Modeling and prediction of corrosion inhibition efficiency of 5-Aminotetrazole on AA6065-AZ31 alloy using electrochemical noise and artificial neural networks with different transfer functions

Different configurations of artificial neural network (ANN) models were developed and evaluated to predict the corrosion inhibition efficiency of 5-Aminotetrazole on AA6065-AZ31 alloy exposed to saline conditions (0.1 M and 0.05 M NaCl) using electrochemical noise. The training dataset consisted of 302,400 measurements from immersion tests with inhibitor concentrations of 2 mM, 4 mM, 6 mM, 8 mM, and 10 mM. The variables time, inhibitor concentration, and electrolyte concentration were used as input variables, while the output variable was electrochemical resistance. A comprehensive analysis was performed using different transfer functions in the hidden layer, including TanSig, LogSig, ElliotSig, Radbas, Softmax, dSiLU, Sqsinc, ReLU, and SoftPlus, all trained with the Levenberg-Marquardt algorithm. Among these configurations, the model employing a 9-neuron hidden layer architecture and dSiLU as transfer function achieved the best performance. The determination coefficient (R²) of 0.9983 obtained by the best model demonstrated an excellent correlation between simulated and experimental data. The corrosion inhibition efficiency predicted by the best ANN model obtained less than 4 % error, confirming the ANN's potential for accurately modeling electrochemical noise. • Corrosion inhibition efficiency predicted using artificial neural networks and electrochemical noise. • Electrochemical resistance modeled with artificial neural networks employing nine transfer functions. • Comparative analysis performed on electrochemical resistance models. • dSiLU transfer function achieved the best results in modeling electrochemical noise with artificial neural networks.