Evaluating embodied energy, carbon impact, and predictive precision through machine learning for pavers manufactured with treated recycled construction and demolition waste aggregate.

This investigation assesses the embodied energy and carbon footprint in the manufacture of pavers using varying proportions of recycled Construction and Demolition Waste (CDW). Additionally, Thin Film Composite Polyamide fiber (TFC PA), extracted from end-of-life Reverse Osmosis (RO) membranes, is introduced as an additive to enhance the concrete's strength. Machine learning techniques, namely Artificial Neural Network (ANN), Support Vector Regression (SVR), and Response Surface Methodology (RSM), are employed to predict the mechanical properties of pavers. The study focuses on examining the energy required and embodied carbon in various mix proportions, as well as the mechanical properties-specifically compressive strength and split tensile strength of concrete with different CDW and TFC PA proportions. Findings reveal that the optimal percentage of TFC PA is 3 % for all CDW replacement proportions, resulting in low carbon content both in terms of energy and embodiment and in mechanical behavior. The implementation of ANN and SVR is conducted in MATLAB, while a Design Expert is employed to generate the experimental design for RSM. The RSM regression model demonstrates a robust correlation between variables and observed outcomes, with optimal p-values, R2 values, and f-values. The ANN model successfully captures the variability in the data. Additionally, the findings indicate a consistent superiority of the Support Vector Regression (SVR) model over both Artificial Neural Network (ANN) and Response Surface Model (RSM) models when considering diverse performance metrics such as residuals and correlation coefficients.