- Email: [email protected]
Early fault diagnosis of bearing and stator faults of the single-phase induction motor using acoustic signals
Accepted Manuscript Early fault diagnosis of bearing and stator faults of the single-phase induction motor using acoustic signals Adam Glowacz, Witold Glowacz, Zygfryd Glowacz, Jaroslaw Kozik PII: DOI: Reference:
S0263-2241(17)30543-2 http://dx.doi.org/10.1016/j.measurement.2017.08.036 MEASUR 4934
To appear in:
Measurement
Received Date: Revised Date: Accepted Date:
3 May 2017 23 August 2017 25 August 2017
Please cite this article as: A. Glowacz, W. Glowacz, Z. Glowacz, J. Kozik, Early fault diagnosis of bearing and stator faults of the single-phase induction motor using acoustic signals, Measurement (2017), doi: http://dx.doi.org/ 10.1016/j.measurement.2017.08.036
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Early fault diagnosis of bearing and stator faults of the singlephase induction motor using acoustic signals Adam Glowacz1, Witold Glowacz1, Zygfryd Glowacz2, Jaroslaw Kozik2 1
AGH University of Science and Technology, Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering, Department of Automatics and Biomedical Engineering, Al. A. Mickiewicza 30, 30-059 Kraków, Poland, [email protected], [email protected] 2
AGH University of Science and Technology, Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering, Department of Power Electronics and Energy Control Systems, Al. A. Mickiewicza 30, 30-059 Kraków, Poland, [email protected], [email protected]
Abstract: An article describes an early fault diagnostic technique based on acoustic signals. The presented technique was used for a single-phase induction motor. The authors measured and analysed following states of the motor: healthy single-phase induction motor, single-phase induction motor with faulty bearing, single-phase induction motor with faulty bearing and shorted coils of auxiliary winding. A feature extraction method called MSAF-20-MULTIEXPANDED (Method of Selection of Amplitudes of Frequency - Multiexpanded) was discussed. The MSAF-20-MULTIEXPANDED was used to create feature vectors. The obtained vectors were classified by NN (Nearest Neighbour classifier), NM (Nearest Mean classifier) and GMM (Gaussian Mixture Models). The proposed technique can be used for diagnosis of the single-phase induction motors. It can be also used for other types of rotating electric motors. Keywords: Fault diagnosis, acoustic signal, bearing, motor, classification 1. Introduction Rotating electric motors are often used in industry, for example in oil refinery, pump oil, steel mill, mine, compressor [1]. Induction motors are widely used electric motors in industry. It is a motivation to analyse such machines. The single-phase induction motor (Fig. 1) is one of the types of induction motor. Diagnostics of rotating electric motors is a normal process of maintenance. A degradation of electric rotating motors depends on environment (heat, moisture) and operation time. Accidents, financial loss, unscheduled downtimes can be predicted based on an early diagnostics of motors. A fault state is a state, which causes adverse effects from the point of view of the correctness of its operation. The early fault state is is the state, in which there are symptoms of characteristic phenomena of the fault state (scratches, short circuits, broken coils, broken bars). In recent years monitoring of machines was developed by engineers [2-3] and companies (Siemens, Dreisilker etc.). Online monitoring of machines also allows for intelligent maintenance with the optimized use of maintenance resources. In the literature the following types of faults of motors were mentioned: stator faults (stator open phase faults, short circuits of windings (Fig. 2), increased resistance of connections), rotor faults (rotor open phase, short circuits of windings, broken bars, faulty ring of squirrel-cage, increased resistance of connections, shaft misalignment, faulty bearings - Fig. 3, 4, 5, rotor eccentricity, bent shaft). Descriptions of diagnostic techniques of machines can be found in the recent literature [4-9]. Diagnostic techniques of bearings are also described [10-14]. The diagnostic techniques are based on various
1
signals such as: electric current [15-18], vibrations [19-26], acoustic signals [27-35], thermal images [36-41], magnetic field. A hybrid EEMD-based SampEn for acoustic signal processing and fault diagnosis was analysed [27]. Diagnosis of stator faults of the single-phase induction motor using acoustic signals was presented [28]. Fault diagnostics of acoustic signals of loaded synchronous motor was also described [29]. Acoustic emission-based condition monitoring was described in the article [30]. Automatic bearing fault localization using vibration and acoustic signals was analysed in the literature [31]. An approach for surface roughness diagnosis in hard turning using acoustic signals was discussed [32]. Identification and monitoring of noise sources of CNC machine tools by acoustic holography methods were also presented [33]. Extraction of fault component from abnormal sound in diesel engines using acoustic signals was described [34]. Analysis of acoustic emission signal for bearing fault was also presented [35]. In the literature there were state of the art methods based on acoustic signals [42-45]. In the paper [42] scientists analysed three-phase induction motors (WEG 00136APE48T). The audio signal was acquired using a condenser microphone JST model CX-509 (cardioid polarization). The scientists used CEEMD (complete ensemble empirical mode decomposition) and TFDG (time-frequency distribution of Gabor). The proposed methodology can be used for fault diagnosis of the motors. However it was not possible to know a priori the number of modes in which the signal was decomposed. Moreover the localization of fault frequencies was related with speed of the motor. Another method based on acoustic signal was described in the article [43]. In this article scientists analysed acoustic signals of squirrel cage induction motors recorded by using 5 microphones simultaneously. The scientists settled 5 different microphones on the motor in hemispherical shape. The analysed motors were 3-phase and 2-pole squirrel cage induction motors rated at 2.2 kW and 380 V. The audio data were digitized with a sampling frequency of 44.1 kHz. Feature extraction was realised by using two different methods: cross correlation, wavelet transformation. To classify data the scientists used Kohonen self-organizing map. The results of analyses were good, however the acoustical approach was sensitive to the interferences of the environmental noises. In the literature there were also papers about applications of using acoustic diagnostics of an internal combustion engine with autoignition [44, 45]. The scientists analysed model of Fiat 1.3 JTD. 12 measurements were carried out for 6 different cases of the engine (engine speeds 1000 rpm, 2000 rpm). The scientists used decision trees and graphs to classify audio data [44, 45]. The proposed decision method of identifying engine failure based on sound emission. It allowed the quick identification of specific damage. However the method may be successfully used for similar engines.
Fig. 1. The healthy single-phase induction motors
2
Fig. 2. The single-phase induction motor with faulty bearing and shorted coils of auxiliary winding
Fig. 3. The single-phase induction motor with faulty bearing
Fig. 4. a) Healthy bearing of the single-phase induction motor, b) Faulty bearing of the single-phase induction motor,
3
Fig. 5. Bearing of the single-phase induction motor
The recognition of acoustic signals of the motor is a difficult task. The first problem is set of microphone. There are many possibilities to do it. The authors put microphone 0.3 m from the rotor of the motor. The second problem is different parameters of the motor such as: size, mass of the motor, rotor speed, nominal current, nominal voltage, power. A lot of samples of acoustic signals from many rotating machines are required for proper diagnostic. Proper diagnostic also requires cooperation with operators of motors and industry. It should be defined what faults and failures are the most essential for diagnosis. The capacitor microphone is less efficient than vibration sensor and thermal camera. An acoustic signal is more interfered than vibration signal and thermal image. It is very difficult to extract characteristic features from the acoustic signal. If we use other features (easier to recognise) we will get better results. Article describes the early fault diagnostic technique based on acoustic signals. The presented technique is used for the single-phase induction motor. Next sections describe proposed technique and analysis. 2. Proposed technique based on acoustic signals The proposed technique was based on acoustic signals of the single-phase induction motor. The technique used various signal processing methods such as: preprocessing, feature extraction and classification. The first step of recognition was data acquisition. A personal computer, ZALMAN ZMMIC1 (Microphone Operation Mode mono, Connectivity Technology Wired, Sensitivity 40 dB, Impedance 2.2 ㏀, Standard power supply 2.0 V DC, S/N 58 dB, Directivity Omnidirectional) and OLYMPUS TP-7 (Microphone Operation Mode mono, Connectivity Technology Wired, Sensitivity 34 dB, Impedance 2.2 Ohm, Frequency Response 50 Hz, Audio Input Details Mono 50 - 20000 Hz Output Impedance 2.2 ㏀) microphones were used for this purpose (Fig. 6). Other types of microphones can be also used. Parameters of soundtrack were following: 44.1 kHz - sampling frequency, 16-bit - bit depth, single channel - number of channels, WAVE PCM - sound file format. Obtained audio data were split into smaller blocks of data. Next these data were normalized. Normalized data were processed by windowing (window size 32768 - 0.74 s, for f=44100 Hz) and the FFT. Obtained amplitudes of frequency (frequency components) were processed by the feature extraction method MSAF-20MULTIEXPANDED. The feature extraction method MSAF-20-MULTIEXPANDED created the feature vectors. The last step of proposed techniques was the classification. The classification step was divided into pattern creation process and test process.
4
Fig. 6. Experimental setup
The pattern creation process used a training set. The test process used a test set. Obtained test feature vectors were classified by Nearest Neighbour classifier, Nearest Mean classifier and Gaussian Mixture Models (Fig. 7).
Fig. 7. Block diagram of the early fault diagnostic technique based on acoustic signals with the use of the MSAF-20-MULTIEXPANDED, Nearest Neighbour classifier, Nearest Mean classifier and Gaussian Mixture Models
2.1. Method of Selection of Amplitudes of Frequency-20-Multiexpanded The feature extraction method - Method of Selection of Amplitudes of Frequency-20-Multiexpanded (MSAF-20-MULTIEXPANDED) was based on processing of the FFT spectra of acoustic signals of the single-phase induction motor. This method can extract characteristic features from the acoustic signals properly. It analysed differences between frequency spectra of states of the single-phase induction motor. It was presented as a block diagram (Fig. 8).
