Sunteți pe pagina 1din 13

ARTICLE IN PRESS

Tribology International 43 (2010) 654666

Contents lists available at ScienceDirect

Tribology International
journal homepage: www.elsevier.com/locate/triboint

Damping and added mass coefcients for a squeeze lm damper using the
full 3-D NavierStokes equation
Changhu Xing a, Minel J. Braun a,, Hongmin Li b
a
b

Department of Mechanical Engineering, The University of Akron, Akron, OH 44325-3903, USA


Ametek Inc., Kent, OH 44240, USA

a r t i c l e in f o

a b s t r a c t

Article history:
Received 6 October 2008
Received in revised form
9 May 2009
Accepted 9 October 2009
Available online 7 November 2009

Direct and cross-coupled damping coefcients are developed for the 2p-lm, p-lm (Gumbel cavitation
condition) and homogeneous two-phase mixture lms in a squeeze lm damper. The numerical
simulation uses the CFD-ACE+ commercial software, which employs a nite volume method for the
discretization of the NavierStokes equations (NSE). In order to determine the dynamic coefcients, the
NSE is combined with a nite perturbation method applied to the equivalent journal of the damper. It
was found that for the 2p-lm and the Gumbel conditions, the damping coefcients exhibit linear
characteristics, while the homogeneous cavitation model yields nonlinear coefcients. Using the CFDACE+ , the inertia/added mass coefcients are derived for the limiting cases of the short and long
dampers, respectively. The rst set of forces is calculated by setting the uid density to its actual value.
The second set of forces is calculated when the density of the uid is set close to zero (1E-10 kg/m3),
thus practically eliminating the effects of the inertia terms. Subtracting the two sets of forces from each
other, allows the determination of the inertia component contribution and the corresponding inertia
coefcients. By varying the density, dynamic viscosity and whirling speed, it was found that the inertia
coefcients follow a single curve represented by a function dependent on the modied Reynolds
number, Re*. The inertia coefcients presented in this study are compared with the ones reported by
other researchers that used the modied Reynolds equation. Some differences were found between the
NSE based results and the Reynolds equation based outcomes. This is attributed to the threedimensional effects introduced by the totality of the terms comprised in the full NSE.
& 2009 Elsevier Ltd. All rights reserved.

Keywords:
Squeeze lm damper
Dynamic coefcients
CFD-ACE +

1. Introduction
The squeeze lm damper (SFD) performance was studied by
many researchers [1,2] and there are well-developed analytical
and perturbation methods to simulate its behavior. Pietra and
Adiletta [3] provided a comprehensive review of the development
of the SFD. Gunter [1] derived the Reynolds equation for a squeeze
lm damper using a short bearing approximation and found that
for a circular centered orbit (CCO) motion, the radial damping
coefcient Brt plays the role of an equivalent stiffness [4]. Hahn [2]
derived a table for damping coefcients for a short, and a nite
length damper by using the Warner approximation [5]. With the
increasing number of practical lubrication problems which
involve moderate to large Reynolds numbers, evaluation of the
inertia effects has become important. Following numerical
simulation and validating experimental work, it is now well
accepted, that the inertia forces exert a noteworthy effect on the

 Corresponding author.

E-mail address: mjbraun@uakron.edu (M.J. Braun).


0301-679X/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.triboint.2009.10.005

pressure distribution and journal stability. Tichy [6] analyzed the


ow patterns in a SFD and found that the shape of the pressure
distribution due to viscous effects follows a sinusoidal behavior,
but inertia affects the pressure according to a cosinusoidal law.
Thus, the damper force would undergo a signicant phase shift in
its shape and magnitude, by comparison with the one caused by
viscous forces alone. San Andres [7] reached a similar conclusion,
and stated that the effect of uid inertia is signicant even for
moderate Reynolds numbers. From the experiments of Jung et al.
[8], it was observed that the positive peak pressure is smaller than
the negative one.
For the evaluation of the inertia force coefcients at moderate
to high modied Reynolds numbers Re b 1, two different
approaches were proposed: (i) momentum based [9] and (ii)
energy approximation [10]. Both methods used the modied
Reynolds equation, including the simplied inertia terms
uj @ui =@xj . Lund et al. [11], used a perturbation method with the
modied Reynolds equation to account for inertia effects, and
related the inertia forces to inertia coefcients. The effect of uid
inertia was further demonstrated experimentally [12], when it
was shown that inertia can totally dominate the dynamic

ARTICLE IN PRESS
C. Xing et al. / Tribology International 43 (2010) 654666

Nomenclature
Arr, Art, Atr, Att inertia coefcients, kg
Brr, Brt, Btr, Btt damping coefcients, N s/m
Brr ; Brt ; Btr ; Btt dimensionless
damping
coefcients,
B B=mor
c
clearance, mm
dx
displacement of the journal surface in x direction, m
dy
displacement of the journal surface in y direction, m
e
eccentricity, mm
fv, fg; av, ag Mass/volume fraction of vapor(v) and gas (g)
F0r, F0t radial and tangential force for inertia-less cases, N
F0rt, F0tt radial and tangential force for inertia-less cases
running at CCO motion, N
Firt, Fitt radial and tangential force at CCO motion with inertia,
N
spring constant, N/m
Kr
Krr, Krt, Ktr, Ktt stiffness coefcients, N/m
L
length, mm

behavior. El-Shafei and Crandall [10] obtained inertia coefcients


for short and long dampers, respectively, using the short (SBA)
and long (LBA) bearing approximation for their energy based
analysis. Szeri et al. [9], adopted an averaging method, integrated
the NavierStokes equation over the lm thickness and obtained a
modied Reynolds equation which incorporated the inertia effect.
Zhang et al. [13] derived another modied Reynolds equation and
assessed the inertia force coefcients by both momentum and
energy methods.
All the literature reviewed above presents models based on the
1-D, or 2-D Reynolds equation, or a modied Reynolds equation.
However, by its physical assumptions and mathematical approach
a Reynolds equation based analysis cannot account for the full
physics of the 3-D effects that can only be modeled by means of
the NSE. The development of computer technology, and the
appearance and wide acceptance of validated 3-D, NSE based
commercial codes (CFD-ACE +, Fluent, CFX, Star-CD), turned the
full 3-D, NSE models incorporating inertia and distributed twophase ow effects into a viable alternative to the Reynolds
equation based models [14,15].
The objective of this work involves a numerical investigation of
the damping and inertia force coefcients for long and short
squeeze lm dampers. The 3-D full NSEs are applied to obtain the
pressure distribution and evaluate the load capacity and dynamic
coefcients of these two types of dampers. They represent two
limiting cases within which the performance of a nite length
damper must fall. By comparing the inertia coefcients from NSE
model with other researchers analytical results, this paper will
synthesize and discuss both the similarities and differences
existing between the NSE based results and those yielded by the
modied Reynolds equation.

