Investigation On Intelligent Recognition Method Based On Feature Extraction Of Hilbert Spectrum And Application

Machinery fault diagnosis is a comprehensive technology. As a matter of fact it is pattern recognition of machine operating conditions. The key is feature extraction which is also the most difficult thing. In some way the success of the fault diagnosis is up to the feature extraction. So it is particular interest to the machinery diagnosis.Based on the actual engineering, the application of Localwave in the feature extraction for non-stationary signal is researched in this dissertation. Compared with other methods, Localwave method has many advantages. But analysis results may be influenced because of mess of time-frequency spectrum. And 3-D diagram of time-frequency spectrum can be affected easily by the observation angle of view. Other more, the identification of the time-frequency spectrum mainly depends on the expert experiences, which limits engineering application of Localwave.To counter these cases, Hilbert time-frequency spectrum(HS) energy barycenter method is discussed to extract the feature for signal in this dissertation. This method is expected to realize quantitative analysis for the time-frequency spectrum. Experiments are carried out to prove the method. And they are done in the different frequencies which are 10Hz, 15Hz, 20Hz and 25Hz. Firstly, datum of bearing are gathered and analyzed when the bearing works under normal condition, inner ring condition and outer ring condition. Secondly, the energy barycenter of HS is calculated. At last energy barycenters are classified and identified by support vector machine. Results of the experiments show that ratio of right classification is high. So it is clear that the method to extract feature by HS energy barycenter is reasonable. Meanwhile, in order to do comparison, time domain index is extracted to do the test, and the result can also prove reliability of this method.Base on HS spectrum energy barycenter method, a fault diagnosis system is developed using LabVIEW. It proves a new method for intelligent fault diagnosis of machinery.