5
Fig. 8. The block diagram of the MSAF-20-MULTIEXPANDED
The steps of the MSAF-20-MULTIEXPANDED were following: 1) Calculate the frequency spectra of acoustic signals of the single-phase induction motor (for example we can use 4 training samples of state S1 and 4 training samples of state S2 and 4 training samples of state S3). The frequency spectrum of acoustic signal of the healthy single-phase induction motor was expressed by vector: fsh=[fsh1, fsh2, ..., fsh16384]. The frequency spectrum of acoustic signal of the single-phase induction motor with faulty bearing was denoted by vector: fsfb=[fsfb1, fsfb2, ..., fsfb16384]. The frequency spectrum of acoustic signal of the single-phase induction motor with faulty bearing and shorted coils of auxiliary winding was expressed by vector: fsfbsc=[fsfbsc1, fsfbsc2, ..., fsfbsc16384]. 2) Calculate differences between frequency spectra of states of the single-phase induction motor. The authors analysed 3 acoustic signals so differences were following: fsh - fsfb, fsh - fsfbsc, fsfb fsfbsc. 3) Calculate absolute values of the obtained differences |fsh - fsfb|, |fsh - fsfbsc|, |fsfb - fsfbsc|. 4) Select 20 maximum amplitudes of frequency for each difference between states of the single-phase induction motor, for example max1(|fsh - fsfb|), max2(|fsh - fsfb|), ..., max20(|fsh - fsfb|). 5) Set a parameter CF-MULTI. The parameter CF-MULTI was expressed as: CF-MULTI=(number of required common frequencies of analysed training sets)/(number of all frequencies of analysed training sets). The parameter CF-MULTI was used for selection of common frequencies. For example, there was 4 training sets each of them had 1 training sample of state S1, 1 training sample of state S2, 1 training sample of state S3 (12 training samples). 12 differences of frequency spectra were calculated, 3 for each training set. The parameter CF-MULTI was equal to 0.65, and 0.65<8/12=0.6666, then 8 of 12 frequencies were required to take this frequency for further processing (9, 10, 11, 12>8 frequencies were also good). When the parameter CF-MULTI was equal to 0.65, we needed 8 of 12 the same frequencies. What would happen to the other parameters CF-MULTI? For example the parameter CF-MULTI was equal to 0.82, then 10 frequencies were selected because 0.82<10/12=0.83333. When the parameter CF-MULTI was equal to 0.57, then 7 frequencies were selected because 0.57<7/12=0.58333. 6) Form the feature vector.
6
The differences of frequency spectra of acoustic signals of the single-phase induction motor |fsh - fsfb|, |fsh - fsfbsc|, |fsfb - fsfbsc| were shown below (Fig. 9-11). The rotor speed was equal to 1390 rpm. The next step was selection of common frequencies.
Fig. 9. The difference between spectra of frequency of acoustic signal of the healthy single-phase induction motor and the single-phase induction motor with faulty bearing (|fsh - fsfb|)
Fig. 10. The difference between spectra of frequency of acoustic signal of the healthy single-phase induction motor and the single-phase induction motor with faulty bearing and shorted coils of auxiliary winding (|fsh - fsfbsc|)
7
Fig. 11. The difference between spectra of frequency of acoustic signal of the single-phase induction motor with faulty bearing and the
single-phase induction motor with faulty bearing and shorted coils of auxiliary winding (|fsfb - fsfbsc|)
The frequency spectrum of acoustic signal of the single-phase induction motor with faulty bearing was presented in figure 12. The MSAF-20-MULTIEXPANDED selected following frequencies 101, 201, 39 Hz (CF-MULTI=0.58), 101, 201 Hz (CF-MULTI=0.66) and 101 Hz (CF-MULTI=0.83). The selected frequencies (39, 101, 201 Hz) of acoustic signal of the single-phase induction motor with faulty bearing were depicted in figure 13. The selected frequencies created feature vectors (vector of 3 elements, vector of 2 elements, vector of 1 element depending on the parameter CF-MULTI ).
Fig. 12. The frequency spectrum of acoustic signal of the single-phase induction motor with faulty bearing
8
Fig. 13. The selected frequencies (39, 101, 201 Hz) of acoustic signal of the single-phase induction motor with faulty bearing
Next the obtained feature vectors were used by Nearest Neighbour classifier, Nearest Mean classifier and GMM (Gaussian Mixture Models). In the paper the authors analysed 3 selected classifiers. However, scientists developed a lot of classifiers in the literature: fuzzy logic [46-49], LVQ (Learning Vector Quantization), neural networks [50-56], self-organizing map, k-means clustering, Naive Bayes classifier [57], SVM (Support Vector Machine) [56, 57], LDA (Linear Discriminant Analysis) [58, 59], classification trees [60], rough sets [61], genetic algorithms. 2.2. Nearest Neighbour classifier Among the various methods of data classification, the Nearest Neighbour classifier achieved high recognition results. This classifier did not have a priori assumptions about the distributions of training and test sets. The Nearest Neighbour classifier involved a training set of both healthy motor and faulty motor cases. A new test feature vector was classified by calculating the distance to the nearest training feature vector. The classifier used distance function such as Manhattan distance, Euclidean distance etc. to calculate distances between test and training feature vectors. The authors used the Manhattan distance, because it was very fast distance function (time complexity O(NlogN)). The Manhattan distance was expressed by formula (1): 3
d (test_fsh, training_fsfb) ∑| (test _ fshi - training _ fsfbi ) | ,
(1)
i 1
where feature vectors test_fsh=[test_fsh29, test_fsh75, ..., test_fsh149] and training_fsfb=[training_fsfb29, training_fsfb75, ..., training_fsfb149]. The analysis were conducted for test feature vectors: test_fsh, test_fsfb, test_fsfbsc and training feature vectors: training_fsh, training_fsfb, training_fsfbsc. However the distance functions such as Chebyshev, Euclidean, Jacquard, cosine distances were also proper for recognition of acoustic signals. More information about the Nearest Neighbour classifier is described in recent literature [62-66]. 2.3. Nearest Mean classifier The Nearest Mean classifier was similar to the Nearest Neighbour classifier. The Nearest Mean used mean vectors instead of training feature vectors. Other steps of the Nearest Mean classifier were the same as for the Nearest Neighbour classifier. Mean vector mean_fsh was expressed by the formula (2):
9
mean_fsh= (training1_fsh + training2_fsh + training3_fsh + training4_fsh)/4
(2)
where feature vector training1_fsh=[training1_fsh29, training1_fsh75, ..., training1_fsh149]. The analysis were conducted for mean vectors: mean_fsh, mean_fsfb, mean_fsfbsc. The Manhattan distance was used for the Nearest Mean classifier. 2.4. Gaussian Mixture Models The Gaussian Mixture Models (GMM) was one of the most popular classification method. These models were used for clustering purpose. It was based on linear combination of finite number of Gaussian distributions. The GMM assessed the partitions of the feature vectors by considering that each component represented a cluster. The clusters had an independent Gaussian distribution, each with their own covariance matrix and mean. The classifier calculated the probability of belonging to each cluster. The GMM used an iterative technique called Expectation Maximization. The classification was achieved by assigning each feature vector to the most likely cluster. The GMM calculated weight of each Gaussian component, covariance matrix and mean. The probability density function of the GMM was expressed by following formula (3): z
f ( x) ∑| wi N ( x; , i ) | ,
(3)
i 1
where z - the number of the Gaussian components, N ( x, , i ) - the probability density function of normal distribution, wi - the weight of each Gaussian component, such that (4):
z
i 1
wi 1, andi : wi 0 ,
(4)
where , i , i=1,2,...,z - parameters of the Gaussian distributions. The GMM used the score function of the model given by the maximum overall probability for the given class. It was defined by following formula (5):
m arg max score(test_fsh, p) , 1 n g
(5)
where test_fsh - test vector, g - the number of output classes and score(test_fsh, p) - the probability value of the GMM classifier for the models trained for the current p-th class [32, 67]. More information about the GMM can be found in the following literature [32, 67]. 3. Analysis of fault diagnostic technique The authors measured and analysed following 3 states of the motor: healthy single-phase induction motor (Fig. 1, 14), single-phase induction motor with faulty bearing (Fig. 3, 15), single-phase induction motor with faulty bearing and shorted coils of auxiliary winding (Fig. 2, 15, 16). Measurements were performed under laboratory conditions. Each induction motor had following parameters: IN=1.06 A, VN=230 V, PN = 0.12 kW, fN= 50 Hz, Ihm = 0.5 A, Imsc = 1.8 A, s = 1390 rpm, M=3.3 kg,
10
where IN - nominal current of the motor, VN - nominal voltage of the motor, PN - nominal power of the motor, fN - nominal frequency of the motor, Ihm - current of the healthy single-phase induction motor, Imsc - current of the single-phase induction motor with shorted coils of auxiliary winding, s - rotor speed, M - mass of the motor.
Fig. 14. Stator windings of the healthy single-phase induction motor
Fig. 15. Faulty bearing of the single-phase induction motor,
Fig. 16. Stator windings of the single-phase induction motor with shorted coils of auxiliary winding
A new induction motor should be similar to the analysed single-phase induction motor (similar operational parameters, size, materials, rotor speed). The authors used 108 one-second test samples of acoustic signals for the test process and 12 one-second training samples for the pattern creation process. The classifiers (NN, NM, GMM) used processed samples of acoustic signals - feature vectors. The
11
feature vectors used frequencies 101, 201, 39 Hz (CF-MULTI=0.58), 101, 201 Hz (CF-MULTI=0.66) and 101 Hz (CF-MULTI=0.83 - see section 2.1) . Efficiency of acoustic signal recognition was defined as (6): E AS
TS 100% AS
(6)
where: EAS – efficiency of acoustic signal recognition, TS – number of test samples recognised properly, AS – number of all test samples. Total efficiency of acoustic signal recognition was expressed as (7): ET
E ASH E ASFB E ASFBSH 3
(7)
Where ET - Total efficiency of acoustic signal recognition, EASH - efficiency of acoustic signal recognition of the healthy single-phase induction motor, EASFB - efficiency of acoustic signal recognition of the single-phase induction motor with faulty bearing, EASFBSH - efficiency of acoustic signal recognition of the single-phase induction motor with faulty bearing and shorted coils of auxiliary winding. The results of acoustic signal recognition of the single-phase induction motor using the MSAF-20MULTIEXPANDED were presented in Tables 1-3. Tab. 1. The results of acoustic signal recognition of the single-phase induction motor using the MSAF-20-MULTIEXPANDED, and the Nearest Neighbour classifier
Type of acoustic signal
healthy single-phase induction motor single-phase induction motor with faulty bearing single-phase induction motor with faulty bearing and shorted coils of auxiliary winding ET [%]
EAS [%] Frequency [Hz] 101 101, 101, 201 201, 39 100 100 100 88.8
100
97.2
63.8
75
58.3
84.1
91.6
85.2
Tab. 2. The results of acoustic signal recognition of the single-phase induction motor using the MSAF-20-MULTIEXPANDED, and the Nearest Mean classifier
Type of acoustic signal
healthy single-phase
EAS [%] Frequency [Hz] 101 101, 101, 201 201, 39 100 100 100
12
induction motor single-phase induction motor with faulty bearing single-phase induction motor with faulty bearing and shorted coils of auxiliary winding ET [%]
83.3
94.4
94.4
86
91.6
86.1
89.7
95.3
93.5
Tab. 3. The results of acoustic signal recognition of the single-phase induction motor using the MSAF-20-MULTIEXPANDED, and the Gaussian Mixture Models
Type of acoustic signal
healthy single-phase induction motor single-phase induction motor with faulty bearing single-phase induction motor with faulty bearing and shorted coils of auxiliary winding ET [%]
EAS [%] Frequency [Hz] 101 101, 101, 201 201, 39 100 100 100 66.6
94.4
100
100
2.7
2.7
88.8
65.7
67.5
The results of all selected classifiers were following: Nearest Neighbour classifier (ET = 84.1-91.6%), Nearest Mean classifier (ET = 89.7-95.3%), Gaussian Mixture Models (ET = 65.7-88.8%). 4. Conclusions The article presents the early fault diagnostic technique based on acoustic signals. The proposed technique was used for the single-phase induction motor. The following states of the motor were analysed: healthy single-phase induction motor, single-phase induction motor with faulty bearing, single-phase induction motor with faulty bearing and shorted coils of auxiliary winding. The authors used feature extraction method called the MSAF-20-MULTIEXPANDED. In this article 3 classifiers were analysed: Nearest Neighbour classifier, Nearest Mean classifier and Gaussian Mixture Models. The results of all selected classifiers were following: Nearest Neighbour classifier (ET = 84.1-91.6%), Nearest Mean classifier (ET = 89.7-95.3%), Gaussian Mixture Models (ET = 65.7-88.8%). The fault diagnostic technique can be used for protection of single-phase induction motors. Other types of rotating machines can be also diagnosed by the described technique. The proposed technique is noninvasive and inexpensive. However the proposed technique was sensitive to the interferences of the environmental noises. The proposed diagnostic technique can be extended by thermal signals and vibrations. In the future the authors will analyse other faults and electric motors with various operational parameters. Next more reliable diagnostic technique will be developed and used by manufacturers of motors.
