See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/229018676

Vibration monitoring for defect diagnosis of rolling element bearings as a

predictive maintenance tool: Comprehensive case studies

Article  in  NDT & E International · November 2005

DOI: 10.1016/j.ndteint.2005.08.008

CITATIONS READS

144 1,368

3 authors:

Sadettin Orhan Nizami Aktürk

Ankara Yıldırım Beyazıt Üniversity Gazi University
37 PUBLICATIONS   464 CITATIONS    38 PUBLICATIONS   660 CITATIONS   

SEE PROFILE SEE PROFILE

Veli Çelik
Ankara Yildirim Beyazit University
18 PUBLICATIONS   347 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Vibrations of a shaft supported by ball bearings View project

Environmental Noise View project

All content following this page was uploaded by Sadettin Orhan on 19 January 2019.

The user has requested enhancement of the downloaded file.

 NDT&E International 39 (2006) 293–298
www.elsevier.com/locate/ndteint

Vibration monitoring for defect diagnosis of rolling element bearings

as a predictive maintenance tool: Comprehensive case studies
Sadettin Orhan a,*, Nizami Aktürk b, Veli Çelik a
a
Faculty of Engineering, Kırıkkale University, 71400 Kırıkkale, Turkey
b
Faculty of Engineering and Architecture, Gazi University, 06570 Maltepe, Ankara, Turkey
Received 26 July 2005; revised 29 August 2005; accepted 29 August 2005
Available online 17 October 2005

Abstract
Vibration monitoring and analysis in rotating machineries offer very important information about anamolies formed internal structure of the
machinery. The information gained by vibration analysis enable to plan a maintenance action. In this study, the vibration monitoring and analysis
case studies were presented and examined in machineries that were running in real operating conditions. Failures formed on the machineries in the
course of time were determined in its early stage by the spectral analysis. It was shown that the vibration analysis gets much advantage in factories
as a predictive maintenance technique.
q 2005 Elsevier Ltd. All rights reserved.

Keywords: Vibration monitoring; Rolling element bearing defects; Predictive maintenance

1. Introduction month in general and vibration is monitored by comparing

previous measurements to new ones. The vibration monitoring
The failures of rotating machineries can be very critical is based on the principle that all systems produce vibration.
because these lead to machinery damage, production losses and When a machine is operating properly, vibration is small and
personnel injury. So, a very important duty of the maintenance constant; however, when faults develop and some of the
department is to prevent these failures when they are in its dynamic processes in the machine change, the vibration
initial stage. The predictive maintenance by vibration analysis spectrum also changes [2]. There are many studies on the
is the best tool for this purpose. The vibration analysis is a vibration monitoring of the rotating machinery. Great amount
technique, which is being used to track machine operating of them concentrate on ball or cylindrical element bearing
conditions and trend deteriorations in order to reduce vibration monitoring [3–6]. In these studies a test rig, includes
maintenance costs and downtime simultaneously [1]. The ball or cylindrical roller bearing, is used and either an artificial
vibration analysis technique consists of vibration measurement defect is formed on bearing parts or the test rig is running until
and its interpretation. Firstly, vibration signals are collected by a desired defect arise on the bearing in laboratory conditions.
means of the vibration analyzer equipped with a sensor in the These studies were partly realized by ideal condition
time domain then, these signals are converted into frequency assumption. However, under conditions of real environment
domain by processing FFT, and the information gained from there are many factors that affect the actual running state of the
the vibration signals can be used to predict catastrophic machinery. Thus, these factors must be taken into consider-
failures, to reduce forced outages, to maximize utilization of ation. Studies related to real operating conditions of
available assets, to increase the life of machinery, and to reduce machineries were quite few. Gluzman [7] monitored vibration
maintenance costs related to health of machinery [2]. The of motor-generator system supported by ball and cylindrical
vibration measurements are taken periodically, one time per roller bearings to predict impending bearing failures. He
successfully identified impending failures of the bearing outer
and inner races. Al-Najjar [8] observed many bearing
* Corresponding author. Tel.: C90 318 357 3575; fax: C90 318 357 2459. vibrations in paper mills for many years to predict remaining
E-mail addresses: [email protected], [email protected] bearing life accurately. He also investigated effectiveness of
(S. Orhan). vibration-based maintenance and proposed some findings.
0963-8695/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. In this study, the application of vibration monitoring and
doi:10.1016/j.ndteint.2005.08.008 analysis was carried out on the machineries in a petroleum
294 S. Orhan et al. / NDT&E International 39 (2006) 293–298

refinery. The vibrations of the machineries were monitored for

a certain period and diagnosing of defects arised on the
bearings of the machineries during this period was aimed. The
study includes three different case studies that were obtained
from the machineries, which run under real operating
conditions.

