Explainable automl and multi-objective optimization for sustainable high-performance geopolymer concrete.

Area of Science:

• Materials Science
• Civil Engineering
• Environmental Science

Background:

• Growing environmental concerns necessitate sustainable alternatives to traditional concrete.
• High-performance geopolymer concrete (HPGC) offers a promising eco-friendly solution.
• Optimizing HPGC mix design for strength, cost, and emissions is challenging with conventional methods.

Purpose of the Study:

• To develop an integrated framework for HPGC mix-design.
• To balance compressive strength, cost, and carbon emissions.
• To enhance the interpretability and accuracy of predictive models for HPGC.

Main Methods:

• Implemented an automated machine learning (AutoML) algorithm for predictive modeling.
• Integrated Shapley Additive Explanations (SHAP) for model interpretability.
• Applied multi-objective optimization (MOO) with Pareto non-dominated sorting for design solutions.

Main Results:

• The AutoML model achieved high accuracy (R²=0.9280) in predicting HPGC compressive strength.
• SHAP analysis identified key factors influencing HPGC performance.
• The optimization framework reduced CO₂ emissions by 23-60% and unit costs by 16-36%.

Conclusions:

• The proposed integrated framework effectively balances multiple objectives in HPGC mix design.
• This approach advances eco-efficient concrete design and supports green building technologies.
• Further data expansion is recommended to improve model generalizability.