Sunteți pe pagina 1din 9

See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/236951584

Ultrasonic Measurement of Rolling Bearing


Lubrication Using Piezoelectric Thin Films
ARTICLE in JOURNAL OF TRIBOLOGY JANUARY 2009
Impact Factor: 0.9 DOI: 10.1115/1.3002324

CITATIONS

DOWNLOADS

VIEWS

183

163

5 AUTHORS, INCLUDING:
Jie Zhang

Katherine Kirk

University of Bristol

University of the West of Scotland

48 PUBLICATIONS 195 CITATIONS

110 PUBLICATIONS 1,113 CITATIONS

SEE PROFILE

SEE PROFILE

Rob Dwyer-Joyce
The University of Sheffield
154 PUBLICATIONS 1,223 CITATIONS
SEE PROFILE

Available from: Jie Zhang


Retrieved on: 15 September 2015

Bruce W. Drinkwater1
e-mail: b.drinkwater@bristol.ac.uk

Jie Zhang
Department of Mechanical Engineering,
University of Bristol,
University Walk,
Bristol, BS8 1TR, UK

Katherine J. Kirk
Jocelyn Elgoyhen
School of Engineering and Science,
University of Paisley,
Paisley, PA1 2BE, UK

Rob S. Dwyer-Joyce
Department of Mechanical Engineering,
University of Sheffield,
Mappin Street,
Sheffield, S1 3JD, UK

Ultrasonic Measurement of
Rolling Bearing Lubrication Using
Piezoelectric Thin Films
This paper describes the measurement of lubricant-film thickness in a rolling element
bearing using a piezoelectric thin film transducer to excite and receive ultrasonic signals.
High frequency (200 MHz) ultrasound is generated using a piezoelectric aluminum nitride film deposited in the form of a very thin layer onto the outer bearing raceway. This
creates a transducer and electrode combination of total thickness of less than 10 m. In
this way the bearing is instrumented with minimal disruption to the housing geometry
and the oil-film can be measured noninvasively. The high frequency transducer generates
a fine columnar beam of ultrasound that has dimensions less than the typical lubricated
contact ellipse. The reflection coefficient from the lubricant-layer is then measured from
within the lubricated contact and the oil-film thickness extracted via a quasistatic spring
model. The results are described on a deep groove 6016 ball bearing supporting an 80
mm shaft under normal operating conditions. Good agreement is shown over a range of
loads and speeds with lubricant-film thickness extracted from elastohydrodynamic lubrication theory. DOI: 10.1115/1.3002324
Keywords: piezoelectric thin films, rolling bearing, oil-film thickness measurement, reflection coefficient, condition monitoring

Introduction

Rolling element bearings rely on the formation of a separating


oil-film for their smooth operation and long life. The lubricated
contact shown schematically in Fig. 1 is typically a few millimeters squared in area and has a thickness of under a micron.
Furthermore, the contact moves at high speed as the bearing rotates under normal operation. Failure of this lubricant-film is one
of the most common reasons for bearing failure, which usually
has serious consequences for machine operation 1. While there
are many models of elastohydrodynamic lubrication EHL, measurement of films of this nature is still a challenging task.
Researchers have measured EHL using optical interferometric
methods, which rely on a transparent window 2 and the insertion
of optical fibers 3. Capacitive techniques have been used in rolling bearings 4 but require electrical isolation or a surface
mounted electrode. These approaches have proved useful in a
laboratory setting but, due to their invasive nature causing significant disruption to the bearing, are not suitable for in situ condition
monitoring of real industrial bearings. Currently, in situ condition
monitoring of bearings and rotating machinery uses either temperature, vibration, or acoustic emission 58, all of which give
warning of an already partially failed bearing. These approaches
rely on the fact that when the lubricant fails, rubbing occurs resulting in heating and vibration over a range of frequencies. Recently Anderson et al. 9 used the transmission and reflection of
shear or transverse ultrasonic waves to monitor the collapse of
oil-layers in thin fluid shaft seals. The principle is that the presence of the lubricant causes significant reflection of the shear
waves and so, on collapse, an increase in the transmitted shear
energy is seen. This approach has been shown to be entirely noninvasive and applicable to many industrial bearings.
The reflection coefficient of longitudinal ultrasonic waves has
1
Corresponding author.
Contributed by the Tribology Division of ASME for publication in the JOURNAL OF
TRIBOLOGY. Manuscript received December 14, 2007; final manuscript received
August 19, 2008; published online December 3, 2008. Assoc. Editor: Shuangbiao
Jordan Liu.

