Research On Coating Control System In Continuous Galvanizing Line Based Neural Network Predictive

The galvanizing on surface of strip is an important method for steel anticorrosion. The anticorrosion time is decided by the coating weight on surface of galvanized strip. It is very important to control coating weight precisely according to customer's demands for improving product quality and reducing cost. Nowadays, the advanced continuous galvanizing lines import from foreign. The research on coating weight control is useful to improve proprietary intellectual property and develop proprietary coating weight control system.In this paper, the coating control system in continuous galvanizing line is analyzed deeply, which include coating model preset, longitudinal coating mean control and transverse coating deviation control. In CLECIM coating model, the coating weight is the function of air pressure, distance from air knives to strip and air knives lip gap. The model preset is used for calculating reference values of control variables. In the longitudinal mean control, air pressure and distance are controlled to achieve precise and mean coating weight. And the transverse deviation controls air knives lip gap to improve the coating uniform.Given the time-delay of coating thickness gauge, Smith predictive control is adopted in coating control system. The linear coating model is used in Smith predictor, which model parameters are depended on present air pressure, line speed etc. In industry fields, air pressure and line speed fluctuate for strong disturbances, so model parameters are changed and control performances of Smith predictor are influenced. The Edward coating weight model is identified using the least square approximation base on real measured data. According identified Edward coating model, the author designs the neural network predictive algorithm for longitudinal coating mean control system.Compared linear coating model adopted in Smith predictor, the neural network coating predictive model is more suitable for nonlinear and variable strip hot dip galvanizing process. The simulation results indicate the controller designed acquires shorter overshoot, shorter adjusting time and greater robustness. And quality of coating weight on strip surface is improved.