Y 2 • 1⫺ ,
3 Analysis ⳵y J1 J1

Geometry and Kinematic Model. A mechanical face seal where I and J are integrals across the film thickness that take into
consists basically of a rotating ring 共rotor, seal seat 2兲 and a static account the turbulent viscosity variation:
ring 共seal ring, 1兲 separated by a thin fluid film 共Fig. 2兲. The seal
冕 z ␰ •d ␰

冉 冊
seat rotates about the (O,z 0 ) axis with a constant angular velocity I共 z 兲⫽ and I 1 ⫽I 共 H 1 兲
␧t
␻. The seal face is assumed perfectly aligned with the rotation H2
␮ • 1⫹ ␦ ␧ •

冕 z d␰

冉 冊
J共 z 兲⫽ and J 1 ⫽J 共 H 1 兲 .
H2 ␧t
␮ • 1⫹ ␦ ␧ •

␦ ␧ is an interpolation factor for transitional flows as used by Szeri
关19兴.
Conservation of Mass. The generalized form of the Rey-
nolds equation results from the mass conservation equation inte-
grated across the film thickness:

⳵x

G 1•
⳵ P̄
⳵x


⳵y
冊 冉 冊
G 1•
⳵ P̄
⳵y



⳵x 冉 冊 冉 冊
V X2 •
I1
J1


⳵y
V Y 2•
I1
J1
⫹V Z1 , (2)

where G 1 is given by

冕冉 冊
H1 I1 z.dz

冉 冊
G 1⫽ z⫺ • .
H2 J1 ␧t
␮ • 1⫹ ␦ ␧ •
Fig. 2 Geometric and kinematic model of the seal ␯

516 Õ Vol. 124, JULY 2002 Transactions of the ASME

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm


Non-Laminar Flows. For non-laminar flows, the turbulent
viscosity is determined using the Elrod and Ng model 关13兴. In the
wall region, the turbulent viscosity expression is based on the
Reichardt formula

␧R


⫽0.4• Z• 冑␶ * ⫺10.7•tanh
Z• 冑␶ *
10.7
, 冊
where Z and ␶* are the dimensionless distance from the nearest
wall and shear stress, respectively, i.e.,

Z⫽min 冉 z⫺H 2 H 1 ⫺z
H
,
H
, 冊 ␶ *⫽
H 2•兩␶兩
␮•␯
and H⫽H 1 ⫺H 2 .

To prevent that turbulent viscosity vanish when the shear stress is


null, the Elrod and Ng model forces an increasing monotonous
evolution of viscosity in each half channel
␧W

共 z 兲 ⫽max
␧R

共z⬘兲冋 册 ᭙z ⬘ ⬍z

In the core region, Elrod and Ng proposed to use a modified


Clauser formula adapted to lubrication
Fig. 3 Interpolation factor ␦ ␧ versus flow factor ␣ for two in-
1 terpolation laws
␧ C⫽
56

⫻ 冑冋 冕 H1

H2
共 V̄ x 共 z 兲 ⫺V XM 兲 •dz ⫹ 册 冋冕
2 H1

H2

共 V̄ y 共 z 兲 ⫺V Y M 兲 •dz ,
2
simer and Thomas and by the sinus law. It can be seen that our
results are close to those of Missimer and Thomas and to Rei-
where V XM and V Y M are the components of the maximum velocity chardt measurements.
across the film. Patel and Head 关21兴 measured skin friction in channel for lami-
Finally, to take into account both the wall and core regions, nar (Rep⬍1300) to turbulent pressure flow (Rep⬎2800). A com-
Elrod and Ng considered that turbulent viscosity ␧ t is the lowest parison of sinus law results and their experimental data is pre-
of the two values previously calculated sented in Fig. 6. The calculated friction factor is close to the Patel
and Head measurements for Rep⬍2000. For higher Reynolds
␧ t ⫽min共 ␧ W ,␧ C 兲 . number values, variations are observed.
Zirkelback and San Andres 关10兴 use a Bulk-flow model to study
The Elrod and Ng turbulent model is typically nonlinear, as op- annular seals operating in the transition to the turbulence regime.
posed to the original model of Ng and Pan, but it offers the ad- In the transitional zone, they interpolated the friction factor ( f 2 )
vantage to correctly describe both Couette and Poiseuille flows. with a cubic polynom. Figure 5 shows that the sinus law for both
Transitional Flows. As proposed by Szeri 关19兴, it is neces- Couette and Poiseuille flows leads to friction factor curves of
sary, for transitional flows to balance turbulent viscosity by an similar form as the formulas proposed by Zirkelback and San
interpolation factor ␦ ␧ . This factor evolves gradually from 0 for Andres.
the laminar zone to 1 for the fully turbulent zone. Szeri used a
power law 共3兲 in the transitional zone to determine the interpola-
tion factor ␦ ␧ :

