A Stacking Ensemble-Based Framework for Predicting the Compressive Strength of Microwave-Cured Geopolymers

Area of Science:

• Materials Science
• Chemical Engineering
• Artificial Intelligence

Background:

• Microwave curing is an energy-efficient method for geopolymer synthesis.
• Optimizing geopolymer compressive strength is complex due to interacting variables.

Purpose of the Study:

• To develop an interpretable stacking ensemble surrogate model for predicting the compressive strength of microwave-cured geopolymers.
• To enhance the generalization and reliability of predictive models for geopolymer synthesis.

Main Methods:

• Compiled a literature-derived dataset of 235 observations (2010-2025) covering diverse precursors and curing parameters.
• Integrated linear, kernel, and tree-based models using a linear meta-learner.
• Implemented source-aware validation for reliability across heterogeneous studies.

Main Results:

• The stacking ensemble model achieved a high coefficient of determination (R^2) of 0.957 and a low RMSE of 5.64 MPa.
• Rigorous analysis confirmed the model's high reliability and stable performance across the strength range.
• Interpretability analyses identified curing time, microwave power, and specimen size as dominant strength-governing factors.

Conclusions:

• Stacking ensemble models are reliable statistical surrogate tools for preliminary experimental screening in geopolymer synthesis.
• The identified dominant factors align with established geopolymerization kinetics.
• The model is a literature-based surrogate tool, not validated for industrial-scale application.