| Bearing is the most commonly used components in mechanical equipment,and applied widely,especially the rolling bearing plays a more and more important role in the application,in the same way in the all kinds of fault of bearing,a proportion of rolling bearing fault is becoming more and more big.As the concept of industrial 4.0 major technical equipment in China will have greater development,and the stability of the mechanical equipment,accuracy,reliability and economic indicators,both to the bearing.Bearing fault is not allowed may cause bearing waste or serious production accidents,therefore,to study the fault diagnosis of rolling bearing has its necessity and importance.For the fault diagnosis of rolling bearing mainly includes signal acquisition,feature extraction,state recognition,diagnosis,analysis,decision-making intervention five links,including feature extraction and state recognition is the core of the bearing fault diagnosis.Traditional diagnostic methods include the shock pulse method,resonance demodulation method,time domain average method and the frequency spectrum analysis method and so on,modern diagnostic algorithm with wavelet analysis and neural network,fuzzy theory and expert system and particle swarm optimization(pso)algorithm,etc.Wavelet transform has good time-frequency local analysis ability,neural network has to handle complex multimodal and lenovo,conjecture and memory function,it is based on the research content of this article is put forward.Firstly,the wavelet transform is used for bearing vibration signal filtering,and then optimize the traditional soft and hard threshold function,design a new threshold function,wavelet threshold denoising on rolling bearing.In using energy normalization method to deal with the fault features,the normalized energy feature vectors as input of neural network,finally,using neural network to the fault characteristics for identification,improve the identification accuracy of bearing failure. |
PDF Full Text Request