Method Of Rotating Machinery Fault Diagnosis Based On Wavelet Packet And Support Vector Machine

Rotating machinery is the most widely used in industrial sector. The fault diagnosis research of large-scale rotating machinery is essential to avoid huge economic losses and the occurrence of disasters. It has important significance. In this paper, combined with Jiangxi Province Natural Science Foundation project"Study of Some Decisive Problems of Intelligent Mechanical Fault Diagnosis Based on Support Vector Machine(0650054)", intelligent fault diagnosis method was studied based on support vector machine theory. Taking rotating machinery as the research object, the author has studied the theory and algorithms of support vector machine, fault diagnosis system, and fault classification method of rotating machinery based on support vector machine, which has been verified by simulation. The rotating machinery vibration data acquisition system has been designed. In the Matlab platform, rotating machinery fault classification prototype system has been built on the basis of support vector machine. Major researches of the paper are as follows:1. Discussed the background and significance of the research. Analyzed the development and application of the support vector machine theory and the feasibility, advantages and shortcomings of its application to mechanical fault diagnosis. Gave an overview of support vector machine research at home and abroad. Finally, some main contents of this article were introduced.2. Discussed and analyzed the main content of machine learning, statistical learning theory and the basic idea of support vector machine algorithm, which were introduced to machine fault diagnosis field. The basic steps and methods of support vector machine applied to machinery fault diagnosis were given.3. For the standard support vector machine cannot be directly used to solve fault diagnosis of such a typical multi-valued classification problems, papers made use of decision directed acyclic graph of the multi-valued classification algorithm to establish a multi-class fault classifier model.4. Classification performance of fault classifier has great relation with the support vector machine kernel function parameters. Taking the Fisher discriminant function as the objective function, the kernel function parameter optimization theory is researched. And the kernel function parameter optimization theory is proposed based on Fisher criterion and the improved genetic algorithm. The method using improved genetic algorithm to optimize the nuclear parameters can identify the most advantage of the global nuclear parameters and improve the classification performance of classifier.5. Using Bently rotor test bed, the rotor test stand vibration data acquisition system has been designed, and some typical rotating machinery failure are simulated, which provide accurate diagnostic data for fault feature extraction of rotating machinery. Then making simulated fault data as the diagnosis object, the author has studied wavelet packet decomposition of fault feature extraction method. Combined with the advantages of Matlab platform, rotating machinery fault classification prototype system has been built based on support vector machine, .and casing study and verification.6. Summed up the preceding work and further research work is prospected.