Research On Improved Weighted Support Vector Machine And Application In Fault Diagnosis Method

Chemical industry is highly continuity of production process, to ensure safety and stability is a very important issue, and it has attracted more and more attention. Along with the development of the application of modern computer monitoring system, chemical enterprise has accumulated abundant data of process variables and control variable in production process.For we employ the small samples data in the real project, and at present some commonly used fault diagnosis method are based on the large samples data. How to make full use of these small samples data for fault diagnosis of chemical process become a hot research issue now. According to the characteristics of chemical reaction process ,with a chemical process simulation: Tennessee-Eastman process (TEP process) as the research object, this paper use the intelligent information processing technology combining with theory of Principal Component Analysis,Independent Component Analysis and classification algorithm based on weighted support vector machine (SVM) to implement the fault diagnosis of chemical process.Besides,to effectively solve the problem of the Small Sample fault recognition in fault diagnosis, due to the problems of results of the classification get by support vector machine biased toward large sample classification for large Scale difference in the training data, In this paper a method of small-sample mechanical failure identification by weighted C - SVM classification algorithm is presented. This method improves the defects of C-SVM in small sample identification by weighting coefficient to different categories together with class-incremental learning and considering the practical situation, realizes the intelligent classification of multiple-fault. Simulations of classification test in robot execution failures are used to verify this algorithm and validate the classification accuracy.Finally, according to the characteristics of TEP process, using the Improved Weighted Supported Vector Machine fault diagnosis system to diagnosis the TEP process respectively by principal component analysis and independent component analysis as preprocessing method. The simulation results improve the rapidity of fault diagnosis, and achieved good effect.