再 ␦ ␧ ⫽0
␦ ␧⫽ f 共 ␣ 兲
␦ ␧ ⫽1
␣⬍␣1
␣ 1⭐ ␣ ⭐ ␣ t
␣⬎␣t
(3)

f 共 ␣ 兲 ⫽1⫺
␣ t⫺ ␣

␣ t⫺ ␣ 1 冊 1/8
.

This interpolation law leads to sharp variations near the turbulent


limit 共Fig. 3兲, so numerical convergence is difficult to reach. To
avoid these abrupt variations, a sinus law 共4兲 that evolves
smoothly from 0 to 1 was selected.

f 共 ␣ 兲⫽
冋 1⫺cos 冉 ␣ 1⫺ ␣
␣ 1⫺ ␣ t
␲ 冊册 (4)
2
To validate the current interpolation law, friction factors obtained
by the sinus law are presently compared to those available in the
literature. Missimer and Thomas 关20兴 studied transitional plane
Couette flow. Their results are in good agreement with Reichardt’s
experimental data. The Fig. 4 presents a comparison of friction Fig. 4 Comparison of friction factor f 1 in transition to turbu-
factor ( f 1 ) versus Couette Reynolds number Rec obtained by Mis- lence versus Couette Reynolds number Rec

Journal of Tribology JULY 2002, Vol. 124 Õ 517

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm


⳵Fz ⳵Fz ⳵Mx
K zz ⫽⫺ , D zz ⫽⫺ , K xx ⫽⫺ , D xx ⫽
⳵h ⳵ ḣ ⳵␹ x
⳵Mx ⳵My
⫺ and K xy ⫽⫺
⳵␹˙ x ⳵␹ x
K y y ⫽K xx , K xy ⫽⫺K yx and D y y ⫽D xx ,
where F z , M x , and M y are, respectively, axial resultant force
共opening force兲, tilting and cross tilting restoring moments of the
fluid film on the seal ring. All dynamic coefficients are determined
numerically as done by Person and al. 关7兴.

4 Numerical Procedure
The pressure field is obtained by a finite element method. The
Galerkin method leads to a discrete form of the generalized Rey-
nolds equation 共2兲 and then, to an algebraic system. If the fluid
flow regime is laminar, the system is linear and the pressure and
velocities calculation is direct.
The numerical procedure of resolution of a non laminar case is
the following. For given operating conditions, the pressure field in
the sealing dam is found assuming a laminar flow regime and then
Fig. 5 Comparison of friction factor f 3 in transition to turbu- estimating the flow factor ␣ and the interpolation factor ␦ ␧ . In an
lence versus Poiseuille Reynolds number Rep iterative process, the turbulent viscosity and new fluid velocities
are next calculated. While ␣ and ␦ ␧ vary, the previous loop is
repeated. We reiterate the entire procedure as long as the pressure
field varies. From the fluid velocities and pressure, the flow rate,
For the three studied cases, significant variations appear at the the friction torque, the opening force 共axial force of the fluid on
beginning of the turbulent zone. The Elrod and Ng turbulence the seal ring兲 and the tilting torques are computed.
model is probably not as accurate for low Reynolds number as for
high Reynolds number. But generally, it can be noted that the
shape of sinus law curves are satisfactory. Moreover it was nec-
5 Results
essary to choose the same transition Reynolds numbers 共Recl , A typical primary hydrostatic seal of a nuclear coolant pump is
Rect , and Rept兲 as authors 关10,20,21兴 in order to carry out com- studied. Dimensions of the studied seal and nominal running con-
parisons. ditions are presented in Table 1. The gap ratio is the ratio of the
flat zone clearance h and the tapered value e 共Fig. 7兲. The effect of
Fluid Film Dynamic Coefficients. Fluid film dynamic coef- two operating condition variations on the seal behavior is pre-
ficients are obtained by small motion perturbations around an axi- sented. From an initial case where the flow regime is laminar, the
symetric equilibrium position. As shown by Green and Etsion 关4兴, seal clearance is gradually increased until the fluid flow becomes
only five dynamic force coefficients are necessary to analyze the
fully turbulent. The dimensionless clearance H̄ varies from 0.25 to
seal ring dynamic behavior, i.e.:
1.25. A similar study is carried out with variation of rotation
speed. All results are presented in the dimensionless form pro-
posed by Koga and Fujita 关22兴. Furthermore, figures present evo-
lution obtained by the sinus interpolation law and those obtained
by assuming the flow regime laminar and fully turbulent.

