13941511 233-241

mme.modares.ac.ir

: Please cite this article using:

Y. Vaghei, A. Farshidianfar, Fault diagnosis and classification of deep groove ball bearings using wavelet transform and adaptive neuro-fuzzy system, Modares MechanicalEngineering Vol. 15 No. 11, pp. 233-241, 2015 (In Persian)

-

1 2*

1- 2- * 1111-91775farshid@um.ac.ir

:15 1394 :18 1394

:06 1394

. .

.

.

- .

. .

-

Fault diagnosis and classification of deep groove ball bearings usingwavelet transform and adaptive neuro-fuzzy system

Yasaman Vaghei, Anooshiravan Farshidianfar *

Department of Mechanical Engineering, Ferdowsi University of Mashhad, Mashhad, Iran.P.O.B 91775-1111, Mashhad, Iran, farshid@um.ac.ir

ARTICLE INFORMATION ABSTRACTOriginal Research PaperReceived 06 July 2015Accepted 09 September 2015Available Online 28 October 2015

Today, fast and accurate fault detection is one of the major concerns in industry. Although manyadvanced algorithms have been implemented in the past decade for this purpose, they were verycomplicated or did not provide the desired results. Hence, in this paper, we have proposed anemerging method for deep groove ball bearing fault diagnosis and classification. In the first step,the vibration test signals, related to the normal and faulty bearings have been used for both thedrive-end and fan-end bearings of an electrical motor. After that, one dimensional Meyer wavelettransform has been employed for signal processing in the frequency domain. Hence, the uniquecoefficients for each kind of fault were extracted and directed to the adaptive neuro-fuzzy systemfor fault classification. The intelligent adaptive neuro-fuzzy system was adopted to enhance thefault classification performance due to its flexibility and ability in dealing with uncertainty androbustness to noise. This system classifies the input data to the faults in the race or the balls ofeach of the fan-end and the drive-end bearings with specific fault diameters. In the final part ofthis study, the new experimental signals were processed in order to verify the results of theproposed method. The results reveal that this method has more accuracy and better classificationperformance in comparison with other methods proposed in the literature.

Keywords:Fault Diagnosis Vibration Signal Wavelet Transform Adaptive Neuro-Fuzzy System

1-

.

.

-

234 13941511

. -

] 1 .[

1 2 ] 2[

. 3 -

-

- .4 - ] 3 .[ .

. ) ( .

] 6-4 .[

.

. - .-

- 5 .

.

-

.

] 5 .[

] 7.[

-

.

1- Hilbert Transform 2- Fast Fourier Transform (FFT) 3- Window Length 4- Wavelet Transform 5- Takagi-Sygeno

:1 -

6 .-2

.-3 .

:2

.3

4 - .5

6

2-

) ( .

.

1

) 1.(

Fig. 1 Schematic view of the bearings and the related faults

1

Fig. 2 The experimental test setup

2 ] 8[ 2

6- Meyer Wavelet

-

13941511 235

) 1 . ( )

. 0.0070.014 0.021

2 .

( ) . 3

5000 V/g 1 .

4 . 48000

1750 8.[

3- )1( .

(1)( ) = ( )

( ) .

) ) ()2((.

(2) = ( ) ( )

] 9 .[

1 Table 1 Bearings’ specifications

SKF 6205SKF 6203

mm( 25 17 mm( 52 40 mm( 15 12

rpm( 14000 19000

1- Encoder 2- Electro Discharging Machine 3- Case Western University 4- MATLAB

1 - .

2 - )

() 2 (

3 - ( )

.

4 - ( )

2 3 5 - 4-1

.

.

.

.

. ) 3 ( ] 10.[

(3)

( )

=

12

sin2

(3| |

1) / ,43

< | | <83

12

cos2

(3| |4

1) / ,3

< | | <3

0,

) 4 ( 5 )5 (

(4)

u(x) =0, x < 0x, 0 < x < 11,

(5)

( )

=

1, | | <

23

12

cos2

(3| |4

1) ,3

< | | <3

0,

3

5- Meyer Scale Function

-

236 13941511

-

4- - -

0.0070.014 0.021

. --

) (

.4 . ] 11:[

1 -

-

2 -

3 - -

4 -

5 -

6-

) 9-6 (

(6)= ( ), = 1,2, … ,

(7)= , = 1,2, … ,

(8)= , = 1,2, … ,

(9)= , = 1,2, … ,

.

