Machine Learning-Guided Screening of Natural Iron Ores for Chemical Looping Combustion: From Algorithms to Experimental Reality

Abstract: This dataset contains a curated collection of 286 experimental entries for Iron (Fe) and Manganese (Mn) based oxygen carriers used in Chemical Looping Combustion (CLC) and Chemical Looping Reforming (CLR). The data was systematically extracted from 45 peer-reviewed articles published between 2010 and 2025. Methodology: Data collection focused on materials tested under standardized thermogravimetric analysis (TGA) conditions: Reduction Atmosphere: 5% H2 + 40% H2O (balance N2) or similar reducing gases. Temperature: Approximately 900°C. Selection Criteria: Only stable cycles (typically >3 cycles) were recorded to ensure performance reliability. File Structure: The dataset includes the following features utilized for Machine Learning modeling (Gradient Boosting / SVR): Chemical Composition (wt%): Normalized mass fractions of active phases (Fe2O3, Mn3O4) and support/gangue materials (MgO, TiO2, SiO2, Al2O3, CaO). Physical Properties: Mechanical Resistance (Crushing Strength) measured in Newtons (N). Performance Metric: Oxygen Transport Capacity (ROC, wt%), calculated based on mass variation between oxidized and reduced states. Applications: This dataset is suitable for training regression models to predict the redox performance of synthetic and natural oxygen carriers, studying the "dilution effect" of inert supports, and benchmarking natural iron ores against synthetic materials.