Answers are grounded in this paper and cite their sources.
The complex concrete behaviour is based on several interrelated parameters such as age, curing conditions, pozzolanic materials, admixtures, and cement content. The concrete compressive strength established using concrete casting and curing is labour-intensive, time-consuming, and delays quality evaluation until the completion of the curing period. Predictive models bridge this gap by evaluating the compressive strength based on various parameters, assisting engineers to optimise designs and fine-tune mixtures proactively. In this research, the compressive strength of concrete containing corn cob ash (CCA) and gum Arabic (GA) was predicted using artificial neural network (ANN), extreme gradient boosting (XGBoost), k-nearest neighbour (KNN), and stacked ensemble models. Subsequently, Taylor analysis was used for visual comparison of multiple model outputs while SHapley Additive exPlanation (SHAP) was synchronised using the best performing model. Surrogate, accumulate local effect (ALE), Cook's distance, as well as residual and leverage values were also applied to capture the simplified explanation of the model features. The results showed that the XGBoost model achieved lowest mean absolute error (MAE) of 0.292 MPa under training stage. The ANN model was the best model during the testing phase. The SHAP parameters identified age as the most critical parameter, contributing towards model prediction followed by GA and cement. The SHAP results suggested that increasing such critical parameters could achieve high model prediction. The global surrogate model analysis revealed that the decision tree (DT) (depth = 8) was the best surrogate model, offering the highest R 2 , lowest root mean square error (RMSE), and ShapGAD values, while exhibiting five variables with significant feature importance values. Therefore, the machine learning algorithms in this study can be adopted as novel predictive approaches, empowering more informed decision-making during high-strength concrete design and quality control.
No takes yet. Share an insight, caveat, or question.
Sinkhonde et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: