Compressive strength prediction for sandcrete blocks with metakaolin: experiment and multiple linear regression analysis

This paper investigates the effects of the partial replacement of cement with metakaolin (MK) on the strength of hollow sandcrete blocks. Hollow sandcrete blocks of size 150 mm × 225 mm × 450 mm were produced using a mix of ratio 1:6 (binder: sand) and cured for 7, 28, 45, and 60 days. For the actual test, the cement was partially replaced with 0–20% metakaolin at 2.5% intervals. Four (4) multiple linear regression (MLR) models were developed for the 7, 28, 45, and 60 days compressive strength of the produced sandcrete blocks. For verification of the MLR models, compressive strength tests were also carried out on samples produced with 4% and 16% metakaolin replacements. Results show that sandcrete block samples produced using 2.5 and 5% metakaolin replacements possess 28 days compressive strengths that are adequate for load-bearing walls in buildings not higher than three (3) storeys, while metakaolin replacements up to 12.5% are acceptable for non-load-bearing purposes. For the MLR models, the coefficient of determination (R2), Root Mean Square Error, Normalized Root Mean Square Error (NRMSE), and absolute prediction error were employed as model evaluation metrics. Although all four (4) MLR models investigated yielded good results, Model M4 has the highest performance (in terms of its maximum R2 value of 0.9907 and least NRMSE of 1.8168%) for the considered 0–20% MK replacement test samples. Model M4 also has the highest prediction accuracies for the compressive strengths of the 4 and 16% MK replacements verification test data. The best MLR model obtained in this study has great potential in predicting the strength of sandcrete blocks with MK replacement, thus reducing the need for multiple laboratory experiments and their associated costs. This study also buttresses the prospect of sustainable construction using waste.