Abstract: Carbonation is a dangerous threat to concrete since it reduces the alkalinity of normal or self-compacting concrete (SCC), allowing iron to corrode and spall the cover. The goal of this research is to use an artificial neural network to create a corrosion prediction model for steel in self-compacting concrete that has been subjected to carbonation. In this study, MATLABR2019a was used to create a feedforward back propagation neural network. As a training function, the researchers utilized the Levenberg-Marquardt back propagation (TRAINLM) which adjusts weights and bias values using Levenberg-Marquardt optimization. The researchers used gradient descent with momentum weight/bias learning (LEARNGDM) for the adaptation learning function, which is a technique that aids the gradient in determining which way to go. The network’s performance was measured using the mean square error (MSE). The Hyperbolic tangent sigmoid transfer function (TANSIG) was also employed as the transfer function since the values obtained by this function range from +1 to -1, considering both the positive and negative aspects of the parameter. To minimize overfitting, the number of hidden nodes should be fewer than the number of input parameters. The researchers tested 4-12 hidden nodes. Modeling was done using data from 102 experimental studies of self-compacting concrete exposed to corrosion. Using feed-forward back propagation ANN with 1 hidden layer and 8 hidden nodes, a Pearson R-value of 0.98748 and a mean square error of 0.5725 were obtained. The factor that most affect the carbonation depth were water-cement ratio and fly ash content. The suggested model was able to analytically describe the connection and behaviors of the various mixtures to the carbonation depth in the parametric investigation. The parameters characteristics were likewise described by the model.