Multi-mode smelting method based on the classification system of molten iron

A method for determining an amount of alloy added during converter tapping, comprising following steps:

• S1, collecting a converter production data set and establish a prediction model database;
• S2, conducting data screening and elutriation on the collected converter production data set, and pre-process the data after screening and elutriation;
• S3, determining process parameter variables that affect a Si yield in a converter tapping process as input variables for establishing a LSTM (Long Short-Term Memory) neural network Si yield prediction model;
• S4, determining process parameter variables that affect a Mn yield in the converter tapping process as input variables for establishing a LSTM neural network Mn yield prediction model;
• S5, determining process parameter variables that affect a Cr yield in the converter tapping process as input variables for establishing a LSTM neural network Cr yield prediction model;
• S6, training and testing the LSTM neural network Si yield prediction model, the LSTM neural network Mn yield prediction model, and the LSTM neural network Cr yield prediction model;
• S7, collecting real-time data of on-site smelting process, and providing the real-time data of on-site smelting process to the LSTM neural network Si yield prediction model, the LSTM neural network Mn yield prediction model, and the LSTM neural network Cr yield prediction model to predict a recovery rate of Si, Mn and Cr elements;
• S8, determining an alloy addition amount through integer linear programming,. specifically including:
• minimizing a total cost Z of alloy addition as an objective function
• min Z=Σi=1nxi·ri=x1·r1+x2·r2+. . . +xn·rn (4)
• where:
• ri represents an unit price of the i-th alloy material;
• xi is the amount of the i-th alloy added;
• and
• S9, adding alloy according to the determined alloy addition amount.