. .

.

- 1 .

1- Nonlinear Aggregation Function

2 ) 10 (

) 11.(

)11(= ( ) 3 - 4

-

.5 2 .

.

.

Fig. 3 The wavelet function (right) and the scale function (left) 3 ( ) ( )

Fig. The adaptive neuro-fuzzy system process 4 -

2- ‘And’ Neurons 3- T-Norm 4- T-Conorm

)10(= + ( 1)

-

13941511 237

Fig. 5 Schematic view of the adaptive neuro-fuzzy system’s sections 5 -

- Table 2 The details of the adaptive neuro-fuzzy system

1

2 3

.-

-

- .

.

. 4

-

6 .

. -

. .

5-

1. Mamdani 2. Subtractive clustering 3. Gradient descend4. Over Fitting

Fig. The algorithm and the total schematic view of the proposedmethod’s process

6 .

7

.

.

) 9,8.(

.

) 11,10.(

89

8 9 5 48000 )240000 =5 × 48000 . (

.

. .

- -

-

238 13941511

Fig. Sample experimental signals

7

Fig. The wavelet transform coefficients for the faults in the race of the bearings, in Drive End (DE) or Fan End (FE) 8 ) ) ) ) FE) (DE (( (.

Fig. The wavelet transform coefficients for the faults in the balls of the bearings, in Drive End (DE) or Fan End (FE)

9 ) ) ) FE) (DE (( (.

.10 11

.

-

13941511 239

Fig. 10 The 3- plots for the continuous wavelet transform for the faults in the race of the bearings with 0.007 (a), 0.014 (b) and 0.021 (c)inch fault diameter

10 a- 0.007 b - 0.014c - 0.021 .

- . -

.108 3

) 3 .(

.

.

3

.

4 .

.

-

240 13941511

Fig. 11 The 3- plots for the continuous wavelet transform for the faults in the balls of the bearings with 0.007 (a), 0.014 (b) and 0.021 (c)

inch fault diameter 11 a- 0.007 b - 0.014 c - 0.021 .

.

6 -

- .

.

-

13941511 241

3 Table 3 The proposed method’s outputs

)

9 100 0.007 9 100 0.007 9 100 0.014 9 100 0.014 9 1000.021 9 1000.021 9 1000.007 9 99.1 0.007 9 1000.014 9 1000.014 9 98.30.021 9 100 0.021 9 100

4 Table 4 The comparison of the proposed method with the

methods in the literature

]12[ 93.054 144

]13[ 77.9853 -(*)

]14[

92.0152 135

]15[ 98.72 -

]16[ 99.99-

10

]17[ 73.014

-

]18[ 99.452- ]19[ 96.7-10 ]20[ 90.5436

]21[ 100 7 63

- 100 12 108 (*)

-

. .

7- [1] Y.G.Lei, J.Lin, Z.J.He, Application of an improved kurtogram method for

fault diagnosis of rolling element bearings, Mechanical Systems andSignal Processing Vol. 25, pp. 1738–1749, 2011.

[2] H. Qiu, J.Lee, J.Lin, G.Yu, Wavelet filter-based weak signature detectionmethod and its application on roller bearing prognostics, Jounal of Soundand Vibration Vol. 289, pp. 1066–1090, 2006.

[3] L. Jedli´nski, J. Jonak, Early fault detection in gearboxes based on supportvector machinesand multilayer perceptron with continuous wavelettransform, Applied Soft Computing Vol. 30, pp. 636–641, 2015.

