Rotor Fault Feature Extraction Method Based On Fractal And Test System Improvements

Effectively quantify feature extraction, particularly non-linear and non-stationary signals quantified characteristics extraction is great significant to diagnose the fault status and fault extent of rotor system accurately. Using conventional time-frequency domain signal analysis methods can not meet the needs of the rotor system status monitoring and diagnosis. Fractal geometry is a new analytical method which can meet the analysis of complex fault signal and have been widely used in the natural sciences and many other fields. In the field of rotating machinery fault diagnosis, people began to use fractal geometry to analysis the vibration signal and achieved some results. But only rely on fractal geometry can not effectively extract the fault feature of the rotor system. Based on this problem, this paper introduces two graphic geometric parameters tightness and abundance together with fractal geometry to study the characteristics of rotating machinery vibration signal extraction problem. Research contents and conclusions obtained are as follows:1) In the theoretical system of fractal geometry, several different methods to estimate the fractal dimension were comparative analyzed in advantages and disadvantages exist on quantify extracting characteristics of vibration signal. With theoretical analysis and experimental verification, some advantages and disadvantages were obtained by comparing box dimension estimation method with several other classic fractal dimension estimation method of quantify extract fault feature of vibration signal.2) For box dimension analysis method uses covering method, the operation method is using the same size box to cover the signal. As the box is the same size, the signal coverage may be uneven and results inaccurate estimates of the required box numbers thereby affecting the estimation accuracy of the fractal dimension. Aim at the shortage of this method, mathematical morphology was introduced. Using its basic operation method to calculate the fractal dimension of the signal, it was called fractal dimension estimation method based on mathematical morphology. This is a new method to estimate the fractal dimension. This method takes the vibration signals as one-dimensional signals and deal with mathematical morphology method. It can avoid the shortcomings of the box dimension estimation method. Therefore, this method has lower computational complexity, greater stability and higher accuracy than box dimension estimates.3) Two geometric parameters compactness and abundance were introduced to fault diagnosis of the rotor system. Two geometric parameters joint with the fractal dimension based on mathematical morphology to extract the fault feature of the fault vibration signals of rotor system. The extracted fault feature was deemed as neural network input characteristic quantities to classification several different running status of the rotor system and achieved good classification results. The result shows that quantify feature extraction method used in this study showed excellent performance in multiple fault diagnosis.4) The function of double cross rotor bench vibration testing and feedback control software platform was extended with the virtual instrument technology. The function was more completed. Mechanical vibration signal testing and analysis system were formed including a set of vibration signal analysis and display, vibration trend display, vibration data and feature data storage, real-time status monitoring, early warning. The system achieved good results in experimental research and engineering applications.The research shows that the unite fractal dimension based on mathematical morphology with compactness and abundance to extract the quantify feature of vibration signal can obtain more accurate feature information than the traditional method. It also can provide a new way for the rotor system fault feature quantify extraction of non-linear, non-stationary vibration signals.