Forecasting the strength of nanocomposite concrete containing carbon nanotubes by interpretable machine learning approaches with graphical user interface

The sustainable development of the construction industry necessitates the utilization of multipurpose Cement Composites (CC). Therefore, the integration of nanomaterials has the potential to provide CC that exhibits superior performance and possesses several functionalities. Hence, the use of Carbon Nanotubes (CNTs) inside the concrete cementitious sector holds significant potential for implementing effective solutions toward creating a sustainable ecosystem characterized by versatile attributes. Nevertheless, the prediction of the characteristics of these composites is a significant challenge owing to their complex composite structure and non-linear response. Furthermore, the process of designing and executing experimental trials on diverse samples and across various age groups is arduous, time-consuming, and financially burdensome. There is currently a dearth of a predictive model capable of estimating the compressive strength of concrete including nanoparticles. The utilization of such models is of significant importance in the project and study of Reinforced Concrete (RC) structures including nanoparticles. Three machine learning algorithms, including Gene Expression Programming (GEP), Gradient Boosting (GB), and Extreme Gradient Boosting (XGB), were utilized in this study to forecast the Compressive Strength (CS) of nanocomposites that incorporate CNTs. The evaluation of the models' reliability was conducted by the utilization of cross-validation with K-folding and subsequent statistical error analysis. According to the results of the coefficient of determination (R2), the XGB model achieved the highest R2 value (0.95), while the GB model and GEP model both earned R2 values of 0.94. Furthermore, the validation method for the models included the implementation of statistical analysis and k-fold cross-validation. Therefore, the XGB model exhibited much lower values for statistical metrics compared to the GEP and GB models. In addition, a GEP empirical equation and a Graphical User Interface (GUI) have been created for practical applications in predicting the strength of concrete. This streamlines the procedure and provides a valuable instrument for harnessing the model's potential in the field of civil engineering. Furthermore, the use of Shapley analysis is conducted to assess the predominant factors in concrete prediction. The findings of this research indicate that the curing time, type of cement, and water-to-cement ratio significantly influence the properties of CNT-based concrete composites.