Vibration and Acoustic Emission Signal Monitoring for Detection of Induction Motor Bearing Fault

Vibration and Acoustic Emission Signal Monitoring for Detection of Induction Motor Bearing Fault

Mohd Sufian Othman, Mohd Zaki Nuawi Department of Mechanical and Material Engineering Faculty of Engineering & Built Environment Universiti Kebangsaan Malaysia

43600 Bangi, Selangor, Malaysia

Ramizi Mohamed

Department of Electrical, Electronic & Systems Engineering

Faculty of Engineering & Built Environment Universiti Kebangsaan Malaysia 43600 Bangi, Selangor, Malaysia

Abstract Vibration and acoustic emission (AE) signal monitoring are popular techniques for detecting bearing fault, the main cause in induction motor failure which can lead to catastrophic damage. This paper presents comparison between vibration and AE signal monitoring as a tool for induction motor bearing fault detection. The effectiveness of time-domain analysis is compared with frequency-domain. Statistical parameters used in time-domain include RMS, crest factor, and kurtosis whereas for frequency-domain, normal spectrum and envelope spectrum using Hilbert transform are applied. The results reveal that vibration and AE signals are effective measurement to detect bearing fault in both time- and frequency-domain.

Keywords Induction motor bearing; fault detection; condition monitoring.

1. INTRODUCTION

   Induction motor is a prominent rotating machine widely used in many industrial, commercial and domestic applications due to its efficiency, reliability, robustness, and economical. It is reported that this type of motor dominate about 96% of energy consumption for all electric motors, which consume about 40% of total electrical energy usage [1]. Although induction motor is reliable and robust, it still expose to catastrophic failure especially when running under heavy loads and long period of time. Generally, induction motor breakdown can be caused by bearing fault, shaft or coupling fault, rotor fault and stator fault [2]. Surveys revealed that bearing fault is the most typical failure in induction motor components [3], [4].

   Bearing defects can be grouped as distributed or

   localized [5][7]. Distributed defects normally caused by manufacturing error, improper installation or abrasive wear. This group of defect include surface roughness, waviness, misaligned races, and off size ball. Whereas, localized defects

   are include cracks, pits, and spalls on the rolling surfaces

   Condition monitoring is one of popular methods in determining bearing faults. Measurements such as vibration, temperature, acoustic, motor current, and wear debris are often used in bearing fault detection [8]. Among these, vibration signal is the most preferred tool in bearing fault monitoring. Fig. 1 shown typical signal condition monitoring procedure [2], [9][11].

   Data acquisition

   Data processing

   Feature extraction

   Diagnosis system

   Fig. 1. Typical process for condition monitoring.

   1. Time-domain Approach

      Visual inspection of vibration and AE signal in time waveform is a fast method for identifying bearing damage especially for inner-race and outer-race fault. However, this method is not reliable because of noise interruption and difficulties when dealing with low amplitude signals. Statistical analysis is other basic procedure in bearing fault detection, used in past literatures considering parameters such as root mean square (RMS), crest factor, and kurtosis [12], [13].

      RMS explains the effective value (magnitude) of the signal, useful to calculate the average value for variation of positive and negative values in sinusoidal signal. For N numbers of data set (y1,y2,yn), RMS is defined as the square root of the arithmetic mean of the squares of original values (y 2,y 2,y 2) as in (1).

      which might be caused by fatigue failure due to overloading or shock loading of the bearings during operation and installation.

      1 2 n

      RMS

      1 N y 2

      N n

      n 1

      The crest factor can be defined as the ratio of the peak value to the RMS value, it yields the information about the spikiness of the measured signal. It is a pure number without any dimensions. Crest factor formula is calculated as:

      Shaft frequency is relative speed difference between inner and outer race which can be obtained from motor rotation speed (RPM).

      Cage frequency or fundamental train frequency (FTF), fC

      is the rotating speed of the ball cage assembly that can be

      cr max( yn )

      RMS

      calculated as:

      2

      f fS 1- d cos

      Kurtosis is the standardized 4th statistical moment of data which indicates the property of signal, whether the signal is

      C D

      peaked or flat relative to a Gaussian distribution. Kurtosis value is given by (3).

      Ball spin frequency (BSF), fB is circular frequency of each

      N

      N

      4

      1

      ball as it spin which is two times the cage frequency:

      kur

      n1

      yn

      4

      fB d

      fS 1-

      2

      2

      d

      2

      cos

      D

      where is the mean value:

      n

      1 N y

      D

      Ball pass frequency of inner race (BPFI), fIR is the frequency appeared when the ball spin across the defect in the inner race which is depend on number of balls:

      N n1

      f Nb fS 1- d cos

      and is signal standard deviation:

      IR 2 D

      1 N

      N n1

      2

      yn

      Ball pass frequency of outer race (BPFO), fOR is the frequency for outer race defect:

      f Nb fS 1 d cos

   2. Bearing Characteristic Frequencies

      There are five characteristic frequencies exist in rotating bearing as shown in Fig. 2: (1) Shaft frequency, fs, (2) Cage frequency, fc, (3) Ball spin frequency, fb, (4) Inner race frequency, fIR, and (5) Outer race frequency, fOR [14][16].

      d

      fS

      OR 2 D

      In these equations, Nb is the number of balls, d is the roller diameter, D is the pitch diameter of the bearing, and is a contact angle.

