Study on neural network proportional-integration-differential (PID) control strategy of the molten metal pool level in the twin roll casting process

Abstract

In the twin-roll casting process, how to accurately control the molten metal pool level is a key problem to produce high quality strip. In this paper, the mathematical model about the pool level control is set up based on the process characteristics. Meanwhile, the limitations of the traditional PID control strategy are analyzed owing to the real-time change of the roll gap and roll speed. Furthermore, the neural network is applied to adaptively optimize the PID control parameters and the simulations show the neural network PID strategy can precisely control the molten metal pool level in twin-roll casting process under the condition of multiple factors interfering.

Keywords

• magnesium strips
• twin-roll casting process
• molten metal level
• PID control
• neural network (NN)