Optimizing sustainable blended concrete mixes using deep learning and multi-objective optimization.

Area of Science:

• Materials Science
• Civil Engineering
• Computational Science

Background:

• Developing sustainable concrete is crucial for reducing environmental impact in construction.
• Optimizing concrete mix designs involves balancing competing objectives like strength, cost, and environmental footprint.
• Traditional methods often struggle to efficiently explore the complex trade-offs in mix design.

Purpose of the Study:

• To develop an integrated data-driven framework for designing environmentally friendly concrete mixes.
• To utilize deep neural networks (DNNs) and multi-objective optimization for optimizing concrete properties.
• To provide practitioners with decision-support tools for cost-effective and sustainable concrete solutions.

Main Methods:

• A deep neural network (DNN) model was trained on a comprehensive dataset of concrete mix parameters and curing conditions.
• Bayesian hyperparameter tuning was employed to optimize the DNN configuration, achieving high predictive accuracy (R²=0.936, RMSE=5.71 MPa).
• The Multi-Objective Particle Swarm Optimization (MOPSO) algorithm was used to identify optimal mix designs balancing strength, cost, and cement reduction.

Main Results:

• Optimized concrete mixes achieved over 50 MPa compressive strength with up to 25% cement reduction.
• A total cost reduction of 15% was realized compared to standard concrete mixes.
• Feature importance analysis identified cement content and concrete age as key factors influencing compressive strength.

Conclusions:

• The proposed framework effectively integrates DNNs and MOPSO for designing high-performance, sustainable concrete.
• The data-driven approach provides valuable insights into green concrete technology and facilitates practical application in construction.
• This methodology enables the discovery of optimal mix proportions that enhance strength, reduce costs, and minimize environmental impact.