A probabilistic framework for early-age concrete strength assessment using infrared thermography-based spatial thermal information

Accurate assessment of early-age concrete strength is critical, yet conventional approaches rely on limited point measurements and produce single deterministic strength estimates, potentially leading to safety risks. This study proposes a probabilistic framework that integrates infrared thermography, deep learning segmentation, and pixel-to-pixel registration to nondestructively capture spatially distributed surface temperature histories. These pixel-level thermal data are transformed into probability distributions of strength through a maturity-based model. Two field placements, one in summer and one in winter, were used to demonstrate the framework and to compare the contribution of the observed surface thermal field with that of the maturity model. In sensitivity analyses of mask contamination and emissivity, the resulting strength deviation remained within approximately 1 MPa beyond the first day, whereas assumed maturity-model coefficients of variation of 10% and 20% produced substantially larger differences in the predicted time of target-strength achievement. Model calibration is therefore identified as the higher-priority route for improving decision confidence. Within the scope examined, the framework supports a shift in quality control from a single pass/fail judgment toward an uncertainty-informed assessment of target-strength achievement; it supplies supplementary surface-temperature information and uncertainty scenarios rather than a validated substitute for established strength testing and structural checks.