Research On Support Vector Machine Method For Rotating Machine And Intelligent Fault Diagnosis System

Research on the large rotating machine fault diagnosis is important for avoiding mass economic loss and disaster accident. In this dissertation, on the basis of the "Research on New Method for Rotating Machine Faults Diagnosis Based on Independent Component Analysis" (National Nature Science Fund Project, No:50205025), we have studied the theory of SVM(Support Vector Machine ), fault diagnosis recognition system and the method for rotating machine faults recognition based on SVM. Then we have made some simulations to testify our research result. In the end, we have realized the acquiring vibrational signals based on embedded system and developed a prototype of rotating machine faults' pattern recognition system based on SVM using MATLAB and LabVIEW. The details are studied as follows:Chapter one discusses the importance of the monitoring and diagnosis technology for large rotating machinery in modern manufacturing and summarizes the research status and development trend of monitoring and diagnosis system of large rotating machinery. We also describe the development of the theories and applications of SLT(Statistical Learning Theory) and SVM. At last the goal and main contents of this dissertation are presented.Chapter two describes the kernel content of SLT and analyzes the structure of the SLT and the problems needed to be solved. Then SVM and main idea of its algorithm are introduced in fault diagnosis domain. We address a structure of the pattern recognition system based on SVM and introduce the components of the system.Chapter three addresses the usual rotating machine's vibral fault mechanism and studys the classified problem of rotating machine fault pattern. On the basis of the rotor test bench, we compare three muti-class SVM classifiers used in rotating machine faults pattern recognition and draw the conclusion that the DAG classifier method is more suitable for rotating machine fault diagnosis.In chapter four, according to rotating machine fault diagnosis and pattern recognition system's requirement, we design a embedded system for rotating machine vibral signal acquiring using doubled CPU architecture of ARM and DSP to realize high speed and full period data acquiring.In chapter five, we take the advantage of MATLAB and LabVIEW and build a prototype system of rotating machine faults recognition based on SVM by mix programming using COM component technology.Chapter six sums up the work of the dissertation and presents conclusion and prospect.