Professional Documents
Culture Documents
2001-GT-0028
John J. Yu
Donald E. Bently
Paul Goldman
Kenwood P. Dayton
Brandon G. Van Slyke
Test Bearing Ball and roller bearings were used for the study of transducer
response to raceway and rolling element defects, respectively.
Approximate ratios of tested bearing frequencies to shaft speed, based
on experimental data from good bearings, are shown in Table 1. Note
that these values may change due to defect effects and vary with
operating conditions. If a defect occurs on inner or outer race,
balls/rollers will contact the defect at frequency iIRBP × (i.e., iIRBP
High-gain displacement transducer
REBAM probes times shaft speed) or iORBP × . The sum of these two ratios is equal to
the number of balls/rollers regardless of changes in speed and slip. If
Fig. 1 Bearing test rig there is a defect on a ball/roller, inner or outer race will contact the
defect at frequency iR × . Notice that since a displacement transducer
2 Test Rig detects outer race deflection facing its tip, parameter iC , the frequency
The test rig used for detecting and diagnosing bearing defects is ratio of the moving cage that holds balls/rollers is related to the
shown in Fig. 1. The test rig consists of a rotor with one support detection of ball/roller defects.
bearing and one test bearing.
The rotor includes a 746 W (1 hp) electric motor, which is connected 3 Vibration Signature for Good Bearings
to the main shaft through a laterally flexible coupling. The step main Both high-gain displacement transducers and accelerometers were
shaft is 419 mm (16.5 in) long with diameters varying from 30 mm installed to observe the corresponding vibration signature for a non-
(1.1811 in) to 41 mm (1.6 in). Attached to the shaft are two mass defective test bearing. The purpose was to find an effective sensor that
disks with 177.8 mm (7 in) diameter and 25.4 mm (1 in) thickness. clearly reflects the bearing operation, and an avenue that can expose
The rotor is balanced by putting mass in one to several holes on the bearing defects.
disks to compensate for unbalance. The high-gain eddy current transducer mounted in the hole of the
A hydraulic pump with 25.4 mm (1 in) diameter cylinder was used housing can detect very small elastic deflection of the outer race
to apply the load to the shaft through a roller bearing with the number penetrated by the transducer as each rolling element under load passes
of elements much higher than the test bearing. In this way, the over the hole. For precaution, the strength of the outer race should be
frequencies of the loading bearing would not interfere with the test considered in order to prevent race failure through cyclic strain when
bearing frequencies. The load on the bearing can be applied either in determining the hole size in terms of the bearing size and loading. The
an upward direction or downward direction, and can be varied between 360o
–2224 N (–500 lbs.) and +2224 N (+500 lbs.). Note that the test deflection has a cycle of shaft rotation. A good bearing
iORBP
bearing is loaded to some extent, even without hydraulic loading, due
to the weight of the mass disks on this overhung rotor. The motor is possesses a smooth deflection curve in time base. Figure 2 shows
connected to a variable frequency drive, which allows for ramped data vibration signature from the transducer mounted at the left side
collection at a controlled acceleration. The bearing housings are fixed 45o from the bottom. Besides the weight of the whole rotor system, a
to the casing, which is fixed to a large concrete foundation. hydraulic load of 1379 kPa (200 psi) downward is applied at the right
The displacement transducers, which had a gain of 78.7 mV/µm (2 end of the shaft as shown in Fig. 1. The test bearing has no defects on
mV/µ in) compared to 7.87 mV/µm (0.2 mV/µ in) for a generic eddy its inner race, outer race, or balls. No lubrication is supplied between
current proximity transducer, were used to measure the outer race the balls and raceways.
deflection of the bearing. Accelerometers were also mounted on the The localized deflection curve of the outer race, facing the probe, is
casing to compare with these transducers. clearly shown in time base where the shaft speed is 680 rpm (see Fig.
2(a)). The deflection has a peak-to-peak amplitude of about 1.65 µm
probe
Amplitude (µm)
20 0.50
high peak points in the time base shown in Fig. 2(a) and is the
10 0.25
maximum deflection towards the probe. When two adjacent balls are
0 0 just equally away from the probe hole, the deflection of the outer race
10 0.25 around the hole is released so that the outer race is deflected away
20 0.50
from the probe. This case corresponds to the low peak points in time
base shown in Fig. 2(a). Due to surface roughness with lack of
0.05 0.1 0.15
One shaft revolution
0.2 0.25 0.3 lubrication, ball tolerances, and ball undulations, there exists slight
Time (sec) 680 rpm
fluctuation on the deflection curve.
