APPLICATION OF BACK-PROPAGATION NEURAL NETWORKS FOR CORROSION BEHAVIOR ESTIMATION OF Ni-TiN COATINGS FABRICATED THROUGH PULSE ELECTRODEPOSITION

Abstract

In this paper, back-propagation (BP) neural network model with 8 hidden layers and 10 neurons was utilized to estimate corrosion behaviors of Ni-TiN coatings, deposited through pulse electrodeposition. Effects of plating parameters, namely, pulse frequency, TiN concentration and current density, on Ni-TiN coatings weight losses were discussed. Results indicated that pulse frequency, TiN concentration and current density had significant effects on weight losses of Ni-TiN coatings. Maximum mean square error of BP model was 9.10%, and this verified that the BP neural network model could accurately estimate corrosion behavior of Ni-TiN coatings. The coating fabricated at pulse frequency of 500[Formula: see text]Hz, TiN particle concentration of 8[Formula: see text]g/L and current density of 4[Formula: see text]A/dm2 consisted of fine grains and compact oxides, demonstrating that the coating displayed best corrosion resistance in this corrosion test. Concentrations of Ti and Ni in Ni-TiN coating prepared at pulse frequency of 500[Formula: see text]Hz, TiN particle concentration of 8[Formula: see text]g/L and current density of 4[Formula: see text]A/dm2 were 18.6[Formula: see text]at.% and 55.4[Formula: see text]at.%, respectively.