Research On The Methods Of The Rough Set And Neural Network In Fault Diagnosis Of Polymerizer

The recent research status of the technology of fault diagnosis and the development of intelligent about fault diagnosis are analyzed in this paper. Various existing fault diagnosis methods are also analyzed, and the basic theories of fault diagnosis are described. Finally, the existing problems and limitations are pointed out.Polymerizer is an important equipment of PVC production line. The stability of the polymerizer derectly affects the production, and the fault of the polymerizer is the main reason of the decline of the quality of PVC. Therefore, the exactness of fault diagnosis of the polymerizer is the key to the quality of the PVC products, so the fault diagfnosis are of great importance. Taking the70m3polymerizer of Jinxi resin factory as an example, rough set theory and neural networks are combined to solve the problem of fault diagnosis:Firstly the improved algorithm based on discernibility matrix of decision-making is used to abtain more concise optimal diagnostic rules in the polymerizer. Secondly the BP neural network for fault diagnosis can be set up based on the optimal decision-making system. Taking the large amount of properties and data into account, the paper deals the discernibility matrix with improved algorithm and highlights the superiority of the improved algorithm in dealing with large amount of data. The results prove that the improved algorithm is simpler and the network structure is simplified.The algorithm of attributes reduction based on the discernibility matrix is improved in this paper, which simplified the elements that contain a single attribute. The result is that the calculated amount is reduced and the process is simplified. The improved algorithm of discernibility matrix attribute reduction is conbined with the BP neural network, and this method is used in the fault diagnosis of the polymerizer. The result of the experiment is proved in MATLAB, and it shows that the proposed method has high accuracy rate and short training time, and all of this reflect the efficiency and effectiveness, which effectively illustrates the advantages of this method.