Development of statistically validated empirical models for rice husk ash and silica fume blended concrete with uncertainty and reliability assessment

The rise in the demand for sustainable concrete necessitates deploying greater levels of industrial and agricultural waste-derived supplementary cementitious materials, whilst achieving predictable engineering performance. Predicting multiple engineering properties of rice husk ash (RHA)-silica fume (SF) blended concrete is complex because of its heterogeneous behaviour, thus highlighting the need for simple and statistically validated predictive tools. RHA and SF have been extensively studied as sustainable supplementary cementitious materials, but few empirical models have been validated that can predict several engineering properties of RHA-SF blended concrete. To address this gap, this study evaluates 36 concrete mixtures including RHA-SF combinations and develops statistically validated multiple regression models for key engineering properties. Residual analysis, 5000 iteration bootstrap resampling, Leave One Out Cross Validation (LOOCV), sensitivity analysis, prediction uncertainty and reliability assessment, and 1000 iteration Monte Carlo simulations are used to assess model reliability. The adjusted R 2 values of the selected models are in the range of 0.833–0.992 and the cross-validation results of all the properties, except RCPT, are satisfactory. The experimental results show that optimized RHA-SF blends show 27% improvement in 28-day compressive strength, 41.52% reduction in water absorption and 72.02% reduction in chloride permeability. The proposed framework provides transparent and statistically robust predictive tools for preliminary assessment, quality control, and broader implementation of sustainable RHA-SF blended concrete.