Prediction of compressive strength of concrete under various curing conditions: a comparison of machine learning models and empirical mathematical models

Considering the pivotal role of compressive strength in assessing concrete quality, accurately predicting it is essential for guiding construction practices. However, conventional techniques have primarily emphasized factors like raw admixture or curing temperature, often neglecting the influence of relative humidity. This study focuses on predicting the compressive strength of concrete, conducting an in-depth analysis of concrete under different curing conditions by combining experimental data and machine learning models. The study found that the XGBoost model performed the best in terms of prediction accuracy, with an R² of 0.9264, RMSE of 2.9255, and MAE of 2.0314. In comparison, traditional mathematical models such as the ACI model and the fib model performed poorly in predicting compressive strength affected by relative humidity. Additionally, the study revealed that curing time is the most critical factor influencing concrete compressive strength, and the contribution of curing relative humidity to strength is close to that of temperature, emphasizing the importance of considering relative humidity when evaluating strength growth. Furthermore, analysis of parameters indicated that temperature, relative humidity, and curing time exhibit diverse trends in their effects on concrete compressive strength under different curing conditions. Despite the effectiveness of machine learning methods in predicting concrete compressive strength, challenges remain in model interpretability and real-world application. Future research could explore more advanced machine learning models and utilize larger and more diverse datasets to enhance prediction capabilities.