Artificial Intelligence Prediction of Carbonate Crystallinity of Carbon Mineralization

Abstract

The importance of carbon capture, utilization, and storage (CCUS) for achieving carbon neutrality is increasingly recognized. Carbonate minerals are currently being manufactured from the abundant calcium-containing wastes and minerals that are generated by carbon mineralization technology in industry. Among these, calcium carbonate, which is highly versatile, generally exists in three crystal forms (vaterite, aragonite, and calcite). These three crystal forms must be freely controllable to increase the value and range of use of calcium carbonate. In this study, the variables of concentration, temperature, pH, stirring speed, and stirring time were changed during the reaction of calcium raw material (i.e., CaCl₂) and carbon raw material (i.e., K₂CO₃). In addition, the phase composition ratios were determined by Rietveld refinement analysis using X-ray diffraction (XRD) patterns. Drawing on an extensive set of experimental data, we constructed data-driven predictive models by training and evaluating multilayer perceptron (MLP), support vector machine (SVM), random forest (RF), and decision tree (DT) algorithms. The best-performing model, selected by k-fold cross-validation, was then applied to determine the optimal operating conditions to control crystallinity. This study provides comprehensive knowledge about a system that allows industries to select, manufacture, and produce calcium carbonate in the crystal form they need. It is anticipated that using carbon mineralization technology, which is part of CCUS technology, will contribute to carbon neutrality, while alleviating waste environmental treatment costs.