Clustering Analysis of Compressive Strength of Structural Recycled Aggregate Concrete

Clustering analysis primarily highlights the in homogeneity of data and can be utilized in structural engineering to demonstrate strength irregularity. It is well-known that strength irregularity between neighboring floors within a structure or among structural elements can lead to non-holistic behavior. Therefore, the clustering of compressive strength holds significant importance. Despite the relevance, only a few studies have addressed the clustering of compressive strength in recycled aggregate concrete (RAC) and proposed potential solutions for clustering issues. This paper aims to investigate the clustering of compressive strength in RAC and explore viable solutions. In this experimental study, four concrete groups were produced under standard conditions. The first group included natural aggregate concretes (NAC) designed with the Absolute Volume Method (AVM) as control concretes. The second group, comprised of RAC, was designed with the equivalent mortar volume method (EVM) as the control RAC. The third group consisted of RAC treated with silica fume (SF) and designed using AVM, while the fourth group included RAC designed with EVM. Statistical analyses were conducted on the 28-day compressive strength test results. The results indicated that the strength class of compressive strength clusters varied among the four groups. The clustering of test results was influenced by the type of concrete components used and the design method employed. Additionally, using silica fume and adopting the Absolute Volume Method reduced strength fluctuation and regulated the strength class of clusters by bringing them closer together. In contrast, the Equivalent Mortar Volume Method resulted in a greater dispersion of strength classes. The clustering effect of recycled aggregate (RA) was more pronounced than that of natural aggregate (NA). Given these findings, it is essential to implement measures when utilizing RAC in sustainable structures to address potential clustering issues.