Machine learning and RSM-CCD analysis of green concrete made from waste water plastic bottle caps: Towards performance and optimization

Abstract

This study aims to serve as a performance indicator for the workability and strength of concrete when coarse aggregate, sand, cement, and water are partially substituted with waste plastic bottle caps. The significance of these alternative plastic water bottle caps is to reduce plastic trash that is difficult to lapse and to prevent waste that can be transformed to something that may be employed in the advancement of technology in the future. The principle of "Reduce, Reuse, and Recycle" is used, which not only lowers environmental pollution but also reduces costs. In the building industry, concrete is the most often utilized material. In order to preserve natural resources and minimize the number of materials that end up in landfills, green construction is becoming a more significant worldwide issue. Empty cans and bottle tops from drinking water bottle produce a lot of garbage. The difficulty of biodegrading plastic trash and the need for techniques for recycling or reuse make this a problem for the environment. Such problems are being investigated in this study in order to determine whether it could be possible to partially replace coarse aggregate with 0, 6 and 12% in the manufacturing of concrete using discarded bottle caps. To evaluate the compressive strength, split tensile, and flexural strength test characteristics in a laboratory setting; waste bottle caps were used as replacements for coarse aggregate at different percentages. The highest compressive strength was determined to be 28.80 MPa with 12% replacement of used plastic bottle caps with a water cement ratio of 0.55. Advanced statistical methods, including RSM-CCD (Response Surface Method-Central Composite Design) and machine learning models ANN-LM (Artificial Neural Network- Levenberg Marquardt), were applied in this study to predict concrete performances based on mix design variations. It was found that ANN-LM model displayed more accurate prediction relative to RSM-CCD method.