Table 1 Dimensions and nominal running conditions

Fig. 6 Comparison of friction factor f 2 in transition to turbu-


lence versus mean Reynolds number Rem

518 Õ Vol. 124, JULY 2002 Transactions of the ASME

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm


Fig. 7 Hydrostatic seal with a flat face

Clearance Effect. The Fig. 8 presents the evolution of the


minimal and maximal value of the flow factor ␣ that indicates the
fluid flow regime, as the function of H̄. For the low value of the
film thickness, the flow is mainly laminar. The flow is fully tur-
bulent when H̄ reached 1.25. It is interesting to note that the flow Fig. 9 The friction torque and the leakage rate versus the
is not completely laminar when the seal operates at nominal func- clearance
tioning conditions (H̄⫽0.5).
The appearance of turbulence in the sealing gap leads to an
increase in apparent viscosity. Consequently the friction torque of opening force curve is S-shaped. Thus, there is an interval of seal
the fluid on the seal ring, which first of all decreases while the gap clearance 共0.9;1.15兲 that leads to unstable behavior, i.e.,: F̄ z in-
ratio rises, increases when the flow becomes turbulent 共Fig. 9兲.
Furthermore, the increase in leakage with the increasing thickness creases with H̄ and K̄ zz is negative. Note that positive stiffness of
is reduced by the effect of turbulence. elastic elements of the seal 共springs and secondary seals兲 can miti-
gate this problem.
Figure 10 presents the evolution of the opening force F̄ z and the As shown by Fig. 10, if the value of the closing force which is
axial stiffness coefficient K̄ zz versus the seal clearance H̄. It is opposed to the opening force is approximately 0.71, two axial
well known that, in laminar flow regime, the axial dynamic be- equilibrium positions of the seal are possible because of the
havior is stable when the gap is convergent in the flow direction S-shape of F̄ z curve 共the third one is not a stable equilibrium
关8兴: F̄ z decreases with an increase in clearance and K̄ zz is positive. position兲. Consequently, a small perturbation can lead to the pas-
As shown by Fig. 10, this typical behavior is observed for the sage of a small thickness equilibrium position (H̄⬵0.85) with a
lower values of the clearance. In turbulent flow regime, it is also larger one (H̄⬵1.25) and then a great increase in leakage rate.
found for the higher values of the clearance. For film thickness The evolution presented on Fig. 10 can be explained in the
corresponding to the transition to turbulence film flow regime, the following way. The increase in film thickness leads to appearance

Fig. 8 The minimal and the maximal value of the flow factor Fig. 10 The opening force and the axial stiffness coefficient
versus the clearance versus the clearance

Journal of Tribology JULY 2002, Vol. 124 Õ 519

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm


Fig. 11 Comparison of pressure and dimensionless viscosity Fig. 13 The axial and the tilting damping coefficients versus
profiles obtained in both laminar and non laminar flow regimes the clearance
with a dimensionless seal clearance of 0.704

