Application of SVM method to operation diagnosis of hot strip rolling

As it is well known, in a factory, there exist many equips with large scale and plants with complexities. For the reason, complete automations of its operations are very difficult. Therefore, the human intervention is remained an important role for smooth operations. The intervening operation reflecting skillful personnel´s experiences and knowledge are inevitable to maintain qualities of products. The typical examples of such operations are those of the hot strip rolling. In this research, the development of the system technology for the operations support of hot strip rolling based on agent method is aimed. First, the simulator of the hot strip rolling is developed, and the rolling phenomena are reproduced. The abnormality is artificially reproduced using the simulator. The diagnosis of the abnormality with a support vector machine (SVM) is studied to prepare the diagnostic system for operations. The identification function to judge the boundary of the abnormality and the normality is made by using the teaching data. It is expected that this diagnostic function can be used for the cause presumption of the anomalous phenomenon. In the experiments, anomalous phenomena are reproduced and the diagnostic test results are shown. Thus, the effect of the proposed operation support method is confirmed.