Sustainable self-compacting concrete: evaluating glass powder and waste marble as cement substitutes through machine learning
摘要
Carbon emissions are generated during cement manufacturing, resulting in global environmental issues. Cement is extensively used in construction processes, particularly in concrete production. The ability of self-compacting concrete (SCC) to flow under its weight without experiencing mechanical vibrations has made it a revolutionary material. Although self-compacting concrete (SCC) offers advantages, including improved workability and reduced labor costs, the production method utilizes a significant amount of cement, enhancing the product's detrimental environmental effects. Researchers examined industrial wastes, such as glass powder and waste marble, as sustainable substitutes for cement in self-compacting concrete (SCC) to address these problems. These materials reduce costs, enhance durability, and minimize their carbon footprint. This study investigates the impact of glass powder and waste marble on the compressive strength (CS) of self-compacting concrete (SCC) for various mix ratios and the effect of different calculated moduli. A total of 498 self-compacting concrete (SCC) mixture compositions, containing glass powder and waste marble, were collected, and a detailed analysis was performed to predict the compressive strength (CS) using five regression and modeling techniques. Linear Regression Model (LR), Non-Linear Regression Model (NLR), Multi-Linear Regression Model (MLR), M5P-tree, and Gaussian Process Regression Squared Exponential (GPR) are utilized in this paper for predicting compressive strength (CS). The independent variables are water-binder ratio, cement, marble powder, glass powder, water, density, and curing time. This study's compressive strength (CS) ranged from 16.12 MPa to 56.56 MPa. The Correlation Coefficient (R2), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Scatter Index (SI), and Objective Function (OBJ) will be utilized to assess model performance across the five models. The Gaussian Process Regression Squared Exponential (GPR) achieved the highest level of accuracy across all assessment criteria. The residual error analysis revealed that the Gaussian process regression squared exponential (GPR) model produced the minimum error among the others. The study determined that curing time was the most significant factor in influencing the compressive strength (CS) of self-compacting concrete (SCC), followed by cement content. The study's results demonstrate that replacing cement with waste marble and glass powder in self-compacting concrete (SCC) is a viable and effective solution for sustainable construction. It highlighted how machine learning techniques can be used to examine construction materials and forecast and optimize concrete qualities. Using machine learning models enables accurately estimating self-compacting concrete's compressive strength (CS) (SCC). These prediction models could lead to more cost-effective, eco-friendly, and efficient concrete mix designs.