[4] H. Khaksari, A. Khoshnood, J. Roshanian, Active Noise Cancelation in Reaction Wheel by simultaneous using of dynamical system identification and online wavelet, Modares Mechanical Engineering Vol.15, No.3, pp. 146-152, 2015. (In Persian

[5] H.Ziaiefar, M.Amiryan, M. Ghodsi, F. Honarvar, Y. Hojjat, UltrasonicDamage Classification in Pipes and Plates using Wavelet Transform andSVM, Modares Mechanical Engineering Vol. 15, No.5, pp. 41-48, 2015. (InPersian

[6] S. A. Atashipour, H. R. Mirdamadi, R. Amirfattahi, S. Ziaei-Rad,Application of wavelet transform in damage identification in thicksteel beam based on ultrasonic guided wave propagation, ModaresMechanical Engineering Vol. 12, No.5, pp. 154-164, 2013. (In Persian

[7] Z. Liang, H. Fei, T. Yifei, L. Dongbo, Fault detection and diagnosis of beltweigher using improved DBSCAN and Bayesian regularized neuralnetwork, MECHANIKA Vol. 21, No. 1, pp. 70-77, 2015.

[8] X. Lou, K.A. Loparo, Bearing fault diagnosis based on wavelet transformand fuzzy inference, Journal of Mechanical Systems and Signal ProcessingVol. 18, pp. 1077–1095, 2004.

[9] M. Misti, Y. Misti, G. Oppnheim, J. Poggi, Wavelet Toolbox User’s GuideThe Mathworks Inc., 2002.

[10] I. Daubechies, Ten lectures on wavelets, CBMS-NSF conference series inapplied mathematics SIAM Edition, Vol 137, pp. 117–119, 1992.

[11] W. Pedrycz, F. Gomide Fuzzy Systems Engineering Toward Human-Centric Computing John Wiley Sons Inc., IEEE Press, Hooboken, NewJersey, USA, 2007.

[12] B. Li, M.-Y. Chow, Y. Tipsuwan, J.C. Hung, Neural network based motorrolling bearing fault diagnosis, IEEE Transaction on Industrial ElectronicsVol. 47, No. 5, 2000.

[13] J. Altmann, J. Mathew, Multiple band-pass autoregressive demodulationfor rolling element bearing fault diagnosis, Mechanical Systems andSignal Processing Vol. 15, No. 5, pp. 963–977, 2001.

[14] B. Samanta, K.R. Al-Balushi, S.A. Al-Araimi, Artificial neural networksand support vector machines with genetic algorithm for bearing faultdetection, Engineering Application of Artificial Intelligent Vol. 16, pp.657–665, 2004.

[15] L. Zhang, L.B. Jack, A.K. Nandi, Fault detection using geneticprogramming, Mechanical Systems and Signal Processing Vol. 19, pp271–289, 2005.

[16] M.-Y. Cho, T.-F. Lee, S.-W. Gau, C.-N. Shih, Power Transformer FaultDiagnosis using Support Vector Machines and Artificial Neural Networkswith Clonal Selection Algorithms Optimization Part I, Springer, Berlin,Heidelberg, pp. 179–186, 2006.

[17] V. Purushotham, S. Narayanan, S.A.N. Prasad, Multi-fault diagnosis ofrolling bearing elements using wavelet analysis and hidden Markovmodel based fault recognition, NDT&E International Vol. 38, No. 8, pp.654–664 2005.

[18] A. Rojas, A.K. Nandi, Practical scheme for fast detection and classificationof rolling-element bearing faults using support vector machines,Mechanical Systems and Signal Processing Vol. 20, No. 7, pp. 1523–1536,2006.

[19] A. Saxena, A. Saad, Evolving an artificial neural network classifier forcondition monitoring of rotating mechanical systems Applied SoftComputing Vol. 7, pp. 441–454, 2007.

[20] X. Lou, K.A. Loparo, Bearing fault diagnosis based on wavelet transformand fuzzy inference, Journal of Mechanical Systems and Signal ProcessingVol. 18, pp. 1077–1095, 2004.

[21] S. Abbasion, A. Rafsanjani, A. Farshidianfar, N. Irani, Rolling elementbearings multi-fault classification based on the wavelet denoising andsupport vector machine, Mechanical Systems and Signal Processing Vol.21, pp. 2933–2945, 2007.