RETRACTED ARTICLE: A new nonlinear formulation-based prediction approach using artificial neural network (ANN) model for rubberized cement composite

Abstract

In this paper, a nonlinear formulation approach using an artificial neural network (ANN) is proposed and applied for the first time to predict the behavior of rubberized cement composite for which there is neither a design formulation nor a closed-form prediction model. For this purpose, the experimental data on the rubberized composite was collected, which includes mix ingredients, curing condition, and age. The statistical analysis of the data was performed to well characterize the interdependencies. Then, the ANN model was built, and the parameters for the optimum ANN model were found. Afterward, based on the relationships used in the ANN architecture at different steps, such as inputs nodes, neurons of hidden layers, activation functions, and output calculations, the ANN model was formulated, and the optimum weights and biases were determined. Once the closed-form formulation was derived, of the model’s performance was assessed through its prediction correlation and error. It was found that the ANN formulation was very robust to predict the complex behavior of rubberized composites with R2 = 0.98. Finally, the normalized importance of the model’s inputs was determined and the parametric study of the ANN model was also performed to show the effect of each variable on the output. The results showed that the relationships between the inputs and the output were mostly nonlinear, indicating the fact that the system’s behavior is a complex problem, which was well captured by formulation-based ANN model.