(a) Figure 2(b) shows the deflection signal in frequency domain. Outer
race ball pass frequency, i.e., iORBP × ( = 3.06 × in this case)
component is dominant, accompanied by its harmonics, 2 (iORBP ×) ,
Amplitude (µ in, pp)
30 0.75
3 (iORBP × ) , etc. The synchronous 1 × component is almost negligible.
Obviously, inner race ball pass frequency, i.e.,
20
Harmonics of iORBP×
0.50
iIRBP × ( = 4.94 × in this case) component is not apparent in frequency
10 0.25 domain.
Data from accelerometers give the casing vibration signals that
0 0
0 10 20 30 40 50 60 70 80 90 include transducer response to rotor unbalance, foundation, and other
Frequency (kcpm) 680 rpm frequency components in addition to the rolling element bearing
(b) operation. Figure 3 shows the signals in both vertical and horizontal
directions for the same operating condition as shown in Fig. 2.
Fig. 2 Outer race deflection measured by displacement Information from rolling element bearing operation is obscured or
transducer. (a)time base, and (b) frequency spectrum. masked with the vibration signature from other sources in the
rotor/bearing/foundation system. In the time domain as shown in Fig.
(65 µ in). For each shaft rotation, the number of periodic waveform 3, spikes or impulses occur although no defects exist in the bearing. In
the frequency domain as shown in Fig. 3, vibration signals include
Good bearing
0.5 5
0.2 2
0.4 4
Vertical Amplitude (g)
0.1 1 0.3 3
0 0.2 2
0
0.1 1
0.1 1
0 0
0.2 2
0.1 1
0.3 3
0.2 2
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Time (sec) 680 rpm Time (sec) 680 rpm
iORBP×
Horizontal Amplitude (m/s2, pk)
0.0025
18iIRBP×
Vertical Amplitude (m/s2, pk)
0.025
iORBP× 9iIRBP×
Vertical Amplitude (g, pk)
0.0020 0.020
0.0020 0.020
0.0015
iIRBP×
0.015
0.0015 0.015
0 0
0 10 20 30 40 50 60 70 80 90 0 0
0 10 20 30 40 50 60 70 80 90
Frequency (kcpm) 680 rpm Frequency (cpm) 680 rpm
(a) (b)
Fig. 3 Casing response in time base and frequency spectrum from accelerometers mounted in (a) vertical,
and (b) horizontal directions.
♦ Ω
1st ♦
∆e e Οc
+e ∗+
c 2 nd
∆
o ∗
Ο
e αeffective
c
∆c
360 e
` 360o ♦c 3 rd
+
iIRBP αdelay
∗
iIRBP Time 4 th
♦a
∆a
+a Οb
+ ∗b
5th Ο
∗Ο a a
♦b ∆ b
b Moving inner
race defect spot . .
REBAM in contact with . .
probe balls/rollers
Outer race . .
deflection 360° (One shaft revolution)
(closer to probe)
No spike in
between
……... Ωt
β1♦
a β2 ∆ a β3 +a β4 ∗ a Οa
360°+ αdelay 360°+ αdelay 360°+ αdelay Contact spot beyond αeffective
Period T
Fig. 4 Diagram of inner race defect detection methodology using high-gain displacement transducer
iIRBP 18
rotation from “♦a”, where the defect spot contacts the next ball at 17
iIRBP
“♦b”, the probe cannot detect a negative spike because the contact Round
Round (i IRBP ) − i IRBP
16
T (shaft revolution)
occurs beyond the effective zone. For the following contact at “♦c”, 15 T=
iIRBP
as shown in Fig.4 when the defect contacts the following ball, the 14
same situation occurs where the contact event is not within the 13
effective zone of the probe. This situation exists until the defect spot 12
contacts the jth (j is a round number for iIRBP ) ball within the effective 11
zone again, as is indicated with “∆∆a” in the effective zone. Note that 10
two conditions must be met for the probe to detect these negative 9
spikes. One is that the inner race defect spot has to contact balls. The 8
other is that the defect spot in contact with balls has to be within the 7
effective zone. 6
4.85 4.86 4.87 4.88 4.89 4.90 4.91 4.92 4.93 4.94 4.95
A delay angle between the two contact spots within the effective
iIRBP
zone “♦a” and “∆ ∆a” can be given by
360o Fig. 5 Relation between iIRBP and detected spike group period T
α delay = Round (iIRBP ) − iIRBP × (1)
i
IRBP
where Round (iIRBP ) represents a round number for iIRBP (for Thus the period T is given by
example, Round (4.9 ) = 5, Round (4.1) = 4 ). In Fig.4, the defect iIRBP
Round
contact spot moves “forward” (from “♦a” to “∆ ∆a”, the same direction Round (iIRBP ) − iIRBP
T=
n
=
as Ω ) within the effective zone as α delay is assumed to be greater than shaft revolutions (6)
iIRBP iIRBP
zero in this case. It would move “backward” in the case of α delay < 0 . Note that a slight change in iIRBP causes a much bigger change in
In time base the two adjacent negative spikes have in terms of shaft detected spike group cycle T, as shown in Fig. 5. Therefore, a change
rotations (degree) an interval in iIRBP , which is related to slip, can be more accurately estimated
360o
α IR = 360o + α delay = Round (iIRBP ) (2) from T .
