Adaptive feed-forward automatic gauge control in hot strip finishing mill

The conventional roll force automatic gauge control (RFAGC) adjusts the roll gap to reduce the delivery thickness variation in the hot strip finishing mills. In this process, the information about the roll force is used to control the roll gap indirectly because the strip gauge cannot be measured owing to insufficient space in the mill and the hostile environment. Moreover, there exists an uncertainty in thickness due to roll eccentricity. Therefore, a small gain is preferable in avoiding product defects when this uncertainty is taken into account in RFAGC. However, in such a case, partially uncontrolled variation still remains on delivery thickness because of insufficient control input. To overcome this drawback, a feed-forward controller (FFAGC) was developed.(12)) This scheme adjusts the roll gap to reduce variations in the current stand by using an estimated uncontrolled delivery thickness variation of the previous stand. Although this FFAGC has an advantage over RFAGC, it still has a problem: The FFAGC cannot cope with the continuous variation of system parameters and circumstance because the existing feed-forward controller has been designed as a low pass filter with constant weights. In other words, desirable control performance in the automatic gauge control system cannot be obtained owing to a phase delay and filtering errors. To overcome the problem, it is essential to estimate the long-term thickness variation exactly in the hard environment, to select a proper gain and to synchronize the feed-forward control input under a time varying speed of the gauge control process. In this paper, an enhanced adaptive filtering structure is proposed to improve the controller performance. And the relation between the gains of RFAGC and FFAGC are discussed to select a proper gain. The proposed controller can readily be implemented on the conventional control system, improving the performance of the system without additional cost. This improvement was demonstrated by using the real coil data which are collected under the proposed control scheme in POSCO.