The Study Of Mathematical Morphology Filtering And Quantized Character Extraction For Vibration Signals In Rotating Machinery

Extraction of vibration signals from mechanical faults is more commonly in mechanical vibration diagnosis. Fault diagnosis involved three parts of the content, that is, signal processing, feature extraction, fault identification. Because of empirical data collected strong noise, making the nature of the signal can not accurately reflect the characteristics of mechanical failure, if the effective filtering to be designed, it will greatly affect the feature extraction and accuracy. Fractal theory is a new mathematical ideas, it's good at characterizing the self-similar and irregular objects, in dealing with nonlinear systems with unique advantages. Rotor test bed for the data model, firstly, de-noising of data by the generalized morphological filter, and then start the feature extraction research to the signal used by the correlation fractal dimension. Main contents and conclusions of the study were as follow:(1) Based on searching a lot of domestic and foreign literature about signal processing, the generalized morphological filter is designed to de-noising the signal. In the design process, mainly discussed the generalized morphological filters structure elements configuration problems before, through the simulation experiments, analyzed different amplitude and width of the structure of the filter element structure, through comparing the filter ability of their respective filtered frequency domain and time domain and signal-to-noise ratio figure find that width and amplitude of the latter structure elements should be twice of the former. But, width is more significant than amplitude.(2) The purity of the vibration signal is obtained, and we can do feature extraction by fractal theory, correlation dimension was be selected as fault type of characteristic, analyzed the affect factors of correlation dimensions accuracy, and explained how to selected parameters of correlation dimensions. G-P algorithm is improved that a relatively stable embedding dimension instead of correlation dimensions. In this way, the blindness of choosing correlation dimensions can be avoid successfully.(3) A simple fault diagnosis system should be developed by MATLAB GUI technology, the system's main function include generalized morphological filters and fractal analyzed, used the system can facility accomplish filtering and feature extraction work. Study found that mathematics morphology and fractal analysis in fault diagnosis has many advantages, will certainly to raise a research boom.