A comprehensive study on the determination of compressive strength of concrete containing fly ash and blast furnace slag based on machine and ensemble learning

The compressive strength (CS) of concrete is a critical parameter for the safety and longevity of structures as it directly affects the load bearing capacity and durability. However, determining the CS by conventional methods is time consuming, costly and requires a large amount of sample preparation. This study aims to quickly and cost-effectively predict and classify the CS of concrete and determine the influence of components on strength using machine learning (ML) algorithms as an alternative to traditional methods. Support Vector Machines (SVM), Decision Trees (DT), Multilayer Perceptron (MLP) machine learning algorithms and Random Forest (RF), Gradient Boosting (GB) and Extreme Gradient Boosting (XGB) ensemble learning models were trained on a dataset consisting of 1030 data points of concrete containing fly ash (FA) and blast furnace slag (BFS). The dataset was split into 75% training and 25% testing, and the Grid Search method and 5-fold cross-validation were applied in the training process. According to the results of the study, the XGB model showed the most robust performance in the prediction and classification of the CS of concrete with an R2 of 0.931 and an accuracy of 0.901. However, SVM and DT demonstrated inferior performance relative to the other four models. In addition, the models were classified normal strength concrete more successfully than low and high strength concrete. It was determined that the two most effective factors on CS were concrete age and cement dosage. While the increase in concrete age and cement dosage increased the strength, the increase in water content decreased the strength.