In this paper, a new gaugemeter automatic gauge control system for a hot strip finishing mill is presented and new suggestions for high gauge accuracy in FSU model are introduced. In the hot strip finishing rolling, there exits a mutual interaction between gauge control and strip tension control. The estimated rolling force and lock on force may be different and there exist input thickness deviations. So the output thickness may have some deviations and offset values. Therefore, AGC system must control the output thickness deviations and offset values and also decouple the mutual interaction between gauge control and strip tension control.
First, some suggestions to improve the control performance are introduced. And this paper extends the diagonal recurrent neural network(DRNN) with multiplication between inputs and self-recurrent neurons. In order to obtain the desired output thickness, new gaugemeter AGC system with neuro-compensator is introduced. The modified diagonal recurrent neural network(MDRNN) will be used in the new AGC system. It is shown by simulations that new gaugemeter AGC system is effective for improving the gauge control accuracy.
