Formulation of estimation models for the compressive strength of concrete mixed with nanosilica and carbon nanotubes

Abstract

New concepts for improving the performance of cementitious materials have recently surfaced due to the advancement in nanotechnology. In this context, nano silica (NS) and carbon nanotubes (CNTs) have been widely used in recent years. These nanomaterials significantly enhance the mechanical and durable performance of cement mixtures, although there are a few drawbacks related to high cost and workability issues. As a result, these components must be consumed at specific rates in order to achieve the desired qualities. This study aimed to create models to forecast the compressive strength (CS) for concrete mixed with two types of nanomaterials utilizing machine-learning-algorithms (MLA), such as the decision tree algorithm (DTA) and random forest algorithm (RFA). The results of both models were compared and verified by external K-fold cross-validation. A comprehensive database was collected containing 72 and 55 data points for the CS of CNTs and NS-modified concrete respectively. Four input variables such as fine aggregate (FA), cement content (CC), coarse aggregate (CA) and water-to-cement ratio (W/C) were used for the calibration of the models. Additionally, predicted results were checked through k-fold-cross-validation and other performance gauges such as mean-absolute-error (MAE), mean-squared-error (MSE), correlation coefficient (R2), root-mean-square-error (RMSE), relative-root-mean-square-error (RRMSE) and performance-index-factor (Pif). The RFA (CNTs and NS) models were found with better performance and accuracy than the DTA models by having the lowest MAE of 3.51, 4.17 RRMSE of 0.0783, 0.0584, and Pif of 0.0398 and 0.0299 respectively. In addition, the value of R2 for RFA models was observed higher such as 0.90 (CNTs) and 0.93 (NS), while for DTA models R2 was found as 0.88 (CNTs) and 0.86 (NS) respectively. The ensembled ML methods demonstrated a better generalization capability, which indicates their better ability for future prediction of CNTs and NS mixed concrete.