Data-Driven Design and Optimization of Sustainable and Low-Carbon Calcium Sulfoaluminate Cement Blends Incorporating Blast Furnace Slag

Ahmad Nawaz^1,*, Hilal Khan², Salamat Ullah³, Hamad Almujibah^4,5, Ali E. A. Elshekh⁵, Maaz Osman Bashir⁵
1 Guangdong Provincial Key Laboratory of Durability for Marine Civil Engineering, College of Civil and Transportation Engineering, Shenzhen University (SZU), Shenzhen, China
2 NUST Institute of Civil Engineering (NICE), School of Civil and Environmental Engineering (SCEE), National University of Sciences and Technology (NUST), Sector H-12, Islamabad, Pakistan
3 Center for Mechanics under Extreme Environments, Ningbo University, Ningbo, China
4 Research Center of Basic Sciences, Engineering and High Altitude, Taif University, Taif, Saudi Arabia
5 Department of Civil Engineering, College of Engineering, Taif University, Taif, Saudi Arabia
* Corresponding Author: Ahmad Nawaz. Email: email
(This article belongs to the Special Issue: Machine Learning, Data-Driven and Novel Approaches in Computational Mechanics)

Computer Modeling in Engineering & Sciences https://doi.org/10.32604/cmes.2026.086781

Received 05 June 2026; Accepted 30 July 2026; Published online 10 August 2026

Download PDF

Abstract

Calcium sulfoaluminate (CSA) cement is considered a promising low-carbon alternative to ordinary Portland cement owing to its lower clinkerization temperature and reduced CO₂ emissions. The incorporation of blast furnace slag can further enhance the sustainability of CSA-based binders by lowering clinker content, reducing cost and embodied carbon emissions, while maintaining satisfactory mechanical performance. However, optimizing CSA-slag systems remains challenging due to the complex interactions among binder composition, clinker mineralogy, and slag replacement levels. This study therefore aims to predict the compressive strength of CSA-slag binders, identify the mixture parameters governing it, and optimize mixture proportions for balanced mechanical, environmental, and economic performance, using a machine learning (ML)-based framework trained on 232 experimental samples compiled from the literature. Three ML models, Decision Tree (DT), Random Forest (RF), and Extreme Gradient Boosting (XGB), were optimized using a Genetic Algorithm and integrated through stacked ensemble learning with a multilayer perceptron meta-learner. Among the developed models, the DT-XGB ensemble achieved the highest predictive accuracy for compressive strength (R² = 0.965, RMSE = 3.168 MPa). Shapley additive explanations (SHAP) analysis identified the water-to-cement ratio as the most influential parameter governing compressive-strength prediction, followed by CSA clinker content, curing age, ye’elimite content, M (sulfate-to-ye’elimite ratio), slag content, and belite content. Furthermore, a multi-objective optimization framework based on NSGA-II was employed to simultaneously optimize compressive strength, embodied CO₂ emissions, and cost. The optimum mixture contained 30% CSA clinker and 70% slag at a water-to-cement ratio of 0.4, achieving 57.9 MPa compressive strength, 379 kg/m³ embodied CO₂ emissions, and a cost of 91.8 USD/m³. These findings provide an effective strategy for designing sustainable, high-performance, and low-carbon CSA-based cementitious materials.