13
Acknowledgments This research has been financed by AGH University of Science and Technology, grant no. 11.11.120.612 (Adam Glowacz), grant no. 11.11.120.815 (Witold Glowacz), grant no. 11.11.120.354 (Zygfryd Glowacz, Jaroslaw Kozik). We thank reviewers for their valuable suggestions. References [1] H. Henao, GA. Capolino, M. Fernandez-Cabanas, F. Filippetti, C. Bruzzese, E. Strangas, R. Pusca, J. Estima, M. Riera-Guasp, S. Hedayati-Kia, Trends in Fault Diagnosis for Electrical Machines A Review of Diagnostic Techniques, IEEE Industrial Electronics Magazine 8 (2) (2014) 31-42. http://dx.doi.org/10.1109/ MIE.2013.2287651 [2] M. Irfan, N. Saad, R. Ibrahim, VS. Asirvadam, An on-line condition monitoring system for induction motors via instantaneous power analysis, Journal of Mechanical Science and Technology 29 (4) (2015) 1483-1492. http://dx.doi.org/10.1007/s12206-015-0321-9 [3] R. Islam, SA. Khan, JM. Kim, Discriminant Feature Distribution Analysis-Based Hybrid Feature Selection for Online Bearing Fault Diagnosis in Induction Motors, Journal of Sensors (2016), Article Number: 7145715. http://dx.doi.org/10.1155/2016/7145715 [4] Y. Jiang, ZX. Li, C. Zhang, C. Hu, Z. Peng, On the bi-dimensional variational decomposition applied to nonstationary vibration signals for rolling bearing crack detection in coal cutters, Measurement Science and Technology 27 (6) (2016), Article Number: 065103. http://dx.doi.org/10.1088/0957-0233/27/6/065103 [5] D. Lopez-Perez, J. Antonino-Daviu, Application of Infrared Thermography to Failure Detection in Industrial Induction Motors: Case Stories, IEEE Transactions on Industry Applications 53 (3) (2017) 1901-1908. http://dx.doi.org/10.1109/TIA.2017.2655008 [6] D. Mika, J. Jozwik, Normative measurements of noise at CNC machines work stations, Advances in Science and Technology-Research Journal 10 (30) (2016) 138-143. http://dx.doi.org/10.12913/22998624/63387 [7] ZX. Li, Y. Jiang, C. Hu, Z. Peng, Recent progress on decoupling diagnosis of hybrid failures in gear transmission systems using vibration sensor signal: A review, Measurement 90 (2016) 4–19. http://dx.doi.org/10.1016/j.measurement.2016.04.036 [8] GM. Krolczyk, JB. Krolczyk, S. Legutko, A. Hunjet, Effect of the disc processing technology on the vibration level of the chipper during operations, Tehnicki Vjesnik-Technical Gazette 21 (2) (2014) 447-450. [9] L. Jedlinski, J. Caban, L. Krzywonos, S. Wierzbicki, F. Brumercik, Application of vibration signal in the diagnosis of IC engine valve clearance, Journal of Vibroengineering 17 (1) (2015) 175-187. [10] G. Perun, Z. Stanik, Evaluation of state of rolling bearings mounted in vehicles with use of vibration signals, Archives of Metallurgy and Materials 60 (3) (2015) 1679-1683. http://dx.doi.org/10.1515/amm-2015-0291 [11] Z. Stanik, G. Perun, T. Matyja, Effective methods for the diagnosis of vehicles rolling bearings wear and damages, Archives of Metallurgy and Materials 60 (3) (2015) 1717-1724. http://dx.doi.org/10.1515/amm2015-0296 [12] B. Van Hecke, J. Yoon, D. He, Low speed bearing fault diagnosis using acoustic emission sensors, Applied Acoustics 105 (2016) 35–44. http://dx.doi.org/10.1016/j.apacoust.2015.10.028 [13] F. Hemmati, W. Orfali, MS. Gadala, Roller bearing acoustic signature extraction by wavelet packet transform, applications in fault detection and size estimation, Applied Acoustics 104 (2016), 101-118. http://dx.doi.org/10.1016/j.apacoust.2015.11.003 [14] P. Rzeszucinski, M. Orman, CT. Pinto, A. Tkaczyk, M. Sulowicz, A signal processing approach to bearing fault detection with the use of a mobile phone, 2015 IEEE 10th International Symposium on Diagnostics For Electric Machines, Power Electronics and Drives (Sdemped) (2015) 310-315. [15] Z. Glowacz, A. Glowacz, Simulation language for analysis of discrete-continuous electrical systems (SESL2), Proceedings of the 26th IASTED International Conference on Modelling, Identification, and Control (2007) 94-99. [16] M. Michalak, B. Sikora, J. Sobczyk, Diagnostic model for longwall conveyor engines, Man-Machine Interactions 4, ICMMI, Book Series: Advances in Intelligent Systems and Computing, 391 (2016) 437-448. http://dx.doi.org/10.1007/978-3-319-23437-3_37
14
[17] C. Da Costa, M. Kashiwagi, MH. Mathias, Rotor Failure Diagnosis of Induction Motors by Wavelet Transform and Fourier Transform in Function of the Load, International Conference on Computer Science and Artificial Intelligence (2015) 109-113. [18] M. Gutten, D. Korenciak, M. Kucera, M. Sebok, M. Opielak, P. Zukowski, T. N. Koltunowicz, Maintenance diagnostics of transformers considering the influence of short-circuit currents during operation. Eksploatacja i Niezawodnosc – Maintenance and Reliability 19, (3) (2017), 459-466. https://doi.org/10.17531/ein.2017.3.17 [19] S. Adamczak, K. Stepien, M. Wrzochal, Comparative study of measurement systems used to evaluate vibrations of rolling bearings, Procedia Engineering 192 (2017) 971-975. http://dx.doi.org/10.1016/j.proeng.2017.06.167 [20] DH. Hwang, YW. Youn, JH. Sun, KH. Choi, JH. Lee, YH. Kim, Support Vector Machine Based Bearing Fault Diagnosis for Induction Motors Using Vibration Signals, Journal of Electrical Engineering & Technology 10 (4) (2015) 1558-1565. http://dx.doi.org/10.5370/JEET.2015.10.4.1558 [21] W. Sawczuk, The Application of Vibration Accelerations in the Assessment of Average Friction Coefficient of a Railway Brake Disc, Measurement Science Review 17 (3) (2017) 125-134. http://dx.doi.org/10.1515/msr-2017-0016 [22] SB. Zhang, SL. Lu, QB. He, FR. Kong, Time-varying singular value decomposition for periodic transient identification in bearing fault diagnosis, Journal of Sound and Vibration 379 (2016) 213-231. http://dx.doi.org/10.1016/j.jsv.2016.05.035 [23] J. Hang, JZ. Zhang, M. Cheng, Application of multi-class fuzzy support vector machine classifier for fault diagnosis of wind turbine, Fuzzy Sets and Systems 297 (2016) 128-140. http://dx.doi.org/ 10.1016/j.fss.2015.07.005 [24] PA. Delgado-Arredondo, A. Garcia-Perez, D. Morinigo-Sotelo, RA. Osornio-Rios, JG. Avina-Cervantes, H. Rostro-Gonzalez, RD. Romero-Troncoso, Comparative study of time-frequency decomposition techniques for fault detection in induction motors using vibration analysis during startup transient, Shock and Vibration 2015, Article Number: 708034. http://dx.doi.org/10.1155/2015/708034 [25] XA. Yan, MP. Jia, L. Xiang, Compound fault diagnosis of rotating machinery based on OVMD and a 1.5dimension envelope spectrum, Measurement Science and Technology 27 (7) (2016) Article Number: 075002. http://dx.doi.org/10.1088/0957-0233/27/7/075002 [26] W. Sawczuk, GM. Szymanski, Diagnostics of the railway friction disc brake based on the analysis of the vibration signals in terms of resonant frequency, Archive of Applied Mechanics 87 (5) (2017) 801-815. http://dx.doi.org/10.1007/s00419-016-1202-0 [27] ZX. Yang, JH. Zhong, A Hybrid EEMD-Based SampEn and SVD for Acoustic Signal Processing and Fault Diagnosis, Entropy 18 (4) (2016) Article Number: 112. http://dx.doi.org/10.3390/e18040112 [28] A. Glowacz, Z. Glowacz, Diagnosis of stator faults of the single-phase induction motor using acoustic signals, Applied Acoustics 117 Part A (2017), 20-27. http://dx.doi.org/10.1016/j.apacoust.2016.10.012 [29] A. Glowacz, Fault diagnostics of acoustic signals of loaded synchronous motor using SMOFS-25EXPANDED and selected classifiers, Tehnicki Vjesnik-Technical Gazette 23 (5) (2016) 1365-1372. http://dx.doi.org/10.17559/TV-20150328135652 [30] W. Caesarendra, B. Kosasih, AK. Tieu, HT. Zhu, CAS. Moodie, Q. Zhu, Acoustic emission-based condition monitoring methods: Review and application for low speed slew bearing, Mechanical Systems and Signal Processing 72-73 (2016) 134-159. http://dx.doi.org/10.1016/j.ymssp.2015.10.020 [31] DP. Jena, SN. Panigrahi, Automatic gear and bearing fault localization using vibration and acoustic signals, Applied Acoustics 98 (2015) 20-33. http://dx.doi.org/10.1016/j.apacoust.2015.04.016 [32] EP. Frigieri, PHS. Campos, AP. Paiva, PP. Balestrassi, JR. Ferreira, CA. Ynoguti, A mel-frequency cepstral coefficient-based approach for surface roughness diagnosis in hard turning using acoustic signals and Gaussian Mixture Models, Applied Acoustics 113 (2016) 230-237. http://dx.doi.org/10.1016/ j.apacoust.2016.06.027 [33] J. Jozwik, Identification and monitoring of noise sources of CNC machine tools by acoustic holography methods, Advances in Science and Technology-Research Journal 10 (30) (2016) 127-137. http://dx.doi.org/10.12913/22998624/63386 [34] DY. Ning, CL. Sun, YJ. Gong, ZM. Zhang, JY. Hou, Extraction of fault component from abnormal sound in diesel engines using acoustic signals, Mechanical Systems and Signal Processing 75 (2016) 544-555. http://dx.doi.org/10.1016/j.ymssp.2015.10.037
15
