Prediction of Mechanical Properties of Fly-Ash/Slag-Based Geopolymer Concrete Using Ensemble and Non-Ensemble Machine-Learning Techniques

Muhammad Nasir Amin; Kaffayatullah Khan; Muhammad Faisal Javed; Fahid Aslam; Muhammad Ghulam Qadir; Muhammad Iftikhar Faraz

doi:10.3390/ma15103478

Prediction of Mechanical Properties of Fly-Ash/Slag-Based Geopolymer Concrete Using Ensemble and Non-Ensemble Machine-Learning Techniques

Materials (Basel). 2022 May 12;15(10):3478. doi: 10.3390/ma15103478.

Authors

Muhammad Nasir Amin¹, Kaffayatullah Khan¹, Muhammad Faisal Javed², Fahid Aslam³, Muhammad Ghulam Qadir⁴, Muhammad Iftikhar Faraz⁵

Affiliations

¹ Department of Civil and Environmental Engineering, College of Engineering, King Faisal University, P.O. Box 380, Al-Ahsa 31982, Saudi Arabia.
² Department of Civil Engineering, Abbottabad Campus, COMSATS University Islamabad, Abbottabad 22060, Pakistan.
³ Department of Civil Engineering, College of Engineering in Al-Kharj, Prince Sattam Bin Abdulaziz University, Al-Kharj 11942, Saudi Arabia.
⁴ Department of Environmental Sciences, Abbottabad Campus, COMSATS University Islamabad, Abbottabad 22060, Pakistan.
⁵ Department of Mechanical Engineering, College of Engineering, King Faisal University, Al-Ahsa 31982, Saudi Arabia.

Abstract

The emission of greenhouse gases and natural-resource depletion caused by the production of ordinary Portland cement (OPC) have a detrimental effect on the environment. Thus, an alternative means is required to produce eco-friendly concrete such as geopolymer concrete (GPC). However, GPC has a complex cementitious matrix and an ambiguous mix design. Aside from that, the composition and proportions of materials utilized may have an impact on the compressive strength. Similarly, the use of robust and efficient machine-learning (ML) approaches is now required to forecast the strength of such a composite cementitious matrix. As a result, this study anticipated the compressive strength of GPC with waste resources using ensemble and non-ensemble ML algorithms. This was accomplished through the use of Anaconda (Python). To build a strong ensemble learner by integrating weak learners, adaptive boosting, random forest (RF), and ensemble learner bagging were employed. Furthermore, ensemble learners were utilized on non-ensemble or weak learners, such as decision trees (DT) and support vector machines (SVM) via regression. The data encompassed 156 statistical samples in which nine variables, namely superplasticizer (kg/m³), fly ash (kg/m³), ground granulated blast-furnace slag (GGBS), temperature (°C), coarse and fine aggregate (kg/m³), sodium silicate (Na₂SiO₃), and sodium hydroxide (NaOH), were chosen to anticipate the results. Exploring it in depth, twenty sub-models with ensemble boosting and bagging approaches were trained, and tuning was performed to achieve the highest possible coefficient of determination (R²). Moreover, cross K-Fold validation analysis and statistical checks were performed via indicators for the evaluation of the models. The result revealed that ensemble approaches yielded robust performance compared to non-ensemble algorithms. Generally, an ensemble learner with the RF and bagging approach on a DT yielded robust performance by achieving a better R² as 0.93, and with the lowest statistical errors. The communal model in artificial-intelligence analysis, on average, improved the accuracy of the model.

Keywords: ensemble approaches; fly ash; machine-learning; parametric analysis; slag; validation.

Grants and funding

This work was supported by the Deanship of Scientific Research, Vice Presidency for Graduate Studies and Scientific Research, King Faisal University, Saudi Arabia [Project No. AN000680]/King Faisal University