Study On Intelligence Model Of Profile Forecast And Profile Control In Hot Strip Mills

The profile control system is a key technology in the modern strip rolling mills, and the profile control system system basically adopts the tradional simplified mathematics model, which is a typical nonlinear process with variable parameters, formidable time-varying and uncertainties in mathematical model, and the problems of flatness hysteresis which flatness detection bring in effects quality, so this model exists larger deviation in the practice rolling mill, and lack adaptability to change during the the reality situation, In recent years, useing neural network and genetic algorithm to control nonlinear systems has become one of the characteristics of the predictional study. This model has received good result especially in nonlinear systems. It has ability to track the target when the sigal is disturbed, and it represent relative good robustness as well.In this paper, to improve the accuracy of profile pre-set model is us goal, based on status of deeply analyzes deomestic and abroad and combined with the 1780 hot strip mill flatness control projects which is a domestic production line, so set up a profile control model based on neuralwork and genetic algorithm and analyze the existing problem.First, we divided finishing mill is into crown areas, straightness controlled areas , so we control the strip crown as well as profile effectively. According to the neural network has robustness, memory capacity, nonlinear mapping ability and strong self-learning ability, we establish the faltness prediction model for the database. The experimental data show that the neural network in predicting the convexity, the calculated value more than high precision.Secondly, we use genetic algorithm in the global search to optimize weights and threshold based on the shape prediction, and it avoid the local minimum problem effectively. We use genetic algorithm to optimize the runout and combine with prediction model, thereby we establish a preset profile model. Through the experimental data analysis that it is feasibility.Finally, we writte a control interface by Matlab software based on prediction model and preset model. Through the interface, we view the data and graphics.simply and directly. Studying on hot Plate-shaped intrllect model provides a new faltness solution to solve the nonlinear and strong coupling problems.