Research

This research proposes a data-driven machine learning approach aimed at optimizing the cement content in concrete used for transportation infrastructure, such as roads, bridges, and retaining walls. The production of cement is a significant contributor to CO₂ emissions and increasing costs in construction projects. By reducing cement content using optimized multi-component mixtures, this method tackles both ecological and performance requirements of transportation structures. A comprehensive dataset of concrete mix designs, which includes supplementary cementitious materials (SCMs) and nanomaterials, was compiled to support the decision-making process. A hybrid machine learning model that combines Artificial Neural Networks (ANN), Genetic Algorithms (GA), and XGBoost was developed to predict compressive strength and determine the best mix combinations for high-load applications, like pavements and bridge decks. To better handle incomplete and diverse data, a binary-feature masking normalization technique was introduced, which improves model accuracy and reliability. The proposed framework effectively identifies concrete mixes that significantly decrease cement usage while maintaining structural strength. The use of SCMs such as fly ash, silica fume, and metakaolin, along with nanomaterials like carbon nanotubes and nanoclay, showed synergistic benefits in enhancing strength and durability, critical factors for enduring transportation infrastructure. This study underscores the promise of AI-driven material design in advancing low-carbon, high performance concrete solutions for roads and bridges, supporting global sustainability objectives and enhancing infrastructure resilience.

This paper introduces an innovative data-driven approach to designing green concrete by minimizing cement content through the optimization of multi-component blends. A comprehensive dataset of 2,332 experimental concrete mix designs incorporating supplementary cementitious materials (SCMs) and nanomaterials was compiled from 31 scientific publications. A hybrid machine learning model combining Artificial Neural Networks (ANN), Genetic Algorithms (GA), and XGBoost was developed to predict compressive strength and identify optimal mix compositions. To address the challenge of incomplete data, a binary-feature masking normalization strategy was implemented, allowing the model to process heterogeneous datasets with improved accuracy and robustness. The proposed framework successfully identifies sustainable concrete mixes that reduce cement consumption by up to 40% without compromising mechanical performance. Hybrid combinations of SCMs such as fly ash, silica fume, and metakaolin, alongside nanomaterials like carbon nanotubes and nanoclay, demonstrated synergistic effects in enhancing strength while lowering environmental impact. The results confirm that intelligent integration of AI techniques with material science can significantly advance the development of high-performance, low-carbon concrete, offering a scalable solution for the construction industry's sustainability goals.