Abstract

One of the most crucial mechanical characteristics determining the stiffness of concrete within its service life is the elastic modulus. This research deployed the support vector regression (SVR) and Back Propagation Neural Network (BPNN) models for predicting the elastic modulus of concrete with sawdust ash as a partial replacement for cement cured at 28 days in 5% sodium chloride using multi-input combinations. Four performance measures were used to analyze the research's findings.: determination coefficient (DC), Root Mean Square Error (RMSE), Mean Square Error (MSE) and Correlation Coefficient (CC). Common features among the experiential and predicted values were inspected and contrasted using a Taylor diagram. Prediction accuracy shows that BPNN-M2 performed better than SVR, with the greatest value of DC= 0. 9995 and the lowest RMSE value of 0.0035 in the testing phase. The overall quantitative comparison of the models revealed an important and narrow increment in the performance of BPNN against the SVR model despite the reliable accuracy of the SVR model. The findings also reveal the proposed AI models' overall ability to predict the 28-day elastic modulus of saw dust ash-based concrete cured in sodium chloride solution.