[35] WJ. Wang, LL. Cui, DY. Chen, Multi-scale morphology analysis of acoustic emission signal and quantitative diagnosis for bearing fault, Acta Mechanica Sinica 32 (2) (2016) 265-272, http://dx.doi.org/10.1007/s10409-015-0529-z [36] AG. Garcia-Ramirez, LA. Morales-Hernandez, RA. Osornio-Rios, JP. Benitez-Rangel, A. Garcia-Perez, RD. Romero-Troncoso, Fault detection in induction motors and the impact on the kinematic chain through thermographic analysis, Electric Power Systems Research 114 (2014) 1-9. http://dx.doi.org/10.1016/ j.epsr.2014.03.031 [37] G. Singh, TCA. Kumar, VNA. Naikan, Induction motor inter turn fault detection using infrared thermographic analysis, Infrared Physics & Technology 77 (2016) 277-282. http://dx.doi.org/10.1016/j.infrared.2016.06.010 [38] M. Sebok, M. Gutten, M. Kucera, Diagnostics of electric equipments by means of thermovision, Przeglad Elektrotechniczny 87 (10) (2011) 313-317. [39] B. Bedkowski, J. Madej, The innovative design concept of thermal model for the calculation of the electromagnetic circuit of rotating electrical machines, Eksploatacja i Niezawodnosc – Maintenance and Reliability 17 (4) (2015) 481–486. http://dx.doi.org/10.17531/ein.2015.4.1 [40] O. Janssens, R. Schulz, V. Slavkovikj, K. Stockman, M. Loccufier, R. Van de Walle, S. Van Hoecke, Thermal image based fault diagnosis for rotating machinery, Infrared Physics & Technology 73 (2015) 7887. http://dx.doi.org/10.1016/j.infrared.2015.09.004 [41] F. Jeffali, B. EL Kihel, A. Nougaoui, F. Delaunois, Monitoring and diagnostic misalignment of asynchronous machines by Infrared Thermography, J. Mater. Environ. Sci. 6 (4) (2015) 1192-1199. [42] PA. Delgado-Arredondo, D. Morinigo-Sotelo, RA. Osornio-Rios, JG. Avina-Cervantes, H. RostroGonzalez, RD. Romero-Troncoso, Methodology for fault detection in induction motors via sound and vibration signals, Mechanical Systems and Signal Processing 83 (2017) 568-589. http://dx.doi.org/10.1016/j.ymssp.2016.06.032 [43] E. Germen, M. Basaran, M. Fidan, Sound based induction motor fault diagnosis using Kohonen selforganizing map, Mechanical Systems and Signal Processing 46 (2014) 45-58. http://dx.doi.org/10.1016/j.ymssp.2013.12.002 [44] A. Deptula, D. Kunderman, P. Osinski, U. Radziwanowska, R. Wlostowski, Acoustic Diagnostics Applications in the Study of Technical Condition of Combustion Engine, Archives of Acoustics 41 (2) (2016) 345-350. http://dx.doi.org/10.1515/aoa-2016-0036 [45] A. Deptula, P. Osinski, U. Radziwanowska, Decision support system for identifying technical condition of combustion engine, Archives of Acoustics 41(3) (2016) 449-460. [46] D. Valis, L. Zak, A. Walek, K. Pietrucha-Urbanik, Selected mathematical functions used for operation data information, Safety, Reliability and Risk Analysis: Beyond the Horizon (2014) 1303-1308. [47] D. Valis, L. Zak, O. Pokora, P. Lansky, Perspective analysis outcomes of selected tribodiagnostic data used as input for condition based maintenance, Reliability Engineering & System Safety 145 (2016) 231-242. http://dx.doi.org/10.1016/j.ress.2015.07.026 [48] Z. Vintr, D. Valis, A Tool for Decision Making in k-out-of-n System Maintenance, Applied Mechanics and Materials 110-116 (2012) 5257-5264. http://dx.doi.org/10.4028/www.scientific.net/AMM.110-116.5257 [49] ZF. Khan, G. NaliniPriya, Automatic Segmentation of Retinal Blood Vessels Employing Textural Fuzzy CMeans Clustering, IEEE International Conference on Emerging Technological Trends in Computing, Communications and Electrical Engineering (ICETT) (2016). [50] R. Koprowski, M. Lanza, C. Irregolare, Corneal power evaluation after myopic corneal refractive surgery using artificial neural networks. BioMedical Engineering OnLine 15 (2016), Article Number: 121. https://doi.org/10.1186/s12938-016-0243-5 [51] R. Tadeusiewicz, Neural networks in mining sciences–general overview and some representative examples, Archives of Mining Sciences 60 (4) (2015) 971-984. http://dx.doi.org/10.1515/amsc-2015-0064 [52] B. Ganovska, M. Molitoris, A. Hosovsky, J. Pitel, JB. Krolczyk, A. Ruggierio, GM. Krolczyk, S. Hloch, Design of the model for the on-line control of the AWJ technology based on neural networks, Indian Journal of Engineering and Materials Sciences 23 (4) (2016) 279-287. [53] Z. B. Hu, J. Su, V. Jotsov, O. Kochan, M. Mykyichuk, R. Kochan, T. Sasiuk, Data Science Applications to Improve Accuracy of Thermocouples. 2016 IEEE 8th International Conference on Intelligent Systems (IS), (2016) 180-188.
16
[54] RJ. O'Brien, JM. Fontana, N. Ponso, L. Molisani, A pattern recognition system based on acoustic signals for fault detection on composite materials, European Journal of Mechanics A-Solids 64 (2017) 1-10. http://dx.doi.org/10.1016/j.euromechsol.2017.01.007 [55] RA. Kanai, RG. Desavale, SP. Chavan, Experimental-Based Fault Diagnosis of Rolling Bearings Using Artificial Neural Network, Journal of Tribology-transactions of The Asme 138 (3) (2016) Article Number: 031103. http://dx.doi.org/10.1115/1.4032525 [56] W. Zhang, MP. Jia, L. Zhu, XA. Yan, Comprehensive Overview on Computational Intelligence Techniques for Machinery Condition Monitoring and Fault Diagnosis, Chinese Journal of Mechanical Engineering 30 (4) (2017) 782-795. [57] RHC. Palacios, IN. da Silva, A. Goedtel, WF. Godoy, A comprehensive evaluation of intelligent classifiers for fault identification in three-phase induction motors, Electric Power Systems Research 127 (2015) 249258. http://dx.doi.org/10.1016/j.epsr.2015.06.008 [58] A. Yadav, A. Swetapadma, A novel transmission line relaying scheme for fault detection and classification using wavelet transform and linear discriminant analysis, Ain Shams Engineering Journal 6 (1) (2015) 199209. http://dx.doi.org/10.1016/j.asej.2014.10.005 [59] A. Glowacz, W. Glowacz, Z. Glowacz, Recognition of armature current of DC generator depending on rotor speed using FFT, MSAF-1 and LDA, Eksploatacja i Niezawodnosc – Maintenance and Reliability 17, (1) (2015), 64-69. https://doi.org/10.17531/ein.2015.1.9 [60] K. Regulski, J. Jakubski, A. Opalinski, M. Brzezinski, Glowacki M., The Prediction of Moulding Sand Moisture Content Based on the Knowledge Acquired by Data Mining Techniques, Archives of Metallurgy and Materials 61 (3) (2016) 1363-1368. http://dx.doi.org/ 10.1515/amm-2016-0277 [61] K. Regulski, Formalization of Technological Knowledge in the Field of Metallurgy using Document Classification Tools Supported with Semantic Techniques, Archives of Metallurgy and Materials 62 (2) (2017) 715-720. http://dx.doi.org/10.1515/amm-2017-0108 [62] SB. Imandoust, M. Bolandraftar, Application of K-Nearest Neighbor (KNN) Approach for Predicting Economic Events: Theoretical Background, Int. Journal of Engineering Research and Applications 3 (5) (2013) 605-610. [63] A. Glowacz, Z. Glowacz, Recognition of rotor damages in a DC motor using acoustic signals, Bulletin of the Polish Academy of Sciences-Technical Sciences 65 (2) (2017) 187-194. http://dx.doi.org/10.1515/bpasts-2017-0023 [64] I. Bandyopadhyay, P. Purkait, C. Koley, A combined image processing and Nearest Neighbor Algorithm tool for classification of incipient faults in induction motor drives, Computers & Electrical Engineering 54 (2016) 296-312. http://dx.doi.org/10.1016/j.compeleceng.2016.01.014 [65] J. Tian, C. Morillo, MH. Azarian, M. Pecht, Motor Bearing Fault Detection Using Spectral Kurtosis-Based Feature Extraction Coupled With K-Nearest Neighbor Distance Analysis, IEEE Transactions on Industrial Electronics 63 (3) (2016) 1793-1803. http://dx.doi.org/10.1109/TIE.2015.2509913 [66] DY. Dou, SS. Zhou, Comparison of four direct classification methods for intelligent fault diagnosis of rotating machinery, Applied Soft Computing 46 (2016) 459-468. http://dx.doi.org/10.1016/ j.asoc.2016.05.015 [67] QC. Jiang, B. Huang, XF. Yan, GMM and optimal principal components-based Bayesian method for multimode fault diagnosis, Computers & Chemical Engineering 84 (2016) 338-349. http://dx.doi.org/ 10.1016/j.compchemeng.2015.09.013
17
Technique of fault diagnosis of the single-phase induction motor was proposed Original method of feature extraction of acoustic signal called MULTIEXPANDED was described Three states of the single-phase induction motor were analysed Efficiency of acoustic signal recognition was analysed
18
MSAF-20-
S0263-2241(17)30543-2 http://dx.doi.org/10.1016/j.measurement.2017.08.036 MEASUR 4934
To appear in:
Measurement
Received Date: Revised Date: Accepted Date:
3 May 2017 23 August 2017 25 August 2017
Please cite this article as: A. Glowacz, W. Glowacz, Z. Glowacz, J. Kozik, Early fault diagnosis of bearing and stator faults of the single-phase induction motor using acoustic signals, Measurement (2017), doi: http://dx.doi.org/ 10.1016/j.measurement.2017.08.036
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Early fault diagnosis of bearing and stator faults of the singlephase induction motor using acoustic signals Adam Glowacz1, Witold Glowacz1, Zygfryd Glowacz2, Jaroslaw Kozik2 1
AGH University of Science and Technology, Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering, Department of Automatics and Biomedical Engineering, Al. A. Mickiewicza 30, 30-059 Kraków, Poland, [email protected], [email protected] 2
AGH University of Science and Technology, Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering, Department of Power Electronics and Energy Control Systems, Al. A. Mickiewicza 30, 30-059 Kraków, Poland, [email protected], [email protected]
