부산대학교 도서관

During copper matte smelting, a lag exists in the analysis of the smelting results, leading to the production parameters not being effectively adjusted in time. As the first smelting process in copper pyrometallurgy, the optimization of the amount of raw material, as well as the prediction and adjustment of slag and matte phases, has become an extremely important step. To address this issue, this study analyzed copper smelting parameters, such as the copper content in the slag, matte grade, slag ratio, and Fe/SiO2 ratio. The Bayesian optimization support vector regression (BO-SVR) model was used to predict the copper content in the slag and matte grade by combining plant production data with machine learning. The data features of the relevant copper smelting parameters were extracted using the BO-SVR model, and the data were fitted to obtain the interaction relationship functions between the variables. The relationships between several variables in the copper smelting process were interpreted in combination with a theoretical analysis and applied in practice. The results showed that the model’s prediction of melting results resolves the issues of lagging and adjustment of the smelting process. The model can provide insights for the analysis and adjustment of critical parameters in the smelting process and promote the development of intelligent analysis and control of copper matte smelting.