of turbulence at the outer radius while the flow remains laminar ment of the stator leads to two zones diametrically opposite where
for lower radius. It is a ‘‘reverse transition’’ that is typical of the film thickness is minimal and maximal. Thus, when the de-
accelerated radial flows 关23兴. The appearance of turbulence leads crease in opening force with the clearance is maximal (H̄
to an increase of apparent viscosity at the outer radius. Conse- ⫽0.75), the tilting stiffness coefficient is also maximal. On the
quently, the pressure profile is modified compared to the one ob- other hand, when F̄ z growths with the gap ratio, the tilting mo-
tained in the isoviscous laminar flow regime that depends only on ment is not opposed any more to the misalignment and tends to
the film thickness 共Fig. 11兲. When the seal clearance increases, the increase it. But, as explained previously, the clearance values cor-
boundary of the ‘‘reverse transition’’ moves progressively toward responding to negative values of stiffness coefficients are not pos-
lower radius values until the film fluid flow regime is fully turbu- sible equilibrium positions. So the seal always operates with posi-
lent throughout the film. Thus the width of the isoviscous laminar tive stiffness coefficients.
zone decreases and the pressure profile evolves according to the The cross tilting stiffness coefficient K̄ xy is due to hydrody-
apparent viscosity profile. namic effect. As expected, the thicker the film is, the more the
Figure 12 shows that the tilting stiffness coefficient K̄ xx that is hydrodynamic effect is attenuated 共Fig. 12兲. Furthermore it is
due to hydrostatic effect evolves similarly as K̄ zz . The misalign- greatly dependent of the rotation velocity and the fluid viscosity.
So turbulence has a beneficial influence on K̄ xy because of the
apparent viscosity increase.
Both axial and tilting damping coefficients D̄ zz and D̄ xx are
typically due to squeeze effect that is mainly influenced by the
film thickness and the fluid viscosity. Hence, the appearance of
turbulence in the sealing dam slows down the decrease of damp-
ing coefficients linked to the clearance increase 共Fig. 13兲. Green
and Etsion 关3兴 show that the ratio K̄ xy /⍀̄•D̄ xx is equal to ⫺0.5 for
narrow aligned face seals operating in laminar flow regime. Al-
though the seal is width, the ratio K̄ xy /⍀̄•D̄ xx is equal to ⫺0.5 in
the laminar flow regime in the present study. When the fluid flow
is non laminar the ratio K̄ xy /⍀̄•D̄ xx stays in an interval of ⫾20
percent around ⫺0.5.
Effects of Rotation Velocity. In this study the rotation veloc-
ity varies from 0.3 to 4.7 times the nominal velocity. It can be
seen on Fig. 14, that the flow is totally laminar when the velocity
is minimal and fully turbulent for the higher value of ⍀̄. As ex-
pected, the higher the velocity value is, the more the friction
torque C̄ f is significant 共Fig. 15兲. The torque rise is amplified by
the increase in apparent viscosity due to turbulence. However the
turbulence has a beneficial influence on leakage rate Q̄ that de-
creases when the fluid flow regime becomes nonlaminar 共Fig. 15兲.
In the laminar region, the opening force F̄ z is independent of
Fig. 12 The tilting and the cross tilting stiffness coefficients the velocity 共Fig. 16兲. In the nonlaminar zone the axial force
versus the clearance evolves similarly as in the previous study. At the beginning of

520 Õ Vol. 124, JULY 2002 Transactions of the ASME

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm


Fig. 14 The minimal and the maximal value of the flow factor Fig. 16 The opening force and the axial stiffness coefficient
versus the angular velocity versus the angular velocity

turbulence are limited. The same comments can be made on the


transition the F̄ z decreases until a minimum and then increases curve of tilting stiffness coefficient K̄ xx versus angular velocity
until fully turbulent zone where it starts a slow decrease. That can presented on Fig. 17.
be explained as previously by the appearance of a ‘‘reverse tran- The increase in velocity leads to a great rise of the cross tilting
sition’’ which is initially localized on the outer radius and progres-
stiffness coefficient K̄ xy because of its hydrodynamic origin 共Fig.
sively moves towards the inner radius while modifying the pres-
17兲. This phenomenon is amplified by the apparent viscosity
sure profile. It is interesting to note that, for a given closing force,
growth due to turbulence.
an increase in velocity leads to a lower clearance equilibrium
The velocity growth and then turbulence appearance has also a
position.
beneficial effect on both axial and tilting damping coefficients D̄ zz
The axial stiffness coefficient K̄ zz evolves similarly as previ-
and D̄ xx 共Fig. 18兲. Whereas D̄ zz and D̄ xx remain constant in the
ously 共Fig. 16兲. At the beginning of transition K̄ zz reaches a maxi-
laminar zone, they start to rise when the fluid flow becomes non
mum and then decreases until a minimal value that is positive. As
laminar. This is due to the increase in apparent viscosity that is
the film remains thin, the axial stiffness coefficient is higher than
in previous study, and the influence of velocity growth and then favourable to squeeze effect. Note that the ratio K̄ xy /⍀̄•D̄ xx re-
mains close to ⫺0.5 as observed previously.

