The application of pseudo-phase portrait in machine condition monitoring

This paper presents a robust phase space reconstruction method based on singular value decomposition technique and its applications to large rotating machine and gear system condition monitoring and fault diagnosis. The singular value decomposition is used to determine the effective embedding space and to reduce the noise level of a measured vibration signal. Following the singular value decomposition, a pseudo-phase portrait can be obtained in the effective embedding space. This pseudo-phase portrait is then used to extract qualitative features of machine faults. Experience has shown that when one compares the pseudo-phase portraits obtained under different machine conditions, it is often possible to detect major differences due to different dynamic and kinematic mechanisms. In the case of gear system condition monitoring, correlation dimension has been introduced to evaluate these differences in order to obtain more accurate and reliable diagnosis. The pseudo-phase portrait is conceptually simple and has been found to be sensitive to some fault types. It is promising therefore that such pseudo-phase portraits can be used to realize real-time, online computer-aided diagnosis of machine faults. (C) 2002 Elsevier Science Ltd. All rights reserved.