Journal of Tribology

been shown to be highly sensitive to lubricant-film thickness 10


and the authors have applied this approach successfully to a range
of bearings 1113 including journal and ball bearings. Most significantly, oil films in thickness region 0.3 1.0 m were measured in rolling element bearings under normal operating conditions. However, these previous measurements used a bulky high
frequency piezoceramic transducer, operating at 50 MHz, that was
focused on the small lubricated contact using an acoustic lens.
Although this approach does not directly disrupt the outerraceway of the bearing, it necessitated significant disruption to the
bearing housing to allow access. This necessitated a 20 mm hole
to be drilled in the bearing housing to locate the transducer, acoustic lens, and cabling. While this is satisfactory for laboratory measurements, the lack of robustness precludes field testing.
The present paper describes the use of a piezoelectric thin film
ultrasonic transducer that operates at very high frequencies 50
400 MHz and has a thickness of a few microns. In this way, the
sensor can be attached to the outer-raceway of the bearing with
minimal disruption to the bearing and housing. Furthermore, the
high frequency of operation removes the need for focusing and
hence the acoustic lens. Instead a fine a columnar ultrasonic beam
is used with beam dimensions less than that of the lubricated
contact. This paper describes this device and demonstrates its application to measurement in a commercial deep groove ball bearing type 6016 under normal operation.

Background Theory

2.1 Spring Model of an Ultrasonic Reflection From a


Lubricant-Film. For the purposes of ultrasonic response modeling, the structure of a lubricated rolling element bearing can be
represented as a multilayered system consisting of outer-racewaylubricant-ball. When an ultrasonic pulse propagates from the
outer-raceway into this structure, the first reflection will come
from the lubricant-layer. If the thickness of the lubricant-layer is
small in comparison with the ultrasonic wavelength, it can be
shown that the reflection of ultrasound is governed only by its
stiffness see for example Ref. 10. The normal stiffness of a
planar fluid layer KN is given by

Copyright 2009 by ASME

JANUARY 2009, Vol. 131 / 011502-1

Downloaded 09 Jan 2009 to 137.222.187.115. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

P=
2b

Raceway
z
Rbx

Rby
Rax

Ray
x
Ball

Fig. 1 Configuration of ball bearing outer-raceway and ball in


contact

KN =

B
h

where h is the lubricant-film thickness, and B is the bulk modulus


of the fluid. The bulk modulus can be expressed in terms of the
density and longitudinal wave speed c of the fluid, such that
B = c2. Assuming that the media on either side of the layer have
identical acoustic properties both are steel in the rolling bearing
case considered in this paper and that the wave is normally incident the lubricant-film thickness can be extracted from the well
known quasistatic spring model as
B
h=
fz

Rf2
1 Rf2

2.2 Ball Bearing Lubrication. Figure 1 shows the configuration of a ball bearing outer-raceway and a ball in contact. As
shown, the ball moves along the groove in the x-direction, the ball
and the groove are conformal in the y-direction, and the load is
applied in the z-direction. Subscripts a and b refer to the two
rolling elements i.e., the ball and the raceway, respectively. In
this work, the regression equations of Dowson and Higginson 14
are used to predict the central film thickness hc in the ball-outerraceway contact.

0.67

E0.53


P
E R 2

0.067

1 0.61e0.73k
3

where U is the mean surface speed, 0 is the lubricant viscosity at


the contact entry, is the pressure-viscosity coefficient, P is the
load on the contact, k is the ellipticity ratio, E is the reduced
elastic modulus, and R is the reduced radius of curvature given
by

2a

2b

1
1 1 1
=
+
E 2 Ea
Eb

1
1
1
1
1
=
+
+
+
4
R Rax Rbx Ray Rby

where E is Youngs modulus, and is Poissons ratio. The lubricated contact area is elliptical in shape with the minor a and
major b semicontact radii given by
a=

6PR
kE

1/3

b=

6k PR
E
2

1/3

where is the complete elliptical integral of the second kind. An


approximate relationship 15 for the maximum load on a single
ball in a radially loaded ball bearing is used.
011502-2 / Vol. 131, JANUARY 2009

2.3 Lubricant Bulk Modulus. The bulk modulus used to determine the oil-film thickness from the layer stiffness Eq. 1
must be that of the oil under contact pressure. In EHL pressures,
the oil compresses significantly and the local bulk modulus is
several times higher than under ambient pressures. In this work,
the compressibility model of Jacobson and Vinet 16 is used to
determine the influence of pressure on the lubricant bulk modulus.
They gave an equation of state to describe the behavior of the
lubricant under pressure p.
p=

3B0
1 xe1x
x2

and the bulk modulus under pressure is given by


B=

B0
2 + 1x x2e1x
x2

where B0 is the bulk modulus at ambient pressure, is a lubricant


specific parameter, and x is a function of the relative compression.

where z is the acoustic impedance of the media surrounding the


lubricant-film, Rf is the amplitude of the measured reflection
coefficient, which is a function of the ultrasonic frequency f.

hc
U0
= 2.69
R
E R

where W is the load on the whole bearing, and nb is the number of


balls.
The inlet lubricant viscosity 0 was assumed to remain constant
during these tests i.e., unaffected by the changes in bearing load
and speed. The tests were performed relatively quickly data recorded within a 5 min interval so the bearing remained at room
temperature throughout. Since there was no instrumentation to
record the temperature of the oil and hence its viscosity at the
contact entry, it was deemed preferable to use a constant value
equal to that of the oil supply temperature.

