Development Of Artificial Intelligence Ai Based Predictive Model And Computational Optimiza Tion Of Concrete Properties

Abstract

This research presents a comprehensive investigation into the development of a robust and efficient multi-objective optimization (MOO) framework for developing sustainable, cost-effective, and environmentally friendly concrete. Concrete is a fundamental building material used extensively in construction projects worldwide. However, traditional concrete production processes often result in high carbon dioxide (and#12310;COand#12311;_2) emissions (resulting from high cement usage) and significant costs. Therefore, there is a growing need to develop innovative approaches that optimize concrete properties while considering environmental and economic factors. The first case study focuses on High-Performance Concrete (HPC), which offers superior mechanical properties and durability compared to conventional concrete. In this case study, a tri-objective optimization approach is employed, aiming to simultaneously maximize Compressive Strength, minimize Cost, and reduce associated and#12310;COand#12311;_2 emissions. These objectives are crucial in the design and production of HPC, as they directly impact structural performance, financial aspects, and environmental sustainability. To address the challenge of accurately predicting Compressive Strength, a novel machine learning (ML) model is proposed as an enhancement to the conventional Backpropagation Neural Network (BPNN). The proposed ensemble BPNN model combines the strengths of multiple individual models, improving accuracy and reliability. Through extensive training and validation using a 10-fold cross-validation approach, the ensemble model exhibits significantly enhanced performance compared to the single BPNN model. It reduces the Root Mean Square Error (RMSE) from 14.48 to 4.009 MPa, indicating a substantial improvement in predicting Compressive Strength. Furthermore, the Coefficient of Determination (R2) increases from 0.14 to 0.93, signifying a strong correlation between predicted and experimental values. The improved performance of the ensemble BPNN model translates into practical benefits in the