Application of multiple linear regression (MLR) on copper ore separation based on grade and density

Abstract Productivity in the mining industry has decreased significantly, partly due to declining level of minerals processed. On a global scale, the commonly researched theme in mineral processing is the comminution process which requires high energy in the production of metallic minerals. The coarse gangue separation method with two strategies, sensor-based separation and gravity separation, is the efficient solution for this issue. However, sensor-based sorting has disadvantages in terms of cost, so gravity separation is considered a more effective separation method. This research aimed to evaluate the efficiency of the gravity separation method using secondary data subjected to Multiple Linear Regression with several processing methods that included Constitution Heterogeneity (CH), R 2 Optimization, Single Linear Model, and Multivariate Linear Regression Model based on particle grade and density data. The results showed that in relation to particle size, CH values was inclined to be very small due to the homogeneity of the materials when comminution ensued. In addition, the analysis shows that the Multivariate Linear Regression Model is unable to predict results based on experiments or real results, so this has an impact on the low effectiveness of separations carried out in the field.