m
p
R
Re*
Sj
t
uj
!
V
xj

mass, kg
pressure, Pa
radius, mm
modied Reynolds number, Re* = (rocR/m)(c/R)
j component of source term, kg/m2/s2
time, s
j component of velocity, m/s
velocity vector, m/s
j component of coordinate system, m
e
dimensionless eccentricity
y0
initial attitude angle, rad
mm, mv, mg viscosity of mixture, vapor and gas, Pa s
rm, rv, rg density of mixture, vapor and gas, kg/m3
r
dimensionless density, r=rref
f
attitude angle, rad
o
rotor rotational speed or the whirling speed of the
journal under synchronous motion, rad/s
or
critical speed, or =(Kr/m)0.5, rad/s
Or = o/or dimensionless whirling speed

baseline density of the lubricant is 870 kg/m3. In order to study


the effects of density on the inertia coefcients through the
variation of the Re*, the density was changed parametrically.
Geometry-wise, two types of dampers will be considered: (i) a
fully open, with axial atmospheric pressure boundary conditions
and (ii) fully sealed and no axial ow. For case (i) the small L/D
ratio (0.175), allowed comparison of the damper performance to
the short bearing theoretical results (SBA case), where the ow in
circumferential direction is neglected. For case (ii), the pressure
distribution and the associated dynamic force coincide with the
results based on the LBA case, where the ow in axial direction
can be ignored.

3. Computational models
The NSE and continuity equations are used to solve the
pressure distribution for the SFD. The commercial program used
in this endeavor is CFD-ACE+ , provided by the ESI Group [17]:
!
@r
rr V 0
@t

!
@ruj
@p
rruj V 
rmruj Sj
@t
@xj

In CFD-ACE + , the homogeneous cavitation model is coupled


with the momentum and continuity equations. The density of the
mixture is calculated by means of
1

fv

rv

The damper parameters used in this analysis are those


published by Taylor [16], and shown in Table 1. The chosen
Table 1
Reference values for the parameters of the squeeze lm damper.
m (kg)

Kr (N/m)

R (mm)

L (mm)

c (mm)

m (Pa s)

33.43

2.154e7

64.8

22.7

0.1

0.00266

fg

rg

1  fv  fg

rl

The void fraction ag can be expressed by

ag fg

2. Physical parameters and uid properties

655

r
rg

Since only gaseous cavitation is considered in this study, in


Eq. (3), fv = 0. The CFD-ACE+ algorithm is designed to use a
constant viscosity uid under isothermal conditions, thus it
becomes necessary to write and add separate code script to
account for the dynamic viscosity variation as a function of void
fraction (or mass fraction) as required by the analysis presented
herein. The equivalent dynamic viscosity of the mixture can be
written as a function of the void/volume fraction ag as

mm ag mg 1  ag ml

ARTICLE IN PRESS
656

C. Xing et al. / Tribology International 43 (2010) 654666

center the journal in the radial clearance. Note that the assembly
of the rotor (1), ball bearings (2) and the whirling ring form an
equivalent journal (EJ) (3).
Using the free body diagram (FBD) of Fig. 1b, one can write the
corresponding radial and tangential overall forces for a uid lm
bearing as

The gas mass fraction spatial distribution is governed by a


transport equation, very similar in its build-up to the NSE
!
@
6
rf r  r V f r  mm rf Re  Rc
@t
In Eq. (6) the Re and Rc terms (kg/m3 s), are the evaporation and
condensation phase terms related to vaporous cavitation. Since
vaporous cavitation is not considered here, the coefcients for
evaporation and condensation in Re and Rc are disabled in the
input sequence. To start the computations, one has to prescribe a
certain initial amount of noncondensable gas mass fraction, f=f0.
As the calculations proceed, the total gas mass fraction while
staying constant at f0, will distribute itself spatially to appropriately match the pressure map around the damper. Such an
approach simulates a steady- or quasi steady-state process known
as pseudo-cavitation when no more air is ingested from the
ambient or diffused out from the existing oil.
The change of clearance is implemented by prescribing the
CCO motion for a given journal eccentricity. Numerically, a grid
deformation module coupled to the NSE algorithm allows for the
incorporation of the motion of the journal and the subsequent
geometry change of the clearance. Thus, while the solid
boundaries are changing due to journal motion, the grid
deformation module automatically re-meshes the uid domain
(the clearance). The motion of the boundaries is resolved into
translational movements in the x- and y-directions as

Stiffness coefficient terms Damping coefficient terms

z}|{
_
 Brr ce_  Brt cef

z}|{
Fr Krr ce  Krt cef

Inertia coefficient terms

z}|{
_ 2  Art cef
2ce_ f
_
 Arr ce  cef

_  Atr ce  cef
_ 2
Ft  Ktr ce  Ktt cef  Btr ce_  Btt cef
2ce_ f
_
 Att cef

10

However, the SFD does not provide stiffness. If in addition, the


inertia effects are neglected (the Aij terms), the total radial and
tangential forces are functions of damping only. They can then be
calculated by integrating the pressures obtained by means of the
well-known classical Reynolds equation [18]. These forces are
_
F0r  Brr ce_  Brt cef

11
12

dx ecos ot  cos y0

_
F0t  Btr ce_  Btt cef

dy esin ot  sin y0

If the EJ of the SFD has a circumferential (CCO) motion only,


then e_ 0, and the tangential and radial forces can be written as

4. Implementation scheme for dynamic coefcients

_ ) Brt  F =cef
_
F0rt  Brt cef
0rt

13

_ ) Btt  F =cef
_
F0tt  Btt cef
0tt

14

If the pressures are calculated using the complete NSE, the


effect of inertia is implicitly incorporated in the nal value of the
resulting forces (Firt and Fitt), and thus an additional set of terms
needs to be added to Eqs. (13) and (14) to account for the
difference in the magnitudes of the inertia-less (F0rt and F0tt)
versus inertia-with (Firt and Fitt) forces.

