Artificial neural network modeling of electrochemical noise for predicting corrosion resistance of AA2024-AZ31 alloy in chloride solutions with cerium(III) nitrate hexahydrate

In this study, artificial neural networks were proposed and applied to estimate the noise resistance ( R n ) of the AA2024–AZ31 alloy exposed to 0.05 M NaCl with 0, 2, 4, 6, 8, and 10 mM of cerium(III) nitrate hexahydrate. Three pseudo-random number generators (MCG16807, MLF6331, and Mersenne Twister), two preprocessing strategies (normalization and standardization), and several transfer functions were evaluated to determine their impact on predictive accuracy. The Mersenne Twister method, combined with standardization, produced the most consistent results. The best performance was achieved with the dSiLU transfer function and seven neurons with a determination correlation ( R 2 ) of 0.998 and an RMSE of 33.363. Meanwhile, the Sinc-Squared function yielded results comparable to the best model. • Electrochemical noise modeling by ANN with different sampling, normalization, and transfer function strategies. • Three pseudo-random number generators (MCG16807, MLF6331, Twister) were evaluated for data sampling. • Normalization versus standardization compared for predictive performance on electrochemical resistance. • dSiLU and Sinc Squared transfer functions demonstrated suitability for modeling electrochemical noise with ANN.