| Phosphor is a harmfulness element for great majority of steel grade, we must removeit to lowest possible content in smelting process; High endpoint remanent manganese canreduce the use quantity of Mn-Fe. Accurate controlling of the endpoint content hasimportant significance for increasing product quality and reducing the cost of steelproduction, So we must precise prediction and control P & Mn endpoint content forconverterThis paper has analyzed the development and actuality of control technology aboutthe modern converter endpoint, because that the most converters are medium or smallconverters , The converter's capacity is small, so can't use the dynamic control technique.The traditional static model's calculation accuracy is bad, target hit rate is low, and it'sapplication effect in actual production is not good. So, this paper take full advantage of theartificial neural network technology developing in recently, has established the staticprediction model of end-point content for oxygen-converter based on neural network withthe Visual Basic programme language.According to the process and data from spot, the paper has analysised the factors forinfluence Mn P endpoint in converter. The control variables for [%P], [%Mn] predictionand control model were determined. And we improve the BP algorithm and establish twoprediction neural networks. The actual data of continuous 200 batch from XG were chosenas example, 30 input variable that influence Mn P endpoint in converter were determined,The learing is 0.01, and the momentum term is 0.8, We establish two three-layers 30-33-1BP prediction neural networks, and predict [Mn] [P] endpoint content.The model of [%P], [%Mn] prediction have been establish, have good performance onprediction and control for [%P], [%Mn] in converter process. The hit rate of the model inthe precision ±0.003%[%P] is 85%. The hit rate of the model in the precision±0.03%[%Mn] is 80%. The performance is close to the performance with dynamic model. |
PDF Full Text Request