iIRBP During a cycle T in time base, the number of consecutive spikes
The interval α IR will be longer than 360o shaft rotation for α delay > 0 nspike is approximately equal to
Amplitude (µm)
Amplitude (µ in)
135°
0 Ω 0
60.1 msec
40
(367.96°°, compared with 360°°+αdelay= 367.35°°) 1.0
α delay
0 0.1 58.8 msec 0.2 0.3 0.4
360o (360°°)
= Round ( 4.9 ) − 4.9 × Time (sec) 1020 rpm
4.9
= 7.35o 11.9 msec Estimated defect orientation:
360o
40 (72.86o , compared with = 73.47o ) 1 4 1.0
iIRBP β defect ≈ ∑ β i = 45o
4 i =1
20
(same as the real location)
0.5
Amplitude (µm)
Amplitude (µ in)
135°
0 Ω 0
and more accurately from T ). Therefore, according to Eqs. (1) and 360o
(2), a delay angle within the detectable zone should be agreement with the expected value of = 73.47o . Note that only
iIRBP
360o
α delay = Round ( 4.9 ) − 4.9 × = +7.35o (Moving forward), 128 samples of data were taken for 360o shaft rotation in this case.
4.9
and in time base the two adjacent negative spikes should have an Using Eq. (8) along with the given β i ’s in the time base for the right
interval of probe, as shown in Fig. 6, the location of the defect spot can be
α IR = 360o + α delay = 367.35o > 360o estimated in the following:
Amplitude (µm)
Amplitude (µm)
Amplitude (µ in)
30 0.75
Amplitude (µ in)
10 0.250
20 0.50
Adeflection
Adeflection
5 0.125
10 0.25
0 0
0 0
5 0.125
10 0.25
10 0.250
20 0.50
15 0.375
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Defect Time (sec) Time (sec)
Removing iORBP× and 2iORBP× components Removing iORBP× and 2iORBP× components
(a) 5 0.125
20 0.50
10 0.25
Amplitude (µm)
Amplitude (µm)
0 0
Amplitude (µ in)
Amplitude (µ in)
50 1.25
0 0
Adeflection=57 µ in (1.45 µm)
5 0.125
40 1.00
10 0.250 10 0.25
Aspike
Amplitude (µ in)
30 0.75
Aspike
Amplitude (µm)
15 0.375 20 0.50
20 0.50
20 0.500 30 0.75
10 0.25 25
T
0.625
T T
40 1.00
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10 0.25 Time (sec) Time (sec)
20 0.50 including the one near the probe tip. Notice that small negative spikes
10 0.25 occur due to a sudden loose contact when the defect is on the right
side.
Aspike =63 µ in (1.6 µm)
0 0
Amplitude (µm)
For the same defect as shown in Fig. 7(a), increases in speed yield
Amplitude (µ in)
10 0.25
20 0.50
slight increases in deflection amplitude Adeflection and spike amplitude
30 0.75 Aspike . As is expected, hydraulic loading is linearly proportional to the
defect spot near the probe tip
40 1.00
deflection amplitude Adeflection . Figures 8(a) and 8(b) show time base
defect spot near the bottom 360o
50 Round (iIRBP ) 1.25
response in different speed and load conditions from those in Fig. 7.
defect spot on the other side iIRBP
60 1.50
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Notice that spike signals appear in a period T that is given by Eq. (3).
Time (sec) In Fig. 8(a) a spike group cycle is T = 12.6 shaft revolutions which
(c) correspond to iIRBP = 4.921 , while in Fig. 8(b) it becomes
Fig. 7 More severe defect case ( Aspike Adeflection = 1.1 ) at speed T = 11.6 shaft revolutions which correspond to iIRBP = 4.914 ,
680 rpm with hydraulic load 1379 kPa (200 psi). (a) real defect, compared with T = 10.6 shaft revolutions ( iIRBP = 4.905 ) in Fig. 7.
(b) time base signal, and (c) time base signal after removing These variations are due to slippage between raceway and balls, which
iORBP × and 2iORBP × components. could be estimated. Regardless of changes in speed and load, ratio
Aspike Adeflection remains almost the same (≈1.1).
