摘要
将描述混沌运动的关联维数用于大机组振动信号的分析 ,并针对故障诊断的实际情况 ,从点间距的计算、动态关联和标度区的判断等方面对关联维数的 G- P算法作了一定的改进。分析结果表明 :由于不同故障的动力学产生机制不同 ,通常也具有不同的关联维数。因此 。
The application of correlation dimension in fault diagnosis for large rotating machinery is reported in this paper. The shortcomings of the traditional G P algorithm are pointed out and modified version is recommended. Firstly, the traditional G P algorithm uses the Euclidean norm to compute the distance between two reconstructed vectors while this paper presents another norm to compute the inter distance between two points. This alternative norm is found to be more practical in on line diagnosis because the computational time can be remarkably reduced. Secondly, a cutoff parameter t min >1 is introduced to avoid dynamic correlation, which can improve the convergence of the standard correlation algorithm and can remove the influence of the locally one dimensional behavior of the trajectory. Finally, this paper suggests that one should plot the D 2(m) ln (r) function to determine the beginning and the end of the scaling region. Non linear time series analysis theory based on correlation dimension for practical application, especially for fault diagnosis of large rotating machinery , is introduced in this paper. The results show that the correlation dimension can usefully reflect the different kinematics mechanisms.
出处
《振动工程学报》
EI
CSCD
2000年第2期229-234,共6页
Journal of Vibration Engineering
基金
中国博士后基金资助项目
关键词
故障诊断
非线性
混沌
关联维数
大型旋转机械
fault diagnosis
nonlinearity
chaos
correlation dimension
large rotating machinery
