A Clustering-analysis-based membership functions formation method for fuzzy controller of ball mill pulverizing system

Abstract: This paper proposes a clustering-analysis-based membership functions formation method for fuzzy controller of ball mill pulverizing system. An improved density-based clustering algorithm is presented to detect the clusters in the field database and the clusters are agglomerated or divided to form the new clusters according to the proximity or the dissimilarity between objects. Then, the new clusters are projected into the domains of the control variables to form the intervals, and the membership functions are established based on the intervals and the projections of the centroids of the clusters. The experiments results verify that the performance of our membership functions formation method is better and the fuzzy controller designed by our method has higher control quality. In addition, the fuzzy controller has been put into practice and the field data verify the effectiveness of the fuzzy controller. [Copyright &y& Elsevier]