      In frequency-domain analysis, Fast-Fourier Transform (FFT) is an established algorithm to convert data from the time-domain. It is difficult to detect bearing fault frequencies in FFT spectrum of vibration signal [17]. However, some researchers have applied successfully [18], [19]. Furthermore, envelope spectrum using Hilbert transform can provide better

      fOR

      fB D

      fIR

      fC

      result to locate fault frequencies and its resonances.

2. EXPERIMENTAL SETUP

   The experimental tests were performed on test rig with a two-speed pole-changing induction motor (See Table 1). The motor is supplied by the 3 phase variable voltage transformer

   Fig. 2. Bearing dimension and characteristic frequencies.

   pre-set at 380V. Tacho generator was connected to the drive- end of motor to measure rotation speed of induction motor (Fig. 3).

   1 2 3

   8

   4 5 6 7

   Fig. 3. Test-rig setup: (1) Computer, (2) Oscilloscope, (3) Control unit, (4) 3 Phase variable voltage supply, (5) Motor protection switch, (6) Squirrel cage induction motor, (7) Tacho generator, and (8) Magnetic powder brake.

   TABLE 1. INDUCTION MOTOR SPECIFICATION

   Description	Value
   Make	Leybold Didactic GmbH
   Mode Number	732 94
   Rated Power	0.7/0.95 kW
   Rated Speed	1405/2825 min-1
   Voltage	3 Phase 380V
   Frequency	50 Hz
   Rated Current	2.0/2.75 A

   (a)

   (b)

   Three bearing condition have been tested, i.e. normal bearing, bearing with inner race fault, and outer-race fault. All faulty bearings have been artificially damaged by axial drilled of 1mm hole through the raceways (see Fig. 4).

   (a)

   (b)

   Fig. 4. Bearing with localized defect of (a) inner race and (b) outer race.

   Tested bearing was placed on the drive-end of the motor. The motor then have been tested at 1450 rpm and 2900 rpm for each bearing conditions. Vibration signal is measured using SKF CMSS 9952 piezoelectric accelerometer with sensitivity of 10mV/g and the DeCI SE1000H acoustic emission transducer used to collect AE signal (Fig. 5). Data are sampled at 12 kHz using Picoscope 7000 oscilloscope.

   Tested bearing location

   Fig. 5. Piezoelectric accelerometer, (a) and acoustic emission transducer,

   (b) axially mounted on the motor casing.

   Characteristic frequencies; fC, fB, fIR, and fOR for the tested bearing are shown in Table 2, calculated based on (6), (7), (8)

   and (9) respectively.

   TABLE 2. TESTED BEARING CHARACTERISTIC FREQUENCIES IN HERTZ

   Frequency Name	1450 rpm	2900 rpm
   Shaft frequency, fS	24.17	48.33
   Fundamental train frequency (FTF), fC	9.21	18.42
   Ball spin frequency (BSF) , fB	96.26	192.51
   Ball pass frequency of inner race (BPFI), fIR	119.55	239.11
   Ball pass frequency of outer race (BPFO), fOR	73.76	147.51

3. RESULTS AND DISCUSSION

   Vibration and AE signal from induction motor running without load at rotational speed of 1450 rpm and 2900 rpm have been analyzed with three methods: (1) visual inspection of time waveform, (2) statistical analysis in time-domain and

   (3) checking for the existence of bearing fault frequencies in normal spectrum and envelope spectrum.

   A. Time-domain Waveform Comparison

   In general, for motor rotational speed of 1450 rpm, AE signal from healthy bearing contained more values compared to acceleration signal [See Fig. 6(a)]. However, waveform for both acceleration and AE signal in 2900 rpm speed displayed in similar pattern [Refer Fig. 6(b)]. Both signals did not showed any repetitive pattern along the timeframe.

   (a)

   (b)

   Fig. 6. Healthy bearing time waveform at (a) 1450 and (b) 2900 rpm.

   Waveform of inner-race faulty bearing for 1450 rpm and 2900 rpm are presented in Fig. 7. Consistent repeating impulse is observed in vibration and AE signals for 1450 rpm, repeated at same slots. Similar pattern also observed in 2900 rpm. Although maximum peak value in vibration waveform is higher than AE signal, the quality of AE signal is better.

