MACHINE LEARNING PREDICTION OF THE BASELINE COMPRESSIVE STRENGTH AND COMPRESSIVE STRENGTH OF SCM CONCRETE SUBJECTED TO CHLORIDE INGRESSION

Abstract

This study explores the potential advantages of using multiple machine learning (ML) models to predict the baseline compressive strength (CS) and CS of SCM concrete subjected to chloride ingression. Four ML models, including Random Forest (RF), Gradient Boosting (GradBoost), Extreme Gradient Boosting (XgBoost), and Adaptive Boosting (AdaBoost) algorithms were employed to construct forward prediction models for predicting the baseline CS and CS of SCM concrete with combinations of fly-ash (FA), silica fume (SF), and metakaolin subjected to chloride ingression. The performance of the models was compared and assessed using the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and mean square error (MSE). The performance of the model was further validated using the k-fold cross-validation approach. Compared to the other employed approaches, the GradBoost model was more effective in predicting results, with an R2 value of 0.8612. To further enhance model interpretability, Shapley Additive Explanation (SHAP) was employed to assess the relative importances of input variables. The use of ML techniques to predict the mechanical properties of SCM concrete subjected to chloride ingression will be beneficial to the field of civil engineering because it will save time, effort, and resources compared to carrying out the laboratory experiments. The proposed techniques pave the way towards the intelligent design of concrete elements and structures.