Research On Optimal Diagnosis Method Of Rotating Machinery System

The research on new rotating machinery fault diagnosis technology is important and meaningful. Sliding bearing rotor system is a common part of rotating machinery, and its shaft centerline orbit contains a wealth of fault information. But shaft centerline orbit of multi-component vibration signals is complex, and it is hard to get the feature of the shaft centerline orbit directly. In order to gain fault information, a new method of shaft centerline orbit purification is presented. And then combine with the image recognition based on moment invariant and the correlation recognition method to diagnose the failure.The main work are as follows.1 A new fault signal processing method is proposed to decompose the multi-component vibration signal to several single-component signals that possess physics sense. Firstly, a sampling frequency range is gained according to the extreme points of the signal. Secondly, take samples at a frequency on the signal to gain sampling mean, and find out the maximum sampling mean in the sampling frequency range. That maximum sampling mean is the amplitude of a single-component and its sampling frequency is the frequency of the single-component. Lastly by changing the first sampling point and sampling length, the time domain wave of the single-component signal in the original signal and its distribution can be got. The theoretical basis of the method and the estimation method of frequency and time domain are introduced, including its decomposing steps.2 Apply this method to the shaft centerline orbit purification. By decomposing the rotating machinery vibration signals, some single-component signals can be obtained to gain shaft centerline orbit in different frequencies. Simulation calculation of some common faults have proved the effectiveness of the method.3 Extract the feature of shaft centerline orbit with the method of the image recognition based on moment invariant. Converter the shaft centerline orbit into two dimensional image, and get the invariant moments of the image as the shape features of the shaft centerline orbit.4 Identify the shaft centerline orbit with the correlation recognition method. Gain the invariant moments of the model to be tested, and compare with the moments of the reference model to get the correlation degree. Identify the failure based on correlation degree. Simulation calculation and actual practice have proved the effectiveness of the method.