   (a)

   (b)

   Fig. 8. Outer-race faulty bearing waveform at (a) 1450 and (b) 2900 rpm.

   B. Statistical Analysis

   Statistical values for vibration and AE signal waveform are calculated using MATLAB 8.3 for time period of 0 to 0.5 seconds, same as in time waveform analysis. Table 3 shows RMS values comparison between healthy and faulty bearing for vibration and AE signals in 1450 and 2900 rpm. RMS values for vibration signal indicate small difference between healthy and defected bearing whereas RMS values for AE signals are close to 0 in all bearing conditions. The highest RMS value is spotted in the outer-race defected bearing running at 2900 rpm.

   TABLE 3. RMS VALUES COMPARISON

   Bearing Condition	Vibration		AE
   	1450 rpm	2950 rpm	1450 rpm	2950 rpm
   Healthy	0.035	0.051	0.003	0.017
   Inner-race defect	0.074	0.112	0.015	0.030
   Outer-race defect	0.119	0.404	0.005	0.031

   (a)

   (b)

   Fig. 7. Inner-race faulty bearing waveform at (a) 1450 and (b) 2900 rpm.

   Fig. 8(a) and Fig. 8(b) show comparison between vibration and AE signal waveform for outer-race defected bearing in 1450 and 2900 rpm respectively. Overall shape for vibration and AE signals are about the same for both rotating speed. Repetitive peak impulse are clearly exists in small period of time compared to inner-race defected bearing.

   Crest factor values for this experimentation data are tabulated in Table 4. There are quite large difference between crest factor values for healthy and faulty bearing for both vibration and AE signal data where the crest factor for normal bearing condition is smaller than inner-race and outer-race faulty bearing in both rotating speed.

   TABLE 4. CREST FACTOR VALUE COMPARISON

   Bearing Condition	Vibration		AE
   	1450 rpm	2950 rpm	1450 rpm	2950 rpm
   Healthy	4.461	6.189	4.906	4.337
   Inner-race defect	8.521	14.830	10.472	12.140
   Outer-race defect	10.571	9.345	14.708	7.836

   The kurtosis values for healthy and defected bearings are presented in Table 5. Generally, kurtosis value for both signal of healthy bearing is smaller than defected bearing which is close to 3, a known value for a normal distribution. The vibration signal provided higher difference between healthy and defected bearings compared to AE signal.

   TABLE 5. KURTOSIS VALUES COMPARISON

   Bearing Condition	Vibration		AE
   	1450 rpm	2950 rpm	1450 rpm	2950 rpm
   Healthy	5.971	4.060	3.088	4.229
   Inner-race defect	70.042	48.413	19.295	16.994
   Outer-race defect	48.695	21.370	30.750	13.350

   1. Envelope Spectrum Analysis

      Normal spectrum and envelope spectrum analysis have been performed using normal FFT and combination of Hilbert transform with FFT respectively. Fig. 9 to Fig. 11 displayed normal spectrum whereas Fig. 12 to Fig. 14 shown envelope spectrum.

      In normal FFT plot, AE spectrum is more stable than vibration spectrum. The shaft frequency, fS is appeared only in AE spectrums at rotational speed of 2900 rpm [Refer Fig. 9(b), Fig. 10(b), and Fig. 11(b)]. Furthermore, only fault frequency of inner-race, fIR is detected (Fig. 10).

      After applying Hilbert transform to vibration and AE signals, the presented envelope spectrum shown more promising results. All fault frequencies appeared in both vibration and AE spectrums (Fig. 13 and Fig. 14) with clearer notification of outer-race fault frequency compared to inner- race.

      Based on time-domain and frequency-domain analysis, the results reveal hat vibration and AE signal monitoring are suitable procedures for detecting inner-race and outer-race bearing fault. In time waveform visual inspection, AE signal provide better information of fault occurrence. However, vibration signal provide clearer statistical parameters of RMS, crest factor, and kurtosis value to differentiate between healthy and faulty bearing. In overall comparison, RMS is the worst feature in detecting bearing fault compared to crest factor and kurtosis.

      For frequency-domain analysis, normal spectrum could not locate fault frequencies except for inner-race defect. The envelope spectrum analysis using Hilbert transform is the best method suitable for detection of bearing fault frequencies in both vibration and AE envelope spectrum.

      1. fIR + fC

         fS

   fS 2fS

   3fS

   4fS

   Fig. 10. Inner-race faulty bearing normal spectrum at (a) 1450 rpm and (b)

   2900 rpm. (a)

   Fig. 9. Healthy bearing normal spectrum at (a) 1450 and (b) 2900 rpm. (a)

   (b)

   fS

   fIR

   (b) Fig. 11. Outer-race faulty bearing normal spectrum at (a) 1450 rpm and (b) 2900 rpm

   (a)

   2fS

   fIR + fC

   fIR + fC

   3fS

   (b)

   2fS

   fS

   3fS

   4fS

   Fig. 13. Inner-race faulty bearing envelope spectrum at (a) 1450 and (b)

   2900 rpm.

