Civil Engineering Faculty Publications
Document Type
Article
Publication Date
4-2026
Abstract
Geopolymer concrete is a promising low-carbon alternative to ordinary Portland cement concrete, but its practical use is limited by complex mix-design requirements and limited cost-aware decision-support tools. This study developed the Intelligent Cost-Optimized Mix Design Prediction and Engineered Strength System (iCOMPRESS), a machine learning-based recommender system that integrates 28-day compressive strength prediction, cost optimization, and compositionally diverse mixture recommendation. A database of 443 literature-derived mixtures was used to train a hyperparameter-optimized Random Forest model with domain-informed features related to binder chemistry, alkaline activation, water content, aggregates, and curing conditions. The model achieved a five-fold cross-validation mean absolute error (MAE) of 6.23 ± 1.18 MPa and an independent test MAE of 7.86 MPa. Feature analysis identified slag fraction, calcium contribution from slag, sodium hydroxide content, binder content, water-to-solids ratio, and curing temperature as key parameters. These results demonstrate iCOMPRESS's potential as a practical tool for cost-efficient and sustainable geopolymer mix design.
Recommended Citation
Natarajan, Yuvaraj, KR Sri Preethaa, V. DanushKumar, Syed Muhammad Oan Naqvi, M. Shyamala Devi, Karen Lozano, Bubryur Kim, and Jinwoo An. "Intelligent cost-optimized mix design prediction and engineered strength system for geopolymer concrete: A machine learning-based recommender system." Developments in the Built Environment (2026): 100951. https://doi.org/10.1016/j.dibe.2026.100951
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Publication Title
Developments in the Built Environment
DOI
10.1016/j.dibe.2026.100951
Comments
© 2026 The Authors.
http://creativecommons.org/licenses/by/4.0/