The uid lm forces are calculated by integrating the pressures


computed by the Reynolds, or NSE equations. According to the
linearized dynamic theory [18], these forces can be also written in
terms of equivalent stiffness, damping and inertia coefcients.
The geometry of the SFD considered in this paper is given in Fig. 1.
The ball bearing (2) is mounted on the rotor (1) with its inner race
rotating with the rotor while the outer one is afxed to the
whirling ring (7). The latters motion is constrained to a
translational movement only, due to the anti-rotation pins (6). A
set of retainer springs, Fig. 1b, with the stiffness of Kr, is used to

_ 2  Art cef

Firt  F0rt Arr cef

15

_  Att cef

Fitt  F0tt Atr cef

16

Line of centers

.
Bttc
3
2
1

.
Bttc
.2
Arrc
Ob
Oj

Fr

.
= 

Ob Ft

.
Atrc2

Oj

Kr

Oil film
4
6

Kr
5

Retainer spring

Fig. 1. Schematic illustration of a squeeze lm damper and CCO motion: (a) damper, (b) FBD 1 Rotor, 2 rolling-element bearing, 3 equivalent journal, 4 squeeze lm,
5 housing, 6 anti-rotation pin, 7 whirling ring Ob, bearing center line, Oj, journal center line.

ARTICLE IN PRESS
C. Xing et al. / Tribology International 43 (2010) 654666

0, Eqs. (15) and (16)


For the steady state case, in which f
become
2

_ ) Arr F  F =cef
_
Firt  F0rt Arr cef
0rt
irt
_ ) Atr F  F =cef
_
Fitt  F0tt Atr cef
0tt
itt

Re*

+
*
O
X

17
4e+5

18

A look at the dimensions of the terms in Eqs. (17) and (18)


_2
reveals that all forces are measured in N, while the quantity cef
2
has the units of m/s . Thus the Aij terms have the units of mass kg.
These inertia terms are also known as added or virtual mass.
The inertia and damping based forces as well as the location of
the CCO are shown in the FBD of Fig. 1b. The computational
procedure owcharts for the computation of the Bij and Aij are
presented in Appendices A and B.

Pressure (Pa)

6e+5

657

=
^
2e+5

Minimum clearance

Increase with Re *

0
1.71
3.43
8.56
17.13
25.69
34.25

^
=

OX
*

increase with Re *

+
Gumbel condition

0
Decrease with Re *

-1e+5
0

2
3
4
Circumferential direction (rad)

5. Numerical experiments

6. Pressure distribution using NSE


The pressure distributions for dampers under SBA and LBA
conditions, running at 10,000 rpm (Re* =3.43) and eccentricity 0.5
are shown in Fig. 2; they are plotted using Re* as a parameter.
Note that while the rotors ((1), Fig. 1) center Oj executes a
rotational motion around Ob inside the rolling element bearings
(2), EJ executes a whirling sinusoidal motion as forced by the pins
(6). The parametric change of the Re* occurs through changes

1e+7

Film thickness

Re*

+
*
O
X

8e+6
Pressure (Pa)

The homogeneous cavitation model was presented and


validated [19] by comparing the CFD-ACE+ calculations with the
experimental results of Diaz and San Andres [20]. It was shown
that for a low to moderate gas volume fraction (0 o ag o0.5), both
mass-fraction (f) and volume-fraction (a) based viscosity models
can simulate the pressure eld very well.
Based on the actual boundary conditions the computation
start-up (initial) condition for pressure distribution can be set to
any reasonable value. However computations have a faster
convergence with a well chosen (hindsight) initial condition. For
the whirling of the equivalent journal at a given eccentricity, the
pressure eld reaches a quasi steady-state, in the sense that the
magnitude and shape of the circumferential prole of the pressure
eld remain constant, but rotate synchronously with the orbit
after a short transient time. The required time period to reach the
quasi steady-state motion, as well as the correct time step size is
determined by preliminary numerical calculations. Grid independence tests were conducted as well, in order to obtain grid
independent solutions. To assess the grid independence, a baseline case was run with 128 grid points in the circumferential
direction, 8 across the damper lm (radial direction) and 12 in the
axial direction. The grid density for two subsequent cases was
increased by 1.5 times in every direction. When two subsequent
calculations yielded a relative deviation smaller than 2%, we
considered that computational grid convergence was achieved.
The authors have carefully investigated the time step size
effect. With a CFL (Courant number) o1, and for Re* 430, the
results were time-step independent and converged with no uid
numerical instability. Subsequent experiments with a CFL numbers varying between 5 and 10 yielded also stable uid
congurations. Integration of pressures in the axial and circumferential directions for 5oCFLo10, resulted in total forces that
were only 0.1% different from those calculated with CFLE1. Thus
all subsequent numerical experiments presented in this paper are
performed with a time step of CFL E5.