   (a)

   fOR

   fOR

   Fig. 12. Healthy bearing envelope spectrum at (a) 1450 and (b) 2900 rpm. (a)

   fIR

   fOR

   (b)

   fOR

   fIR

   (b)

   Fig. 14. Outer-race faulty bearing envelope spectrum at (a) 1450 and (b)

   2900 rpm.

4. CONCLUSION

Visual inspection of acceleration and AE signal waveform, statistical analysis and envelope spectrum monitoring are acceptable methods in detecting bearing fault. However, this procedures are time consuming and need relevant knowledge to analyze the results. Therefore, an automatic diagnosis system is essential and will be considered for future research.

REFERENCES

1. A. T. de Almeida, F. J. T. E. Ferreira, J. Fong, and P. Fonseca, EUP Lot 11 Motors, Coimbra, 2008.

2. T. Han, B.-S. Yang, and Z.-J. Yin, Feature-based fault diagnosis system of induction motors using vibration signal, Journal of Quality in Maintenance Engineering, vol. 13, no. 2. pp. 163175, 2007.

3. Motor Reliability Working Group, Report of large motor reliability survey of industrial and commercial installations, Part 1, 1985.

4. A. Bonnett and C. Yung, Increased efficiency versus increased reliability, Ind. Appl. Mag. IEEE, vol. 1992, pp. 2936, 2008.

5. N. Tandon and A. Choudhury, A review of vibration and acoustic measurement methods for the detection of defects in rolling element bearings, Tribol. Int., vol. 32, no. 1999, pp. 469480, 2000.

6. L. Batista, B. Badri, R. Sabourin, and M. Thomas, A classifier fusion system for bearing fault diagnosis, Expert Syst. Appl., vol. 40, no. 17, pp. 67886797, Dec. 2013.

7. M. D. Prieto, J. C. i Roura, J. L. R. Martínez, J. Cusidó, J. Luis, and R. Martínez, Bearings fault detection using inference tools, in Vibration Analysis and Control – New Trends and Developments, F. B.- Carbajal, Ed. InTech, 2011, pp. 263280.

8. P. P. Kharche, Review of Fault Detection in Rolling Element Bearing, Int. J. Innov. Res. Adv. Eng., vol. 1, no. 5, pp. 169174, 2014.

9. V. Sugumaran and K. I. Ramachandran, Fault diagnosis of roller bearing using fuzzy classifier and histogram features with focus on automatic rule learning, Expert Syst. Appl., vol. 38, no. 5, pp. 4901 4907, 2011.

10. H.-E. Kim, A. C. C. Tan, J. Mathew, and B.-K. Choi, Bearing fault prognosis based on health state probability estimation, Expert Syst. Appl., vol. 39, no. 5, pp. 52005213, Apr. 2012.

11. B. Li, P. Zhang, Z. Wang, S. Mi, and D. Liu, A weighted multi-scale morphological gradient filter for rolling element bearing fault detection., ISA Trans., vol. 50, no. 4, pp. 599608, Oct. 2011.

12. Z. Kiral and H. Karagulle, Simulation and analysis of vibration signals generated by rolling element bearing with defects, Tribol. Int., vol. 36, pp. 667678, 2003.

13. Y. Lei, Z. He, and Y. Zi, Application of an intelligent classification method to mechanical fault diagnosis, Expert Syst. Appl., vol. 36, no. 6, pp. 99419948, Aug. 2009.

14. M. J. Devaney and L. Eren, Detecting motor bearing faults, IEEE Instrumentation Measurement Magazine, vol. 7, no. 4, IEEE, pp. 30 50, 2005.

15. J. Dybaa and R. Zimroz, Rolling bearing diagnosing method based on empirical mode decomposition of machine vibration signal, Appl. Acoust., vol. 77, pp. 195203, 2014.

16. S. Nandi, H. A. Toliyat, and X. Li, Condition monitoring and fault diagnosis of electrical motors – a review, IEEE Trans. Energy Convers., vol. 20, no. 4, pp. 719729, 2005.

17. J. Zarei, Induction motors bearing fault detection using pattern recognition techniques, Expert Syst. Appl., vol. 39, no. 1, pp. 6873, 2012.

18. A. V Dube, L. S. Dhamande, and P. G. Kulkarni, Vibration based condition assessment of rollingelement bearings with localized defects, Int. J. Sci. Technol. Res., vol. 2, no. 4, pp. 149155, 2013.

19. K. Raghavendra and B. N. Karabasanagouda, Frequency response analysis of deep groove ball bearing, Int. J. Sci. Res., vol. 3, no. 8, pp. 19201926, 2014.