Fig. 15 The friction torque and the leakage rate versus the Fig. 17 The tilting and the cross tilting stiffness coefficients
angular velocity versus the angular velocity

Journal of Tribology JULY 2002, Vol. 124 Õ 521

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm


Nomenclature
A f ⫽ ␲ (r 2o ⫺r i2 ) seal interface section, 共m2兲
C f ⫽ friction torque, 共N•m兲
C̄ f ⫽ 2C f h/ 关 ␲ ␮ ␻ (r 4o ⫺r i4 ) 兴 dimensionless fric-
tion torque
⫺⳵M x
D xx ⫽ tilting damping coefficient,
⳵␹˙ x
共N•m/rad•s⫺1兲

D̄ xx ⫽ ⫺ ⳵ M̄ x dimensionless tilting damping coef-


⳵␹˙ x ␻
ficient
⫺⳵Fz
D zz ⫽ axial damping coefficient, 共N/m•s⫺1兲
⳵ ḣ
D̄ zz ⫽ ⫺ ⳵ F̄ z dimensionless axial damping coeffi-
⳵Hថ␻
cient
e ⫽ tapered value, 共m兲
F z ⫽ opening force, 共N兲
F̄ z ⫽ F z /( P o A f ) dimensionless opening force
f 1 ⫽ 2 ␶ W /( ␳␻ 2 r 2 ) friction factor
Fig. 18 The axial and the tilting damping coefficients versus f 2 ⫽ h 2 / ␳ G 冑(Ṽ 2 ⫹Ṽ 2 ) friction factor
1 x y
the angular velocity f 3 ⫽ 2 ␶ W /( ␳ Ṽ r2 ) friction factor
h ⫽ centerline seal clearance, 共m兲
6 Conclusions h 0 ⫽ e reference centerline seal clearance, 共m兲
Face seals such as primary hydrostatic seals of reactor coolant H̄ ⫽ h/h 0 dimensionless seal clearance
pump operate at the limit of laminar regime. A modification of H i ⫽ axial position of ring i, m
running conditions can lead to a non laminar fluid flow in the K xx ⫽ ⫺ ⳵ M x / ⳵␹ x tilting stiffness coefficient,
sealing dam. With an aim of studying such phenomena, a numeri- 共N•m/rad兲
cal model has been developed. The fluid flow regime is deter- K̄ xx ⫽ ⫺ ⳵ M̄ x dimensionless tilting stiffness coeffi-
⳵␹ x
mined by means of a flow factor that takes into account both
cient
Couette and Poiseuille flows. The Elrod and Ng model 关13兴
⫺⳵M y
coupled to an interpolation law is used to study non laminar flows. K xy ⫽ ⳵␹ cross tilting stiffness coefficient,
The influence of an increase in seal clearance and in angular x
共N•m/rad兲
velocity on the seal performances has been studied.
Generally speaking, the appearance of turbulence results in an K̄ xy ⫽ ⫺ ⳵ M̄ y dimensionless cross tilting stiffness
apparent viscosity rise. As expected, the dissipated power in- ⳵␹ x
creases but the leakage rate is slowed down. For the two studied coefficient
cases, numerical results show that opening force reaches a mini- ⫺⳵Fz
mal value in the transitional zone. In addition, axial and tilting K zz ⫽ ⳵ h axial stiffness coefficient, 共N/m兲
stiffness coefficients present an optimal value at the beginning of
transition to turbulence. Unfortunately, these coefficients decrease K̄ zz ⫽ ⫺ ⳵ F̄ z dimensionless axial stiffness coeffi-
sharply until a minimal value that can be negative, just before ⳵ H̄
fully turbulent regime. Generally, numerical predictions point out cient
that nonlaminar flow has a beneficial influence on damping and M x , M y ⫽ tilting torque on x and y axis, 共N•m兲
hydrodynamic stiffness coefficients. M̄ x ⫽ M x /( P o A f r m ) dimensionless tilting torque
As discussed by Yasuna and Hughes 关9兴, this study highlights on x axis
the fact that transition to turbulence cannot be neglected. Indeed, M̄ y ⫽ M y /( P o A f r m ) dimensionless tilting torque
the transition from laminar to fully turbulent flow regime due to on y axis
an increase in seal clearance cannot be a smooth evolution be-
P̄ ⫽ mean fluid pressure
cause of unstable positions which occur in the second half of
Q ⫽ leakage rate, 共m3/s兲
transition. Furthermore, fluid film dynamic coefficients are greatly
influenced by the change of fluid flow regime. Our results are in Q̄ ⫽ 6 ␮ r i Q/ 关 ␲ e 3 r o ( P o ⫺ P i ) 兴 dimensionless
agreement with predictions made for annular seals operating in leakage rate
transition to turbulence 关10兴. r e ⫽ inner radius of the tapered zone, 共m兲
In order to obtain more accurate results for large seal clearance, ReC ⫽ ␳␻ rh/ ␮ Couette Reynolds number
it would be interesting to take into account fluid inertia effects and Rem ⫽ ␳ 冑(Ṽ 2 ⫹Ṽ 2 )h/ ␮ mean Reynolds number
x y
use a turbulence model adapted to low Reynolds number. On the Rep ⫽ ␳ Ṽ r h/ ␮ Poiseuille Reynolds number
other hand, a numerical and experimental comparison is necessary r m ⫽ (r o ⫹r i )/2 mean radius, 共m兲
to validate our model but, unfortunately, no experimental data are V̄ x , V̄ y ⫽ mean fluid velocity components, 共m/s兲
now available in the literature. V xi , V yi , V zi ⫽ ring i velocity component, 共m/s兲
␣ ⫽ flow factor
Acknowledgment ␤ i ⫽ taper angle of ring i, 共rad兲
Professors T. A. Davis and I. S. Duff are gratefully acknowl- ␹ x , ␹ y ⫽ seal ring tilting angles on x and y axis,
edged for their work on sparse matrix that is available on line 共rad兲
共www.cis.ufl.edu/⬃davis/techreports.html兲. ␦ ␧ ⫽ interpolation factor

