Molten Steel Level Control of Strip Casting Process Monitoring by Using Self-Learning Fuzzy Controller

The twin-roll strip casting process is a typical steel-strip production method which combines continuous casting and hot rolling process. The production line from molten liquid steel to the final steel-strip is shortened and the production cost is reduced significantly compared to the conventional continuous casting. The twin-roll strip casting process can produce 1-5 mm thin steel strip directly from the molten steel. Furthermore, since the strip casting process has high cooling rate, it can improve the mechanical properties of steel (Liang et al. 1997; Cook et al. 1995). Usually, the molten steel level is controlled at a preset desired level to monitor the normal strip casting operation. During the roll casting process, once the molten metal contacts with the rotating rolls, a thin solidification shell is formed on the surface of each roll. The shell thickness gradually grows from each roll surface, finally contacts with each other and weld together at a position around the roll exit, called the solidification final point. If the molten metal level is higher than the specified value, the solidification final point will occur at a point above the roll exit. That will result in heat cracking and damage to the cooling roll surface in addition to material structural abnormalities of the steel strip. If the molten metal level is lower than the desirable value, the solidification final point will occur at a point below the roll exit. The steel strip surface will have inferior quality due to the breakout and oxidation. Hence, the molten metal level is an important process control parameter to guarantee the solidification final point and rolling strip quality. The molten steel level must maintain within a specific range during the full casting process except the initial startup operating mode by filling the molten steel into the twin roll cylinders from the tundish. Since the strip casting process has nonlinear dynamics uncertainty and coupled behaviors, accurate molten steel level control problem is still an important research topic to guarantee the steel strip quality. Graebe et al. (1995) verified the dynamic model and various nonlinearities appearing in the continuous casting process and proposed different issues that had to be solved in controller design. Hesketh et al. (1993) applied an adaptive control strategy for the mold level control of a continuous steel slab casting. Hong et al. (2001) investigated the modeling and control problem of a twin-roll strip caster. They analyzed different critical dynamics, including molten steel pool leveling, and developed a two-level