V.V. Kondrat\'ev - Corresponding Member of the Russian Academy of Sciences, Dr.Sc. (Eng.), Professor, Nizhny Novgorod State Technical University n.a. R.E. Alekseev. E-mail: vyachkon@sandy.ru V.A. Utrobin - Dr.Sc. (Eng.), Professor, Nizhny Novgorod State Technical University n.a. R.E. Alekseev. E-mail: utrobin-va@yandex.ru N.N. Makarov - Ph.D. (Eng.), Associate Professor, Nizhny Novgorod State Technical University n.a. R.E. Alekseev. E-mail: Maknik46@yandex.ru V.E. Gai- Ph.D. (Eng.), Associate Professor, Nizhny Novgorod State Technical University n.a. R.E. Alekseev. E-mail: iamuser@inbox.ru

In this work is considered vibration-based diagnostics method the state of rolling element bearing of engines on the basis of the theory of active perception. A classification system of the bearing\'s condition by its vibration signals can be represented as a system of image recognition. From the aspect of system analysis the problem of recognition (to wide extent) is a scope of three stages: preprocessing, calculation features and decision-making. There are known problems which linked with application of existing methods of pattern recognition solving vibrodiagnostics problems: the problem of initial description formation, the problem of sign\'s system formation, and problem of decision making under uncertainty a priori. Using the methods of the theory of active perception can solve the described problems. This paper deals with the application of this theory to the analysis of vibration signals. In the paper is proposed implementation of formation stages of feature system and decision making. The implementation stage of calculating the feature descriptions of the vibration signal consists in the following: 1) vibration signal readings are divided into a multitude of segments 2) to each segment is used U-transform (U-transformation is a basic in the theory of active perception), consequently forming a spectral representation of each segment 3) by calculated spectral presentation of segment determines a closed groups 4) on the basis of closed groups, calculated on the previous stage, forming a histogram of closed groups, which is indicative of the description of vibration signal. It is used 840 closed groups to create a features description of vibration signal. Stage of classification is based on the method of template matching (see. Fig. 2 is used measure of proximity Euclidean distance). As an input signal class is chosen the class of that standard, to which the distance from input signal was the least. A computational experiment was carried out using a database of vibration signals presented by Case Western Reserve University Bearing Data Center. This database contains records of rolling bearing vibration signals with defects (on outside and inside track and on the rolling body), and without defects. It was carried out testing of existing and proposed method of the bearing condition detection. It is followed from these results that for twenty-one class of possible states of bearing one hundred percent classification accuracy is only achieved by using the proposed of recognition method. The influence of various method parameters on the accuracy of solving the problem of vibration diagnostics is considered. The proposed method ensure a high precision the state\'s classification of rolling element bearing by vibration signal, as compared with existing methods.

 

1. Panov V.A. Vibroakusticheskaja diagnostika nerazbornykh podshipnikov pri remonte aviacionnykh dvigatelejj // Dvigatel. 2011. T. 76. № 4. S. 28-31.
2. Nelwamondo F.V., Marwala T., Mahola U. Early classifications of bearing faults using hidden Markov models, Gaussian mixture models, mel-frequency cepstral coefficients and fractals // International Journal of Innovative Computing, Information and Control. 2006. V. 2. №. 6. P. 1281-1299.
3. Boldt F. de A., Rauber T.W., Varejão F.M. Feature Extraction and Selection for Automatic Fault Diagnosis of Rotating Machinery // X Encontro Nacional de Inteligência Artificial e Computacional (ENIAC). Fortaleza, Ceará. 2013. P. 213-220.
4. Al-Raheem K.F., Abdul-Karem W. Rolling bearing fault diagnostics using artificial neural networks based on Laplace wavelet analysis // International Journal of Engineering, Science and Technology. 2010. V. 2. № 6. P. 278-290.
5. Baranov V.G., Kondratev V.V., Milov V.R., Zaripova JU.KH. Nejjrosetevye algoritmy raspoznavanija obrazov // Nejjrokompjutery: razrabotka, primenenie. 2007. № 11. S. 20-27.
6. Utrobin V.A. EHlementy teorii aktivnogo vosprijatija izobrazhenijj // Trudy Nizhegorodskogo gosudarstvennogo tekhnicheskogo universiteta im. R.E. Alekseeva. 2010. T. 81. № 2. S. 61-69.
7. Gajj V.E. Metod ocenki chastoty osnovnogo tona v uslovijakh pomekh // Cifrovaja obrabotka signalov. 2013. № 4. S. 65-71.
8. Gajj V.E., Ajjnullin D.T. Algoritm segmentacii zvukovogo signala po priznaku ton/shum v uslovijakh apriornojj neopredelennosti  // Nejjrokompjutery: razrabotka, primenenie. 2014. № 11. S. 32-37.
9. Baranov V.G., Milov V.R., Zaripova JU.KH., EHpshtejjn A.JU. Intellektualizacija sistemy raspoznavanija obrazov na osnove sravnenija ehffektivnosti metodov klassifikacii // Informacionno-izmeritelnye i upravljajushhie sistemy. 2010. T. 8. № 2. S. 35-39.
10. Baranov V.G., Kolcov V.A., Milov V.R., Milov D.V., Gajj V.E. EHkspertnaja sistema ocenki sostojanija ehlementov transportno-tekhnologicheskogo oborudovanija v processe peregruzochnykh operacijj // Informacionno-izmeritelnye i upravljajushhie sistemy. 2013. №7. S. 67-71.