Evaluation and Estimation of Compressive Strength of Concrete Using Hybrid Modeling Techniques

The paper presents the research work carried out to predict the 28-day compressive strength of concrete with supplementary materials such as ash (fly ash, bottom ash) and silica fume using various data mining techniques. It mimics the decision-making ability of humans in imprecise and incomplete information situations. The model developed consists of 7 input parameters i.e., contents of cement, fine aggregates, coarse aggregates, silica fume, ash, water to cement ratio, superplasticizers, and one output parameter that is compressive strength at 28 days. The models used i.e., Gaussian Process (GP), Random Forest (RF), Artificial Neural Network (ANN) and ANN-Fire fly Algorithm (ANN-FFA) to estimate the compressive strength (MPa) at 28 days. The model developed is completely based on experimental data obtained from creditable literature available. The result of modeling techniques suggests that ANN-FFA based model works better than the other modeling techniques used in this study with Mean Square Error = 1.8099, Root Mean Square Error = 2.6584, and Coefficient of Correlation = 0.9370 with the testing dataset. Hence, these computational techniques suggest that it can be used to estimate the compressive strength of concrete at any stage. The sensitivity study concludes that RF is more sensitive to the absence of important parameters and less sensitive to the lack of less important parameters. Sensitivity analysis results using RF model suggest that Silica fume (kg/m 3 ) is the most important parameter for estimate the compressive strength of concrete using this data set.