The Study On Temperature Modeling Methods Of Regenerative Reheating Furnace

Regenerative reheating furnace is an importan part of iron and steel metallurgical industry, whose main feature is heating the steel blank on the rolling line according to our process heating requirements. As a result of heating process in the reheating furace is very complex, obviously, it is hard to establish a practical and accurate model. In other words, it is difficult to accurately control the temperature in this process. Therefore, the establishment of an accurate temperature model of the reheating furnace has a great significance for better control of that.The regenerative furnace is used as the backgroud for the thesis. First of all, modeling is established by the method based on multivariate statistical technique. The modeling have been got respectively by the method of Least Squares Support Vector Machine (LSSVM) and the Kernel Principal Componet Analysis (KPCA) method to deal with our large amount of datas, following by the application of Least Squares Support Vector Machine (LSSVM) on regression temperature modeling of reheating furnace. In the modeling process, wo found that an accurate temprtature model establishment has a closely relation with the gas and air tempeture before and after the regenerator through the course of choosing procedure variables, but some of the temperature datas in the process can not directly measured and be looked as modeling parameters. In order to achieve better modeling results, further, we studied the mechanism of regenerator, and estabished a corresponding mechanism model in accordance with the parameters we need, through what the temperature of the gas after regenerator preheating can be indirectly got according to the paremeters of the smoke temperature and so on.Based on above, combining with the mechanism model of regenerator, temperature modeling were re-estabishde respectively by LSSVM method and KPCA-LSSVM method. Compared with the previous mothed using multivariate statistical technique only, the results show that the method of the mechanistic modeling combined with multivariate statistica modeling can obtain a better result than the method of using multivariate statistical technique only. The conclusion also provides a useful attempt for the study of temperature modeling.