Abstract: An article describes an early fault diagnostic technique based on acoustic signals. The presented technique was used for a single-phase induction motor. The authors measured and analysed following states of the motor: healthy single-phase induction motor, single-phase induction motor with faulty bearing, single-phase induction motor with faulty bearing and shorted coils of auxiliary winding. A feature extraction method called MSAF-20-MULTIEXPANDED (Method of Selection of Amplitudes of Frequency - Multiexpanded) was discussed. The MSAF-20-MULTIEXPANDED was used to create feature vectors. The obtained vectors were classified by NN (Nearest Neighbour classifier), NM (Nearest Mean classifier) and GMM (Gaussian Mixture Models). The proposed technique can be used for diagnosis of the single-phase induction motors. It can be also used for other types of rotating electric motors. Keywords: Fault diagnosis, acoustic signal, bearing, motor, classification 1. Introduction Rotating electric motors are often used in industry, for example in oil refinery, pump oil, steel mill, mine, compressor [1]. Induction motors are widely used electric motors in industry. It is a motivation to analyse such machines. The single-phase induction motor (Fig. 1) is one of the types of induction motor. Diagnostics of rotating electric motors is a normal process of maintenance. A degradation of electric rotating motors depends on environment (heat, moisture) and operation time. Accidents, financial loss, unscheduled downtimes can be predicted based on an early diagnostics of motors. A fault state is a state, which causes adverse effects from the point of view of the correctness of its operation. The early fault state is is the state, in which there are symptoms of characteristic phenomena of the fault state (scratches, short circuits, broken coils, broken bars). In recent years monitoring of machines was developed by engineers [2-3] and companies (Siemens, Dreisilker etc.). Online monitoring of machines also allows for intelligent maintenance with the optimized use of maintenance resources. In the literature the following types of faults of motors were mentioned: stator faults (stator open phase faults, short circuits of windings (Fig. 2), increased resistance of connections), rotor faults (rotor open phase, short circuits of windings, broken bars, faulty ring of squirrel-cage, increased resistance of connections, shaft misalignment, faulty bearings - Fig. 3, 4, 5, rotor eccentricity, bent shaft). Descriptions of diagnostic techniques of machines can be found in the recent literature [4-9]. Diagnostic techniques of bearings are also described [10-14]. The diagnostic techniques are based on various
1
signals such as: electric current [15-18], vibrations [19-26], acoustic signals [27-35], thermal images [36-41], magnetic field. A hybrid EEMD-based SampEn for acoustic signal processing and fault diagnosis was analysed [27]. Diagnosis of stator faults of the single-phase induction motor using acoustic signals was presented [28]. Fault diagnostics of acoustic signals of loaded synchronous motor was also described [29]. Acoustic emission-based condition monitoring was described in the article [30]. Automatic bearing fault localization using vibration and acoustic signals was analysed in the literature [31]. An approach for surface roughness diagnosis in hard turning using acoustic signals was discussed [32]. Identification and monitoring of noise sources of CNC machine tools by acoustic holography methods were also presented [33]. Extraction of fault component from abnormal sound in diesel engines using acoustic signals was described [34]. Analysis of acoustic emission signal for bearing fault was also presented [35]. In the literature there were state of the art methods based on acoustic signals [42-45]. In the paper [42] scientists analysed three-phase induction motors (WEG 00136APE48T). The audio signal was acquired using a condenser microphone JST model CX-509 (cardioid polarization). The scientists used CEEMD (complete ensemble empirical mode decomposition) and TFDG (time-frequency distribution of Gabor). The proposed methodology can be used for fault diagnosis of the motors. However it was not possible to know a priori the number of modes in which the signal was decomposed. Moreover the localization of fault frequencies was related with speed of the motor. Another method based on acoustic signal was described in the article [43]. In this article scientists analysed acoustic signals of squirrel cage induction motors recorded by using 5 microphones simultaneously. The scientists settled 5 different microphones on the motor in hemispherical shape. The analysed motors were 3-phase and 2-pole squirrel cage induction motors rated at 2.2 kW and 380 V. The audio data were digitized with a sampling frequency of 44.1 kHz. Feature extraction was realised by using two different methods: cross correlation, wavelet transformation. To classify data the scientists used Kohonen self-organizing map. The results of analyses were good, however the acoustical approach was sensitive to the interferences of the environmental noises. In the literature there were also papers about applications of using acoustic diagnostics of an internal combustion engine with autoignition [44, 45]. The scientists analysed model of Fiat 1.3 JTD. 12 measurements were carried out for 6 different cases of the engine (engine speeds 1000 rpm, 2000 rpm). The scientists used decision trees and graphs to classify audio data [44, 45]. The proposed decision method of identifying engine failure based on sound emission. It allowed the quick identification of specific damage. However the method may be successfully used for similar engines.
Fig. 1. The healthy single-phase induction motors
2
Fig. 2. The single-phase induction motor with faulty bearing and shorted coils of auxiliary winding
Fig. 3. The single-phase induction motor with faulty bearing
Fig. 4. a) Healthy bearing of the single-phase induction motor, b) Faulty bearing of the single-phase induction motor,
3
Fig. 5. Bearing of the single-phase induction motor
The recognition of acoustic signals of the motor is a difficult task. The first problem is set of microphone. There are many possibilities to do it. The authors put microphone 0.3 m from the rotor of the motor. The second problem is different parameters of the motor such as: size, mass of the motor, rotor speed, nominal current, nominal voltage, power. A lot of samples of acoustic signals from many rotating machines are required for proper diagnostic. Proper diagnostic also requires cooperation with operators of motors and industry. It should be defined what faults and failures are the most essential for diagnosis. The capacitor microphone is less efficient than vibration sensor and thermal camera. An acoustic signal is more interfered than vibration signal and thermal image. It is very difficult to extract characteristic features from the acoustic signal. If we use other features (easier to recognise) we will get better results. Article describes the early fault diagnostic technique based on acoustic signals. The presented technique is used for the single-phase induction motor. Next sections describe proposed technique and analysis. 2. Proposed technique based on acoustic signals The proposed technique was based on acoustic signals of the single-phase induction motor. The technique used various signal processing methods such as: preprocessing, feature extraction and classification. The first step of recognition was data acquisition. A personal computer, ZALMAN ZMMIC1 (Microphone Operation Mode mono, Connectivity Technology Wired, Sensitivity 40 dB, Impedance 2.2 ㏀, Standard power supply 2.0 V DC, S/N 58 dB, Directivity Omnidirectional) and OLYMPUS TP-7 (Microphone Operation Mode mono, Connectivity Technology Wired, Sensitivity 34 dB, Impedance 2.2 Ohm, Frequency Response 50 Hz, Audio Input Details Mono 50 - 20000 Hz Output Impedance 2.2 ㏀) microphones were used for this purpose (Fig. 6). Other types of microphones can be also used. Parameters of soundtrack were following: 44.1 kHz - sampling frequency, 16-bit - bit depth, single channel - number of channels, WAVE PCM - sound file format. Obtained audio data were split into smaller blocks of data. Next these data were normalized. Normalized data were processed by windowing (window size 32768 - 0.74 s, for f=44100 Hz) and the FFT. Obtained amplitudes of frequency (frequency components) were processed by the feature extraction method MSAF-20MULTIEXPANDED. The feature extraction method MSAF-20-MULTIEXPANDED created the feature vectors. The last step of proposed techniques was the classification. The classification step was divided into pattern creation process and test process.
4
Fig. 6. Experimental setup
The pattern creation process used a training set. The test process used a test set. Obtained test feature vectors were classified by Nearest Neighbour classifier, Nearest Mean classifier and Gaussian Mixture Models (Fig. 7).
Fig. 7. Block diagram of the early fault diagnostic technique based on acoustic signals with the use of the MSAF-20-MULTIEXPANDED, Nearest Neighbour classifier, Nearest Mean classifier and Gaussian Mixture Models
2.1. Method of Selection of Amplitudes of Frequency-20-Multiexpanded The feature extraction method - Method of Selection of Amplitudes of Frequency-20-Multiexpanded (MSAF-20-MULTIEXPANDED) was based on processing of the FFT spectra of acoustic signals of the single-phase induction motor. This method can extract characteristic features from the acoustic signals properly. It analysed differences between frequency spectra of states of the single-phase induction motor. It was presented as a block diagram (Fig. 8).
