[%P] prediction and control model for oxygen-converter process at the end point based on adaptive neuro-fuzzy system

According to the process and data from spot, the methodology for [%P] prediction and control is discussed. A self-organizing network has been utilized to classify 303 heats from spot, which makes the analysis of the influence of steelmaking variables on [%P] possible. The control variables for the [%P] prediction and control model were determined with the analysis. A model of [%P] prediction and control has been established for BOF at the end point based on an adaptive neuro-fuzzy system. The results show that this model has good performance on prediction and control for [%P] in the BOF process. The R-value of model output and actual [%P] in the experiment reaches 0.5867. The hit rate of the model in the precision ±0.003% [%P] is 79.21%. With this model, if the [%P] was controlled by the model with the value less than target by 0.004%, 91% of heats are up to grade in regard to [%P].