Particle-scale magnetic susceptibility estimation from automated mineralogy data: A workflow for complex multi-mineral particle systems

Magnetic susceptibility is one of the key properties influencing the response of mineral particles in magnetic separation processes, alongside factors such as particle size, shape, and magnetic field conditions. While particle-scale properties such as composition and size can be directly measured using automated mineralogical techniques, magnetic susceptibility remains inaccessible at the individual particle level and is typically available only as a bulk value, limiting process understanding and predictive modeling in heterogeneous systems. This study presents a particle-based magnetic susceptibility estimation method using a linear ridge regression model calibrated on quantitative mineralogical data derived from automated analysis of susceptibility classes generated by a Frantz Isodynamic Separator. The model establishes statistical relationships between mineralogical composition and measured susceptibilities, enabling estimation of apparent susceptibility values for individual mineral phases and particles. The approach is demonstrated on Norra Kärr ore and a WEEE slag system. For Norra Kärr, predicted and measured susceptibilities show good agreement. For the WEEE slag, the model captures a broad susceptibility range despite complex ferro- and ferrimagnetic behavior. The method provides a material-specific framework to estimate particle-scale susceptibilities, addressing a key measurement gap and supporting improved magnetic separation modeling and process optimization.