This thesis is focused on the research of both the strip steel thickness and the rolling force correction of Hot Strip Mills. The Neural Network, combining with the mathematic model of the rolling force is introduced to this study. In this paper, multiple correction method for the strip steel thickness and the rolling force deviation in the single and tandem rolling mill is established. The RL-BP neural network is used to predict the multiple deviations in 2050 tandem rolling and finishing mill group. The off-line simulation of the multiple correction method in 2050mm hot tandem rolling and finishing mill group shows that the rolling force deviation is decreased by 5%; the strip steel thickness deviation is decreased by 20%; the correction is obviously effective. After the neural network prediction system of rolling force is used in the 2050mm HSM, from the data we can see that the strip steel thickness precision has improved. The application of artificial neural network in the control of rolling force in 2050mm hot tandem rolling and finishing mill group provides a method to combine the neural network and mathematical models. keyword: hot rolling strip steel, mathematic model, neural network, rolling force, rolling force deviation, thickness deviation.
展开▼
