A hybrid prediction and multi-objective optimization framework for limestone calcined clay cement concrete mixture...

Area of Science:

• Materials Science
• Civil Engineering
• Environmental Science

Background:

• Limestone calcined clay cement (LC³) offers comparable mechanical properties to ordinary Portland cement (OPC) with a significantly lower carbon footprint.
• Large-scale LC³ implementation can reduce embodied CO₂ emissions by over 30% compared to OPC.
• Developing cost-effective and sustainable LC³ mixtures is crucial for widespread adoption.

Purpose of the Study:

• To propose a hybrid framework combining machine learning (ML) and multi-objective optimization (MOO) for designing LC³ mixtures.
• To predict the compressive strength of LC³ mixtures and identify key influencing factors.
• To optimize LC³ mix designs for reduced cost and embodied carbon while meeting performance requirements.

Main Methods:

• A dataset of 387 LC³ specimens was used to train ML models for compressive strength prediction.
• The Multivariate Imputation by Chained Equations-Extreme Gradient Boosting (MICE-XGBoost) model achieved high prediction accuracy (R² = 0.928).
• SHAP analysis identified critical factors affecting strength, and Non-dominated Sorting Genetic Algorithm-II was used for multi-objective optimization.

Main Results:

• The MICE-XGBoost model accurately predicted LC³ compressive strength.
• Key factors influencing strength, such as water-to-binder ratio and kaolinite content, were identified.
• Multi-objective optimization yielded Pareto fronts, demonstrating balanced reductions in cost (13.06%) and embodied carbon (14.83%).

Conclusions:

• The hybrid ML-MOO framework effectively designs sustainable and cost-effective LC³ mixtures.
• Inflection points on Pareto fronts provide guidance for selecting optimal low-medium grade LC³ formulations.
• The study offers practical solutions for sustainable LC³ mix design, contributing to reduced environmental impact in construction.