Compressive behavior of eco-friendly concrete containing glass waste and recycled concrete aggregate using experimental investigation and machine learning techniques

Abstract

This paper presents an experimental investigation on the compressive behavior of eco-friendly concrete containing glass waste (GW) and recycled concrete aggregate (RCA). In all mixtures, cement and water are stayed constant while fine and coarse aggregates were varied. The studied parameters are: utilizing fine recycled glass (FRG) instead of fine natural aggregate (FNA) at substitution levels of 0, 10, 25, 50 and 100 %, performing coarse recycled glass (CRG) instead of coarse natural aggregate (CNA) at replacement ratios of 0, 10, 20 and 40 %, conducting both CRG at constant ratio (20 %) as well as RCA at varied levels (0, 16, 40 and 80 %) and replacing CNA by RCA at varied levels (0, 16, 40 and 80 %). The experimental results showed that replacing 10, 25, 50 and 100 % of FNA by FRG declined the compressive strength of concrete by 12.8, 18.5, 24.5 and 49.8 %, respectively. Also, the compressive strength of concrete decreased by 18.5, 23.3, and 32.83 % when 10, 20, and 40 % of CNA were substituted by CRG, respectively. Additionally, the current study investigates the efficacy of machine learning (ML) techniques in evaluating the compressive strength of eco-friendly concrete containing GW and RCA separately. Four tree-based ensemble methods—decision trees, random forest, gradient boosted regression trees, and extreme gradient boosting—were trained and tested using 241 and 319 datasets to predict the compressive strength of eco-friendly concrete containing GW and RCA, respectively. The data employed in constructing the ML model were gathered from both the current experimental study and previous studies. The findings indicate that the XGBoost model demonstrates exceptional accuracy. The impact of each parameter on the predicted outcomes was discussed.