Abstract
In hot strip rolling mills, the looper control system is automated. However, the looper’s behavior tend to be unstable in threading. Therefore, human expert never fail to intervene and stabilize the looper’s behavior by tuning PID gain and interposing manipulated variable of looper control system. We try to express PID gain tuning action by human expert using recurrent neural networks. Furthermore, we propose a method to update the model by reinforcement learning. To check the effectiveness of the proposed learning, numerical simulation of the looper control is carried out.
