Abstract
Sulfur is amongst the most important and useful industrial materials. A key product of sulfur is what we know as sulfur concrete (SC). However, the estimation of the compressive strength of sulfur concrete through testing may become time- and cost-intensive. Therefore, the application of soft computing techniques can help accelerate and simplify this process. Accordingly, in the present research, artificial neural networks (ANNs) and gene expression programming (GEP) were applied to predict the compressive strength of the sulfur concrete. Experimental data on the compressive strength of 33 concrete samples with different mix designs were used to develop the ANN and GEP with four input parameters, namely filler, sulfur, sand, and aggregate contents, with the model output (i.e., compressive strength) being classified under 5 different classes of G1, G2, G3, G4, and G5. To study the proposed models in terms of accuracy, comparative analyses were conducted in the form of statistical indices of R2, RMSE, and MAE. Next, analysis of variance (ANOVA) was performed using two-factor analysis at the confidence level of 95% (α = 0.05), indicating high potentials of the ANN and GEP for the prediction of compressive strength of sulfur concrete based on comparisons to experimental data. The findings of the present research can serve as a reliable alternative to time- and cost-intensive tests for obtaining the compressive strength of sulfur concrete