Performance evaluation of M30 grade concrete under carbonation influenced by humidity and temperature variations using artificial neural network

This paper examines the interactive effects of relative humidity, curing temperature, and curing age on the performance and carbonation resistance of M30-grade concrete through experimental analysis and artificial neural network (ANN) modelling. Specimens of concrete were cured under controlled conditions at 27 °C and 50 °C, with relative humidities of 40, 60, 80, and 100, at curing ages of 7 and 28 days. Non-destructive testing methods such as ultrasonic pulse velocity (UPV) and rebound hammer tests, as well as carbonation depth analysis using a phenolphthalein indicator solution, were used to assess concrete performance. The experimental evidence shows that the most important parameters determining concrete performance are relative humidity and curing age, with a slight influence of curing temperature within the studied range. The values of UPV were higher in relation to high relative humidity, which pointed to the high density of microstructure, and the rebound hammer strength was lower with high relative humidity, which meant the saturation of pore-water. When the relative humidity was increased, the carbonation depth steadily declined, with the minimum observed at 100% RH, confirming that the moisture-controlled diffusion mechanism of CO₂ ingress was at play. The depth of carbonation was high at older curing ages and under reduced humidity. ANN-based predictive models have also been developed to estimate UPV, compressive strength, and carbonation depth using humidity, temperature, and curing age as predictors. Compared with linear regression models, the ANN models were highly predictive, with lower MAE and RMSE values and higher R² coefficients (R² > 0.97), especially for predicting carbonation depth. One important result of this research is that curing temperature showed no statistically significant influence within the limited range (27–50 °C) and the short conditioning duration considered in this study. The results provide important insights into moisture-dependent carbonation behaviour and demonstrate the effectiveness of ANN models for predictive durability assessment.