Dataset and Python Codes for Data-Driven Identification of Mechanical Performance Thresholds and Transition Regions in Recycled Aggregate Concrete

The repository was developed to ensure transparency, reproducibility, and open access to the computational workflow used in the study.

The provided materials include:

• Global recycled aggregate concrete (RAC) database compiled from published experimental studies.
• Processed datasets containing normalized compressive strength (NCS), normalized splitting tensile strength (NSTS), normalized flexural strength (NFS), normalized bulk density (NBD), and normalized modulus of elasticity (NME).
• JupyterLab notebooks and Python scripts used for data preprocessing and statistical analysis.
• LOWESS-based nonlinear trend modeling workflows.
• Bootstrap uncertainty quantification procedures and confidence interval analyses.
• Critical transition and threshold identification algorithms.
• Kruskal–Wallis significance testing and effect-size assessment.
• K-Means clustering and dimensionality-reduction analyses.
• Multi-property performance mapping and visualization scripts.
• Performance classification framework and engineering design framework outputs.
• High-resolution publication-ready figures generated for the manuscript.
• Additional analyses, intermediate outputs, and supplementary visualizations that were not included in the final published article.

The repository enables full reproduction of the reported results and provides a reusable framework for future data-driven investigations involving recycled aggregate concrete and other sustainable cementitious materials.