Machine learning and multicriteria analysis for prediction of compressive strength and sustainability of cementitious materials

Abstract

Achieving an optimal concrete mix design is critical for mechanical performance and sustainability, particularly by incorporating supplementary cementitious materials to promote eco-friendly concrete. This study introduces an intelligent concrete mix design method that optimizes performance and integrates machine learning and multi-criteria decision-making techniques. In the initial phase, three machine learning models—Decision Tree, Random Forest, and Multi-layer Perceptron—were developed and trained on a dataset of 1030 records to predict sustainable concrete's compressive strength accurately. Among these models, the Random Forest model demonstrated the highest accuracy, exceeding 90 % in testing, affirming its superior predictive capability for concrete strength compared to other models reported in the literature. In Tthe optimization phase, multi-objective optimization was involved. Thus, compressive strength was optimized alongside sustainability criteria, including CO₂ emissions and cost-effectiveness. This optimization was processed by Pareto analysis and the Technique for Order of Preference by Similarity to the Ideal Solution (TOPSIS) to identify the most effective mix design, which was achieved with a binary combination of supplementary cementitious materials. This combination was identified as the top-ranked mix regarding sustainability and performance metrics. For the validation in practical cases involving commercial buildings, it has acheived 27 % cost reduction and 63 % decrease in CO₂ emissions compared to conventional concrete mixing. This intelligent mix design approach significantly advances sustainable concrete development in reducing environmental impact as well as promoting cost-effective.

Keywords

• Machine learning
• Supplementary cementitious materials
• Multi-objective Optimization
• Compressive strength
• Enviro-economic analysis