Least Square Fitting for Adaptive Wavelet Generation and Automatic Prediction of Defect Size in the Bearing Using Levenberg–Marquardt Backpropagation

详细信息    查看全文

• 作者：Anil Kumar ; Rajesh Kumar
• 关键词：Bearing defect size prediction ; Continuous wavelet transform (CWT) ; Adaptive wavelet ; Time marginal integration (TMI) ; Undecimated wavelet transform (UWT) ; Artificial neural network (ANN) ; Levenberg–Marquardt back propagation
• 刊名：Journal of Nondestructive Evaluation
• 出版年：2017
• 出版时间：March 2017
• 年：2017
• 卷：36
• 期：1
• 全文大小：
• 刊物类别：Engineering
• 刊物主题：Structural Mechanics; Characterization and Evaluation of Materials; Vibration, Dynamical Systems, Control; Classical Mechanics;
• 出版者：Springer US
• ISSN：1573-4862
• 卷排序：36

文摘

In this communication, an attempt has been made to develop an algorithm for automatic prediction of the size of the bearing defect during operation of a machine. Features for the purpose are meticulously designed so as defect commencement and termination events in the signal could be easily spotted. Information on commencement of defect in the signal, in general, is very weak. It is enhanced by approximating the burst in the signal to a wavelet, making use of least squares fitting. Levenberg–Marquardt back propagation network is used for prediction of defect size from defect features. The comparison shows that the Levenberg–Marquardt back propagation network outperforms another network in terms of accuracy. The experimental validation of the proposed scheme is carried out for four different defect sizes each for the inner race, outer race, and roller defect. The maximum deviation in the width measurement result is 5.35% which occurs in the case of bearing with roller defect of width 1.12 mm. The performance evaluation of the method is also carried out using t test. The result of t test validates the accuracy of proposed method in the prediction of defect width.