2. Vibration measurement

Vibration data were gathered and processed using CSI 2110

[9] machinery analyzer. It is well known that a machinery
analyzer generally consists of a sensor, a memory in which the
signals are stored, electrical circuits that convert time domain
signals to frequency domain signals (FFT process) and a port
by which vibration signals were transfered into a computer.
The sensor used was an accelerometer (CSI 350) with a
sensitivity of 0.1 V/EU. Parameters for collecting of vibration
signals were given in Table 1. Vibration was measured in axial, Fig. 1. Rolling element bearing geometry.
horizontal and vertical directions. Obtained data from the
vertical direction was dominant compared with the other two because equations use shaft speed by declared manufacturer
directions so the vibration data measured in the vertical but real speed of the shaft may be different from this speed in
direction was used to characterize the health of the machinery. the moment of vibration measurement.
When a defect is formed on one of these parts of the bearing,
3. Defect frequencies of rolling element bearings related frequency, its orders, its sidebands, etc. may arise in a
spectrum graph. Sometimes these frequencies may exist for
The rolling element bearing defect produces certain healthy rolling element bearing due to some manufacturing
frequencies that depend on rolling element bearing geometry, errors. The defect frequencies of rolling element bearings used
which is shown in (Fig. 1), number of rolling element, and shaft in this study were given in Table 2.
speed. These frequencies are expressed in Eqs. (1)–(4).
 
n d 4. Case study I
wc Z 1K cosðaÞ (1)
2 D
The vibration of a huge centrifugal pump with nine vaned
   2 
n d d 2
was monitored. The power and revolution of the pump motor
wb Z 1K cos ðaÞ (2) are 160 kW, 2975 rpm, respectively. The reference measure-
2 D D
ment was taken on 17 October 2001 (Figs. 2 and 3). As shown
  from reference measurement, vibration behaviour of the pump
n d
wbp Z Nb 1K cosðaÞ (3) inner bearing composes of multipliers of the shaft rotation in
2 D
the spectrum graph and irregular impact signals in the time
  domain waveform graph.
n d
wbpi Z Nb 1 C cosðaÞ (4) This situation indicates that rotating equipment, namely ball
2 D
bearing, is looseness on the housing [10,11]. The vibration
where n, Nb, wc, wb, wbp, wbpi represent shaft speed, number of amplitudes are in low level. For this reason, maintenance of the
rolling elements, cage frequency, rolling element spin pump is not required in this stage. The vibration monitoring
frequency, rolling element pass frequency outer race, rolling was continued and after three weeks, vibration amplitude
element pass frequency inner race, respectively. Sometimes the increased suddenly (Figs. 4 and 5). This situation shows that
defect frequencies computed by above equations deviate to ball bearing looseness developed. The maintenance was
some amount from obtained-ones by measurement. This is planned and then ball bearing was pulled out on 13 February
2002, finally it was shown that the housing was worn out and
Table 1 the outer race of ball bearing was corroded (Fig. 6). After
Parameters for vibration measurement
Table 2
Number of spectral lines 400 Defect frequencies of rolling element bearings
Number of average 6
Number of gathered data 1024 Bearing type Shaft speed Nb wc (Hz) wb (Hz) wbp wbpi
Window type Hanning (Hz) (Hz) (Hz)
Spectral average mode Normal SKF NU 224 12.35 17 5.385 43 91.55 122.14
Frequency range 60–3000 Hz SKF 6222 16.41 10 6.78 43.56 67.82 98.48
 S. Orhan et al. / NDT&E International 39 (2006) 293–298 295

Fig. 2. Reference time waveform graph of the pump inner bearing.

Fig. 5. Vibration time waveform graph resulted from increasing of looseness.