Maximum clearance
0.1
0.05
0
0
1

=
^

6e+6

6 2

0
1.71
3.43
8.56
17.13
25.69
34.25

Increase then
decrease with Re*

+ *

^
=

increase with Re *

4e+6

Minimum clearance

2e+6
Decrease with Re*
Gumbel condition

0
0

2
3
4
Circumferential direction (rad)

6 2

Fig. 2. Pressure distributions along the circumferential direction in the axial plane
of symmetry: (a) SBA and (b) LBA.

effected in density, while all the other variables are kept constant.
The baseline case is calculated for Re* E0. This indicates that the
density was set up approximately to zero (1.E 10), and therefore
the inertia terms in the NSE (which multiply the density), have
negligible magnitudes ( E0). Both Figs. 2a and 2b show the solid
line prole to be the classical full Sommerfeld curve for an
instantaneous position of the EJ, as would be yielded by a
Reynolds equation solution (no inertia, Re* E0). For both the SBA
and LBA cases one can see that with the increase in the inertia
effect (as Re* increases), the positive peak pressures grow in
magnitude while shifting away from the minimum clearance; at
the same time the negative pressures peaks would increase as
well while moving towards the minimum clearance. There are
also other differences in pressure behavior between the SBA and
LBA models for the damper. For the damper with SBA, both the
positive and negative peak pressures increase with the increase of
Re*. For the damper with LBA, the positive peak pressure
decreases rst, and then increases with the increase of Re*.
Under the Gumbel cavitation condition as applied to NSE
calculations, the gauge cavitation pressure is set to p= pcav =0.
This is consistent with this papers original assumption that only
gaseous cavitation is considered. One can see a second threshold

ARTICLE IN PRESS
C. Xing et al. / Tribology International 43 (2010) 654666

line at 105 Pa, which represents the absolute zero pressure. It


should be noted that when pressures fall below the gas saturation
pressure, gaseous cavitation ensues. When pressures fall at the
levels of vapor pressures then vaporous cavitation occurs. In a
regular oil containing dissolved gases, these gases saturation
pressures are much above any of the component oil hydrocarbons
vapor pressures [21,22].

7. Damping coefcients variation


Fig. 3 presents a comparison of direct and cross-coupled
damping coefcients for an eccentricity of e = 0.5, for the SBA,
when the dampers EJ is running at different whirling speeds and
the total mass fraction of gas, f0, is varied on a parametric basis.
The legend volume denotes that the dynamic viscosity was
modied according to Eq. (5). The damping coefcients for the
p-lm assumption (Gumbel) are constant for the eccentricities
used in these studies. This can be assigned to the fact that Fij
varies linearly with o (Eqs. (14) and (15)), thus causing Bij to
practically remain constant. If one were to compare the p- and
2p-lm results one would nd that most generally Brt (2p-lm)

Dimensionless damping coefficient (Brt)

0.20
0.18
0.16
0.14

Brt constant

0.12
0.10
0.08
0.06
0.04

-film
1E-5 gas_volume
1E-4 gas_volume
5E-4 gas_volume
1E-3 gas_volume

0.02
0.00
0.0

0.5

1.0
1.5
2.0
2.5
Dimensionless whirling speed (r)

3.0

3.5

Dimensionless damping coefficient (Btt)

0.40
-film
1E-5 gas_volume
1E-4 gas_volume
5E-4 gas_volume
1E-3 gas_volume

0.35
0.30
0.25

0.13

0.20
Btt constant

0.15
0.10

Dimensionless damping coefficient (Brt, Btt)

658

50

Brt, NSE
Brt, Reynolds Equ.
Btt, NSE
Btt, Reynolds Equ.

40

30

20

10

0
0.0

0.2

0.4
0.6
Eccentricity ()

0.8

1.0

Fig. 4. Dimensionless direct- and cross-coupled damping coefcient variation


with eccentricity for the LBA case under the Gumbel cavitation condition.

E0, and Btt (2p-lm) E2Btt (p-lm). For the homogeneous twophase cavitation, various gaseous void fraction cases yield variable
Brt and Btt coefcients. When the whirling speed is low (most
commonly associated with low eccentricities, e o0.5), the
pressure buildup around the circumference is small, and the
instantaneous pressures formed in the convergent and divergent
regions do not differ greatly. Consequently, any mass of gas
existing in the lm will distribute rather homogeneously around
the circumference causing large direct damping and small crosscoupled damping. With the increase in the whirling speed, at the
chosen eccentricity of 0.5, the pressure buildup around the
circumference becomes signicant causing specic migration of
the gas towards the low pressure zones and leaving little gas in
the high pressure regions. Under this scenario, the cross-coupled
damping increases until it reaches almost an asymptotic-like
growth, Fig. 3a. The direct damping, Btt decreases almost
monotonically, Fig. 3b. One should note that the cross-coupled
damping at whirling speeds beyond Or = 0.75 is larger than the one
from the p-lm simulation, Fig. 3a. The direct damping at low gas
concentration is larger than the one obtained from the p-lm
calculation and decreases with the increase of the gas void
fraction, Fig. 3b. This is due to the fact that as the void fraction
increases, the pressure build-up is hampered, and thus the
resulting forces responsible for the Btt magnitude are getting
smaller.
Fig. 4 presents the comparison of dimensionless damping
coefcients for the p-lm assumption as they are calculated by
the NSE and Reynolds equations for the LBA case. The EJ whirling
speed is 10,000 rpm. The results presented in this gure are based
on Eqs. (13) and (14), under the assumption that the Re* is
negligible and the inertia terms effect can be neglected. One can
see that only a small relative difference, o2%, exists between the
two groups of results (NSE and Reynolds), indicating that the
Reynolds equation represents a reasonable and quicker alternative
to the NSE, as long as the inertia effects are not signicant.

0.05
0.00
0.0

0.5

1.0
1.5
2.0
2.5
Dimensionless whirling speed (r)

3.0

3.5

Fig. 3. Dimensionless direct- and cross-coupled damping coefcient variation


with the whirling speed for the Gumbel and homogeneous cavitation models:
(a) Brt and (b) Btt.

