Abstract: Flatness is an important quality indicator in strip rolling, and tilting roll and bending roll are important means of flatness control. Now conventional PID control algorithm is widely applied to tilting roll and bending roll flatness control system for strip mill, but it is very difficult to build an exact math model and conventional PID control algorithm is hard to meet the need of high-precision flatness control because of actual system’s severe random disturbance and multivariable, non-linear and strong coupling characters. Fuzzy PID tilting roll and bending roll flatness control model based on neural network is built on the ground of conventional PID control algorithm. An optimized non-linear combined control rule of P, I, and D is built by the self-learning ability of neural network and the “conception” abstraction ability of fuzzy control to strengthen the ability of self-adaptation and self-leaning to the change of control environment. Simulation results of the model indicate that flatness target value is tracked well with rapid response, small overshot and strong robust. In conclusion, flatness control precision of tilting roll and bending roll is increased and a new method is provided for high-precision flatness control.

Key words: Bending roll, Flatness control, Neural fuzzy, PID, Tilting roll