Prediction of compressive and flexural strength of coal gangue-based geopolymer using machine learning method

Document Type

Article

Publication Date

3-2025

Department

Department of Civil, Environmental, and Geospatial Engineering

Abstract

The mechanical properties of coal gangue-based geopolymers are influenced by various factors, including source location, activator type, and liquid-to-solid ratio. Among these, the precursor Si/Al ratio exhibits vary significantly based on the source location. This study focuses on investigating the effectiveness of machine learning models in predicting the mechanical properties of coal gangue-based geopolymers and offers guidance for mix design tailored to coal gangue from different sources. By utilizing machine learning models, this research aims to optimize the mix proportions of geopolymer materials, enhancing the sustainable recycling of coal gangue. To achieve this, orthogonal experiments were conducted by adjusting the Si/Al ratio of coal gangue-based geopolymers by incorporating fly ash and slag. The experimental results, combined with data from the literature, formed a dataset that was analyzed using various machine learning techniques. Generalization tests were also conducted to assess the predictive accuracy of the best-performing models. The results indicate that XGBoost and Random Forest exhibited strong predictive accuracy, with R² values of 0.865 and 0.882 for 3-day and 28-day compressive strength, and 0.788 for 28-day flexural strength. Prediction errors remained minimal, with MAE values of 2.78 MPa, 3.38 MPa, and 0.594 MPa for 3-day compressive, 28-day compressive, and 28-day flexural strength, respectively, confirming model reliability. Furthermore, the generalization test results indicate that Random Forest is less sensitive to variations in precursor material composition under identical Si/Al conditions, exhibiting more excellent generalization capabilities. This study provides theoretical support for the intelligent optimization of geopolymer mix design, which not only reduces experimental costs but also contributes to the sustainable utilization of coal gangue as a cementitious material alternative, thereby mitigating environmental impacts associated with conventional cement production.

Publisher's Statement

© 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

Publication Title

Materials Today Communications

Link to Full Text

Share

COinS