• Quantified the effect of w/c ratio on multivariate NDT strength prediction accuracy. • Combined rebound hammer, UPV, and curing age in a mix-sensitive regression model. • Proposed model achieved low prediction errors (RMSE = 1.46 MPa; MAE = 1.26 MPa). • Demonstrated limitations of generic SonReb correlations for local materials. • Developed a locally calibrated correction factor for rebound-based strength estimates. Accurate estimation of concrete compressive strength using non-destructive testing (NDT) methods remains challenging because widely used empirical correlations often neglect key mix design parameters, particularly the water–cement (w/c) ratio. This study investigates the influence of the w/c ratio on concrete compressive strength and on the predictive performance of combined rebound hammer (RH) and ultrasonic pulse velocity (UPV) models through controlled laboratory experiments using locally sourced Nigerian materials. Six concrete mixes with w/c ratios ranging from 0.45 to 0.70 were prepared with a constant 1:1:2 mix proportion and tested at curing ages of 7, 14, and 28 days. RH and UPV measurements were conducted on laboratory-cured cube specimens and correlated with compressive strength obtained from standard destructive tests. Pearson correlation analysis revealed strong positive relationships between compressive strength and RH (r = 0.96) and UPV (r = 0.93), while increasing w/c ratio consistently reduced both strength and NDT responses. A multivariate regression model incorporating RH, UPV, and curing age was developed to predict compressive strength across the investigated mixes. The model achieved a coefficient of determination (R²) of 0.79 with low prediction errors (RMSE = 1.46 MPa; MAE = 1.26 MPa). Comparative analysis showed that the proposed model outperformed single-parameter approaches and highlighted the limitations of generic SonReb correlations when applied without local calibration. The results demonstrate that reliable NDT-based strength prediction is strongly mix-sensitive and requires locally calibrated models for accurate laboratory-based quality control.
Hassan et al. (Sun,) studied this question.