5
Fig. 8. The block diagram of the MSAF-20-MULTIEXPANDED
The steps of the MSAF-20-MULTIEXPANDED were following: 1) Calculate the frequency spectra of acoustic signals of the single-phase induction motor (for example we can use 4 training samples of state S1 and 4 training samples of state S2 and 4 training samples of state S3). The frequency spectrum of acoustic signal of the healthy single-phase induction motor was expressed by vector: fsh=[fsh1, fsh2, ..., fsh16384]. The frequency spectrum of acoustic signal of the single-phase induction motor with faulty bearing was denoted by vector: fsfb=[fsfb1, fsfb2, ..., fsfb16384]. The frequency spectrum of acoustic signal of the single-phase induction motor with faulty bearing and shorted coils of auxiliary winding was expressed by vector: fsfbsc=[fsfbsc1, fsfbsc2, ..., fsfbsc16384]. 2) Calculate differences between frequency spectra of states of the single-phase induction motor. The authors analysed 3 acoustic signals so differences were following: fsh - fsfb, fsh - fsfbsc, fsfb fsfbsc. 3) Calculate absolute values of the obtained differences |fsh - fsfb|, |fsh - fsfbsc|, |fsfb - fsfbsc|. 4) Select 20 maximum amplitudes of frequency for each difference between states of the single-phase induction motor, for example max1(|fsh - fsfb|), max2(|fsh - fsfb|), ..., max20(|fsh - fsfb|). 5) Set a parameter CF-MULTI. The parameter CF-MULTI was expressed as: CF-MULTI=(number of required common frequencies of analysed training sets)/(number of all frequencies of analysed training sets). The parameter CF-MULTI was used for selection of common frequencies. For example, there was 4 training sets each of them had 1 training sample of state S1, 1 training sample of state S2, 1 training sample of state S3 (12 training samples). 12 differences of frequency spectra were calculated, 3 for each training set. The parameter CF-MULTI was equal to 0.65, and 0.65<8/12=0.6666, then 8 of 12 frequencies were required to take this frequency for further processing (9, 10, 11, 12>8 frequencies were also good). When the parameter CF-MULTI was equal to 0.65, we needed 8 of 12 the same frequencies. What would happen to the other parameters CF-MULTI? For example the parameter CF-MULTI was equal to 0.82, then 10 frequencies were selected because 0.82<10/12=0.83333. When the parameter CF-MULTI was equal to 0.57, then 7 frequencies were selected because 0.57<7/12=0.58333. 6) Form the feature vector.
6
The differences of frequency spectra of acoustic signals of the single-phase induction motor |fsh - fsfb|, |fsh - fsfbsc|, |fsfb - fsfbsc| were shown below (Fig. 9-11). The rotor speed was equal to 1390 rpm. The next step was selection of common frequencies.
Fig. 9. The difference between spectra of frequency of acoustic signal of the healthy single-phase induction motor and the single-phase induction motor with faulty bearing (|fsh - fsfb|)
Fig. 10. The difference between spectra of frequency of acoustic signal of the healthy single-phase induction motor and the single-phase induction motor with faulty bearing and shorted coils of auxiliary winding (|fsh - fsfbsc|)
7
Fig. 11. The difference between spectra of frequency of acoustic signal of the single-phase induction motor with faulty bearing and the
single-phase induction motor with faulty bearing and shorted coils of auxiliary winding (|fsfb - fsfbsc|)
The frequency spectrum of acoustic signal of the single-phase induction motor with faulty bearing was presented in figure 12. The MSAF-20-MULTIEXPANDED selected following frequencies 101, 201, 39 Hz (CF-MULTI=0.58), 101, 201 Hz (CF-MULTI=0.66) and 101 Hz (CF-MULTI=0.83). The selected frequencies (39, 101, 201 Hz) of acoustic signal of the single-phase induction motor with faulty bearing were depicted in figure 13. The selected frequencies created feature vectors (vector of 3 elements, vector of 2 elements, vector of 1 element depending on the parameter CF-MULTI ).
Fig. 12. The frequency spectrum of acoustic signal of the single-phase induction motor with faulty bearing
8
Fig. 13. The selected frequencies (39, 101, 201 Hz) of acoustic signal of the single-phase induction motor with faulty bearing
Next the obtained feature vectors were used by Nearest Neighbour classifier, Nearest Mean classifier and GMM (Gaussian Mixture Models). In the paper the authors analysed 3 selected classifiers. However, scientists developed a lot of classifiers in the literature: fuzzy logic [46-49], LVQ (Learning Vector Quantization), neural networks [50-56], self-organizing map, k-means clustering, Naive Bayes classifier [57], SVM (Support Vector Machine) [56, 57], LDA (Linear Discriminant Analysis) [58, 59], classification trees [60], rough sets [61], genetic algorithms. 2.2. Nearest Neighbour classifier Among the various methods of data classification, the Nearest Neighbour classifier achieved high recognition results. This classifier did not have a priori assumptions about the distributions of training and test sets. The Nearest Neighbour classifier involved a training set of both healthy motor and faulty motor cases. A new test feature vector was classified by calculating the distance to the nearest training feature vector. The classifier used distance function such as Manhattan distance, Euclidean distance etc. to calculate distances between test and training feature vectors. The authors used the Manhattan distance, because it was very fast distance function (time complexity O(NlogN)). The Manhattan distance was expressed by formula (1): 3
d (test_fsh, training_fsfb) ∑| (test _ fshi - training _ fsfbi ) | ,
(1)
i 1
where feature vectors test_fsh=[test_fsh29, test_fsh75, ..., test_fsh149] and training_fsfb=[training_fsfb29, training_fsfb75, ..., training_fsfb149]. The analysis were conducted for test feature vectors: test_fsh, test_fsfb, test_fsfbsc and training feature vectors: training_fsh, training_fsfb, training_fsfbsc. However the distance functions such as Chebyshev, Euclidean, Jacquard, cosine distances were also proper for recognition of acoustic signals. More information about the Nearest Neighbour classifier is described in recent literature [62-66]. 2.3. Nearest Mean classifier The Nearest Mean classifier was similar to the Nearest Neighbour classifier. The Nearest Mean used mean vectors instead of training feature vectors. Other steps of the Nearest Mean classifier were the same as for the Nearest Neighbour classifier. Mean vector mean_fsh was expressed by the formula (2):
9
mean_fsh= (training1_fsh + training2_fsh + training3_fsh + training4_fsh)/4
(2)
where feature vector training1_fsh=[training1_fsh29, training1_fsh75, ..., training1_fsh149]. The analysis were conducted for mean vectors: mean_fsh, mean_fsfb, mean_fsfbsc. The Manhattan distance was used for the Nearest Mean classifier. 2.4. Gaussian Mixture Models The Gaussian Mixture Models (GMM) was one of the most popular classification method. These models were used for clustering purpose. It was based on linear combination of finite number of Gaussian distributions. The GMM assessed the partitions of the feature vectors by considering that each component represented a cluster. The clusters had an independent Gaussian distribution, each with their own covariance matrix and mean. The classifier calculated the probability of belonging to each cluster. The GMM used an iterative technique called Expectation Maximization. The classification was achieved by assigning each feature vector to the most likely cluster. The GMM calculated weight of each Gaussian component, covariance matrix and mean. The probability density function of the GMM was expressed by following formula (3): z
f ( x) ∑| wi N ( x; , i ) | ,
(3)
i 1
where z - the number of the Gaussian components, N ( x, , i ) - the probability density function of normal distribution, wi - the weight of each Gaussian component, such that (4):
z
i 1
wi 1, andi : wi 0 ,
(4)
where , i , i=1,2,...,z - parameters of the Gaussian distributions. The GMM used the score function of the model given by the maximum overall probability for the given class. It was defined by following formula (5):
m arg max score(test_fsh, p) , 1 n g
(5)
where test_fsh - test vector, g - the number of output classes and score(test_fsh, p) - the probability value of the GMM classifier for the models trained for the current p-th class [32, 67]. More information about the GMM can be found in the following literature [32, 67]. 3. Analysis of fault diagnostic technique The authors measured and analysed following 3 states of the motor: healthy single-phase induction motor (Fig. 1, 14), single-phase induction motor with faulty bearing (Fig. 3, 15), single-phase induction motor with faulty bearing and shorted coils of auxiliary winding (Fig. 2, 15, 16). Measurements were performed under laboratory conditions. Each induction motor had following parameters: IN=1.06 A, VN=230 V, PN = 0.12 kW, fN= 50 Hz, Ihm = 0.5 A, Imsc = 1.8 A, s = 1390 rpm, M=3.3 kg,
10
where IN - nominal current of the motor, VN - nominal voltage of the motor, PN - nominal power of the motor, fN - nominal frequency of the motor, Ihm - current of the healthy single-phase induction motor, Imsc - current of the single-phase induction motor with shorted coils of auxiliary winding, s - rotor speed, M - mass of the motor.
Fig. 14. Stator windings of the healthy single-phase induction motor
Fig. 15. Faulty bearing of the single-phase induction motor,
Fig. 16. Stator windings of the single-phase induction motor with shorted coils of auxiliary winding
A new induction motor should be similar to the analysed single-phase induction motor (similar operational parameters, size, materials, rotor speed). The authors used 108 one-second test samples of acoustic signals for the test process and 12 one-second training samples for the pattern creation process. The classifiers (NN, NM, GMM) used processed samples of acoustic signals - feature vectors. The
11
feature vectors used frequencies 101, 201, 39 Hz (CF-MULTI=0.58), 101, 201 Hz (CF-MULTI=0.66) and 101 Hz (CF-MULTI=0.83 - see section 2.1) . Efficiency of acoustic signal recognition was defined as (6): E AS
TS 100% AS
(6)
where: EAS – efficiency of acoustic signal recognition, TS – number of test samples recognised properly, AS – number of all test samples. Total efficiency of acoustic signal recognition was expressed as (7): ET
E ASH E ASFB E ASFBSH 3
(7)
Where ET - Total efficiency of acoustic signal recognition, EASH - efficiency of acoustic signal recognition of the healthy single-phase induction motor, EASFB - efficiency of acoustic signal recognition of the single-phase induction motor with faulty bearing, EASFBSH - efficiency of acoustic signal recognition of the single-phase induction motor with faulty bearing and shorted coils of auxiliary winding. The results of acoustic signal recognition of the single-phase induction motor using the MSAF-20MULTIEXPANDED were presented in Tables 1-3. Tab. 1. The results of acoustic signal recognition of the single-phase induction motor using the MSAF-20-MULTIEXPANDED, and the Nearest Neighbour classifier
Type of acoustic signal
healthy single-phase induction motor single-phase induction motor with faulty bearing single-phase induction motor with faulty bearing and shorted coils of auxiliary winding ET [%]
EAS [%] Frequency [Hz] 101 101, 101, 201 201, 39 100 100 100 88.8
100
97.2
63.8
75
58.3
84.1
91.6
85.2
Tab. 2. The results of acoustic signal recognition of the single-phase induction motor using the MSAF-20-MULTIEXPANDED, and the Nearest Mean classifier
Type of acoustic signal
healthy single-phase
EAS [%] Frequency [Hz] 101 101, 101, 201 201, 39 100 100 100
12
induction motor single-phase induction motor with faulty bearing single-phase induction motor with faulty bearing and shorted coils of auxiliary winding ET [%]
83.3
94.4
94.4
86
91.6
86.1
89.7
95.3
93.5
Tab. 3. The results of acoustic signal recognition of the single-phase induction motor using the MSAF-20-MULTIEXPANDED, and the Gaussian Mixture Models
Type of acoustic signal
healthy single-phase induction motor single-phase induction motor with faulty bearing single-phase induction motor with faulty bearing and shorted coils of auxiliary winding ET [%]
EAS [%] Frequency [Hz] 101 101, 101, 201 201, 39 100 100 100 66.6
94.4
100
100
2.7
2.7
88.8
65.7
67.5
The results of all selected classifiers were following: Nearest Neighbour classifier (ET = 84.1-91.6%), Nearest Mean classifier (ET = 89.7-95.3%), Gaussian Mixture Models (ET = 65.7-88.8%). 4. Conclusions The article presents the early fault diagnostic technique based on acoustic signals. The proposed technique was used for the single-phase induction motor. The following states of the motor were analysed: healthy single-phase induction motor, single-phase induction motor with faulty bearing, single-phase induction motor with faulty bearing and shorted coils of auxiliary winding. The authors used feature extraction method called the MSAF-20-MULTIEXPANDED. In this article 3 classifiers were analysed: Nearest Neighbour classifier, Nearest Mean classifier and Gaussian Mixture Models. The results of all selected classifiers were following: Nearest Neighbour classifier (ET = 84.1-91.6%), Nearest Mean classifier (ET = 89.7-95.3%), Gaussian Mixture Models (ET = 65.7-88.8%). The fault diagnostic technique can be used for protection of single-phase induction motors. Other types of rotating machines can be also diagnosed by the described technique. The proposed technique is noninvasive and inexpensive. However the proposed technique was sensitive to the interferences of the environmental noises. The proposed diagnostic technique can be extended by thermal signals and vibrations. In the future the authors will analyse other faults and electric motors with various operational parameters. Next more reliable diagnostic technique will be developed and used by manufacturers of motors.