2a

5W
nb

x=

o
p

where 0 is the density at ambient pressure, and p at pressure p.


These equations are used for the oil both in its liquid state and
after it undergoes solidification by employing different values of
and B0. Values for these parameters were estimated for the test oil
used in these experiments Shell Turbo T68 from generically
similar oils tested by Jacobson and Vinet using a high pressure
chamber. The bulk modulus for T68 at ambient pressure was determined from the speed of sound and density B = c2 of the oil
and found to be B = 1.84 GPa. Using Eqs. 79, the bulk modulus at a pressure of 1.5 GPa was found to increase to 21.2 GPa.
2.4 Deformed Surface Geometry. Given the dimensions of
the lubricated contact a, b, and hc, it is now possible to develop
equations for the deformed geometries of the ball and outerraceway, both inside and outside the lubricated contact. The starting point is the geometry of the undeformed ball surface and
outer-raceway. Taking the initial point of contact between the ball
and the raceway as the origin, the geometry of the ball and raceway surfaces under no load is given by
za = R2ax x2 y 2 Rax

10

2
y 2 Rbx Rby
zb = R2bx x2 + Rby

11

The total elastic deflection at the central point of the contact i.e.,
at x , y = 0 is given by 17.

c =


4.5
nR

P
E

1/3

12

where is the complete elliptic integral of the first kind. The


gap between the ball and raceway filled with lubricant in the
contact region and either lubricant, air, or a mixture of the two
outside this region is given by
Transactions of the ASME

Downloaded 09 Jan 2009 to 137.222.187.115. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Protective shroud

Electrode(0.3x3mm)

Bearing shell

AlN Piezo - film

Electrical connection

Ball
Cage
Housing
80 mm

Fig. 2 Schematic of the piezo thin film transducer on a ball


bearing

h = hc + hg + e c

13

where hg is the difference in the undeformed shapes of the ball


and raceway, i.e., hgx , y = zbx , y zax , y and e is the total
elastic deflection of the surfaces, i.e., ex , y = ax , y + bx , y,
where a and b are the elastic deflections of the ball and raceway,
respectively. Because the ball and outer-raceway are typically
made from the same material, their elastic deflections are equal
i.e., a = b. Assuming the oil-film thickness is the constant in the
lubricated contact region, from Eq. 13 the elastic deflection of
outer-raceway in the contact region is given by

a = b =

e
2

1
= c h g
2

where
re =

for

re 1

14

x2 y 2
+
a2 b2

A Hertzian contact model then allows the elastic deflection of the


ball and outer-raceway outside the contact region to be written as
18

a = b =

1
c
2 r2e sin1
+ r2e 1

re

for re 1 15

Finally, the deformed geometry of the ball and outer-raceway can


be written as

= za,b a,b
za,b

16

where a prime is now used to denote the deformed geometry.

Experimental Apparatus and Procedure

A piezoelectric thin film transducer was manufactured on the


outer surface of the outer-raceway of a deep groove 6016 ball
bearing. Figures 2 and 3 show a schematic and photograph of the
instrumented bearing. Aluminum nitride AlN was deposited by
radio frequency RF magnetron sputtering to form the piezoelectric film. The aluminum target was presputtered under a vacuum
to remove the surface oxide. Nitrogen gas was then fed into the
chamber and a film of AlN deposited on the bearing outerraceway. A coating thickness of approximately 4 6 m was
achieved after several hours of sputtering. The substrate needed
no preheating so there was no danger of the bearing steel tempering. Scanning electron microscope SEM examinations of the
coating demonstrate that the film has a highly columnar structure.
This gives them a strong piezoelectric property.
An electrode was then deposited by evaporating aluminum
through a hole in a mask that measured 3 0.3 mm2 producing
Journal of Tribology

