Sustainable predictive model of concrete utilizing waste ingredient: Individual alogrithms with optimized ensemble approaches

Abstract

Silica fume (SF) is a prominent mineral ingredient used to manufacture sustainable concrete in the construction industry. The use of silica fume as a partial substitute for cement has numerous advantages, including lower carbon dioxide (CO2) excretion, more economical concrete, higher strength, and improved mechanical capabilities. As the effects of climate change continue to deteriorate, it has become crucial to create machine learning approaches with predictive capacities. Therefore, the purpose of this investigation is to develop models for evaluating the compressive strength (CS) of silica fume concrete (SF-C). This study utilizes various stand alone algorithms and modified like decision tree (DT), random forest (RF), mulilayer perceptron neural network (ANN), support vector regression. Afterwards, ensembling of the algorithms is done with bagging and boosting with adaboot to minimize the variance and biasness. To develop models, comprehensive published data is gathered containing 283 samples with superplastisizer (kg/m3), cement (kg/m3), silica fume (kg/m3), fine aggregate (kg/m3), coarse aggregate (kg/m3), and water (kg/m3) as input parameters. Moreover, statistical approaches is employed to assess the models capacity, including mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2). The result reveals that the ensemble approach on standalone models and modified bagging depicts robust performance with R2 > 0.9. Infact, RF gives a most accurate and precise measures with R2 = 0.95. Similarly, cross-validation measures was utilized to eliminate fitting issues and evaluate the generalized modeling technique's results. The permutation importance reveals that silica fume have major influence to its outcome. Thus, uitlization of MLA in concrete industry will save time and give robust results in making an eco-friendly concrete.