Early Detection of Bearing Failure Using Least Squares.pdf

第十五屆中華民國振動與噪音工程學術研討會 中國文化大 中華民國九十六年一月二十四日 中華民國振動與噪音工程學會 Chinese Society of Sound and Vibration Early Detection of Bearing Failure Using Least Squares Harmonic Analysis Chao-Shih Liu1 and Per G. Reinhall2 1Department of Power Vehicle and Systems Engineering, Chung-Cheng Institute of Technology, National Defense University, 2Department of Mechanical Engineering, University of Washington, USA E-mail: chao@.tw Abstract The early detection of bearing failure requires methods that are sensitive to impulsive signals and their changes. The difficulty of early bearing defect detection is that the signal component generated by early defect is very small compared to other vibration sources. One of the most popular techniques to detect early bearing failure is the high frequency resonance technique (HRFT). However, the performance of the HFRT deteriorates at low SNR level. When the SNR level decreased to 0.1, the individual pulses can not be identified from the envelope of the signal. We propose a least square harmonic analysis algorithm (LSH) to enhance the performance of HFRT. The LSH has been widely applied in speech coding, compression, enhancement, and synthesis. The idea of the harmonic modeling is to represent a signal by a harmonic component and a noise component. Since an impulse train can be approximated as a finite number of harmonics, we can use harmonic analysis as a pre-processing tool to extract the impulse component generated by bearing defects from a noisy signal. Both the simulation and experimental data show that LSH can increase the sensitivity of HFRT for early bearing defect detection. Keywords: HFRT, harmonic analysis, bearing defect detection, least square algorithm 1. Introduction Nearly all pumps and motors of different sizes operate with rolling element bearings and bearing failure accounts for 51% of electric motor failure [1]. In addition, bearing defects can also be a warning signal for other fault symptoms becau