8. Inertia (added mass) coefcients


Figs. 5 and 6 present, for the Gumbel condition, the variation of
the inertia coefcients Arr and Atr, with respect to the Re*, and thus
implicitly with the whirling speed (the whirling speed was
converted to an equivalent Re*). Three different eccentricity

ARTICLE IN PRESS
C. Xing et al. / Tribology International 43 (2010) 654666

1.04

-0.31

= 0.2

NSE agree with El-Shafei

= 0.2

0.98
10.9%

0.96
16.3%, NSE results as the reference

0.94
0.92
N-S equation
Vance model
El-Shafei model

0.90
0.88

Inertia force coefficient (Atr)

Inertia force coefficient (Arr)

1.02
1.00

659

-0.32

11.7%

-0.33

-0.34

N-S equation
Vance & El-Shafei model

0.86
-0.35

0.84
0

See Fig. 13

10
20
Modified Reynolds' number (Re*)

-0.86
= 0.5

0.90

N-S equation
Vance model
El-Shafei model

Agreement

0.85

15.3%
17.9%

0.80

0.75

See Fig. 13 10

20
Modified Reynolds' number (Re*)

= 0.5

N-S equation
Vance & El-Shafei model

-0.90
-0.92
16%

-0.94
-0.96
-0.98
-1.00
-1.02

0.70
0

10
20
See Fig. 13 Modified Reynolds' number (Re*)

30

See Fig. 13 10

20
30
Modified Reynolds' number (Re*)

-1.9
0.35
0.30

57.8%

0.25
82.7%

0.20

N-S equation
Vance & El-Shafei model

= 0.8

NSE agree with Vance

N-S equation
Vance model
El-Shafei model

0.15

72.9%

0.10
0.05

Inertia force coefficient (Atr)

0.40

Inertia force coefficient (Arr)

30

-0.88
Inertia force coefficient (Atr)

0.95

Inertia force coefficient (Arr)

30

-2.0
= 0.8

-2.1
23.6%

-2.2
9.2%

-2.3

-2.4
0

0.00
0

10
20
See Fig. 13
Modified Reynolds' number (Re*)

30

Fig. 5. Variation of inertia coefcients Arr for SBA case with the change of Re*
under the Gumbel cavitation condition at several eccentricities: (a) 0.2, (b) 0.5, and
(c) 0.8.

cases are considered: e = 0.2, 0.5 and 0.8. For comparison, the
results obtained by Vance [18] and El-Shafei [23] are presented as
well. Based on the modied Reynolds equation, the inertia
coefcients presented by these two authors, while different

10
20
30
See Fig. 13
Modified Reynolds' number (Re*)

Fig. 6. Variation of inertia coefcients Atr for SBA case with the change of Re*
under the Gumbel cavitation condition at several eccentricities: (a) 0.2, (b) 0.5, and
(c) 0.8.

from each other in magnitude, are both independent of the


whirling speed (Assumption made by Vance is that Re b 1).
However, according to the full NSE model presented in this paper
it was found that for 0.3 oRe* o35, these coefcients do show a

ARTICLE IN PRESS
660

C. Xing et al. / Tribology International 43 (2010) 654666

1.0
Inertia force coefficient (Arr, Atr)

Inertia force coefficient (Arr, Atr)

2
Arr

0.5
Arr, Variation with whirling,
Arr, Variation with dimensionless density,
Arr, Variation with viscosity,

0.0

Atr, Variation with whirling,


Atr, Variation with dimensionless density,
Atr, Variation with viscosity,

-0.5

Atr

-1.0
10
20
Modified Reynolds number (Re*)

30

Fig. 7. Variation of inertia coefcients Arr and Atr for SBA case under the Gumbel
cavitation condition when o, r and m are changed in parametric fashion, one at a
time at e = 0.5.

3.6
3.4
Inertia force coefficient (Arr)

Arr, Variation with dimensionless density,

Arr

Arr, Variation with viscosity,

Atr, Variation with whirling,


Atr, Variation with dimensionless density,
Atr, Variation with viscosity,

0
Atr

-1

-2
0

3.2
3.0

Eccentricity 0.2
Eccentricity 0.5
Eccentricity 0.8

2.8

Arr increases
with increase

2.6
2.4
2.2
2.0
1.8
0

20
See Fig. 13 10
Modified Reynolds' number (Re*)

30

Inertia force coefficient (Atr)

Arr, Variation with whirling,

-1

-2

Atr decreases
with increase

-3

Eccentricity 0.2
Eccentricity 0.5
Eccentricity 0.8

-4

-5

See Fig. 13 10

20

30

Modified Reynolds' number (Re*)


Fig. 8. Variation of inertia coefcients Arr and Atr for SBA case with the change of
Re* for the 2p-lm at several eccentricities (0.2, 0.5, and 0.8): (a) Arr and (b) Atr.

10
20
Modified Reynolds number (Re*)

30

Fig. 9. Variation of inertia coefcients Arr and Atr for SBA case under 2p-lm when
o, r and m are changed in a parametric fashion, one at a time at e = 0.5.

dependence with the whirling speed. In Fig. 5, for all


eccentricities, the direct inertia coefcient Arr increases, as the
whirling speed increases, reaches a maximum value, and then
decreases even though the whirling speed continues to increase.
The cross-coupled inertia coefcient Atr, Fig. 6, follows a different
trend. They show a moderate increase for the low Re* numbers
(low whirling speed), and generally plateau as the whirling speed
increases further. There is an exception at e = 0.2, where Atr
decreases with a further increase of the whirling speed. In both
Figs. 5 and 6, at all eccentricities, a strong nonlinear behavior can
be ascertained when compared with the results of El-Shafei and
Vance. This can be assigned to the fact that while the 3-D NSE are
fully populated and account for the inertia effects, the modied
Reynolds equation is integrated across the clearance, thus
becoming two-dimensional and neglecting some viscous and
third direction inertia terms.
For Arr, at low eccentricity (0.2) and Re* between 0.3 and 10,
the NSE model agrees well with El-Shafeis in the neighborhoods
of the intersection between the two curves. The maximum
deviation between the two models is 10.9%. The NSE results vary
nonlinearly and deviate more (17.9%) from the constant Arr values
when e = 0.5. When e = 0.8 the deviation is even larger, at about
72.9%. The NSE model deviates from the Vance model results by as
much as 16.3% for e = 0.2, 15.3% for e = 0.5 and 57.8% for e = 0.8,
respectively. For Atr, the NSE model deviates by a minimum of 9.2%
and a maximum of 23.6% from the two models of Vance and
El-Shafei whose results are still linear and independent of the
whirling speed.
From the NSE based results, it is reasonable to assume that the
nonlinearities of Aij coefcients are dependent of the modied
Reynolds number, Re*. While the whirling speed was used as a
parameter in Fig. 5 and 6, in Fig. 7 density (r) and viscosity (m) are
used as parameters. The inertia coefcients, Aij, are presented as a
function of the Re*, when r and m are changed one at a time.
When density is used as the parameter, to better expose their
nonlinearities, the Aij terms were normalized with respect to a
reference density, r r=rref . In doing so, one can see that
whether o, r, or m were varied, the Aij values fall all on their
same respective curves. This further conrms the ndings of
Figs. 5 and 6, in that the inertia coefcients are nonlinear in
character and variable in nature.
Fig. 8 presents the inertia force coefcients Arr and Atr obtained
at different whirling speeds under the 2p-lm condition