522 Õ Vol. 124, JULY 2002 Transactions of the ASME

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm


␧ t , ␧ W , ␧ R , ␧ C ⫽ turbulent diffusivity, wall turbulent diffusiv- 关6兴 Wileman, J., and Green, I., 1995, ‘‘The Rotordynamics Coefficients of Eccen-
tric Mechanical Seals,’’ ASME J. Tribol., 118, No. 1, pp. 215–224.
ity, Reichardt turbulent diffusivity, Clauser
关7兴 Person, V., Tournerie, B., and Frêne, J., 1997, ‘‘A Numerical Study of the
turbulent diffusivity Stable Dynamic Behavior of Radial Face Seals With Grooved Faces,’’ ASME
␮ ⫽ fluid dynamic viscosity, 共Pa•s兲

冉 冊
J. Tribol., 119, No. 3, pp. 507–514.
H1 ␧t dz 关8兴 Green, I., and Etsion, I., 1985, ‘‘Stability Threshold and Steady-state Response

¯ ⫽ 兰H ␮• 1⫹␦␧• 共H ⫺H2兲 of Noncontacting Coned-Face Seals,’’ ASLE Trans., 28, No. 4, pp. 449– 460.
2 ␯ ␮ 1 关9兴 Yasuna, J. A., and Hughes, W. F., 1994, ‘‘Squeeze Film Dynamics of Two-
mean value across the film of the dimen- phase Seals: Part II—Turbulent Flow,’’ ASME J. Tribol., 116, No. 3, pp. 479–
sionless apparent viscosity 488.
␯ ⫽ ␮/␳ fluid kinematic viscosity 关10兴 Zirkelback, N., and San Andres, L., 1996, ‘‘Bulk-Flow Model for the Transi-
tion to Turbulence Regime in Annular Pressure Seals,’’ Tribol. Trans., 39, No.
␳ ⫽ fluid density, 共kg/m3兲 4, pp. 835– 842.
␶ ⫽ shear stress, 共N/m2兲 关11兴 Ransom, D. L., and San Andres, L., 1999, ‘‘Identification of Force Coefficients
␶* ⫽ H 2 • 兩 ␶ 兩 / ␮ • ␯ dimensionless shear stress from a Gas Annular Seal—Effect of Transition Flow Regime to Turbulence,’’
␶ W ⫽ wall shear stress, 共N/m2兲 Tribol. Trans., 42, No. 3, pp. 487– 494.
␻ ⫽ angular velocity, 共rad/s兲 关12兴 Mayer, E., 1989, ‘‘Performance of Rotating High Duty Nuclear Seals,’’ Lubr.
Eng. 45, No. 5, pp. 275–286.
␻ 0 ⫽ nominal angular velocity, 共rad/s兲 关13兴 Elrod, H. G., and Ng, C. W., 1967, ‘‘A Theory for Turbulent Films and its
⍀̄ ⫽ ␻ / ␻ 0 dimensionless angular velocity Application to Bearings,’’ ASME J. Lubr. Technol. 89共3兲, pp. 346 –362.
关14兴 Lin, J. F., and Yao, C. C., 1996, ‘‘Hydrodynamic Lubrication of Face Seal in a
Subscripts Turbulent Flow Regime,’’ ASME J. Tribol., 118, No. 3, pp. 589– 601.
H 关15兴 Gregory, R. S., 1975, ‘‘Operating Characteristics of Fluid-film Thrust Bearing
f̃ ⫽ 兰 H 1 f dz/(H 1 ⫺H 2 ) mean value of f across the film Subjected to High Shaft Speeds,’’ 2nd Leeds-Lyon Symposium, Super Lami-
2