Fig. 3 Photograph of the instrumented ball bearing

an electroded region of the same size. An electrical connection


was then made between the electrode and a fine wire using conductive epoxy. The bearing raceway was grounded to form the
other electrical connection. Finally, a protective shroud made from
an electrically insulating material FR-4 PCB material was
bonded to the outer-raceway to protect the electrode and the electrical connection during insertion of the bearing into the housing.
More details of the coating process and transducer development
are given in Ref. 19.
The instrumented bearing was inserted into one of four bearing
housings in a bearing test apparatus that supported an 80 mm
diameter shaft see Ref. 20 for fuller details. This used a gravity
feed system to lubricate the bearings with a mineral turbine oil
Shell Turbo T68. This apparatus enabled vertically upwards radial loads to be applied to the shaft and hence the ball at the top of
the instrumented raceway was the most heavily loaded. Rotary
shaft speed was controllable in the range 1002900 rpm by a 7.5
kW electric motor. This control of load and speed then enabled
control of the resultant oil-film thickness typically in the range
0 1 m.
An optical sensor was used, both to allow accurate triggering of
the ultrasonic instrumentation and to measure shaft speed. This
was triggered by reflective tape attached to the ball cage which
rotates at half the shaft speed. When this tape passed the optical
sensor it triggered a signal generator Agilent 33220A. After the
addition of an adjustable delay the signal generator triggered a
pulser-receiver Panametrics 5072PR at its maximum pulse repetition frequency, which was 20 kHz.
The pulses pass through the outer-raceway and reflect from the
inner surface of the raceway below the transducer. If the ball is
not located beneath the contact then the pulses are almost completely reflected back from the raceway due to the high reflectivity
of a steel-air interface. When the trigger was set appropriately the
pulses reflected back from the oil-film that forms between the ball
and the raceway as it moves beneath the measurement location.
The pulser-receiver outputs a voltage spike containing frequencies centered on 200 MHz and had a receive bandwidth of up to
400 MHz. The reflected signals were then passed to a digital oscilloscope sampling frequency 5 GS/s and PC for storage and
analysis. Figure 4 shows the reflected signal amplitude recorded
for 500 successive data points the maximum capacity of the storage oscilloscope used. In this example, the data storage was triggered such that two ball passages can be seen.
Figure 5 shows typical time and frequency domain plots of the
signal reflected from the lubricant-film using the piezo thin film
transducer on the bearing. In the time domain Fig. 5a, a series
of reverberations can be seen that, after experimentation, were
found to be caused by electrical impedance mismatch between the
cable and the receiver part of the pulser-receiver. To overcome
JANUARY 2009, Vol. 131 / 011502-3

Downloaded 09 Jan 2009 to 137.222.187.115. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Table 1 Acoustic properties of lubricant and bearing steel

Reflected signal amplitude (arb.)

1.8
1.6

Density
kg/ m3

Longitudinal wave
velocity cl
m/s

Bulk modulus B
GPa

876
1044
7900

1460
4500
5900

1.84
21.2
172

1.4

T68 at 0.1 MPa


T68 at 1.5 GPa
Bearing steel

1.2
1
0.8
0.6

Rf =

0.4
0.2
0

50

100

150

200

250

300

350

400

450

500

Data point number

Fig. 4 Amplitude of the reflected signal for 500 successive


data points. The oil-film generation from two ball passages
have been captured.

Results

The reflection coefficient data obtained from the piezo thin film
transducer under various load and speed conditions is shown in
Fig. 6. A distance of zero corresponds to the center of the lubricated contact. Away from the central region i.e., at distances
greater than 300 m the reflection coefficient increases toward
unity. Note that at distances of over 1 mm from the center of the
lubricated contact the reflection coefficient remained within
10% of unity. Remote from the contact, signals are reflected
from a steel-air or steel-oil interface depending on whether the
bearing cavity between the balls remains partially filled or fully
flooded with oil. The former would have a reflection coefficient
R 1, the latter of R 0.95.
When the contact is under the measurement location, the reflection coefficient reduces distinctly as a greater proportion of the
signal is transmitted through the oil-film. It might be expected that
since the oil-film in the contact region is largely parallel 17 that
the reflection coefficient should remain at a constant level corresponding to a value predicted by Eq. 2. However, it can be seen
from Fig. 6 that the results particularly at W = 2.5 and 5 kN
appear to exhibit a local reflection coefficient maximum at the

b)

a)

Spectrum (arbitrary unit)

0.4

Amplitude (V)

0.2
0
-0.2
-0.4

-0.6
1.32

1.34

1.36

1.38

1.4

1.42

Time (s)

1.44

17

where Amf is the amplitude of the signal reflected from the lubricant film layer, Areff is the amplitude of the reference signal,
and Rref is the reflection coefficient of the reference interface. The
reflection coefficient calculated from Eq. 17 can then be used in
Eq. 2 to extract the lubricant-film thickness assuming all other
material constants, acoustic properties, and the reference reflection coefficient are known. The relevant acoustic properties used
for bearing steel and the lubricant are given in Table 1.

4
this problem a 3 m cable was used to enable the cable reverberations to be separated and hence permit extraction of the first signal, corresponding to the lubricant-layer reflection. Note that the
distance from the transducer to the inner surface of the outerraceway distance was 4.5 mm and so the total return path of the
ultrasonic echo was twice this. The time domain data of the oilfilm reflection shown in Fig. 5a was converted, via a fast Fourier
transform, to the frequency domain. It can then be seen from Fig.
5b that the piezo thin film transducer is extremely wide band,
with energy in the range 50350 MHz and has a center frequency
of around 200 MHz.
The reflection coefficient was measured by comparing the signal reflected from the interface of interest to that from a known
reference interface. In this case the interface of interest was the
lubricant-oil-film between the outer-raceway and the ball and the
reference interface was that from the inner surface of the outerraceway in the absence of both ball and lubricant i.e., a steel-air
interface and so Rref 1. The reflection coefficient was then calculated from

Amf
Rref
Areff

4
3.5
3
2.5
2
1.5
1
0.5

1.46

1.48

1.5

50

100

150

200

250

300

350

400

Frequency (MHz)

Fig. 5 Response of pulse-echo signal obtained from a piezo thin film transducer deposited onto the bearing outer-raceway
total propagation distance 9.0 mm. Signal A is used to extract lubricant-layer thickness and signals B are reverberations
in the cable; a time domain and b frequency domain.

011502-4 / Vol. 131, JANUARY 2009

Transactions of the ASME

Downloaded 09 Jan 2009 to 137.222.187.115. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

1
0.9

Reflection Coefficient, R

0.8

W=2.5 kN, w=506rpm


+ W=5 kN, w=364rpm
o W=10 kN, w=166rpm
x W=15 kN, w=106rpm

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-500 -400 -300 -200 -100

100

200

300

400

500

Distance, x (m)

Fig. 6 Reflection coefficient profile measured as the bearing


rotates. The center of the lubricated contact occurs at x = 0. The
horizontal dashed lines indicate the predictions based on Eq.
3.

center of the lubricated contact. The central film thickness for


each test case was predicted using Eq. 3; this value was then
used to determine the reflection coefficient using the spring model
from Eq. 2. The predictions are shown on the plot as a series of
horizontal lines and the predicted reflection coefficient can be
seen to correlate better with the experimental value at the contact
center.
It is illuminating to consider the reflection coefficient behavior
seen in Fig. 6 in more detail. First, note that the film thickness
within the contact region is known to be approximately parallel
with a small horseshoe shaped constriction on the exit side. Second that the bulk modulus increases with pressure 16 so it might
be expected that the oil at the center of the contact is slightly
stiffer than that at the edges, which would lead to further reduced
reflection coefficient. Consideration of these facts and the spring
model Eq. 2 suggests that the reflection coefficient distribution
should be a minimum at the center of the contact, where the
pressure and bulk modulus are largest, and increase smoothly and
monotonically toward the edges of the contact as the contact pressure falls to zero. If the horseshoe constriction were visible, it
would appear as a local minimum in reflection coefficient toward
the exit edge of the contact right hand side. The experimentally
observed reflection coefficient distribution within the contact differs from this expected trend, particularly at low loads. The reason
for the anomalous central maximum is explained by the deflection
of the raceway under ball passage load as discussed in following
section.
z

Discussion

Under normal operation, the inner surface of the outer-raceway


and the ball undergo displacement due to the contact stress. This
is shown schematically in Fig. 7. The reference reflection Rref is
recorded from the undeformed raceway-air interface when the ball
is not in position beneath the transducer Fig. 7a. When the ball
passes underneath the transducer, the oil-film reflection R is measured, which corresponds to when the raceway is subjected to
deformation by the contact Fig. 7b. Thus the reflected signal is
recorded from a deformed raceway surface, while the reference
signal is from an undeformed surface.
When the ball is directly below the transducer the reflection is
from a symmetrical surface that has a lower radius of curvature
relative than the undeformed surface. This smaller radius will
cause a more divergent reflection relative to the undeformed
case and hence decrease the measured reflection coefficient.
When the ball is on either side of this central alignment Fig.
7c, the reflection is from an asymmetric surface and so this will
tend to skew the ultrasonic beam away from the transducer. Again,
this situation would be expected to cause a reduction in the measured reflection coefficient relative to the undeformed case. This
source of discrepancy is analyzed here and shown to be the cause
of the anomalous reflection maximum at the contact center.
The lubricated contact is elliptical and in the 6016 bearing
tested, the major to minor radius aspect ratio is 11:1. For the
purposes of this study, the contact can therefore be modeled as a
two-dimensional 2D line contact with reasonable accuracy. In
this section a 2D contact model and a 2D ultrasonic wave propagation model are used to investigate the effects of bearing curvature and local deformation on the measurement of oil-film thickness.
Equation 16 is used to compute the deformed geometry of the
ball and raceway under various operating conditions load and
speed using the data shown in Tables 1 and 2. The calculated
deformed and undeformed surfaces for the 6016 bearing under the
conditions used in this paper are shown in Fig. 8.
The effect of the bearing deformation on the ultrasonic reflection was modeled using a simple ray approach. In this way the
transducer was considered to output a number N typically, N
= 40 of rays normal to its surface. These rays are then traced to
the surface of the outer-raceway and are assumed to be reflected
from a flat surface that is the tangent of the surface at that point.
At a given location on the outer-raceway the tangent, , is given
by

= tan1

18

Tangent to deformed surface


Ray path

Outer-raceway

zb
x

The reflected rays are then traced back to the transducer plane
i.e., z = d and their location on the x-axis xR is given by

Ultrasonic transducer
d

Air
Ball Rax

(a)

Rbx
P

(b)

(c)

Fig. 7 The effect of ball location on ultrasonic reflection; a the reference signal is
recorded when the ball is remote from the transducer location, b the ball directly beneath the transducer, and c the ball just past the transducer location.

