Research On Billet Temperature Modeling Method Of Regenerative Reheating Furnace

The reheating furnace is a large energy consumer on steel rolling line. The quality of billet heating in it has a direct effect on product quality. Only reasonable control of billet’s temperature and its distribution can assure rolling quality. But refer to billet temperature field, the surface and inner temperature of billet now can’t be easily reached or measured by using some instruments because of the limition of testing technology and the high test cost and the complexity of the furnace. Therefore, an accurate temperature prediction model is particularly important, an accurate model can calculate the steel temperature, and have important reference value for the control system of reheating furnace, and to improve the quality of billet heating and reduce the fuel consumption in reheating furnace.There are a brief overview of regenerative reheating furnace and fundamental algorithm of system modeling in this paper. Heat transfer knowledge is taken as a supporting, the mechanism model based on overall heat absorption rate is established, because the parameters of this model need to be determined according to empirical formula repeatedly, the adaptive genetic algorithm(AGA) is used to find the best parameters of the mechanism model in order to avoid the tedious process. Secondly, the mean impact value(MIV) algorithm is used to screen out the input variables which have the biggest impact on the output variables of the neural network. The MIV algorithm can reduce the dimension of the input variables and simplify the structure of the network model. Then the method of momentum-adaptive learning rate is used to train the neural network, the billet temperature prediction model which is based on the improved BP neural network is established. Finally, the adaptive genetic simulated annealing algorithm is used to optimize the improved BP neural network model, the adaptive genetic simulated annealing algorithm improve prediction accuracy and ensure the stability of the improved BP neural network model compared with the model optimized by the adaptive genetic algorithm.The model simulation results show that the solution process of the mechanism model is more complicated and the prediction accuracy is low, but the mechanism model can predict the final temperature of the billet, and obtain the approximate temperature of the billet in the reheating process in the absence of the test cases. It is difficult for the intelligent method to predict the temperature of the billet in the reheating process, the intelligent method can predict the final temperature of the slab precisely, The convergence of conventional BP neural network is slow, the training procese has low precision and the prediction ability is poor, the improved BP neural network model can overcome these shortcomings. The adaptive genetic algorithm is used to optimize the improved BP neural network model, this model can overcome the shortcoming of easy to fall into the local minimum, further improve the model’s training accuracy and generalization ability, but still needs to further improve the stability of this model. The adaptive genetic simulated annealing algorithm is used to optimize the improved BP neural network model, the prediction accuracy and stability of this model is further improved.