Bond graph modeling and experimental verification of a novel scheme for fault diagnosis of rolling element bearings in special operating conditions
详细信息 查看全文
- 作者：C. Mishra ; A.K. Samantaray ; samantaray@mech.iitkgp.ernet.in" class="auth_mail" title="E-mail the corresponding author ; G. Chakraborty
- 关键词：Rolling element bearing ; Bond graph model ; Variable and slow speed operation ; Condition monitoring ; EMD-based de-noising
- 刊名：Journal of Sound and Vibration
- 出版年：2016
- 出版时间：1 September 2016
- 年：2016
- 卷：377
- 期：Complete
- 页码：302-330
- 全文大小：3948 K
文摘
Vibration analysis for diagnosis of faults in rolling element bearings is complicated when the rotor speed is variable or slow. In the former case, the time interval between the fault-induced impact responses in the vibration signal are non-uniform and the signal strength is variable. In the latter case, the fault-induced impact response strength is weak and generally gets buried in the noise, i.e. noise dominates the signal. This article proposes a diagnosis scheme based on a combination of a few signal processing techniques. The proposed scheme initially represents the vibration signal in terms of uniformly resampled angular position of the rotor shaft by using the interpolated instantaneous angular position measurements. Thereafter, intrinsic mode functions (IMFs) are generated through empirical mode decomposition (EMD) of resampled vibration signal which is followed by thresholding of IMFs and signal reconstruction to de-noise the signal and envelope order tracking to diagnose the faults. Data for validating the proposed diagnosis scheme are initially generated from a multi-body simulation model of rolling element bearing which is developed using bond graph approach. This bond graph model includes the ball and cage dynamics, localized fault geometry, contact mechanics, rotor unbalance, and friction and slip effects. The diagnosis scheme is finally validated with experiments performed with the help of a machine fault simulator (MFS) system. Some fault scenarios which could not be experimentally recreated are then generated through simulations and analyzed through the developed diagnosis scheme.