Optimized Machine Learning Model for Predicting Compressive Strength of Alkali-Activated Concrete Through...

Area of Science:

• Materials Science and Engineering
• Sustainable Construction Materials
• Computational Materials Science

Background:

• Alkali-activated concrete (AAC) utilizes industrial by-products (fly ash, slag) as sustainable alternatives to Portland cement, significantly reducing carbon emissions.
• Predicting AAC compressive strength is challenging due to formulation variability, hindering widespread adoption.
• Accurate strength prediction is crucial for reliable performance and material optimization.

Purpose of the Study:

• To develop a robust machine learning (ML) model for accurately predicting the compressive strength of alkali-activated concrete (AAC).
• To identify optimal input variable schemes and machine learning algorithms for AAC strength prediction.
• To enhance model interpretability and provide insights into AAC formulation-strength relationships.

Main Methods:

• Curated an extensive dataset of 1756 unique AAC mixtures.
• Evaluated four input variable schemes and compared ML algorithms: Random Forest, AdaBoost, Gradient Boosting Regression Trees, and Extreme Gradient Boosting (XGBoost).
• Optimized the best-performing XGBoost model using Gray Wolf Optimizer (GWO), Whale Optimization Algorithm (WOA), Beetle Antennae Search (BAS), and Bayesian Optimization (BO).
• Applied SHapely Additive exPlanations (SHAP) for model interpretability.

Main Results:

• The XGBoost model demonstrated superior performance over other ML algorithms.
• Optimized XGBoost achieved a coefficient of determination (R²) of 0.99 on the training set and 0.94 on the entire dataset.
• SHAP analysis provided insights into the influence of various input parameters on AAC compressive strength.

Conclusions:

• A synergistic approach combining careful input selection, hyperparameter optimization, and model interpretability significantly enhances AAC strength prediction accuracy.
• The optimized ML model offers a robust and scalable solution for predicting AAC performance.
• This study validates the potential of ML in advancing sustainable construction materials like alkali-activated concrete.