Machine learning optimization and prediction of waste glass used as partial replacement of coarse aggregate in concrete

Waste glass of the same property replaces coarse aggregate in this project. Waste glass will be replaced at 5%, 10%, 15%, 20%, and 25%, and the cubes will be cast and tested for compressive strength and workability at 7, 14, and 28 days. The project’s goal is to compare the strength of normal concrete to concrete with varied percentages of discarded glass. This research article presents a comprehensive analysis of the replacement of coarse aggregate in concrete, employing Response Surface Methodology (RSM), Multiple Linear Regression (MLR), and Artificial Neural Networks (ANN) as predictive models. To determine physical qualities, cement, fine aggregate, and coarse aggregate are tested. Slump cone and compressive strength tests are performed on fresh and hardened concrete, respectively. This casted M20 grade cube of 150 mm side undergoes all the aforesaid tests according to Indian standards. 5%, 10%, 15%, 20%, and 25% waste glass is added to M20 grade concrete. The investigation showed that the proposed equations generated more accurate compressive strength in optimized RSM and forecasts than MLR and ANN. The proposed equations have lower MAE, RMSE, mean, and R2 values than the LRM and ANN. The compressive strength testing machine measures cube compressive strength. The data's variability was captured by the ANN model's effective R2threshold.