27-Year Multi-Producer European Cement Characterization Dataset: Chemistry, Fineness, and Performance

This record contains a processed, anonymized cement characterization datasetassembled from routine laboratory measurements collected between 1997 and2024. It contains 476 cement samples from 23 anonymized producers and 50variables. The release comprises 311 CEM I samples and 165 other cementrecords, primarily CEM II and CEM III, together with a small number of CEM IV,CEM VI, and other special cements. The variables cover cement type and strength class, oxide chemistry, loss onignition, density, Blaine fineness, particle-size-distribution descriptors,setting time, water demand, Le Chatelier soundness, and flexural andcompressive strength at 2, 7, and 28 days. Calculated descriptors includeequivalent alkali, clinker moduli, corrected Bogue phase fractions, thesetting-time ratio, and derived particle-size percentiles. Each row represents one cement sample. Producer names and internal cementcodes are not included. Empty fields denote unavailable or non-applicablemeasurements. Oxide, loss-on-ignition, and Bogue quantities are stored as massfractions from 0 to 1. Units, storage scales, derivations, and missing-valueconventions are documented in data_dictionary.csv anddataset_information.md. The processed dataset supports the analyses reported in “Machine LearningInference Limits of Routine Cement Characterization for CEM I Performance:Evidence From a Multi-Producer Dataset” (arXiv:2607.22512).