Rotating Machinery Fault Diagnosis Based On SVM And Fuzzy Neural Network

Rotating machinery is the key component of industrial sector, its running statedirectly affect the production of the industrial sector. If it runs in abnormal state, lightwill cause huge economic losses, cause casualties and serious social impact. So it isimportant to state the monitoring and diagnose the fault of rotating machinery.This dissertation found the support vector machine (SVM) and fuzzy neural net-work fault diagnosis model. It is based on five kinds of typical mechanical state whichis simulated by rotating machinery fault experiment platform. Five kinds of mechanic-al state of the diagnosis is good. The five kinds of mechanical state is rotor normal,rotor unbalance, bearing inner ring cracks and bearing outer ring cracks. The mainconclusions are presented as follows:1. To analyze application of modern signal processing method and traditionalfault diagnosis, and restructure the signal by using frequency components which islarger amplitude, extract31characteristic value which is for restructuring signal. Atthe same time, correlation analysis for restructuring signal which is restructured thesignal for vibration acceleration signals of five kinds of five kinds of typicalmechanical state. The result is that the same type of correlation coefficient is close to1, belong to the strong correlation, and the different types of correlation coefficient issmall, belong to the weak correlation, shows that the extracted characteristic value hasbetter pertinence.2. According to binary tree support vector machine (SVM) classificationalgorithm for fault diagnosis model, it is difficult to determine the priority, this paperproposed the multiple support vector machine (SVM) classification algorithm todetermine the priority, and founded the binary tree support vector machine (SVM)model to use for fault diagnosis. The test result is good, shows that the method iseffective and has good engineering application value.3. The series connection of fuzzy neural network sequential fault diagnosismodel is build. The test result is good, shows that the method is effective. Then it iscompared with binary tree support vector machine (SVM) model for fault diagnosisresults, further shows that the binary tree support vector machine (SVM) theapplication value for the five state of mechanical fault diagnosis in this paper.