Prediction Model of Antimony Content in Copper Matte Based on PCA-NIPSO-SVM
Copper matte is the product of matte smelting and also the raw material for the production of blister copper through converting processes.Antimony,as a difficult-to-remove impurity in copper matte,significantly influences the grade of copper,the lifespan of anode furnaces in pyrometallurgical refining,and the quality of cathode plates in electrolytic refining.Given the limited research on predictive control of antimony content in copper smelting,and issues such as poor feature selection and prediction results,a predictive model for antimony content in copper matte was developed,utilizing Principal Component Analysis(PCA)and Support Vector Machine(SVM).The model utilized Grey Relational Analysis(GRA)and Principal Component Analysis(PCA)for feature selection and dimensionality reduction,and integrated with a Novel Improved Particle Swarm Optimization(NIPSO)algorithm enhanced by an improved Sine chaos map for SVM optimization.The model was tested,trained,and validated using metallurgical data.Experimental results demonstrate that the PCA-NIPSO-SVM model significantly outperforms PCA-PSO-SVM,PSO-SVM,and standard SVM in predicting antimony content.The evaluation metrics-Mean Absolute Error(MAE),Mean Squared Error(MSE),and Root Mean Square Error(RMSE)—are 0.011 815 5,0.000 225 6,and 0.015 019 7 respectively.This SVM model,optimized with the improved Sine chaos map-based particle swarm algorithm,effectively predicts the antimony content in copper matte,providing valuable insights for ingredient planning and process control in oxygen bottom-blown copper smelting.
万方数据
维普
