Dynamic Pattern Recognition Method Research And Applications

The fault diagnosis technology for rotating machinery is developing rapidly from theory research to practical application in parallel with the new achievement of modern science and technology,and becomes a new interdisciphne composing of mathematics,computer,signal processing and artificial intelligence,etc.Rolling bearing is an universal part of rotating machinery.Therefore,fault monitoring and diagnosing of rolling bearing is a hot research in the advanced mechanical field.Most of the traditional pattern recognition methods deal with the static problem without the consideration that object may take some short-term changes.In this paper,we study the dynamic model of class methods through static and dynamic samples.Dynamic clustering algorithm in general are directed at static sample data,the clustering results not only depend on the initial classification,fall into local minimum easily,but also not very satisfactory for the results of the classification from a classification system without clear boundaries.In such cases,this paper improves an unsupervised optimal fuzzy clustering algorithm which is a very good solution to the above problems.However many data are dynamic,such as some mechanical problem,large databased and information processing on the Internet and so on,the data is a curve that changes along the timeline.In this paper,combine with the wavelet analysis and the unsupervised optimal fuzzy clustering algortihm,we have got a better recognition effect.The main research work of full text are as follows:(1) The unsupervised optimal fuzzy clustering algorithm has been improved,by modifying the constraint conditions,which has reduced the computational complexity;by increasing a step which can discriminate the Local optimum and global optimum,then it can avoid falling into local minimum.The simulation results have proved that the improved method is more effective and has a higher recognition rate.(2) This paper introduces some dynamic pattern recognition algorithms systematically,such as recurrent neural networks,dynamic time warping,hidden Markov model,and dynamic fuzzy pattern recognition algorithm and so on,which can deal with dynamic samples.Then it summarizes the advantages and disadvantages of each.(3) The wavelet analysis method is introduced into rolling beatings intelligent fault diagnosis,the wavelet packet denoising is used to the eigenvector extracting is presented.A method which combined wavelet analysis with improved unsupervised optimal fuzzy clustering algorithm appling to the rolling beating fault diagnosis is presented.The results show that the proposed method can recognize the fault pattern accurately.