The Study And Application Of The Hot Metal Desulfurization Forecast Model Based On RBF Neural Network

Sulphur is a detrimental element in most steel grades. Currently the domestic level of the hot iron desulfurization control is relatively low with high fluctuation of the S content in the hot iron, big consumption of the desulfurization powder and then giant desulfurization cost. The fact that the hot iron desulfurization process is a complicated multi-nonlinear reaction one makes it hard to adapt to multi-parameters and non-linear characters of the process by applying the conventional technological theory to create the model. The article, under the background of the hot iron desulfurization production, analyses the characteristics of the desulfurization process and establishes its desulfurization forecast model of the neural network based on RBF.The main tasks of the paper are listed as follows:(1) Based on the hot iron desulfurization system owned by the Pudong Iron&Steel Co., Ltd-a subsidiary of Baosteel Group, the paper clearly depicts the technological process of the desulfurization production, analyses the mechanism of the desulfurization process and determines the elements which affect the S content.(2) Selecting ROLS as the training algorithm, the thesis puts forward a forecast model of the desulfurization powder volume based on the RBF neural network.(3) In order to overcome the model errors caused by the volume variation of the desulfurization powder, it is necessary to rectify the model data according to the predicted errors. The paper designs a comprehensive intelligent desulfurization forecast model combining data collecting, pre-handling, statistical analysis, calculating of the desulfurization powder volume and the technical specifications verifications.(4) The paper realizes the function of the intelligent desulfurization of the hot iron by using the Oracle database to store and handle the data and the language of C++combined with MATLAB to program. The results gained from the scene undoubtedly demonstrate that satisfying desulfurization effects can be achieved by computing the desulfuration powder volume according to the model advanced by the paper.