Assessment of the strength and behaviour of high-strength concrete improved with metakaolin

Abstract

Concrete is one of the most widely used materials in construction, with ongoing efforts to enhance its performance while minimizing environmental impact. High Strength Concrete (HSC), incorporating mineral admixtures such as metakaolin, has emerged as a promising solution for improving strength, durability, and sustainability. This study investigates the mechanical behavior of HSC with metakaolin replacement ranging from 0% to 25%, using M50 grade concrete and a constant water-cement ratio of 0.32. Experimental tests were conducted to evaluate compressive strength, flexural strength, and tensile strength. The results indicate that 15% metakaolin replacement yielded optimal performance, with compressive strength improving by up to 22.8% compared to the control mix. To further enhance predictive capability, the Efficient CapsNet with Sparse Attention (ECNet-SA) model was developed. When benchmarked against traditional machine learning models, the ECNet-SA model achieved superior Root Mean Square Error (RMSE) of 30.9%, 28.3%, 24%, and 22% better than Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), and Neural Network (NN) for strength prediction. These findings support the potential of both metakaolin as a partial cement replacement and deep learning for performance forecasting in high-strength concrete applications.