Journal of Tribology

JANUARY 2009, Vol. 131 / 011502-5

Downloaded 09 Jan 2009 to 137.222.187.115. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Table 2 Parameters used to calculate contact dimensions, surface deflection, and oil-film
thickness in the test 6016 ball bearing
Reduced
modulus E
GPa
228

Reduced
radius R
mm

Second elliptic
integral

Ellipticity
ratio
k

5.85

1.014

11.5

xR = x + d tan2

19

where d is the thickness of the outer-raceway. The number of rays


NR that return to the transducer location x d / 2 is then
counted. The geometrical reflection coefficient Rg can then be
calculated as the ratio of the number of reflected rays in the deformed and undeformed cases,
Rgx =

NR
NR

20

Figure 9 shows the variation in tangent angles across the surface


of the outer-raceway for different bearing loads. Figure 10 shows
the simulated reflection coefficient, given by Eq. 20, which
would be caused by the surface deformation alone. From Fig. 9, it
can be seen that for the region x = 200 m the tangent angle

Effective viscosity
for T68

First elliptic
integral

Pressure viscosity
coefficient
for T68 a
GPa1

3.81

20

0.2

increases linearly with distance for both the unloaded and loaded
cases. It can also be seen that, for the loads plotted, the gradient of
this linear increase is almost independent of load. This means that
for the different load cases considered in this paper the geometrical reflection coefficient Fig. 10 has little load dependency in
this central region. In this central region, the geometrical reflection coefficient is 0.7, meaning that the deformed raceway causes
a 30% reduction in the signal amplitude, relative to the undeformed surface.
Away from the central region, the behavior is more complex
and can be seen to exhibit the same local minima as was observed
experimentally see Fig. 6 and also Rg tends to unity for large x.

12

2
Tangent angle, (degrees)

Geometry of the outer-raceway, zb (m)

10
8
6
4

N / m2 s

No load
After load

W=2.5 kN, + W=5 kN


o W=10 kN, x W=15 kN

-1

-2
0
No load
After load

-2
-4
-500

-400

-300

(a)

-3
-500 -400 -300 -200 -100

W=2.5 kN, + W=5 kN


o W=10 kN, x W=15 kN

-200 -100

100

200

100

200

300

400

500

Distance, x (m)

300

400

500

Distance, x (m)

Fig. 9 The tangent angle as a function of distance across the


outer-raceway for a range of bearing loads
1

0.9
0.8

-10

-15

-20

-25
-500 -400 -300

(b)

Geometrical Reflection Coefficient

Geometry of the ball, za (m)

-5

No load
Under load

W=2.5 kN, + W=5 kN


o W=10 kN, x W=15 kN

-200 -100

100

200

0.6
0.5
0.4
0.3

W=2.5 kN
+ W=5 kN
o W=10 kN
x W=15 kN

0.2
0.1

300

400

500

Distance, x (m)

Fig. 8 The undeformed and deformed shape of the contact


surfaces; a outer-raceway and b ball

011502-6 / Vol. 131, JANUARY 2009

0.7

0
-500 -400 -300

-200 -100

100

200

300

400

500

Distance, x (m)

Fig. 10 Simulated geometrical reflection coefficient Rg


caused by the surface deformation alone

Transactions of the ASME

Downloaded 09 Jan 2009 to 137.222.187.115. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

0.8

0.9

0.7
0.6
0.5
0.4
0.3
0.2
0.1

0.8

100

200

300

400

l
l

0.6

0.4
0.3

l ll

l
l

l
ll l

Rx = RgxRexx

21

A simple form of reflection coefficient profile is assumed in which


the reflection is constant and given by Eq. 2 inside the contact
and is unity outside the contact:
1

when x a

w = 166 rpm

AlN:

W=2.5 kN, + W=5 kN


O W=10 kN, x W=15 kN
Focsd:
W=2.5 kN, W=5 kN
W=10 kN,
W=15 kN

0.2

0
0

0.2

0.4

0.6

0.8

1.2

w = 106 rpm

1.4

1.6

1.8

Measured oil-film thickness (m)

As a final step, the expected reflection coefficient profile at a


given load and speed Rex can then be multiplied by the geometrical reflection coefficient at the same load to predict the reflection
coefficient that would be recorded by the transducer.

Rh when x a

w = 364 rpm

0.5

500

Fig. 11 The reflection coefficient profiles obtained when an


assumed reflection coefficient given by Eq. 22 is distorted
by the reflection effect caused by surface deformation. The
dashed lines indicate the theoretical values based on the oilfilm thickness calculated from Eq. 3.

