Developing a data-driven framework to predict the compressive strength of manufactured sand and fly ash concrete under varying curing ages

Abstract

The depletion of river sand supplies due to overexploitation and illegal sand mining has become a growing concern. Additionally, the production of one ton of Portland cement, the primary binder in concrete, releases an equivalent amount of carbon dioxide into the atmosphere. In response, this study aimed to replace conventional cement and river sand with recycled waste materials, such as fly ash and manufactured sand (M sand). Concrete mix proportions were designed for M40 grade, with Ordinary Portland Cement (OPC) and river sand being substituted with 0–85 wt% fly ash and 0–100 wt% M sand. The concrete samples were tested for compressive strength after curing for 3–90 days. Machine learning (ML) models, such as “Bat research algorithm (Bat)”, “Cuckoo research algorithm (Cuckoo), “Elephant research algorithm (Elephant), “FireFly research algorithm (FireFly)”, “Rhinoceros research algorithm (Rhino)” and “GrayWolf research algorithm” (Wolf), were utilized to predict the compressive strength of the samples. The results showed that the incorporation of fly ash and M sand improved the compressive strength of the concrete. After 28 days of curing, partial replacement of OPC and river sand with 25 wt% fly ash and 50 wt% M sand achieved the targeted strength for M40 grade concrete. Among the ML models, present models in the summary table, which use decision tree-based techniques like DT-Bat, DT-Cuckoo, DT-Elephant, and others, display varied performance across training and validation datasets. Therefore, the metaheuristic methods prove to be a reliable technique for predicting the compressive strength of concrete containing fly ash and M sand.