Filtering And Pattern Recognition Method Of Rotor Fault Signals

In recent years, with the development of machinery industry and computer technology, the research about equipment condition monitoring and fault diagnosis more and more subject to various aspects of attention, then theoretical study and practical application have to catch up with schedule. In the uncertainty of the system to solve the small sample of the pattern recognition problem, the gray system theory showed a good performance. It studies uncertainty system, include "small sample"of section information is know and unknown,"poor information". Mainly through the section known information to generate, develop, extract valuable information, and run on the system behavior, the evolution of the correct description and effective monitoring.In this paper, the common faults of the rotor system on the rotor test equipment were simulated. Use discrete gray-scale morphological filters to filtering signal of fault vibration. After, frequency domains transform and calculate the gray correlation of frequency-domain amplitude, discriminant matrix compared with the standard failure mode to distinguish the type of fault. Research and experimental analysis of the above methods, the major work done and conclusions are as follows:(1) In the stage of laboratory experiments to simulate the normal state of imbalance, misalignment, rub, bearing loose four typical fault. Aanalyzed the vibration signal of the four failure mechanism and failure filter denoising, spectral analysis, pattern recognition analysis. Which focuses on the signal filtering denoising and pattern recognition, then discussion the slightly superior of method about conventional filtering of fault diagnosis and pattern recognition.(2) In order to overcome the traditional morphological filters, a single opening and closing of the expansion of operations, anti-expansionary, use open-close and close-open computing two cascade combination of discrete grayscale morphological filters on the fault signal is filteredstudy. Through analysis of the original signal and filtered signal waveform and spectrum, prove that the discrete grayscale morphological filter for rotor vibration signal filtering with good results.(3) In order to achieve accurate and quick to distinguish the various types of composite failure, the concept ofgray correlation of frequency-domain amplitude with the fault pattern recognition have been proposed. It is in the standard failure mode and test signal calculation process, the two signals to determine the degree of geometric similarity of the data sequence polyline; this data column is the amplitude vector consisting of the frequency spectrum after feature extraction. Preliminary experiments the amplitude of frequency-domain amplitude is too large, resulting in a certain degree of fault types of discrimination error. Later, during the dimensionless normalized before calculating the relevancy signal to control the signal amplitude range between (0,1], to improve the recognition accuracy.(4) Since this study is an overall system, all in the rotor test rig to complete the experimental data collection, based on the rotor test bed developed a set of control data processing system is particularly important. Microsoft's C sharp language selected through needs analysis to prepare a set of Windows Forms applications, to achieve a test-bed and its lubrication system start-stop control, speed adjustment, the ADLINK DAQ2214data acquisition card sampling parameters set, the vibration signal waveformreal-time display, the signal data processing and other basic functions.