13
Acknowledgments This research has been financed by AGH University of Science and Technology, grant no. 11.11.120.612 (Adam Glowacz), grant no. 11.11.120.815 (Witold Glowacz), grant no. 11.11.120.354 (Zygfryd Glowacz, Jaroslaw Kozik). We thank reviewers for their valuable suggestions. References [1] H. Henao, GA. Capolino, M. Fernandez-Cabanas, F. Filippetti, C. Bruzzese, E. Strangas, R. Pusca, J. Estima, M. Riera-Guasp, S. Hedayati-Kia, Trends in Fault Diagnosis for Electrical Machines A Review of Diagnostic Techniques, IEEE Industrial Electronics Magazine 8 (2) (2014) 31-42. http://dx.doi.org/10.1109/ MIE.2013.2287651 [2] M. Irfan, N. Saad, R. Ibrahim, VS. Asirvadam, An on-line condition monitoring system for induction motors via instantaneous power analysis, Journal of Mechanical Science and Technology 29 (4) (2015) 1483-1492. http://dx.doi.org/10.1007/s12206-015-0321-9 [3] R. Islam, SA. Khan, JM. Kim, Discriminant Feature Distribution Analysis-Based Hybrid Feature Selection for Online Bearing Fault Diagnosis in Induction Motors, Journal of Sensors (2016), Article Number: 7145715. http://dx.doi.org/10.1155/2016/7145715 [4] Y. Jiang, ZX. Li, C. Zhang, C. Hu, Z. Peng, On the bi-dimensional variational decomposition applied to nonstationary vibration signals for rolling bearing crack detection in coal cutters, Measurement Science and Technology 27 (6) (2016), Article Number: 065103. http://dx.doi.org/10.1088/0957-0233/27/6/065103 [5] D. Lopez-Perez, J. Antonino-Daviu, Application of Infrared Thermography to Failure Detection in Industrial Induction Motors: Case Stories, IEEE Transactions on Industry Applications 53 (3) (2017) 1901-1908. http://dx.doi.org/10.1109/TIA.2017.2655008 [6] D. Mika, J. Jozwik, Normative measurements of noise at CNC machines work stations, Advances in Science and Technology-Research Journal 10 (30) (2016) 138-143. http://dx.doi.org/10.12913/22998624/63387 [7] ZX. Li, Y. Jiang, C. Hu, Z. Peng, Recent progress on decoupling diagnosis of hybrid failures in gear transmission systems using vibration sensor signal: A review, Measurement 90 (2016) 4–19. http://dx.doi.org/10.1016/j.measurement.2016.04.036 [8] GM. Krolczyk, JB. Krolczyk, S. Legutko, A. Hunjet, Effect of the disc processing technology on the vibration level of the chipper during operations, Tehnicki Vjesnik-Technical Gazette 21 (2) (2014) 447-450. [9] L. Jedlinski, J. Caban, L. Krzywonos, S. Wierzbicki, F. Brumercik, Application of vibration signal in the diagnosis of IC engine valve clearance, Journal of Vibroengineering 17 (1) (2015) 175-187. [10] G. Perun, Z. Stanik, Evaluation of state of rolling bearings mounted in vehicles with use of vibration signals, Archives of Metallurgy and Materials 60 (3) (2015) 1679-1683. http://dx.doi.org/10.1515/amm-2015-0291 [11] Z. Stanik, G. Perun, T. Matyja, Effective methods for the diagnosis of vehicles rolling bearings wear and damages, Archives of Metallurgy and Materials 60 (3) (2015) 1717-1724. http://dx.doi.org/10.1515/amm2015-0296 [12] B. Van Hecke, J. Yoon, D. He, Low speed bearing fault diagnosis using acoustic emission sensors, Applied Acoustics 105 (2016) 35–44. http://dx.doi.org/10.1016/j.apacoust.2015.10.028 [13] F. Hemmati, W. Orfali, MS. Gadala, Roller bearing acoustic signature extraction by wavelet packet transform, applications in fault detection and size estimation, Applied Acoustics 104 (2016), 101-118. http://dx.doi.org/10.1016/j.apacoust.2015.11.003 [14] P. Rzeszucinski, M. Orman, CT. Pinto, A. Tkaczyk, M. Sulowicz, A signal processing approach to bearing fault detection with the use of a mobile phone, 2015 IEEE 10th International Symposium on Diagnostics For Electric Machines, Power Electronics and Drives (Sdemped) (2015) 310-315. [15] Z. Glowacz, A. Glowacz, Simulation language for analysis of discrete-continuous electrical systems (SESL2), Proceedings of the 26th IASTED International Conference on Modelling, Identification, and Control (2007) 94-99. [16] M. Michalak, B. Sikora, J. Sobczyk, Diagnostic model for longwall conveyor engines, Man-Machine Interactions 4, ICMMI, Book Series: Advances in Intelligent Systems and Computing, 391 (2016) 437-448. http://dx.doi.org/10.1007/978-3-319-23437-3_37
14
[17] C. Da Costa, M. Kashiwagi, MH. Mathias, Rotor Failure Diagnosis of Induction Motors by Wavelet Transform and Fourier Transform in Function of the Load, International Conference on Computer Science and Artificial Intelligence (2015) 109-113. [18] M. Gutten, D. Korenciak, M. Kucera, M. Sebok, M. Opielak, P. Zukowski, T. N. Koltunowicz, Maintenance diagnostics of transformers considering the influence of short-circuit currents during operation. Eksploatacja i Niezawodnosc – Maintenance and Reliability 19, (3) (2017), 459-466. https://doi.org/10.17531/ein.2017.3.17 [19] S. Adamczak, K. Stepien, M. Wrzochal, Comparative study of measurement systems used to evaluate vibrations of rolling bearings, Procedia Engineering 192 (2017) 971-975. http://dx.doi.org/10.1016/j.proeng.2017.06.167 [20] DH. Hwang, YW. Youn, JH. Sun, KH. Choi, JH. Lee, YH. Kim, Support Vector Machine Based Bearing Fault Diagnosis for Induction Motors Using Vibration Signals, Journal of Electrical Engineering & Technology 10 (4) (2015) 1558-1565. http://dx.doi.org/10.5370/JEET.2015.10.4.1558 [21] W. Sawczuk, The Application of Vibration Accelerations in the Assessment of Average Friction Coefficient of a Railway Brake Disc, Measurement Science Review 17 (3) (2017) 125-134. http://dx.doi.org/10.1515/msr-2017-0016 [22] SB. Zhang, SL. Lu, QB. He, FR. Kong, Time-varying singular value decomposition for periodic transient identification in bearing fault diagnosis, Journal of Sound and Vibration 379 (2016) 213-231. http://dx.doi.org/10.1016/j.jsv.2016.05.035 [23] J. Hang, JZ. Zhang, M. Cheng, Application of multi-class fuzzy support vector machine classifier for fault diagnosis of wind turbine, Fuzzy Sets and Systems 297 (2016) 128-140. http://dx.doi.org/ 10.1016/j.fss.2015.07.005 [24] PA. Delgado-Arredondo, A. Garcia-Perez, D. Morinigo-Sotelo, RA. Osornio-Rios, JG. Avina-Cervantes, H. Rostro-Gonzalez, RD. Romero-Troncoso, Comparative study of time-frequency decomposition techniques for fault detection in induction motors using vibration analysis during startup transient, Shock and Vibration 2015, Article Number: 708034. http://dx.doi.org/10.1155/2015/708034 [25] XA. Yan, MP. Jia, L. Xiang, Compound fault diagnosis of rotating machinery based on OVMD and a 1.5dimension envelope spectrum, Measurement Science and Technology 27 (7) (2016) Article Number: 075002. http://dx.doi.org/10.1088/0957-0233/27/7/075002 [26] W. Sawczuk, GM. Szymanski, Diagnostics of the railway friction disc brake based on the analysis of the vibration signals in terms of resonant frequency, Archive of Applied Mechanics 87 (5) (2017) 801-815. http://dx.doi.org/10.1007/s00419-016-1202-0 [27] ZX. Yang, JH. Zhong, A Hybrid EEMD-Based SampEn and SVD for Acoustic Signal Processing and Fault Diagnosis, Entropy 18 (4) (2016) Article Number: 112. http://dx.doi.org/10.3390/e18040112 [28] A. Glowacz, Z. Glowacz, Diagnosis of stator faults of the single-phase induction motor using acoustic signals, Applied Acoustics 117 Part A (2017), 20-27. http://dx.doi.org/10.1016/j.apacoust.2016.10.012 [29] A. Glowacz, Fault diagnostics of acoustic signals of loaded synchronous motor using SMOFS-25EXPANDED and selected classifiers, Tehnicki Vjesnik-Technical Gazette 23 (5) (2016) 1365-1372. http://dx.doi.org/10.17559/TV-20150328135652 [30] W. Caesarendra, B. Kosasih, AK. Tieu, HT. Zhu, CAS. Moodie, Q. Zhu, Acoustic emission-based condition monitoring methods: Review and application for low speed slew bearing, Mechanical Systems and Signal Processing 72-73 (2016) 134-159. http://dx.doi.org/10.1016/j.ymssp.2015.10.020 [31] DP. Jena, SN. Panigrahi, Automatic gear and bearing fault localization using vibration and acoustic signals, Applied Acoustics 98 (2015) 20-33. http://dx.doi.org/10.1016/j.apacoust.2015.04.016 [32] EP. Frigieri, PHS. Campos, AP. Paiva, PP. Balestrassi, JR. Ferreira, CA. Ynoguti, A mel-frequency cepstral coefficient-based approach for surface roughness diagnosis in hard turning using acoustic signals and Gaussian Mixture Models, Applied Acoustics 113 (2016) 230-237. http://dx.doi.org/10.1016/ j.apacoust.2016.06.027 [33] J. Jozwik, Identification and monitoring of noise sources of CNC machine tools by acoustic holography methods, Advances in Science and Technology-Research Journal 10 (30) (2016) 127-137. http://dx.doi.org/10.12913/22998624/63386 [34] DY. Ning, CL. Sun, YJ. Gong, ZM. Zhang, JY. Hou, Extraction of fault component from abnormal sound in diesel engines using acoustic signals, Mechanical Systems and Signal Processing 75 (2016) 544-555. http://dx.doi.org/10.1016/j.ymssp.2015.10.037
