Automatic knowledge acquisition system for blast furnace burden distribution operation

At Keihin Works, NKK, case-based reasoning and explanation based learning has been applied to the online control system of blast furnace, and an automatic knowledge acquisition system is adopted. In this system, features of operation values before and after the manipulation are stored as cases, and the cases are evaluated utilizing the membership functions and processed for a guidance message to the operator. The cases are also used for the control rule acquisition. For the similar cases, causality relationships among the operation values are acquired based on the entropy and static gain calculation. The control rules are generated from each case by means of deductive learning using the causality relationships. Thus, the quantitative knowledge of blast furnace control can be acquired automatically, which is difficult to attain by human operators judgement.<>