2iORBP × components due to normal loading deflection makes spikes For a less severe defect case as shown in Fig. 9, ratio
pronounced without influence of their magnitude. For a given defect, Aspike Adeflection is decreased to around 0.4. This ratio, therefore, can
the ratio of spike-to-deflection amplitude seems to have little change.
be used to measure defect severity.
Figure 7 shows a more severe inner race defect case. The defect as
shown in Fig. 7(a) is located at the horizontal position ( 90o from 0o )
when the shaft reference signal is triggered. The left probe detects the
corresponding response in time base as shown in Fig. 7(b). The spikes
due to the defect become much pronounced after removing
iORBP × and 2iORBP × components due to normal loading deflection, as
shown in Fig. 7(c). For this severe defect, more negative spikes are
present as the defect contact spot passes near the probe tip. The spike-
to-deflection amplitude ratio is
Aspike Adeflection = 63 57 = 1.1
Defect
In this case, positive spikes also appear. This is because the severe
defect spot contacts a ball near the bottom where the maximum Fig.9 Less severe defect case ( Aspike Adeflection = 0.4 )
loading is located, thus impacting balls at other contact locations
Round (iIRBP )
20 0.50
16 0.40
harmonics corresponding to spike signals due to the inner race defect
0.30
12 as shown in Fig. 9. At speed 680 rpm, a frequency range of 10~20
8 0.20 kcpm is divided by about 15 intervals with harmonics of
iIRBP
4 0.10 × ( = 0.98 × ) for iIRBP = 4.9 . Observing whether these
Round (iIRBP )
0 0
0 10 20 30 40 50 60 70 80 90
fault frequency components occur is an additional tool to diagnose
Frequency (kcpm)
inner race defects in a rolling element bearing.
0.20 0.005
iIRBP
Harmonics of × 4.2 Outer Race Defect
Round (iIRBP ) An outer race defect spot in contact with balls/rollers does not move
against probes for a rolling element bearing with its outer ring fixed to
0.12 0.003 the casing. Figure 11 shows an outer race defect case at speed 680 rpm
with 1379 kPa (200 psi) hydraulic loading. Since the defect spot faces
0.08 0.002
the left probe, a sudden release of pressure between races and the
0.04 0.001
passing ball leads to a negative spike on the outer race deflection
curve. The number of spikes is the same as that of deflection cycles in
0 0 time base. Thus, a spike interval for an outer race defect is given by
10 11 12 13 14 15 16 17 18 19 20
Frequency (kcpm) 360o
α OR = (9)
iORBP
Fig. 10 Fault frequency components due to inner race defect
using displacement transducer Therefore, the fundamental frequency of spikes due to an outer race
defect is equal to that of the outer race deflection, i.e., iORBP × . Due to
4.1.4 Fault Frequency Components
the defect, many of its harmonics are remarkable in a frequency range
360o
For spikes with an interval of Round (iIRBP ) shaft rotation in higher than 3iORBP × . Since ball sizes vary, harmonics of the cage
iIRBP
Ω
Defect
Amplitude (µ in, pp)
15 0.375
Amplitude (µm)
10 0.250 8 0.20
5 0.125 6 0.15
0 0 4 0.10
5 0.125 2 0.05
0 0
0.1 0.15 0.2 0.25 0.3 0 10 20 30 40 50 60 70 80 90
Time (sec) Frequency (kcpm)
0.8 0.020
Amplitude (µ in)
Amplitude (µm)
0 0
0.6 0.015
5 360 o 0.125
iO R B P 0.4 0.010
10 0.250
0.2 0.005
Defect
5 0.125
7 0.175
6 0.150
Amplitude (µm)
0 0 5 0.125
4 0.100
4(4') 3 0.075
5
1(1') 0.125
2i 360o
Round R − 0.5 2 0.050
2i 360o iC 2 iR
Round R + 0.5 3(3')
iC 2 iR 1 0.025
2(2')
Spike due to rolling element defect
10 0.250 0
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 10 20 30 40 50 60 70 80 90 100
Time (sec) Frequency (kcpm)
αeffective Ω
Defect/raceway contact spots
0.12 0.0030
near the probe:
1→2→3→4 0.10 Harmonics of approximately iC × 0.0025
Amplitude (µm, pp)
Amplitude (µ in, pp)
4 or 0.08 0.0020
(a) (b)
Fig. 13 Rolling element defect signal detected by the probe at 700 rpm with 1379 kPa (200 psi). (a) time base after removing
iORBP × and 2iORBP × components with corresponding defect contact locations near the probe, and (b) frequency spectrum