15
[35] WJ. Wang, LL. Cui, DY. Chen, Multi-scale morphology analysis of acoustic emission signal and quantitative diagnosis for bearing fault, Acta Mechanica Sinica 32 (2) (2016) 265-272, http://dx.doi.org/10.1007/s10409-015-0529-z [36] AG. Garcia-Ramirez, LA. Morales-Hernandez, RA. Osornio-Rios, JP. Benitez-Rangel, A. Garcia-Perez, RD. Romero-Troncoso, Fault detection in induction motors and the impact on the kinematic chain through thermographic analysis, Electric Power Systems Research 114 (2014) 1-9. http://dx.doi.org/10.1016/ j.epsr.2014.03.031 [37] G. Singh, TCA. Kumar, VNA. Naikan, Induction motor inter turn fault detection using infrared thermographic analysis, Infrared Physics & Technology 77 (2016) 277-282. http://dx.doi.org/10.1016/j.infrared.2016.06.010 [38] M. Sebok, M. Gutten, M. Kucera, Diagnostics of electric equipments by means of thermovision, Przeglad Elektrotechniczny 87 (10) (2011) 313-317. [39] B. Bedkowski, J. Madej, The innovative design concept of thermal model for the calculation of the electromagnetic circuit of rotating electrical machines, Eksploatacja i Niezawodnosc – Maintenance and Reliability 17 (4) (2015) 481–486. http://dx.doi.org/10.17531/ein.2015.4.1 [40] O. Janssens, R. Schulz, V. Slavkovikj, K. Stockman, M. Loccufier, R. Van de Walle, S. Van Hoecke, Thermal image based fault diagnosis for rotating machinery, Infrared Physics & Technology 73 (2015) 7887. http://dx.doi.org/10.1016/j.infrared.2015.09.004 [41] F. Jeffali, B. EL Kihel, A. Nougaoui, F. Delaunois, Monitoring and diagnostic misalignment of asynchronous machines by Infrared Thermography, J. Mater. Environ. Sci. 6 (4) (2015) 1192-1199. [42] PA. Delgado-Arredondo, D. Morinigo-Sotelo, RA. Osornio-Rios, JG. Avina-Cervantes, H. RostroGonzalez, RD. Romero-Troncoso, Methodology for fault detection in induction motors via sound and vibration signals, Mechanical Systems and Signal Processing 83 (2017) 568-589. http://dx.doi.org/10.1016/j.ymssp.2016.06.032 [43] E. Germen, M. Basaran, M. Fidan, Sound based induction motor fault diagnosis using Kohonen selforganizing map, Mechanical Systems and Signal Processing 46 (2014) 45-58. http://dx.doi.org/10.1016/j.ymssp.2013.12.002 [44] A. Deptula, D. Kunderman, P. Osinski, U. Radziwanowska, R. Wlostowski, Acoustic Diagnostics Applications in the Study of Technical Condition of Combustion Engine, Archives of Acoustics 41 (2) (2016) 345-350. http://dx.doi.org/10.1515/aoa-2016-0036 [45] A. Deptula, P. Osinski, U. Radziwanowska, Decision support system for identifying technical condition of combustion engine, Archives of Acoustics 41(3) (2016) 449-460. [46] D. Valis, L. Zak, A. Walek, K. Pietrucha-Urbanik, Selected mathematical functions used for operation data information, Safety, Reliability and Risk Analysis: Beyond the Horizon (2014) 1303-1308. [47] D. Valis, L. Zak, O. Pokora, P. Lansky, Perspective analysis outcomes of selected tribodiagnostic data used as input for condition based maintenance, Reliability Engineering & System Safety 145 (2016) 231-242. http://dx.doi.org/10.1016/j.ress.2015.07.026 [48] Z. Vintr, D. Valis, A Tool for Decision Making in k-out-of-n System Maintenance, Applied Mechanics and Materials 110-116 (2012) 5257-5264. http://dx.doi.org/10.4028/www.scientific.net/AMM.110-116.5257 [49] ZF. Khan, G. NaliniPriya, Automatic Segmentation of Retinal Blood Vessels Employing Textural Fuzzy CMeans Clustering, IEEE International Conference on Emerging Technological Trends in Computing, Communications and Electrical Engineering (ICETT) (2016). [50] R. Koprowski, M. Lanza, C. Irregolare, Corneal power evaluation after myopic corneal refractive surgery using artificial neural networks. BioMedical Engineering OnLine 15 (2016), Article Number: 121. https://doi.org/10.1186/s12938-016-0243-5 [51] R. Tadeusiewicz, Neural networks in mining sciences–general overview and some representative examples, Archives of Mining Sciences 60 (4) (2015) 971-984. http://dx.doi.org/10.1515/amsc-2015-0064 [52] B. Ganovska, M. Molitoris, A. Hosovsky, J. Pitel, JB. Krolczyk, A. Ruggierio, GM. Krolczyk, S. Hloch, Design of the model for the on-line control of the AWJ technology based on neural networks, Indian Journal of Engineering and Materials Sciences 23 (4) (2016) 279-287. [53] Z. B. Hu, J. Su, V. Jotsov, O. Kochan, M. Mykyichuk, R. Kochan, T. Sasiuk, Data Science Applications to Improve Accuracy of Thermocouples. 2016 IEEE 8th International Conference on Intelligent Systems (IS), (2016) 180-188.
16
[54] RJ. O'Brien, JM. Fontana, N. Ponso, L. Molisani, A pattern recognition system based on acoustic signals for fault detection on composite materials, European Journal of Mechanics A-Solids 64 (2017) 1-10. http://dx.doi.org/10.1016/j.euromechsol.2017.01.007 [55] RA. Kanai, RG. Desavale, SP. Chavan, Experimental-Based Fault Diagnosis of Rolling Bearings Using Artificial Neural Network, Journal of Tribology-transactions of The Asme 138 (3) (2016) Article Number: 031103. http://dx.doi.org/10.1115/1.4032525 [56] W. Zhang, MP. Jia, L. Zhu, XA. Yan, Comprehensive Overview on Computational Intelligence Techniques for Machinery Condition Monitoring and Fault Diagnosis, Chinese Journal of Mechanical Engineering 30 (4) (2017) 782-795. [57] RHC. Palacios, IN. da Silva, A. Goedtel, WF. Godoy, A comprehensive evaluation of intelligent classifiers for fault identification in three-phase induction motors, Electric Power Systems Research 127 (2015) 249258. http://dx.doi.org/10.1016/j.epsr.2015.06.008 [58] A. Yadav, A. Swetapadma, A novel transmission line relaying scheme for fault detection and classification using wavelet transform and linear discriminant analysis, Ain Shams Engineering Journal 6 (1) (2015) 199209. http://dx.doi.org/10.1016/j.asej.2014.10.005 [59] A. Glowacz, W. Glowacz, Z. Glowacz, Recognition of armature current of DC generator depending on rotor speed using FFT, MSAF-1 and LDA, Eksploatacja i Niezawodnosc – Maintenance and Reliability 17, (1) (2015), 64-69. https://doi.org/10.17531/ein.2015.1.9 [60] K. Regulski, J. Jakubski, A. Opalinski, M. Brzezinski, Glowacki M., The Prediction of Moulding Sand Moisture Content Based on the Knowledge Acquired by Data Mining Techniques, Archives of Metallurgy and Materials 61 (3) (2016) 1363-1368. http://dx.doi.org/ 10.1515/amm-2016-0277 [61] K. Regulski, Formalization of Technological Knowledge in the Field of Metallurgy using Document Classification Tools Supported with Semantic Techniques, Archives of Metallurgy and Materials 62 (2) (2017) 715-720. http://dx.doi.org/10.1515/amm-2017-0108 [62] SB. Imandoust, M. Bolandraftar, Application of K-Nearest Neighbor (KNN) Approach for Predicting Economic Events: Theoretical Background, Int. Journal of Engineering Research and Applications 3 (5) (2013) 605-610. [63] A. Glowacz, Z. Glowacz, Recognition of rotor damages in a DC motor using acoustic signals, Bulletin of the Polish Academy of Sciences-Technical Sciences 65 (2) (2017) 187-194. http://dx.doi.org/10.1515/bpasts-2017-0023 [64] I. Bandyopadhyay, P. Purkait, C. Koley, A combined image processing and Nearest Neighbor Algorithm tool for classification of incipient faults in induction motor drives, Computers & Electrical Engineering 54 (2016) 296-312. http://dx.doi.org/10.1016/j.compeleceng.2016.01.014 [65] J. Tian, C. Morillo, MH. Azarian, M. Pecht, Motor Bearing Fault Detection Using Spectral Kurtosis-Based Feature Extraction Coupled With K-Nearest Neighbor Distance Analysis, IEEE Transactions on Industrial Electronics 63 (3) (2016) 1793-1803. http://dx.doi.org/10.1109/TIE.2015.2509913 [66] DY. Dou, SS. Zhou, Comparison of four direct classification methods for intelligent fault diagnosis of rotating machinery, Applied Soft Computing 46 (2016) 459-468. http://dx.doi.org/10.1016/ j.asoc.2016.05.015 [67] QC. Jiang, B. Huang, XF. Yan, GMM and optimal principal components-based Bayesian method for multimode fault diagnosis, Computers & Chemical Engineering 84 (2016) 338-349. http://dx.doi.org/ 10.1016/j.compchemeng.2015.09.013
17
Technique of fault diagnosis of the single-phase induction motor was proposed Original method of feature extraction of acoustic signal called MULTIEXPANDED was described Three states of the single-phase induction motor were analysed Efficiency of acoustic signal recognition was analysed
18
MSAF-20-