Prediction of ball mill power in iron ore concentration plants: A comparison between radial basis functions and linear regression

• Data from an iron ore concentration plant was collected, including feed bond index (ranging from 12.5 to 13.5 kWh/t), feed size distribution (F80: 5000 to 14,600 µm), product size distribution (P80: 80 to 180 µm), feed rate (171 to 339 t/h), grinding media weight (varying between 75 and 175 tons), and ball mill power draw (1.91 to 2.66 MW) to train the models. • Integration of various operating conditions and parameters into the models through precise data collection emphasizes the potential for enhancing energy efficiency and sustainability in mineral processing operations. • These advanced modeling techniques offer practical applicability in optimizing ball mill performance and energy efficiency in the mining industry. The iron and steel industry is a major energy consumer, with the need for optimization in energy consumption becoming more apparent as high-grade minerals become scarcer. This study uses linear regression and radial basis function (RBF) models to predict ball mill power draw in iron ore concentration plants. Data from an iron ore concentration plant were collected, including feed bond index, size distribution, feed rate, and grinding media weight. A total of 1438 datasets from industrial ball mills were utilized, with 80 % allocated for training and 20 % for testing. Findings indicate that the RBF model outperforms linear regression in terms of accuracy and precision, exhibiting a lower standard deviation (0.0782 MW for RBF compared to 0.1207 MW for linear regression) and a smaller average absolute error (0.0585 MW for RBF compared to 0.0899 MW for linear regression).. By integrating various operating conditions and parameters into the models via precise data collection, opportunities to enhance energy efficiency and sustainability in mineral processing operations are emphasized. The RBF model's non-linear nature makes it particularly well-suited for capturing complex relationships in the data, which contributes to its superior performance.