Emerging infrastructure development demands the use of advanced construction materials such as nanomaterials and high-performance concrete (HPC). Simultaneously, assessing the environmental impact (EI) of these materials is critically important for ensuring sustainable development. In this study, an artificial neural network (ANN), a well-established modeling approach, was applied to assess the EI and eco-friendliness of various nanomaterial-impregnated HPC (NHPC). A data set comprising 713 data points of material quantities and associated EI metrics was compiled from a combination of secondary data sources such as peer-reviewed published literature, life cycle assessment (LCA) case studies, and standardized environmental databases such as Ecoinvent and GaBi. In this study, ANN was employed for both regression and classification tasks. Using a multilayer back-propagation ANN approach, the regression model was trained to predict values across nine midpoint EI categories defined by the LCA method. Based on these outputs, a classification model was developed to categorize the environmental risk levels of NHPC as eco-friendly, moderate risk, and high risk. Among five ANN configurations, ANN-V exhibited the lowest prediction errors root mean square error (RMSE) = 0.027 and mean absolute error (MAE) = 0.021 with the highest coefficient of correlation (R2=0.93) even when subjected to fivefold cross-validation. Several alternative ANN architectures with varying hidden layers (one to three) and neurons per layer (5–15) were initially tested. The configuration of two hidden layers with nine neurons each (ANN-V) was chosen because it consistently produced the lowest error metrics and stable convergence across folds, which balanced model complexity with generalization ability. Also, the uncertainty and robustness of ANN regression and classification models were validated using Monte Carlo simulation and confusion matrix, respectively. Environmental risk in this study was defined as a normalized cumulative score across the nine midpoint impact categories, which is expressed as a percentage. Based on the predicted risk scores and computed risk priority numbers, a risk ranking system was developed for selecting the viable nanomaterial. The results revealed that the commercial nanomaterials such as nanosilica, nano calcium carbonate, and nanoalumina each posed a high environmental risk of 9% mainly due to their energy-intensive production processes. Although consistent with earlier studies, the outcome was validated across nine EI categories using ANN, which ensured a more rigorous evidence base and facilitated the development of a robust risk-ranking system. These findings indicated that the waste-derived nanomaterials are more sustainable alternatives for next-generation HPC applications.
Kalimuthu et al. (Fri,) studied this question.