Fig. 6. Corrosion on the ball bearing outer race due to looseness.

maintenance, on 7 March 2002 and following measurements,

vibration amplitude decreased to the normal value (Fig. 7). By
means of vibration monitoring ball bearing looseness, which
may lead to break up bearing, later the pump, was easily
determined, thus, possible damage and production loss was
Fig. 3. Reference spectrum graph of the pump inner bearing. prevented.

Fig. 4. Vibration time waveform graph resulted from increasing of looseness. Fig. 7. Overall vibration level trend of the pump inner bearing.
296 S. Orhan et al. / NDT&E International 39 (2006) 293–298

Fig. 8. Reference time domain waveform graph of the fan motor inner bearing. Fig. 11. Reference time waveform graph of the fan motor outer bearing.

Fig. 9. Reference spectrum graph of the fan motor inner bearing. Fig. 12. Reference spectrum graph of the fan motor outer bearing.

5. Case study II

Cylindrical rolling element bearing is often used to support

motors in industrial applications. Inner bearing vibration of the
fan motor, which supported by SKF NU 224 types cylindrical,
rolling bearing, was monitored periodically. The power and
revolution of the motor are 345 kW, 741 rpm, respectively. The
reference measurement was taken on 20 July 2001 (Figs. 8
and 9). By analyzing the vibration signatures, it is shown that
the bearing condition was critical. Existing of multipliers of the
cylindrical rolling bearing outer race defect frequency
(90.67 Hz) on the spectrum graph and impact signals whose
frequency is equal to outer race defect frequency (ubp), in the
time domain waveform graph indicate that ball bearing outer
race has a defect but it does not require to take apart the bearing
in this stage because the vibration amplitude is on a normal
level [13]. This situation continued in this way for 4 months. It
Fig. 10. Overall vibration level trend of the fan motor inner bearing. is shown from the next measurement that vibration amplitude
 S. Orhan et al. / NDT&E International 39 (2006) 293–298 297

Table 3
Frequency compositions of the fan motor outer bearing

Frequency (Hz) Amplitude Corresponding frequency

(mm/s-Peak) (!ubp)
66.69 0.25 1
131.75 0.06 2
202.2 0.1 3
271.91 0.1 4
339.2 0.11 5
406.64 0.04 6
473.84 0.14 7
542.01 0.08 8
609.4 0.1 9
677.75 0.11 10
744.41 0.05 11
812.98 0.064 12
880.62 0.039 13
949.26 0.03 14
1016.7 0.04 15
1084.4 0.04 16
Fig. 14. Spectrum graph of the fan motor outer bearing on 3 October 2001.
1152.9 0.057 17
1220.8 0.036 18
vibration measurements, vibration levels some varied due to
1292.3 0.032 19
1356.6 0.02 20 loading conditions. This bearing may be used effectively under
1490.7 0.01 22 the condition that the vibration level is accurately monitored.
1558.8 0.02 23 From this time on, any increase in the vibration level must be
1626.6 0.01 24 carefully evaluated because otherwise the bearing will be brake
1694.9 0.02 25
down at any moment.
1761.9 0.02 26
1829.8 0.02 27
1898.2 0.01 28 6. Case study III
1961.4 0.0045 29
2311.9 0.1 34
2379.6 0.011 35
Ball bearings were widely used in any type of rotating
2583.5 0.01 38 machinery. Thus, determining the health of the ball bearings is
2652.5 0.01 39 very important. Outer bearing vibration of a fan motor, which
2723.2 0.03 40 is supported by SKF 6222 type ball bearing, was monitored
periodically. The power and revolution of the motor are
suddenly increased (Fig. 10), the vibration level was still 200 kW, 985 rpm, respectively. The reference measurement
normal so not any action was initialized but vibration must be was taken on 6 September 2001 (Figs. 11 and 12). By spectral
measured regularly from this time on and any increase in analysis of vibration signatures, it is shown that the bearing
condition was normal. Vibration frequencies on the spectrum
amplitude has accurately to be evaluated. In the next and other
graph consist of ball bearing outer race defect frequency and its

Fig. 13. Time waveform graph of the fan motor outer bearing on 3 October
2001. Fig. 15. Time waveform graph of the fan motor outer bearing on 6 July 2002.
298 S. Orhan et al. / NDT&E International 39 (2006) 293–298

see bearing defect because maintenance staff of the refinery did

not accept a maintenance action.

