Modeling of Mechanical Properties of Silica Fume-Based Green Concrete Using Machine Learning Techniques

Abstract

Silica fume (SF) is a frequently used mineral admixture in producing sustainable concrete in the construction sector. Incorporating SF as a partial substitution of cement in concrete has obvious advantages, including reduced CO2 emission, cost-effective concrete, enhanced durability, and mechanical properties. Due to ever-increasing environmental concerns, the development of predictive machine learning (ML) models requires time. Therefore, the present study focuses on developing modeling techniques in predicting the compressive strength of silica fume concrete. The employed techniques include decision tree (DT) and support vector machine (SVM). An extensive and reliable database of 283 compressive strengths was established from the available literature information. The six most influential factors, i.e., cement, fine aggregate, coarse aggregate, water, superplasticizer, and silica fume, were considered as significant input parameters. The evaluation of models was performed by different statistical parameters, such as mean absolute error (MAE), root mean squared error (RMSE), root mean squared log error (RMSLE), and coefficient of determination (R2). Individual and ensemble models of DT and SVM showed satisfactory results with high prediction accuracy. Statistical analyses indicated that DT models bested SVM for predicting compressive strength. Ensemble modeling showed an enhancement of 11 percent and 1.5 percent for DT and SVM compressive strength models, respectively, as depicted by statistical parameters. Moreover, sensitivity analyses showed that cement and water are the governing parameters in developing compressive strength. A cross-validation technique was used to avoid overfitting issues and confirm the generalized modeling output. ML algorithms are used to predict SFC compressive strength to promote the use of green concrete.

A Comparison of Machine Learning Tools That Model the Splitting Tensile Strength of Self-Compacting Recycled Aggregate Concrete

“…Currently, the R 2 value is thought to be the best metric for assessing the model [95,97]. Table 4 shows the range of R 2 values for prediction model evaluations [54,98,99]. On the other hand, the closer the root mean square error and mean average error values are to zero, the better the ML model's performance is at predicting the splitting tensile strength of SCC made with RA after 28 days [14,21,55,100].…”

A Systematic Review of the Research Development on the Application of Machine Learning for Concrete

“…Some supplementary elements such as fly ash, silica fume, or chemical combinations have been utilized to enhance the performance and strength of concrete [32][33][34][35][36]. The literature indicates that researchers are concentrating increasingly on these supplementary materials, since they are often waste materials created as a result of industrial, agricultural, and municipal processes [37][38][39][40][41]. The recycling and the widespread stockpiling of these leftovers for bulk use raise serious social and environmental problems on a global scale [42].…”

Assessment of Artificial Intelligence Strategies to Estimate the Strength of Geopolymer Composites and Influence of Input Parameters

“…Furthermore, regression approaches may overestimate the contribution of certain elements [48]. Artificial intelligence methods, such as supervised machine learning (ML), are among the most refined modeling methods used in the present study area [49][50][51][52][53]. These methods utilize input variables to model responses, and the yield models are supported by testing.…”