Research on nondestructive testing of compressive strength of rapid-hardening cement-based material based on rebound method and random forest model

To address the challenge of on-site compressive strength detection of rapid-hardening cement-based materials (RHCBM), this study develops a non-destructive testing method and an accurate prediction model. Four RHCBM mixtures were prepared, and standard cubic specimens were tested at 13 curing ages from 1 to 28 days. A multi-dimensional database including rebound numbers, compressive strengths, and curing ages was established using rebound and compressive strength tests. Results show that both rebound numbers and compressive strengths exhibit similar age-dependent growth, with rapid increases in the first 7 days and stabilization after 21 days. Five traditional mathematical models and a random forest regression (RFR) algorithm were applied to model the relationship between rebound numbers and compressive strengths. A parabolic function was identified as the optimal rebound strength measurement curve, and a C-R-M RFR model was developed using k-fold cross-validation. Comparative analysis indicates that the C-R-M RFR model improves prediction reliability by approximately 15% compared with the parabolic model, with R2 values of 0.98 for both training and test sets. The results demonstrate that the proposed C-R-M RFR model offers higher accuracy and reliability, providing a practical solution for on-site non-destructive compressive strength prediction of RHCBM.