ARTICLE IN PRESS
C. Xing et al. / Tribology International 43 (2010) 654666

100

661

200

Eccentricity 0.2
Eccentricity 0.5
Eccentricity 0.8
Variation with dimensionless density
Variation with viscosity

60

40
Trend of Arr as going up

20

Inertia force coefficient (Arr)

Inertia force coefficient (Arr)

180
80

160
140
120
100

Eccentricity 0.2
Eccentricity 0.5
Eccentricity 0.8
Variation with dimensionless density
Variation with viscosity

80
60
40
20

0
20
See Fig. 12 10
Modified Reynolds' number (Re*)

30

-1

See Fig. 12

10
20
Modified Reynolds' number (Re*)

30

See Fig. 12

0
-2

Trend of Atr as going up

-4
-6
Eccentricity 0.2
Eccentricity 0.5
Eccentricity 0.8
Variation with dimensionless density
Variation with viscosity

-8
-10
-12
0

See Fig. 12 10

20

30

Modified Reynolds' number (Re*)


Fig. 10. Variation of inertia coefcients Arr and Atr for LBA case with the change of
Re* under the Gumbel cavitation condition: (a) Arr and (b) Atr.

(full Sommerfeld condition). Like in Figs. 6 and 7, the Re* is varied


in terms of the whirling speed o. The NSE based calculations
found that for the 2p-lm the Arr values are much larger than
those obtained for the p-lm. Thus, the direct inertia coefcients
Arr, Fig. 8a, have only positive values and, while showing a weak
dependence of the Re* number, increase strongly as the
eccentricity increases. The cross-coupled inertia coefcient Atr,
Fig. 8b, is different from zero and decreasing with eccentricity; its
magnitude, is larger by approximately a factor of 2 when
compared to the p-lm case (Fig. 6).
For the 2p-lm cases, Fig. 9 shows the variation of direct and
cross-coupled inertia coefcients with the Re* number. It can be
seen that based on density, dynamic viscosity and whirling speed
changes, the inertia coefcients follow the same variation trend
and fall on a unique curve. As Re* varies from 0.3 to 35, the Arr
decreases by approximately 5%, while Atr increases by almost
13.8%. The trends observed here re-conrm the trends and
observations discussed in Fig. 7.
Based on the SBA case results, it is reasonable to extrapolate
that the inertia coefcients for the LBA case will exhibit a
dependency on the Re* number as well. For the p-lm assumption
and for eccentricities of 0.2, 0.5 and 0.8. Fig. 10 shows the inertia
coefcients behavior with respect to changes in the whirling

Inertia force coefficient (Atr)

Inertia force coefficient (Atr)

-2
-3
-4
-5
Eccentricity 0.2
Eccentricity 0.5
Eccentricity 0.8
Variation with dimensionless density
Variation with viscosity

-6
-7
-8
0

10
20
Modified Reynolds' number (Re*)

30

Fig. 11. Variation of inertia coefcients Arr and Atr for LBA case with the change of
Re* under the 2p-lm condition: (a) Arr and (b) Atr.

speed, density and viscosity. Generally, as the eccentricity is going


up, both the direct and the cross-coupled inertia coefcients
decrease in magnitude. One can further note that nonlinear
characteristics are present in the behavior of these coefcients as
well. Thus, for all eccentricities presented, the Arr coefcients
decrease monotonically with the increase of Re*, while the Atr
coefcients have a more mixed behavior. At e = 0.2, Atr decreases
monotonically, while for e =0.8 it increases asymptotically. For
e = 0.5, the Atr presents an inection point around Re* = 10. It should
be mentioned that for the p-lm LBA case the results for the
inertia coefcients yielded by the NSE model deviate considerably
from those obtained by the El-Shafei [24] model.
Fig. 11 further presents the inertia coefcients variation with
Re* for the 2p-lm case. For the Arr coefcients, Fig. 11a, the trends
are similar to those observed in the p-lm case (Fig. 10a). For all
eccentricities presented, there is roughly a factor of 2 increase in
magnitude in favor of the 2p-lm case. The nonlinear character of
Arr is weak when compared to that of the Atr, Fig. 11b.
Fig. 12 presents the variation of the inertia coefcients with
respect to eccentricity for the LBA case when the damper EJ is
whirling at 20,000 rpm (Re* = 6.85). The results for both the p- and
2p-lm cases are superposed for comparison. The direct inertia
coefcients Arr for both the p- and 2p-lm cases show a

ARTICLE IN PRESS
662

C. Xing et al. / Tribology International 43 (2010) 654666

6
2 and case i

Inertia force coefficient (Arr)

Inertia force coefficient (Arr)

200

150
case iii

case ii

100

50

Xing et al., -film


Xing et al., 2-film
San Andres, case i
San Andres, case ii
San Andres, case iii

0
0.0

0.2

0.4

0.6

-film, NSE
-film, Vance model
-film, El-Shafei model
2-film, NSE

Better comparison with El-Shafei

-2
Better comparison with Vance

0.8

-4

1.0

0.0

0.2

Eccentricity ()

Inertia force coefficient (Atr)

Inertia force coefficient (Atr)

-2
-4
-6
-8

0.0

-film
2-film

0.2

0.4

0.6

0.8

1.0

?? film

-10

0.8

San Andres 2-film

0.4
0.6
Eccentricity ()

1.0

Eccentricity ()

-2

-4
2film

-6
-film, NSE
-film, Vance & El-Shafei model
2-film, NSE

-8
0.0

0.2

0.4
0.6
Eccentricity ()

0.8

1.0

Fig. 12. Variation of inertia coefcients Arr and Atr for LBA case with respect to
eccentricity, when Re*=6.85 under the Gumbel cavitation and 2p-lm conditions:
(a) Arr and (b) Atr.

