Study On The Application Of Fractal Theory In Rotating Machines Fault Diagnosis

In this paper a automatic method for the correlation dimension estimation was introduced,which can improve the precision and speed of the correlation dimension calculation comparedwith traditional Grassberger and Procaccia algorithm. Firstly, the influence of noise on thecomputation of correlation dimension was discussed, and the iterative singular valuedecomposition (ISVD) method was introduced. Secondly, the calculation of the minimalsufficient embedding dimension and selection of lag time can be automatically carried out byusing the false nearest neighbors and across displacement respectively. Thirdly, in order todistinguish the scaling region of the bi-logarithmic plot of the correlation integral (log-logplot) ,a methods by means of searching for the widest intervals in the derivative of the log–log plot was proposed. The effectiveness of the introduced method was tested by means of thecalculation of well-know models as Logistic attractor and Henon attractor. As a typicalexample, this automatic method was applied to real vibration signals of different rotatingmachines. Analysis result demonstrates the applicability of this method in distinguishing themachines conditions with normal status and default status.