Distance, x (m)

Rexx =

w = 506 rpm

l
l

0.7

0.1

0
-500 -400 -300 -200 -100

l
l

Theoretical oil-film thickness (m)

0.9

Reflection coefficient

W=2.5 kN, w=506rpm


+ W=5 kN, w=364rpm
o W=10 kN, w=166rpm
x W=15 kN, w=106rpm

22

where h is calculated using Eq. 2. Then the ray model is used to


quantify the effect of the geometry on this expected reflection
coefficient profile using Eqs. 20 and 21. Figure 11 shows the
result of this prediction using the data shown in Table 3. This plot
is then the expected reflection coefficient i.e. from an undeformed surface distorted by the surface deformation effect on
reflection. This can now be compared directly with the equivalent
experimental result shown in Fig. 6.
The agreement is reasonable, demonstrating that the change in
bearing geometry between the reference signal unloaded and the
measurement signal loaded was the cause of the unexpected
form of the experimentally observed reflection coefficient distribution. More importantly, as the desired measurement is the central oil-film thickness this can be obtained simply from the central
reflection coefficient divided by relevant local geometrical reflection coefficient in this case 0.7. This provides the approach for
correcting a measured reflection coefficient to accommodate the
effect of surface deformation.

Fig. 12 A comparison of experimentally measured oil-film


thickness including a correction for surface deformation with
EHL theoretical solution Eq. 3 for a range of bearing loads
W and speeds w. Note that label AlN refers to measurements made using the piezo thin film transducer and label
Focsd refers to data from Ref. 19 obtained using a highly
focused 50 MHz transducer.

The test bearing was run at a range of loads and speeds during
which measurements of the reflection coefficient were recorded.
The central reflection coefficient was then extracted and corrected
for geometry by division by the relevant geometrical reflection
coefficient 0.7. The result was then used in Eq. 2 to determine
the oil-film thickness. Figure 12 shows the results. At each test
condition, 20 repeat measurements were made and the average
and error bars representing three standard deviations are shown.
The experimental results are compared with the theoretical solution from Eq. 3. Also shown on the graph are results obtained in
previous work where a conventional highly focused, water
coupled, 50 MHz ultrasonic transducer was used 20. From Fig.
12 it can be seen that not only are the piezo thin film results in
agreement with the previous results, they are somewhat more
tightly grouped around the theoretical line than those from the
focused 50 MHz transducer. It should also be noted that the theoretical prediction from Eq. 3 was based on an assumption that
the viscosity of the oil remained constant during the tests. This
was assumed, in the absence of any further information, and because the tests were conducted in rapid succession before the
bearing could equilibrate after an increase in load or speed. It is
likely that at higher speed the oil entering the contact would be
hotter and hence at a lower viscosity. This would mean a thinner
film being formed than predicted, explaining why for the data at
higher speeds the measured film thickness is above the predicted
value. Overall, this demonstrates that slightly improved results
can be achieved using the piezo thin film device. More importantly, the piezo thin film transducer is significantly more convenient due to its low profile nature and direct contact with the

Table 3 Minor semicontact radius a and central oil-film thickness hc for various bearing operating conditions
Bearing load W kN
Minor semicontact radius
a m
Central film thickness hc m
at various bearing speeds w rpm

Journal of Tribology

w = 106
w = 166
w = 364
w = 506

2.5

10

15

163
0.3
0.42
0.71
0.89

210
0.28
0.4
0.67
0.84

267
0.28
0.38
0.64
0.8

307
0.27
0.37
0.62
0.78

JANUARY 2009, Vol. 131 / 011502-7

Downloaded 09 Jan 2009 to 137.222.187.115. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

bearing. This negates the requirement for a bulky immersion type


transducer with the necessity of a water bath within the bearing
housing.

Conclusions

A novel transducer for the measurement of lubricant-film thickness in bearings has been described. This used the thin film piezoelectric material aluminum nitride that was sputtered onto the outside of the outer-raceway of a deep groove ball bearing. This
resulted in a transducer and electrode combination of less than
10 m in thickness and so offers significant potential for integration in bearing systems with minimal disruption. The active area
of the transducer measured 3 0.3 mm2, and it operated with a
center frequency of 200 MHz. Crucially, this resulted in a narrow
columnar ultrasonic beam that had spatial dimensions less than
those of the lubricated contact. Ultrasonic reflection measurements from the lubricant-film formed as the ball passed under the
transducer location were recorded. This reflection was shown to
be susceptible to small geometrical changes due to ball-raceway
contact. A calibration procedure was then described whereby these
geometrical effects can be reduced and the desired lubricant-film
thickness extracted. The results demonstrated that the lubricantfilm thicknesses can be measured to reasonable accuracy and are
comparable, if not an improvement, over those made with the
more bulky ultrasonic transducer systems used previously.

Acknowledgment
This work was funded by the U.K. Engineering and Physical
Sciences Research Council Grant Nos. GR/S46963/01 and GR/
S46956/01.

