Prediction of X-ray fluorescence copper grade using regularized stochastic configuration networks

Abstract

In copper ore flotation processes, accurate measurement and estimation of the copper grade of the final products are crucial for field operations and control. Due to the presence of multicollinearity and outliers in the measured data, the existing on-stream X-ray fluorescence grade analyzer is inaccurate. To address these issues, this study proposes a robust data modeling algorithm to improve the performance of copper grade prediction. Specifically, a regularized stochastic configuration network and a generalized M−estimation method are used to overcome the uncertainties in the processing data. The experimental results clearly demonstrate that the proposed model outperforms other modeling algorithms in terms of prediction accuracy and robustness.