Fig. 13. Variation of Inertia coefcients Arr and Atr for SBA case with respect to
eccentricity when Re* = 6.85 under the Gumbel cavitation and 2p-lm conditions:
(a) Arr and (b) Atr.

monotonically decreasing trend with increasing eccentricity. San


Andres [25] proposed for the 2p-lm case piecewise direct inertia
coefcient models for Arr: (i) Re 51; (ii) Re* at moderate number
and (iii) Re b1. The results presented in this paper are very close
to his case (i). A re-inspection of Fig. 11 shows that the NSE based
Arr for the 2p-lm varies o2% for all Re*, so their values that
compare well with San Andres case (i) are valid only for Re 5 1.
In Fig. 12b, the cross-coupled coefcients Atr, change with the
increase in eccentricity, shows a strong nonlinear character and
are out of phase with each other as shown by the different
locations of their peak values. The 2p-lm case shows a less
nonlinear character than the p-lm case, but it hovers in negative
territory and never crosses the Atr = 0 line. Thus for Atr, the NSE
based results present a nonlinear character and do not match the
value of Atr = 0 as proposed by San Andres.
Figs. 13 presents comparison plots of the inertia force
coefcients for SBA case as functions of eccentricity under the
p- and 2p-lm conditions when the EJ whirls at 20,000 rpm
(Re* = 6.85). This represents just one point, Re* =6.85, on the
graphs of Figs. 5 and 6. The values for both Arr and Atr obtained
through the NSE simulation, compare well with those of El-Shafei
and Vance models by following the same trend. When e o0.5,

both the NSE and El-Shafei models render very close Arr. When
e 40.5, the NSE model yields Arr values that compare well with
those of Vance model. More generally, the Atr coefcients obtained
with the NSE model follow those of both El-Shafei and Vance
models very closely at low eccentricity but deviate as much as 25%
when eccentricity reaches 0.95.
The NSE based inertia coefcients calculated for the 2p-lm
with SBA are plotted also in Fig. 13. Thus they can be directly
compared to the NSE based results for the p-lm. For the 2p-lm
case the Arr coefcients show an opposite trend to that of the plm case. For the cross-coupled coefcient Atr, the values decrease
with the increase in eccentricity for both 2p- and p-lm cases. For
the Arr in the 2p-lm, San Andres [25] proposed a piecewise
model for three Re* regimes as well: (i) Re 51, (ii) Re* at
moderate number and (iii) Re b 1. He found that for the 2p-lm,
the Arr is different by a factor of 2 when compared with the
corresponding Arr values for the p-lm, and that Atr =0.
Figs. 12 and 13 presented information relating to the variation
of the inertia coefcients with respect to eccentricity. These
coefcient values are based on the magnitude of the forces
obtained from the integration of the pressure around the
circumference. Since the pressure differences between NSE and

ARTICLE IN PRESS
C. Xing et al. / Tribology International 43 (2010) 654666

1.2e+5

o
+

Pressure (Pa)

1.0e+5

Withou inertia
With inertia

= 0.2

8.0e+4
6.0e+4
Minimum clearance

4.0e+4
2.0e+4
0.0
0

2
3
4
5
Circumferential direction (rad)

6 2

6e+5

o
+

Withou inertia
With inertia

5e+5

Pressure (Pa)

= 0.5

4e+5
3e+5

Minimum clearance

2e+5
1e+5

663

The gure presents, for three different eccentricities (e = 0.2, 0.5


and 0.8), the pressure built-up in the damper, when the Gumbel
lm rupture conditions (p-lm) are considered. The pressures are
calculated under two assumptions: (i) no-inertia contribution (the
Reynolds equation) and (ii) with inertia (the NSE). For case (i)
Re 5 1, the pressures are balanced only by viscous terms and they
represent the limit towards which the NSE based cases will trend.
For case (ii), the inertia terms uj @ui =@xj are considered. Under the
same speed, when eccentricities are small, the amount of mass
carried through the clearance is large, thus the inertia effects are
large when compared to the viscous effects. As the eccentricity
increases, the amount of mass carried through the clearance
becomes much smaller and the inertia effects get proportionally
smaller as well; the viscous effects start to dominate. This
interpretation is appropriate for the convergent region under
Gumbel p-lm condition. Thus for a small eccentricity (0.2),
Fig. 14a, the NSE based pressures are larger than the Reynolds
equation based ones. As the eccentricity increase to 0.5
(Fig. 14b) and 0.8 (Fig. 14c), respectively, the NSE based results
tend towards the limiting cases presented by the Reynolds
equation. The net radial forces Fr are a summation of the
components due to the viscous and inertia effects. At low
eccentricity, the inertia terms uj @ui =@xj dominate causing the
inertia induced radial component to be positive. As eccentricity
increases, this component decreases and eventually becomes
negative, subtracting from the radial force component caused
by the viscous terms m @2 ui =@x2i . As a result, the net radial force
becomes smaller. That is the reason why the direct
inertia coefcient Arr changed from positive to negative values
(Fig. 13a). The variation of the magnitude of tangential force, Ft,
when considering inertia, is always larger than that when
only viscous effects are considered, therefore, the cross-coupled
inertia coefcient Atr acts as an additional damping force
coefcient.

0
0

2
3
4
5
Circumferential direction (rad)

6 2
9. Conclusion

o
+

6e+6

= 0.8

5e+6
Pressure (Pa)

Withou inertia
With inertia

4e+6
3e+6

Minimum clearance

2e+6

+
1e+6
0
0

2
3
4
5
Circumferential direction (rad)

6 2

Fig. 14. Pressure distribution for the SBA case under the Gumbel cavitation
condition, with and without inertia terms, along the circumferential direction in
the axial plane of symmetry. (eccentricities are e = 0.2, 0.5 and 0.8 and o = 20,000
rpm): (a) 0.2, (b) 0.5, and (c) 0.8.

Reynolds equation based results are strongly dependent on the


changes in eccentricity, the inertia coefcients themselves
become strongly dependent on the eccentricity. Within this
context, Fig. 14, helps explaining the results of Figs. 12 and 13.