References
1 Stachowiak, F. W., and Andrew, W. B., 2001, Engineering Tribology, 2nd ed.,
Butterworth-Heinemann, Boston, Chap. 7, pp. 281313.
2 Cann, P. M., Spikes, H. A., and Hutchinson, J., 1996, The Development of a
Spacer Layer Imaging Method SLIM for Mapping Elastohydrodynamic Contacts, Tribol. Trans., 394, pp. 915921.
3 Hbel, M., and Haffner, K., 1999, An On-Line Monitoring System for OilFilm, Pressure and Temperature Distributions in Large-Scale Hydro-Generator
Bearings, Meas. Sci. Technol., 105, pp. 393402.

011502-8 / Vol. 131, JANUARY 2009

4 Baly, H., Poll, G., Cann, P. M., and Lubrecht, A. A., 2006, Correlation Between Model Test Devices and Full Bearing Tests Under Grease Lubricated
Conditions, Proceedings of the IUTAM Symposium on Elastohydrodynamics
and Micro-Elastohydrodynamics, R. W. Snidle and H. P. Evans, eds., Springer,
Dordrecht, Vol. 134, pp. 229249.
5 Jardine, A. K. S., Lin, D., and Banjevic, D., 2006, A Review on Machinery
Diagnostics and Prognostics Implementing Condition-Based Maintenance,
Mech. Syst. Signal Process., 207, pp. 14831510.
6 Glavatskih, S. B., 2004, A Method of Temperature Monitoring in Fluid Film
Bearings, Tribol. Int., 372, pp. 143148.
7 Wang, W. Q., Ismail, F., and Golnaraghi, M. F., 2001, Assessment of Gear
Damage Monitoring Techniques Using Vibration Measurements, Mech. Syst.
Signal Process., 155, pp. 905922.
8 Miettinen, J., Andersson, P., and Wikstrom, V., 2001, Analysis of Grease
Lubrication of a Ball Bearing Using Acoustic Emission Measurement, Proc.
Inst. Mech. Eng., Part J: J. Eng. Tribol., 2156, pp. 535544.
9 Anderson, W., Jarzynski, J., and Salant, R. F., 2000, Condition Monitoring of
Mechanical Seals: Detection of Film Collapse Using Reflected Ultrasonic
Waves,Proc. Inst. Mech. Eng., Part C: J. Mech. Eng. Sci., 214, pp. 1187
1194.
10 Dwyer-Joyce, R. S., Drinkwater, B. W., and Donohoe, C. J., 2003, The Measurement of Lubricant Film Thickness Using Ultrasound, Proc. R. Soc. London, Ser. A, 459, pp. 957976.
11 Dwyer-Joyce, R. S., Harper, P., and Drinkwater, B. W., 2004, A Method for
the Measurement of Hydrodynamic Oil Films Using Ultrasonic Reflection,
Tribol. Lett., 172, pp. 337348.
12 Dwyer-Joyce, R. S., Harper, P., Pritchard, J., and Drinkwater, B. W., 2006,
Oil Film Measurement in PTFE-Faced Thrust Pad Bearings for Hydrodynamic Applications, Proc. Inst. Mech. Eng., Part A, 220, pp. 619628.
13 Zhang, J., Drinkwater, B. W., and Dwyer-Joyce, R. S., 2006, Monitoring
Lubricant Film Failure in a Ball Bearing Using Ultrasound, ASME J. Tribol.,
1283, pp. 612618.
14 Dowson, D., and Higginson, G. R., 1977, Elasto-Hydrodynamic Lubrication
(SI Edition), Pergamon, Oxford, Chap. 7, pp. 78105.
15 Harris, T. A., 2001, Rolling Bearing Analysis, 4th ed., Wiley, New York, Chap.
25, pp. 9631010.
16 Jacobson, B. O., and Vinet, P. A., 1987, Model for the Influence of Pressure
on the Bulk Modulus and the Influence of Temperature on the Solidification
Pressure for Liquid Lubricants, ASME J. Tribol., 109, pp. 709714.
17 Hamrock, B. J., and Dowson, D., 1981, Ball Bearing Lubrication: The Elastohydrodynamics of Elliptical Contacts, Wiley, New York, Chap. 3, pp. 6879.
18 Cameron, A., 1966, The Principles of Lubrication, Longmans, London, Chap.
8, pp. 189212.
19 Lee, C.K., Cochran, S., Abrar, A., Kirk, K.J. and Placido, F., 2004, Thick
Aluminium Nitride Films Deposited by Room-Temperature Sputtering for Ultrasonic Applications, Ultrasonics, 4219, pp. 485490.
20 Zhang, J., Drinkwater, B. W., and Dwyer-Joyce, R. S., 2006, Acoustic Measurement of Lubricant-Film Thickness Distribution in Ball Bearings, J.
Acoust. Soc. Am., 1192, pp. 863871.

Transactions of the ASME

Downloaded 09 Jan 2009 to 137.222.187.115. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

S-ar putea să vă placă și