Optimising Metal and Mineral Recovery from Waste Streams : A Data-Driven Approach

Global primary material use is expected to double by 2060, contributing to increased landfilled waste, environmental pollution and reduced access to metals and minerals. In order to address this challenge, a transition to a circular economy is essential. However, global circularity is decreasing annually, and circularity within the EU remains largely stagnant, indicating that this shift is not occurring at scale. Data-driven methods could be used as decision making support for enabling development of waste recovery technologies to increase the input rate of secondary raw materials into the global economy.This thesis presents a data-driven methodology consisting of: correlation matrix, clustering algorithms, domain confirmation and regression modeling to predict the oxide composition of ladle furnace slags more precisely compared to basic statistical methods. Evaluation using randomised datasets shows that the model improves prediction accuracy of oxide concentration by 37–45% for 4 out of 5 oxides with a variance reduction of 23–54%, compared to using statistical means. The result is lower error margins when estimating available quantity of oxides in ladle furnace slags,showing that the methodology can be used to better identify high potential mineral waste streams, improve the recovery of metals and minerals and thereby support the transition to a circular economy.