Concrete Spalling Data

This dataset compiles comprehensive experimental results on the thermal-induced spalling behavior of concrete from 55 published studies covering a broad range of material compositions, specimen geometries, and fire exposure conditions.
It integrates 1,430 samples featuring multi-scale descriptors at the material, geometric, and thermal levels — including parameters such as binder ratios (W/B, FA/B, CA/B), fiber geometry, density, moisture content, specimen dimensions, heating rate, and peak temperature.

The dataset is primarily used to support the development and validation of a hierarchical, data-driven framework for predicting and interpreting post-fire concrete spalling. It enables the study of both binary classification (spalling occurrence and explosivity) and regression analysis (residual mass loss and spalling depth) through explainable machine learning models such as XGBoost and SHAP.

All variables are standardized in consistent physical units and organized according to three feature groups — Material, Physical & Geometric, and Thermal. This dataset provides a unified benchmark for future research on data-driven fire performance assessment and mechanistic understanding of spalling phenomena in cementitious materials.