Using Correlation Dimension to Analyze the Signal of Microscale Crack Effect Induced Self-Excited Torsional Vibration on the Experimental Platform

Abstract

Signal feature extraction of self-excited torsional vibration had been a effective ways to identify the large-scale rotating machinery dynamic fault characteristics and analysis. For solving the problem of feature extraction of signal on transient impact torsional vibration, correlation dimension was studied to be used for signal processing. Based on fractal theory, self-similarity statistical feature of signal had fractal characteristics in a certain scale, and it could be used for extracting non-stationary signal feature of the complex nonlinear system, also be for analyzing the dynamic characteristics of nonlinear, complex nonlinear or chaotic of transient torsional vibration signals of the rolling mill’s main driving system. Firstly, using principles of physical similarity of similar engineering, our study work set up experimental platform to simulate the self-excited torsional vibration induced by microscale effect of crack of the rolling mill’s main driving shaft. Secondly, correlation dimension about fractal had be used for analysis of nonlinear and non-stationary dynamic torque signals of the experimental platform. The result of theoretical analysis and experimental showed that correlation dimension to be as quantitative or qualitative analysis of dynamic fault characteristics of the vibration systems was effective. At last, the result could be concluded that correlation dimension could reflect information of fault state or dynamic characteristics. Correlation dimension was smaller in normal state than that of in fault state associated with a larger dimension. It should be convergence of to judge which system was deterministic or stochastic. Under varying degrees of fault, correlation dimension was with the obvious difference, which was with the deepening of fault, the correlation dimension would gradually increase.