ABSTRACT
It is well known that the friction vibration originates from the wear process of the tribological pairs, and thus, it can provide information about the wear states of the tribological pairs (Sun 2015). Therefore, the friction vibration can be used to monitor the changes in the wear states of the tribological pairs by complete feature extraction of the measured signals. Many tribology scientists performed thorough research for years from the aspect of the feature extractions of friction vibration (Li et al. 2003; Li & Peng 2009; Wan et al. 2009; Kang & Luan 2006; Chen & Zhou 2003; Chen & Zhou 2006; Zhu et al. 2007; Ji 2012). Chen & Zhou (2003, 2006) extracted the time-frequency characteristics of both TFV and NFV by methods of short-time Fourier transform, Zhao-AtlasMarks Distribution, and db4 Wavelet Transform. Zhu et al. (2007) performed researches on the fractal feature of NFV, studied the variations in the correlation dimension in different wear processes, and found that the Correlation Dimension of normal friction vibration was in decline in the “divergence” wear process, while there was an increase in the running-in wear process. In order to discuss the wear state of a piston-cylinder, Ji (2012) extracted the chaos features and the fractal characteristics of NFV, discovered that all of the Correlation Dimen-
sion, the Largest Lyapunov Exponent (LL E), and the Kolmogorov Entropy could reflect the friction and wear information of tribological pairs. It was also observed that both the Correlation Dimension and the Kolmogorov Entropy showed a “Positive Bathtub Curve” from the running-in wear process to a stable wear stage and then to a severe wear state, while the LLE showed a “Inverted Bathtub Curve”. As shown earlier, the friction vibration feature extraction was studied using the applications of wavelet transform, fractal, and chaos, and the friction vibration feature extraction was successfully achieved in a single direction. However, both the tangential friction vibration and the normal friction vibration can be excited simultaneously in the wear process, and they both contain the information of the variation in the wear state. Hence, it is incomplete when extracting the friction vibration feature in a single direction, and difficult to accurately reflect the wear state of the tribological pairs. Unfortunately, few have been done to research the coupling of TFV and NFV, and then extract the feature information, which can reflect the wear state of the tribological pairs.
