Desulphurization Prediction Model For Ladle Furnace Refining Process

The demand for high quality steel, has been increasing continuously with the development of industries and the process of technologies, particularly, high-level pipeline steel. In order to response to the demand, LF refining is playing an important role in entire steelmaking process due to it’s special function of desulphurization whith white slag refining. Therefore, it is very necessary to study the factors influencing desulphurization and predict the sulfur content whith refining desulphurization model precisely in order to have a stability of steel products.In the present work, mechanism model and "black-box" model for predicting desulphurization of LF reefing process were established according the condition of135t LF in some steel plant, and the two models were validated by the measured data. The main conclusions were drawn as follows:(1) The desulphurization process is controlled by different steps. In the initial stage of refining (before20min), the mass transfer of [S] is found out to be the ultimate restrictive step. In the final stage of refining, the ultimate restrictive step changes to the mass transfer of (S2-).(2) Mechanism desulphurization model is developed by VB6.0. The results show that if the error is allowed within5×10-6, the prediction hitting ratio is80%and the prediction hitting ratio may arrive92%when the error is allowed within10×10-6.(3) Based on BP neural network, desulphurization "black-box" model is established by MATLAB. The results of the model show that the prediction performance of BP "black-box" is best when the numbers of hidden layers is15. If the error permission is restricted within5×10-6, the prediction hitting ratio of the black box model is84%and the prediction hitting ratio may arrive94%when the error is allowed with10×10-6.(4) Both the mechanism and "black-box" desulphurization models have their own defects which cannot be conquered by themselves. The "grey-box" can be developed through the combination of mechanism and "black-box" desulphurization models. Before the application of "black-box" desulphurization model, the sample data is pre-judged and invalid data is rejected through the mechanism desulphurization model to provide high-quality data for the "black-box" desulphurization model which is the further direction of this project.