| At present, it is very easy to get a vast amount of process parameter data, however, it takes both time and efforts to process these data in the traditional way. Data mining is to'find these valuable information from the vast amount of data. The typical knowledge we can find through data mining technology, which mainly includes: association rules, classification and cluster analysis. Data mining technology has been widely used in the business sphere, but seldom used during the industry process control. This paper attempts to provide a view of applying the cluster analysis to the industry process control. The important research is about the theory and methods of the cluster analysis in view of statistical theory, the theory and methods of fuzzy cluster analysis, the FKN' s structure and the FKN's study algorithm (FKN, fuzzyKohonen network)-the organic fusion of the fuzzy c-means algorithm andself-organized feature map neural network. The paper proposes the IFKN (Improved FKN) on the basis of the hard classification idea and the soft classification idea, then carries on the cluster analysis of the artificial synthetic control chart time series through Matlab program and tt?cluster result matches the cluster result of the famous DataEngine' s software of the intellectual data analysis and data mining from German MIT Company. Finally, the paper discusses the applying of the cluster analysis to the control process, which can be widely applied to the pattern recognition of the parameter' s changing trend during the control process and the image partition processing, and utilizes the IFKN to recognize the thermotechnical parameter' s changing trend based on the engineering of clinker sintering rotary kiln automatic control system of Guizhou' s Aluminium Factory, through which good effect is obtained. Applying the data mining technology to the industrial process control is not only the brand-new research content, but a new subject much worth studying in automationfields. |
PDF Full Text Request