Development of Converter Alloying Model Based on Cascaded BP Network
Deoxygenation alloying is an important part of the converter smelting process,and the accuracy of composition stability directly affects the subsequent refining effect.Due to the influence of oxygen content in the steel,there are many factors that affect the yield of alloy elements,making it difficult to create deoxygenation alloying models.Many traditional converters are also not equipped with alloying models.This article is based on cascaded BP neural network technology,using data analysis to screen important influencing factors as network inputs,steel content and alloy content as outputs to establish a network.Finally,an alloying application software is formed using VBA program,which has made significant progress in practical production.The predicted and actual deviation of steel content is within 2 tons,with a hit rate of 97.5％,the predicted and actual deviation of silicon manganese alloy is within 50 kilograms,with a hit rate of 94.6％,and the deviation of silicon iron content and actual value is achieved,Within 10 kilograms,achieve a hit rate of 98.5％,with a deviation between the amount of carburetor added and the actual value,and a hit rate of 89.8％within 5 kilograms.
NETL
NSTL
万方数据