7. Conclusions

In this study, diagnosing techniques of the ball and

cylindrical roller element bearing defects were investigated
by vibration monitoring and spectral analysis as a predictive
maintenance tool. Ball bearing looseness, a ball bearing outer
race defect and a cylindrical bearing outer race defect were
successfully diagnosed. It was shown that ball and cylindrical
roller bearing defects were progressed in identical manner
Fig. 16. Multiple spectrum graph of the fan motor outer bearing. without depending on rolling element type. Furthermore, it was
experienced that when a bearing defect reaches an advanced
stage, high frequency amplitude levels often decrease due to
‘self-peening’ of the bearing flaws. Remaining life of the
bearings can be estimated using vibrational behaviour and
running time of the bearings. Diagnosing of defects on multiple
parts of the bearing may also be investigated in a real running
condition. Also, it can be concluded that if vibration
monitoring is applied within regular selected periods, capable
instrumentation and if vibration analysis is performed by
experienced personnel, impending failures can be easily
detected.

References

[1] Renwick JT, Babson PE. Vibration analysis-a proven technique as a

predictive maintenance tool. IEEE Trans Ind Appl 1985;IA-21(2):
324–32.
Fig. 17. Overall vibration level trend of the fan motor outer bearing. [2] Alguindigue IE, Buczak LA, Uhrig RE. Monitoring and diagnosis of
rolling element bearings using artificial neural networks. IEEE Trans Ind
harmonics from 1!ubp to 40!ubp as shown in the Table 3. Electron 1993;40(2):209–17.
This situation indicates the sign of a outer race defect but there [3] McFadden PD, Smith JD. Model for the vibration produced by a
single point defect in a rolling element bearing. J Sound Vib 1984;
are no impact signals in the time domain waveform graph. This 96(1a):69–82.
indicates that the defect did not still formed, namely, is in its [4] McFadden PD, Smith JD. The vibration produced by a multiple point
early stage. In the time domain waveform graph (Fig. 13) defect in a rolling element bearing. J Sound Vib 1985;98(2):263–73.
obtained after about one month (3 October 2001) periodical [5] Gohar R, Aktürk N. Vibrations associated with ball bearings. ImechE
impact signals are formed whose period was related to ball 1998;43–64.
[6] Aktürk N, Uneeb M, Gohar R. The effects of number of balls and preload
bearing outer race defect frequency (66.69 Hz) and vibration on vibrations associated with ball bearings. J Tribol 1997;119:747–53.
frequencies in the spectrum graph (Fig. 14) were the same as in [7] Gluzman D. Recognizing impending bearing failure. Reliab Mag 2001;
the previous situation. At this moment, the defect became June:41–7.
evident, the amplitude of the frequencies was on a low level, [8] Al-Najjar B. Accuracy, effectiveness and improvement of vibration-based
thus a maintenance action was no required. From following maintenance in paper mills: case studies. J Sound Vib 2000;229(2):
389–410.
measurements, for example 6 July 2002, impact signals were
[9] Computational Systems Inc., Knoxville,TN,USA.
evident as shown in Fig. 15. In contrast to these facts, it is [10] Mobley RK. An introduction to predictive maintenance. New York: Van
shown that vibration amplitudes had been decreased on 5 June Nostrand Reinhold; 1990.
2002 and in following measurements (Figs. 16 and 17). This [11] Overton, B. Basic vibration primer. DoctorKnow Application Paper,
indicates that bearing entered a catastrophic stage and due to Computational Systems Incorporated. Knoxille, TN, USA
[12] Predictive maintenance through the monitoring and diagnostics of rolling
‘self-peening’ of the bearing flaws, high frequency amplitude
element bearings. Application note. Bently Nevada Corporation. http://
levels often decreased. The occurrence of this ‘self-peening’ www.bently.com/articles/apnotes/an044.asp.
phenomenon is especially true for low speed machines [12]. [13] ISO standard 2372: vibration amplitude acceptance guidelines for rotating
There was no chance to pull out ball bearing, consequently, to machinery operating from 600 to 12000 RPM.

View publication stats