Railway Rolling Bearing Faults Diagnosis Based on Wavelet Packet and EKF Training RBF Neural Network

• Main Authors: Xing WANG, Yuan-Jing ZHAO
• Format: Article
• Language: English
• Published: IFSA Publishing, S.L. 2013-11-01
• Series: Sensors & Transducers
• Subjects: Wavelet packet; Extended Kalman filter; RBF neural network; Railway rolling bearing; Fault diagnosis.
• Online Access: http://www.sensorsportal.com/HTML/DIGEST/november_2013/PDF_vol_158/P_1522.pdf
• ISSN: 2306-8515; 1726-5479

Based on wavelet packet and extended Kalman filter (EKF) training RBF neural network method, a method for the fault diagnosis of railway rolling bearing is proposed in this paper. The wavelet packet and RBFNN are introduced. The wavelet packet is used to translate raw vibration signals of a railway rolling bearing into time-scale representation. Then, the wavelet packet energy eigenvector is constructed, next, those wavelet packet energy eigenvectors as fault samples for training RBF neural network. To ameliorate the algorithm, EKF is exploited to optimize the algorithm so as to determine the best values for "network connection weight", finally the fault patterns of the railway rolling bearings are identified. The results show that the proposed method is superior to the RBF neural network in extracting the fault characteristics of roller bearings. This method is effective and can be used for automotive recognition to rotary machine faults.