i ⫽ inner nar Flow in Bearing, Paper VIII 共4兲, pp. 154 –162.
关16兴 Frêne, J., 1977, ‘‘Tapered Land Thrust Bearing Operating in Both Turbulent
l ⫽ laminar and Laminar Regimes,’’ ASLE Trans., 21, No. 3, pp. 243–249.
o ⫽ outer 关17兴 Souchet, D., 1991, ‘‘Comportement Thermohydrodynamique des Butées à Pa-
t ⫽ turbulent tins Oscillants en Régime Laminaire et Turbulent,’’ Ph.D. Thesis, University of
0 ⫽ nominal or reference value Poitiers.
1 ⫽ stator 关18兴 Bassani, R., and Piccigallo, B., 1992, Hydrostatic Lubrication, Tribology Se-
ries, 22, Elsevier.
2 ⫽ rotor 关19兴 Szeri, A. S., 1980, Turbulence, Inertia, and Thermal Effects in Fluid Film
Bearings, Chap. 5, ‘‘Tribology: Friction, Lubrication, and Wear,’’ Mc Graw-
References Hill B.C., pp. 229–294.
关20兴 Missimer, J. R., and Thomas, L. C., 1983, ‘‘Analysis of Transitional and Fully
关1兴 Etsion, I., 1982, ‘‘Dynamic Analysis of Noncontacting Face Seals,’’ ASME J.
Turbulent Plane Couette Flow,’’ ASME J. Lubr. Technol., 105, No. 2, pp.
Lubr. Technol., 104, No. 4, pp. 460– 468.
364 –368.
关2兴 Tournerie, B., and Frêne, J., 1984, ‘‘Computer Modelling of the Functioning
Modes of Noncontacting Face Seals,’’ Tribol. Int., 17, No. 5, pp. 269–276. 关21兴 Patel, V. C., and Head, M. R., 1969, ‘‘Some Observations on Skin Friction and
关3兴 Green, I., and Etsion, I., 1986, ‘‘Nonlinear Dynamic Analysis of Noncontact- Velocity Profiles in Fully Developed Pipe and Channel Flows,’’ J. Fluid Mech.,
ing Coned-face Mechanical Seals,’’ ASLE Trans., 29, No. 3, pp. 383–393. 38, pp. 181–201.
关4兴 Green, I., and Etsion, I., 1983, ‘‘Fluid Film Dynamic Coefficients in Mechani- 关22兴 Koga, T., and Fujita, T., 1986, ‘‘The Hydrostatic Noncontact Seal Including
cal Face Seals,’’ ASME J. Lubr. Technol., 105, No. 2, pp. 297–302. Fluid Inertia Effect,’’ ASLE Trans., 29, No. 1, pp. 35– 42.
关5兴 Wileman, J., and Green, I., 1991, ‘‘The Rotordynamics Coefficients of Me- 关23兴 Singh, A., Vyas, B. D., and Powle, U. S., 1999, ‘‘Investigations on Inward
chanical Seals Having Two Flexibly Mounted Rotors,’’ ASME J. Tribol., 113, Flow Between Two Stationary Parallel Disks,’’ Int. J. Heat Fluid Flow, 20, pp.
No. 4, pp. 795– 804. 395– 401.

Journal of Tribology JULY 2002, Vol. 124 Õ 523

Downloaded From: http://tribology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jotre9/28707/ on 04/15/2017 Terms of Use: http://www.asme.org/about-asm

You might also like