Methods of diagnosis and intervention for agent of hot rolling operation support

In the last two decades, it becomes possible to automate operations of various steel plants especially in rolling mills. As the results, stabilization of productivity and improvement of product quality have been attained. On the while, in these years, many skilled engineers and operators who actively promoted economical growth of steel industries will retire due to their age limits. Thus, the inheritance of the high level technology and know-how has becomes a serious problem. To overcome the problem, it is necessary to extract knowledge of the skilled persons and make technical textbook reducing tacit knowledge. In this paper, rules are extracted from the operation data of hot strip rolling applicable to the operation diagnosis and intervention during operation. To attain the object, agent based simulator of hot strip rolling has been developed to prepare various rolling data for extraction of diagnosis and intervention rules in rolling operations. As for the selection of normal and abnormal data, SVM algorithm is tested before rules extraction. Rules are written in Fuzzy logic forms and its parameters are optimized by GA algorithm. These technologies are involved in the operation support agent system of hot strip rolling mills together with RNN for automatic gain tuning of mill controller.