Abstract: In order to realize the precise control of alloys in the steelmaking process, firstly, the physical and chemical properties of all alloys were analyzed, including the physical form, strength, wear resistance and composition of alloys, and an alloy database was established based on the analytical results to guide the addition of alloys. Secondly, the alloy yield of typical steel grades was analyzed, and a correlation analysis was carried out between the alloy yield and the process parameters using Pearson correlation coefficient, and  the following results was obtained. The Mn element yield increased with the reduction of oxygen content at the end of the converter and the increase of endpoint temperature. T-LSTM neural network and K-means clustering algorithm were used to predict the alloy yield of converter tapping, increasing the temperature of tapping, reducing the end of the oxygen content was conducive to improving the alloy yield. Finally, the intlinprog function based on matlab was used to solve the linear programming for the dosage scheme with the lowest cost, and the intelligent optimization control system of steelmaking alloy was developed to realize the lowest cost of alloy. The alloy intelligent optimization control system was applied in a  120 t converter at a steel mill, the model was operated online  for 8 months and  verified data  from 26 000 furnaces. The average alloy consumption of steel reduced by 0.46 kg/t, and the alloy cost of steel grades reducted by 1.21—6.58 yuan/t,  with an average reduction of  3.71 yuan/t.

Key words: converter, alloy yield, alloy database, linear programming, alloy model