The dynamic coefcients of a squeeze lm damper for the SBA


and LBA cases are presented in this paper. The development
and understanding of these coefcients are necessary in order to
perform a damper stability analysis. It has been shown [19]
that the inertia effect of the lubricant, as well as the type and
extent of the cavitation inuence signicantly the damper
performance.
Due to cavitation (p-lm) the magnitude of the direct
damping, Btt, reduces to as much as half of the value for the 2plm. The cross-coupled damping, Brt, takes values different from
zero, thus causing multiple-valued transient and steady-state
responses for the damper. For the homogeneous cavitation model,
Fig. 3 presents damping coefcients that vary with the increase of
the whirling speed; the direct damping coefcient decreases with
the increase of gas mass fraction, while the cross-coupled
damping coefcient increases at the same time.
The inertia coefcients obtained from the full 3-D NSE have
both similarities and differences with the results derived by
previous researchers. By varying the density, viscosity and
whirling speed of the damper, it was found that for both LBA
and SBA cases, the nonlinear characteristics of the inertia
coefcients are generally a function of the modied Reynolds
number Re*, even though at times they are only a weak function
thereof. As illustrated in Fig. 14, the variation of the inertia
coefcient can be explained by the changes in the pressure
distribution and resulting forces depending on whether the inertia
or viscous terms are dominant.

ARTICLE IN PRESS
664

C. Xing et al. / Tribology International 43 (2010) 654666

Appendix. Flowchart for damping and inertia coefcients computations


Appendix A. Damping coefcient (kg/s) evaluation without inertia (CCO motion)

ARTICLE IN PRESS
C. Xing et al. / Tribology International 43 (2010) 654666

665

Appendix B. Inertia coefcient (kg) evaluation with CCO motion

References
[1] Gunter EJ, Barrett LE, Allaire PE. Design of nonlinear squeeze-lm dampers for
aircraft engines. J Lubric Technol 1977;99(1):5764.
[2] Hahn EJ. Stability and unbalance response of centrally preloaded rotors mounted
in journal and squeeze lm bearings. J Lubric Technol 1979;101(2):1208.
[3] Pietra LD, Adiletta G. The squeeze lm damper over four decades of
investigations. Part I: characteristics and operating features. Shock Vib Dig
2002;34:326.
[4] Gunter EJ, Barrett LE, Allaire PE. Design and application of squeeze lm
dampers for turbomachinery stabilization. In: Proceedings of the fourth
turbomachinery symposium, 1975, p. 12741.

[5] Warner PC. Static and dynamic properties of partial journal bearings. J Basic
Eng 1963;85(2):24757.
[6] Tichy JA. The effect of uid inertia in squeeze lm damper bearings: a
heuristic and physical description. ASME Pap., 1983, 83-GT-177.
[7] San Andres LA, Vance JM. Effect of uid inertia on squeeze-lm damper forces
for small-amplitude circular-centered motions. ASLE Trans 1986;30(1):
638.
[8] Jung SY, San Andres LA, Vance JM. Measurements of pressure distributions
and force coefcients in a squeeze lm damper part I: fully open ended
conguration. Tribol Trans 1991;34(3):37582.
[9] Szeri AZ, Raimondi AA, Giron-Duarte A. Linear force coefcients for squeezelm dampers. ASME Pap., 1982, 82-Lub-30.

ARTICLE IN PRESS
666

C. Xing et al. / Tribology International 43 (2010) 654666

[10] El-Shafei A, Crandall SH. Fluid inertia forces in squeeze lm dampers. Rotating
machinery and vehicle dynamics, DE-35. 1991, p. 21928.
[11] Lund JW, Smalley AJ, Tecza JA, Walton JF. Squeeze-lm damper technology:
part 1prediction of nite length damper performance. ASME Pap., 1983, 83GT-247.
[12] Lund JW, Smalley AJ, Tecza JA, Walton JF. Squeeze-lm damper technology:
part 2experimental verication using a controlled-orbit test rig. ASME Pap.,
1983, 83-GT-248.
[13] Zhang J, Ellis J, Roberts JB. Observations on the nonlinear uid force in short
cylindrical squeeze lm dampers. J Tribol 1993;115(4):6928.
[14] Guo Z, Hirano T, Kirk RG. Application of CFD analysis for rotating machinery,
part 1: hydrodynamic, hydrostatic bearings and squeeze lm damper. In:
Proceedings of ASME Turbo expo, 2003, p. 6519.
[15] Chen PYP, Hahn EJ. Use of computational uid dynamics in hydrodynamic
lubrication. Proc Inst Mech Eng Part J 1998;212(6):42736.
[16] Taylor DL, Kumar BRK. Nonlinear response of short squeeze lm dampers. J
Lubric Technol 1980;102(1):518.
[17] ESI Group. CFD-ACE +V2006 modules manual. ESI CFD Inc.; 2006.
[18] Vance JM. Rotordynamics of turbomachinery. New York: Wiley; 1988.

[19] Xing C. Analysis of the characteristics of a squeeze lm damper by threedimensional NavierStokes equations. Doctoral thesis, The University of
Akron; 2009.
[20] Diaz SE, San Andres LA. A model for squeeze lm dampers operating with air
entrainment and validation with experiments. J Tribol 2001;123(1):
12533.
[21] Braun MJ, Henricks RC. An experimental investigation of the vaporous/
gaseous cavity characteristics of an eccentric journal bearing. ASLE Trans
1984;27(1):114.
[22] Dowson D, Godet M, Taylor CM. Cavitation and related phenomena in
lubrication. In: First Leeds Lyon symposium on tribology. University of Leeds,
Leeds, England; 1974.
[23] El-Shafei A. Unbalance response of a jeffcott rotor incorporating short
squeeze lm dampers. J Eng Gas Turb Power 1990;112(4):44553.
[24] El-Shafei A. Unbalance response of a jeffcott rotor incorporating long squeeze
lm dampers. ASME Pap., 1989, p. 14958.
[25] San Andres LA, Vance JM. Effects of uid inertia and turbulence on the force
coefcients for squeeze lm dampers. J Eng Gas Turb Power 1986;108(2):
332339.

S-ar putea să vă placă și